DK3009_half-series-title
3/7/05
12:52 PM
Page A
Synaptic Plasticity
DK3009_half-series-title
3/7/05
12:52 PM
P...
30 downloads
1467 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
DK3009_half-series-title
3/7/05
12:52 PM
Page A
Synaptic Plasticity
DK3009_half-series-title
3/7/05
12:52 PM
Page B
NEUROLOGICAL DISEASE AND THERAPY Advisory Board Louis R. Caplan, M.D. Professor of Neurology Harvard University School of Medicine Beth Israel Deaconess Medical Center Boston, Massachusetts
William C. Koller, M.D. Mount Sinai School of Medicine New York, New York
John C. Morris, M.D. Friedman Professor of Neurology Co-Director, Alzheimer’s Disease Research Center Washington University School of Medicine St. Louis, Missouri
Bruce Ransom, M.D., Ph.D. Warren Magnuson Professor Chair, Department of Neurology University of Washington School of Medicine Seattle, Washington
Kapil Sethi, M.D. Professor of Neurology Director, Movement Disorders Program Medical College of Georgia Augusta, Georgia
Mark Tuszynski, M.D., Ph.D. Associate Professor of Neurosciences Director, Center for Neural Repair University of California–San Diego La Jolla, California
DK3009_half-series-title
3/7/05
12:52 PM
Page C
1. Handbook of Parkinson’s Disease, edited by William C. Koller 2. Medical Therapy of Acute Stroke, edited by Mark Fisher 3. Familial Alzheimer’s Disease: Molecular Genetics and Clinical Perspectives, edited by Gary D. Miner, Ralph W. Richter, John P. Blass, Jimmie L. Valentine, and Linda A. Winters-Miner 4. Alzheimer’s Disease: Treatment and Long-Term Management, edited by Jeffrey L. Cummings and Bruce L. Miller 5. Therapy of Parkinson’s Disease, edited by William C. Koller and George Paulson 6. Handbook of Sleep Disorders, edited by Michael J. Thorpy 7. Epilepsy and Sudden Death, edited by Claire M. Lathers and Paul L. Schraeder 8. Handbook of Multiple Sclerosis, edited by Stuart D. Cook 9. Memory Disorders: Research and Clinical Practice, edited by Takehiko Yanagihara and Ronald C. Petersen 10. The Medical Treatment of Epilepsy, edited by Stanley R. Resor, Jr., and Henn Kutt 11. Cognitive Disorders: Pathophysiology and Treatment, edited by Leon J. Thal, Walter H. Moos, and Elkan R. Gamzu 12. Handbook of Amyotrophic Lateral Sclerosis, edited by Richard Alan Smith 13. Handbook of Parkinson’s Disease: Second Edition, Revised and Expanded, edited by William C. Koller 14. Handbook of Pediatric Epilepsy, edited by Jerome V. Murphy and Fereydoun Dehkharghani 15. Handbook of Tourette’s Syndrome and Related Tic and Behavioral Disorders, edited by Roger Kurlan 16. Handbook of Cerebellar Diseases, edited by Richard Lechtenberg 17. Handbook of Cerebrovascular Diseases, edited by Harold P. Adams, Jr. 18. Parkinsonian Syndromes, edited by Matthew B. Stern and William C. Koller 19. Handbook of Head and Spine Trauma, edited by Jonathan Greenberg 20. Brain Tumors: A Comprehensive Text, edited by Robert A. Morantz and John W. Walsh 21. Monoamine Oxidase Inhibitors in Neurological Diseases, edited by Abraham Lieberman, C. Warren Olanow, Moussa B. H. Youdim, and Keith Tipton
DK3009_half-series-title
3/7/05
12:52 PM
Page D
22. Handbook of Dementing Illnesses, edited by John C. Morris 23. Handbook of Myasthenia Gravis and Myasthenic Syndromes, edited by Robert P. Lisak 24. Handbook of Neurorehabilitation, edited by David C. Good and James R. Couch, Jr. 25. Therapy with Botulinum Toxin, edited by Joseph Jankovic and Mark Hallett 26. Principles of Neurotoxicology, edited by Louis W. Chang 27. Handbook of Neurovirology, edited by Robert R. McKendall and William G. Stroop 28. Handbook of Neuro-Urology, edited by David N. Rushton 29. Handbook of Neuroepidemiology, edited by Philip B. Gorelick and Milton Alter 30. Handbook of Tremor Disorders, edited by Leslie J. Findley and William C. Koller 31. Neuro-Ophthalmological Disorders: Diagnostic Work-Up and Management, edited by Ronald J. Tusa and Steven A. Newman 32. Handbook of Olfaction and Gustation, edited by Richard L. Doty 33. Handbook of Neurological Speech and Language Disorders, edited by Howard S. Kirshner 34. Therapy of Parkinson’s Disease: Second Edition, Revised and Expanded, edited by William C. Koller and George Paulson 35. Evaluation and Management of Gait Disorders, edited by Barney S. Spivack 36. Handbook of Neurotoxicology, edited by Louis W. Chang and Robert S. Dyer 37. Neurological Complications of Cancer, edited by Ronald G. Wiley 38. Handbook of Autonomic Nervous System Dysfunction, edited by Amos D. Korczyn 39. Handbook of Dystonia, edited by Joseph King Ching Tsui and Donald B. Calne 40. Etiology of Parkinson’s Disease, edited by Jonas H. Ellenberg, William C. Koller, and J. William Langston 41. Practical Neurology of the Elderly, edited by Jacob I. Sage and Margery H. Mark 42. Handbook of Muscle Disease, edited by Russell J. M. Lane 43. Handbook of Multiple Sclerosis: Second Edition, Revised and Expanded, edited by Stuart D. Cook
DK3009_half-series-title
3/7/05
12:52 PM
Page E
44. Central Nervous System Infectious Diseases and Therapy, edited by Karen L. Roos 45. Subarachnoid Hemorrhage: Clinical Management, edited by Takehiko Yanagihara, David G. Piepgras, and John L. D. Atkinson 46. Neurology Practice Guidelines, edited by Richard Lechtenberg and Henry S. Schutta 47. Spinal Cord Diseases: Diagnosis and Treatment, edited by Gordon L. Engler, Jonathan Cole, and W. Louis Merton 48. Management of Acute Stroke, edited by Ashfaq Shuaib and Larry B. Goldstein 49. Sleep Disorders and Neurological Disease, edited by Antonio Culebras 50. Handbook of Ataxia Disorders, edited by Thomas Klockgether 51. The Autonomic Nervous System in Health and Disease, David S. Goldstein 52. Axonal Regeneration in the Central Nervous System, edited by Nicholas A. Ingoglia and Marion Murray 53. Handbook of Multiple Sclerosis: Third Edition, edited by Stuart D. Cook 54. Long-Term Effects of Stroke, edited by Julien Bogousslavsky 55. Handbook of the Autonomic Nervous System in Health and Disease, edited by C. Liana Bolis, Julio Licinio, and Stefano Govoni 56. Dopamine Receptors and Transporters: Function, Imaging, and Clinical Implication, Second Edition, edited by Anita Sidhu, Marc Laruelle, and Philippe Vernier 57. Handbook of Olfaction and Gustation: Second Edition, Revised and Expanded, edited by Richard L. Doty 58. Handbook of Stereotactic and Functional Neurosurgery, edited by Michael Schulder 59. Handbook of Parkinson’s Disease: Third Edition, edited by Rajesh Pahwa, Kelly E. Lyons, and William C. Koller 60. Clinical Neurovirology, edited by Avindra Nath and Joseph R. Berger 61. Neuromuscular Junction Disorders: Diagnosis and Treatment, Matthew N. Meriggioli, James F. Howard, Jr., and C. Michel Harper 62. Drug-Induced Movement Disorders, edited by Kapil D. Sethi
DK3009_half-series-title
3/7/05
12:52 PM
Page F
63. Therapy of Parkinson’s Disease: Third Edition, Revised and Expanded, edited by Rajesh Pahwa, Kelly E. Lyons, and William C. Koller 64. Epilepsy: Scientific Foundations of Clinical Practice, edited by Jong M. Rho, Raman Sankar, and José E. Cavazos 65. Handbook of Tourette’s Syndrome and Related Tic and Behavioral Disorders: Second Edition, edited by Roger Kurlan 66. Handbook of Cerebrovascular Diseases: Second Edition, Revised and Expanded, edited by Harold P. Adams, Jr. 67. Emerging Neurological Infections, edited by Christopher Power and Richard T. Johnson 68. Treatment of Pediatric Neurologic Disorders, edited by Harvey S. Singer, Eric H. Kossoff, Adam L. Hartman, and Thomas O. Crawford 69. Synaptic Plasticity : Basic Mechanisms to Clinical Applications, edited by Michel Baudry, Xiaoning Bi, and Steven S. Schreiber 70. Handbook of Essential Tremor and Other Tremor Disorders, edited by Kelly E. Lyons and Rajesh Pahwa
DK3009_half-series-title
3/7/05
12:52 PM
Page i
Synaptic Plasticity Basic Mechanisms to Clinical Applications
edited by
Michel Baudry Xiaoning Bi Steven S. Schreiber
Boca Raton London New York Singapore
DK3009_Discl Page 1 Wednesday, March 9, 2005 1:11 PM
Published in 2005 by Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2005 by Taylor & Francis Group, LLC No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-10: 0-8247-5900-1 (Hardcover) International Standard Book Number-13: 978-0-8247-5900-1 (Hardcover) This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data Catalog record is available from the Library of Congress
Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com Taylor & Francis Group is the Academic Division of T&F Informa plc.
Published in 2005 by Taylor & Francis Group 6000 Broken Sound Parkway NW Boca Raton, FL 33487–2742 # 2005 by Taylor & Francis Group, LLC No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-10: 0–8247–5900–1 (Hardcover) International Standard Book Number-13: 978–0–8247–5900–1 (Hardcover) This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http:==www.copyright.com=) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978–750–8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data Catalog record is available from the Library of Congress
Visit the Taylor & Francis Web site at http :==www.taylorandfrancis.com
Preface
Although tremendous progress has been made over the last decade in our understanding of the molecular and cellular mechanisms underlying synaptic plasticity, this term has been so widely used that it has lost most of its original meaning. The editors of this book use this term to refer to fundamental processes that are responsible for adaptive properties of the central nervous system (CNS) under physiological as well as pathological conditions. Thus, many mechanisms regarding the way physical and nervous activity, diseases, and insults regulate synaptic transmission in the adult CNS have been characterized, and this understanding has started to be implemented in the design of new treatments for neuronal diseases and age-related loss of cognitive function. Likewise, basic processes used by developing nervous systems to establish their adult patterns of connectivity have been described in exquisite detail. In particular, the roles of growth factors and their effects on gene regulation are leading the way to new avenues for the treatment of several genetic and environmental defects associated with the inappropriate establishment of neuronal iii
iv
Preface
connectivity. Furthermore, the field of neuronal degeneration has seen a renewed excitement with advances in stem cell research. At the other end of the spectrum of scientists interested in improving alterations in various CNS functions, neurologists, neurosurgeons, occupational therapists, and other caregivers are developing new approaches to help patients recover lost functions using the very same principles of synaptic plasticity studied by molecular and cellular neurobiologists. In particular, several methods of rehabilitation therapy and repair are in various stages of implementation. The premise here is that the plastic, adaptive properties of neuronal networks will allow for compensatory mechanisms in a way that might provide new functionalities to replace damaged circuitries. Furthermore, the use of genetically engineered cells or neurons targeted at specific locations or for specific functions might well facilitate recovery of lost functions. A third category of scientists interested in brain repair is slowly emerging out of the new disciplines of neural engineering and neural computation. Their explicit goals are to use principles extracted from neurobiological systems to reverse engineer neuronal circuits that are damaged as a result of genetic defect, trauma, and other disease processes. This requires not only a detailed description of neuronal circuitries in various CNS structures as well as the rules regulating synaptic transmission and activity-dependent changes in synaptic transmission, but also the ability to implement neuronal circuits and rules of plasticity into silicon-based circuits and devices. In addition, this approach will require new technologies to permit long-term interactions between silicon-based devices and neural tissues. We reasoned that the time was ripe for presenting a summary of the most recent advances in these three lines of research. Each of them has progressed along paths that superficially look quite different, but nevertheless they all are converging toward the same fundamental goal; that is, to provide new therapeutics for alleviating impairments of CNS function resulting from gene defects, injury, or disease. Clearly, some of the research is years away from providing
Preface
v
any concrete device or treatment, but technological as well as empirical advances move so fast that it is not too early to think about the clinical and ethical applications of these lines of work. Similarly, gene therapy was thought to remain for years in an incubation phase, but advances in this field have been so rapid that ethical concerns as well as practical issues have now moved to the front stage of the discussion. Because of the apparent distance between these fields, they are rarely associated in books, and this was another incentive for us to bring together leaders in these fields to provide their views, ideas, and suggestions for future directions. The book is therefore organized in sections that reflect the three approaches mentioned above. The first section is devoted to basic mechanisms of synaptic plasticity and is subdivided into two sections. The first section reviews various forms of plasticity in synaptic and nonsynaptic transmission. In particular, one of us (MB) provides an update on the mechanism of long-term potentiation (LTP) and on the development of a new class of molecules, the ampakines (positive modulators of AMPA receptors), which are making their way to the clinic for various indications, including mild-cognitive impairment associated with aging. The group of Dominique Muller presents convincing evidence that neural activity-induced plasticity is associated with formation of new synapses. On the other hand, Bach-yRita reminds us that nonsynaptic transmission, also called volume transmission, plays an important role in the brain under physiological and pathological conditions and that it may play a very significant role in tactile vestibular substitution following bilateral vestibular damage. The second section deals with various factors that participate in mechanisms regulating neuronal sprouting after lesions, age-related neurodegeneration, and exercise-induced improvement of cognitive function. One of us (XB) reviews the role of lysosomal dysfunction in aging and presents an hypothesis linking dysfunction in this cellular system to impairment in trophic signaling. The group of Marie-Francoise Chesselet summarizes the evidence that axonal sprouting is much more widespread than initially thought and discusses potential
vi
Preface
mechanisms underlying this phenomenon. Gomez-Pinilla and one of his associates provide an extensive review of the characteristics and potential functions of neurotrophins, and in particular, brain-derived neurotrophic factor (BDNF). They then illustrate how the understanding of these functions could be translated into new ways to stimulate brain function and the potential therapeutic benefits of exercise, enriched environment, or diet. Finally, Roberta Brinton takes the difficult task of explaining the recent failures of estrogen therapy for the treatment or prevention of Alzheimer’s disease. She identifies potential factors that could account for these failures and proposes a new framework to understand the mechanisms of estrogen action. The second section of the book is directed at clinical applications of synaptic plasticity processes and is subdivided into three sections. The first one deals with various aspects of injury. Kornblum’s group describes the advantages of using imaging techniques and in particular, positron emission tomography (PET), to evaluate the role of neural plasticity in recovery from damage in rodent models of brain injury. Jim Simpkins and colleagues review the current evidence supporting neuroprotective roles for estrogens in stroke and also discuss the argument for the use of estrogen as a neuroprotectant for Alzheimer’s disease. Finally, one of us (SS) describes the current knowledge of the mechanisms underlying neuronal death, summarizes the evidence supporting a role for apoptotic cell death in spinal cord injury and outlines potential approaches to interfere with these processes following spinal cord injury. The second part of this section focuses on rehabilitation following injury, mostly stroke, one of the major sources of brain injuries. Hesse and Werner provide a broad review of all the rehabilitation approaches that have been used to promote recovery of motor function following stroke. Nudo and his colleagues then focus on the neural bases for new therapies for maximizing the recovery process. Bruce Dobkin discusses the challenges that remain ahead of us to link plasticity processes at various levels of analysis to recovery of motor function after stroke. This theme is further developed in the following chapter by Winstein and Prettyman
Preface
vii
who focus on constraint-induced (CI) therapy for functional recovery after stroke. After discussing the detailed principles underlying this therapy, they review the potential biological mechanisms that could account for the effects of CI therapy. Finally, Cynthia Thompson reviews neuroimaging studies applied to understand recovery of language functions in aphasic patients with left or right hemispheric lesions. The third part of this section covers various approaches using stem cell grafting coupled with different gene technologies that are being explored to repair CNS damage. Whittemore and colleagues review numerous studies using a variety of stem cells to repair spinal cord injury, and they stress the fact that we need to better understand the factors involved in the differentiation of these cells in various locations. This is followed by a chapter by Popovic, Petersen, and Brundin reviewing a large number of studies performed in rodents, nonhuman primates, and humans to use stem cell implants to treat Huntington’s disease. Although the authors are optimistic that a grafting approach will ultimately work, they do stress the need to obtain a better source of stem cells for implantation and better understanding of the factors needed for optimal survival of the grafted cells. The final chapter of this section by Do¨bro¨ssy and Dunnett further elaborates on this theme and discusses how interactions between external manipulations might provide optimal conditions for survival of grafted cells, thus linking the rehabilitation and the cell therapies together. The last section of the book is devoted to the final line of research discussed above, namely engineering approaches to understand mechanisms of plasticity and develop new therapies based on these principles. The group of Sam Deadwyler reviews some of the recent work applying a variety of algorithms to examine the reliability and dynamics of hippocampal neural representations of information. They demonstrate that the analysis of the activity of a relatively small group of neurons is sufficient to provide an adequate prediction of the animal’s behavior. David Krupa summarizes on-going work in a number of laboratories directed at developing so-called brain–machine interfaces (BMI) that would eventually be
viii
Preface
used to restore lost brain functions. The need for further basic research before such devices can be successfully implanted in humans is stressed. Finally, the group of Ted Berger provides an update on their on-going work directed at developing implantable biomimetic devices that can replace damaged brain regions involved in cognitive functions. Their work represents the perfect illustration of the overall theme of the book, as it starts from the understanding of basic mechanisms of synaptic transmission and information processing, processes through the development of VLSI devices reproducing the functions of large groups of neurons, and ends at the clinical level to improve the conditions of large numbers of patients suffering loss of cognitive abilities. Overall, we wish to thank each of the contributors who have provided what we believe will be extremely useful information for a wide range of neuroscientists and clinicians interested in understanding the basic mechanisms of CNS plasticity and the potential clinical applications of these mechanisms to recover functions that are lost as a result of disease or injury. Michel Baudry Xiaoning Bi Steven S. Schreiber
Contents
Preface . . . . iii Contributors . . . . xvii 1. Memory: From Molecular Mechanisms to Clinical Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Michel Baudry 1. Introduction . . . . 1 2. LTP: A Cellular Mechanism for Learning and Memory . . . . 2 3. Current Hypotheses for LTP=LTD Mechanisms . . . . 9 4. Ampakines, LTP, and Memory . . . . 13 5. Clinical Trials of Ampakines . . . . 15 6. Conclusions . . . . 16 References . . . . 16 2. Synaptogenesis as a Correlate of Activity-Induced Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Irina Nikonenko, Pascal Jourdain, Bemadett Boda, and Dominique Muller 1. Introduction . . . . 25 2. Changes in Spine Dynamics with Activity . . . . 26 3. Modulation of Spine Morphology by Activity . . . . 26 4. Evidence for Activity-Induced Synaptogenesis . . . . 28 5. Postsynaptic Mechanisms Controlling Activity-Induced Synaptogenesis . . . . 30 6. Activity-Induced Remodeling of Presynaptic Terminals . . . . 31 ix
x
Contents
7. Conclusions . . . . 32 References . . . . 33 3. Volume Transmission: Role in the Carry-Over Functional Effect in Tactile Vestibular Substitution? . . . . . . . . . . . . . Paul Bach-y-Rita 1. Volume Transmission . . . . 39 2. The History of Volume Transmission . . . . 42 3. Computational Neuroscience VT Studies . . . . 44 4. Carry-Over Functional Effect in Tactile Vestibular Substitution? . . . . 47 References . . . . 48
39
4. Lysosomal Dysfunction in Brain Aging and Neurodegeneration: Roles in Trophic Signaling and Neuroinflammation . . . . . . 53 Xiaoning Bi 1. Lysosomal Dysfunction and Brain Aging . . . . 53 2. Lysosomal Dysfunction and Trophic Endosomal Signaling . . . . 57 3. Lysosomal Dysfunction and Neuroinflammation . . . . 58 4. Roles of Lysosomal Dysfunction in Brain Aging and Neurodegeneration: A Hypothesis . . . . 62 References . . . . 64 5. Axonal Sprouting in the Adult Brain: Mechanisms of Lesion-Specific Sprouting of the Corticostriatal Pathway in Adult Rats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marie-Francoise Chesselet, S. Thomas Carmichael, Marc Shomer, Dorothy Harris, and Veronique Riban 1. Introduction . . . . 75 2. Lesion-Specific Axonal Sprouting of the Corticostriatal Pathway . . . . 76 3. The Role of Electrophysiological Activity in Axonal Sprouting in the Adult Brain . . . . 80 4. Myelin-Associated Proteins and Axonal Sprouting in the Adult Brain . . . . 85 5. Conclusion . . . . 85 References . . . . 86
75
Contents
6. Behavioral Implications for the Activity Dependent Regulation of Neurotrophins . . . . . . . . . . . . . . . . . . . . . . Fernando Gomez-Pinilla and Shoshanna Vaynman 1. Introduction . . . . 89 2. Cell Biology of NTs . . . . 90 3. Neurotrophins and Neural Activity . . . . 95 4. Neurotrophins and Learning and Memory . . . . 101 5. Neurotrophins and Behavior and Lifestyle: Experience-Dependent Plasticity . . . . 107 6. Conclusions and Research Demands . . . . 118 References . . . . 118
xi
89
7. Estrogen Therapy for Prevention of Alzheimer’s Disease But Not for Rehabilitation Following Onset of Disease: The Healthy Cell Bias of Estrogen Action . . . . . . . . . . . 131 Roberta Diaz Brinton 1. Introduction . . . . 131 2. Alzheimer’s Disease: A Potential Health Crisis . . . . 132 3. Hormone Replacement Therapy (HRT) and Risk of Developing AD: Importance of the Time of Onset and the Type of HRT . . . . 133 4. The Role of Estrogen in Cognitive Functions . . . . 140 5. How Might Estrogen-Replacement Therapy Reduce the Risk of AD? . . . . 146 6. Speculations on When to Intervene with Estrogen-Replacement Therapy . . . . 149 7. Summary . . . . 151 References . . . . 152 8. The Use of PET to Evaluate Neural Plasticity and Repair in Rodent Brain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Harley I. Kornblum, Keith J. Tatsukawa, S. Thomas Carmichael, and Simon R. Cherry 1. Introduction . . . . 159 2. Principles of PET . . . . 160 3. Dedicated Lab Animal PET Scanners . . . . 163 4. Studies of Glucose Utilization and Neuroplasticity . . . . 164
159
xii
Contents
5. Other Potential Applications of PET in Neuroplasticity . . . . 168 6. Current Limitations for the Use of PET in Studies of Rodent CNS Plasticity . . . . 169 7. Comparison of PET to Other In Vivo Functional Imaging Modalities . . . . 171 References . . . . 173 9. Use of Estrogens as Neuroprotectants in Stroke and Alzheimer’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 James W. Simpkins, Shao-Hua Yang, and Yi Wen 1. Introduction . . . . 175 2. Estrogen Physiology and Pharmacology . . . . 176 3. Current Clinical Uses of Estrogens . . . . 176 4. Evidence for Efficacy of Estrogens and Estrogen Analogs in Stroke Neuroprotection . . . . 177 5. Transient Cerebral Ischemia as a Model for Alzheimer’s Disease Neuropathology . . . . 180 6. Need for Further Studies . . . . 183 References . . . . 183 10. Cell Death After Spinal Cord Injury: Basic Mechanisms and Potential Therapeutic Approaches to Promote Functional Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 Steven S. Schreiber 1. Apoptosis vs. Necrosis . . . . 190 2. Molecular and Cellular Mechanisms of Apoptosis . . . . 191 3. Evidence for Apoptosis After SCI . . . . 195 4. Prevention of Apoptosis After SCI . . . . 198 5. Conclusion . . . . 200 References . . . . 200 11. Motor Rehabilitation After Stroke: Novel Physical Treatment Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . Stefan Hesse and Cordula Werner 1. Introduction . . . . 209 2. General Principles of Physical Therapy . . . . 210
209
Contents
xiii
3. Special Aspects of Upper Limb Rehabilitation . . . . 210 4. Special Aspects of Gait Rehabilitation . . . . 218 5. Summary . . . . 224 References . . . . 224 12. Neural Bases for Rehabilitation After Stroke . . . . . . . . . 229 Randolph J. Nudo, Ann M. Stowe, Ines Eisner-Janowicz, and Numa Dancause 1. Animal Models of Stroke and Rehabilitation . . . . 230 2. Effects of Motor Injury on Behavioral Function: Recovery and Compensation . . . . 231 3. Neurophysiological Consequences of Motor Cortex Injury . . . . 233 4. Influence of Postinjury Motor Experience on Reorganization of Motor Maps . . . . 238 5. Anatomical Consequences Following Motor Cortex Injury . . . . 239 6. Influence of Postinjury Motor Experience on Neuroanatomical Plasticity . . . . 241 7. New Approaches to Maximize Neuroplasticity and Recovery After Stroke . . . . 243 8. Summary . . . . 245 References . . . . 245 13. Plasticity of Language Networks . . . . . . . . . . . . . . . . . . Cynthia K. Thompson 1. Neuroimaging Studies . . . . 256 2. Factors Related to Neuroplastic Processes . . . . 260 3. Conclusion . . . . 267 References . . . . 267
255
14. Locomotor Training-Induced Plasticity . . . . . . . . . . . . . . Bruce H. Dobkin 1. Introduction . . . . 271 2. The Distributed Locomotor System . . . . 271 3. Interventions to Drive Activity-Dependent Plasticity . . . . 275
271
xiv
Contents
4. Summary . . . . 276 References . . . . 277 15. Constraint-Induced Therapy for Functional Recovery After Brain Injury: Unraveling the Key Ingredients and Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 C. J. Winstein and M. G. Prettyman 1. Introduction . . . . 281 2. Essentials of Constraint-Induced Therapy: What is the Evidence? . . . . 284 3. Two Complementary Mechanisms Underlying CI Therapy: What is the Evidence? . . . . 310 References . . . . 317 16. Stem Cell Grafting and Spinal Cord Injury . . . . . . . . . . . Richard L. Benton, Gaby U. Enzmann, and Scott R. Whittemore 1. Introduction . . . . 329 2. Neural Stem Cell Transplantation in the Spinal Cord . . . . 330 3. Immortalized Cell Transplantation in the Spinal Cord . . . . 342 4. Bone Marrow and Stromal Stem Cell Transplantation in the Spinal Cord . . . . 347 5. Conclusions . . . . 351 References . . . . 351
329
17. Stem Cells and Huntington’s Disease . . . . . . . . . . . . . . . N. Popovic, A˚. Peterse´n, J.-Y. Li, and P. Brundin 1. Features of Huntington’s Disease . . . . 361 2. Symptoms of HD . . . . 364 3. Strategies in the Treatment of HD . . . . 365 4. Striatal Neural Transplantation in Nonhuman Primates . . . . 372 5. Clinical Transplantation Trials in HD . . . . 374 6. Risk Factors of Neural Transplantation . . . . 378
361
Contents
xv
7. Neurogenesis in HD . . . . 378 8. Conclusion . . . . 379 References . . . . 381 18. Plasticity, Cell Transplantation, and Brain Repair . . . . . . Maˇte´ D. Do¨bro¨ssy and Stephen B. Dunnett 1. Introduction . . . . 395 2. Cell Transplantation and Plasticity . . . . 396 3. Enhancing Graft Function . . . . 399 4. The Influence of Behavioral Experience and Training on Graft Function . . . . 404 5. The Choice of Cells . . . . 410 6. Conclusion: Plasticity and the Clinical Application of Cell Replacement Therapy . . . . 412 References . . . . 413
395
19. Brain-Implantable Biomimetic Electronics as Neural Prostheses to Restore Lost Cognitive Function . . . . . . . . 423 Theodore W. Berger, Ashish Ahuja, Patrick Nasiatka, Spiros H. Courellis, Gopal Erinjippurath, Ghassan Gholmieh, John J. Granacki, Min Chi Hsaio, Jeffrey LaCoss, Vijayaraghavan Srinivasan, Vasilis Z. Marmarelis, Dong Song, Armand R. Tanguay, Jr., and Jack Wills 1. Introduction . . . . 424 2. The System: Hippocampus . . . . 425 3. General Strategy and System Requirements . . . . 427 4. Proof-of-Concept: Replacement of the CA3 Region of the Hippocampal Slice with a Biomimetic Device . . . . 429 5. Experimental Characterization of Nonlinear Properties of the Hippocampal Trisynaptic Pathway . . . . 430 6. Nonlinear Dynamic Modeling of CA3 Input/Output Properties . . . . 432 7. Microcircuitry Implementation of the CA3 Input/Output Model . . . . 439 8. Conformal, Multisite Electrode Array Interface . . . . 447 9. System Integration: Current Status of the Hippocampal Slice Prosthetic . . . . 451 References . . . . 453
xvi
Contents
20. Behaviorally Relevant Neural Codes in Hippocampal Ensembles: Detection on Single Trials . . . . . . . . . . . . . . 459 John D. Simeral, Robert E. Hampson, and Sam A. Deadwyler 1. Introduction . . . . 459 2. Recording and Analysis Methods . . . . 460 3. Representation of Phase and Position . . . . 465 4. Detection and Dynamics of Representations . . . . 471 References . . . . 473 21. Recent Advances Toward Development of Practical Brain–Machine Interfaces . . . . . . . . . . . . . . . . . . . . . . . David J. Krupa 1. Brain–Machine Interfaces . . . . 477 2. Input BMIs . . . . 480 3. Output BMIs . . . . 484 4. Closed-Loop Input–Output BMIs . . . . 492 5. Conclusions . . . . 495 References . . . . 495 Index . . . . 503
477
Contributors
Ashish Ahuja Department of Electrical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Paul Bach-y-Rita Departments of Orthopedics and Rehabilitation Medicine, and Biomedical Engineering, University of Wisconsin, Madison, Wisconsin, U.S.A. Michel Baudry Department of Biology, Neuroscience Program, University of Southern California, Los Angeles, California, U.S.A. Richard L. Benton Kentucky Spinal Cord Injury Research Center (KSCIRC), Department of Neurology, University of Louisville School of Medicine, Louisville, Kentucky, U.S.A. Theodore W. Berger Department of Biomedical Engineering, Center for Neural Engineering, and Program in Neuroscience, University of Southern California, Los Angeles, California, U.S.A. Xiaoning Bi Department of Psychiatry & Human Behavior, University of California, Irvine, Irvine, California, U.S.A. Bemadett Boda Neuropharmacology, Centre Medical Universitaire, Geneva, Switzerland Roberta Diaz Brinton Department of Molecular Pharmacology and Toxicology and Program in Neuroscience, Pharmaceutical Sciences Center, University of Southern California, Los Angeles, California, U.S.A. xvii
xviii
Contributors
P. Brundin Section for Neuronal Survival, Wallenberg Neuroscience Center, Department of Physiological Sciences, Lund University, Lund, Sweden S. Thomas Carmichael Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A. Simon R. Cherry Department of Biomedical Engineering, University of California Davis, Davis, California, U.S.A. Marie-Francoise Chesselet Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A. Spiros H. Courellis Department of Biomedical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Maˇte´ D. Do¨bro¨ssy The Brain Repair Group, Cardiff School of Biosciences, Cardiff University, Cardiff, Wales, U.K. Numa Dancause Center on Aging and Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas, U.S.A. Sam A. Deadwyler Department of Physiology and Pharmacology, Wake Forest University School of Medicine, Winston-Salem, North Carolina, U.S.A. Bruce H. Dobkin Department of Neurology, Geffen School of Medicine, Neurologic Rehabilitation and Research Program, University of California Los Angeles, Los Angeles, California, U.S.A. Stephen B. Dunnett The Brain Repair Group, Cardiff School of Biosciences, Cardiff University, Cardiff, Wales, U.K. Ines Eisner-Janowicz Center on Aging and Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas, U.S.A. Gaby U. Enzmann Kentucky Spinal Cord Injury Research Center (KSCIRC), Department of Neurology, University of Louisville School of Medicine, Louisville, Kentucky, U.S.A.
Contributors
xix
Gopal Erinjippurath Department of Biomedical Engineering, University of Southern California, Los Angeles, California, U.S.A. Ghassan Gholmieh Department of Biomedical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Fernando Gomez-Pinilla Department of Physiological Science, University of California at Los Angeles, Los Angeles, California, U.S.A. John J. Granacki Departments of Biomedical Engineering and Electrical Engineering, Center for Neural Engineering, and Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A. Robert E. Hampson Department of Physiology and Pharmacology, Wake Forest University School of Medicine, Winston-Salem, North Carolina, U.S.A. Dorothy Harris Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A. Stefan Hesse Klinik Berlin, Department of Neurological Rehabilitation, Free University Berlin, Berlin, Germany Min Chi Hsaio Department of Biomedical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Pascal Jourdain Switzerland
Neuropharmacology, Centre Medical Universitaire, Geneva,
Harley I. Kornblum Department of Pharmacology, University of California at Los Angeles, School of Medicine, Los Angeles, California, U.S.A. David J. Krupa Department of Neurobiology, Duke University Medical Center, Durham, North Carolina, U.S.A. Jeffrey LaCoss Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A.
xx
Contributors
J.-Y. Li Section for Neuronal Survival, Wallenberg Neuroscience Center, Department of Physiological Sciences, Lund University, Lund, Sweden Vasilis Z. Marmarelis Departments of Biomedical Engineering and Electrical Engineering, and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Dominique Muller Neuropharmacology, Centre Medical Universitaire, Geneva, Switzerland Patrick Nasiatka Department of Electrical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Irina Nikonenko Neuropharmacology, Centre Medical Universitaire, Geneva, Switzerland Randolph J. Nudo Center on Aging and Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas, U.S.A. ˚ . Peterse´n Section for Neuronal Survival, Wallenberg Neuroscience Center, A Department of Physiological Sciences, Lund University, Lund, Sweden N. Popovic Section for Neuronal Survival, Wallenberg Neuroscience Center, Department of Physiological Sciences, Lund University, Lund, Sweden M. G. Prettyman Department of Biokinescology Physical Therapy, University of Southern California, Los Angeles, California, U.S.A. Veronique Riban Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A. Steven S. Schreiber Department of Neurology, and Anatomy and Neurobiology, UCI College of Medicine, Irvine; and Neurology Section, VA Long Beach Healthcare System, Long Beach, California, U.S.A. Marc Shomer Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A.
Contributors
xxi
John D. Simeral Department of Physiology and Pharmacology, Wake Forest University School of Medicine, Winston-Salem, North Carolina, U.S.A. James W. Simpkins Department of Pharmacology and Neuroscience, University of North Texas Health Science Center, Fort Worth, Texas, U.S.A. Dong Song Department of Biomedical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A. Vijayaraghavan Srinivasan Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A. Ann M. Stowe Center on Aging and Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas, U.S.A. Armand R. Tanguay, Jr. Departments of Electrical Engineering, Biomedical Engineering, and Materials Science; Center for Neural Engineering; and Program in Neuroscience, University of Southern California, Los Angeles, California, U.S.A. Keith J. Tatsukawa Department of Pediatrics, University of California at Los Angeles, School of Medicine, Los Angeles, California, U.S.A. Cynthia K. Thompson Department of Communication Sciences and Disorders, and Neurology, Cognitive Neurology and Alzheimer’s Disease Center, Northwestern University, Evanston, Illinois, U.S.A. Shoshanna Vaynman Department of Physiological Science, University of California at Los Angeles, Los Angeles, California, U.S.A. Yi Wen Department of Pharmacology and Neuroscience, University of North Texas Health Science Center, Fort Worth, Texas, U.S.A. Cordula Werner Klinik Berlin, Department of Neurological Rehabilitation, Free University Berlin, Berlin, Germany Scott R. Whittemore Kentucky Spinal Cord Injury Research Center (KSCIRC), Department of Neurology, University of Louisville School of Medicine, Louisville, Kentucky, U.S.A.
xxii
Contributors
Jack Wills Department of Electrical Engineering and Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A. C. J. Winstein Department of Biokinescology Physical Therapy, University of Southern California, Los Angeles, California, U.S.A. Shao-Hua Yang Department of Pharmacology and Neuroscience, University of North Texas Health Science Center, Fort Worth, Texas, U.S.A.
1 Memory: From Molecular Mechanisms to Clinical Applications Michel Baudry Department of Biology, Neuroscience Program, University of Southern California, Los Angeles, California, U.S.A.
1.
INTRODUCTION
Memory is a phenomenon that has fascinated scientists as well as writers, philosophers, and common people since the beginning of mankind. As has been repeatedly underlined, memory is both incredibly powerful and fragile; we all have memories that date back to our infancy, and at the same time, we can forget what we did yesterday or where we parked our car this morning. For the biologist, memory is a remarkable feature of the nervous system, and its unique properties have attracted the interest of scientists from every field, ranging from molecular biologists to psychologists and even to physicists and mathematicians. Ever since the identification of the synaptic contacts as the sites of communication between neurons, the prevalent notion for the cellular mechanisms underlying the storage of information in the CNS has been some form of activity-dependent modification of synaptic efficacy. This idea, already present in Ramon y Cajal’s writing, was further popularized by Hebb (1), and, many years after Hebb’s death led to the now famous concept of hebbian synapse (2,3). Unfortunately, the search for the cellular mechanisms of memory got a bad reputation following a series of experiments attempting to demonstrate that, while the genes held the memories of the species, proteins were the repositories for the memories of the individuals. These attempts resulted in the infamous 1
2
Baudry
transfer of memory experiments, in which extracts from trained worms were fed to naive worms or brain extracts from trained mice were injected into naive mice (4–7). While experimental psychologists were making great progress at providing a classification of various forms of memory (8), the search for the cellular mechanisms remained limited to the study of some simple systems such as the sensitization of the gill withdrawal reflex in Aplysia (9) or the acquisition of classical conditioning in Hermissenda (10). However, an important discovery, published in 1973, forever changed our views of the cellular mechanisms of learning and memory (11). After failing to find synaptic plasticity in neocortex as a Ph.D. student, Bliss (12) moved to Per Andersen’s laboratory. There, Bliss and Lomo discovered a form of activitydependent modification of synaptic plasticity at hippocampal synapses that exhibited several features expected of a cellular mechanism of learning and memory. They called it long-lasting potentiation, but it soon became known and referred to as long-term potentiation or LTP. Over the last 30 years, this phenomenon has become the epicenter of the discussions related to the mechanisms of learning and memory, and, as we will discuss below, has generated a wealth of information regarding not only the basic mechanisms of learning and memory but also the first rationale design of new molecules directed at improving learning disabilities resulting from diseases or old age. In this chapter, we will first review the features of LTP, and discuss why these features are well suited for a learning mechanism. We will then summarize the current hypotheses that have been proposed to explain LTP, and to account for the links between LTP and learning and memory (Fig. 1). This will be followed by a review of the properties of positive AMPA receptor modulators, the ampakines, and of the studies indicating that they facilitate LTP formation and improve learning in a variety of tasks. The review will conclude by a brief survey of the various clinical trials currently evaluating these molecules for selected indications.
2.
LTP: A CELLULAR MECHANISM FOR LEARNING AND MEMORY
As often the case in experimental sciences, LTP was discovered by serendipity when Bliss and Lomo were trying to study the characteristics of the perforant path input to the dentate gyrus in anesthetized rabbits (see Ref. 13). In order to maintain the size of the evoked responses, they noticed that when they delivered brief episodes of high frequency stimulation, evoked responses would get bigger and remained bigger for long periods of time. Thus, the expression ‘‘long lasting potentiation’’ was created; when similar experiments were performed in acute hippocampal slices and resulted in a similar long-lasting enhancement of excitatory synaptic transmission (14–16), the expression long-term potentiation (LTP) took over, and, with
Memory: From Molecular Mechanisms to Clinical Applications
3
Figure 1 Links between LTP, learning and memory, and cellular processes. The relationships between LTP and learning and memory have been generally explored with three complementary approaches. The features of various forms of LTP have been compared with those of various forms of learning and memory. Pharmacological tools derived from the molecular=cellular studies of LTP have been applied on various forms of learning and memory. Finally, targeted mutations of specific genes identified from molecular studies of LTP have been tested for their effects on various forms of learning and memory.
a few exceptions, is now universally used to design a long-lasting increase in monosynaptic transmission resulting from a brief train of high frequency stimulation, typically 100 Hz. 2.1.
Features of LTP
As already mentioned, three of the initial features of the LTP phenomenon were immediately stressed as potential links with memory processes; rapid induction, long duration and prominent expression in the hippocampus, a structure that the work of Scoville and Milner (17) had previously demonstrated to be critical in memory formation. Shortly after came the demonstration that LTP exhibited some forms of associativity that could
4
Baudry
account for some of the associative properties of learning and memory (2,18), although this point has recently been challenged (19). It was also initially recognized that LTP was prominently expressed at excitatory synapses using glutamate as a neurotransmitter. The existence of an LTP-like phenomenon at synapses using a different neurotransmitter such as GABA has been difficult to demonstrate convincingly until now (see Refs. 20–22). For years, an intense debate opposed the supporters of a presynaptic hypothesis for the site of expression of LTP to those of the postsynaptic localization of the changes in synaptic transmission underlying LTP (see Refs. 23–25). For the former, increased glutamate release or increase in the probability of release was the likely explanation for increased transmission (26). For the later, increased number of glutamate receptors or increased currents gated by glutamate receptors were more likely to account for the characteristics of LTP. Based on a number of studies directed at understanding the regulation of the properties of glutamate receptors, we proposed in 1984 that LTP was due to changes in glutamate receptors triggered by the activation of a calcium-dependent protease, calpain (27). In our model, we also discussed the possibility that calpain activation resulted in modifications of the structures of the dendritic spines, making them more stable and providing for the long duration of LTP. As we will discuss later, this idea is still one of the most widely accepted idea for the cellular mechanisms responsible for the increased synaptic transmission in LTP. At the time we proposed our LTP model, little was known regarding the characteristics of the glutamate receptors. Dramatic progress were made in the mid-1980s following the discovery that glutamate receptors could be subdivided into three classes of receptors, the N-methyl-d-aspartic acid (NMDA) receptors, the a-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors and the kainate receptors (28,29), and that a specific antagonist of the NMDA receptors, 2-amino-phosphonovalerate (APV) did not affect basal synaptic transmission but completely prevented LTP induction in CAl (30). This apparently paradoxical result was beautifully explained by Mayer et al. (31) and Nowack et al. (32), who independently found that magnesium blocked NMDA receptor channels in a voltage-dependent manner. When it was later on found that NMDA receptor channels were also calcium channels (33,34), most of the elements responsible for LTP induction were understood. These successive discoveries provided a general scheme for LTP formation, which is still currently valid (Fig. 2). As it was thus clear that an initial influx of calcium triggered by activation of NMDA receptors was necessary and sufficient to generate LTP (35–38), a myriad of calcium-dependent processes and enzymes have been proposed to play some role in LTP formation. The exact roles played by these different processes are still debated and will be discussed in later sections. In any event, this framework provided a rich field for pharmacologists and molecular biologists to use their tools to attempt at dissecting out the roles of these processes in LTP, and thereby in learning and memory.
Memory: From Molecular Mechanisms to Clinical Applications
5
Figure 2 Schematic representation of the elements necessary and sufficient to induce LTP. It is now generally agreed that activation of NMDA receptors and the resulting increase in intracellular calcium are the necessary and sufficient events to induce LTP. The number of AMPA receptors present at the surface of dendritic spines is regulated by both a constitutive insertion of receptors and a regulated insertion, which could be the target of increased calcium concentration.
2.2.
Pharmacology of LTP
As mentioned above, a very large number of biochemical events have been proposed to participate in one way or another in LTP induction or expression over the last 20 years (39). Among these, several calcium-dependent processes have been identified (Fig. 3). Activation of the calcium-dependent protease, calpain, results in modifications of the structure of several cytoskeletal proteins and cell adhesion molecules and therefore is ideally suited to
6
Baudry
Figure 3 Various calcium-dependent processes triggered during LTP induction. Several calcium-dependent enzymes and their main substrates have been proposed to participate in LTP.
promote modifications in the structure of synaptic contacts (40). Activation of several protein kinases, and in particular, calcium=calmodulin kinase type II (CamKII), is also a key process to phosphorylate numerous proteins in synaptic contacts, including glutamate receptors (Table 1) (41). Autophosphorylation of CamKII might also provide for a form of short-term plasticity that exceeds the duration of the LTP-inducing stimulus (42,43). Finally, activation of phospholipases provides for both the synthesis of lipid breakdown products with various signaling properties as well as for the breakdown of membrane lipids, an essential element for structural modifications (44). It is interesting to point out that most of these cell biological processes are not unique to neurons but are present in many cell types where they participate in regulation of structure and function. 2.3.
Genetic Manipulations of LTP
A much-debated issue in the LTP literature concerns the role of gene expression and protein synthesis and the mechanisms linking LTP induction to the
Memory: From Molecular Mechanisms to Clinical Applications
Table 1
7
a-CamKII and Synaptic Plasticity
a-CamKII inhibitors block LTP induction in CA1 a-CamKII ko have impaired LTP and LTD in hippocampus and neocortex a-CamKII ko are severely impaired in Morris water maze; no problem to learn visible platform a-CamKII ko exhibit abnormal plasticity in visual cortex and somatosensory cortex Single-site mutation (Thr286!Ala) impairs LTP and spatial learning Active a-CamKII ki also disrupts LTP and learning See Ref. 43 for a recent review of the molecular basis for a role CamKII in LTP and learning and memory.
protein synthesis machinery of neurons. Numerous studies have attempted to identify genes activated as a result of LTP induction (45–47). In most cases, several transcription factors are induced as a result of trains of electrical activity producing LTP. However, it has been quite difficult to further establish the relationship between gene induction with mechanisms underlying LTP expression and maintenance. Advances in molecular biology techniques, and in particular, the possibility of knocking out or knocking in specific genes have produced a large number of mutant mice with various ‘‘learning and memory’’ impairments, in addition to problems with LTP induction or maintenance (Table 2) (48,49). In addition, the mutations that have been generated often affect enzymes, receptor=channels, or other cellular elements, which participate in ‘‘normal’’ cellular functions, and are likely to modify even in a subtle way, various aspects of brain structure and function. The limitation on the role of gene expression in synaptic plasticity resulting from the cell-wide modifications it generally produces would be seriously eliminated if there were mechanisms coupling gene expression and local translation of the corresponding mRNAs. For local protein synthesis to take place requires the dendritic targeting of mRNA and there are now a number of examples of such mRNAs (50–52). Furthermore, there is also experimental evidence for links between synaptic transmission and regulation of local protein synthesis as activation of glutamate metabotropic receptors in synaptoneurosomes has been shown to rapidly stimulate protein synthesis (53). There are therefore two types of mechanisms by which synaptic activity can produce localized modifications of synaptic properties restricted to activated synapses. In the first one, a small population of mRNAs is constitutively present in dendrites, and in response to an appropriate signal, these mRNAs are locally translated, and the newly synthesized proteins are incorporated into the activated synapses. Note that this process is rapid and does not require gene expression, as it only depends on constitutively present mRNAs. In the second one, some locally generated signals are transmitted to the cell nucleus, trigger gene transcription, and the synthesis of mRNAs.
8
Baudry
Table 2
Effects of Gene Manipulation on Plasticity and Learning and Memory
Gene modification
Effect on synaptic plasticity
Effect on learning
a-CamKII Fyn PKCg MGluRl PKA GAP-43 overexpression t-PA ko
Impaired Impaired Impaired Impaired Normal Increased LTP Decreased hippocampal LTP Decreased striatal LTD Increased LTP Decreased CA1 LTP Decreased LTP
Impaired learning Impaired learning Impaired learning Impaired learning Normal learning Increased learning Impaired learning
t-PA overexpression BDNF overexpression BDNF ko HD mutation NR2b overexpression Bcl-2 overexpression Human APP mutation HO-1 overexpression Kvl.l antisense Treatment Nociceptin receptor ko N-CAM ko
Increased plasticity Decreased LTP No effect on LTP Impaired LTP Impaired LTP
Increased learning Impaired learning Normal learning Cognitive impairment Increased learning Increased learning Decreased learning Impaired learning Decreased learning Impaired learning Impaired learning
These mRNAs are targeted to the dendrites and locally translated at the appropriate site. Note that this requires the existence of a signal or signals that remain present at activated synapses for quite some time as the transcription of genes, and their targeting to the dendrites necessitate a significant period of time. This mechanism is somewhat related to the synaptic tagging described by Frey and Morris (54). In their model, these authors propose that LTP induction is accompanied by the activation of signals they call ‘‘synaptic tags,’’ which interact with plasticity-related newly synthesized proteins and stabilize the formation of LTP. Clearly, much more work is needed to clarify the exact mechanisms involved in this type of regulation, and to determine the types of mRNAs present in the dendrites, the mechanisms involved in targeting them to dendrites and the signals responsible for tagging activated synapses. Finally, the types of proteins involved and their roles in regulating synaptic structure and function need to be further investigated. 2.4.
Is LTP a Cellular Mechanism for Learning and Memory?
Long-term potentiation is the most widely proposed mechanism of memory storage in the hippocampus and neocortex. Although this issue is still debated, evidence supporting this hypothesis comes from a variety of
Memory: From Molecular Mechanisms to Clinical Applications
Table 3
9
Pharmacology of LTP and of Hippocampal-Dependent Learning
Drug
Target
LTP
Learning
AP-5 Ampakines Benzodiazepines Leupeptin Protein synthesis inhibitors
NMDAr AMPAr GABAr Calpain ?
Blockade Facilitation Impaiment Blockade Blockade
Impairment Facilitation Impairment Impairment Blockade
experimental data and theoretical models (23,55–63). Long-term potentiation is prevalent in hippocampal and cortical networks and exhibits many properties required for a large capacity information storage device: rapid induction, associativity, long duration, and links with brain rhythms [in particular, the theta rhythm (64)]. Pharmacological manipulations or gene mutations interfering with LTP also interfere with various forms of learning and memory, while pharmacological agents facilitating LTP formation also facilitate learning (Table 3). Finally, incorporating LTP-based rules in biologically realistic neuronal networks produces large capacity storage devices (65). 3. CURRENT HYPOTHESES FOR LTP=LTD MECHANISMS 3.1. Induction The cellular events that are necessary and sufficient to produce LTP are now well understood (57,66). What is required is sufficient postsynaptic depolarization (ergo the high frequency bursts of stimulation) to relieve the NMDA receptor=channel from a voltage-dependent magnesium blockade and to produce a postsynaptic influx of calcium at the right place and of the right amplitude. The NMDA receptor thus functions as an AND gate (presynaptic release of neurotransmitter and postsynaptic depolarization) and accounts for most features of LTP, such as associativity, calcium dependency, regulation by the degree of inhibition, etc. . . . It should be noted, however, that some forms of LTP do not require NMDA receptor activation, and thus there may be multiple forms of LTP as well as multiple ways of triggering the biochemical cascades leading to LTP (67). 3.2.
Maintenance
It was clear from the outset that LTP could be due to changes in either the release characteristics of the presynaptic terminals or the postsynaptic responsiveness to the neurotransmitter or some combination of both factors. As some reports suggested that glutamate release was enhanced following LTP induction (68), the postsynaptic localization of NMDA
10
Baudry
receptors imposed the existence of some retrograde signals that could link LTP induction to LTP expression mechanisms. Several such signals were proposed such as arachidonic acid, nitric oxide (NO), or carbon monoxide (CO) (69). On the other hand, many reports have also indicated several postsynaptic modifications including changes in the properties of another type of glutamate receptors, the a-amino-3-hydroxy-5-methylisoxazole propionic acid (AMPA) receptors, and in the morphology and number of dendritic spines (70–72). There is now some general agreement that LTP expression is due to several modifications of the distribution of AMPA receptors and the structure of synaptic contacts (57). 3.2.1. Presynaptic Mechanisms An increase in transmitter release obviously could account for an increase in synaptic transmission; it is generally admitted, for instance, that short-term potentiation is due to increased transmitter release (73). Biochemical and electrophysiological evidence has been obtained in support of the hypothesis that increased transmitter release accounts for LTP (68). Glutamate is the neurotransmitter used by synapses exhibiting LTP in various hippocampal pathways, and increased glutamate levels have been found after LTP induction in perfusates obtained with push–pull cannulas implanted in the dentate gyrus (74). Increased glutamate release was also found after LTP induction in hippocampal slices and in synaptosomes prepared from hippocampus of rats in which LTP had been induced. Since the induction of LTP involves the postsynaptic activation of NMDA receptors, a retrograde signal has been postulated that relays the postsynaptic activation to the presynaptic terminal. Arachidonic acid was initially proposed to be such a retrograde signal since activation of NMDA receptors stimulates phospholipase A2 (PLA2), and PLA2 inhibitors prevent LTP formation (75). Arachidonic acid could also provide a link with the presynaptic machinery involved in the regulation of transmitter release because it activates protein kinases that have been shown to participate in phosphorylation reactions important for transmitter release. Nitric oxide as well as carbon monoxide were later proposed to be retrograde signals although the evidence for this conclusion remains highly controversial (57,76). While this mechanism could conceivably provide a satisfactory explanation for a short-lasting enhancement of transmitter release, in its present version it does not account for a longlasting increase. 3.2.2. Quantal Analysis Quantal analysis of synaptic transmission (77) is an approach that, in principle, could provide an unambiguous answer to the question of the locus of the changes underlying LTP. It uses the intrinsic variability of transmitter release and statistical methods to determine the parameters generally thought to govern synaptic transmission, i.e., the probability of
Memory: From Molecular Mechanisms to Clinical Applications
11
release, the number of release sites, and the elementary size of a postsynaptic response elicited by a quantum of neurotransmitter. Several groups have reported results obtained with quantal analysis in field CA1 before and after LTP induction in hippocampal slices. Two groups observed an increase in quantal content (78,79), whereas one group observed an increase in quantal size (80), and another one both (81). In addition to this ambiguity in the results, quantal analysis requires a number of assumptions that might not be satisfied at hippocampal synapses. In particular, several studies have indicated the existence of synapses with functional NMDA receptors but few, if any, functional AMPA receptors (the so-called silent synapses) (82,83). Transformation of silent synapses into active synapses by LTP-inducing stimulation could account for the inconsistencies in results from quantal analysis (57). Finally, increased transmitter release could be accounted for by an increase in the number of synapses. Anatomical studies using quantitative electron microscopy have provided evidence that the number of certain categories of synapses, in particular sessile synapses, is increased in field CA1 following LTP induction (71,72). Although an increased number of synapses could explain the duration of LTP and account for its maintenance, it is not yet clear whether the increase in sessile synapses represents the formation of new synapses or the transformation of existing synapses by shape modifications (84,85) (see Nikonenko et al., this volume not available). 3.2.3. Changes in Spine Electrical Properties Since their discovery, there has been much speculation concerning the roles of dendritic spines in synaptic transmission and the possibilities that synaptic plasticity could be due to alterations in their electrical properties (86,87). Because of their shape—large heads relative to long, narrow necks—they are considered to be high-resistance elements coupling voltage changes at the synapses with voltage changes in dendritic shafts. Based upon calculation and computer simulation, it has been proposed that decreased neck resistance could account for increased synaptic responses (88), which could affect the AMPA receptors more specifically than the NMDA receptors (89). Such a decrease in spine resistance could be due to an increase in neck dimension, and anatomical evidence indicates that LTP is accompanied by an increase in spines with wider and shorter necks. The recent discoveries of active elements (channels or pumps) in dendritic spines provide alternative means of modifying their electrical properties (90), although there is no evidence that these elements are modified following LTP induction. 3.2.4. Changes in Receptor Properties One of the most important characteristics of LTP is that it is expressed by an increase in the component of the synaptic response resulting from the activation of the AMPA receptors with little changes in the component
12
Baudry
generated by the activation of NMDA receptors. Thus an alteration of the properties of the AMPA receptors has been proposed as a potential maintenance mechanism. Several studies have shown that both the ligand binding and the ionic conductance properties of the AMPA receptors are affected by a variety of manipulations. In particular, changes in the lipid environment of the AMPA receptors modify its affinity for agonists, an observation that could account for the involvement of PLA2 in LTP. More recently, the molecular biology of the AMPA receptors has provided exciting results concerning the nature and properties of these receptors. AMPA receptors belong to a family of receptors encoded by at least four related genes, each existing in two closely similar versions generated by alternative splicing of the genes, and designated flip and flop. It has been proposed that AMPA receptors exist as oligomers composed of a number of possible arrangements of flip and flop elements, with the flip elements providing larger currents than the flop (91). Recent evidence has suggested that LTP could result from a change in the composition of receptor subunits or in changes in the rates of receptor insertion or internalization in the synaptic membrane from a subsynaptic pool of receptors (92). Another dimension related to the stabilization of the changes underlying LTP maintenance has recently been included to the search for the mechanisms of LTP. It is now apparent that LTP stabilization involves adhesion molecules and that there is a time window following LTP induction during which processes leading to LTP consolidation can be reversed, thus producing depotentiation (93). There has been considerable discussion to determine whether depotentiation represents the same process as LTD of synaptic transmission. Although both processes share a number of features, there is not yet a general consensus as to the mechanisms and the significance of these forms of synaptic plasticity. 3.3.
Stabilization
It has only recently been recognized that the extracellular matrix plays an equally important role in regulating synaptic morphology and properties. Antibodies to N-CAM and peptide inhibitors of integrin receptors inhibit the formation of stable LTP (94,95). In several current models of LTP, stabilization of the modifications triggered during the induction phase of LTP is accomplished first by the disruption of the adhesive properties of synaptic contacts, possibly mediated through the proteolysis of CAM molecules, followed by the activation of integrins. In this way, synaptic contacts undergo a cycle of destabilization resulting from intracellular as well as extracellular proteolysis of proteins contributing to the morphology of synaptic contacts, followed by restabilization of a new synaptic structure implicating integrin activation and extracellular matrix components (93). Recent data have confirmed that a3 and a5 integrins are essential members of the integrin family to consolidate LTP and memory (96).
Memory: From Molecular Mechanisms to Clinical Applications
13
4. AMPAKINES, LTP, AND MEMORY 4.1. Ampakine Definition, Structure, Classes Following the discovery that aniracetam increased AMPA receptormediated synaptic transmission in CA1 but not in CA3 (97), a number of laboratories have explored the possibility that small molecules could modulate AMPA receptor function in a way similar to the way benzodiazepines modulate GABAA receptor-mediated inhibitory transmission. A number of different structural classes of molecules have thus been identified, including some benzothiadiazides, and biarylpropylsulfonamides, which are positive modulators of AMPA receptors (Fig. 4) (98–104). The term Ampakine2 was introduced by Gary Lynch to designate molecules able to allosterically modulate the function of AMPA receptors. It was soon realized that these agents modulate AMPA receptors through several mechanisms: Desensitization Blockade. Compounds like cyclothiazide increase AMPA receptor-mediated responses by blocking the desensitization which normally follows receptor activation by agonists; although these compounds have a very large effects on responses elicited by exogenous application of
Figure 4 Structure of various positive modulators of AMPA receptors. A number of positive AMPA receptor modulators have been synthesized and shown to have various effects of synaptic transmission, network function and behaviors.
14
Baudry
agonists, they have a very limited effect on AMPA receptor-mediated synaptic responses, as desensitization does not appear to play a role in determining the time-course of synaptic responses, at least at low frequency stimulation (105,106). Activation Facilitation or Deactivation Inhibition. More typical ampakines appear to increase AMPA receptor-mediated responses by either increasing the rate of activation of the receptor following agonist binding to its recognition site or by slowing the rate of inactivation. In contrast to cyclothiazide, these ampakines do increase the amplitude and=or half-width of AMPA receptor-mediated synaptic responses. Detailed analysis of the effects of CX516 and CX546 has suggested that there might be two separate sites on the AMPA receptors responsible for these distinct effects of ampakines on the kinetics of the responses (107). Recent studies have identified the sites within the receptor where ampakines bind, and suggest that these sites are localized at the interface between subunits constituting the receptors (108,109). 4.2.
Physiological Effects
As mentioned above, the main feature of AMPA receptor modulators is to increase the amount and=or the duration of depolarization produced by AMPA receptor activation. These dual effects of ampakines have interesting consequences for information processing and storage. First, it facilitates the transfer of information across networks, and in particular, tri-synaptic responses in hippocampus are much more sensitive to ampakines than mono-synaptic responses (110). Second, it facilitates LTP induction, as short bursts of electrical stimulation are more likely to produce depolarization large enough to overcome the voltage-dependent magnesium blockade of NMDA receptors (111). Finally, by a mechanism that is not well elucidated, semi-chronic ampakine treatment of cultured hippocampal slices or in vivo administration of ampakines results in increase in BDNF expression (112,113). These effects have raised the possibility of using ampakines to stimulate endogenous production of neurotrophins for long-term treatment of neurodegenerative diseases. 4.3.
Behavioral Effects
A number of studies have now shown the beneficial effects of ampakines and other AMPA receptor modulators in a variety of learning tasks. Thus, IDRA-21 improved rat performance in the water maze learning task (114), while CX516 improved performance in a two-odor discrimination task in rats and in delayed non-match-to-sample performance in rats (115–118). CX516 also enhanced memory encoding in several tests in humans and in delayed recall in aged humans (119–121). Interestingly, ampakines were also shown to have synergistic effects with typical and
Memory: From Molecular Mechanisms to Clinical Applications
15
atypical antipsychotic in methamphetamine-induced hyperactivity in rats (122,123). Finally, ampakines also showed some antidepressant effects in a submissive model of depression in rats (124). All these results strongly supported the notion that ampakine-mediated facilitation of AMPA receptor function and thereby of information encoding is the key mechanism underlying the behavioral effects of this new type of drugs and were critical elements in initiating a number of clinical trials sponsored by various pharmaceutical companies. 5. CLINICAL TRIALS OF AMPAKINES 5.1. Alzheimer’s Disease Alzheimer’s disease (AD) is a neurodegenerative disease characterized by loss of cognitive function, massive brain atrophy, and ultimately death. Currently, very few treatments have been available and they have mostly targeted the cholinergic system based upon the assumption that degeneration of cholinergic inputs to hippocampus and neocortex was critical for the loss of cognition in AD patients. While a plethora of drugs are currently being considered and tested for targeting various aspects of the pathology (125,126), the ampakine CX516 was actually tested in a small pilot study. Positive results were obtained and led to the development of a much larger scale study for another indication characterized by loss of memory function, mild cognitive impairment, or MCI (see the following section). 5.2.
Mild Cognitive Impairment
Mild cognitive impairment has recently been recognized as a clinical disease that affects a large portion of the aging population. Mild cognitive impairment affects an estimated 2.5–4 million people in the United States, with an equivalent number in Western Europe (127). MCI patients exhibit difficulty with recent memory tasks when compared to age-matched people with similar educational background. More importantly, between 60% and 80% of people with MCI develop Alzheimer’s disease over a five-year period. CX516 is currently under investigation in a world-wide clinical investigation and the results of this trial should become available by the end of the year. 5.3.
Schizophrenia
Schizophrenia is characterized by positive symptoms such as hallucinations and paranoia, negative symptoms such as social withdrawal, and cognitive deficits. Current drug treatments with typical and atypical neuroleptics are mostly directed at positive treatments, and there is a clear need for new treatments directed at other symptoms of the disease. Furthermore, in addition to the hypothesis that a hyperfunctioning of some dopaminergic
16
Baudry
systems plays a critical role in the etiology of schizophrenia (128), several reports have proposed that the hypofunctioning of certain glutamatergic pathways might be responsible for some of the symptoms (129,130). As mentioned previously, CX516 was found to attenuate methamphetamineinduced hyperactivity, an animal model widely used to evaluate antipsychotic compounds. This has led to the idea that positive modulators of AMPA receptors could be useful as potential treatment for schizophrenia (131); in particular, clinical testing of CX516 in combination with clozapine was initiated. The results of the small study indicated a significant improvement in both cognitive and negative symptoms of schizophrenia (132).
6.
CONCLUSIONS
Although some of the elements of the puzzle constituting human memory are still missing, the main elements have been identified over the last 25 years. Thus, it is now clear that the remarkable synaptic plasticity discovered by Bliss and Lomo 30 years ago has been an Arianna thread that has led scientists to decipher the machinery used by a large number of synapses to modify their characteristics as a function of activity. There is little doubt now that the cellular constituents responsible for this activitydependent synaptic plasticity have been identified. The critical elements appear to be the main subtypes of glutamate receptors, i.e., the AMPA and the NMDA receptors. Attempts to use modulators of NMDA receptors as memory enhancers have not been very successful yet. On the other hand, positive modulators of AMPA receptors, the so-called Ampakines are rapidly moving from the laboratory to the clinics (133). Recent reports of their clinical successes not only justify the efforts devoted by scientists involved in their discovery, but also strengthen the idea that it is now possible to design rational cognitive enhancers and to start to address the issues of remedying for cognitive impairment.
REFERENCES 1. 2. 3. 4.
5.
Hebb DO. The Organization of Behavior. New York: Wiley, 1949. Levy WB, Steward O. Synapses as associative memory elements in the hippocampal formation. Brain Res 1979; 175(2):233–245. Brown TH, Kairiss EW, Keenan CL. Hebbian synapses: biophysical mechanisms and algorithms. Annu Rev Neurosci 1990; 13:475–511. Babich FR, Jacobson AL, Bubash S. Cross-species transfer of learning: effect of ribonucleic acid from hamsters on rat behavior. Proc Natl Acad Sci USA 1965; 54(5):1299–1302. Byrne WL, Samuel D, Bennett EL, Rosenzweig MR, Wasserman E. Memory transfer. Science 1966; 153(736):658–659.
Memory: From Molecular Mechanisms to Clinical Applications 6. 7. 8. 9.
10. 11.
12. 13.
14.
15.
16.
17. 18.
19. 20.
21.
22.
23. 24.
17
Jacobson AL, Fried C, Horowitz SD. Planarians and memory. Nature 1966; 209(23):599–601. Lesk AM. Transfer experiments: a plea for tolerance. Science 1968; 159(818):922–923. Squire LR. Mechanisms of memory. Science 1986; 232(4758):1612–1619. Carew TJ, Castellucci VF, Kandel ER. An analysis of dishabituation and sensitization of the gill-withdrawal reflex in Aplysia. Int J Neurosci 1971; 2(2): 79–98. Alkon DL, Sakakibara M, Naito S, Heldman E, Lederhendler I. The role of neurochemical modulation in learning. Neurosci Res 1986; 3(6):487–497. Bliss TV, Lomo T. Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol 1973; 232(2):331–356. Bliss TV. A journey from neocortex to hippocampus. Philos Trans R Soc Lond B Biol Sci 2003; 358(1432):621–623. Andersen P. LTP-An exciting continuing Saga. In: Baudry M, Davis JL, eds. Long-Term Potentiation: A Debate of Current Issues. Cambridge, MA: MIT Press, 1991:xiii–xvii. Lynch GS, Gribkoff VK, Deadwyler SA. Long term potentiation is accompanied by a reduction in dendritic responsiveness to glutamic acid. Nature 1976; 263(5573):151–153. Andersen P, Sundberg SH, Sveen O, Wigstrom H. Specific long-lasting potentiation of synaptic transmission in hippocampal slices. Nature 1977; 266(5604):736–737. Dunwiddie T, Lynch G. Long-term potentiation and depression of synaptic responses in the rat hippocampus: localization and frequency dependency. J Physiol 1978; 276:353–367. Scoville WB, Milner B. Loss of recent memory after bilateral hippocampal lesions. J Neuropsychiatry Clin Neurosci 1957; 12(1):103–113. McNaughton BL, Douglas RM, Goddard GV. Synaptic enhancement in fascia dentata: cooperativity among coactive afferents. Brain Res 1978; 157(2):277–293. Shors TJ, Matzel LD. Long-term potentiation: what’s learning got to do with it? Behav Brain Sci 1997; 20(4):597–614; discussion 614–555. Oda Y, Kawasaki K, Morita M, Korn H, Matsui H. Inhibitory long-term potentiation underlies auditory conditioning of goldfish escape behaviour. Nature 1998; 394(6689):182–185. Caillard O, Ben-Ari Y, Gaiarsa JL. Long-term potentiation of GABAergic synaptic transmission in neonatal rat hippocampus. J Physiol 1999; 518(Pt 1):109–119. Kawaguchi SY, Hirano T. Signaling cascade regulating long-term potentiation of GABA(A) receptor responsiveness in cerebellar Purkinje neurons. J Neurosci 2002; 22(10):3969–3976. Baudry M, Davis JL, eds. Long-term Potentiation. Vol. III. Cambridge: MIT Press, 1996. Baudry M, Davis JL, eds. Long-term Potentiation: a Debate of Current Issues. Cambridge, MA: MIT Press, 1991.
18
Baudry
25. Baudry M, Davis JL, eds. Long-term Potentiation. Vol. II. Cambridge: MIT Press, 1994. 26. Bliss TV, Douglas RM, Errington ML, Lynch MA. Correlation between long-term potentiation and release of endogenous amino acids from dentate gyrus of anaesthetized rats. J Physiol 1986; 377:391–408. 27. Lynch G, Baudry M. The biochemistry of memory: a new and specific hypothesis. Science 1984; 224:1057–1063. 28. McLennan H. Receptors for the excitatory amino acids in the mammalian central nervous system. Prog Neurobiol 1983; 20(3–4):251–271. 29. Dingledine R, Boland LM, Chamberlin NL, Kawasaki K, Kleckner NW, Traynelis SF, Verdoorn TA. Amino acid receptors and uptake systems in the mammalian central nervous system. Crit Rev Neurobiol 1988; 4(1):1–96. 30. Collingridge GL, Kehl SJ, McLennan H. The antagonism of amino acidinduced excitations of rat hippocampal CA1 neurones in vitro. J Physiol Lond 1983; 334(1):19–31. 31. Mayer ML, Westbrook GL, Guthrie PB. Voltage-dependent block by Mgþþ of NMDA responses in spinal cord neurones. Nature 1984; 309:261–263. 32. Nowack L, Bregestovski P, Ascher P, Herbert A, Prochiantz A. Magnesium gates glutamate-activated channels in mouse central neurones. Nature 1984; 307:462–465. 33. McDermott AB, Mayer ML, Westbrook GL, Smith SJ, Barker JL. NMDAreceptor activation increases cytoplasmic calcium concentration in cultured spinal cord neurons. Nature 1986; 321:519–522. 34. Jahr CE, Stevens CF. Glutamate activates multiple single channel conductances in hippocampal neurons. Nature 1987; 325(6104):522–525. 35. Lynch G, Larson J, Kelso S, Barrionuevo G, Schottler F. Intracellular injections of EGTA block induction of hippocampal long-term potentiation. Nature 1983; 305(5936):20–26. 36. Muller D, Joly M, Lynch G. Contributions of quisqualate and NMDA receptors to the induction and expression of LTP. Science 1988; 242(4886): 1694–1697. 37. Malenka RC, Kauer JA, Zucker RS, Nicoll RA. Postsynaptic calcium is sufficient for potentiation of hippocampai synaptic transmission. Science 1988; 242:81–84. 38. Kauer JA, Malenka RC, Perkel DJ, Nicoll RA. Postsynaptic mechanisms involved in long-term potentiation. Adv Exp Med Biol 1990; 268:291–299. 39. Sanes JR, Lichtman JW. Can molecules explain long-term potentiation? Nat Neurosci 1999; 2(7):597–604. 40. Baudry M. An integrated biochemical model for long-term potentiation. In: Baudry M, Davis JL, eds. Long-Term Potentiation: A Debate of Current Issues. Cambridge: MIT Press, 1991:169–182. 41. Kennedy MB. Signal transduction molecules at the glutamatergic postsynaptic membrane. Brain Res Rev 1998; 26(2–3):243–257. 42. Lisman J. The CaM kinase II hypothesis for the storage of synaptic memory. Trends Neurosci 1994; 17(10):406–412. 43. Lisman J, Schulman H, Cline H. The molecular basis of CaMKII function in synaptic and behavioural memory. Nat Rev Neurosci 2002; 3(3):175–190.
Memory: From Molecular Mechanisms to Clinical Applications
19
44. Massicotte G, Baudry M. Triggers and substrates of hippocampai synaptic plasticity. Neurosci Biobehav Rev 1991; 15(3):415–423. 45. Cole AJ, Saffen DW, Baraban JM, Worley PF. Rapid increase of an immediate early gene messenger RNA in hippocampai neurons by synaptic NMDA receptor activation. Nature 1989; 340:474–476. 46. Abraham WC, Mason SE, Demmer J, Williams JM, Richardson CL, Tate WP, Lawlor PA, Dragunow M. Correlations between immediate early gene induction and the persistence of long-term potentiation. Neuroscience 1993; 56(3):717–727. 47. Qian Z, Gilbert ME, Colicos MA, Kandel ER, Kuhl D. Tissue-plasminogen activator is induced as an immediate-early gene during seizure, kindling and long- term potentiation. Nature 1993; 361(6411):453–457. 48. Silva AJ, Smith AM, Giese KP. Gene targeting and the biology of learning and memory. Ann Rev Genet 1997; 31(1997):527–546. 49. Mayford M, Kandel ER. Genetic approaches to memory storage. Trends Genet 1999; 15(11):463–470. 50. Martin KC. Arc mRNA dynamics: return to sender—the NMDA receptor provides the targeting address for Arc mRNA. Trends Neurosci 2001; 24(11):621–623. 51. Steward O. mRNA at synapses, synaptic plasticity, and memory consolidation. Neuron 2002; 36(3):338–340. 52. Steward O, Worley P. Local synthesis of proteins at synaptic sites on dendrites: role in synaptic plasticity and memory consolidation? Neurobiol Learn Mem 2002; 78(3):508–527. 53. Weiler IJ, Greenough WT. Metabotropic glutamate receptors trigger postsynaptic protein synthesis. Proc Natl Acad Sci USA 1993; 90(15):7168–7171. 54. Frey U, Morris RG. Synaptic tagging and long-term potentiation. Nature 1997; 385(6616):533–536. 55. Teyler TJ. Long-term potentiation and memory. Int J Neurol 1987; 21–22:163–171. 56. Bliss TV, Collingridge GL. A synaptic model of memory: long-term potentiation in the hippocampus. Nature 1993; 361(6407):31–39. 57. Malenka RC, Nicoll RA. Long-term potentiation—a decade of progress? Science 1999; 285(5435):1870–1874. 58. Baudry M. Synaptic plasticity and learning and memory: 15 years of progress. Neurobiol Learn Mem 1998; 70(1=2):113–118. 59. Lynch G, Kessler M, Arai A, Larson J. The nature and causes of hippocampal long-term potentiation. Prog Brain Res 1990; 83:233–250. 60. Martin SJ, Grimwood PD, Morris RG. Synaptic plasticity and memory: an evaluation of the hypothesis. Annu Rev Neurosci 2000; 23:649–711. 61. Martin SJ, Morris RG. New life in an old idea: the synaptic plasticity and memory hypothesis revisited. Hippocampus 2002; 12(5):609–636. 62. Goosens KA, Maren S. Long-term potentiation as a substrate for memory: evidence from studies of amygdaloid plasticity and Pavlovian fear conditioning. Hippocampus 2002; 12(5):592–599. 63. Morris RG. Long-term potentiation and memory. Philos Trans R Soc Lond B Biol Sci 2003; 358(1432):643–647.
20
Baudry
64. Larson J, Wong D, Lynch G. Patterned stimulation at the theta frequency is optimal for the induction of hippocampal long-term potentiation. Brain Res 1986; 368(2):347–350. 65. Granger R, Whitson J, Larson J, Lynch G. Non-Hebbian properties of longterm potentiation enable high-capacity encoding of temporal sequences. Proc Natl Acad Sci USA 1994; 91(21):10104–10108. 66. Brown TH, Chapman PF, Kairiss EW, Keenan CL. Long-term synaptic potentiation. Science 1988; 242(4879):724–728. 67. Grover LM, Teyler TJ. Two components of long-term potentiation induced by different patterns of afferent activation. Nature 1990; 347(6292):477–479. 68. Bliss TVP. Maintenance is presynaptic. Nature 1990; 346:698–699. 69. O’Dell TJ, Hawkins RD, Kandel ER, Arancio O. Tests of the roles of two diffusible substances in long-term potentiation: evidence for nitric oxide as a possible early retrograde messenger. Proc Natl Acad Sci 1991; 88:11285–11289. 70. Maren S, Tocco G, Standley S, Baudry M, Thompson RF. Postsynaptic factors in the expression of long-term potentiation (LTP): increased glutamate receptor binding following LTP induction in vivo. Proc Natl Acad Sci USA 1993; 90(20):9654–9658. 71. Lee K, Oliver M, Lynch G. Brief bursts of high-frequency stimulation produce two types of structural change in rat hippocampus. J Neurophysiol 1980; 44:247–258. 72. Chang F, Greenough WT. Transient and enduring morphological correlates of synaptic activity and efficacy change in the rat hippocampal slice. Brain Res 1984; 309:35–46. 73. Zucker RS. Short-term synaptic plasticity. Annu Rev Neurosci 1989; 12: 13–22. 74. Dolphin AC, Errington ML, Bliss TV. Long-term potentiation of the perforant path in vivo is associated with increased glutamate release. Nature 1982; 297(5866):496–498. 75. Bliss TVP, Lynch MA. Long-term potentiation: mechanisms and properties. In: Landfield P, Deadwyler S, eds. Long-term Potentiation: From Biophysics to Behavior. New York: Alan Liss, 1988:3–72. 76. Hawkins RD, Son H, Arancio O. Nitric oxide as a retrograde messenger during long-term potentiation in hippocampus. Prog Brain Res 1998; 118: 155–172. 77. Martin AR. Quantal nature of synaptic transmission. Physiol Rev 1966; 46:51–66. 78. Malinow R, Tsien RW. Presynaptic enhancement shown by whole-cell recordings of long-term potentiation in hippocampal slices. Nature 1990; 346(6280):177–180. 79. Bekkers JM, Stevens CF. Presynaptic mechanism for long-term potentiation in the hippocampus. Nature 1990; 346(6286):724–729. 80. Foster TC, McNaughton BL. Long-term synaptic enhancement in CA1 is due to increased quantal size, not quantal content. Hippocampus 1991; 1:79–91. 81. Voronin LL. Quantal analysis of hippocampal long-term potentiation. Rev Neurosci 1994; 5(2):141–170.
Memory: From Molecular Mechanisms to Clinical Applications
21
82. Isaac JTR, Nicoll RA, Malenka RC. Evidence for silent synapses: implications for the expression of LTP. Neuron 1995; 15:427–434. 83. Liao DZ, Hessler NA, Malinow R. Activation of postsynaptically silent synapses during pairing-induced ltp in cal region of hippocampal slice. Nature 1995; 375:400–404. 84. Geinisman Y, Detoledo-Morrell L, Morrell F, Heller RE, Rossi M, Parshall RF. Structural synaptic correlate of long-term potentiation—formation of axospinous synapses with multiple, completely partitioned transmission zones. Hippocampus 1993; 3(4):435–446. 85. Toni N, Buchs P-A, Nikonenko I, Bron CR, Muller D. LTP promotes formation of multiple spine synapses between an axon terminal and a dendrite. Nature 1999; 401:421–425. 86. Rail W. Dendritic spines and synaptic potency. In: Porter R, ed. Studies in Neurophysiology. Cambridge: Cambridge University Press, 1978:203–209. 87. Coss RG, Perkel DH. The function of dendritic spines: a review of theoretical issues. Behav Neur Biol 1985; 44:151–185. 88. Wilson CJ. Cellular mechanisms controlling the strength of synapses. J Electron Microsc Tec 1988; 10:293–313. 89. Lynch G, Baudry M. Reevaluating the constraints on hypotheses regarding LTP expression. Hippocampus 1991; 1(1):9–14. 90. Johnston D, Magee JC, Colbert CM, Cristie BR. Active properties of neuronal dendrites. Annu Rev Neurosci 1996; 19:165–186. 91. Sommer B, Keinanen K, Verdoorn TA, Wisden W, Burnashev N, Herb A, Kohler M, Takagi T, Sakmann B, Seeburg PH. Flip and flop: a cell-specific functional switch in glutamate-operated channels of the CNS. Science 1990; 249:1580–1585. 92. Malinow R, Mainen ZF, Hayashi Y. LTP mechanisms: from silence to fourlane traffic. Curr Opin Neurobiol 2000; 10(3):352–357. 93. Lynch G. Memory and the brain: unexpected chemistries and a new pharmacology. Neurobiol Learn Mem 1998; 70:80–100. 94. Staubli U, Vanderklish P, Lynch G. An inhibitor of integrin receptors blocks long-term potentiation. Behav Neural Biol 1990; 53:1–5. 95. Luthi A, Laurent JP, Figurov A, Muller D, Schachner M. Hippocampal longterm potentiation and neural cell-adhesion molecules LI and NCAM. Nature 1994; 372(6508):777–779. 96. Chan CS, Weeber EJ, Kurup S, Sweatt JD, Davis RL. Integrin requirement for hippocampal synaptic plasticity and spatial memory. J Neurosci 2003; 23(18):7107–7116. 97. Ito I, Tanabe S, Khoda A, Sugiyama H. Allosteric potentiation of quisqualate receptors by a nootropic drug aniracetam. J Physiol 1990; 424:533–543. 98. Patneau DK, Vyklicky L Jr, Mayer ML. Hippocampal neurons exhibit cyclothiazide-sensitive rapidly desensitizing responses to kainate. J Neurosci 1993; 13(8):3496–3509. 99. Impagnatiello F, Oberto A, Longone P, Costa E, Guidotti A. 7-Chloro-3methyl-3,4-dihydro-2H-l,2,4-benzothiadiazine S,S-dioxide: a partial modulator of AMPA receptor desensitization devoid of neurotoxicity. Proc Natl Acad Sci USA 1997; 94(13):7053–7058.
22
Baudry
100. Sekiguchi M, Fleck MW, Mayer ML, Takeo J, Chiba Y, Yamashita S, Wada K. A novel allosteric potentiator of AMPA receptors: 4-2-(phenylsulfonylamino)ethylthio-2,6-difluoro-phenoxyaceta mide. J Neurosci 1997; 17(15):5760–5771. 101. Pirotte B, Podona T, Diouf O, de Tullio P, Lebrun P, Dupont L, Somers F, Delarge J, Morain P, Lestage P, Lepagnol J, Spedding M. 4H-1,2,4Pyridothiadiazine 1,1-dioxides and 2,3-dihydro-4H-l,2, 4-pyridothiadiazine 1,1- dioxides chemically related to diazoxide and cyclothiazide as powerful positive allosteric modulators of (R=S)-2-amino-3-(3-hydroxy-5-methylisoxazoI-4-yl)propionic acid receptors: design, synthesis, pharmacology, and structure–activity relationships. J Med Chem 1998; 41(16):2946–2959. 102. Miu P, Jarvie KR, Radhakrishnan V, Gates MR, Ogden A, Ornstein PL, Zarrinmayeh H, Ho K, Peters D, Grabell J, Gupta A, Zimmerman DM, Bleakman D. Novel AMPA receptor potentiators LY392098 and LY404187: effects on recombinant human AMPA receptors in vitro. Neuropharmacology 2001; 40(8):976–983. 103. Gates M, Ogden A, Bleakman D. Pharmacological effects of AMPA receptor potentiators LY392098 and LY404187 on rat neuronal AMPA receptors in vitro. Neuropharmacology 2001; 40(8):984–991. 104. Baumbarger P, Muhlhauser M, Yang CR, Nisenbaum ES. LY392098, a novel AMPA receptor potentiator: electrophysiological studies in prefrontal cortical neurons. Neuropharmacology 2001; 40(8):992–1002. 105. Arai A, Kessler M, Ambros-Ingerson J, Quan A, Yigiter E, Rogers G, Lynch G. Effects of a centrally active benzoylpyrrolidine drug on AMPA receptor kinetics. Neuroscience 1996; 75(2):573–585. 106. Arai AC, Kessler M, Rogers G, Lynch G. Effects of the potent ampakine CX614 on hippocampal and recombinant AMPA receptors: interactions with cyclothiazide and GYKI 52466. Mol Pharmacol 2000; 58(4):802–813. 107. Arai AC, Xia YF, Rogers G, Lynch G, Kessler M. Benzamide-type AMPA receptor modulators form two subfamilies with distinct modes of action. J Pharmacol Exp Ther 2002; 303(3):1075–1085. 108. Sun Y, Olson R, Horning M, Armstrong N, Mayer M, Gouaux E. Mechanism of glutamate receptor desensitization. Nature 2002; 417(6886):245–253. 109. Leever JD, Clark S, Weeks AM, Partin KM. Identification of a site in GluRl and GluR2 that is important for modulation of deactivation and desensitization. Mol Pharmacol 2003; 64(1):5–10. 110. Sirvio J, Larson J, Quach CN, Rogers GA, Lynch G. Effects of pharmacologically facilitating glutamatergic transmission in the trisynaptic intrahippocampal circuit. Neuroscience 1996; 74(4):1025–1035. 111. Arai A, Guidotti A, Costa E, Lynch G. Effect of the AMPA receptor modulator IDRA 21 on LTP in hippocampal slices. Neuroreport 1996; 7(13): 2211–2215. 112. Lauterborn JC, Lynch G, Vanderklish P, Arai A, Gall CM. Positive modulation of AMPA receptors increases neurotrophin expression by hippocampal and cortical neurons. J Neurosci 2000; 20(1):8–21. 113. Mackowiak M, O’Neill MJ, Hicks CA, Bleakman D, Skolnick P. An AMPA receptor potentiator modulates hippocampal expression of BDNF: an in vivo study. Neuropharmacology 2002; 43(1):1–10.
Memory: From Molecular Mechanisms to Clinical Applications 114.
115.
116. 117.
118.
119.
120.
121.
122.
123.
124.
125. 126. 127.
128.
23
Zivkovic I, Thompson DM, Bertolino M, Uzunov D, DiBella M, Costa E, Guidotti A. 7-Chloro-3-methyl-3-4-dihydro-2H-l,2,4 benzothiadiazine S,Sdioxide (IDRA 21): a benzothiadiazine derivative that enhances cognition by attenuating DL-alpha-amino-2,3-dihydro-5-methyl-3-oxo-4-isoxazolepropanoic acid (AMPA) receptor desensitization. J Pharmacol Exp Ther 1995; 272(1):300–309. Granger R, Staubli U, Davis M, Perez Y, Nilsson L, Rogers GA, Lynch G. A drug that facilitates glutamatergic transmission reduces exploratory activity and improves performance in a learning-dependent task. Synapse 1993; 15(4):326–329. Staubli U, Rogers G, Lynch G. Facilitation of glutamate receptors enhances memory. Proc Natl Acad Sci 1994; 91:777–781. Larson J, Lieu T, Petchpradub V, LeDuc B, Ngo H, Rogers GA, Lynch G. Facilitation of olfactory learning by a modulator of AMPA receptors. J Neurosci 1995; 15(12):8023–8030. Hampson RE, Rogers G, Lynch G, Deadwyler SA. Facilitative effects of the ampakine CX516 on short-term memory in rats: enhancement of delayednonmatch-to-sample performance. J Neurosci 1998; 18(7):2740–2747. Lynch G, Kessler M, Rogers G, Ambros-Ingerson J, Granger R, Schehr RS. Psychological effects of a drug that facilitates brain AMPA receptors. Int Clin Psychopharmacol 1996; 11(1):13–19. Ingvar M, Ambros-Ingerson J, Davis M, Granger R, Kessler M, Rogers GA, Schehr RS, Lynch G. Enhancement by an ampakine of memory encoding in humans. Exp Neurol 1997; 146(2):553–559. Lynch G, Granger R, Ambros-Ingerson J, Davis CM, Kessler M, Schehr R. Evidence that a positive modulator of AMPA-type glutamate receptors improves delayed recall in aged humans. Exp Neurol 1997; 145(1):89–92. Johnson SA, Larson J, Luu NT, Herbst TA, Rogers G, Lynch G. Reduction of metamphetamine hyperactivity by AMPA receptor facilitation: positive interaction with typical and atypical antipsychotics. Pharmacologist 1997; 39:120. Johnson SA, Luu NT, Herbst TA, Knapp R, Lutz D, Arai A, Rogers GA, Lynch G. Synergistic interactions between ampakines and antipsychotic drugs. J Pharmacol Exp Ther 1999; 289(1):392–397. Knapp RJ, Goldenberg R, Shuck C, Cecil A, Watkins J, Miller C, Crites G, Malatynska E. Antidepressant activity of memory-enhancing drugs in the reduction of submissive behavior model. Eur J Pharmacol 2002; 440(1):27–35. Ferris SH. Clinical trials in AD: are current formats and outcome measures adequate? Alzheimer Dis Assoc Disord 2002; 16(suppl 1):S13–S17. Herrmann N. Cognitive pharmacotherapy of Alzheimer’s disease and other dementias. Can J Psychiatry 2002; 47(8):715–722. Petersen RC, Doody R, Kurz A, Mohs RC, Morris JC, Rabins PV, Ritchie K, Rossor M, Thai L, Winblad B. Current concepts in mild cognitive impairment. Arch Neurol 2001; 59:1985–1992. Carlsson A. The current status of the dopamine hypothesis of schizophrenia. Neuropsychopharmacology 1988; 1(3):179–186.
24
Baudry
129. Carlsson M, Carlsson A. Interactions between glutamatergic and monoaminergic systems within the basal ganglia—implications for schizophrenia and Parkinson’s disease. Trends Neurosci 1990; 13(7):272–276. 130. Tsai G, Coyle JT. Glutamatergic mechanisms in schizophrenia. Annu Rev Pharmacol Toxicol 2002; 42:165–179. 131. Danysz W. Positive modulators of AMPA receptors as a potential treatment for schizophrenia. Curr Opin Investig Drugs 2002; 3(7):1062–1066. 132. Goff DC, Leahy L, Berman I, Posever T, Herz L, Leon AC, Johnson SA, Lynch G. A placebo-controlled pilot study of the ampakine CX516 added to clozapine in schizophrenia. J Clin Psychopharmacol 2001; 21(5):484–487. 133. Lynch G. Memory enhancement: the search for mechanism-based drugs. Nat Neurosci 2002; (5 suppl):1035–1038.
2 Synaptogenesis as a Correlate of Activity-Induced Plasticity Irina Nikonenko, Pascal Jourdain, Bemadett Boda, and Dominique Muller Neuropharmacology, Centre Medical Universitaire, Geneva, Switzerland
1.
INTRODUCTION
Plasticity of excitatory synapses is believed to underlie important aspects of signal integration by neuronal networks. Until recently, plasticity was essentially considered in terms of functional changes in the efficacy of transmission, focusing particularly on long-term potentiation (LTP), an increase in synaptic strength induced by a brief period of high frequency stimulation (1). Decades of intense research on the molecular mechanisms contributing to this form of plasticity have revealed the central role played by postsynaptic receptors in this process (1–4). Following activation of NMDA receptors during high frequency stimulation, calcium triggers signaling cascades that modify the properties or number of postsynaptic AMPA type receptors expressed at the synaptic membrane (5–8). Phosphorylation-dependent mechanisms that affect receptor properties and recycling or diffusion in the membrane appear therefore to be critically implicated in the changes of synaptic efficacy. However, these functional analyses have been paralleled in last years by accumulating evidence indicating that spine synapses are also characterized by a high degree of morphological plasticity (9–13). The development of time lapse confocal video microscopy associated to more traditional morphological approaches including electron microscopy and three-dimensional reconstruction has revealed several different types of structural changes taking place at 25
26
Nikonenko et al.
the level of dendrites or spines under the influence of synaptic activity. These changes range from modification of the shape of spines to the budding of dendritic filopodia or the formation of new spines or new types of synapses (14–19). More recently, changes affecting the presynaptic varicosities have also been described (20–24). In this case, too synaptic activation seems to be an important triggering mechanism. Together, these observations have strengthened the interpretation that synaptic activity is able to induce a process of synaptogenesis by stimulating growth mechanisms at both pre- and postsynaptical levels. In addition, activity appears to affect the proportion of dynamic synapses characterizing a synaptic network, a factor that certainly contributes to enhance remodeling. We review here these new data and discuss the implications and open issues that these fascinating results have unraveled.
2.
CHANGES IN SPINE DYNAMICS WITH ACTIVITY
Observation of dendritic spines using time-lapse fluorescent confocal video microscopy revealed that they are highly dynamic structures. They change shape over a time scale of seconds without major modifications in the size of their head (25). These movements can be blocked by cytochalasin D and probably reflect a continuous rearrangement of the actin filaments that are particularly enriched in spine heads (25,26). The role of this spine motility remains unclear, but interestingly, it is regulated by synaptic activation (26,27). Application of small concentrations of AMPA or anesthetics was found to fully block the movements of these dancing spines. Curiously and in contrast with many other aspects of structural plasticity, no effects of NMDA receptor agonists or antagonists were so far observed. Among possible explanations, it has been proposed that the dynamic movements of the spine head could affect the morphological characteristics of the spine neck and thereby modulate the diffusional coupling between spines and dendrites (28–30). Another possibility is that these movements relate somehow to the stability and size of spines. Interestingly, it is not only the spines that are dynamic, but so are also the postsynaptic densities (PSDs). These complex structures are made by the clustering of postsynaptic receptors together with many scaffolding proteins. A recent analysis of PSDs using over-expression of a fluorescently tagged PSD-95 molecule revealed that they may change shape, dissociate or form de novo within time frames of minutes (31,32). It appears therefore that the structural organization of spines may change rapidly in a way that can have significant consequences for their functioning.
3.
MODULATION OF SPINE MORPHOLOGY BY ACTIVITY
In addition to or as a result of continuous dynamic movements, dendritic spines appear to change shape and size. This has particularly been studied
Synaptogenesis and Activity-Induced Plasticity
27
using electron microscopy by comparing synaptic types under various conditions of stimulation. Different forms of structural remodeling have been described, including enlargements of the spine head, formation of protrusions such as spinules, or modifications in the convexity or concavity of spines (33–37). Whether these observations effectively reflect true and stable structural changes or only transitory states through which spines evolve due to their dynamic movements is still unclear. There is however one specific type of spine synapses that has attracted particular interest: the so-called perforated synapses that exhibit a discontinuity in their PSD or a PSD distributed on two or more different locations (38–42). The morphology of these perforated synapses is peculiar in several respects. They usually have large, mushroom-type spine heads, and the areas of their PSD and spine head average about three times those of simple spines (43–46). This could imply that the formation of perforated synapses requires addition of membrane and PSD to a simple spine. There are indeed examples observed using 2-photon confocal microscopy that show such enlargements of dendritic spines following stimulation (47). Perforated synapses are also characterized by a high content in coated vesicles. Under control conditions, spines very rarely include coated vesicles (only about 15% of them). In contrast, we found that a greater proportion of them showed one or several coated vesicles, and this parameter was markedly increased in the population of perforated synapses observed after LTP induction (43). While the reasons for the presence of so many endocytotic vesicles remain unclear, one obvious interpretation is that they could reflect an ongoing process of synaptic membrane remodeling. These observations therefore lead to the conclusion that perforated synapses could represent a state of a spine where membrane recycling is enhanced. In many studies, the proportion of perforated synapses has been found to increase under various experimental conditions. This includes rats raised in a complex environment (39), development, models of regeneration (42,48) and, more recently, conditions involving increased neuronal activity such as kindling, LTP or brief anoxia=hypoglycemia (19,36,47,49–52). The timing of appearance of these perforated synapses is variable, but recent studies suggest that the change may be quite rapid about 15 min after bicuculline treatment or following LTP induction and anoxia=hypoglycemia (19,47,52). Formation of these perforated synapses appears to require NMDA receptor activation and could also depend upon calcium. There is indeed evidence that spines may change shape and size under the influence of calciumdependent mechanisms (12,16,47). Formation of perforated synapses could thus represent a process of transformation of activated spine synapses that would result in a more efficient synapse due to the reorganization and enlargement of the PSD and eventually the incorporation of new receptors. The larger size of these synapses could promote a greater stability through an improved capacity to handle calcium transients. Larger spines have indeed
28
Nikonenko et al.
a higher proportion of intracellular organelles such as endoplasmic reticulum, spine apparatus, and their increased volume could help reduce or dampen the calcium transient induced by activity. In some way, formation of perforated synapses could reflect some kind of maturation process allowing important synapses to be stabilized.
4.
EVIDENCE FOR ACTIVITY-INDUCED SYNAPTOGENESIS
In addition to changes in size or structural organization of spine synapses, activity was also recently found to promote a process of synaptogenesis. This conclusion is essentially based on results obtained through the development of 2-photon confocal microscopy techniques, but also on some electron microscopic observations. In confocal studies, it was found that induction of LTP resulted within 10–15 min in the growth of dendritic filopodia (Fig. 1A; see Ref. 17). These are believed to be precursors of dendritic spines (15,53,54), and recent observations support the interpretation that filopodia, after elongation and retraction, might stabilize, exhibit PSDs and form new synapses (31,32). Growth of filopodia is an NMDA receptor-sensitive and calcium-dependent process (17). They might thus reflect an activity-dependent growth mechanism associated with LTP induction that leads to the formation of new synapses. Another result obtained at the same time and supporting the same interpretation was the observation that high frequency stimulation directly triggered the formation of new spines (Fig. 1B; see Ref. 18). This phenomenon was however more delayed in time, occurring between 30 and 60 min after stimulation. It was probably unrelated to the formation of filopodia, since the new spines seen in this study appeared de novo within a few minutes on dendritic segments without evidence for preexisting filopodia. Interestingly, formation of these new spines was also NMDA receptor and calcium dependent. Further supporting these results, an analysis of spine morphology carried out using electron microcopy before and at different times after LTP induction revealed similar changes very suggestive for a process of synaptogenesis (19). Between 30 and 60 min after stimulation, quantitative morphometric analyses of activated synapses showed a marked increase in images of multiple spine boutons, synapses where one nerve terminal contacts two different spines. After 3D reconstruction of a number of these images, it appeared that the increase in multiple spine boutons observed after LTP induction was mainly due to cases where the two spines originated from the same dendrite (Fig. 1C). This result is interesting because it strongly suggested a process of spine duplication, providing thereby a mechanism of synaptogenesis that enhances synaptic transmission in a specific manner between activated cells. It is also important to mention that although in most cases these duplicated synapses included a smaller spine, these contacts exhibited all the
Synaptogenesis and Activity-Induced Plasticity
29
Figure 1 Activity-dependent structural remodeling of postsynaptic structures. (A) Illustration of the growth of two filopodia (arrows) induced by application of short period (2 min) of anoxia=hypoglycemia to CAl hippocampal slice cultures. The filopodia started to grow about 10 min after synaptic activation by anoxia= hypoglycemia, they elongated to reach a maximal size at about 20 min and then retracted to stabilize as new spines (bar: 2 mm). (B) Illustration of the rapid appearance of two new spines (arrows) 35 min after a short anoxia=hypoglycemia (bar: 2 mm). (C) Three-dimensional reconstruction of a simple spine (left) and a multiple spine synapse (right). Note that the duplicated synapse involves a single presynaptic terminal that contacts two distinct spines originating from the same dendrite (bar: 0.5 mm).
characteristics of morphologically mature synapses. There was even some indirect evidence that, in some cases at least, these duplicated synapses were functional (19). Again, these changes were clearly activity and LTPdependent, as they could be blocked by NMDA receptor antagonists and inhibitors of LTP. Taken together, these results strongly suggested that synaptic stimulation through NMDA receptor activation and calcium influx could drive the formation of filopodia, new spines, and most likely new synapses. The importance of these mechanisms has since been confirmed in several other conditions. In dissociated cultures, potassium depolarization produces a synaptic activation and an enhancement comparable in several respects to LTP. This treatment is also associated with growth of
30
Nikonenko et al.
filopodia and formation of new spines (55). Similarly, in organotypic slice cultures, we found that application of a short period of anoxia= hypoglycemia reproduced the same phenomenology with a very comparable time course (47). A few minutes of glucose and oxygen deprivation resulted in a lasting and NMDA receptor-dependent potentiation of synaptic transmission, but also in a marked structural remodeling including formation of perforated synapses and of multiple synapse boutons, growth, within 10– 20 min, of dendritic filopodia (Fig. 1A) and, after 30 min, formation of new spines (Fig. 1B). Finally, in a very elegant in vivo study, it was recently shown that sensory stimulation of vibrissae is associated with an increased turnover of spines in the somatosensory cortex (56). The authors could even observe the formation of new spines and subsequently demonstrate at the electron microscopic level that these new spines corresponded to the creation of new, morphologically mature synapses. Thus, in the last years, evidence has accumulated demonstrating that activity can directly trigger a process of synaptogenesis within a time frame of 30–60 min.
5.
POSTSYNAPTIC MECHANISMS CONTROLLING ACTIVITY-INDUCED SYNAPTOGENESIS
The mechanisms of activity-induced structural plasticity mentioned above essentially concerned postsynaptic aspects. The formation of particular types of synapses, the enlargement of spines, the growth of filopodia, and the direct formation of new spines are all changes that imply a reorganization of postsynaptic structures. These changes were all shown to require NMDA receptor activation, to depend upon calcium influx in postsynaptic spines and to take place within tens of minutes. Evidence from analyses of PSDs using tagged molecules further indicates that restructuring of the PSD as well as formation of new PSDs are also rapid events that take place within the same time frame (31,32). Taken together, these results support the interpretation that a major aspect of activity-induced synaptogenesis is controlled by postsynaptic mechanisms. They suggest further that the same signaling mechanisms involved in the induction of LTP are also responsible for triggering structural plasticity. The possibility to now directly visualize and stimulate growth processes and spine formation in dissociated or slice cultures opens new avenues to investigate the signaling cascades that contribute to structural plasticity. In preliminary experiments, we have investigated whether calcium-cahnodulin dependent protein kinase II, which plays a critical role in LTP induction, also participates in regulating these mechanisms. We found that intracellular injection of an active form of the kinase reproduced the changes seen following LTP induction and stimulated the growth of filopodia and formation of new spines. In contrast, blockade of the enzyme prevented activity-induced structural remodeling.
Synaptogenesis and Activity-Induced Plasticity
31
Similar analyses were also performed while looking at spine formation in cells over-expressing various components of the PSD or scaffolding proteins regulating PSD formation. Over-expression or elimination of various components of the PSD, including receptor subunits, have been found to affect spine formation or result in the formation of aberrant spines (57–59). These experiments thus start to reveal a role played by the numerous components of synapses not only in terms of function, but also in the mechanisms and dynamics of synaptic contact formation.
6.
ACTIVITY-INDUCED REMODELING OF PRESYNAPTIC TERMINALS
While the data reviewed above strongly argue for a postsynaptic control of activity-induced synaptogenesis, it seems obvious that there must also be a presynaptic counterpart to this structural plasticity as new release sites must also be created. Evidence obtained through analysis of tagged components of the release machinery revealed the existence of preformed release packages that are transported through axonal flow (60–62). These packages can be seen moving along the axon and, upon activation, may be inserted quite rapidly at new release sites. More recently, analysis of slice cultures prepared from transgenic mice expressing plasmalemma or synaptic vesicle-associated fluorescent markers allowed for the first time to visualize on a time scale of days the behavior and turn-over of presynaptic varicosities. This study revealed that in mature, adult synaptic networks, the total number of presynaptic varicosities is stable and remains constant, that the majority of varicosities do not change over time, but that there exists a small proportion of dynamic boutons that undergo continuous remodeling, disappearing, and reappearing within days (23). Interestingly, the fraction of these dynamic presynaptic varicosities also corresponds to a slightly larger proportion of highly dynamic spines identified in adult cortex. Furthermore, as for spines, this study revealed that synaptic activation by high frequency stimulation, glutamate, or AMPA treatments markedly increased the fraction of these dynamic boutons, thereby demonstrating a clear role of activation in varicosity turnover. Another recent study revealed the highly dynamic nature of presynaptic filopodia in mossy fibers. This motility was also shown to be activity dependent and, curiously, developmentally regulated (24). In experiments carried out by assessing the morphology of varicosities in the CA1 area following synaptic activation and LTP induction, we found very comparable results and observed images of presynaptic filopodia growth or varicosity splitting induced by activity (Fig. 2A). Using electron microscopy, we reconstructed these presynaptic filopodia-like protrusions and found that within 30 min they established contacts with postsynaptic membrane and induced the formation of a PSD (Fig. 2B). Furthermore,
32
Nikonenko et al.
Figure 2 Activity-dependent growth of presynaptic filopodia. (A) Illustration of the growth of a presynaptic filopodia-like protrusion (arrow) induced by a short (2 min) anoxia=hypoglycemia in the CAl region of hippocampal organotypic slice cultures (bar: 2 mm). (B) Electron microscopic image of a presynaptic filopodia (star) taken 30 min after a short (2 min) anoxia=hypoglycemia. Note the presence of clustered synaptic vesicles at the tip of the filopodia, including vesicles docked to the synaptic membrane facing the PSD on the dendritic spine (bar: 0.5 mm). (C) Three dimensional reconstruction of a presynaptic filopodia observed on serial sections obtained 20 min after a short anoxia=hypoglycemia (bar: 0.5 mm).
while most postsynaptic structures contacted initially were dendritic shafts (Fig. 2C), it appeared that after 30 min almost all filopodia-like protrusions had established synapses with dendritic spines (Fig. 2B). These results thus confirmed the notion that activity also triggers presynaptic growth mechanisms, and that presynaptic remodeling could rapidly lead to formation of new, morphologically mature synaptic contacts, and eventually account for the formation of some of the new spines observed with confocal microscopy. 7.
CONCLUSIONS
Several decades of studies have contributed to highlight the mechanisms through which synaptic activation modulates the efficacy of signal transmission. In the last years, these studies have also demonstrated that, in addition, synaptic activation can result in growth mechanisms and formation of new synaptic contacts. The evidence reviewed here indicates that various types of structural changes can be observed. On the one hand, there are modifications that affect the morphology of existing synapses. These include changes in the shape of the spine head or redistribution of the PSD at the synaptic membrane. One of the simplest interpretations of these images would be
Synaptogenesis and Activity-Induced Plasticity
33
that they represent the morphological correlate of the reorganization or changes in number of receptors that most likely account for the increase in synaptic efficacy. On the other hand, many aspects of the structural studies reported above indicate that activity also triggers a process of synaptogenesis. Based on these results, an emerging concept is that synaptic networks are characterized both by stability and dynamic remodeling. While most synapses probably do not change over time, there seems to be a small fraction of them that continuously disappear and reappear. This seems to be the case for spines as well as for presynaptic varicosities. In both cases activation and particularly high frequency stimulation increases the fraction of dynamic synapses and within 30 min stimulate the formation of new contacts. Important questions however remain unanswered. While the new synapses appear to be morphologically mature, evidence that they are functional is still lacking. It is also unclear how specific this process of activityinduced synaptogenesis is: the new synapses could be formed exclusively between activated terminals and dendrites, but the growth process might also be more general and unspecific. Other intriguing questions are what defines a dynamic synapse and how long it remains dynamic; what are the molecular mechanisms that control these events and whether this remodeling process really contributes to information processing (63). These are exciting times, new methodologies are now available to address these questions and one can predict major fascinating advances in the coming years. ACKNOWLEDGMENTS This work was supported by the Swiss National Science Foundation, the SCOPE program, the Ott Foundation, the De Reuter Foundation, and the Novartis Foundation. REFERENCES 1. 2.
3.
4. 5. 6.
Malenka RC, Nicoll RA. Long-term potentiation—a decade of progress? Science 1999; 285:1870–1874. Shi S, Hayashi Y, Esteban JA, Malinow R. Subunit-specific rules governing AMPA receptor trafficking to synapses in hippocampal pyramidal neurons. Cell 2001; 105:331–343. Shi SH, Hayashi Y, Petralia RS, Zaman SH, Wenthold RJ, Svoboda K, et al. Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 1999; 284:1811–1816. Song I, Huganir RL. Regulation of AMPA receptors during synaptic plasticity. Trends Neurosci 2002; 25:578–588. Lisman JE, Mclntyre CC. Synaptic plasticity: a molecular memory switch. Curr Biol 2001; 11(19):R788–R791. Lisman J, Schulman H, Cline H. The molecular basis of CamKII function in synaptic and behavioural memory. Nat Rev Neurosci 2002; 3:175–190.
34
Nikonenko et al.
7.
Soderling TR, Derkach VA. Postsynaptic protein phosphorylation and LTP. Trends Neurosci 2000; 23:75–80. Malinow R, Malenka RC. AMPA receptor trafficking and synaptic plasticity. Annu Rev Neurosci 2002; 25:103–126. Edwards FA. Anatomy and electrophysiology of fast central synapses lead to a structural model for long-term potentiation. Physiol Rev 1995; 75:759–787. Luscher C, Nicoll RA, Malenka RC, Muller D. Synaptic plasticity and dynamic modulation of the postsynaptic membrane. Nat Neurosci 2000; 3(6):545–550. Matus A. Actin-based plasticity in dendritic spines. Science 2000; 290(5492):754–758. Segal I, Korkotian I, Murphy DD. Dendritic spine formation and pruning: common cellular mechanisms? Trends Neurosci 2000; 23:53–27. Harris KM. Structure, development, and plasticity of dendritic spines. Curr Opin Neurobiol 1999; 9:343–348. Hosokawa T, Rusakov DA, Bliss TV, Fine A. Repeated confocal imaging of individual dendritic spines in the living hippocampal slice: evidence for changes in length and orientation associated with chemically induced LTP. J Neurosci 1995; 5:5560–5573. Ziv NE, Smith SJ. Evidence for a role of dendritic filopodia in synaptogenesis and spine formation. Neuron 1996; 17(l):91–102. Korkotian E, Segal M. Release of calcium from stores alters the morphology of dendritic spines in cultured hippocampal neurons. Proc Natl Acad Sci USA 1999; 96:12068–12072. Maletic-Savatic M, Malinow R, Svoboda K. Rapid dendritic morphogenesis in CA1 hippocampal dendrites induced by synaptic activity. Science 1999; 283:1923–1927. Engert F, Bonhoeffer T. Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature 1999; 399:66–70. Toni N, Buchs PA, Nikonenko I, Bron CR, Muller D. LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature 1999; 402:421–425. Ahmari SE, Smith SJ. Knowing a nascent synapse when you see it. Neuron 2002; 34:333–336. Chang S, De Camilli P. Glutamate regulates actin-based motility in axonal filopodia. Nat Neurosci 2001; 4(8):787–793. Colicos MA, Collins BE, Sailor MJ, Goda Y. Remodeling of synaptic actin induced by photoconductive stimulation. Cell 2002; 107(5):605–616. De Paola V, Arber S, Caroni P. AMPA receptors regulate dynamic equilibrium of presynaptic terminals in mature hippocampal networks. Nat Neurosci 2003; 6(5):491–500. Tashiro A, Dunaevsky A, Blazeski R, Mason CA, Yuste R. Bidirectional regulation of hippocampal mossy fiber filopodial motility by kainate receptors. A two-step model of synaptogenesis. Neuron 2003; 38(5):773–784. Fischer M, Kaech S, Knutti D, Matus A. Rapid actin-based plasticity in dendritic spines. Neuron 1998; 20:847–854.
8. 9. 10.
11. 12. 13. 14.
15. 16.
17.
18. 19.
20. 21. 22. 23.
24.
25.
Synaptogenesis and Activity-Induced Plasticity 26. 27. 28. 29. 30.
31.
32. 33. 34.
35.
36.
37.
38. 39. 40.
41. 42.
43.
35
Kaech S, Brinkhaus H, Matus A. Volatile anesthetics block actin-based motility in dendritic spines. Proc Natl Acad Sci USA 1999; 96:10433–10437. Fischer M, Kaech S, Wagner U, Brinkhaus H, Matus A. Glutamate receptors regulate actin-based plasticity in dendritic spines. Nat Neurosci 2000; 3:887–894. Majewska A, Tashiro A, Yuste R. Regulation of spine calcium dynamics by rapid spine motility. J Neurosci 2000; 20(22):8262–8268. Kim CH, Lisman JE. A role of actin filament in synaptic transmission and long-term potentiation. J Neurosci 1999; 19:4314–4324. Kim CH, Lisman JE. A labile component of AMPA receptor-mediated synaptic transmission is dependent on microtubule motors, actin, and N-ethylmaleimide-sensitive factor. J Neurosci 2001, 2002; 21:4188–4194. Prange O, Murphy TH. Modular transport of postsynaptic density-95 clusters and association with stable spine precursors during early development of cortical neurons. J Neurosci 2001; 21:9325–9333. Marrs GS, Green SH, Dailey ME. Rapid formation and remodeling of postsynaptic densities in developing dendrites. Nat Neurosci 2001; 4(10):1006–1013. Fifkova E, Anderson CL. Stimulation-induced changes in dimensions of stalks of dendritic spines in the dentate molecular layer. Exp Neurol 1981; 74:621–627. Chang PL, Isaacs KR, Greenough WT. Synapse formation occurs in association with the induction of long-term potentiation in two-year-old rat hippocampus in vitro. Neurobiol Aging 1991; 2:517–522. Chang F-LF, Greenough WT. Transient and enduring morphological correlates of synaptic activity and efficacy change in the rat hippocampal slice. Brain Res 1984; 309:35–46. Weeks AC, Ivanco TL, LeBoutillier JC, Racine RJ, Petit TL. Sequential changes in the synaptic structural profile following long-term potentiation in the rat dentate gyrus: I. The intermediate maintenance phase. Synapse 1999; 31:97–107. Desmond NL, Levy WB. Morphological correlates of long-term potentiation imply the modification of existing synapses, not synaptogenesis, in the hippocampal dentate gyrus. Synapse 1990; 5:139–143. Calverley RK, Jones DG. Contributions of dendritic spines and perforated synapses to synaptic plasticity. Brain Res Brain Res Rev 1990; 15:215–249. Greenough WT, West RW, DeVoogd TJ. Subsynaptic plate perforations: changes with age and experience in the rat. Science 1978; 202:1096–1098. Geinisman Y. Perforated axospinous synapses with multiple, completely partitioned transmission zones: Probable structural intermediates in synaptic plasticity. Hippocampus 1993; 3:417–434. Itarat W, Jones DG. Perforated synapses are present during synaptogenesis in rat neocortex. Synapse 1992; 11:279–286. Nieto-Sampedro M, Bussineau CM, Cotman CW. Turnover of brain postsynaptic densities after selective deafferentation: detection by means of an antibody to antigen PSD- 95. Brain Res 1982; 251:211–220. Toni N, Buchs PA, Nikonenko I, Povilaitite P, Parisi L, Muller D. Remodeling of synaptic membranes after induction of long-term potentiation. J Neurosci 2001; 21(16):6245–6251.
36
Nikonenko et al.
44. Jones DG, Calverley RK. Perforated and non-perforated synapses in rat neocortex: three-dimensional reconstructions. Brain Res 1991; 556:247–258. 45. Calverley RK, Jones DG. A serial-section study of perforated synapses in rat neocortex. Cell Tissue Res 1987; 247:565–572. 46. Geinisman Y, DeToledo-Morrell L, Morrell F, Heller RE, Rossi M, Parshall RF. Structural synaptic correlate of long-term potentiation: formation of axospinous synapses with multiple, completely partitioned transmission zones. Hippocampus 1993; 3:435–446. 47. Jourdain P, Nikonenko I, Alberi S, Muller D. Remodeling of hippocampal synaptic networks by a brief anoxia-hypoglycemia. J Neurosci 2002; 22(8): 3108–3116. 48. Carlin RK, Siekevitz P. Plasticity in the central nervous system: do synapses divide? Proc Natl Acad Sci USA 1983; 80:3517–3521. 49. Geinisman Y, deToledo Morrell L, Morrell F. Induction of long-term potentiation is associated with an increase in the number of axospinous synapses with segmented postsynaptic densities. Brain Res 1991; 566:77–88. 50. Geinisman Y, Morrell F, deToledo Morrell L. Increase in the number of axospinous synapses with segmented postsynaptic densities following hippocampal kindling. Brain Res 1992; 569:341–347. 51. Buchs P-A, Muller D. LTP induction is associated with major ultrastructural changes of activated synapses. Proc Natl Acad Sci USA 1996; 96:8040–8045. 52. Neuhoff H, Roeper J, Schweizer M. Activity-dependent formation of perforated synapses in cultured hippocampal neurons. Eur J Neurosci 1999; 11:4241–4250. 53. Dailey ME, Smith SJ. The dynamics of dendritic structure in developing hippocampal slices. J Neurosci 1996; 16:2983–2994. 54. Fiala JC, Feinberg M, Popov V, Harris KM. Synaptogenesis via dendritic filopodia in developing hippocampal area CA1. J Neurosci 1998; 18:8900–8911. 55. Goldin M, Segal M, Avignone E. Functional plasticity triggers formation and pruning of dendritic spines in cultured hippocampal networks. J Neurosci 2001; 21(1):186–193. 56. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E, et al. Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature 2002; 420:788–794. 57. Hering H, Sheng M. Dendritic spines: structure, dynamics and regulation. Nat Rev Neurosci 2001; 2(12):880–888. 58. Sala C, Piech V, Wilson NR, Passafaro M, Liu G, Sheng M. Regulation of dendritic spine morphology and synaptic function by Shank and Homer. Neuron 2001; 31:115–130. 59. Pak DT, Yang S, Rudolph-Correia S, Kim E, Sheng M. Regulation of dendritic spine morphology by SPAR, a PSD-95-associated RapGAP. Neuron 2001; 31:289–303. 60. Kraszewski K, Mundigl O, Daniell L, Verderio C, Matteoli M, De Camilli P. Synaptic vesicle dynamics in living cultured hippocampal neurons visualized with CY3- conjugated antibodies directed against the lumenal domain of synaptotagmin. J Neurosci 1995; 15:4328–4342.
Synaptogenesis and Activity-Induced Plasticity 61. 62. 63.
37
Okabe S, Miwa A, Okado H. Spine formation and correlated assembly of presynaptic and postsynaptic molecules. J Neurosci 2001; 21:6105–6114. Ahmari SE, Buchanan J, Smith SJ. Assembly of presynaptic active zones from cytoplasmic transport packets. Nat Neurosci 2000; 3(5):445–451. Mel BW. Have we been hebbing down the wrong path? Neuron 2002; 34: 175–177.
3 Volume Transmission: Role in the Carry-Over Functional Effect in Tactile Vestibular Substitution? Paul Bach-y-Rita Departments of Orthopedics and Rehabilitation Medicine, and Biomedical Engineering, University of Wisconsin, Madison, Wisconsin, U.S.A.
This chapter will explore an alternative neurotransmission mechanism to synaptic transmission: volume transmission (VT). In addition to the possibility that it is a major mechanism of mass, sustained functions such as pain, mood, and hunger, it may play a role in brain plasticity. Volume transmission in brain reorganization following brain damage has been reviewed elsewhere (1). In this brief review, VT will be discussed, and its possible role in a carry-over functional effect in tactile vestibular substitution will be explored.
1.
VOLUME TRANSMISSION
Volume transmission (VT), which has also been called nonsynaptic diffusion neurotransmission (NDN), includes the diffusion through the extracellular fluid of neurotransmitters released at points that may be remote from the target cells, with the resulting activation of extrasynaptic receptors (and possible intrasynaptic receptors reached by diffusion into the synaptic cleft). Volume transmission also includes the diffusion of substances such as nitric oxide (NO) and carbon monoxide (CO) through both the extracellular fluid and cellular membranes. In contrast to the one-to-one, point-to-point ‘‘private’’ intercellular synaptic communication in the brain, VT is a slow, 39
40
Bach-y-Rita
one-to-many, widespread intercellular communication (2). Volume transmission appears to play multiple roles in the brain in normal and abnormal activity, brain plasticity, and drug actions (reviewed in Refs. 1–5); in some, such as psychotropic drugs, VT mechanisms may predominate (6). Combinations of both synaptic and diffusion neurotransmission are important in vision, and many other functions. Although the synaptic basis of information flow (also called the Cajal–Sherington paradigm of interneuronal communication) has been the major focus of neuroscience research for more than a century, experimental results obtained in VT studies have led various scientists, including Agnati et al. (3), to consider that it is time to revise that paradigm. They noted that although much work remains to be accomplished to firmly establish the concept, VT must now be considered to play an important role in neuronal communication. Volume transmission can be modeled by students in a university classroom (7), who can be equated to neurotransmitter molecules in a vesicle. Upon release, they must go to specific other classrooms spread throughout the campus (receptor sites). They flow out into the halls and grounds between buildings (extracellular fluid), where they mix with other students (neurotransmitter molecules) from other classrooms (vesicles) going to other target classrooms. They walk (diffuse) to their specific classrooms (receptors) which they enter (bind). In contrast, synaptic communication would require each student to travel to the next class on a motorized separate pathway, much like the enclosed people-transporter bands at the De Gaulle airport in Paris. Among the range of functions in which VT plays a role are many of those mediated by noradrenaline: in the autonomic nervous system (e.g., see Ref. 8), in human omental veins (9), and in the central nervous system (CNS), such as in sleep (cf., Ref. 10). Volume transmission may also play a role in functional reorganization following brain damage (4). Volume transmission appears to be the principal means of neurocommunication in many invertebrates (cf., Refs. 4,11), and may be more common in certain mammalian brain regions. Volume transmission may be the primary information transmission mechanism in certain normal mass, sustained functions, such as sleep, vigilance, hunger, brain tone, and mood (12) and certain responses to sensory stimuli, as well as several abnormal functions, such as mood disorders, spinal shock, spasticity, shoulder-hand, and autonomic dysreflexia syndromes, and drug addiction (4,13,14). There have been several names proposed for diffusion neurotransmission (4), the two principal ones being VT and NDN; neither is completely appropriate. Nonsynaptic diffusion neurotransmission (NDN) has the disadvantage of being a negative term, and of not including diffusion neurotransmission that leaks out of or into synapses. Fuxe and Agnati (2) and Agnati et al. (15) proposed ‘‘Volume Transmission’’ (VT) for electrical
Volume Transmission: Role in Vestibular Substitution
41
and chemical communication in the brain via the extracellular fluid, considering VT to be analogous to the term ‘‘Volume Conduction’’; Nicholson (16) noted that ‘‘volume conductors’’ had been used in 1947 by Lorente de No´, in discussing the way current distributes in tissues. The term ‘‘VT’’ may create the image of massive undirected neurotransmission. However, work carried out in Routtenberg’s laboratory in the late 1960s suggested that transmitters could readily move in the extracellular space (ECS) (the fluid that surrounds neurons in the brain) and travel over distances by directional flow (this has been confirmed and extended by Bjelke et al. (17). Furthermore, the large number of receptor subtypes provides a mechanism for selective activation even if the VT neurotransmitter is massively diffused over an area. High affinity receptors on cells distant from the transmitter release can bind the neurotransmitter and cause a neuron response, while closer neurons without those specific receptor subtypes will not respond (18). Nicholson (16) (p. 439) stated: ‘‘ . . . there is merit . . . .in using a commonly agreed upon term to identify and categorize diverse research . . . ’’. Volume transmission has clearly been the term preferred by investigators, and thus will be used here. Volume transmission may be more common in certain brain regions; nonsynaptic interneuronal communication is very common in the greater limbic system (19) and may play an important role in the organization and regulation of behavior by the core and paracore regions of the brain (11). Volume transmission may also be an important mechanism in the highest levels of the human brain: physical anthropologists are puzzled by the fact that the human brain does not use more energy than the smaller brains of animals of comparable corporal weight; Bach-y-Rita and Aiello (18) have suggested that the parts of the human brain that show the greatest size increase over other animals, such as prefrontal cortex, may be exactly those parts in which highly nonsynaptic-based functions have their neuronal representation. Music appreciation and intellectual functions, for example, may not require the high-frequency (and energetically costly) alternating cycles of activation–inactivation of synaptic transmission, and may be largely replaced by VT (18). Volume transmission may play a role in the recovery from brain damage, which causes changes in neurotransmitter levels. Some neurotransmitter systems are upregulated, while others are downregulated (20). Drug therapy and rehabilitation may induce functional recovery by influencing the affected neurotransmitter systems by both synaptic and VT mechanisms. Studies on the ECS, calculating the diffusion of ions in the brain stem microenvironment in the living brain (21), confirm the conclusions of van Harreveld (22) and others that about 20% of brain tissue is ECS, and microdialysis studies indicate that virtually all the neurotransmitters are found in the extracellular fluid (23,24), which confirms that the conditions exist for VT.
42
Bach-y-Rita
The concept of a neurotransmitter has broadened recently to include atypical neural messengers such as gasses (NO and others) as well as an exclusively glial neurotransmitter, d-serine (25). Fuxe and Agnati (2) pointed out the importance of understanding the structure of the ECS with its intercellular matrix of polyionic molecules; the structure of the brain cell microenvironment may depend on physical and chemical constraints represented by cellular elements such as astroglia as well as the extracellular matrix. Glial cells constitute a high proportion of the total cell number and have a strategic position with processes reaching both blood vessels and synaptic regions. Sykova´ and Chva´tal (26) (p. 397) noted that neurotransmitters in the extracellular fluid bind to extrasynaptically located high affinity binding sites on glia as well as on neurons. They noted that glial cells ‘‘ . . . play an important role in extrasynaptic transmission, for example in functions such as vigilance, sleep, depression, chronic pain, LTP, TTD, memory formation, and other plastic changes in the CNS.’’ Thus, glia appear to have a critically important role in neurocommunication. Volume transmission provides a conceptual framework for understanding how glia can play such an important role in the absence of nerve fibers connecting glia to neurons.
2.
THE HISTORY OF VOLUME TRANSMISSION
Neurotransmitters and what have been called ‘‘modulators’’ are volume transmitters, as are gases such as NO that diffuse across the cell membrane [Snyder (27) has called NO ‘‘an important neurotransmitter’’]. The concept of VT has slowly entered the mainstream of the neurosciences. The delay in acceptance is not surprising; even the concept of the synapse took many years to become accepted, and in fact in 1987 the Editor of the journal SYNAPSE considered the synapse itself to be a ‘‘concept in evolution’’ (28). Schmitt (29) considered VT (he called it ‘‘parasynaptic’’ information transmission) to represent new ways of conceptualization of informationtransactional chemical processes as applied to basic concepts of neurobiology. He hypothesized that the parasynaptic system (in which informational substances reach specific target cell receptors by diffusion through the extracellular fluid from release points) has the same specificity as synaptic circuitry, but possesses, in addition, a domain of versatility and plasticity lacking in ‘‘hardwired’’ circuitry. Schmitt (30) considers that we are emerging from the ‘‘(classical) neurotransmitter-centered’’ era of neurophysiology and neurochemistry, and that we are now embarking on a historic period of discovery. He considers that this may represent a paradigm shift, or in the Kuhn (31) view, ‘‘revolutionary’’ science. In the 1950s, in order to study the action of GABA analogues on thalamic mechanisms, Killam and his colleagues (32,33) injected them into
Volume Transmission: Role in Vestibular Substitution
43
cat ventricles, since they did not cross the blood–brain barrier. At the time, we did not consider diffusion neurotransmission as a mechanism of action. However, later intra- and extracellular microelectrode studies in cat brainstem of highly convergent (responding to stimulation of several modalities from various parts of the body) brain cells following sensory stimulation, revealed very long latency responses, of up to 4 secs and more. These could not be ascribed to classical synaptic mechanisms in spite of many surgical and physiological experimental efforts to do so. This resulted in the suggestion that diffusion neurotransmission was the mechanism, and that it could play a role in the multiplexing of the polysensory cells (34). Neurohumoral transmission by local diffusion had been proposed as a process of the long-latency activity. This led to a proposed law of the conservation of space and energy in the brain (35). These conceptual issues have been discussed elsewhere (4,36,37). In 1968, Fuxe et al. (38) began a series of immuno-histochemical studies that led to highly productive VT studies (reviewed in Refs. 2, 3). Beaudet and Descarries (39) showed in 1978 that the biogenic amines released from nonsynaptic varicosities may act not only upon adjacent postsynaptic surfaces, but also in tissue of more distant receptor elements. Descarries has repeatedly confirmed and extended his earlier findings (cf., Ref. 3). Intrinsic and relational features of the monoamine and acetylcholine innervations are comparable during their early postnatal development to what they described in the adult, including their largely asynaptic character, which is apparent from the very beginning. This is now clear from a recent report (40) of the acetylcholine innervation in the developing parietal cortex of rat. Neurochemical transmission by extrasynaptic diffusion had also been noted by Vizi (41,42), whose studies on the nonsynaptic modulation of transmitter release led him to propose (43) that the essence of brain function (e.g., learning, and thinking) lies not in variations of neuronal circuitry (hardware) but rather within the chemical communication itself which is partly wired (synaptic) and unwired (nonsynaptic) transmission. The important role of glia in neuronal communication has been increasingly recognized in recent years. Sykova´ and her associates (cf., Ref. 3) have shown how VT provides a conceptual framework for understanding how glia can participate in neurotransmission in the absence of nerve fibers connecting glia to neurons. Sykova´ and Chva´tal (26) also noted that during activation, neurons release not only neurotransmitters, but also a variety of other neuroactive substances such as ions that diffuse through the extracellular fluid to other cells. Individual movements or functions, such as playing the piano, or watching a tennis game, require great selectivity, rapid initiation, and rapid ending; for such functions, synaptic action is essential. However, for mass sustained functions (e.g., sleep, mood, and hunger), sustained, widespread activity (rather than speed and selectivity) is required, which appear to be
44
Bach-y-Rita
largely mediated by VT (4,12). Many functions may be produced by combinations of both types of neurotransmission. We have calculated space and energy figures for synaptic vs. nonsynaptic neurotransmission (see below) and have concluded that it is very unlikely that synaptic neurotransmission could be the exclusive means of information transmission in the brain. In the piano playing example presented above, in addition to the relevant synaptic mechanisms, the finger movements can be more precise in the presence of adequate preparation including changes in brain tone (probably mediated by noradrenaline), and the visual perception of the tennis game may require neuronal receptivity to be set at a high level, probably involving several neurotransmitters, including NO (44) and dopamine (45) in the retina, serotonin and histamine in the lateral geniculate nucleus (46,47), and noradrenaline in the visual cortex (48,49). These effects appear to be primarily nonsynaptically mediated. Some of them have been called modulation; the modulation of synaptic activity by diffusion outside the synaptic gap is also a nonsynaptic, diffusion-mediated activity. 3.
COMPUTATIONAL NEUROSCIENCE VT STUDIES
How information is carried from one part to another in the brain has a great influence on the brain’s space and energy. Volume transmission may provide a low metabolic-cost alternative to synaptic communication. The cost of an action potential has been calculated (50) to be about 1011–1012 ATP per cm2 of depolarized membrane, with a minimum cost of 106 ATP at a node of Ranvier. Fully synaptically connected cell-assemblies—of the type Hebb (51) envisioned to explain mental states, such as thought, expectancy, interest, and attention—might exceed metabolic resources: a small (10,000 neuron) module would require approximately 1 J=L of brain. Furthermore, the volume of the nerve fibers necessary to fully connect the brain synaptically would create a brain so large that it would not fit in the head (7). However, rather than purely synaptic or purely diffusive, it is likely that hybrid neuronal networks, exhibiting varying combinations of hard-wired (synaptic) and wireless (VT) connectivities, would fit into the available volume in the brain, with a significant cut in space and energetic costs resulting from removing a significant number of ‘‘hard’’ links (and the nerve fibers needed to connect them) and replacing them with ‘‘soft,’’ diffusive pathways. 3.1.
Space Issues
Hebb imagined cell-assemblies as modules comprising enormous numbers of neurons, each connected directly to every other. Hebb stated that the connections within a cell-assembly are by means of nerve fibers and synapses.
Volume Transmission: Role in Vestibular Substitution
45
He considered cell-assemblies to be diffuse, anatomically irregular structures that function briefly as closed systems, and do so only by virtue of the time relations in the firing of constituent cells. He could not explain the up to half second delays, which, he noted, is characteristic of thought. However, delays are a characteristic of VT model of cell-assemblies; as noted elsewhere, the initial VT studies revealed delays up to 4 sec, which were considered to be related to multiplexing of polysensory reducing the numbers of cells needed for sensory messages (34). Hebb’s view suggests that connectivity should be ‘‘high’’ in cell-assemblies, possibly up to the extreme case of a fully connected cell module. Such connectivity, however, must be sought beyond the classical connectionist scheme. It is quite likely that the brain does not waste energy or space. Synaptic transmission requires nerve fibers and alternating cycles of activation– inactivation (nerve spike), which allows the nerve spikes to occur in rapid sequence, providing fine functional control. When such fine control is not needed (e.g., for mass sustained functions such as hunger) it appears to be excessively costly to use the synaptic mechanism. Thus, I suggested, based on these studies, a Law of the Conservation of Space and Energy in the Brain (Bach-y-Rita, 1996a). Rather than purely synaptic or purely diffusive, it is likely that hybrid neuronal networks, exhibiting varying combinations of hard-wired (synaptic) and wireless (VT) connectivities, would fit into the available volume in the brain, with a significant cut in space and energetic costs resulting from removing a significant number of ‘‘hard’’ links (and the nerve fibers needed to connect them) and replacing them with ‘‘soft,’’ diffusive pathways. The burgeoning field of Artificial Neural Nets is based on the concept of a fully synaptically connected brain. However, in view of the role VT may play in the brain, we have explored the development of partially synaptically connected neural nets (52) 3.2.
Energy Issues
In Hebb’s view (51), waves of depolarization spread along an impressive network of links, carrying information all over the assembly. Hebb did not consider how much energy would be necessary to drive such a network. The estimate depends on the nature of the links. Unmyelinated links require depolarization of the entire link, while myelinated nerves require only depolarization of the short unmyelinated segments at the nodes of Ranvier. The cost of an action potential (AP) has been calculated (50) to be about 1011–1012 ATP per cm2 of depolarized membrane, with a minimum cost of 106 ATP at a node of Ranvier. In the most conservative scenario of fine myelinated links, with Ranvier nodes spaced d apart, the axonal cost of circulating a single AP through a network of total length L would exceed 106 L=d ATP (1013 L=d Joules). With d ¼ 1 mm (the average distance between
46
Bach-y-Rita
nodes), a network of 105 cells connected in a ratio 1:127, would require at least 1.27 106 J. The cost per liter (assuming that all 1010 neurons in a liter are arranged in 105 modules, each comprising 105 cells) would be 0.127 J=L=AP. A connectivity ratio of 1:127 is quite conservative. For example, a spike rate of the order of 100 Hz would be required in order to achieve the measured 40 W=L brain metabolic power at rest, determined from glucose consumption rates in rats (53). On the other hand, full-connectivity would yield an energetic figure of 100 W=L=AP, with an impossible metabolic demand of 1 KW=L at a spike rate as low as 10 Hz. Thus, the hard-wired scheme may easily conflict with brain metabolic resources, unless it is limited to modules of a few thousands cells, with limited branching. Brain functions requiring the participation of large modules and profuse branching would be incompatible with brain metabolic resources. In this case, additional mechanisms of a ‘‘wireless’’ type may offer a less expensive alternative to synaptic communication. The ECS in the brain plays a role in many functions. In brain cellassemblies, the distance between neurons can be reduced by 50% with neuronal activity that causes cell swelling (53,54). A decreased volume fraction has a direct effect on the ionic concentrations, and may affect cell excitability and metabolism (55). Neurotransmitters that may have escaped the synaptic cleft, or survived enzyme degradation, or have originated extrasynaptically (4,39) are also found diffusing freely in the ECS. They carry information by means of VT. We have studied a model of a diffusion-only artificial neural net (56,57), which would comply with space and energy constraints in brain, yet one that allows the cells to directly communicate with each other, although via diffusive paths. More realistically, a hybrid neuronal network, exhibiting varying combinations of hard-wired (‘‘synaptic’’) and wireless (VT) connectivities, would fit into the available volume in the brain, with a significant cut in the energetic costs. The energetic cost of handling information through diffusive pathways (the VT cost) is basically that of restoring the presynaptic membrane’s free energy after neurotransmitter release. The synaptic vesicle cycle (58) forms the basis for neurotransmitter release, therefore all the events in this cycle should be checked for energy cost. Evaluation of all these contributions to the cerebral metabolism is quite difficult. However, uncertain and controversial the available data may be, they seem to support the conclusion of Albers and Siegel (59) that the energy expenditure for biosynthetic processes, like vesicle recycling and turnover of neurotransmitters, is relatively small, probably less than 10% of the total ATP utilization. Furthermore, the process of restoring the free energy of the presynaptic terminal is carried out by secondary transport systems, that do not require a local source of energy (as occurs in ATP-coupled primary transport) but are fuelled indirectly by the activity of the Na,K-ATPase.
Volume Transmission: Role in Vestibular Substitution
47
It is estimated that 25–40% of total brain energy is related to Na,KATPase activity. Most of this energy appears to be utilized for recharging the Nernst batteries (60) after membrane depolarization, which approximately equals the axonal cost (50). Thus, the VT cost is expected to be small (although perhaps significant) compared to the axonal cost. Therefore, removing a few axons and replacing them with diffusion pathways would have an energy saving effect.
4.
CARRY-OVER FUNCTIONAL EFFECT IN TACTILE VESTIBULAR SUBSTITUTION?
Tactile sensory substitution studies have included visual (61–63) and vestibular (64) substitution, with both behavioral, evoked potential, and Positron Emission Tomography (PET) scan evidence of late brain reorganization (1,64–66). This brief report will concentrate on tactile vestibular substitution because a remarkable finding has been the carry-over of functional improvement for a period of minutes to hours after the substitution system is removed. The persistence of functional improvement will be discussed in the context of possible VT mechanisms of brain reorganization. Persons with bilateral vestibular damage (BVD), such as from the ototoxicity of gentamicin, experience functional difficulties that include postural ‘‘wobbling,’’ unstable gait, and oscillopsia. Bilateral vestibular damage patients can stand only with difficulty, making conscious efforts to integrate a range of visual and tactile cues. The tactile vestibular substitution system (TVSS) includes a head-mounted accelerometer and an electrotactile human machine interface (HMI), through the tongue. The use of the TVSS produces a strong stabilization effect on head–body coordination in subjects with bilateral vestibular damage. Under these conditions, we identified three characteristic and unique head motion features (drift, sway, and long-period perturbations) that consistently appear in the head-postural behavior of subjects with bilateral vestibular damage, but are greatly reduced or eliminated with the TVSS. Head motion studies of the subjects show that the stability that the victim obtains from the accelerometer data approaches that of the person with normal vestibular function. Other head motion studies show that it is not a placebo effect (64). Dynamic and spectrographic analysis of head posturograms accompanying functional vestibular substitution through the tongue may provide clues regarding neural mechanisms of sensory substitution. Power spectra of the head posturogram in normal subjects consist of patterns of discrete peaks in a range from 1 to 15 Hz. In BVD patients, the normal spectral pattern was severely depressed in the 2–15 Hz frequency range, and the power in the spectra shifted to the 0.15–2.0 Hz range. Shift to lower frequency correlated with the severity of the vestibular damage (67).
48
Bach-y-Rita
An unexpected remarkable finding in this study has been the carryover of functional improvement after removing the sensory substitution system (64). This effect lasts a few minutes after a short (e.g., 3 min) sensory substitution session (64) and up to several hours after a 20 min TVSS session. The improved function during this carry-over period is occurring in the apparent absence of the vestibular system. As far as is known, it is impossible to have vestibular function in the absence of a vestibular system. It has previously been shown that as little as 2% of surviving tissue in a system may be sufficient for functional reorganization (reviewed in Ref. 68). As an example, Galambos et al. (69) destroyed up to 98.5% of the optic track fibers in the cat and a few weeks after the lesions, they found almost normal visual behavior. Pattern discrimination was present and visual evoked potentials had approximately the same amplitude as preoperatively. Previously, Lashley (70) had demonstrated that complex visual discrimination and visual behavior could be maintained in the rat if only 2% of the visual cortex remained. Thus, our working hypothesis is that remnants of the system escape destruction by the ototoxic substance. In the absence of valid vestibular data, the system appears inherently noisy and unstable, and may be operating as an open-loop control system. The TVSS appears to close the loop. The brain reorganization necessary for the functional carry-over occurs quickly, and it is likely that synaptic as well as VT receptor up- and downregulation are involved in the changes. A goal of this review is to emphasize the existence of nonsynaptic mechanisms that should be considered when the neural mechanisms of brain plasticity such as the vestibular functional carry-over are explored.
REFERENCES 1. 2. 3. 4. 5.
6. 7.
Zoli M, Jansson A, Sykova E, Agnatti LF, Fuxe K. Volume transmission in the CNS and its relevance for neuropsychopharmacology. TIPS 1999; 20:142–150. Fuxe K, Agnati LF. Volume Transmission in the Brain. New York: Raven Press, 1991. Agnati L, Fuxe K, Nicholson C, Sykova´ E. Volume transmission revisited. Prog Brain Res 2000; 125. Bach-y-Rita P. Nonsynaptic Diffusion Neurotransmission and Late Brain Reorganization. New York: Demos-Vermande, 1995. Vizi ES. Role of high affinity receptors and membrane transporters in nonsynaptic communication and drug action in the central nervous system. Pharmacol Rev 2000; 52:63–90. Bach-y-Rita P. Psychopharmacologic drugs: mechanisms of action (Letter). Science 1994; 264:642–644. Bach-y-Rita P, Aiello GL. Nerve length and volume in synaptic versus diffusion neurotransmission: a model. NeuroReport 1996; 7:1502–1504.
Volume Transmission: Role in Vestibular Substitution 8.
9.
10. 11.
12.
13. 14.
15.
16. 17.
18. 19. 20.
21.
22.
23. 24.
49
Stjarne L, Lundbert JM. On the possible roles of noradrenaline, adenosine 50 -triphosphate and neuropeptide Y as sympathetic cotransmitters in the mouse vas deferens. Brain Res 1986; 68:263–278. ˚ , Stja¨rne L. Dense cored vesicles in actively secreting noradreThureson-Klein A ˚ , eds. nergic neurons. In: Stja¨rne L, Hedqvist P, Lagercrantz H, Wenmalm A Chemical Neurotransmission 75 Years. London: Academic Press, 1981:153–164. Bach-y-Rita P. Nonsynaptic diffusion neurotransmission (NDN) in the brain. Neurochem Int 1993; 23:297–318. Nieuwenhuys R. Comparative aspects of volume transmission, with sidelight on other forms of intercellular communication. Volume Transmission Revisited. 2000; 125:49–126. Bach-y-Rita P. Thoughts on the role of volume transmission in normal and abnormal mass sustained functions. In: Fuxe K, Agnati LF, eds. Volume Transmission in the Brain. New York: Raven Press, Ltd, 1991:489–496. Bach-y-Rita P, Illis L. Spinal shock: possible role of receptor plasticity and non synaptic transmission. Paraplegia 1993; 31:82–87. Bach-y-Rita P. Possible role of nonsynaptic diffusion neurotransmission (NDN) and receptor plasticity in shoulder-hand and autonomic dysreflexia syndromes. Scand J Rehab Med 1994; 26:33–35. Agnati LF, Fuxe K, Zoli M, Toffano G, Ferraguti F. A correlation analysis of the regional distribution of central enkephalin and betaendorphin immunoreactive terminal and of opiate receptors in adult and old male rats. Evidence for the existence of two main types of communication in the central nervous system: the volume transmission and the wiring transmission. Acta Physiol Scand 1986; 128:201–207. Nicholson C. Volume transmission in the year 2000. Prog Brain Res 2000; 125:437–446. Bjelke B, England R, Nicholson C, Rice ME, Lindberg J, Zoli M, Agnati LF, Fuxe K. Long distance pathways of diffusion for dextran along fibre bundles in brain. NeuroReport 1995; 6:1005–1009. Bach-y-Rita P, Aiello GL. Brain energetics and evolution. Brain Behav Sci 2001; 24:280. Nieuwenhuys R. Chemoarchitecture of the Brain. Berlin: Springer, 1985. Westerberg E, Monaghan DT, Kalimo H, Cotman CW, Wieloch TW. Dynamic changes of excitatory amino acid receptors in the rat hippocampus. J Neurosci 1989; 9:798–805. Nicholson C, Rice ME. Diffusion of ions and transmitters in the brain cell microenvironment. In: Fuxe K, Agnati L, eds. Volume Transmission in the Brain. New York: Raven, 1991:279–294. Van Harreveld A. The extracellular space in the vertebrate central nervous system. In: Bourne GH, ed. The Structure and Function of Nervous Tissue. Vol. 4. New York: Academic Press, 1972:447–511. Di Chiara G. In-vivo brain dialysis of neurotransmitters. Trends Pharmacol Sci 1990; 11:116–121. Ungerstedt U. Measurement of neurotransmitter release by intracranial dialysis. In: Marsden CA, ed. Measurement of Neurotransmitter Release in vivo. Chichester: John Wiley & Sons Ltd, 1984:81–105.
50
Bach-y-Rita
25. Baran˜ano DE, Ferris CD, Snyder SH. Atypical neural messengers. TINS 2001; 24:99–106. 26. Sykova´ E, Chva´tal A. Glial cells and volume transmission in the CNS. Neurochem Int 2000; 36:397–409. 27. Snyder SH. Janis faces of nitric oxide. Nature 1993; 364:577. 28. Johnson JE. Editorial. Synapse 1987; 1:1–2. 29. Schmitt FO. Molecular regulators of brain function: a new view. Neuroscience 1984; 13(4):991–1001. 30. Schmitt FO. Chemical information processing in the brain: prospect from retrospect. In: Iverson LL, Goodman E, eds. Fast and Slow Chemical Signalling in the Nervous System. New York: Oxford University Press, 1986:239–243. 31. Kuhn TS. The Structure of Scientific Revolutions. 2nd ed. Chicago: University of Chicago Press, 1970. 32. Bach-y-Rita P, Hance AJ, Winters WD. Some effects of atropine and eserine on thalamocortical pathways. Proc Western Pharmacol Soc 1960; 3:69–79. 33. Hance AJ, Winters WD, Bach-y-Rita P, Killam KF. A neuropharmacological study of gamma-aminobutyrylcholine, gamma-aminobutyric acid, phyostigmine and atropine. J Pharmacol Exp Therapeut 1963; 140:385–395. 34. Bach-y-Rita P. Convergent and long latency unit responses in the reticular formation of the cat. Exp Neurol 1964; 9:327–344. 35. Bach-y-Rita P. Conservation of space and energy in the brain. Restor Neurol Neurosci 1996; 10:1–3. 36. Bach-y-Rita P. Nonsynaptic diffusion neurotransmission and some other emerging concepts. Proc Western Pharmacol Soc 1998; 41:211–218. 37. Bach-y-Rita P. Nonsynaptic diffusion neurotransmission in the brain: functional considerations. Neurochem Res 2001; 26:871–873. 38. Fuxe K, Hamberger B, Ho¨kfelt T. Distribution of noradrenaline nerve terminals in cortical areas of the rat. Brain Res 1968; 8:125–131. 39. Beaudet A, Descarries L. The monoamine innervation of rat cerebral cortex: synaptic and nonsynaptic axon terminals. Neuroscience 1978; 3:851–860. 40. Mechawar N, Watkins KC, Descarries L. Ultrastructural features of the acetylcholine innervation in the developing parietal cortex of rat. J Comp Neurol 2002; 443:250–258. 41. Vizi ES. Presynaptic modulation of neurochemical transmission. Prog Neurobiol 1979; 12:181–290. 42. Vizi ES. Non-synaptic modulation of transmitter release: pharmacological implications. Trends in Pharmacol Sci 1980; 2:172–175. 43. Vizi ES. Nonsynaptic inhibitory signal transmission between axon terminals: physiological and pharmacological evidence. In: Fuxe K, Agnati LF, eds. Volume Transmission in the Brain. New York: Raven Press, 1991:89–96. 44. Bredt DS, Hwang P, Snyder SH. Localization of nitric oxide synthase indicating a neural role for nitric oxide. Nature 1990; 347:768–770. 45. Daw NW, Brunken WJ, Parkinson D. The functions of synaptic transmitters in the retina. Annu Rev Neurosci 1989; 12:205–225. 46. McCormick DA. Cholinergic and noradrenergic modulation of thalamocortical processing. TINS 1989; 12:215–221.
Volume Transmission: Role in Vestibular Substitution 47. 48.
49. 50. 51. 52.
53. 54. 55.
56. 57. 58.
59.
60. 61. 62. 63. 64. 65. 66.
51
Wilson JR, Hendrickson AE. Seratonergic axons in the monkey’s lateral geniculate nucleus. Vis Neurosci 1988; 1:125–133. Kasamatsu T, Pettigrew JD. Depletion of brain catecholamines: failure of ocular dominance shift after monocular occlusion in kittens. Science 1976; 194:206–209. Pettigrew JD. The locus Coeruleus cortical plasticity. Trends Neurosci 1978; 1:73–14. Aiello GL, Bach-y-Rita P. The cost of an action potential. J Neurosci Methods 2000; 103:145–149. Hebb DO. The Organization of Behavior. New York: Willey, 1949. Aiello GL, Bach-y-Rita P. Hebbian brain cell-assemblies: nonsynaptic neurotransmission, space and energy considerations. In: Cihan D, Buczak AL, Embrechts MJ, Ersoy O, Ghosh J, Kercel SW, eds. Intelligent Engineering Systems through Artificial Neural Networks Vol. 11. New York: ASME Press, 2001:441–447. Aiello GL, Bach-y-Rita P. Brain cell micro environment effects on neuron excitability and basal metabolism. NeuroReport 1997; 8:1165–1168. Dietzel MA, Heinemann I. Dynamic variations of the brain cell microenvironment in relation to neuronal hyperactivity. Ann NY Acad Sci 1986; 481:72–86. Dietzel MA, Heinemann I, Hofmeier U, Lux HD. Transient changes in the size of the extracellular space in the sensorimotor cortex of cats in relation to stimulus-induced changes in potassium concentration. Exp Brain Res 1980; 40: 432–439. Aiello GL, Bach-y-Rita P. Brain cell microenvironment effects on neuron excitability and basal metabolism. NeuroReport 1997; 8:1165–1168. Aiello GL. Diffusive Neural Network. Neurocomputing 2001; In press. Su¨dhof TC. Intracellular trafficking. In: Siegel GJ, Agranov BW, Albers RW, Fisher SK, Uhler MD, eds. Basic Neurochemistry, Lippincott Williams & Wilkins, 1999:175–187 Albers RW, Siegel GJ. Membrane transport. In: Siegel GJ, Agranov BW, Albers RW, Fisher SK, Uhler MD, eds. Basic Neurochemistry, Lippincott Williams & Wilkins, 1999:95–118. Laughlin SB, Stevenick RRdRv, Anderson JC. The metabolic cost of neural information. Nat Neurosci 1998; 1:36–41. Bach-y-Rita P, Collins CC, Saunders F, White B, Scadden L. Vision substitution by tactile image projection. Nature 1969; 221:963–964. Bach-y-Rita P. Brain Mechanisms in Sensory Substitution. New York: Academic Press, 1972. Bach-y-Rita P, Tyler ME, Kaczmarek KA. Seeing with the brain. Int J Human—Computer Interac 2003; 15:287–297. Tyler ME, Danilov Y, Bach-y-Rita P. Closing an open-loop control system: vestibular substitution through the tongue. J Integr Neurosci 2003; 2:1–6. Sadato N, Pascual-Leone A, Grafman J, Deiber MP, Ibanez V, Hallett M. Neural networks for Braille reading by the blind. Brain 1998; 121:1213–1229. DeVolder AG, Catalan-Ahumada M, Robert A, Bol A, Labar D, Coppens A, Michel C, Veraart C. Changes in occipital cortex activity in early blind humans using a sensory substitution device. Brain Res 1999; 826:129–134.
52
Bach-y-Rita
67. Danilov Y, Tyler ME, Bach-y-Rita P. Analysis of head stability in normal and bilateral vestibular dysfunction subjects. J Neurosci Abstr Soc Neurosci Meeting, 2003. 68. Bach-y-Rita P. Brain plasticity as a basis for therapeutic procedures. In: Bachy-Rita P, ed. Recovery of Function: Theoretical Considerations for Brain Injury Rehabilitation. Bern, Switzerland: H. Huber, 1980:225–263. 69. Galambos R, Norton TT, Frommer GP. Optic tract lesions sparing pattern vision in cats. Exp Neurol 1967; 18:8–25. 70. Lashley KS. The mechanisms of vision. XVI. The functions of small remnants of the visual cortex. J Comp Neurol 1939; 70:45–67.
4 Lysosomal Dysfunction in Brain Aging and Neurodegeneration: Roles in Trophic Signaling and Neuroinflammation Xiaoning Bi Department of Psychiatry & Human Behavior, University of California, Irvine, Irvine, California, U.S.A.
Brain aging, defined as decline in brain function as a result of biological aging, is closely associated with lysosomal dysfunction. Changes in the numbers and enzymatic activities of lysosomes appear early during brain aging process; these changes appear to be common in several species of mammals and occur specifically in brain regions that are vulnerable to pathogenesis in age-related neurodegenerative diseases, including Alzheimer’s disease (AD). The present review first summarizes the correlative and experimental evidence for lysosomal dysfunction in brain aging and age-related neurodegeneration, then assesses the roles of lysosomal dysfunction in trophic factor signaling and in glia-mediated inflammation. Finally, the potential pathways through which lysosomal dysfunction induces neurodegeneration are outlined so as to provide novel avenues for therapeutic approaches for age-related neurodegeneration.
1. LYSOSOMAL DYSFUNCTION AND BRAIN AGING 1.1. Lysosomal Dysfunction as the Earliest Sign of Brain Aging Brain aging is characterized by decline in brain function produced as a result of biological aging, and is a complex process that involves a variety of 53
54
Bi
processes, including reduced mitochondrial function (1–7), changes in endosomal=lysosomal function (3,8), inflammatory responses (9–11), and altered regulation of Ca2þ and of other signaling factors (12,13). Perhaps one of the earliest phenomena of brain aging represents alterations in endosomal–lysosomal function, e.g., marked increases in the numbers of lysosomes that contain lipofuscin (also called the age-pigment) (14–17). Changes in the levels and activities of lysosomal enzymes also gradually develop over the course of aging. Activity of cathepsin L decreases by 90% from 2 to 28 months in rat brain (18) while the activity and concentration of cathepsin D increase from young adulthood to middle age and still further with old age (18–22). Changes in lysosomal enzymes, especially in the aspartyl protease cathepsin D, are also common in aged human brains. Intraneuronal concentrations of cathepsin D are increased in vulnerable regions of AD in advance of overt pathology (23,24). Increased levels of cathepsin D proteins correlate on a cell-by-cell basis with decreases in the vesicular protein synaptophysin and with the presence of intraneuronal neurofibrillary tangles, a hallmark of AD (25). A more recent study using in situ hybridization and combined immunohistochemistry confirmed the correlation of upregulation of cathepsin D with the presence of neurofibrillary tangles (26). There is also evidence that the compartmentalization of lysosomal hydrolases begins to break down as aging progresses: cytosolic cathepsin D activity increases significantly from 2 to 6 months of age in rat brain and by 36 months is twice that found within the lysosomal=mitochondrial fraction (19). Cathepsin D has also been found in the extracellular space and cerebrospinal fluid of individuals with AD (27). 1.2.
Age-Related Changes in Human Brain Aging Are Region Specific
Intriguingly brain aging and age-related neurodegenerative disease-induced changes are not randomly distributed in the brain; they occur preferentially in certain brain regions. Neurofibrillary tangles, a defining feature of AD, occur earlier and with much greater frequency in the superficial layers of perirhinal and entorhinal cortex than in most areas of neocortex (28–35). Regional differentiation is also well documented for hippocampus, where tangles are significantly more frequent in CA1 than in CA3 neurons (33,36–38). Aberrant swellings of the initial segments of axons (‘‘meganeurites’’), a feature that occurs with increasing frequency in human brains beginning at about age 50, show a striking regional selectivity (e.g., in CA1 but not in CA3 of hippocampus), which matches well with the distribution of tangles (17). Dendritic abnormalities (39–41), including the loss of branches and spines, associated with AD have a distribution that either overlaps that for tangles per se (e.g., in CA1 and subiculum) or the targets
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
55
(e.g., dentate granule cells) of projecting neurons that are prone for tangles (see Ref. 42 for a recent review). 1.3.
Lysosomal Dysfunction Is Common in Brain Aging Processes Among Mammals
If lysosomal dysfunction plays important roles in brain aging, pathologic changes in lysosomes as well as their regional specific distribution are expected to exist in other mammalian species as well. This assumption was first tested in rat brains using cathepsin D as a marker (43). A marked increase in cathepsin D immunostaining was found in the brains of 12-month-old rats compared to those of 2-month-old young adults. The changes were particularly pronounced in the superficial layers of entorhinal cortex and in hippocampal field CA1, but much less in other areas of neocortex and field CA3 of hippocampus. Changes in cathepsin D entitled three categories: an increase in the intraneuronal area occupied by labeled lysosomes, the formation of lysosomal clusters within neurons, and, in some neurons, the appearance of intense cytoplasmic cathepsin D staining. Apparent shrinkage was also observed in entorhinal cortical neurons exhibiting changes in cathepsin D immunostaining but not in other neocortical regions. These results provided the first indication that aspects of brain aging might be common to mammals. Regional brain aging patterns were further tested in another mammal that has a much longer lifespan, the dog, again using cathepsin D immunostaining as a marker (44). Accumulations of cathepsin D immunopositive material were evident in neuronal cell bodies in many areas of telencephalon of dogs from middle age to old dogs (6 years). Three types of changes could be distinguished: (1) dense aggregates with no particular position within the cell body; (2) crescent-shaped ‘‘caps’’ that occupied one pole of the cell body; and (3) very dense ‘‘spikes’’ that extended from the cell body for variable distances into the apical dendrites. Of particular interest were the regional specific localization of spikes that were found only in the external pyramidal layer of the subiculum and layer V of neocortex. Underlining the significance of these findings, an identical dendritic agglomeration of ‘‘age pigments’’ was reported in the same layer of subiculum in human brains (45), while dendritic regression was found selectively in layer V of aged human brain (46). The numbers of spikes were further increased in the subiculum in AD brains. These results establish that brain aging in dogs is (1) well advanced by middle age, (2) varies markedly across regions, and (3), in at least some of its aspects (dystrophic dendrites), is prominent in areas known to exhibit pathology early in the course of AD. Findings from the second mammal study further confirm that changes in lysosomal enzymes, in particular in the temporal lobe, reflect a generalized mammalian pattern of brain aging.
56
1.4.
Bi
Lysosomal Dysfunction Induces Characteristic Features of the Aged Human Brain
Experimental testing of the hypothesis that lysosomal perturbations contribute in an important way to brain aging began with studies using infusions of two functionally distinct, broad spectrum inhibitors, leupeptin (a thiol proteinase inhibitor) or chloroquine (a general lysosomal enzyme inhibitor), into the ventricles of young, adult rats (47). Both drugs caused a rapid and dramatic increase in the (i) number of lysosomes and (ii) concentration of ceroid-lipofuscin. In subsequent studies, Ivy et al. (48) demonstrated that these effects were followed shortly by intraneuronal accumulations of hyperphosphorylated tau and distended initial axon segments. More recently, studies using cultured hippocampal slice preparation showed that partial lysosomal dysfunction caused by suppression of cathepsins B and L generated characteristic features of the aging brain, including the following: (i) Cterminal fragments of amyloid precursor proteins (49), (ii) endosomal–lysosomal hyperplasia (50–52), (iii) increased levels of cathepsin D (53,54), (iv) leakage of cathepsin D into the cytoplasm (54), and (v) meganeurites (50,52). Meganeurites are common features of the human brain after age 50 (17), and are very large swellings filled with fused lysosomes found within the initial segments of axons. More importantly, meganeurites were found only in CA1 but not in CA3 of hippocampus (50), in layer III but not in layer V of neocortex (52) and in layer II–III but not in deep layers of entorhinal cortex (51), which precisely recapitulates what happens in aged human brains. In this paradigm at least, lysosomal dysfunction in rat brain slices not only reproduced a characteristic of brain aging in humans but also the regional pattern of aging as well. It would appear that experimentally induced lysosomal dysfunction facilitates some age-related process(es) in cultured rat slices at such a rate that the resulting changes are more than ‘‘natural brain aging’’ in the rat and closer to the process of human brain aging. Evidence reviewed in the previous sections demonstrated that early onset alterations in endosomal=lysosomal function and certain hydrolases in the telencephalon are a common brain-aging phenomenon in mammals while regional specificity varies among different species. In the rat, the most severely affected zones are the entorhinal cortex and CA1 region of hippocampus, whereas in dog the subiculum, hippocampus, and layer V of the neocortex are the first regions exhibiting lysosomal abnormality. In human brains, all these regions show lysosomal dysfunction and are particularly vulnerable to brain aging and AD-related pathology. Accordingly, regionally differentiated neurosenescence in humans (35,46,55,56) could, in part, be the outcome of a general feature of mammalian brain aging. Thus, mounting evidence implicates lysosomes in brain aging and related neurodegeneration; however, the precise cellular and molecular links are still missing.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
57
2.
LYSOSOMAL DYSFUNCTION AND TROPHIC ENDOSOMAL SIGNALING 2.1. Lysosomal Dysfunction and Axonal Transport of Neurotrophins Neurotrophins are essential factors for neuronal survival and differentiation, neurite outgrowth and synapse formation during development and for maintaining intact neuronal circuits in adulthood as well. Downregulation of neurotrophins has been implicated in several neurodegenerative diseases and applications of exogenous neurotrophins ameliorate neuronal degeneration in many model systems for neurological diseases. Neurotrophin signaling is initiated by binding of the neurotrophin to specific receptors, which results in activation of local signaling cascades and induction of the expression of genes that are important for cell survival. To induce gene expression, neurotrophin–receptor complexes are first internalized through endocytosis followed by retrograde transport to cell bodies since receptors for neurotrophins are usually located at axonal terminals. Perturbation of either endocytosis or intra-axonal transport would certainly affect trophic factor-mediated neuroprotection. Lysosomal dysfunction impairs both dendritic (57) and axonal transport (58) possibly through truncation of tau, disruption of microtubules, and formation of neurofibrillary tangles. Furthermore, lysosomal dysfunction could also impair endocytosis.
2.2.
Lysosomal Dysfunction and Endosomal Signaling
Endocytosis has been linked to a variety of receptor-mediated signaling pathways including those for receptor tyrosine kinases, G-protein coupled receptors, and cell adhesion receptors (59). While the traditional view is that endocytosis through clathrin-coated pits functions as a mechanism to switch off plasma membrane receptor signaling, more recent evidence indicates that retrogradely transported endosomes might also mediate signal propagation, the so called ‘‘signaling endosome hypothesis.’’ Several lines of evidence support this hypothesis. First, ligand–receptor complexes with receptors in their active states are found inside endosomes. Howe et al. (60) showed that nerve growth factor (NGF) treatment resulted in the formation of an endocytotic complex consisting of clathrin, clathrin adaptor protein2 (AP2), and activated TrkA, and that NGF signaling was blocked by inhibition of endocytosis. The same group also showed that clathrin-coated vesicles contained NGF-activated TrkA, extracellular signal-regulated kinase (ERK), phosphatidylinositol-3 kinase, and phospholipase C-g signaling cascades (60). Second, activation of ERKs by receptor tyrosine kinases and G protein-coupled receptors was decreased by expressing dominant-negative mutants of dynamin or b-arrestin, proteins that play important roles in intracellular trafficking (see Refs. 61,62
58
Bi
for recent reviews). Finally, this hypothesis is further supported by the finding that NGF internalization from distal neurites was necessary for the activation of cAMP-response-element-binding protein in cell bodies, which involves dynamin-dependent endocytosis of activated Trk receptors, relocation of Trk receptors to cell bodies, and activation of an ERK5dependent pathway (63,64). The association of Trk receptors with the dynein motor machinery implies a direct link between Trk activation, endocytosis, and retrograde transport (65). More recently, Wang et al. (66,67) developed a system to test trophic signaling of growth factor receptors, specifically from endosomes, without activation of membraneassociated receptors. The system works by inducing internalization of growth factor receptors in the presence of receptor tyrosine kinase inhibitors and receptor recycling blockers, which leads to the accumulation of inactive receptor–ligand complexes in endosomes. Upon removal of tyrosine kinase inhibitors, receptors inside the endosomes become active and trigger signaling cascades. Using this system, these authors provided evidence that both the epidermal growth factor receptor (EGFR) and the platelet-derived growth factor receptor (PDGFR) were able to form signaling complexes with various downstream proteins within endosomes, and endosomal EGFR=PDGFR signaling was sufficient to activate major signaling pathways involved in cell proliferation and survival (66–68). While the mechanisms involved in retrograde neurotrophic signaling are still elusive, activated receptors are ultimately dissociated from their ligands and are inactivated in the endosomal=lysosomal system. Trophic factor receptors are then recycled back to cell membranes or undergo degradation; perturbation of receptor recycling or degradation results in deregulation of trophic signaling. Endosomal=lysosomal milieu plays important roles in the dissociation of receptors from their ligands, as in vitro studies indicate that inhibition of endosomal acidification leads to a decrease in insulin-IRK dissociation and insulin degradation (69). Trophic signaling depends not only on the levels of neurotrophins but also on the availability of their receptors and the responsiveness of the downstream factors. Interestingly, impairment of trophic signaling due to lack of responsiveness has been reported in neurons prepared from a mouse model of Niemann–Pick Type C disease (70). These results raise the possibility that endosomal= lysosomal dysfunction impairs neurotrophic signaling. 3.
LYSOSOMAL DYSFUNCTION AND NEUROINFLAMMATION 3.1. Neuroinflammation in Age-Related Neurodegenerative Diseases Brain aging is also associated with early activation of both astrocytes and microglia, before any detectable neuropathology. Levels of mRNA and
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
59
proteins for glial fibrillary acidic protein (GFAP), a marker of astrocytes, increase progressively during aging in both humans and rodents (71–76). Morphologic analyses showed that changes in astrocytes during aging included increases in the size (77–79) and numbers (71,80–82) of astrocytes. Interestingly, the reactive properties of astrocytes from aged animals were maintained even after being cultured in vitro. Primary cortical astrocytes prepared from 12- and 24-month old rats showed increased proliferation and transcription of GFAP (twofold higher in astrocytes from 12- and 24-month vs. 3-month-old) (83). Likewise, increases in the number and size of microglia are also present in aged humans and mammals. Furthermore, expression of a number of proteins that are linked to phagocytosis and antigenic presentation is also enhanced in microglia from aged brains compared to those from young animals. As is the case for astrocytes, in vitro experiments showed that microglia prepared from old rodents exhibited an amoeboid-like morphology and expressed antigens characteristic of an activated state (e.g., major histocompatibility complex class II). It is noteworthy that the aforementioned changes in astrocytes and microglia were closely associated with increases in lipofuscin and increased levels of cathepsins (18) during the aging process as well as in brains from patients with lysosomal storage diseases (see Sec. 3.2). Alzheimer’s disease is also characterized with marked inflammation in the brain, and treatment with anti-inflammatory drugs reduces the risk factor for AD and decline in cognitive performance of AD patients (84–89). Increased levels of cathepsins in activated microglia were found to be widely distributed, especially in the white matter of AD brains (44,90,91). However, the precise roles of microglia in AD pathogenesis are not clear. It has been shown that Ab peptides are taken up via scavenger receptors and are degraded in the lysosomes of microglia (92,93). Data from recent immunization studies indicate that microglia are also involved in vaccine-induced clearance of Ab peptides in brains of transgenic mice used as AD models (94). Thus, microglia-mediated degradation of Ab peptides may play an important role in the immunotherapy of AD. However, an ample literature also implicates microglia in Ab-induced neurotoxicity (95–98). Ab peptides have been reported to induce neuronal death indirectly by activating microglia to produce inflammatory mediators such as nitric oxide, cytokines, and reactive oxygen intermediates (95,97). Brain inflammation has also been implicated in other neurodegenerative diseases, including Parkinson’s disease (99,100), multiple sclerosis (101–103), amyotrophic lateral sclerosis (104–106), prion disease (86,107,108), HIV infection-induced neurodegeneration (109–111), stroke, and trauma (112– 114). As neurodegeneration is generally associated with glial activation and proliferation, it is unclear whether these pathological changes in glial cells
60
Bi
are simply a consequence of or contribute causally to neuronal degeneration in these diseases. Results from studies of microglial reaction in a lysosomal storage disease discussed below seem to favor the latter interpretation. 3.2.
Neuroinflammation in Niemann–Pick Type C Disease
Niemann–Pick type C disease (NPC) is a rare and fatal neurovisceral storage disorder that is currently untreatable. Patients affected with the disease commonly exhibit ataxia, dystonia, vertical supranuclear gaze palsy, and dementia (115,116). In most cases (95%), the disease is caused by mutations in the NPC1 gene (117), while the reminders by mutations in the NPC2 gene; both genes encode proteins that play important roles in intracellular cholesterol transport (118–120). Accumulation of unesterified cholesterol and other lipids in endosomes and lysosomes of peripheral and brain cells is the pathological hallmark of the disease (118,121–123). In Npc1= mice, this accumulation occurs in neurons as early as postnatal day 9, as evidenced by Reid et al. (124), using a novel cholesterol stain. NPC brains are also characterized by axonal abnormalities, neurofibrillary tangles (NFTs), and progressive neurodegeneration (125–129). The mechanisms leading to pathology in NPC are not well understood. Cholesterol may play a role in NPC pathogenesis since defects in cholesterol metabolism are also evident in brains from patients with AD, another neurodegenerative disorder characterized by dementia, plaques, and NFTs (121,130–134). This suggestion is supported by the observation that, in both AD and NPC brains, free cholesterol levels are higher in neurons that contain NFTs than in those that do not (135,136). Altered cholesterol metabolism in NPC1 mutant cells leads to axonal and neuronal degeneration as well as to hyperphosphorylation of the tau protein, which is the major component of NFTs (137,138). Interestingly, the NFTs observed in NPC and AD human brains are chemically and morphologically indistinguishable (125,128), and, in AD, their numbers correlate with the severity of dementia (33,139). Furthermore, presenilin, a protein associated with increased Ab peptides in AD, accumulates in vesicular compartments in Chinese hamster ovary cells expressing mutant NPC1 (140). Redistribution of presenilin to endosomal system was also observed in Npc1= mice and was associated with enhanced production of b-amyloid peptides (141). Pathological changes in glial cells have also been reported in some mouse models of NPC (127,142–144). Recently, we have studied the inflammatory reaction in brains of Npc1= mice during early postnatal development and the results showed that numbers of microglia were markedly increased in several brain regions of Npc1= mice as early as one week of age (145). Microglial reaction was most prominent in regions vulnerable to the disease, including ventral lateral thalamus, medial geniculate nucleus,
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
61
and cerebellum. Astrogial reaction started later than microglia: at 2 weeks postnatal, reactive astrocytes were only observed in the ventral lateral thalamus while they were present throughout the brain of Npc1= mice at 4 weeks of age. Moreover, the astroglial reaction coincided with upregulation of the cytokine, interleukin-1b, in most, but not all brain regions. In particular, no interleukin-1b upregulation was observed in regions devoid of neuronal degeneration. These results suggest that microglial activation precedes and might be causally related to neuronal degeneration, while astrocyte activation might be a consequence of neuronal degeneration. Thus, disturbances in microglia function might play a critical role in the development of pathology observed in the disease.
3.3.
Neuroinflammation in Experimentally Induced Lysosomal Dysfunction
The involvement of lysosomal dysfunction in microglial reaction was further tested using experimentally induced partial lysosomal dysfunction in organotypic brain tissue cultures. Hippocampal slices prepared from 10- to 12- day-old rats were incubated in vitro for 12–14 days and then treated with the cathepsin B and L inhibitor, N-CBZ-L-phenylalanyl-l-alanine-diazo-methyl-ketone (ZPAD; 20 mM) (146,147), for 6 days. After treatment, cultured hippocampal slices were processed for immunohistochemistry using monoclonal antibodies ED-1 that detect reactive microglia. A few ramified microglia with slim cell bodies were present in CA1 region of hippocampal slices treated with vehicle (Fig. 1A). Inhibition of lysosomal function by ZPAD induced a marked increase in the number and size of microglia in rat hippocampal slices; microglia from the ZPAD-treated slices were larger and stained darker, suggesting they were in a ‘‘phagocytic state’’ (Fig. 1B). It is well known that lysosomes are actively involved in antigen processing and presentation; both are critical steps in initiating immune responses (148). Earlier studies showed that perturbation of lysosomal function by increasing lysosomal pH or inhibiting lysosomal proteases blocked MHCII-mediated antigen presentation. More recently, studies with knockout mice revealed that lysosomal cysteine proteases, especially cathepsin S and L, play important roles in both MHCII- and MHCI-mediated antigen presentation: deficiency in these two enzymes caused marked delay in antigen processing and defect in CD4þ T-cell selection (see Ref. 149 for a recent review). Although the precise implication of exogenous antigen presentation in the CNS is not fully understood, inflammatory cytokines secreted from activated helper T cells and microglia activated through their interaction may contribute to inflammation-induced brain damage and=or, in some cases, brain repair.
62
Bi
Figure 1 Lysosomal dysfunction-induced microglial activation in cultured hippocampal slices. Microglia labeled with ED-1 antibody in hippocampal slices cultured from 10- to 12-day-old rats were treated with vehicle (A) or 20 mM ZPAD (B) for 6 days.
4.
ROLES OF LYSOSOMAL DYSFUNCTION IN BRAIN AGING AND NEURODEGENERATION: A HYPOTHESIS
Although accumulation of lipofuscin has been recognized as the earliest sign of human brain aging for several decades (17,150), how lysosomal dysfunction results in decline in brain function remains unknown (14). It is noteworthy that accumulation of lipofuscin=cathepsin D positive lysosomes= endosomes occurs in most neurons throughout mammalian telencephalon; however, formation of intracellular clusters varies greatly among different regions and is often associated with neurodegeneration (17,43,44,150). Clusters of lipofuscin exist mainly in cortical projecting neurons with long
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
63
axons and spindle-shaped axons filled with pigments are only in layer V of neocortex and field CA1 of hippocampus of human brains (17,150). Dendritic pigment spindles are also found only in the subiculum of human telencephalic cortex (45) while they are in both the subiculum and layer V of neocortex of canine brains (44). The topology of postmitotic neurons is closely associated with their function; therefore it is not difficult to speculate the consequence of clustering of ‘‘leaking endosomes=lysosomes’’ in critical subcellular compartments such as the initial segment of an axon or the principle trunk of an apical dendrite. For instance, accumulation of lipofuscin-positive secondary lysosomes in the axon hillock would lead to the formation of meganeurites, which could then pinch off the cells and result in axotomy, as was previously proposed (50). Furthermore, aggregation of these elements in the axon hillock and apical dendrites would affect intraneuronal trafficking, and could result in degeneration of axonal terminals and dendritic spines. Along this line, it has been shown that the microtubule-associated protein tau can be cleaved by cathepsin D at neutral pH (54,151), a process that would not only impair intracellular trafficking, but also facilitate formation of neurofibrillary tangles (152–154). Signal transduction between neurons involves continuous exocytotic– endocytotic activities at both pre- and postsynaptic locations, which require normal endosomal=lysosomal function. Likewise, long distance trophic signaling also depends on an intact endosomal=lysosomal system. Sequestration of endosomes=lysosomes into ‘‘aggregates’’ would undoubtedly affect neuronal function and ultimately lead to synaptic and neuronal degeneration. Lysosomal=endosomal dysfunction would also affect other brain cell types. For instance, microglia, the brain residential macrophages, undergo a constant recycling of membrane constituents as part of their normal function, and it is conceivable that they are extremely susceptible to defect in the intracellular flow of cholesterol. Activation of microglia, due to intrinsic endosomal=lysosomal stress or reactive to damaged neurons, leads to release of oxidative free radicals and various cytokines, and further facilitates neurodegeneration. Furthermore, data from more recent studies indicate that microglial phagocytosis may play important roles in various neurodegenerative diseases (155,156). Potential pathways through which lysosomal= endosomal dysfunction may contribute to neurodegeneration are illustrated in Fig. 2. The hypothesis we have developed in this chapter also satisfies a very important feature of the aging process, that is the very slow time-course of the various alterations in cellular structure and functions. Although little is known regarding the mechanisms involved in lysosomes turn-over, it is quite clear that a slow accumulation of damaged proteins and non-degradable material might result in the various manifestations of lysosomal
64
Bi
Figure 2 Potential pathways of neurodegeneration induced by lysosomal dysfunction. Slowly progressive lysosomal dysfunction caused by genetic or environmental factors directly induces neurodegeneration (Neuron) or indirectly through activation of glia (Glia). Lines ending with arrows denote stimulatory effects, while lines ending in closed circles denote inhibitory effects.
dysfunctions we discussed above. Our hope is that this line of investigation will lead to the development of new therapeutics directed at preventing or curing age-related lysosomal dysfunction. ACKNOWLEDGMENT The author would like to thank Drs. Michel Baudry and Gary Lynch for their critical discussions regarding the manuscript and Dr. Jihua Liu and Ms. Yueqin Yao for their excellent work. Research in the author’s laboratory is supported by grants from UC BioSTAR, Alzheimer’s Association, NINDS, and Thuris Corp. REFERENCES 1. 2.
Mattson MP, Liu D. Energetics and oxidative stress in synaptic plasticity and neurodegenerative disorders. Neuromolecular Med 2002; 2(2):215–231. Floyd RA, Hensley K. Oxidative stress in brain aging. Implications for therapeutics of neurodegenerative diseases. Neurobiol Aging 2002; 23(5):795–807.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration 3.
4.
5. 6. 7. 8.
9. 10.
11.
12.
13. 14. 15. 16. 17. 18.
19.
20. 21.
65
Brunk UT, Terman A. The mitochondrial-lysosomal axis theory of aging: accumulation of damaged mitochondria as a result of imperfect autophagocytosis. Eur J Biochem 2002; 269(8):1996–2002. Liu J, Atamna H, Kuratsune H, Ames BN. Delaying brain mitochondrial decay and aging with mitochondrial antioxidants and metabolites. Ann N Y Acad Sci 2002; 959:133–166. Pollack M, Leeuwenburgh C. Apoptosis and aging: role of the mitochondria. J Gerontol A Biol Sci Med Sci 2001; 56(11):B475–B482. Samson FE, Nelson SR. The aging brain, metals and oxygen free radicals. Cell Mol Biol (Noisy-le-grand) 2000; 46(4):699–707. Sastre J, Pallardo FV, Garcia de la Asuncion J, Vina J. Mitochondria, oxidative stress and aging. Free Radic Res 2000; 32(3):189–198. Nixon RA, Cataldo AM, Mathews PM. The endosomal-lysosomal system of neurons in Alzheimer’s disease pathogenesis: a review. Neurochem Res 2000; 25(9–10):1161–1172. Wyss-Coray T, Mucke L. Inflammation in neurodegenerative disease—a double-edged sword. Neuron 2002; 35(3):419–432. Gemma C, Mesches MH, Sepesi B, Choo K, Holmes DB, Bickford PC. Diets enriched in foods with high antioxidant activity reverse age-induced decreases in cerebellar beta-adrenergic function and increases in proinflammatory cytokines. J Neurosci 2002; 22(14):6114–6120. Wilson CJ, Finch CE, Cohen HJ. Cytokines and cognition—the case for a head-to-toe inflammatory paradigm. J Am Geriatr Soc 2002; 50(12): 2041–2056. Landfield PW. Increased hippocampal Ca2þ channel activity in brain aging and dementia Hormonal and pharmacologic modulation. Ann NY Acad Sci 1994; 747:351–364. Foster TC, Kumar A. Calcium dysregulation in the aging brain. Neuroscientist 2002; 8(4):297–301. Brunk UT, Terman A. Lipofuscin: mechanisms of age-related accumulation and influence on cell function. Free Radic Biol Med 2002; 33(5):611–619. Porta EA. Pigments in aging: an overview. Ann NY Acad Sci 2002; 959:57–65. Terman A, Brunk UT. Lipofuscin: mechanisms of formation and increase with age. Apmis 1998; 106(2):265–276. Braak H. Architectonics as seen by lipofuscin stains. In: Peters A, Jones EG, eds. Cerebral Cortex. New York: Plenum Press, 1984:59–104. Nakanishi H, Tominaga K, Amano T, Hirotsu I, Inoue T, Yamamoto K. Age-related changes in activities and localizations of cathepsin D, E, B, and L in the rat brain tissues. Exp Neurol 1994; 126:119–128. Nakamura Y, Takeda M, Suzuki H, Morita H, Tada K, Hariguchi S, Nishimura T. Lysosome instability in aged rat brain. Neurosci Lett 1989; 97 (1–2):215–220. Kenessey A, Banay-Schwartz M, DeGuzman T, Lajtha A. Increase in cathepsin D activity in rat brain in aging. J Neurosci Res 1989; 23(4):454–456. Nakamura Y, Takeda M, Suzuki H, Morita H, Tada K, Hariguchi S, Nishimura T. Age-dependent change in activities of lysosomal enzymes in rat brain. Mech Ageing Dev 1989; 50(3):215–225.
66
Bi
22. Nakanishi H, Amano T, Sastradipura DF, Yoshimine Y, Tsukuba T, Tanabe K, Hirotsu I, Ohono T, Yamamoto K. Increased expression of cathepsins E and D in neurons of the aged rat brain and their colocalization with lipofuscin and carboxy-terminal fragments of Alzheimer amyloid precursor protein. J Neurochem 1997; 68(2):739–749. 23. Troncoso JC, Cataldo AM, Nixon RA, Barnett JL, Lee MK, Checler F, Fowler DR, Smialek JE, Crain B, Martin LJ, Kawas CH. Neuropathology of preclinical and clinical late-onset Alzheimer’s disease. Ann Neurol 1998; 43(5):673–676. 24. Cataldo AM, Paskevich PA, Kominami E, Nixon RA. Lysosomal hydrolases of different classes are abnormally distributed in brains of patients with Alzheimer disease. Proc Natl Acad Sci USA 1991; 88(24):10998–11002. 25. Callahan LM, Vaules WA, Coleman PD. Quantitative decrease in synaptophysin message expression and increase in cathepsin D message expression in Alzheimer disease neurons containing neurofibrillary tangles. J Neuropathol Exp Neurol 1999; 58(3):275–287. 26. Ginsberg SD, Hemby SE, Lee VM, Eberwine JH, Trojanowski JQ. Expression profile of transcripts in Alzheimer’s disease tangle-bearing CA1 neurons. Ann Neurol 2000; 48(1):77–87. 27. Schwagerl AL, Mohan PS, Cataldo AM, Vonsattel JP, Kowall NW, Nixon RA. Elevated levels of the endosomal–lysosomal proteinase cathepsin D in cerebrospinal fluid in Alzheimer disease. J Neurochem 1995; 64(1):443–446. 28. Arendt T, Bruckner MK, Gertz HJ, Marcova L. Cortical distribution of neurofibrillary tangles in Alzheimer’s disease matches the pattern of neurons that retain their capacity of plastic remodelling in the adult brain. Neuroscience 1998; 83(4):991–1002. 29. Arnold SE, Hyman BT, Flory J, Damasio AR, Van Hoesen GW. The topographical and neuroanatomical distribution of neurofibrillary tangles and neuritic plaques in the cerebral cortex of patients with Alzheimer’s disease. Cereb Cortex 1991; 1(1):103–116. 30. Ball MJ. Neuronal loss, neurofibrillary tangles and granulovacuolar degeneration in the hippocampus with ageing and dementia. A quantitative study. Acta Neuropathol (Berl) 1977; 37(2):111–118. 31. Brun A, Gustafson L. Distribution of cerebral degeneration in Alzheimer’s disease. A clinico-pathological study. Arch Psychiatr Nervenkr 1976; 223(1): 15–33. 32. Casanova MF, Carosella NW, Gold JM, Kleinman JE, Weinberger DR, Powers RE. A topographical study of senile plaques and neurofibrillary tangles in the hippocampi of patients with Alzheimer’s disease and cognitively impaired patients with schizophrenia. Psychiatry Res 1993; 49(1):41–62. 33. Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol (Berl) 1991; 82(4):239–259. 34. Braak H, Braak E. On areas of transition between entorhinal allocortex and temporal isocortex in the human brain. Normal morphology and laminaspecific pathology in Alzheimer’s disease. Acta Neuropathol (Berl) 1985; 68(4):325–332.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
67
35. Braak H, Del Tredici K, Schultz C, Braak E. Vulnerability of select neuronal types to Alzheimer’s disease. Ann NY Acad Sci 2000; 924:53–61. 36. Bobinski M, Wegiel J, Wisniewski HW, Tarnawski M, Reisberg B, De Leon MJ, Miller DC. Neurofibrillary pathology—correlation with hippocampal formation atrophy in Alzheimer disease. Neurobiol Aging 1996; 17(6):909–919. 37. Fukutani Y, Cairns NJ, Shiozawa M, Sasaki K, Sudo S, Isaki K, Lantos PL. Neuronal loss and neurofibrillary degeneration in the hippocampal cortex in late-onset sporadic Alzheimer’s disease. Psychiatry Clin Neurosci 2000; 54(5): 523–529. 38. Corder EH, Woodbury MA, Volkmann I, Madsen DK, Bogdanovic N, Winblad B. Density profiles of Alzheimer disease regional brain pathology for the huddinge brain bank: pattern recognition emulates and expands upon Braak staging. Exp Gerontol 2000; 35(6–7):851–864. 39. Coleman PD, Flood DG. Neuron numbers and dendritic extent in normal aging and Alzheimer’s disease. Neurobiol Aging 1987; 8(6):521–545. 40. Flood DG, Buell SJ, Horwitz GJ, Coleman PD. Dendritic extent in human dentate gyrus granule cells in normal aging and senile dementia. Brain Res 1987; 402(2):205–216. 41. Flood DG, Guarnaccia M, Coleman PD. Dendritic extent in human CA2-3 hippocampal pyramidal neurons in normal aging and senile dementia. Brain Res 1987; 409(1):88–96. 42. Anderton BH, Callahan L, Coleman P, Davies P, Flood D, Jicha GA, Ohm T, Weaver C. Dendritic changes in Alzheimer’s disease and factors that may underlie these changes. Prog Neurobiol 1998; 55(6):595–609. 43. Bi X, Yong AP, Zhou J, Gall CM, Lynch G. Regionally selective changes in brain lysosomes occur in the transition from young adulthood to middle age in rats. Neuroscience 2000; 97(2):395–404. 44. Bi X, Head E, Cotman CW, Lynch G. Spatial patterns of mammalian brain aging: distribution of cathepsin D-immunoreactive cell bodies and dystrophic dendrites in aging dogs resembles that in Alzheimer’s disease. J Comp Neurol 2003; 464(3):371–381. 45. Braak H. Architectonics of the Human Telencephalic Cortex. Berlin: SpringerVerlag, 1980. 46. Uylings HB, de Brabander JM. Neuronal changes in normal human aging and Alzheimer’s disease. Brain Cogn 2002; 49(3):268–276. 47. Ivy GO, Schottler F, Wenzel J, Baudry M, Lynch G. Inhibitors of lysosomal enzymes: accumulation of lipofuscin-like dense bodies in the brain. Science 1984; 226(4677):985–987. 48. Ivy GO, Kitani K, Ihara Y. Anomalous accumulation of tau and ubiquitin immunoreactivities in rat brain caused by protease inhibition and by normal aging: a clue to PHF pathogenesis? Brain Res 1989; 498(2):360–365. 49. Bahr BA, Abai B, Gall CM, Vanderklish PW, Hoffman KB, Lynch G. Induction of beta-amyloid-containing polypeptides in hippocampus: evidence for a concomitant loss of synaptic proteins and interactions with an excitotoxin. Exp Neurol 1994; 129(1):81–94.
68
Bi
50. Bednarski E, Ribak CE, Lynch G. Suppression of cathepsins B and L causes a proliferation of lysosomes and the formation of meganeurites in hippocampus. J Neurosci 1997; 17(11):4006–4021. 51. Bi X, Zhou J, Lynch G. Lysosomal protease inhibitors induce meganeurites and tangle-like structures in entorhinohippocampal regions vulnerable to Alzheimer’s disease. Exp Neurol 1999; 158(2):312–327. 52. Yong AP, Bednarski E, Gall CM, Lynch G, Ribak CE. Lysosomal dysfunction results in lamina-specific meganeurite formation but not apoptosis in frontal cortex. Exp Neurol, 1999; 157(1):150–160. 53. Bi X, Haque TS, Zhou J, Skillman AG, Lin B, Lee CE, Kuntz ID, Ellman JA, Lynch G. Novel cathepsin D inhibitors block the formation of hyperphosphorylated tau fragments in hippocampus. J Neurochem 2000; 74(4):1469– 1477. 54. Bednarski E, Lynch G. Cytosolic proteolysis of tau by cathepsin D in hippocampus following suppression of cathepsins B and L. J Neurochem 1996; 67(5):1846–1855. 55. Hyman BT, Gomez-Isla T. Alzheimer’s disease is a laminar, regional, and neural system specific disease, not a global brain disease. Neurobiol Aging 1994; 15(3):353–354; discussion 379–380. 56. Morrison BM, Hof PR, Morrison JH. Determinants of neuronal vulnerability in neurodegenerative diseases. Ann Neurol 1998; 44(3 suppl 1):S32–S44. 57. Bendiske J, Caba E, Brown QB, Bahr BA. Intracellular deposition, microtubule destabilization, and transport failure: an ‘‘early’’ pathogenic cascade leading to synaptic decline. J Neuropathol Exp Neurol 2002; 61(7):640–650. 58. Bi X, Pinkstaff J, Nguyen K, Gall CM, Lynch G. Experimentally induced lysosomal dysfunction disrupts processing of hypothalamic releasing factors. J Comp Neurol 1998; 401(3):382–394. 59. Barker PA, Hussain NK, McPherson PS. Retrograde signaling by the neurotrophins follows a well-worn trk. Trends Neurosci 2002; 25(8):379–381. 60. Howe CL, Valletta JS, Rusnak AS, Mobley WC. NGF signaling from clathrin-coated vesicles: evidence that signaling endosomes serve as a platform for the Ras-MAPK pathway. Neuron 2001; 32(5):801–814. 61. Miller WE, Lefkowitz RJ. Expanding roles for beta-arrestins as scaffolds and adapters in GPCR signaling and trafficking. Curr Opin Cell Biol 2001; 13(2):139–145. 62. Di Fiore PP, De Camilli P. Endocytosis and signaling. An inseparable partnership. Cell 2001; 106(1):1–4. 63. Watson FL, Heerssen HM, Bhattacharyya A, Klesse L, Lin MZ, Segal RA. Neurotrophins use the Erk5 pathway to mediate a retrograde survival response. Nat Neurosci 2001; 4(10):981–988. 64. Watson FL, Heerssen HM, Moheban DB, Lin MZ, Sauvageot CM, Bhattacharyya A, Pomeroy SL, Segal RA. Rapid nuclear responses to target-derived neurotrophins require retrograde transport of ligand–receptor complex. J Neurosci 1999; 19(18):7889–7900. 65. Yano H, Lee FS, Kong H, Chuang J, Arevalo J, Perez P, Sung C, Chao MV. Association of Trk neurotrophin receptors with components of the cytoplasmic dynein motor. J Neurosci 2001; 21(3):RC125.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
69
66. Wang Y, Pennock S, Chen X, Wang Z. Endosomal signaling of epidermal growth factor receptor stimulates signal transduction pathways leading to cell survival. Mol Cell Biol 2002; 22(20):7279–7290. 67. Wang Y, Pennock SD, Chen X, Kazlauskas A, Wang Z. Platelet-derived growth factor receptor-mediated signal transduction from endosomes. J Biol Chem 2004; 279(9):8038–8046. 68. Pennock S, Wang Z. Stimulation of cell proliferation by endosomal epidermal growth factor receptor as revealed through two distinct phases of signaling. Mol Cell Biol 2003; 23(16):5803–5815. 69. Balbis A, Baquiran G, Dumas V, Posner BI. Effect of inhibiting vacuolar acidification on insulin signaling in hepatocytes. J Biol Chem 2004; 279(38): 39348–39357. 70. Henderson LP, Lin L, Prasad A, Paul CA, Chang TY, Maue RA. Embryonic striatal neurons from Niemann–Pick type C mice exhibit defects in cholesterol metabolism and neurotrophin responsiveness. J Biol Chem 2000; 275(26): 20179–20187. 71. Hansen LA, Armstrong DM, Terry RD. An immunohistochemical quantification of fibrous astrocytes in the aging human cerebral cortex. Neurobiol Aging 1987; 8(1):1–6. 72. Kohama SG, Goss JR, McNeill TH, Finch CE. Glial fibrillary acidic protein mRNA increases at proestrus in the arcuate nucleus of mice. Neurosci Lett 1995; 183(3):164–166. 73. Nichols NR, Day JR, Laping NJ, Johnson SA, Finch CE. GFAP mRNA increases with age in rat and human brain. Neurobiol Aging 1993; 14(5): 421–429. 74. O’Callaghan JP, Miller DB. The concentration of glial fibrillary acidic protein increases with age in the mouse and rat brain. Neurobiol Aging 1991; 12(2):171–174. 75. Kohama SG, Goss JR, Finch CE, McNeill TH. Increases of glial fibrillary acidic protein in the aging female mouse brain. Neurobiol Aging 1995; 16(1):59–67. 76. Laping NJ, Teter B, Nichols NR, Rozovsky I, Finch CE. Glial fibrillary acidic protein: regulation by hormones, cytokines, and growth factors. Brain Pathol 1994; 4(3):259–275. 77. Landfield PW, Rose G, Sandles L, Wohlstadter TC, Lynch G. Patterns of astroglial hypertrophy and neuronal degeneration in the hippocampus of ages, memory-deficient rats. J Gerontol 1977; 32(1):3–12. 78. Lindsey JD, Landfield PW, Lynch G. Early onset and topographical distribution of hypertrophied astrocytes in hippocampus of aging rats: a quantitative study. J Gerontol 1979; 34(5):661–671. 79. Geinisman Y, Bondareff W, Dodge JT. Hypertrophy of astroglial processes in the dentate gyrus of the senescent rat. Am J Anat 1978; 153(4):537–543. 80. Dalton MM, Hommes OR, Leblond CP. Correlation of glial proliferation with age in the mouse brain. J Comp Neurol 1968; 134(4):397–400. 81. Bjorklund H, Eriksdotter-Nilsson M, Dahl D, Rose G, Hoffer B, Olson L. Image analysis of GFA-positive astrocytes from adolescence to senescence. Exp Brain Res 1985; 58(1):163–170.
70
Bi
82. Mouton PR, Long JM, Lei DL, Howard V, Jucker M, Calhoun ME, Ingram DK. Age and gender effects on microglia and astrocyte numbers in brains of mice. Brain Res 2002; 956(1):30–35. 83. Rozovsky I, Finch CE, Morgan TE. Age-related activation of microglia and astrocytes: in vitro studies show persistent phenotypes of aging, increased proliferation, and resistance to down-regulation. Neurobiol Aging 1998; 19(1): 97–103. 84. Gupta A, Pansari K. Inflammation and Alzheimer’s disease. Int J Clin Pract 2003; 57(1):36–39. 85. Liu B, Hong JS. Role of microglia in inflammation-mediated neurodegenerative diseases: mechanisms and strategies for therapeutic intervention. J Pharmacol Exp Ther 2003; 304(1):1–7. 86. Eikelenboom P, Bate C, Van Gool WA, Hoozemans JJ, Rozemuller JM Veerhuis R, Williams A. Neuroinflammation in Alzheimer’s disease and prion disease. Glia 2002; 40(2):232–239. 87. McGeer PL, McGeer EG. Mechanisms of cell death in Alzheimer disease– immunopathology. J Neural Transm Suppl 1998; 54:159–166. 88. Bamberger ME, Landreth GE. Inflammation, apoptosis, and Alzheimer’s disease. Neuroscientist 2002; 8(3):276–283. 89. Hoozemans JJ, Veerhuis R, Rozemuller AJ, Eikelenboom P. The pathological cascade of Alzheimer’s disease: the role of inflammation and its therapeutic implications. Drugs Today (Barc) 2002; 38(6):429–443. 90. Yoshiyama Y, Arai K, Oki T, Hattori T. Expression of invariant chain and pro-cathepsin L in Alzheimer’s brain. Neurosci Lett 2000; 290(2):125–128. 91. Bernstein HG, Wiederanders B. An immunohistochemical study of cathepsin E in Alzheimer-type dementia brains. Brain Res 1994; 667(2):287–290. 92. Paresce DM, Chung H, Maxfield FR. Slow degradation of aggregates of the Alzheimer’s disease amyloid beta-protein by microglial cells. J Biol Chem 1997; 272(46):29390–29397. 93. Paresce DM, Ghosh RN, Maxfield FR. Microglial cells internalize aggregates of the Alzheimer’s disease amyloid beta-protein via a scavenger receptor. Neuron 1996; 17(3):553–565. 94. Schenk D, Barbour R, Dunn W, Gordon G, Grajeda H, Guido T, Hu K, Huang J, Johnson-Wood K, Khan K, Kholodenko D, Lee M, Liao Z, Lieberburg I, Motter R, Mutter L, Soriano F, Shopp G, Vasquez N, Vandevert C, Walker S, Wogulis M, Yednock T, Games D, Seubert P. Immunization with amyloid-beta attenuates Alzheimer-disease-like pathology in the PDAPP mouse. Nature 1999; 400:173–177. 95. Ii M, Sunamoto M, Ohnishi K, Ichimori Y. Beta-Amyloid protein-dependent nitric oxide production from microglial cells neurotoxicity. Brain Res 1996; 720(1–2):93–100. 96. Harrigan MR, Kunkel DD, Nguyen LB, Malouf AT. Beta amyloid is neurotoxic in hippocampal slice cultures. Neurobiol Aging 1995; 16(5):779–789. 97. Meda L, Cassatella MA, Szendrei GI, Otvos L Jr, Baron P, Villalba M Ferrari D, Rossi F. Activation of microglial cells by beta-amyloid protein and interferon-gamma. Nature 1995; 374(6523):647–650.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration
71
98. Bi X, Gall CM, Zhou J, Lynch G. Uptake and pathogenic effects of amyloid beta peptide 1–42 are enhanced by integrin antagonists and blocked by NMDA receptor antagonists. Neuroscience 2002; 112(4):827–840. 99. Czlonkowska A, Kurkowska-Jastrzebska I, Czlonkowski A, Peter D, Stefano GB. Immune processes in the pathogenesis of Parkinson’s disease — a potential role for microglia and nitric oxide. Med Sci Monit 2002; 8(8):RA165–177. 100. McGeer PL, Yasojima K, McGeer EG. Inflammation in Parkinson’s disease. Adv Neurol 2001; 86:83–89. 101. Bjartmar C, Wujek JR, Trapp BD. Axonal loss in the pathology of MS: consequences for understanding the progressive phase of the disease. J Neurol Sci 2003; 206(2):165–171. 102. Steinman L, Martin R, Bernard C, Conlon P, Oksenberg JR. Multiple sclerosis: deeper understanding of its pathogenesis reveals new targets for therapy. Annu Rev Neurosci 2002; 25:491–505. 103. Rieckmann P, Maurer M. Anti-inflammatory strategies to prevent axonal injury in multiple sclerosis. Curr Opin Neurol 2002; 15(3):361–370. 104. Weydt P, Weiss MD, Moller T, Carter GT. Neuro-inflammation as a therapeutic target in amyotrophic lateral sclerosis. Curr Opin Investig Drugs 2002; 3(12):1720–1724. 105. McGeer PL, McGeer EG. Inflammatory processes in amyotrophic lateral sclerosis. Muscle Nerve 2002; 26(4):459–470. 106. Ghezzi P, Mennini T. Tumor necrosis factor and motoneuronal degeneration: an open problem. Neuroimmunomodulation 2001; 9(4):178–182. 107. Perry VH, Cunningham C, Boche D. Atypical inflammation in the central nervous system in prion disease. Curr Opin Neurol 2002; 15(3):349–354. 108. Parker JC Jr, Snyder JW. Prion infections in Creutzfeldt–Jakob disease and its variants. Ann Clin Lab Sci 1999; 29(2):112–117. 109. Pasquier RA, Koralnik IJ. Inflammatory reaction in progressive multifocal leukoencephalopathy: harmful or beneficial? J Neurovirol 2003; 9(suppl 1): 25–31. 110. Minagar A, Shapshak P, Fujimura R, Ownby R, Heyes M, Eisdorfer C. The role of macrophage=microglia and astrocytes in the pathogenesis of three neurologic disorders: HIV-associated dementia, Alzheimer disease, and multiple sclerosis. J Neurol Sci 2002; 202(1–2):13–23. 111. Epstein LG, Gelbard HA. HIV-1-induced neuronal injury in the developing brain. J Leukoc Biol 1999; 65(4):453–457. 112. Danton GH, Dietrich WD. Inflammatory mechanisms after ischemia and stroke. J Neuropathol Exp Neurol 2003; 62(2):127–136. 113. Emsley HC, Tyrrell PJ. Inflammation and infection in clinical stroke. J Cereb Blood Flow Metab 2002; 22(12):1399–1419. 114. Clark WM, Lutsep HL. Potential of anticytokine therapies in central nervous system ischaemia. Expert Opin Biol Ther 2001; 1(2):227–237. 115. Vanier MT, Rodriguez-Lafrasse C, Rousson R, Gazzah N, Juge MC, Pentchev PG, Revol A, Louisot P. Type C Niemann–Pick disease: spectrum of phenotypic variation in disruption of intracellular LDL-derived cholesterol processing. Biochim Biophys Acta 1991; 1096(4):328–337.
72
Bi
116. Vanier MT, Suzuki K. Recent advances in elucidating Niemann–Pick C disease. Brain Pathol 1998; 8(1):163–174. 117. Carstea ED, Morris JA, Coleman KG, Loftus SK, Zhang D, Cummings C, Gu J, Rosenfeld MA, Pavan WJ, Krizman DB, Nagle J, Polymeropoulos MH, Sturley SL, Ioannou YA, Higgins ME, Comly M, Cooney A, Brown A, Kaneski CR, Blanchette-Mackie EJ, Dwyer NK, Neufeld EB, Chang TY, Liscum L, Tagle DA, et al. Niemann–Pick C1 disease gene: homology to mediators of cholesterol homeostasis. Science 1997; 277(5323):228–231. 118. Blom TS, Linder MD, Snow K, Pihko H, Hess MW, Jokitalo E, Veckman V, Syvanen AC, Ikonen E. Defective endocytic trafficking of NPC1 and NPC2 underlying infantile Niemann–Pick type C disease. Hum Mol Genet 2003; 12(3):257–272. 119. Garver WS, Heidenreich RA. The Niemann–Pick C proteins and trafficking of cholesterol through the late endosomal=lysosomal system. Curr Mol Med 2002; 2(5):485–505. 120. Ory DS. Niemann–Pick type C: a disorder of cellular cholesterol trafficking. Biochim Biophys Acta 2000; 1529(1–3):331–339. 121. Burns M, Duff K. Cholesterol in Alzheimer’s disease and tauopathy. Ann N Y Acad Sci 2002; 977:367–375. 122. Liscum L. Niemann–Pick type C mutations cause lipid traffic jam. Traffic 2000; 1(3):218–225. 123. Morris JA, Carstea ED. Niemann–Pick C disease: cholesterol handling gone awry. Mol Med Today 1998; 4(12):525–531. 124. Reid PC, Sakashita N, Sugii S, Ohno-Iwashita Y, Shimada Y, Hickey WF, Chang TY. A novel cholesterol stain reveals early neuronal cholesterol accumulation in the Niemann–Pick type C1 mouse brain. J Lipid Res 2004; 45(3):582–591. 125. Auer IA, Schmidt ML, Lee VM, Curry B, Suzuki K, Shin RW, Pentchev PG, Carstea ED, Trojanowski JQ. Paired helical filament tau (PHFtau) in Niemann–Pick type C disease is similar to PHFtau in Alzheimer’s disease. Acta Neuropathol (Berl) 1995; 90(6):547–551. 126. Suzuki K, Parker CC, Pentchev PG, Katz D, Ghetti B, D’Agostino AN, Carstea ED. Neurofibrillary tangles in Niemann–Pick disease type C. Acta Neuropathol (Berl) 1995; 89(3):227–238. 127. Suzuki H, Sakiyama T, Harada N, Abe M, Tadokoro M. Pathologic changes of glial cells in murine model of Niemann–Pick disease type C: immunohistochemical, lectin-histochemical and ultrastructural observations. Pediatr Int 2003; 45(1):1–4. 128. Love S, Bridges LR, Case CP. Neurofibrillary tangles in Niemann–Pick disease type C. Brain ; 1995; 118(Pt 1):119–129. 129. Patterson MC, PentchevC. PG. Niemann–Pick; type C. Neurology 1996; 46(6):1785–1786. 130. Grundke-Iqbal I, Iqbal K, Quinlan M, Tung YC, Zaidi MS, Wisniewski HM. Microtubule-associated protein tau. A component of Alzheimer paired helical filaments. J Biol Chem 1986; 261(13):6084–6089. 131. Strittmatter WJ. Apolipoprotein E and Alzheimer’s disease. Ann NY Acad Sci 2000; 924:91–92.
Lysosomal Dysfunction in Brain Aging and Neurodegeneration 132. 133.
134. 135.
136.
137.
138.
139.
140.
141.
142.
143.
144.
145.
146.
73
Michikawa M. Cholesterol paradox: is high total or low HDL cholesterol level a risk for Alzheimer’s disease? J Neurosci Res 2003; 72(2):141–146. Prasad KN, Cole WC, Prasad KC. Risk factors for Alzheimer’s disease: role of multiple antioxidants, non-steroidal anti-inflammatory and cholinergic agents alone or in combination in prevention and treatment. J Am Coll Nutr 2002; 21(6):506–522. Wolozin B. Cholesterol and Alzheimer’s disease. Biochem Soc Trans 2002; 30(4):525–529. Distl R, Meske V, Ohm TG. Tangle-bearing neurons contain more free cholesterol than adjacent tangle-free neurons. Acta Neuropathol (Berl) 2001; 101(6):547–554. Distl R, Treiber-Held S, Albert F, Meske V, Harzer K, Ohm TG. Cholesterol storage and tau pathology in Niemann–Pick type C disease in the brain. J Pathol 2003; 200(1):104–111. Sawamura N, Gong JS, Chang TY, Yanagisawa K, Michikawa M. Promotion of tau phosphorylation by MAP kinase Erk1=2 is accompanied by reduced cholesterol level in detergent-insoluble membrane fraction in Niemann–Pick C1-deficient cells. J Neurochem 2003; 84(5):1086–1096. Fan QW, Yu W, Senda T, Yanagisawa K, Michikawa M. Cholesterol-dependent modulation of tau phosphorylation in cultured neurons. J Neurochem 2001; 76(2):391–400. Arriagada PV, Growdon JH, Hedley-Whyte ET, Hyman BT. Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer’s disease. Neurology 1992; 42(3 Pt 1):631–639. Runz H, Rietdorf J, Tomic I, de Bernard M, Beyreuther K, Pepperkok R, Hartmann T. Inhibition of intracellular cholesterol transport alters presenilin localization and amyloid precursor protein processing in neuronal cells. J Neurosci 2002; 22(5):1679–1689. Burns M, Gaynor K, Olm V, Mercken M, LaFrancois J, Wang L, Mathews PM, Noble W, Matsuoka Y, Duff K. Presenilin redistribution associated with aberrant cholesterol transport enhances beta-amyloid production in vivo. J Neurosci 2003; 23(13):5645–5649. German DC, Liang CL, Song T, Yazdani U, Xie C, Dietschy JM. Neurodegeneration in the Niemann–Pick C mouse: glial involvement. Neuroscience 2002; 109(3):437–450. German DC, Quintero EM, Liang C, Xie C, Dietschy JM. Degeneration of neurons and glia in the Niemann–Pick C mouse is unrelated to the low-density lipoprotein receptor. Neuroscience 2001; 105(4):999–1005. German DC, Quintero EM, Liang CL, Ng B, Punia S, Xie C, Dietschy JM. Selective neurodegeneration, without neurofibrillary tangles, in a mouse model of Niemann–Pick C disease. J Comp Neurol 2001; 433(3):415–425. Baudry M, Yao Y, Simmons D, Liu J, Bi X. Postnatal development of inflammation in a murine model of Niemann–Pick type C disease: immunohistochemical observations of microglia and astroglia. Exp Neurol 2003; 184(2): 887–903. Richardson JM, Woychik NA, Ebert DL, Dimond RL, Cardelli JA. Inhibition of early but not late proteolytic processing events leads to the missorting
74
147. 148. 149. 150.
151.
152.
153.
154.
155.
156.
Bi and oversecretion of precursor forms of lysosomal enzymes in Dictyostelium discoideum. J Cell Biol 1988; 107(6 Pt 1):2097–2107. Shaw E, Dean RT. The inhibition of macrophage protein turnover by a selective inhibitor of thiol proteinases. Biochem J 1980; 186(2):385–390. Gatti E, Pierre P. Understanding the cell biology of antigen presentation: the dendritic cell contribution. Curr Opin Cell Biol 2003; 15(4):468–473. Honey K, Rudensky AY. Lysosomal cysteine proteases regulate antigen presentation. Nat Rev Immunol 2003; 3(6):472–482. Braak H. Spindle-shaped appendages of IIIab-pyramids filled with lipofuscin: a striking pathological change of the senescent human isocortex. Acta Neuropathol (Berl) 1979; 46(3):197–202. Kenessey A, Nacharaju P, Ko LW, Yen SH. Degradation of tau by lysosomal enzyme cathepsin D: implication for Alzheimer neurofibrillary degeneration. J Neurochem 1997; 69:2026–2038. von Bergen M, Friedhoff P, Biernat J, Heberle J, Mandelkow EM, Mandelkow E. Assembly of tau protein into Alzheimer paired helical filaments depends on a local sequence motif ((306)VQIVYK(311)) forming beta structure. Proc Natl Acad Sci USA 2000; 97(10):5129–5134. Ugolini G, Cattaneo A, Novak M. Co-localization of truncated tau and DNA fragmentation in Alzheimer’s disease neurones. Neuroreport 1997; 8(17): 3709–3712. Garcia-Sierra F, Ghoshal N, Quinn B, Berry RW, Binder LI. Conformational changes and truncation of tau protein during tangle evolution in Alzheimer’s disease. J Alzheimer’s Dis 2003; 5(2):65–77. Marin-Teva, JL, Dusart I, Colin C, Gervais A, van Rooijen N, Mallat M. Microglia promote the death of developing Purkinje cells. Neuron 2004; 41(4):535–547. Ashwell K. Microglia and cell death in the developing mouse cerebellum. Brain Res Dev Brain Res 1990; 55(2):219–230.
5 Axonal Sprouting in the Adult Brain: Mechanisms of Lesion-Specific Sprouting of the Corticostriatal Pathway in Adult Rats Marie-Francoise Chesselet, S. Thomas Carmichael, Marc Shomer, Dorothy Harris, and Veronique Riban Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A.
1.
INTRODUCTION
Axonal sprouting is a form of neuronal plasticity defined as the ability of spared axons to grow new ‘‘sprouts’’ to reinnervate neurons deprived of their normal inputs by a lesion of their afferents. Therefore, it fundamentally differs from the regrowth of an axon that has been injured (axonal regeneration). Axonal sprouting has been described more than 25 years ago in the brain of adult mammals (1). However, until very recently it was thought to be restricted to regions such as the hippocampus that still express a number of ‘‘immature’’ proteins in adulthood. This view has changed dramatically in the last few years with the demonstration that axonal sprouting can occur in adults in a number of brain regions, including the cerebral cortex and the striatum (2–5). The mechanisms that trigger axonal sprouting, however, remain largely unknown. In this review, we will focus on findings related to a model of axonal sprouting of the corticostriatal pathway that presents unique opportunities to elucidate these mechanisms.
75
76
2.
Chesselet et al.
LESION-SPECIFIC AXONAL SPROUTING OF THE CORTICOSTRIATAL PATHWAY
The cerebral cortex provides a massive glutamatergic projection to the caudate-putamen, commonly known as the striatum in rats and mice (6). This projection is topographically organized, with the sensorimotor cortex projecting to the dorsolateral part of the striatum, which corresponds to the putamen in primates. In adults, this projection is largely ipsilateral, whereas a minor component, less than 10% of fibers at most, innervates the contralateral striatum by way of the corpus callosum (7). Earlier studies have shown that aspiration lesions of the cerebral cortex were followed by an enlargement of asymmetric synapses, presumably glutamatergic, in the dorsolateral striatum on the side of the lesion, without evidence of new axonal growth (8). Except for one study in which a minor increase in crossed-corticostriatal connections was observed in very young adult rats (9), the absence of significant axonal sprouting after cortical lesions in the adult has been confirmed by several groups (2,4,5,10). Based on this evidence, one could conclude that the striatum differs from the hippocampus, where robust axonal sprouting and reactive synaptogenesis are well documented. Studies in our laboratory in the midnineties revealed that, contrary to this belief, corticostriatal axons are capable of sprouting and can reinnervate the dorsolateral striatum after cortical lesions (2). This effect, however, was only observed when the cerebral cortex was destroyed by ischemia-inducing thermocoagulation of the pial blood vessels. Lesions of the cerebral cortex by thermocoagulation or devascularization were originally developed to provide a way to selectively destroy the cerebral cortex without running the risk of damaging the underlying corpus callosum or the striatum with the aspiration pipette (11). These lesions are made possible by the organization of the blood supply to the cerebral cortex, which is provided entirely by superficial blood vessels coursing in the pia mater. In contrast, blood flow to the corpus callosum and striatum is provided by the anterior cerebral artery and deep subcortical branches of the middle cerebral artery. As a result, removing or coagulating the pial blood vessels on the cortical surface creates an ischemic lesion restricted to the cerebral cortex, which degenerates in 3–5 days (12–14). Although devascularization by peeling off the pia has been successfully used, our laboratory has used thermocoagulatory lesions because they produce extremely reproducible cortical lesions both in terms of extent and location. We have shown that with this method pycnotic neurons can be seen in all layers of the sensorimotor cortex as early as 6-hr postsurgery (14). After about 5 days, all cortical layers have degenerated and the lesion is surrounded by reactive astrocytes (12,14). The corpus callosum, although thinned, is not interrupted or directly injured (4,5). Furthermore, activation of c-fos and increased MAP2 immunoreactivity were limited to the cerebral
Axonal Sprouting in the Adult Brain of Rats
77
cortex surrounding the lesion, indicating that the dorsolateral striatum does not exhibit the characteristics of a ‘‘penumbra’’ after these lesions (14). Evidence that axonal sprouting takes place in the corticostriatal pathway after thermocoagulatory lesions of the cerebral cortex comes from several lines of evidence. First, immunolabeling for GAP-43, an axonal protein present in corticostriatal terminals in adults (15), remains unchanged in the dorsolateral striatum after ischemic lesions whereas it decreases after aspiration lesions (14). Since both lesions similarly remove ipsilateral corticostriatal fibers that contain GAP-43 in the dorsolateral striatum, this result suggests a redistribution of the protein after the thermocoagulatory lesions. Indeed, ultrastructural studies have revealed the presence of growth cone-like structures rich in GAP-43 in dorsolateral striatum after thermocoagulatory cortical lesions (4). These structures are absent in controls or after cortical lesions by aspiration. Interestingly, a much smaller number of GAP-43 containing growth cones were found in the striatum just under the corpus callosum after aspiration lesions. These growth cone-like structures were much fewer, of a smaller size, and very restricted in location compared with those growth cones found in the dorsolateral striatum after thermocoagulatory lesions (4). Taken together, these data indicate that axonal growth occurs in the dorsolateral striatum of adult rats after ischemic lesions of the cerebral cortex. In contrast, axonal sprouting is very limited and fails to invade the denervated striatum after aspiration lesions. Quantitative analysis of asymmetric synapses after both lesions confirmed the difference between the two types of lesions and suggested that the sprouting axons may reinnervate striatal neurons. Indeed, aspiration lesions are followed by a significant decrease in the number of asymmetric synapses in the dorsolateral striatum 16-day postlesion whereas no significant decrease was observed after thermocoagulatory lesions (4). These experiments indicate that axonal sprouting can occur in dorsolateral striatum, even in adults, but do not reveal the origin of the sprouting axons. We tested the hypothesis that crossed-corticostriatal fibers were, at least in part, the origin of the sprouting axons (2,5). When injected in the sensorimotor cortex of adult rats, the anterograde tracers fluororuby or biotinylated dextran amine (BDA) predominantly labels the ipsilateral dorsolateral striatum, which receives the bulk of the corticostriatal fibers. The same pattern was found after aspiration lesions. In contrast, after thermocoagulatory lesions, a dense axonal network of fibers was labeled in the contralateral, i.e., denervated, dorsolateral striatum (Fig. 1). This indicates that axons from the crossed-corticostriatal pathway originating in the homotypic contralateral cortex sprout into the denervated striatum after ischemic but not aspiration cortical lesions. We have recently confirmed that a similar crossed-corticostriatal sprouting occurs after thermocoagulatory lesions of the sensorimotor cortex in adult mice (Fig. 2) (16).
78
Chesselet et al.
Figure 1 Axonal sprouting within the corticostriatal system. A–C. Coronal sections through rat frontal cortex of a control animal (B), an animal with an aspiration lesion (A) and an animal with an ischemic lesion (C). The dark region in the cortex of each section (arrowhead in A) is an injection of BDA. In control, axonal labeling can be seen in the ipsilateral striatum and contralateral cortex, with a small projection to the contralateral striatum. After ischemic lesions, there is a substantial increase in the axonal projection to the contralateral striatum and perilesion cortex (arrows in C). Bar in C ¼ 500 uM. D. Schematic representation of axonal sprouting after ischemic cortical lesions.
The maintenance of a normal number of asymmetric synapses in the dorsolateral striatum after thermocoagulatory but not aspiration lesions suggests that the sprouting axons reoccupy the synaptic sites denervated by the loss of ipsilateral corticostriatal neurons (4). It is unclear, however, whether this anatomical plasticity is able to fully restore the function of the normal corticostriatal connections. Indeed, we have found long-term changes in gene expression in the dorsolateral striatum after thermocoagulatory lesions, whereas similar changes are not observed after aspirations lesions (2,12,13). The functional consequences of these changes in gene expression, however, are unknown. A better index of the functional consequences of axonal sprouting would be evidence of behavioral recovery after thermocoagulatory but not aspiration lesions. Unfortunately, lesions limited to the sensorimotor cortex and sparing the striatum minimally affect behavior in rats, and the small deficits observed recover spontaneously after both types of lesion (17). More
Axonal Sprouting in the Adult Brain of Rats
79
Figure 2 Axonal sprouting after cortical lesion in mice. Quantification of corticostriatal projections in sham and TCL animals. Corticostriatal projections are revealed by BDA injection in the cortex. Surface area of projection is measured (NIH Image analysis 1.6) and the ratio of stained projection on the contralateral side per projections on the ipsilateral side is compared between the two groups. : p < 0.05 (Mann Whitney test, n ¼ 5)
detailed behavioral analyses will be necessary to determine the functional consequences of axonal sprouting in rats and mice (16). In summary, several lines of evidence indicate that, contrary to previous findings, the corticostriatal pathway is capable of axonal plasticity in the striatum. Failure to demonstrate this plasticity was related to the method used to perform the lesion. Only ischemic lesions secondary to thermocoagulation of pial blood vessels induce this plasticity. Thus, axonal sprouting in adult brain is not limited to the hippocampus, a finding that also applies to intracortical connections (3). Importantly, the data also indicate that different types of lesion differ in their ability to trigger axonal sprouting. Identifying the mechanisms underlying this differential effect may help uncover the mechanisms responsible for axonal sprouting in adult brain. The comparison between aspiration and thermocoagulatory lesions offers a unique model system to accomplish this goal. Indeed, our results suggest that mechanisms elicited specifically by the ischemic lesion may enable axonal sprouting. Alternatively, it is possible that mechanisms selectively triggered by aspiration prevent the sprouting from occurring. Recent work from our laboratory has provided evidence that both types of mechanisms contribute to lesion-specific sprouting of the corticostriatal pathway in adult brain.
80
3.
Chesselet et al.
THE ROLE OF ELECTROPHYSIOLOGICAL ACTIVITY IN AXONAL SPROUTING IN THE ADULT BRAIN
A similar region of the striatum loses its input form the sensorimotor cortex after both types of lesions. Although the time course of input degeneration may vary, it is unlikely that signals originating in the denervated targets of damaged axons, the striatal output neurons, differ in both conditions. Indeed, astrogliosis as well as the changes in the expression of growth factors (FGF2) or adhesion molecules (tenascin, CSPG, PSA-NCAM) were similar after the two types of lesions (14). Similarly, DNA microarray studies did not identify significant differences in changes of gene expression after the two lesions (Shomer and Chesselet, unpublished observations). Therefore, the mechanisms triggering the differential axonal sprouting after aspiration vs. thermocoagulatory lesions may originate in the cortex surrounding the lesion rather than in the dorsolateral striatum. Our tracing studies have shown that at least some of the sprouting axons originate in the cortex contralateral to the damaged cortex (2,5). This suggests that a signal originating from the lesion site triggers a sprouting response in neuronal cell bodies located in the other hemisphere. A possible mechanism would be a difference in cellular activity transmitted through corticocortical connections, or with relay in subcortical structures. To test this hypothesis, we recorded multiunit activity from the cortex surrounding the lesion and from the contralateral cortex in awake rats after sham surgery, aspiration, or thermocoagulatory lesions of the sensorimotor cortex. These experiments revealed a major difference in cellular activity in these regions after both types of lesions (Fig. 3) (5). Sham surgery or aspiration lesions did not significantly alter the pattern of cellular activity observed in cortex surrounding the lesion or in contralateral cortex compared to prelesion recordings. In contrast, thermocoagulatory lesions strongly disrupted this activity. These changes were first observed in cortex surrounding the lesion, as early as 1-day postsurgery. In this region, rats with a thermocoagulatory lesion exhibited slow wave activity in the 0.2–2 Hz range, with a peak activity at 1 Hz. This pattern was significantly different than the multiunit activity observed in rats with either sham surgery or an aspiration lesion of the same cortical region. It persisted until day 3 postsurgery. Lesion-specific slow wave activity was also observed in the contralateral cortex, where it reached its maximum 3 days postsurgery and persisted until day 5. Its frequency was slower than in cortex surrounding the lesion, with a maximum between 0.1 and 0.4 Hz. Bursts of action potentials could be recorded during the ascending phase of the slow waves. When activity was compared between distant cortical regions, it became apparent that it was significantly correlated with the slow wave activity over long distance in cortex both ipsi- and contralateral to the thermocoagulatory lesion.
Axonal Sprouting in the Adult Brain of Rats
81
Figure 3 Patterned spontaneous activity in association with axonal sprouting. The top figure is a schematic dorsal view of the rat brain, showing the location of the ischemic lesion and electrode placement. Cortical activity was recorded in vivo in the awake animal. The vertical line corresponds to the position of the coronal sections in Fig. 1. Normal cortical activity is low voltage and desynchronized (scale in this trace is the same as for perilesion cortex). There is no difference between this pattern of normal cortical activity in control animals, and the pattern of cortical activity after aspiration lesions. However, ischemic lesions induce two patterns of slow, synchronized neuronal activity. In the first pattern, ischemic lesions induce high voltage slow waves in perilesion cortex at 0.2–2 Hz, maximal 1-day postlesion. Higher resolution views of these waves (to the right in the figure) show these slow waves have a burst of multiunit neuronal discharges on the negative phase of the wave. In the second pattern, ischemic cortical lesions induce a slow wave in contralateral cortex that is maximal on day 3 postlesion. These are biphasic waves occurring with a frequency of 0.1–0.4 Hz, with wavelets of 6–8 Hz on the negative phase. Multiunit neuronal discharges are synchronized to the negative phase of the waves.
82
Chesselet et al.
The presence of these slow wave activities only after the type of lesions that induced axonal sprouting suggests that they may play a role in its induction, or at least enable the process. To further establish a role for lesion-specific changes in cellular activity in axonal sprouting, we have blocked cellular activity by infusing tetrodotoxin in the cortex surrounding the lesion and examined the effect of this treatment on the cellular activity and axonal sprouting induced by the thermocoagulatory lesions (5). Local infusion of tetrodotoxin in the cortex surrounding the lesion blocked local axon potentials but not action potentials in the contralateral cortex, indicating that the toxin did not diffuse to the other hemisphere. However, the lesion-induced slow wave activity was blocked in both hemispheres, indicating that it originates in the cortex surrounding the lesion. Vehicle infusion in rats with thermocoagulatory lesions did not affect either slow wave activity or lesion-induced sprouting. In contrast, no sprouting was observed when slow waves were blocked by the infusion of tetrodotoxin in the cortex surrounding the lesion. Together with the absence of slow wave activity after aspiration lesions (that do not induce axonal sprouting), this result strongly suggests that this lesion-induced cellular activity is critical for the corticostriatal axonal sprouting observed after ischemic cortical lesions. Two major questions remain unanswered: what triggers the slow wave activity in the cortex surrounding the lesion and by what mechanism this cellular activity induces or enables axonal sprouting in the corticostriatal neurons contralateral to the lesion. The widespread slow wave activity induced by thermocoagulatory lesions clearly involves cortical and subcortical sites (18). The fact that local TTX infusion into the cortex surrounding the lesion blocks the synchronous neuronal activity implicates this region in initiating this patterned activity. The cortex surrounding the lesion is the site of profound alterations in electrophysiological properties and neurotransmitter signaling systems that may provide the substrate for slow wave generation. Focal cortical ischemic lesions induce sustained increases in baseline neuronal firing frequency, prolonged EPSPs, diminished IPSPs and paired-pulse inhibition, and a facilitated induction of LTP in periinfarct cortex (reviewed in Ref. 19). These electrophysiological changes are associated with diminished GABAa receptor subunit levels and GABAb receptor-binding sites and an increase in NMDA receptor-binding sites (19,20). The net effect of enhanced excitability and diminished inhibition is important in the generation of synchronous neuronal activity in epileptogenesis. While there was no electrographic or behavioral evidence for seizure activity during the period of axonal sprouting after thermocoagulatory lesions, it is likely that similar alterations in neuronal network properties in cortex surrounding the lesion are important in generating the synchronous neuronal activity that triggers axonal sprouting.
Axonal Sprouting in the Adult Brain of Rats
83
The molecular mechanisms linking slow wave activity and axonal sprouting remain unknown. Although extensive research has focused on the role of patterns of electrophysiological activity during neuronal development, earlier work concluded that activity was critical for establishing specific synaptic connections but not for axonal growth. This view has been challenged, however, by more recent results. Synchronous slow wave activity and neuronal discharges are present throughout the neural axis during development, such as in retina, lateral geniculate, spinal cord, hippocampus, and cortex, during periods of active axonal sprouting and synaptogenesis in these structures (21). While quite slow in subcortical structures, the pattern and frequency of this synchronous neural activity in developing cortex during axonal sprouting and synaptogenesis are the same as that observed in adult cortex during axonal sprouting after thermocoagulatory lesions (22). Importantly, it has recently been shown that this frequency of neuronal activity in vitro allows growing axons to traverse inhibitory substrates, such as are present in myelin (23). Thus, the synchronous neuronal activity induced by thermocoagulatory lesions during axonal sprouting is similar to the patterned electrical activity in the cortex during axonal sprouting in the developing animal, and may allow growth cones to traverse the inhibitory microenvironments of the adult brain. Although thermocoagulatory lesions do not represent a perfect model of stroke since no reperfusion occurs, this model shares with stroke the main features of ischemia, breakdown of blood–brain barrier, and progressive neuronal degeneration. Recent evidence indicates that axonal sprouting also occurs in the cortex after ischemic stroke in adult rats. Poststroke axonal sprouting has been described in cortex adjacent to the infarct, and is substantial—producing a wholesale remapping of somatosensory cortex (3). Axonal sprouting also occurs into cortex adjacent to the thermocoagulatory lesions (5). It is not clear whether or not axonal sprouting similar to that seen in rats after these lesions occurs in humans. However, stroke in humans strongly induces the growth-associated protein GAP43 in cortex adjacent to the infarct (24). Also, GAP43 is tightly linked to growth cones and its induction has been a reliable marker for axonal sprouting (4,25,26). In this regard, the slow wave activity that induces axonal sprouting after thermocoagulatory lesions in rats is the same in morphology and frequency to episodes of slow activity noted in the electroencephalogram of stroke patients. This is the pattern of polymorphic delta activity (Fig. 4) (27). Limited functional recovery often occurs days or weeks after a stroke and it is possible that axonal sprouting and reactive synaptogenesis, if they take place in humans, play a role in this delayed recovery. If this is the case, then it will be important to avoid blocking cellular mechanisms that, according to our results, may be critical for stroke-induced axonal plasticity.
84
Chesselet et al.
Figure 4 Synchronous, slow neuronal activity after stroke in humans and ischemic lesions in rats. EEG recordings from patients after stroke frequently show rhythmic slow wave activity. The top of the figure shows a representative EEG trace from a patient after stroke. Slow wave activity, termed polymorphic delta activity, is present in the right hemisphere, near the stroke. An enlarged view of the region within the red box is shown in the middle trace. The polymorphic delta activity after stroke in humans has the same frequency, wave morphology, and topographic relationship to the infarct as that seen after ischemic lesions in rats. In rats, these waves are part of a highly synchronous pattern of extracellular field potentials and neuronal discharges that serves as a signal for axonal sprouting. The EEG trace was taken from Ref. 27.
Axonal Sprouting in the Adult Brain of Rats
4.
85
MYELIN-ASSOCIATED PROTEINS AND AXONAL SPROUTING IN THE ADULT BRAIN
Numerous studies in brain and spinal cord have concluded that inhibitory proteins associated with myelin sheets provide a major impediment to the growth of axons in the adult central nervous system. If this is the case, one may expect that the ischemic lesions interfere with these inhibitory mechanisms to permit the robust axonal growth we have observed in dorsolateral striatum after these lesions. Previous work by Kartje et al. (10) suggested that the myelin-associated inhibitory protein nogo may be of particular importance for preventing axonal sprouting after cortical lesions by aspiration. Indeed, this group was able to elicit sprouting of the crossed-corticostriatal pathway after aspiration of the cerebral cortex by implanting in the lesion cavity, cells secreting an antibody (IN1) that blocks nogo activity. Because constitutive expression of nogo is believed to block axonal growth, these data suggest that perhaps thermocoagulatory lesions depress nogo production. To test this hypothesis, we have measured nogo A mRNA by northern blot and in situ hybridization histochemistry after both types of lesions. Contrary to expectations, we found that thermocoagulatory lesions did not change nogo A mRNA expression on the side of the lesion. However, aspiration lesions induced a marked increase in nogo A mRNA 3 days after surgery. The increase observed by northern blot analysis was confirmed by in situ hybridization histochemistry, which showed a marked increase in mRNA in neurons of the entire ipsilateral cortex at 1- and 3- day postsurgery (28; Harris and Chesselet, unpublished observations). Experiments are in progress to determine whether this neuronal increase is accompanied by an increased expression in oligodendrocytes of the corpus callosum and an increase in protein expression. Together with Kartje’s results, these data suggest that an increased expression of nogo may contribute to the absence of sprouting of the corticostriatal pathway after aspiration lesions. A most interesting result is the absence of increase in nogo mRNA expression after thermocoagulatory lesions. Indeed, we have observed that even superficial abrasions of cortex are sufficient to induce nogo A mRNA expression. This suggests that mechanisms triggered by the ischemic lesion may block the induction of nogo A mRNA normally induced by even mild brain lesions. These inhibitory mechanisms remain unknown but their identification may lead to the development of therapeutic strategies to enhance axonal growth by controlling nogo A expression after brain injury.
5.
CONCLUSION
It is now well established that more anatomical plasticity than previously thought can occur after neuronal injury in adult brain. Interestingly, this
86
Chesselet et al.
anatomical plasticity is exquisitely lesion-dependent, suggesting that lesionspecific mechanisms play a critical role in eliciting and=or enabling axonal growth. The mechanisms regulating this plasticity remain poorly understood. However, they seem to involve lesion-specific changes in neuronal activity that are poised to transmit signals over long distances. These may elicit the molecular machinery necessary for axonal sprouting in distant neuronal populations connected to the injured brain region. Regulation of inhibitory myelin-associated protein is likely to also play a critical role. Unexpectedly, regulation rather than constitutive expression of these factors may be responsible for the limited axonal growth observed after most brain lesions. Although the existence of similar phenomena in humans remains speculative, the occurrence of changes in cellular activity after cortical stroke that are very similar to those associated with axonal sprouting in rats raises the possibility that similar mechanisms may contribute to functional recovery after stroke in humans. Neuroprotective therapies to improve stroke outcome have failed in clinic because they need to be administered within the first few hours after the ischemic episode. Therefore, new strategies for improving functional recovery are critical for the treatment of stroke. The model we have described offers unique opportunities, not only to elucidate the mechanisms of axonal sprouting after ischemic cortical lesions, but also to test new therapies for eliciting axonal sprouting when it normally does not occur, i.e., after cortical aspiration.
REFERENCES 1.
2.
3. 4.
5.
6. 7.
Steward O, Cotman CW, Lynch GS. Re-establishment of electrophysiologically functional entorhinal cortical input to the dentate gyrus deafferented by ipsilateral entorhinal lesions: innervation by the contralateral entorhinal cortex. Exp Brain Res 1973; 18:396–414. Napieralski JA, Butler AK, Chesselet MF. Anatomical and functional evidence for lesion-specific sprouting of corticostriatal input in the adult rat. J Comp Neurol 1996; 373:484–497. Carmichael ST, Wei L, Rovainen CM, Woolsey TA. New patterns of intracortical projections after focal cortical stroke. Neurobiol Dis 2001; 8:910–922. Uryu K, MacKenzie L, Chesselet MF. Ultrastructural evidence for differential axonal sprouting in the striatum after thermocoagulatory and aspiration lesions of the cerebral cortex in adult rats. Neuroscience 2001; 105:307–316. Carmichael ST, Chesselet MF. Synchronous neuronal activity is a signal for axonal sprouting after cortical lesions in the adult. J Neurosci 2002; 22: 6062–6070. Chesselet MF. Mapping the basal ganglia. In: Toga AW, Mazziotta JC, eds. Brain Mapping: The Systems. San Diego: Academic Press, 2000:177–206. Wilson CJ. Morphology and synaptic connections of crossed corticostriatal neurons in the rat. J Comp Neurol 1987; 263:567–580.
Axonal Sprouting in the Adult Brain of Rats 8. 9. 10.
11.
12.
13.
14.
15.
16.
17.
18. 19.
20.
21. 22. 23. 24.
87
Chen S, Hillman DE. Robust synaptic plasticity of striatal cells following partial deafferentation. Brain Res 1990; 520:103–114. Cheng HW, Tong J, McNeill TH. Lesion-induced axon sprouting in the deafferented striatum of adult rat. Neurosci Lett 1998; 242:69–72. Kartje GL, Schulz MK, Lopez-Yunez A, Schnell L, Schwab ME. Corticostriatal plasticity is restricted by myelin-associated neurite growth inhibitors in the adult rat. Ann Neurol 1999; 45:778–786. Errami M, Nieoullon A. Development of a micromethod to study the Naþindependent L-[3H]glutamic acid binding to rat striatal membranes. II. Effects of selective striatal lesions and deafferentations. Brain Res 1986; 366:178–186. Salin P, Chesselet MF. Paradoxical increase in striatal neuropeptide gene expression following ischemic lesions of the cerebral cortex. Proc Natl Acad Sci USA 1992; 89:9954–9958. Salin P, Chesselet MF. Expression of GAD (M(r) 67,000) and its messenger RNA in basal ganglia and cerebral cortex after ischemic cortical lesions in rats. Exp Neurol 1993; 119:291–301. Szele FG, Alexander C, Chesselet MF. Expression of molecules associated with neuronal plasticity in the striatum after aspiration and thermocoagulatory lesions of the cerebral cortex in adult rats. J Neurosci 1995; 15:4429–4448. DiFiglia M, Roberts RC, Benowitz LI. Immunoreactive GAP-43 in the neuropil of adult rat neostriatum: localization in unmyelinated fibers, axon terminals, and dendritic spines. J Comp Neurol 1990; 302:992–1001. Riban V, Fleming S, Chesselet MF. Thermocoagulatory lesion of the sensorimotor cortex induces corticostriatal sprouting in the adult mice. Neuroscience [Abst] 2003. Napieralski JA, Banks RJ, Chesselet MF. Motor and somatosensory deficits following uni- and bilateral lesions of the cortex induced by aspiration or thermocoagulation in the adult rat. Exp Neurol 1998; 154:80–88. Steriade M. Corticothalamic resonance, states of vigilance and mentation. Neuroscience 2000; 101:243–276. Witte OW, Bidmon HJ, Schiene K, Redecker C, Hagemann G. Functional differentiation of multiple perilesional zones after focal cerebral ischemia. J Cereb Blood Flow Metab 2000; 20:1149–1165. Redecker C, Wang W, Fritschy JM, Witte OW. Widespread and long-lasting alterations in GABA(A)-receptor subtypes after focal cortical infarcts in rats: mediation by NMDA-dependent processes. J Cereb Blood Flow Metab 2002; 22:1463–1475. Feller MB. Spontaneous correlated activity in developing neural circuits. Neuron 1999; 22:653–656. Weliky M, Katz LC. Correlational structure of spontaneous neuronal activity in the developing lateral geniculate nucleus in vivo. Science 1999; 285:599–604. Ming G, Henley J, Tessier-Lavigne M, Song H, Poo M. Electrical activity modulates growth cone guidance by diffusible factors. Neuron 2001; 29:441–452. Ng SC, de la Monte SM, Conboy GL, Karns LR, Fishman MC. Cloning of human GAP-43: growth association and ischemic resurgence. Neuron 1988; 1: 133–139.
88
Chesselet et al.
25. Benowitz LI, Goldberg DE, Madsen JR, Soni D, Irwin N. Inosine stimulates extensive axon collateral growth in the rat corticospinal tract after injury. Proc Natl Acad Sci USA 1999; 96:13486–13490. 26. Chen P, Goldberg DE, Kolb B, Lanser M, Benowitz LI. Inosine induces axonal rewiring and improves behavioral outcome after stroke. Proc Natl Acad Sci USA 2002; 99:9031–9036. 27. Daly D, Markand O. Focal brain lesions. In: Daly D, Pedley T, eds. Current Practice of Clinical Electroencephalography. 2nd ed. New York: Raven Press, 1990:335–370. 28. Shomer M, Chesselet MF. Lesion-specific upregulation of nogo-A mRNA in adult brain. Neuroscience [Abst]. 2001; 27:258.7.
6 Behavioral Implications for the Activity Dependent Regulation of Neurotrophins Fernando Gomez-Pinilla and Shoshanna Vaynman Department of Physiological Science, University of California at Los Angeles, Los Angeles, California, U.S.A.
1.
INTRODUCTION
The neurotrophin (NT) family of signaling molecules is a diffusible class of gene products whose role in the central nervous system (CNS) is quickly expanding under the helmsman of both scientific development and interest. Recent findings have extended their capacity for neuron–target interaction from their classical role in neuronal differentiation and survival (1), to that of CNS synapse formation, maturation, and stabilization. In fact, the extensity of NT research, such as the discovery that NT synthesis and release in the brain is regulated by neuronal activity (2,3) and that NTs themselves in turn acutely potentiate synaptic transmission (4,5), makes them perfectly suited for and possibly contributing to their own activity-dependent synthesis and release. Taken together, these findings have led to coin ‘‘the NT hypothesis of synaptic plasticity.’’ The activity-dependent regulation of NTs is inherent to their ability to modulate the transference of information across the synapse and may powerfully contribute to, if not provide for, a cellular correlate to information processing (such as learning and memory) in the mammalian brain. In particular, the NT brain-derived neurotrophic factor (BDNF) has received much experimental support for promoting formation of new synaptic 89
90
Gomez-Pinilla and Vaynman
contacts and increasing synaptic efficacy in both developing and adult nervous systems. 2. CELL BIOLOGY OF NTs 2.1. Neurotrophin Family The family of NTs is a set of homodimeric proteins, which consists of nerve growth factor (NGF), BDNF, neurotrophin-3 (NT-3), neurotrophin-4=5 (NT-4=5), and neurotrophin-6 (NT-6). The effects of NTs are mediated by their binding to tyrosine kinase receptors (Trks) and to the low-affinity p75 NT receptor. The different NTs bind to their cognate Trk receptors; NGF to TrkA, BDNF and NT-4=5 to TrkB, and NT-3 to Trk-C (6). Additionally the different NTs share a similar affinity for the low-affinity p75 NT receptor. Thus, the ability to act through their selective receptors, the nonselective p75 receptor, or in conjoint activation, capacitates the NTs to exert their effects. 2.2.
Expression of NTs and their Receptors in the CNS
The distribution of NTs in the brain can yield important information regarding their function. The expression of NTs and their receptors have mainly been studied at the mRNA level using in situ hybridization. BDNF, NT-3, and NGF have been found to be prominently expressed in neurons of the hippocampus and neocortex (7). Both Trk-B (specific to BDNF and NT4=5) and Trk-C (specific to NT-3) are also abundantly expressed in the hippocampus and neocortex. Conversely, both the Trk-A receptor (specific for NGF) and the p75 NT receptor are very scarce in these brain regions. However, both are found in the afferent cholinergic projections into the hippocampus and neocortex (6). Immunohistochemical studies researching the distribution of NT protein have shown that not always is there a correspondence between NT mRNA expression and their bioactive protein distribution. Although the hippocampus and neocortex have shown corresponding protein to mRNA localization, neuronal populations have been found to solely contain the bioactive NT proteins (8), suggesting that NT proteins may be transported from their site of synthesis to their site of action (reviewed in Refs. 9–11). 2.3.
Neurotrophins and Maturation of Neuronal Connectivity
The role of NTs was initially believed to be delimited to promoting neuronal survival and differentiation during development (12). This plasticity of the nervous system depends, to a large extent, on the properties endowed by NTs. In fact, now we know that NTs modulate synapse development and function (13) and are prominent in mediating activity-dependent functional
Behavioral Implications of Neurotrophic Actions
91
and structural plasticity in both embryonic and mature CNS (13–16). Particularly, the story that has emerged for BDNF has provided interesting insights into how neurotrophins can alter neuronal connectivity and modulate the functional complexity of neuronal circuits. Regulation by activity is an integral property for BDNF and can be orchestrated at the level of individual synapses; activity induces both BDNF vesicular release from synaptic membranes (17) and BDNF-induced protein synthesis in dendrites (18). This neurotrophic effect is further illustrated by findings that BDNF participates in the regulation of axonal and dendritic branching and remodeling (19–22), augments the efficacy of synaptic transmission (4,5,23,24), and participates to modulate the functional maturation of excitatory and inhibitory synapses (25–27). Neurotrophins are capable of regulating both the number of synapses and the complexity of axonal arborization. Studies conducted using transgenic mice have shown that BDNF regulates the expression of synaptic vesicle proteins and the density of synaptic innervation: BDNF knockout mice are deficient compared with their control counterparts in the number of innervating axons and synapses in the sympathetic system (28). In vivo imaging experiments have taken us further by enabling us to view the ability of BDNF to regulate synaptogenesis in arborizing axon terminals (29) (Fig. 1). Studies using electron microscopy (EM) analysis in TrkB and TrkC knockout mice revealed a significant reduction in axonal arborization and the number of synapses in the postnatal hippocampus (30). These findings were important for establishing NT involvement in the regulation of synapse formation, as they were observed in the postnatal developmental stage (P12–P13), when synaptogenesis is normally taking place.
Figure 1 BDNF increases retinal ganglion cell (RGC) axon arborization and green fluorescent protein (GFP)-synaptobrevin puncta in vivo. Tracings of threedimensional arbors illustrate the effects of BDNF on axon arbor complexity and synapse number at 0, 4, and 24 hr following BDNF application. Confocal microscopy was used to visualize individual RGC axons in the live, developing tadpole after direct tectal injection of vehicle solution (control) or BDNF. BDNF not only increases axon arbor complexity, but also increases the number and density of GFP-synaptobrevin clusters per axon terminal. Note the increased proportion of branch points with synaptic clusters in the arbors exposed to BDNF when compared with controls. Scale bar: 20 mm. (From Ref. 29.)
92
Gomez-Pinilla and Vaynman
Neurotrophins may mediate the maturation of synapses by both preand postsynaptic actions (Figs. 2,3). The increase in docked vesicles occurs presynaptically and is a significant indicator of increased transmitter release. Increased postsynaptic membrane thickness may be due to the incorporation of structural machinery, thus enhancing sensitivity to presynaptic release. In fact, it has been suggested that the ability of BDNF to regulate AMPA receptor expression in central glutamatergic synapses (33) may be responsible for BDNF’s ability to promote the transformation of immature ‘‘silent’’ synapses into mature synapses, which manifest functional AMPA receptors (34). The story narrating the effects of NTs on inhibitory synapses is based upon evidence for the action of BDNF on GABAergic synapses, both in vitro and in vivo (26,35–37). BDNF, but not NT-3, has been found to promote the establishment of inhibitory synapses in hippocampal and cerebellar synapses in cell culture (25,26,38–40). BDNF has been found to promote axonal length and branching of GABAergic synapses in hippocampal slice and cell culture and in the visual cortex in vivo (26,38,40,41), a mechanism which has been suggested to increase GABAergic synapses (42). Interestingly, a difference in BDNF and NT-3 effects have been observed for inhibitory synapses; BDNF elicits both axonal and dendritic changes whereas NT-3 seems specific for increasing dendritic length and branching (26).
Figure 2 (Facing page) Neurotrophins mediate synapse maturation. The ability of NTs to regulate synapse maturation is best illustrated by a study performed in hippocampal cultures, exploring the effects of BDNF and NT-3. By embryonic day 18 (E18), hippocampal cultures, taken from rat embryos, develop normal network connections to exhibit spontaneous synaptic activity (26). Interestingly, although by E16, cultural neurons have developed the architectural designs necessary for proper synaptic function, such as complex dendritic and axonal arborizations, EM identifiable synapses, and synaptic vesicle related protein, and glutamate receptors (26,31,32), most of them remain presynaptically functionally silent synapses. This means that without exogenous depolarization or glutamate application, to generate postsynaptic action potentials, the presynaptic synapse is functionally silent in its ability to do so. Interestingly, NT exposure of E16 hippocampal cultures to BDNF or NT-3, for as brief a period as a single day, was found to be sufficient to induce maturation of these synapses to functional status (a). Specifically, NT’s application significantly increased the number of presynaptically docked vesicles (b) and the thickness of the postsynaptic membrane (c). Notably, BDNF was more successful than NT-3 in augmenting both the number of presynaptic docked vesicles and the postsynaptic membrane thickness. As E18 neurons contain functional synapses without exposure to exogenous BDNF or NT3, it is believed that endogenous NTs may be responsible for mediating the synaptic maturation of these cells. (d) Electron micrographs illustrating that homozygous knockout mice for Trkb= and Trkc=, which are deprived of experiencing the effects of endogenous NTs due to their lack of functional receptors, exhibit decreased number of synaptic vesicles and postsynaptic membrane thickness. (From Ref. 26.)
Behavioral Implications of Neurotrophic Actions
Figure 2 (Caption on facing page)
93
94
Gomez-Pinilla and Vaynman
Figure 3 Neurotrophins as synaptic modulators. Neurotrophic regulation of synaptic function ranges from a constitutive to an activity-dependent role. Constitutive secretion of low levels of NTs from postsynaptic dendrites is believed to be required for the maintenance of normal synaptic function (a, top). Once intense synaptic activity is provided, the level of NTs transiently increases (a, middle), resulting in a local sprouting of nerve terminals to accommodate new synapse formation. In addition to regulating synapse morphogenesis, NTs participate in an activity-dependent refinement of connections. The low level of neurotrophic level in the constitutive state is sufficient to maintain multiple axons coinnervating the same postsynaptic dendrite (b, top). Synchronous activity at axons 1 and 2 results in a large postsynaptic depolarization immediately after synaptic activation of axons 1 and 2, subsequently producing a large amount of NT secretion at synapses 1 and 2 (b, middle). Axon 3, whose activity is uncorrelated with axons 1 and 2, does not experience postsynaptic spiking with its synaptic activation, thereby forgoing a large secretion in NTs at its synapse. The high level of NTs at axons 1 and 2 results in the sprouting of new spines (b, bottom). Conversely, the synapse containing axon 3 may experience synaptic weakening and axonal withdrawal from the nerve terminal if it continues to experience a loss of neurotrophic supply to adjacent synapses. (From Ref. 5,13.)
It is important to stress that the neurotrophic effect on the level of the synapse extends to the changes seen on the level of neuronal networks. This can be better understood when one considers the role that NTs play in forming neuronal circuitry as most prominently demonstrated in their
Behavioral Implications of Neurotrophic Actions
95
effects on the formation of ocular dominance columns in the cat visual cortex (43,44). During development, it is believed that ocular dominance columns are formed by activity-dependent competition between thalamic glutamatergic afferents from both eyes. This model stresses that each eye competes for a limited supply of NT factor, namely BDNF, which is necessary to stabilize the synaptic contact of the thalamic afferent to one eye (45). Studies have shown that either limiting BDNF with BDNF functionblocking immunoglobulins, which bind endogenously released BDNF (43), or infusing an excess of BDNF (44) into the developing kitten visual cortex impairs the development of ocular dominance columns. In particular, infusion of either BDNF or NT-4, but not NGF or NT-3, into a localized area of the striate cortex during the period when geniculate neuronal projections derived from the two eyes segregate into eye specific regions prevents the formation of ocular dominance columns (43). Taken together, these studies show that although the release of endogenous BDNF is necessary for the formation of these ocular dominance columns, it is the activity, namely the delivery of the limited endogenous BDNF supply to synapses exposed to afferent thalamic activity from visual stimuli, which structures the patterns of these columns. Not too long ago, when one conceived of neuronal circuitry, one regarded it as a rigid nonmalleable entity, once it had gone through its formative stages during development. Now, when one conceives of neuronal circuitry, one has become aware that this very act of conception may be actively contributing to changing and transforming neuronal networks, from the level of synapse to system. It is especially critical to note that NT activity mediates plasticity in the adult nervous system. As in the developing nervous system, BDNF has been found to regulate axonal sprouting and fiber density in the adult CNS. Tropea et al. (46) found that delivery of BDNF in conjunction with a drug application to reduce the inhibitory properties of neurite growth of the adult superior colliculus (chondroitinase; C-ABC) synergistically enhanced axonal growth of retinal ganglion axons after partial retinal lesion in the adult rat (Fig. 4). Although BDNF infusion alone, directly into the retinal ganglion cell (RGC) bodies (eye) or into the tectum via minipump, promoted sprouting of RGCs, a greater efficacy of sprouting was achieved by BDNF in conjunction with C-ABC (Fig. 4).
3. NEUROTROPHINS AND NEURAL ACTIVITY 3.1. Regulation of NTs by Neural Activity The NT hypothesis of synaptic plasticity proposes that neuronal activity enhances the expression, secretion, and=or actions of NTs at the synapse to result in the modification of synaptic transmission and connectivity.
96
Gomez-Pinilla and Vaynman
Figure 4 Confocal images of representative fields showing the density of RGC fibers at the border of the scotoma in control and treated rats, showing the effect of BDNF combined with various treatment on fiber density and sprouting in the adult nervous system. Retinal fibers are labeled with Alexa 594-conjugated subunit B of Cholera toxin (CTB). (A) Untreated animals, (B) the effect of C-ABC, (C) effect of intraocular injection of BDNF, (D) effect of the combined treatment with BDNF and C-ABC, and (E) mean values of RGC fiber density observed in the different experimental groups. D and E illustrate that fiber sprouting is maximally promoted by BDNF plus C-ABC. Scale bar: 25 mm. Numbers on the columns represent the fold increase in fiber density with respect to the mean value observed in untreated lesioned animals at 1 week. Error bars indicate SE. For each histogram, n ¼ 5–8 rats. (From Ref. 46.)
Different forms of activity have shown to be able to increase BDNF expression. The concept of activity-dependent BDNF expression was initially introduced in findings from experiments conducted on animals undergoing experimental seizures, which showed that seizure dramatically
Behavioral Implications of Neurotrophic Actions
97
increased expression of NGF (47) and BDNF mRNA in the hippocampus (3). In what follows, other stimuli were found to increase BDNF levels: It was discovered that inducing long-term potentiation (LTP) an electrophysiological correlate of learning and memory in the hippocampus also increases BDNF mRNA (48,49). Sensory stimulation was found to regulate BDNF with visual input in the visual cortex (50) and whisker stimulation in the barrel cortex (51). Additionally, physiological activities such as exercise (52,53), learning (54), and sleep and circadian rhythm (55,56) were found to increase BDNF. The mRNAs encoding the various NTs and their Trk receptors show selective expression in distinct brain regions (57). Interestingly, the regulation of NT transcription is dependent upon neuronal activity. The best-known site for NT regulation is the hippocampus, an area where already high levels of NTs are present in the basal condition. The BDNF and NGF mRNA levels are rapidly and significantly upregulated by epileptiform activity in both hippocampus and cerebral cortex (3), but an increase in BDNF mRNA is much more conspicuous in the hippocampus, increasing by over sixfold in the dentate gyrus within a 30-min time frame postseizure induction (3). In the hippocampus, BDNF mRNA has been localized to the dendrites of pyramidale cells (58). Moreover, neuronal activity may be responsible for this localization of BDNF, as high Kþ depolarization has been shown to translocate BDNF mRNA to hippocampal dendrites (59). The functionality of BDNF mRNA translocation, which constitutes local protein translation, has been proposed to enable synapse specificity in hippocampal LTP (60). Interestingly, BDNF itself has been shown to regulate its own mRNA translocation to dendrites (61). 3.2.
Activity-Dependent Secretion of NTs
The secretion of NTs can be either regulated or constitutive. Regulated secretion is dependent upon an extracellular or an intracellular signal to trigger secretion (62). In excitable cells such as neurons and neuroendocrine cells, neuropeptides are cleaved by prohormone convertase 1 and 2 (PC1 and PC2), a regulated pathway used for the secretion of proteins triggered by intracellular and extracellular signals (63). Likewise, NTs have been found to be cleaved by PC1 (64,65), indicating that they are secreted by a regulated pathway. Thus, it is no surprise that NTs can be secreted in response to depolarization in both neurons and neuroendocrine cells (66,67). Interestingly, protease cleavage is not necessary for NT secretion, as proforms of NGF and BDNF have been found to be secreted and then cleaved extracellularly (68).
98
Gomez-Pinilla and Vaynman
Experiments using cultured hippocampal neurons, in which the mRNAs for the precursor proteins pro-BDNF and pro-NGF were overexpressed, demonstrated that NGF was constitutively secreted while BDNF remained in the cytoplasm under resting conditions. Conversely, activity in the form of depolarization triggered the release of BDNF but not that of NGF (65). These findings illustrate that there is a differential regulation of NTs, i.e., BDNF secretion is activity dependent while NGF secretion is constitutive. Although other studies have found that both NT-3 and NT-4 expressions do not seem to be activity dependent (69), the regulation of neurotrophic expression remains as yet unsettled. In hippocampal neurons, experiments using a vaccina virus expression system showed that BDNF is sorted into the regulated pathway, whereas other NTs are mainly sorted into the constitutive pathway (65,70). Interestingly, NT-3 can employ the regulated pathway if it forms a heterodimer with BDNF (65,70). Studies visualizing the localization of BDNF have furthered our understanding of activity-induced secretion of BDNF. Transfection experiments employing BDNF–GFP (green fluorescent protein) fusion constructs have shown that BDNF is packaged in secretory vesicles, which are predominately transported to somatodendritic compartments and to a lesser extent to axons (71,72). The BDNF–GFP fluorescence was found to be concentrated at synaptic junctions, revealed by colocalization with the presynaptic secretory protein synapsin I and the postsynaptic scaf-folding protein PSD95 (71,72). Upon depolarization or high-frequency stimulation, the BDNF–GFP fluorescence spots quickly disappeared, suggesting the secretion of BDNF in synaptically localized secretory vesicles (17,72). Experiments conducted in central synapses have also shown that activity can concurrently increase the expression and the release of NTs; epileptogenic activity was shown to induce an increase in the expression of BDNF mRNA, which was accompanied by BDNF release (3). Given the involvement of NTs in neuronal protection following brain insults (73), it is particularly interesting that BDNF is the most abundantly expressed NT in the postnatal neocortex and hippocampus and shows enhanced expression in response to physiological levels of neuronal activity (48,49). 3.3.
Neurotrophins Modulate Synaptic Efficacy
Experiments conducted in Xenopus nerve–muscle culture were the first to demonstrate that the application of exogenous NTs could modify synaptic efficacy (5). When these cultures were acutely treated with BDNF or NT-3, the frequency of spontaneous quantal transmitter release and the amplitude of evoked postsynaptic responses increased (without change in quantal size). These findings indicated that NTs could enhance presynaptic-transmitter release. Studies conducted in cultured hippocampal neurons show that
Behavioral Implications of Neurotrophic Actions
99
BDNF, NT-3, or NT-4 application induces a rapid potentiation of glutamatemediated synaptic transmission (74). Concurring, studies performed in hippocampal slices demonstrate that perfusion with exogenous BDNF and NT-3 can dramatically potentiate glutamatergic transmission in the Schaffer collateral-CA1 synapses (4). Additionally, in vivo BDNF infusion has been found to induce a long-lasting potentiation of perforant path-dentate gyrus connections (75). Another way BDNF infusion can increase excitability in the hippocampus is by depressing GABAergic transmission in the CA1 region, thus effectively reducing synaptic inhibition (76,77). 3.4.
Pre- or Postsynaptic Action?
An important question to ask about the nature of NTs on synaptic plasticity is whether their release and their mode of action is pre- or postsynaptic. A presynaptic release would entail the anterograde transportation of NTs from the soma to the presynaptic terminal where they would be released, as a consequence of neuronal spiking, to be received by the postsynaptic cell. Alternatively, a postsynaptic regulation would require that neuronal activity acts on the postsynaptic neuron to induce NT secretion from dendrites, which would in turn acts in a retrograde fashion on the presynaptic neuron (Fig. 5). Studies using Xenopus nerve–muscle cultures showed that postsynaptically secreted NTs from dendrites could induce presynaptic transmitter release. Terminals that were in contact with myocytes overexpressing NT-4 were found to contain a higher frequency of spontaneous acetylcholine (ACH) secretion, and this effect on acetylcholine (ACH) secretion was blocked by scavenging NT-4 with TrkB-IgGs (78). It has also been suggested that activity-dependent postsynaptic signals can regulate presynaptic
Figure 5
Illustration of anterograde and retrograde action of neurotrophins.
100
Gomez-Pinilla and Vaynman
excitability, and specifically, that NT-3 can act as a retrograde signal to affect motor neuron excitability (79). Studies in cultured CNS neurons have also found that NTs can be both distributed and secreted postsynaptically (17,66,67,72). Studies employing BDNF–GFP constructs have morphologically illustrated, in dissociated hippocampal neurons, BDNF–GFP clustering to postsynaptic density-95, thus providing evidence of a postsynaptic localization for BDNF (72). Interestingly, BDNF–GFP imaging has also provided evidence to suggest the activity-dependent transfer of BDNF to postsynaptic neurons: BDNF-GFP was anterogradely transported to axon terminal and taken up by postsynaptic neurons (80). Studies by Levine et al. (81), using BDNF bath application and whole cell patch clamp recordings, demonstrated that BDNF could act postsynaptically. Application of exogenous BDNF produced a two fold increase in synaptic charge (a measure of integrated synaptic current) within 5 min. This effect was shown to be specific, as neither denatured BDNF nor NGF was effective. Furthermore, this BDNF effect was shown to be mediated by its cognate Trk receptor, as application of k252a, an antagonist of Trk tyrosine kinase activity, blocked the effect. Conversely, increasing phosphorylation activity by injecting okadaic acid, a phosphatase inhibitor, increased BDNF’s effect on synaptic charge by two fold. Interestingly, these intracellular agents did not affect excitatory postsynaptic current (EPSC) frequency, although they had a robust effect on EPSC amplitude; k252a eliminated the BDNF-mediated effect on EPSC amplitude while okadaic acid increased the amplitude two fold. The findings suggest that the BDNF-mediated increase in EPSC amplitude may be accomplished postsynaptically, which entail phosphorylation-dependent mechanisms. The hippocampus studies have found evidence to support both preand postsynaptic action of NTs. BDNF mRNA was found to be localized in the granule layer of the dentate gyrus while immunoreactivity experiments displayed high BDNF protein levels in the mossy fibers (82,83). Taken together, these findings suggested that BDNF protein was anterogradely transported from the soma to the presynaptic terminal. Additionally, both kindling (83) and electroconvulsive seizures (84) increase BDNF levels in the mossy fibers while epileptogenic activity induces TrkB receptor activation in CA3 (85). As an increase in BDNF levels in the mossy fibers indicates an accumulation of the protein at the presynaptic terminal and an increase in TrkB receptor activation on CA3 neurons denotes a readiness of the postsynaptic terminal for the secretion of NT from the presynaptic terminal, both of these findings suggest an activity-dependent secretion of BDNF from the presynaptic terminal. Conversely, studies conducted in hippocampal-cultured neurons, showing the soma-dendritic localization of BDNF, have suggested that both BDNF localization and its activitydependent secretion may be postsynaptically conducted in hippocampal neurons (67).
Behavioral Implications of Neurotrophic Actions
3.5.
101
Neuronal Activity Facilitates Neurotrophic Action
Neuronal activity can enhance the effects of neurotrophic action such as NT-potentiated synaptic transmission. Studies performed using Xenopus neuromuscular junctions have found that presynaptic stimulation greatly enhances the synaptic potentiation induced by exogenous BDNF application (23). This synergistic effect of presynaptic depolarization was found to be abolished by inhibiting cAMP signaling (23), indicating that depolarization elevates intracellular cAMP levels. In RGCs, neuronal depolarization in conjunction with elevated intracellular cAMP levels increased the number of TrkB receptors at the plasma membrane (86). Thus, a possible mechanism responsible for this activity-induced efficacy of NT action may be an increment in the number of TrkB receptors, which would provide an increased substrate for BDNF action.
4. NEUROTROPHINS AND LEARNING AND MEMORY 4.1. Involvement of NTs in Learning and Memory The plethora of scientific research conducted on the involvement of BDNF in activity-dependent synaptic plasticity has generated the consensus that BDNF is important for learning and memory. Both learning and memory tasks (87) and LTP (the synaptic candidate underlying learning and memory) (48,88) selectively increase BDNF mRNA levels in the hippocampus, an area believed to be critical for learning and memory. A long-line of scientific research in animal models has been supplemented with the finding that BDNF is important for human learning. An important connection between cognitive function and BDNF can be illustrated in a clinical study performed on humans, which found that individuals expressing a specific polymorphism in the BDNF gene have learning impairments (89). This association between BDNF and learning and memory was found to exist when measuring the performance of rats on the Morris water maze (MWM) task (90). On the MWM task, the rat learns to locate the hidden platform by using spatial cues located around the water maze pool. The latency to find the platform (A) or the distance swam in the platform quadrant after the platform was removed (B) was correlated with levels of BDNF for individual animals. It was discovered that levels of hippocampal BDNF are proportional to learning performance in the water maze for individual animals. The results of this study suggest that hippocampal levels of BDNF are directly related to learning efficiency (Fig. 6). Studies have reported that BDNF mRNA levels are increased in the hippocampi of rats that have undergone 3 or 6 days of MWM training (54). Other hippocampal-dependent learning paradigms, such as contextual fear conditioning, have reported increases in BDNF mRNA levels in the CA1 region of the hippocampus (91). These studies may implicate BDNF
102
Gomez-Pinilla and Vaynman
Figure 6 Association between BDNF and spatial learning for individual animals. (A) Significant negative correlation between hippocampal BDNF protein levels and the latency required to find the hidden platform (r ¼ 0.94, p < 0.01). (B) Significant positive correlation between hippocampal BDNF protein and the swimming distance within the platform quadrant (r ¼ 0.95, p < 0.01) after the platform was removed. (From Ref. 54.)
in learning and memory functions through association, while more direct evidence of BDNF’s involvement comes from BDNF knockout and inhibition studies. Quenching endogenous BDNF with function-blocking antiBDNF antibodies impairs spatial learning and memory in rats. Mu et al. (92) found that rats who received 7 days of intracerebroventricular infusions of anti-BDNF antibodies showed poorer performance on the MWM task, longer escape latencies during the training trials, and poorer task completion during the probe trial (Fig. 7). Performing a more specific inhibition of BDNF, by injecting BDNF antisense oligonucleotides into the dentate gyrus of the hippocampus, Ma et al. (93) showed that blocking BDNF in a specific subfield of the hippocampus was sufficient to decrease hippocampal BDNF mRNA levels and memory consolidation on an inhibitory avoidance task. Furthermore, it was found that this same blocking protocol for BDNF was able to significantly reduce LTP, considered an electrophysiological correlate of hippocampal learning and memory, in perforant path-dentate gyrus synapses (93). BDNF has been shown to be important for both reference memory, which comprises trial independent information, and working memory, in which the information retained changes across repeated trials. Mizuno et al. (94) found that after a 28-day training period on the eight-arm radial Figure 7 (Facing page) (A) Average escape latencies in rats treated with normal sheep IgG (n ¼ 6) and antibody to BDNF (n ¼ 7) in the place of navigation tests. (B) The learning curves plotted as percentages of the first block. Each point represents mean SE; p < 0.05, p < 0.01. (C) Representative trails of the spatial probe tests, recorded by a video camera. Rats treated with normal sheep IgG spends more time in quadrant P than the other three quadrants. (D) Rats treated with antibody to BDNF do not prefer quadrant P over the other quadrants. (From Ref. 92.)
Behavioral Implications of Neurotrophic Actions
Figure 7 (Caption on facing page)
103
104
Gomez-Pinilla and Vaynman
maze tasks, rats given continuous intracerebroventricular infusion of BDNF antisense oligonucleotides, starting 4 days prior and during a testing period, performed worse on both reference and working memory (Fig. 8). Interestingly, the importance of BDNF in learning and memory may be age dependent. A study conducted by Linnarsson et al. (95) found that young (6–8 weeks) heterozygous BDNF knockout mice showed no deficits in memory retention compared with their age-matched wild-type controls. Alternatively, Linnarsson et al. found that older BDNF knock mice (10 months) and the age-matched wild-type mice were impaired on the MWM compared to their young counterparts, with impairments being more severe in the KO mice. Studies manipulating the function of the TrkB receptor, the cognate receptor for BDNF, have also found impairments in hippocampal-mediated learning. Mice with a time regulated TrkB knockout, induced by using the CRE-lox P recombination and the a–calcium–calmodulin protein kinase (a-CAMKII) promoter (96), and mice overexpressing a truncated noncatalytic form of the TrkB receptor, TrkB.T1 (97), show an impaired performance on the MWM task. The importance of BDNF in mediating the effects of a synaptic correlate of learning and memory, LTP, can be seen from the results of studies employing BDNF inhibition or knockout. Transgenic animals with
Figure 8 Spatial reference (A) and working (B) memory formation in nonoperated controls rats (n=7) and rats that were continuously infused with BD7NF antisense (n=6) and sense oligonucleotides (n=6) into the cerebral ventrical. Maze training was started 4 days after the start of continuous intracerebroventricular infusion of the oligonucleotide. Each value represents the mean SE. A and B illustrate that rats infused with antisense BDNF oligonucleotides had a greater number of errors on the water maze performance than their counnterparts (control and sense oligonucleotides) indicationg that antisense BDNF treatment significantly impaired both reference and working memory. (From Ref. 34.)
Behavioral Implications of Neurotrophic Actions
105
diminished BDNF expression lose the ability to induce LTP (98). Reinstating BDNF into the depleted hippocampus has been shown to augment these deficits. The ability to induce LTP can be amended with exogenous BDNF application (98) or transfection of hippocampal slices with a BDNFexpressing adenovirus (99). In order to properly examine the role of BDNF in learning and memory, it is important to consider that there is dissociation between spatial learning and hippocampal LTP. Impairments in LTP are not always associated with spatial learning deficits, as indicated by studies which showed that spatial learning does not require hippocampal CA1 LTP (100–102). In the TrkB.T1 mutants, learning deficits were exhibited on the MWM maze, but CA1 LTP was normal (97). Alternatively, one may argue that LTP in a specific hippocampal subfield may be important for hippocampaldependent learning, as evidence from van Praag et al. (103) showed that mice who had poor performance on the MWM task also had decreased BDNF levels and impaired dentate gyrus LTP. Interesting findings concerning the role of BDNF and other trophic factors in memory consolidation have shown that BDNF, but not NGF or NT-3, plays a role in consolidating short-term memories into long-term memories (104). Antisense oligonucleotides for BDNF, NGF, and NT-3 were administered in day-old chicks and their performance was tested on the one-trial avoidance task with each respective administration. Although each of the treatments inhibited both the mRNA and protein levels of the respective trophic factor during the time of training, only inhibition of BDNF produced amnesia for the avoidance behavior 3–24 hr after training. Interestingly, BDNF inhibition did not produce amnesia 1 hr following the avoidance training, a finding that has been suggested by some to indicate that BDNF may not be necessary for some forms of long-term memory in this task. Unfortunately, there are inconsistencies in the findings regarding neurotrophic factor effects on learning and memory. For instance, another study found that NT-3 overexpression prevented age-related memory deficits in mice (105). Others found that high levels of hippocampal NGF mRNA levels are correlated with poorer MWM performance (106). Therefore, it is important to consider that these discrepancies may be due to the differences in protocol, species, memory task, age, in vitro vs. in vivo conditions, among others. 4.2.
Mechanisms Mediating BDNF Neurotrophic Action: Relevance to Neuronal Plasticity and Learning and Memory
Through what molecular mechanisms does BDNF mediate its action on neuronal circuits relevant to synaptic plasticity and learning and memory? Neurotrophins bind to their cognate Trk receptors to cause dimerization and kinase activation. The cytoplasmic domain of each tyrosine kinase
106
Gomez-Pinilla and Vaynman
receptor contains 10 conserved tyrosines. Phosphorylation of the 3-autoregulatory loop amino acid residues of the kinase domain further activates the kinase (1,107). Phosphorylation of the other remaining residues creates docking sites for adaptor proteins that couple the activation of Trk receptors to intracellular signaling cascades. The intracellular signaling proteins activated by the autophosphorylated Trk receptors include phosphainositol-3-OH kinase (PI3K)=Akt kinase (108,109), and phospholipase Cg (PLCg) (109,110) and the adaptor proteins Shc, rAPS, and SH2-B (109,110). The BDNF activation of its TrkB receptor has been shown to lead to Ras-dependent activation of mitogen-activated protein (MAP) kinase (109). The MAP kinase=ERK cascade is consequential to the integration of multiple intracellular signals (111) and leads to the activation of a diverse set of targets, such as cAMP response element binding protein (CREB) mediated transcription, protein synthesis, and voltage=ion gated channels. In fact, the task MAP kinase serves as an integrator is appreciated when one considers the ability of protein kinase C (PKC) to converge on MAPK (112). Thus, the BDNF=TrkB signal, which activates PLCg to hydrolyze phosphotidylinositol-4,5-bisphosphate to inositoltriphosphate, subsequently triggers the release of calcium from intracellular stores and diacylglycerol and activates PKC (108). Intracellular calcium also activates calcium–calmodulin dependent kinases (CAMK) (113), through which BDNF has been shown to induce the phosphorylation of CREB, subsequently triggering gene transcription (113). If one were to simplify the pathways by which BDNF action exerts its effects, one would say that BDNF relies on CAMK- and MAPK-dependent pathways. Both of these pathways lead to CREB-mediated gene transcription (113,114) and phosphorylation of synapsin I (115), two prominent end-products of BDNF action. CREB is one of the best-described stimulus-induced transcription factors involved in adaptive responses (113,116) and long-term memory (117). CREB may provide a self-perpetuating loop for BDNF action, as it has been found to regulate BDNF transcription (118) and is, in turn, itself regulated by BDNF (113,119). On the other hand, synapsin I may be a part of a number of molecules that BDNF controls, important for synaptic transmission. This is illustrated with the finding that presynaptic neurotransmitter release via MAPK phosphorylation of synapsin I is coupled with the ability of BDNF action through its TrkB-R to enhance communication between neurons (120). In fact, deletion of the BDNF gene in mice results in a reduction in synaptic proteins, sparsely docked vesicles, and impaired neurotransmitter release (121). The BDNF activation of synapsin I may exert synaptic plasticity changes in additional alternate ways. Besides modulating transmitter release (120), synapsin I is involved in the formation and maintenance of the presynaptic structure (122,123) and in axonal elongation (124). Another way by which BDNF action may lead to CAMK- and MAPK-induced plasticity changes is by interacting with the N-methyl-d-
Behavioral Implications of Neurotrophic Actions
107
aspartate receptor (NMDA-R). NMDA-R activation (125) can initiate calcium–calmodulin protein kinase II (CAMKII) (126) to converge on the MAPK cascade (127), leading to the expression of plasticity-dependent genes (128). BDNF can potentiate synaptic transmission through the NMDA-R (129), providing an alternative path leading to CAMKII- and MAPK-mediated changes. The mechanisms through which BDNF affects synaptic plasticity may underlie hippocampal-dependent learning, conceded by the observation that molecules comprising these pathways are important for both neuronal plasticity and learning and memory. The NMDA-R plays a critical role in neural plasticity, by modulating LTP (130), mediating learning and memory processes (128), and regulating synaptogenesis (131). Protein kinase C is important for neurite outgrowth (112) and its gene deletion results in mild spatial and contextual learning impairments (132). CAMKII is believed important to participate in mechanisms underlying learning and memory (117) especially, as mice with gene deletions of a-CAMKII show an impaired performance on spatial learning tasks (133). MAPK is involved in nuclear signaling (134), LTP production (135), and learning and memory formation (136). In fact, the capacity of MAPK to induce long-lasting changes in synaptic plasticity is attributed to its ability to regulate (113) and prolong the transcriptionally active state of CREB (137), important for various forms of learning and memory (117,138).
5.
NEUROTROPHINS AND BEHAVIOR AND LIFESTYLE: EXPERIENCE-DEPENDENT PLASTICITY 5.1. Behaviorally Induced NTs Improve Neural Functions: Exercise, Synaptic Plasticity, and Learning and Memory Studies over the last decade have demonstrated the beneficial effects of exercise on improving or maintaining cognition, especially in aged populations. Human studies, involving a large number of subjects, have shown that a provision of exercise reduces the normal decay of cognitive function observed during aging (139) and may reduce the risk of getting Alzheimer’s disease. Exercise can also increase the quality of life of individuals, raising alertness and responsiveness to various stimuli and situations. The practice of some forms of exercise is becoming common in many senior citizen homes, in hope of reducing the risk of developing cognitive impairments characteristic of Alzheimer’s disease. The current view is that exercise, as other types of behavioral stimuli, can activate specific neural circuits, involving modifications in the way that information is transmitted across cells at the synapse and the action of specialized molecules. Interestingly, voluntary wheel running in rodents elevates levels of BDNF and synaptic proteins in select regions of the brain and
108
Gomez-Pinilla and Vaynman
spinal cord. The selective effect of exercise on the synaptic machinery within the hippocampus has provided the initial molecular support to previous observations showing the beneficial effects of exercise on cognitive function. Clinical studies in humans (139–141) and studies in rodents (103,142,143) demonstrate that exercise can benefit learning and memory. The finding that both exercise (52,144,145) and learning (54,87) induce an increase in hippocampal BDNF levels propones that BDNF may be the candidate most likely to mediate these benefits of exercise on cognitive function. Exercise induces an increase in BDNF levels in the area vital for learning and memory formation, the hippocampus (Fig. 9). In fact, the hippocampus seems ideal for utilizing the exercise-induced increases in BDNF, as it exhibits a high level of the receptor (TrkB receptor) through which BDNF exerts its action. The use of a novel microbead injection technique has shed light on the contribution of different pathways to the exerciseinduced increases in the mRNA levels of BDNF, TrkB, and molecules
Figure 9 Mechanistic illustration of BDnf interaction at the synapse, emphasizing its ability to impact synaptic plastictiy on both the pre- and post-synaptic membrane. BDNF activation of its TrkB receptor is shown activating downstream intracellular signaling cascades, CAMK and MAPK. Signals leading from BDNF=TrkB and NMDA-R activation can integrate downstrean to impact CREB mediated transcription, protein synthesis, and channel activity. BDNF also has a large impact on presynaptic transmitter release by activating MAPK to regulate the presynaptic phosphoprotein synapsin I.
Behavioral Implications of Neurotrophic Actions
109
involved in transcription and synaptic transmission, CREB, and synapsin I (53). It was found that besides BDNF, additional molecules, i.e., the N-NMDA-R, CAMKII, and the MAP-K cascade, mediate the effects of exercise on hippocampal synaptic plasticity. These molecules have a welldescribed association to BDNF action, comprising a basic mechanism through which exercise may promote synaptic plasticity in the adult brain (Fig. 10). Particularly interesting was the finding that BDNF is capable of concurrently increasing the mRNA levels of both itself and its TrkB receptor, suggesting that exercise may employ a feedback loop to augment the effects of BDNF on synaptic plasticity (Fig. 10). Additionally, evaluation of an array of genes activated by exercise has corroborated the involvement of these molecules in exercise-induced synaptic plasticity and have provided insight into additional mechanisms subserving the exercise effect. In addition to synapsin I, exercise increases molecules involved in synaptic function, such as the mRNAs for syntaxin
Figure 10 Diagram of proposed basic mechanism, by which exercise can affect the synapse, using BDNF as a central player. Exercise may employ both BDNF and NMDA-R, possibly by using BDNF to potentiate transmission through NMDAR. Exercise may activate BDNF through TrkB-R and NMDA-R, acting through CAMKII, to converge on the MAPK cascade. The CAMKII–MAPK pathway seems to be important for the regulation of CREB mRNA levels during exercise. In contrast, BDNF may interact with NMDA-R to regulate the exercise-induced mRNA levels of BDNF, TrkB, and synapsin I, by employing CAMKII without MAPK activation. The importance of BDNF as a central player of the effects of exercise is noteworthy given the fact that as exercise increases both BDNF mRNA and proteins levels, BDNF may have the ability to maintain the effects of exercise on synaptic plasticity by a selfperpetuating loop. (From Ref. 53.)
110
Gomez-Pinilla and Vaynman
Figure 11 Genes affected by exercise in the hippocampus as assessed by microarray. Animals were exposed to running wheels for 3, 7, and 28 days and their hippocampal mRNA was used for microarray (Clontech) analysis. Out of a pool of 1200 genes, a subgroup of genes that belong to the categories of trophic factors and synaptic plasticity were upregulated. Genes which showed a ratio 1.8 for exercise=sedentary were considered differentially expressed. For illustrative purposes, genes were grouped according to their biochemical function. (From Ref. 90.)
and synaptotagmin (145). Furthermore, this study found that exercise increased the glutamate transporter, EAAC1, and downregulated specific components of the GABAergic system, indicating that exercise may enhance neuroprotective mechanisms. EAAC1 is important in removing extracellular glutamate from the synaptic cleft (146), without which excessive glutamate induces neuronal death (147). GABA inhibition has been shown to enhance functional recovery after CNS injury (148). A prominent finding from this study showed that although exercise affects the expression of other neurotrophic factors, BDNF is the only neurotrophic factor consistently elevated after a few weeks of continuous exercise (145) (Fig. 11).
Behavioral Implications of Neurotrophic Actions
111
Another positive and prominent effect of exercise is that it increases neurogenesis in the adult dentate gyrus (103). To date, scientific findings have established that the adult mammalian brain contains limited pockets of neural stem cells that are capable of undergoing neurogenesis, i.e., dividing and differentiating into neurons. This process of neurogenesis has been found to be delimited to the subventricular zone, lying adjacent to the lateral ventricle, and the subgranular zone of the hippocampal dentate gyrus (149). The nascent neurons, which constantly arise from the precursor cells in the subgranular zone of the dentate gyrus, have been found to migrate into the subgranular cell layer and incorporate into existing synaptic circuitry (150). As hippocampal neurogenesis has been shown to actively play a role in the formation of hippocampal-dependent memories (151), this exercise-induced neurogenesis may contribute to remodeling hippocampal synaptic circuitry and enhancing cognitive function in the adult brain. Indeed, exercise-induced hippocampal neurogenesis is accompanied by an enhancement in learning and memory and its electrophysiological correlate, LTP (103).
5.2.
Neurotrophins and Neurotransmitter Systems: Stress, Depression, and Mood Disorders
The ability of neurotrophic factors to regulate intracellular messenger cascades enables them to influence neuronal function, spanning a control over activities involving gene expression, neuronal morphology, protein synthesis, activity, maturation, maintenance, and survival. Because these signal transduction cascades are also regulated by neurotransmitter systems, NTs essentially interact with neurotransmitter systems to affect brain function. This interaction can be seen with the intersection of intracellular signaling cascades; G-protein coupled second messenger cascades, regulated by neurotransmitters, with the protein tyrosine kinase-activated signal transduction cascades, regulated by NTs. The interaction of BDNF with neurotransmitter systems becomes especially relevant when examining its role in stress, depression, and mood disorders. Multiple findings have indicated that BDNF serves as a target for antidepressant treatment in both animal and human studies. Chronic, but not acute, administration of various classes of antidepressant drugs increases the expression of BDNF and its cognate TrkB receptor in the hippocampus (152). The induction of BDNF and its TrkB receptor mRNA overlaps with the time course for which antidepressant treatments show therapeutic effects. Direct infusion of BDNF into the midbrain exerts antidepressant effects in two animal models of depression, the forced swim model and the learned helplessness model (153,154).
112
Gomez-Pinilla and Vaynman
Pretreatment with antidepressants has been shown to block the stress-induced down regulation of hippocampal BDNF mRNA levels (152). Decreased BDNF serum levels are found in major depressed subjects (155). Postmortem studies show that BDNF immunoreactivity is increased in the hippocampus of human subjects treated with antidepressant medications (156). Interestingly, stress and corticosterone treatments decrease BDNF expression in the CA1 and CA3 subfields and dentate gyrus of the hippocampus (157) and could be relevant to contributing to our understanding of stress-related depressive disorders. To the extent that depression is considered resultant from a deficiency in one or more monoamines [serotonin, 5-HT; norepinephrine (NE); dopamine, (DA)], BDNF may play a pivotal role in modulating neurotransmitter systems when one considers the following findings. Infusion of exogenous BDNF was found to increase monoamine activity, especially serotonin levels in the brain (153). A more recent study conducted using striatum slices showed that BDNF modulates the stimulated release of serotonin, DA, and GABA through its TrkB receptor, but that this modulation is dependent on the presence of depolarizing activity (158). Additionally, BDNF exerts a potent trophic effect on serotonergic and noradrenergic neurons to regulate their morphology, neurotransmitter metabolism, and firing patterns (159,160). Other findings indicate that BDNF lies downstream to monoamine activity and the effects of antidepressant medication. The repeated administration of antidepressants, which selectively inhibit the reuptake of 5-HT or NE, increases, the expression of BDNF in the hippocampus (152,161). The increases in BDNF expression may be due to an increase in CREB-mediated transcription following monoamine application (162). Chronic antidepressant treatment, consisting of 5-HT and NE reuptake inhibitors, increased the expression and phosphorylation state of CREB in the cerebral cortex and the hippocampus (163). Additionally, the induced expression of CREB, using viral gene transfer technology, singularly into the dentate gyrus of the hippocampus was sufficient to produce antidepressant effects in animal models of depression (164). Findings from studies exploring the combination of antidepressant medication and physical activity show an additive and accelerated increase in the levels of BDNF mRNA in the hippocampus, suggesting a convergence on the molecular level (144). Thus, G-coupled neurotransmitter channels and tyrosine kinase receptors may intercept at intracellular signaling cascades to induce BDNF expression. 5.3.
Environment Enrichment and Alternative Lifestyles
The intrinsic limitation of mature neurons to divide and replace lost neurons has been accepted as dogma in the neuroscience community, until recently.
Behavioral Implications of Neurotrophic Actions
113
Studies exploring the effects of stimulating environments on neuronal plasticity and learning and memory have found that exposure to an enriched environment induces the formation of new neurons in the hippocampus. In the enriched environment condition, housing rats in units of 6–12 numbers per cage encourage social interactions. The provision of permanent objects, such as ladders, running wheels, ropes, platforms, tunnels, and boxes, is provided for sensory and physical stimulation (165). Many studies have found that enriched environment promotes neuronal plasticity such as neurogenesis in the adult hippocampal dentate gyrus (166,167) and an upregulated expression of BDNF and NGF (87). Given that both neurogenesis and neurotrophic factor seem important in learning and memory process (92,151), it is not surprising that enriched environment animals perform better on spatial tests of learning and memory than their counterparts housed in standard or impoverished conditions (166,168). In fact, exposure to an enriched environment improved spatial memory on the MWM task in rats and correlated with an increased BDNF expression in their hippocampi (87). Moreover, exposure of rats to enriched environment conditions during their adult lives has the ability to counteract the decreased learning performance associated with nonhandling in infancy (165). Additional promising findings come from studies exploring the effects of enriched environment on aging. Aging seems to be accompanied by a decrease in baseline hippocampal neurogenesis (167,169,170). Kempermann et al. (171) have found that aged mice living in enriched environment conditions show a fivefold increase in neurogenesis in the hippocampus. This increase in hippocampal neurogenesis in the senescent animals did not wear off with prolonged exposure to the stimulating environment, suggesting that exposure to more stimulating conditions can be restorative of baseline levels of hippocampal neurogenesis. Moreover, these changes in neuronal plasticity were accompanied with behavioral modification, as enriched environment mice showed enhanced ability for task acquisition (171). Finally, these findings illustrate the ability of the aged brain to actively respond by upregulating neuronal birth in response to a more challenging environment, bear the witness that the brain remains plastic even in old age. 5.4.
Consequences of a NT Deficiency: The Effects of a Saturated High-Fat Diet
Recent studies in rodents indicate that a diet rich in saturated high-fat refined sugar (HFS), similar in composition to the average popular diet of most industrialized western societies, can threaten neuronal health and reduce the capacity of the brain for learning. These studies have demonstrated that the consumption of an HFS diet decreases BDNF, CREB, and synapsin I mRNA selectively in the hippocampus (90) (Fig. 12). In addition, the hippocampal expression of the growth-associated protein 43 (GAP-43), important
114
Gomez-Pinilla and Vaynman
Figure 12 Effect of high-fat refined sugar (HFS) diet on BDNF mRNA and protein levels in the hippocampus. (A) Rats maintained on the HFS diet showed decreased levels of BDNF mRNA after 2 months (75%), 6 months (77%), and 2 years (68%). (B) BDNF protein levels of rats maintained on the HFS diet were reduced after 6 months when compared with rats maintained on a low-fat complex carbohydrate (LFCC) diet (33%), as measured by ELISA. Immunohistochemistry for BDNF performed in saggital hippocampal sections from rats exposed to the HFS diet (D) revealed decreased BDNF immunostaining when compared with LFCC controls (C). Values represent mean SE; p < 0.05, p < 0.01. (From Ref. 90.)
for axonal growth, neurotransmitter release, and learning and memory, was also reduced after the HFS diet. Furthermore, the HFS diet reduced the capacity to learn a spatial memory task in the water maze. Interestingly, access to voluntary physical activity for as short as 2 months was sufficient to revert the effects of the HFS diet (172). That is, rodents fed the HFS diet, which were exposed to exercise, did not experience neither the cognitive decline nor the deficiencies in BDNF, synapsin I, CREB, and GAP-43 seen in sedentary HFS-diet rats. These findings emphasize the value of exercise as a compensatory strategy, functioning to ameliorate the effects of an unhealthy lifestyle on both cognition and neural plasticity. There is unequivocal evidence that exercise can protect from various types of diseases such as cancer, diabetes, and insults related to lifestyle. It is also likely that other dietary factors and regimens can have a benevolent effect on brain function.
Behavioral Implications of Neurotrophic Actions
115
Dietary factors are considered to be important predictors of the general health of individuals. It has recently been shown that the consumption of an HFS diet aggravates the harmful outcome of traumatic brain injury (TBI) on hippocampal plasticity to reduce BDNF, as well as CREB and synapsin I, molecules through which BDNF is known to exert changes in hippocampal synaptic plasticity (173). Not only was there a decrease in these molecules, but also the normal relationship between BDNF and the active phosphorylated state of these molecules was disrupted with the presence of the HFS diet in the TBI brain. These reductions were accompanied with a decreased performance on the MWM task, which effected latency to find the platform and probe trial performance (Fig. 13). Other alarming findings about the effects that detrimental behaviors may have on brain function and health come from studies on eating disorders. Both anorexia and bulimia nervousa decrease serum levels of BDNF in humans (174). As BDNF can cross the blood–brain barrier (175), it has been suggested that serum BDNF levels may reflect brain BDNF levels (174). Given the preponderance of BDNF in cognitive and neuronal plasticity and new findings showing that humans with a specific BDNF mutation are compromised in cognitive function (89), this finding may provide additional evidence for the importance of a healthy lifestyle for maintaining a healthy brain. 5.5
Neurotrophins and the Spinal Cord
A robust body of evidence indicates that repetitive locomotor activity can improve functional recovery following different types of injuries to the spinal cord in humans and animals. It appears that functional recovery is highly task specific, such that rehabilitative strategies that simulate walking are particularly effective in promoting the recovery of locomotion. In addition, to the effects of exercise training on recovery of locomotion per se, even minimal exposure to exercise has a strong effect on improving the quality of life of individuals with reduced mobility. It remains to be elucidated that what type of interventions, in terms of exercise routines, can accelerate recovery processes. Voluntary wheel running and forced treadmill exercise elevate the expression of BDNF and molecules important for synaptic function and neurite outgrowth in the spinal cord and innervated skeletal muscle (176). These molecules may comprise the phenomenon of learning found to occur in the spinal cord of fully transected animals (177). It is likely that neurotrophic factors being produced in muscle as a result of exercise can be transported to the spinal cord via motoneuron axons and help spinal cells (178). Specifically, it was found that BDNF and NT-3 mRNA levels increased in the lumber region of the spinal cord and in the soleus muscle, whose innervating motoneurons are located in the lumbar region. The findings that the
116
Gomez-Pinilla and Vaynman
Figure 13 Effect of fluid percussion injury (FPI) on BDNF mRNA and protein levels in the hippocampus of reduced fat diet (RD) and high-fat refined sugar (HFS) deit rats. The combination of FPI and HFS produced a decrease in BDNF mRNA (A) and protein (B). Effect of FPI and HFS on spatial learning and memory retention. (C) HFS and RD intact rats had similar escape latencies. (D) FPI resulted in increasing the escape latency time for both HFS and RD animals, with the combination of FPI producing the longest escape latency times. (E, F) Probe trial, as a test of memory retention, showed that HFS rats exposed to FPI spent the least amount of time in target zone when compared with other groups. Values represent mean SE; p < 0.05, p < 0.01. (From Ref. 93.)
Behavioral Implications of Neurotrophic Actions
117
Figure 14 BDNF immunohistochemistry performed in coronal sections of the lumber spinal cord of sedentary and exercised rats. Low (A) and high (B) magnification of these coronal sections show the typical BDNF immunohistochemistry in sedentary rats. Sedentary animals have light BDNF immunostaining in their motor neuron cell bodies. Low (C) and high (D) magnifications show BDNF immunostaining in the exercised rats. Exercise rats have stronger BDNF immunostaining, especially in the cell bodies (arrows) and axons (arrow heads). Calibration is 20 mm (bar shown in B and D), n ¼ 5 per group. (From Ref. 52.)
soleus muscle contained significant increases in BDNF mRNA levels without concurrent increases in protein and that the spinal cord protein levels far exceeded the small increases in spinal BDNF mRNA levels have lead to the suggestion that neuromuscular activity might increase retrograde transport of BDNF from the muscle (178). The results of several studies in which BDNF has been added to the neural milieu support the possibility that these factors promote survival and growth of brain and spinal cord neurons affected by several types of insults (179,180). It has been shown that BDNF administration after midthoracic complete spinal cord transection improves the functional recovery of hindlimb stepping and that these
118
Gomez-Pinilla and Vaynman
changes appear to be associated with neuronal sprouting at the injury site (181,182). Some of the protective effects of exercise on the spinal system have been recently illustrated in the dorsal root ganglion, a relay station of sensory information going to the spinal cord. Exercising prior to a sciatic nerve lesion has been shown to increase the capacity for regeneration of the severed sciatic nerve (183). Therefore, exercise via NTs can be a suitable therapy to promote functional restoration following injury to the spinal cord.
6.
CONCLUSIONS AND RESEARCH DEMANDS
The brain can change to cope with challenging situations to remain healthy in a capacity known as brain plasticity. Several research fronts indicate that NTs can enhance brain plasticity. It is fascinating that select behaviors such as exercise, environmental enrichment, and diet can translate into changes in neurotrophic factor expression to help maintain and strengthen neural circuits. Recent advances in the understanding of the mechanisms through which experience can impact neural function can help develop therapies aimed at maximizing the beneficial effects of neurotrophic factors. Indeed, most of the current therapies, which make use of the ability of neurotrophic factors to help neural function, are based on the exogenous delivery of these factors into the brain or spinal cord. In turn, the induction of neurotrophic factors by physiological means has the advantage of using the pharmacology that is intrinsic to the systems, thereby maximizing physiological effects. For example, the capacity of exercise to compensate for the decrease in hippocampal BDNF levels and poor cognitive function, caused by the consumption of a ‘‘bad diet,’’ can belong to a general neuroprotective mechanism. Strategies based on neurotrophic factor induction by exercise can lead to new treatments for maintaining a healthy brain and facilitating repair processes after trauma or disease.
REFERENCES 1. 2.
3.
4.
Huang EJ, Reichardt LF. Neurotrophins: roles in neuronal development and function. Annu Rev Neurosci 2001; 24:677–736. Zafra F, Hengerer B, Leibrock J, Thoenen H, Lindholm D. Activitydependent regulation of BDNF and NGF mRNAs in the rat hippocampus is mediated by non-NMDA glutamate receptors. EMBO J 1990; 9:3545–3550. Ernfors P, Bengzon J, Kokaia Z, Persson H, Lindvall O. Increased levels of messenger RNAs for neurotrophic factors in the brain during kindling epileptogenesis. Neuron 1991; 7:165–176. Kang H, Schuman EM. Long-lasting neurotrophin-induced enhancement of synaptic transmisssion in the adult hippocampus. Science 1996; 273: 1402–1406.
Behavioral Implications of Neurotrophic Actions 5. 6. 7.
8.
9. 10. 11. 12. 13. 14. 15. 16.
17.
18.
19.
20. 21. 22.
23.
119
Lohof AM, Ip NY, Poo MM. Potentiation of developing neuromuscular synapses by the neurotrophins NT-3 and BDNF. Nature 1993; 363:350–353. Barbacid, M. The Trk family of neurotrophin receptors. J Neurobiol 1994; 25:1386–1403. Ernfors P, Merlio JP, Persson H. Cells expressing mRNA for neurotrophins and their receptors during embryonic rat development. Eur J Neurosci 1992; 4:1140–1158. Friedman WJ, Black IB, Kaplan DR. Distribution of the neurotrophins brainderived neurotrophic factor, neurotrophin-3, and neurotrophin-4=5 in the postnatal rat brain: an immunocytochemical study. Neuroscience 1998; 84: 101–114. Barde YA. The nerve growth factor family. Prog Growth Factor Res 1990; 2:237–248. Meakin SO, Shooter EM. The nerve growth factor family of receptors. Trends Neurosci 1992; 15:323–331. Ebendal T. Function and evolution in the NGF family and its receptors. J Neurosci Res 1992; 32:461–470. Barde YA. Neurotrophins: a family of proteins supporting the survival of neurons. Prog Clin Biol Res 1994; 390:45–56. Poo, MM. Neurotrophins as synaptic modulators. Nature Rev Neurosci 2001; 2:24–32. Snider WD. Functions of the neurotrophins during nervous system development: what the knockouts are teaching us. Cell 1994; 77:627–638. McAllister AK, Katz LC, Lo DC. Neurotrophins and synaptic plasticity. Annu Rev Neurosci 1999; 22:295–318. Parakash N, Cohen-Cory S, Frostig RD. RAPID and opposite effects of BDNF and NGF on the functional organization of the adult cortex in vivo. Nature 1996; 381:702–706. Hartmann H, Heumann R, Lessmann V. Synaptic secretion of BDNF after high frequency stimulation of glutamatergic synapses. EMBO J 2001; 20:5887–5897. Aakalu G, Smith WB, Nguyen N, Jiang C, Schuman EM. Dynamic visualization of local protein synthesis in hippocampal neurons. Neuron 2001; 30: 489–502. Lom B, Cohen-Cory S. Brain-derived neurotrophic factor differentially regulates retinal ganglion cell dendritic and axonal arborization in vivo. J Neurosci 1999; 19:9928–9938. McAllister AK, Katz LC, Lo DC. Neurotrophin regulation of cortical dendritic growth requires activity. Neuron 1996; 17:1057–1064. Yacoubian TA, Lo DC. Truncated and full-length TrkB receptors regulate distinct modes of dendritic growth. Nat Neurosci 2000; 3:342–349. Shimada A, Mason CA, Morrison ME. TrkB signaling modulates spine density and morphology independent of dendrite structure in cultured neonatal Purkinje cells. J Neurosci 1998; 18:8559–8570. Boulanger L, Poo M. Gating of BDNF-induced synaptic potentiation by cAMP. Science 1999; 284:1982–1984.
120
Gomez-Pinilla and Vaynman
24. Kafitz KW, Rose CR, Thoenen H, Konnerth A. Neurotrophin-evoked rapid excitation through TrkB receptors. Nature 1999; 401:918–921. 25. Seil FJ, Drake-Baumann R. TrkB receptor ligands promote activitydependent inhibitory synaptogenesis. J Neurosci 2000; 20:5367–5373. 26. Vicario-Abejon C, Collin C, McKay RDG, Segal M. Neurotrophins induce formation of functional excitatory and inhibitory synapses between cultural hippocampalm neurons. J Neurosci 1998; 18:7256–7271. 27. Rutherford LC, Nelson SB, Turrigiano GG. BDNF has opposite effects on the quantal amplitude of pyramidal neuron and interneuron excitatory synapses. Neuron 1998; 21:521–530. 28. Causing CG, Gloster A, Aloyz R, Bamji SX, Chang E, Fawcett J, Kuchel G, Miller FD. Synaptic innervation density is regulated by neuron-derived BDNF. Neuron 1997; 18:257–267. 29. Alsina B, Vu T, Cohen-Cory S. Visualizing synapse formation in arborizing optic axons in vivo: dynamics and modulation by BDNF. Nat Neurosci 2001; 4:1093–1101. 30. Martinez A, Alcantara S, Borrell V, Del Rio JA, Blasi J, Otal R, Campos N, Boronat A, Barbacid M, Silos-Santiago I, Soriano E. TrkB and TrkC signaling are required for maturation and synaptogenesis of hippocampal connections. J Neurosci 1998; 19:7336–7350. 31. Collin C, et al. Neurotrophins act at presynaptic terminals activate synapses among cultured hippocampal neurons. Eur J Neurosci 2001; 13:1273–1282. 32. Blondel O, et al. A glial derived signal regulating neuronal differentiation, J Neurosci. 2000; 20:812–820. 33. Narisawa-Saito M, Carnahan J, Araki K, Yamaguchi T, Nawa H. Brainderived neurotrophic factor regulates the expression of AMPA receptor proteins in neocortical neurons. Neurosci 1999; 88:1009–1014. 34. Itami C, Mizuno K, Kohno T, Nakamura S. Brain-derived neurotrophic factor requirement for activity-dependent maturation of glutamatergic synapse in developing mouse somatosensory cortex. Brain Res 2000; 857: 141–150. 35. Ventimiglia R, Mather PE, Jones BE, Lindsay RM. The neurotrophins BDNF, NT-3 and NT-4=5 promote survival and morphological and biochemical differentiation of striatal neurons in vitro. Eur J Neurosci 1995; 7: 213–222. 36. Marty S, Carroll P P, Cellerino A, Castren E, Staiger V, Thoenen H, Lindholm D. Brain-derived neurotrophic factor promotes the differentiation of various hippocampal nonpyramidal neurons, including Cajal-Retzius cells, in organotypic slice cultures. J Neurosci 1996; 16:675–687. 37. Cellerino A, Pinzon-Duarte G, Carroll P, Kohler K. Brain-derived neurotrophic factor modulates the development of the dopaminergic network in the rodent retina. J Neurosci 1998; 18:3351–3362. 38. Marty S, Wehrle R, Sotelo C. Neuronal activity and brain-derived neurotrophic factor regulate the density of inhibitory synapses in organotypic slice cultures of postnatal hippocampus. J Neurosci 2000; 20:8087–8095.
Behavioral Implications of Neurotrophic Actions
121
39. Vicario-Abejon C, Collin C, Tsoulfas P, McKay RD. Hippocampal stem cells differentiate into excitatory and inhibitory neurons. Eur J Neurosci 2000; 12:677–688. 40. Yamada MK, Nakanishi K, Ohba S, Nakamura T, Ikegaya Y, Nishiyama N, Matsuki N. Brain-derived neurotrophic factor promotes the maturation of GABAergic mechanisms in cultured hippocampal neurons. J Neurosci 2002; 22:7580–7585. 41. Bolton MM, Lo DC, Sherwood NT. Long-term regulation of excitatory and inhibitory synaptic transmission in hippocampal cultures by brain-derived neurotrophic factor. Prog Brain Res 2000; 128:203–218. 42. Vicario-Abejon C, Owens D, McKay R, Segal M. Role of neurotrophins in central synapse formation and stabilization. Nat Rev Neurosci 2002; 3: 965–974. 43. Cabelli RJ, Hohn A, Shatz CJ. Inhibition of ocular dominance column formation by infusion of NT-4=5 or BDNF. Science 1995; 267:1662–1666. 44. Cabelli RJ, Shelton DL, Segal RA, Shatz CJ. Blockade of endogenous ligands of TrkB inhibits formation of ocular dominance columns. Neuron 1997; 19:63–76. 45. Katz LC, Shatz CJ. Synaptic activity and the construction of cortical circuits. Science 1996; 274:1133–1138. 46. Tropea D, Caleo M, Maffei L. Synergistic effects of brain-derived neurotrophic factor and chondroitinase ABC on retinal giber sprouting after denervation of the superior colliculus in adult rats. J Neurosci 2003; 23:7034–7044. 47. Gall CM, Isackson PJ. Limbic seizures increase neuronal production of messenger RNA for nerve growth factor. Science 1989; 245:758–761. 48. Patterson SL, Grover LM, Schwartzkroin PA, Bothwell M. Neurotrophin expression in rat hippocampal slices: a stimulus paradigm inducing LTP in CA1 evokes increases in BDNF and NT-3 mRNAs. Neuron 1992; 9: 1081–1088. 49. Castren E, Pitkanen M, Sirvio J, Parsadanian A, Lindholm D, Thoenen H, Riekkinen PJ. The induction of LTP increases BDNF and NGF mRNA but decreases NT-3 mRNA in the dentate gyrus. Neuroreport 1993; 4: 895–898. 50. Castren E, Zafra F, Thoenen H, Lindholm D. Light regulates expression of brain-derived neurotrophic factor mRNA in rat visual cortex. Proc Natl Acad Sci USA 1992; 89:9444–9448. 51. Rocamora N, Pascual M, Acsady L, de Lecea L, Freund TF, Soriano E. Expression of NGF and NT3 mRNAs in hippocampal interneurons innervated by the GABAergic septohippocampal pathway. J Neurosci 1996; 16:3991–4004. 52. Neeper SA, Gomez-Pinilla F, Choi J, Cotman CW. Physical activity increases mRNA for brain-derived neurotrophic factor and nerve growth factor in rat brain. Brain Res 1996; 726:49–56. 53. Vaynman S, Ying Z, Gomez-Pinilla F. Interplay between brain-derived neurotrophic factor and signal transduction modulators in the regulation of the effects of exercise on synaptic-plasticity. Neuroscience. 2003; 122:647–657.
122
Gomez-Pinilla and Vaynman
54. Kesslak JP, So V, Choi J, Cotman CW, Gomez-Pinilla F. Learning upregulates brain-derived neurotrophic factor messenger ribonucleic acid: a mechanism to facilitate encoding of circuit maintenance. Beh Neurosci 1998; 112:1012–1019. 55. Bova R, Micheli MR, Qualadrucci P, Zucconi GG. BDNF and TrkB mRNAs oscillate in rat brain during the light-dark cycle. Brain Res Mol Brain Res 1998; 57:321–324. 56. Liang FQ, Walline R, Earnest DJ. Circadian rhythm of brain-derived neurotrophic factor in the rat suprachiasmatic nucleus. Neurosci Lett 1998; 242: 89–92. 57. Snider WD, Lichtman JW. Are neurotrophins synaptotrophins? Mol Cell Neurosci 1996; 7:433–442 58. Crino PB, Eberwine J. Molecular characterization of the dendritic growth cone: regulated mRNA transport and local protein synthesis. Neuron 1996; 17:1173–1187. 59. Tongiorgi E, Righi M, Cattaneo A. Activity-dependent dendritic targeting of BDNF and TrkB mRNAs in hippocampal neurons. J Neurosci 1997; 17: 9492–9505. 60. Steward O, Schuman EM. Protein synthesis at synaptic sites on dendrites. Annu Rev Neurosci 2001; 24:299–325. 61. Righi M, Tongiorgi E, Cattaneo A. Brain-derived neurotrophic factor (BDNF) induces dendritic targeting of BDNF and tyrosine kinase B mRNAs in hippocampal neurons through a phosphatidylinositol-3 kinase-dependent pathway. J Neurosci 2000; 20:3165–3174. 62. Glombik MM, Gerdes HH. Signal-mediated sorting of neuropeptides and prohormones: secretory granule biogenesis revisited. Biochemistry 2000; 82:315–326. 63. Rouille Y, Duguay SJ, Lund K, Furuta M, Gong Q, Lipkind G, Oliva AA Jr, Chan SJ, Steiner DF. Proteolytic processing mechanisms in the biosynthesis of neuroendocrine peptides: the subtilisin-like proprotein convertases. Front Neuroendocrinol 1995; 16:322–361. 64. Seidah NG, Benjannet S, Pareek S, Savaria D, Hamelin J, Goulet B, Laliberte J, Lazure C, Chretien M, Murphy RA. Cellular processing of the nerve growth factor precursor by the mammalian pro-protein convertases. Biochem J 1996; 314:951–960. 65. Mowla SJ, Pareek S, Farhadi HF, Petrecca K, Fawcett JP, Seidah NG, Morris SJ, Sossin WS, Murphy RA. Differential sorting of nerve growth factor and brain-derived neurotrophic factor in hippocampal neurons. J Neurosci 1999; 19:2069–2080. 66. Blochl A, Thoenen H. Localization of cellular storage compartments and sites of constitutive and activity-dependent release of nerve growth factor (NGF) in primary cultures of hippocampal neurons. Mol Cell Neurosci 1996; 7:173–190. 67. Goodman LJ, Valverde J, Lim F, Geschwind MD, Federoff HJ, Geller AI, Hefti F. Regulated release and polarized localization of brain-derived neurotrophic factor in hippocampal neurons. Mol Cell Neurosci 1996; 7:222–238. 68. Lee R, Kermani P, Teng KK, Hempstead BL. Regulation of cell survival by secreted proneurotrophins. Science 2001; 294:1945–1948.
Behavioral Implications of Neurotrophic Actions
123
69. Chen WP, Chang YC, Hsieh ST. Trophic interactions between sensory nerves and their targets. J Biomed Sci 1999; 6:79–85. 70. Farhadi HF, Mowla SJ, Petrecca K, Morris SJ, Seidah NG, Murphy RA. Neurotrophin-3 sorts to the constitutive secretory pathway of hippocampal neurons and is diverted to the regulated secretory pathway by coexpression with brain-derived neurotrophic factor. J Neurosci 2000; 20:4059–4068. 71. Haubensak W, Narz F, Heumann R, Lessmann V. BDNF-GFP containing secretory granules are localized in the vicinity of synaptic junctions of cultured cortical neurons. J Cell Sci 1998; 111:1483–1493. 72. Kojima M, Takei N, Numakawa T, Ishikawa Y, Suzuki S, Matsumoto T, Katoh-Semba R, Nawa H, Hatanaka H. Biological characterization and optical imaging of brain-derived neurotrophic factor-green fluorescent protein suggest an activity-dependent local release of brain-derived neurotrophic factor in neurites of cultured hippocampal neurons. J Neurosci Res 2001; 64: 1–10. 73. Lindvall O, Kokaia Z, Bengzon J, Elmer E, Kokaia M. Neurotrophins and brain insults. Trends Neurosci 1994; 17:490–496. 74. Lessmann V, Heumann R. Modulation of unitary glutamatergic synapses by neurotrophin-4=5 or brain-derived neurotrophic factor in hippocampal microcultures: presynaptic enhancement depends on pre-established paired-pulse facilitation. Neurosci 1998; 86:399–413. 75. Messaoudi E, Bardsen K, Srebro B, Bramham CR. Acute intrahippocampal infusion of BDNF induces lasting potentiation of synaptic transmission in the rat dentate gyrus. J Neurophysiol 1998; 79:496–499. 76. Tanaka T, Saito H, Matsuki N. Inhibition of GABAA synaptic responses by brain-derived neurotrophic factor (BDNF) in rat hippocampus. J Neurosci 1997; 17:2959–2966. 77. Frerking M, Malenka RC, Nicoll RA. Brain-derived neurotrophic factor (BDNF) modulates inhibitory, but not excitatory, transmission in the CA1 region of the hippocampus. J Neurophysiol 1998; 80:3383–3386. 78. Wang XH, Poo MM. Potentiation of developing synapses by postsynaptic release of neurotrophin-4. Neuron 1997; 19:825–835. 79. Nick TA, Ribera AB. Synaptic activity modulates presynaptic excitability. Nature 2000; 3:142–149. 80. Kohara K, Kitamura A, Morishima M, Tsumoto T. Activity-dependent transfer of brain-derived neurotrophic factor to postsynaptic neurons. Science 2001; 291:2419–2423. 81. Levine ES, Dreyfus CF, Black IB, Plummer MR. Differential effects of NGF and BDNF on voltage-gated calcium currents in embryonic basal forebrain neurons. J Neurosci 1995; 15:3084–3091. 82. Conner JM, Lauterborn JC, Yan Q, Gall CM, Varon S. Distribution of brainderived neurotrophic factor (BDNF) protein and mRNA in the normal adult rat CNS: evidence for anterograde axonal transport. J Neurosci 1997; 17:2295–2313. 83. Yan Q, Rosenfeld RD, Matheson CR, Hawkins N, Lopez OT, Bennett L, Welcher AA. Expression of brain-derived neurotrophic factor protein in the adult rat central nervous system. Neuroscience 1997; 78:431–448.
124
Gomez-Pinilla and Vaynman
84. Smith MA, Makino S, Kvetnansky R, Post RM. Stress glucocorticoids affect the expression of brain-derived neurotrophic factor and neurotrophin-3 mRNAs in the hippocampus. J Neurosci 1995; 15:1768–1777. 85. Binder DK, Routbort MJ, McNamara JO. Immunohistochemical evidence of seizure-induced activation of Trk receptors in the mossy fiber pathway of adult rat hippocampus. J Neurosci 1999; 19:4616–4626. 86. Meyer-Franke A, Wilkinson GA, Kruttgen A, Hu M, Munro E, Hanson MG Jr, Reichardt LF, Barres BA. Depolarization and cAMP elevation rapidly recruit TrkB to the plasma membrane of CNS neurons. Neuron 1998; 21: 681–693. 87. Falkenberg T, Mohammed AK, Henriksson B, Persson H, Winblad B, Lindefors N. Increased expression of brain-derived neurotrophic factor mRNA in rat hippocampus is associated with improved spatial memory and enriched environment. Neurosci Lett 1992; 138:153–156. 88. Patterson SL, Pittenger C, Morozov A, Martin KC, Scanlin H, Drake C, Kandel ER. Some forms of cAMP-mediated long-lasting potentiation are associated with release of BDNF and nuclear translocation of phosphoMAP kinase. Neuron 2001; 11:123–140. 89. Egan MF, Kojima M, Callicott JH, Goldberg TE, Kolachana BS, Bertolino A, Zaitsev E, Gold B, Goldman D, Dean M, Lu B, Weinberger DR. The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell 2003; 112:257–269. 90. Molteni R, Barnard RJ, Ying Z, Roberts CK, Gomez-Pinilla F. A high-fat refined sugar diet reduces hippocampal brain-derived neurotrophic factor neuronal plasticity, and learning. Neuroscience 2002a; 112:803–814. 91. Hall J, Thomas KL, Everitt BJ. Rapid and selective induction of BDNF expression in the hippocampus during contextual learning. Nat Neurosci 2000; 3:533–535. 92. Mu J, Li W, Yao Z, Zhou X. Deprivation of endogenous brain-derived neurotrophic factor results in impairment of spatial learning and memory in adult rats. Brain Res 1999; 835:259–265. 93. Ma YL, Wang HL, Wu HC, Wei CL, Lee EH. Brain-derived neurotrophic factor antisense oligonucleotide impairs memory retention and inhibits longterm potentiation in rats. Neuroscience 1998; 82:957–967. 94. Mizuno M, Yamada K, Olariu A, Nawa H, Nabeshima T. Involvement of brain-derived neurotrophic factor in spatial memory formation and maintenance in a radial arm maze test in rats. J Neurosci 2000; 20:7116–7121. 95. Linnarsson S, Bjorklund A, Ernfors P. Learning deficit in BDNF mutant mice. Eur J Neurosci 1997; 9:2581–2587. 96. Minichiello L, Korte M, Wolfer D, Kuhn R, Unsicker K, Cestari V, RossiArnaud C, Lipp HP, Bonhoeffer T, Klein R. Essential role for TrkB receptors in hippocampus-mediated learning. Neuron 1999; 24:401–414. 97. Saarelainen T, Pussinen R, Koponen E, Alhonen L, Wong G, Sirvio J, Castren E. Transgenic mice overexpressing truncated TrkB neurotrophin receptors in neurons have impaired long-term spatial memory but normal hippocampal LTP. Synapse 2000; 38:102–104.
Behavioral Implications of Neurotrophic Actions
125
98. Patterson SL, Abel T, Deuel TAS, Martin KC, Rose JC, Kandel ER. Recombinant BDNF rescues deficits in basal synaptic transmission and hippocampal LTP in BDNF knockout mice. Neuron 1996; 16:1137–1145. 99. Korte M, Carroll P, Wolf E, Brem G, Thoenen H, Bonhoeffer T. Hippocampal long-term potentiation is impaired in mice lacking brain-derived neurotrophic factor. Proc Natl Acad Sci USA 1995; 92:8856–8860. 100. Bannerman DM, Good MA, Butcher SP, Ramsay M, Morris RG. Distinct components of spatial learning revealed by prior training and NMDA receptor blockade. Nature 1995; 378:182–186. 101. Huang YY, Kandel ER, Varshavsky L, Brandon EP, Qi M, Idzerda RL, McKnight GS, Bourtchouladze R. A genetic test of the effects of mutations in PKA on mossy fiber LTP and its relation to spatial and contextual learning. Cell 1995; 83:1211–1222. 102. Okabe S, Collin C, Auerbach JM, Meiri N, Bengzon J, Kennedy MB, Segal M, McKay RD. Hippocampal synaptic plasticity in mice overexpressing an embryonic subunit of the NMDA receptor. J Neurosci 1998; 18:4177–4188. 103. van Praag H, Christie BR, Sejnowski TJ, Gage FH. Running enhances neurogenesis, learning, and long-term potentiation in mice. PNAS 1999; 96: 13427–13434. 104. Johnston Rose SPR. Memory consolidation in day-old chicks requires BDNF but not NGF or NT-3: an anti-sense study. Mol Brain Res 2001; 88:26–36. 105. Kaisho Y, Ohta H, Miyamato M, Igarashi K. Nerve growth factor promotor driven neurotrophin-3 overexpression in the mouse and the neuroprotective effects of transgene on age related behavioral deficits. Neurosci Lett 1999; 277:181–184. 106. Sugaya K, Greene R, Personett D, Robbins M, Kent C, Bryan D, Skiba E, Gallagher M, McKinney M. Septo-hippocampal cholinergic and neurotrophin markers in age-induced cognitive decline. Neurobiol Aging 1998; 19:351–361. 107. Segal RA, Greenberg ME. Intracellular signaling pathways activated by neurotrophic factors. Annu Rev Neurosci 1996; 19:463–489. 108. Kaplan DR, Stephens RM. Neurotrophin signal transduction by the Trk receptor. J Neurobiol 1994; 25:1404–1417. 109. Stephens L, Smrcka A, Cooke FT, Jackson TR, Sternweis PC, Hawkins PT. A novel phosphoinositide 3 kinase activity in myeloid-derived cells is activated by G protein beta gamma subunits. Cell 1994; 77:83–93. 110. Obermeier A, Bradshaw RA, Seedorf K, Choidas A, Schlessinger J, Ullrich A. Neuronal differentiation signals are controlled by nerve growth factor receptor=Trk binding sites for SHC and PLC gamma. EMBO J 1994; 13:1585–1590. 111. Sweatt JD. The neuronal MAP kinase cascade: a biochemical signal integration system subserving synaptic plasticity and memory. J Neurochem 2001; 76:1–10. 112. Corbit KC, Foster DA, Rosner MR. Protein kinase C-d mediates neurogenic but not mitogenic activation of mitogen-activated protein kinase in neuronal cells. Mol Cell Biol 1999; 19:4209–4218. 113. Finkbeiner S, Tavazoie SF, Maloratsky A, Jacobs KM, Harris KM, Greenberg ME. CREB: a major mediator of neuronal neurotrophin responses. Neuron 1997; 19:1031–1047.
126
Gomez-Pinilla and Vaynman
114. Ying S, Futter M, Rosenblum K, Webber MJ, Hunt SP, Bliss TV, Bramham CR. Brain-derived neurotrophic factor induces long-term potentiation in intact adult hippocampus: requirement for ERK activation coupled to CREB and upregulation of Arc synthesis. J Neurosci 2002; 22:1532–1540. 115. Matsubra M, Kusubata M, Ishiguro K, Uchida T, Titani K, Taniguchi H. Site-specific phosphorylation of synapsin I by mitogen-activated protein kinase and Cdk5 and its effects on physiological functions. J Biol Chem 1996; 271:21108–21113. 116. Finkbeiner S. CREB couples neurotrophin signals to survival messages. Neuron 2000; 25:11–14. 117. Yin JC, Tully T. CREB and the formation of long-term memory. Curr Opin Neurobiol 1996; 6:264–268. 118. Tao X, Finkbeiner S, Arnold DB, Shaywitz A, Greenberg ME. Ca2þ influx regulates BDNF expression by a CREB dependent mechanism. Neuron 1998; 20:709–726. 119. Tully T. Regulation of gene expression and its role in long-term memory and synaptic-plasticity. Proc Natl Acad Sci USA 1997; 94:4239–4241. 120. Jovanovic JN, Czernik AJ, Fienberg AA, Greengard P, Sihra TS. Synapsins as mediators of BDNF-enhanced neurotransmitter release. Nat Neurosci 2000; 3:323–329. 121. Pozzo-Miller LD, Gottschalk W, Zhang L, McDermott K, Du J, Gopalakrishnan R, Oho C, Sheng ZH, Lu B. Impairments in high-frequency transmission, synaptic vesicle docking, and synaptic protein distribution in the hippocampus of BDNF knockout mice. J Neurosci 1999; 19:4972–4983. 122. Melloni RH Jr, Apostolides PJ, Hamos JE, DeGennaro LJ. Dynamics of synapsin I gene expression during the establishment and restoration of functional synapses in the rat hippocampus. Neurosci 1994; 58:683–703. 123. Takei Y, Harada A, Takeda S, Kobayashi K, Terada S, Noda T, Taka Hirokawa N. Synapsin I deficiency results in the structural change in the presynaptic terminals in the murine nervous system. J Cell Biol 1995; 131: 1789–1800. 124. Akagi S, Mizoguchi A, Sobue K, Nakamura H, Ide C. Localization of synapsin I in normal fibers and regenerating axonal sprouts of the rat sciatic nerve. Histochem Cell Boil 1996; 105:365–373. 125. Ghosh A, Ginty DD, Bading H, Greenberg ME. Calcium regulation of gene expression in neuronal cells. J Neurobiol 1994; 25:294–303. 126. Platenik J, Kuramoto N, Yoneda Y. Molecular mechanisms associated with long-term consolidation of the NMDA signals. Life Sci 2000; 67:335–364. 127. Bading H, Greenberg ME. Stimulation of protein tyrosine phosphorylation by NMDA receptor activation. Science 1991; 253:912–914. 128. Cammarota M, de Stein ML, Paratcha G, Bevilaqua LR, Izquierdo I, Medina JH. Rapid and transient learning-associated increase in NMDA NR1 subunit in the rat hippocampus. Neurochem Res 2000; 25:567–572. 129. Song D, Choe B, Bae JH, Park WK, Han IS, Ho W, Earm YE. Brain-derived neurotrophic factor rapidly potentiates synaptic transmission through NMDA, but suppresses it through non-NMDA receptors in rat hippocampal neuron. Brain Res 1998; 799:176–179.
Behavioral Implications of Neurotrophic Actions 130. 131.
132.
133. 134.
135.
136.
137.
138. 139.
140.
141.
142.
143.
144.
145.
127
Bliss TVP, Collingridge GL. A synaptic model of memory: long-term potentiation in the hippocampus. Nature 1973; 361:31–39. Hollmann M, Boulter J, Maron C, Heinemann S. Molecular biology of glutamate receptors: potentiation of N-methyl-d-aspartate receptor splice variants by zinc. Ren Physiol Biochem 1994; 17:182–183. Abeliovich A, Paylor R, Chen C, Kim JJ, Wehner JM, Tonegawa S. PKCg mutant mice exhibit mild deficits in spatial and contexual learning. Cell 1993; 75:1263–1271. Silva AJ, Paylor R, Wehnet JM, Tonegawa S. Impaired spatial learning in acalcium-calmodulin kinase II mutant mice. Science 1992; 257:206–211. Adams J, Roberson ED, English JD, Selcher JC, Sweatt JD. MAPK regulation of gene expression in the central nervous system. Acta Neurobiol Exp 2000; 60:377–394. English JD, Sweatt JD. A requirement for the mitogen-activated protein kinase cascade in hippocampal long term potentiation. J Biol Chem 1997; 272:19103–19106. Blum S, Moore AN, Adams F, Dash PK. A mitogen-activated protein kinase cascade in the CA1=CA2 subfield of the dorsal hippocampus is essential for long-term spatial memory. J Neurosci 1999; 19:3535–3544. Hardingham G, Arnold FJ, Bading H. Nuclear calcium signaling controls CREB-mediated gene expression triggered by synaptic activity. Nat Neurosci 2001; 4:261–267. Silva AJ, Kogan JH, Frankland PW, Kida S. CREB and memory. Annu Rev Neurosci 1998; 21:127–148. Kramer AF, Hahn S, Cohen NJ, Banich MT, McAuley E, Harrison CR, Chason J, Vakil E, Bardell L, Boileau RA, Colombe A. Aging, fitness, and neurocognitive function. Nature 1999; 400:418–419. Suominen-Troyer S, Davis KJ, Ismail AH, Salvendy G. Impact of physical fitness on strategy development in decision-making tasks. Percept Mot Skills 1986; 62:71–77. Rogers RL, Meyer JS, Mortel KF. After reaching retirement age physical activity sustains cerebral perfusion and cognition. J Am Geriatr Soc 1990; 38:123–128. Fordyce DE, Farrar RP. Enhancement of spatial learning in F344 rats by physical activity and related learning-associated alterations in hippocampal and cortical cholinergic functioning. Behav Brain Res 1991; 46:123–133. Samorajski T, Delaney C, Durham L, Ordy JM, Johnson JA, Dunlap WP. Effect of exercise on longevity, body weight, locomotor performance, and passive-avoidance memory of C57BL=6J mice. Neurobiol Aging 1985; 6:17–24. Russo-Neustadt AA, Beard RC, Huang YM, Cotman CW. Physical activity and antidepressant treatment potentiate the expression of specific brainderived neurotrophic factor transcripts in the rat hippocampus. Neurosci. 2000; 101:305–312. Molteni R, Ying Z, Gomez-Pinilla F. Differential effects of acute chronic exercise on plasticity-related genes in the rat hippocampus revealed by microarray. Eur J Neurosci 2002; 16:1107–1116.
128
Gomez-Pinilla and Vaynman
146. Gadea A, Lopez-Colome’ AM. Glial transporters for glutamate, glycine and GABA 1. J Neurosci Res 2001; 63:453–460. 147. Rossi DJ, Oshima T, Attwell D. Glutamate release in severe brain ischaemia is mainly by reversed uptake. Nature 2000; 403:316–321. 148. Raineteau O, Schwab ME. Plasticity of motor systems after incomplete spinal cod injury. Nature Rev Neurosci 2001; 2:263–273. 149. Gage FH. Mammalian neural stem cells. Science 2000; 287:1433–1438. 150. Gould E, Tanapat P, Rydel T, Hastings N. Regulation of hippocampal neurogenesis in adulthood. Biol Psychiatry 2000; 48:715–720. 151. Shors TJ, Miesegaes G, Beylin A, Zhao M, Rydel T, Gould E. Neurogenesis in the adult is involved in the formation of trace memories. Nature 2001; 410:372–376. 152. Nibuya M, Morinobu S, Duman RS. Regulation of BDNF TrkB mRNA by chronic anticonvulsive s seizure antidepressant drug treatments. J Neurosci 1995; 15:7539–7547. 153. Siuciak JA, Boylan C, Fritsche M, Altar CA, Lindsay RM. BDNF increases monoamine activity in the rat brain following intracerebroventricular or intraparenchymal administration. Brain Res 1996; 710:11–20. 154. Shirayama Y, Chen ACH, Nakagawa S, Russell DS, Duman RS. Brainderived neurotrophic factor produces anti-depressant effects in behavioral models of depression. J Neurosci 2002; 22:3251–3261. 155. Chen B, Dowlatshahi D, MacQueen GM, Wang J, Young LT. Increased hippocampal BDNF immunoreactivity in subjects treated with antidepressant medication. Bio Psychiatry 2001; 50:260–265. 156. Karege F, Perret G, Bondolfi G, Schwald M, Bertschy G, Aubry G. Decreased serum brain-derived neuortrophic factor levels in major depressed patients. Psychiatry Res 2002; 109:143–148. 157. Smith MA, Makino S, Kvetnansky R, Post RM. Stress and glucocorticoids affect the expression of brain-derived neurotrophic factor and neurotrophin3 mRNAs in the hippocampus. J Neurosci 1995; 15:1768–1777. 158. Goggi J, Pullar IA, Carney SL, Bradford HFZ. Modulation of neurotransmitter release induced by brain-derived neurotrophic factor in rat brain striatal slices in vitro. Brain Res 2002; 941:34–42. 159. Sklair-Tyron L, Nestler EJ. Opposing effects of morphine and the neurotrophins, NT-3, NT-4, and BDNF, on locus coeruleus neurons in vitro. Brain Res 1995; 702:117–125. 160. Mamounas LA, Blue ME, Siuciak JA, Altar CA. Brain-derived neurotrophic factor promotes the survival and sprouting of serotonergic axons in rat brain. J Neurosci 1995; 15:7929–7939. 161. Russo-Neustadt AA, Beard RC, Cotman CW. Exercise antidepressant medications enhance BRAIN derived neurotrophic factor. Neuropharmacol 1999; 21:679–682. 162. Duman RS, Malberg J, Nakagawa S, D’Sa C. Neuronal plasticity and survival in mood disorders. Biol Psychiatry 2000; 48:732–739. 163. Thome J, Sakai N, Shin K, Steffan C, Zhang YJ, Impey S, et al. cAMP reponse element-mediated gene transcription is upregulated by chronic antidepressant treatment. J Neurosci 2000; 20:4030–4036.
Behavioral Implications of Neurotrophic Actions 164.
165. 166.
167. 168. 169.
170. 171.
172.
173.
174.
175.
176.
177.
178.
179.
129
Chen A-H, Shirayama Y, Skin KH, Neve RL, Duman RS. Expression of the cAMP response element binding protein (CREB) in hippocampus produces antidepressant effect. Biol Psychiatry 2001; 49:753–762. Pham TM, Winblad B, Granholm A, Mohammed AH. Environmental influences of brain neurotrophins in rats. Pharm Biochem Beh 2002; 73:167–175. Nilsson M, Perifilieva E, Johansson U, Orwar O, Eriksson PS. Enriched environment increases neurogenesis in the adult rat dentate gyrus and improves spatial memory. J Neurobiol 1999; 39:569–578. Kempermann G, Kuhn HG, Gage FH. Experience-induced neurogenesis in the senescent dentate gyrus. J Neurosci 1998; 18:3206–3212. Janus C, Copervas JS, Janus M, Roder J. Rearing environment and radial maze exploration in mice. Behav Process 1995; 34:129–140. Kuhn HG, Dickinson-Anson H, Gage FH. Neurogenesis in the dentate gyrus of the adult rat: age related decrease in neuronal progenitor proliferation. J Neurosci 1996; 16:2027–2033. Cameron HA, McKay RD. Restoring production of hippocampal neurons in old age. Nat Neurosci 1999; 2:894–897. Kempermann G, Gast D, Gage FH. Neuroplasticity in old age: sustained fivefold induction of hippocampal neurogenesis by long-term environment enrichment. Ann Neurol 2002; 52:135–143. Molteni R, Wu A, Vaynman S, Ying Z, Barnard RJ, Gomez-Pinilla F. Exercise reverses the harmful effects of consumption of high-fat diet on synaptic and behavioral plasticity associated to the action of brain-derived neurotrophic factor. Neuroscience. 2004; 123:429–440. Wu A, Molteni R, Ying Z, Gomez-Pinilla F. A saturated-fat diet aggravates the outcome of traumatic brain injury on hippocampal plasticity and cognitive function by reducing brain-derived neurotrophic factor. Neuroscience 2003; 119:365–375. Nakazato M, Hashimoto K, Shimizu E, Kumakiri C, Koizumi H, Okamura N, Mitsumori M, Komatsu N, Iyo M. Decreased levels of serum brain-derived neurotrophic factor in female patients with eating disorders. Biol Psychiatry 2003; 54:485–490. Pan W, Banks WA, Fasold MB, Bluth J, Kastin AJ. Transport of brain–derived neurotrophic factor across the blood–brain barrier. Neuropharmacology 1998; 37:1553–1561. Gomez-Pinilla F, Ying Z, Roy RR, Molteni R, Edgerton R. Voluntary exercise induces a BDNF-mediated mechanism that promotes neuroplasticity. J Neurophysiol 2002; 88:2187–2195. De Leon RD, Hodgson JA, Roy RR, Edgerton VR. Locomotor capacity attributable to step training vs. spontaneous recovery following spinalization in adult cats. J Neurophysiol 1998; 79:1329–1340. Gomez-Pinilla F, Ying Z, Opazo P, Roy RR, Edgerton R. Differential regulation of BDNF and NT-3 in the spinal cord and skeletal muscle by exercise in the rat. Eur J Neurosci 2001; 13:1–8. Koliatsos VE, Clatterbuck RE, Winslow JW, Cayouetter MH, Price DL. Evidence that brain-derived neurotrophic factor is a trophic factor for mote neurons in vivo. Neuron 1993; 10:359–367.
130
Gomez-Pinilla and Vaynman
180. Xu XM, Gue´nard V, Kleitman N, Aebischer P, Bunge MB. A combination of BDNF and NT-3 promotes supraspinal axonal regeneration in Schwann cell grafts in adult rat thoracic spinal cord. Exp Neurol 1995; 134:261–272. 181. Jakeman LB, Wei P, Guan Z, Strokes BT. Brain-derived neurotrophic factor stimulates hindlimb stepping and sprouting of cholinergic fibers after spinal cord injury. Exp Neurol 1998; 154:170–184. 182. Bregman BS, McAtee M, Dai HN, Kuhn PL. Neurotrophic factors increase axonal growth after spinal cord injury and transplantation in the adult rat. Exp Neurol 1997; 148:475–494. 183. Molteni R, Zheng JQ, Ying Z, Gomez-Pinilla F, Twiss JL. Voluntary exercise increases axonal regeneration from sensory neurons. Proc Natl Acad Sci USA. 2004 Jun 1; 101:8473–8478.
7 Estrogen Therapy for Prevention of Alzheimer’s Disease But Not for Rehabilitation Following Onset of Disease: The Healthy Cell Bias of Estrogen Action Roberta Diaz Brinton Department of Molecular Pharmacology and Toxicology and Program in Neuroscience, Pharmaceutical Sciences Center, University of Southern California, Los Angeles, California, U.S.A.
1.
INTRODUCTION
Results of recent clinical studies have challenged our previously held view that estrogen therapy (ET) promotes neurological health and prevents or ameliorates Alzheimer’s disease (AD). A major question emerging from these studies is how can there be such disparity between the basic science and epidemiological data which show that estrogen can protect neurons against degenerative insults and reduce the risk of AD and the recent data from the Women’s Health Initiative Memory Study (WHIMS) trial and the trial of estrogen treatment for AD which show that hormone therapy (HT) in the WHIMS and estrogen treatment in women with existing AD showed no benefit and even a potential deleterious effect of the interventions? Which set of data is correct? The proposition put forth in this review is that both sets of data are correct and that two major factors determine the efficacy of ET or HT. First is the time at which ET is initiated. The data indicate that initiation of therapy early in menopause, in a healthy 131
132
Brinton
neuron state, reduces the risk of AD, whereas ET initiated after the disease has developed or decades following menopause, is without benefit. Second, ET is not the same as HT and the type of progestin used determines the outcome of the therapeutic intervention. Insights into estrogen and progestin mechanisms of action in the brain provide a framework for understanding the paradox of estrogen benefit in prevention of AD vs. the lack of benefit in treatment trials and in trials when HT is instituted years following the menopause. On the basis of estrogen-inducible mechanisms, which have been elucidated in healthy neuron model systems, it would be predicted that ET could be highly efficacious in preventing neurodegenerative diseases by promoting neuronal defense and memory mechanisms. The mechanisms of estrogen action also predict that ET would be an ineffective strategy for reversing the pathology of AD. Factors affecting the efficacy of ET, such as duration of estrogen deficiency, age, the hormonal status when ET or HT is begun, and the type of progestin used in hormone combination therapy, are considered. The issue of when to intervene with ET or HT, whether during perimenopause or following menopause, is also discussed. Finally, an estrogen-advantage hypothesis is put forth, which provides a unifying mechanism of estrogen action with implications for both the benefits and the risks of ET.
2.
ALZHEIMER’S DISEASE: A POTENTIAL HEALTH CRISIS
Of the age-associated maladies, neurodegenerative diseases are among the most devastating and costly. They typically have a prolonged gestational period until frank symptomatology develops, followed by an equally slow rate of debilitation (1). Of the age-associated neurodegenerative diseases, AD is perhaps the most devastating for both the victim and the family (2). Age, which is the most frequent cause of dementia and the leading cause of a loss in independent living and institutionalization, remains the greatest risk factor for developing AD (1,3). The care and treatment of persons afflicted with AD result in health care costs of about $100 billion dollars per year in the United States alone (1). Conservative projections predict that the prevalence of AD will nearly quadruple in the next 50 years, by which time 1 in 45 Americans will be afflicted with the disease (4). It is estimated that there are 360,000 new cases of AD each year, with a range of 200,000–600,000 cases per year. The annual number of new cases could be expected to rise more than 3-fold, from 360,000 cases in 1997 to 1.14 million new cases per year in 2047. Despite high profile cases such as Ronald Reagan and Charlton Heston, women are by far the principal victims of AD (5). In 1997, the prevalence of AD was 2.32 million (with a range of 1.09–4.58 million); of those affected, 68% were female and 32% were male (4). Because women
Estrogen Therapy for Prevention of Alzheimer’s Disease
133
have a longer life expectancy than men, it would be anticipated that the absolute number of women with AD would exceed that of men. However, a double danger exists for women. Several studies indicate that at any given age, women exhibit a higher incidence of AD than their age-matched male counterparts (6), and that a 1.6:1.0 female to male ratio in the incidence of AD persists, even when controlling for greater female longevity (7). Results of this study confirmed a striking gender bias for the development of AD; in that, women had a 2.5-fold higher incidence of AD than their age-matched male counterparts. However, studies done by Lindsay et al. (8) indicate that this ostensible gender bias may be related to other factors, such as age and low educational level. Interestingly, in this study, when the odds ratio was adjusted for both age and education, a gender bias for the incidence of AD was not apparent (8). Worldwide, there are currently more than 470 million women aged >50 years, and 30% of those are projected to live into their 80s (9). At the turn of the new millennium in the United States, there were nearly 42 million women over the age of 50 years, and of these, more than 31 million women were over the age of 55 years (9). By the year 2020, it is estimated that there will be nearly 50 million women over the age of 55 years in the United States. Currently, a woman’s average life expectancy is estimated at 79.7 years. However, a woman who reaches the age of 54 years can expect to live to the age of 84.3 years. Nearly two-thirds of the total U.S. population will survive to 85 years of age. On the basis of these data, women can anticipate spending one-third to one-half of their lifetime in the menopausal state. On the basis of current epidemiological data, of the 18 million American women in their 70s–80s, 40–50% can be expected to manifest the histopathological changes of AD (10). The increasing number of individuals vulnerable to developing AD is even more staggering when one considers that globally, in every month, 800,000 people reach the age of 65 years (3,11). The potential impact of AD on society as a whole is overwhelming, as the disease, although more prevalent in women, affects both sexes. The projected exponential increase in the prevalence of AD, along with the anticipated impact on families and society, highlights the imperative for developing strategies for the prevention and treatment of AD.
3.
HORMONE REPLACEMENT THERAPY (HRT) AND RISK OF DEVELOPING AD: IMPORTANCE OF THE TIME OF ONSET AND THE TYPE OF HRT
Studies of the use of HT have suggested that HT reduces the risk of developing AD. Multiple epidemiological studies indicate that ET=HT can significantly reduce the risk of developing AD (10,12–15) (Table 1).
76.2 years (ET=HT users) 74.5 years— all women in study
Age of women at time of enrollment Age at initiation of HT
Duration of HT use
ET vs. HT use
Prospective study of incident AD
Type of study
50–75 years (extrapolated by HT use of 1–10 years prior to time of enrollment: presumption that some women began HT at the time of menopause) 72% of HT group used unopposed ET 28% of HT used estrogen plus progestin 1–3 years (310 women) 3–10 years (319 women) >10 years (427 women)
Cache County (18)
Comparison of the Cache County Study vs. WHIMS
Variable
Table 1
71% 60% 54% 50% 44% 10%
of of of of of of
cohort cohort cohort cohort cohort cohort
for for for for for for
1 year (1583=2229) 2 years (1337=2229) 3 years (1204=2229) 4 years (1115=2229) 5 years (0981=2229) 6 years (0223=2229)
65–79 years (At time of enrollment in trial: 3-month washout if using HT at time of initial assessment) 100% of HT group used CEE plus MPA
Randomized, double-blind, placebo-controlled clinical trial of incident dementia Range: 65–79 years 81.7% of cohort between 65–75 years of age
WHIMS (21)
134 Brinton
Predictions based on data
Outcome
Former users with >10 years of exposure showed a 5-fold (500%) reduction in risk of AD (0.17 RR). All former users showed a reduced risk of AD (67% reduction in risk: 33% relative risk) Current users with >10 years of exposure showed 50% reduction in risk Current users with <10 years of exposure showed a doubling of the risk for developing AD 2.12–2.41 Because 72% of women in Cache County study were on ET and only 28% were on HT, the greatest contribution to the data are from ET users. These data lead to the prediction that women started on ET at the time of menopause and who continue the therapy for 10 years will show a substantial reduction in the incidence of AD, whereas women who begin ET or HT at age 65 years will show an increased incidence of AD
Data on the 44% of women who remained on HT for 5 years is currently unavailable If HT instituted at age 65 years leads to increased incidence of AD, then women on HT for greater duration of time should contribute the greatest number of AD cases
Ever HT users showed a 2.05% increase in the risk for developing AD These data are consistent with Cache County data for current users with <10 years of exposure
Estrogen Therapy for Prevention of Alzheimer’s Disease 135
136
Brinton
Estrogen-replacement therapy was the first therapeutic intervention found to significantly reduce the risk of AD and it has remained the best characterized of several therapeutic interventions, such as nonsteroidal anti-inflammatory agents, which are also proposed to decrease the risk of AD (15). If the epidemiological projections prove correct, the widespread use of ET during menopause could reduce the number of women in the United States with AD by more than 1 million (10). Moreover, ET and other gonadal hormone replacement regimens, such as testosterone replacement therapy for men, may reduce the overall incidence of AD in aging populations (16,17). Two recent studies on the impact of HT (combination of estrogen and a progestin agent or estrogen only) provide crucial insights into the time of initiation of therapy, the duration of therapy, and the formulation of the therapy. A summary comparison of some of the details of these two studies is presented in Table 1. The first study conducted by Zandi et al. (18) was a prospective study of incident dementia in older women who were either currently using or had formerly used HT. It is important to note that 72% of the HT users in this study actually used unopposed ET. Only 28% of the women had or were using a combination of estrogen and progestin (HT). Results of this study confirmed a striking gender bias for development of AD; in that, women had a 2.5-fold higher incidence of AD than their age-matched male counterparts. The greater incidence of AD in women was evident in the nonusers group, in which incidence of the disease increased in women over the age of 80 years and exceeded the risk in men of similar age. Remarkably, the female-specific increase in AD risk was completely abolished in women who had formerly used ET=HT for 10 years. In this group, their risk of AD was the same as that of men. Nearly, all of the reduced risk reflected former use of ET=HT. Women who were currently using ET=HT had a similar benefit only if they had used HT for >10 years. Secondly, women who had used ET=HT had a lower incidence of AD than those who had never used ET=HT. Thirdly, women who used ET=HT showed a 60% reduction in the incidence of AD over a 3-year follow-up than nonusers (18). Further, the magnitude of the HT effect was dependent upon duration of use. Overall, risk of AD was reduced from 20% to 60% among former users of HT. Cache County women who developed AD, but who had begun HT 10 years prior to the diagnosis of AD, experienced no benefit of the hormone intervention (comparable with the women in the WHIMS trial, see what follows). Why? One reason might be that the degenerative process of AD was already underway, thereby abrogating the benefits of ET=HT. Histological analyses indicate that the degenerative process of AD begins well before the clinical symptomatology (19,20). The data from the Cache County study are consistent with other studies. Data from the retrospective and prospective epidemiological trials for the most part indicated that ET=HT reduced the risk of AD (15)
Estrogen Therapy for Prevention of Alzheimer’s Disease
137
(see also Fig. 1, for a theoretical model of advantage of ET). Thus, it came as a shock when results of the double-blind, placebo-controlled, multicenter clinical trial of WHIMS of the National Institutes of Health reported that women on HT (PremProÕ ), consisting of conjugated equine estrogens (CEEs) and medroxyprogesterone acetate (MPA), showed a 2-fold increased risk in developing AD (21). Because of the increased risk of cardiovascular events, such as stroke, and breast cancer in the hormone-replacement arm of the Women’s Health Initiative (WHI), this arm of the trial was prematurely terminated (22,23). Shumaker et al. (21) conducted an analysis of the risk of developing dementia in these women as part of the WHIMS. Results of the analysis of the hormone-replacement component of the WHIMS revealed that women receiving PremProÕ had twice the risk of developing AD than those receiving placebo. Among the 4532 women enrolled in the trial, a total of 61 were diagnosed with probable dementia. In the placebo group, 21 out of 2303 women were diagnosed with probable dementia, whereas 40 out of
Figure 1 Model of the estrogen-inducible advantage. Neurons treated with estrogen prior to toxic insults, such as free radicals, b-amyloid, ischemia, or glutamate excitotoxicity, are able to prevent damage by exerting an enhanced defense response, indicated by an increase in the antiapoptotic protein Bcl-2 (although this is not the only activated defense mechanism), thereby surviving the assault. In contrast, neurons treated with estrogen, after damage has occurred, are not imbued with an estrogen advantage, as estrogen-inducible defense mechanisms require an intact membrane and low intracellular calcium for activation. Treatment following toxic insult fails to activate defense mechanisms and results in cell death.
138
Brinton
2229 women treated with PremProÕ were diagnosed with probable dementia. Alzheimer’s disease was the most prevalent diagnosis in both groups, accounting for 50% of the diagnoses of dementia. Twenty of the 40 cases within the PremProÕ group were diagnosed with AD while 20 were diagnosed with vascular dementia. Of women receiving placebo, 12 of the 21 cases were diagnosed with AD, whereas the remaining 9 were diagnosed with vascular dementia. The increased risk of dementia associated with PremProÕ was projected to result in 23 additional cases of dementia per 10,000 women per year (21). In a related study, also from WHIMS, Rapp et al. (24) determined the effect of HT on global cognitive function and decline as measured by the Modified Mini-Mental State Examination (3MSE), the same cognitive test used to screen for dementia in WHIMS. Because of the repeated testing over 4 years of follow-up, women’s performance on this exam tended to improve over time. However, women on PremProÕ exhibited slightly less improvement relative to the placebo group. Moreover, women assigned to the hormone-replacement arm had a greater likelihood of declining two standard deviations on the 3MSE relative to placebo. In the third study reported in this series, Wasertheil-Smoller et al. (22) reported the impact of PremProÕ on the risk of stroke among 16,608 women enrolled in the WHI hormone-replacement arm. The reported 31% increased risk for combined ischemic and hemorrhagic stroke among women taking PremProÕ is consistent with earlier reports from the WHI based on fewer cases of stroke. What are the factors that might account for the adverse effect on incidence of AD found in the WHIMS trial? First among these is the age at which HT was initiated. All women participating in the WHIMS study were 65 years of age and began HT at the time of enrollment into the trial (Table 1). Resnick et al. (25) have addressed the issue of a critical time window for efficacy for initiating ET=HT for prevention of AD. On the basis of their analysis of case–control studies, such as the Cache County study discussed earlier, and prospective analyses, these investigators suggested that early intervention with ET or HT during the climacteric change is predictive of reduced risk of AD, whereas ET or HT intervention at later times or when AD has developed is without benefit. A second potential reason for the adverse impact of HT (PremProÕ ) on the incidence of AD is the progestin used in the formulation. The WHIMS data showing that HT was associated with a 2-fold increase in the risk of AD are consistent with our findings that MPA, the progestin used in the trial, completely antagonizes the neuroprotective and cognitive mechanisms of estrogen (26–28). Results of our studies indicate that progestins differ dramatically in their effects on neuronal responses and their interaction with estrogen. Specifically, in hippocampal neurons, 17b-estradiol (E2), progesterone, and 19-norprogesterone, alone or in combination with
Estrogen Therapy for Prevention of Alzheimer’s Disease
139
E2, are neuroprotective against insults associated with AD and increase the expression of the antiapoptotic protein Bcl-2 (26). In striking contrast, MPA was not neuroprotective nor did induce Bcl-2 expression. Moreover, MPA completely antagonized E2-induced neuroprotection and Bcl-2 expression. We also found that MPA blocked E2-inducible memory mechanisms, whereas progesterone exerted a mixed agonist=antagonist effect (27). Although the mechanism(s) underlying the disparity between progestins remains to be fully determined, translocation of signaling molecules (ERK 1=2) to the nucleus appears to be crucial and is predictive of neuroprotective efficacy (28). E2 and progesterone, alone or in combination, translocate the MAP kinase ERK to the nucleus, whereas MPA does not and MPA blocked E2-induced translocation of ERK (28). These studies conducted while WHI was in progress, and provide a cellular profile upon which current clinical progestins and those in development can be tested for their impact on estrogen-inducible effects in neurons. Results of these analyses could then be used to predict efficacy of HT formulations for prevention of AD. Blockade of estrogen-inducible mechanisms of neuroprotection and cognition would be expected to obstruct estrogen protection against insults that culminate in AD thereby increasing the risk for the disease (26–28). Together, these data show that there is no disparity between the observations derived from cellular analyses and those derived from humans (26–29). As MPA antagonized estrogen-induced protection against degenerative insults in neurons, we propose that the HT trial of the WHI supports a vital role for estrogen in the prevention of AD. Our data and those of the WHI predict that the progestin within an HT will be pivotal for preventing neurodegenerative diseases and sustaining cognitive function throughout the menopausal years. The challenge remains to develop a therapeutic strategy for promoting the beneficial effects of estrogen in brain while preventing untoward consequences of the hormone in other organ systems. The continuation of the estrogen-only arm would suggest that ET is drastically different from the combination of estrogen and MPA therapy. The estrogen-only arm of the WHIMS, which will evaluate the role of ET in 7525 women on both symptom onset and progression of AD, is still ongoing (21). If the estrogen-replacement arm of the WHIMS continues to completion, results of this study will begin to emerge in the year 2006. However, the results of the Cache County study, showing that women who had recently begun ET=HT prior to the time of assessment were at higher risk of developing AD (Table 1), maybe predictive of an outcome comparable to the HT arm of WHIMS. Basic science analyses of the ET formulation of the WHI and WHIMS, which comprises CEEs, predict that ET will protect neurons against degenerative insults associated with AD. Our data show that CEEs are both neurotrophic and neuroprotective against toxic insults associated
140
Brinton
with AD in hippocampal, cortical, and basal forebrain neurons (30). An extremely important aspect of these studies and virtually all of the basic science in vitro and in vivo analyses is that the neurons are healthy prior to estrogen exposure and prior to exposure to the neurotoxic insult. Moreover, most of the studies use a prevention model—i,e., neurons are pretreated with an estrogen prior to exposure to the neurological insult. The healthy neuron=prevention model is used to reduce the variables contributing to the experimental outcome so that results can be directly attributable to one experimental condition, but this model obviously suffers from the generalizability to unhealthy neurons. Results from the healthy neuron model lead to the prediction that, in healthy women, ET and some formulations of HT would result in a reversal of neurological deficits associated with estrogen deficiency and would reduce the risk of neurodegenerative disease. Simply put, healthy women at the time of the perimenopause or menopause transition would show the greatest benefit of estrogen receptor (ER) or HT. The following sections further discuss relevant basic science data, as well as clinical data, related to the role of estrogen in cognitive functions and AD.
4.
THE ROLE OF ESTROGEN IN COGNITIVE FUNCTIONS
Clinically, changes in cognitive functioning have been associated with menopause (31). The relationship between estrogen and cognition was first noted, as women entering menopause frequently voiced complaints of memory and concentration difficulties (32). The first experimental data to indicate a link between estrogen and cognitive functions were the findings of Luine and colleagues (33). In the 1980s, they found that 17b-estradiol increased choline acetyltransferase activity, which led to increased acetylcholine levels. This finding provided the impetus for studies of cognitive and memory function in neurologically normal women and for the eventual use of ET in women with AD. In the mid-1980s, Sherwin and colleagues (34) began a systematic analysis of the impact of estrogen loss and replacement on the cognitive function of memory. Results of these analyses demonstrated that verbal memory declined with loss of estrogen, but could be restored to premenopausal levels when ET was rapidly instituted following surgically induced menopause. More recent studies by Maki et al. (35) found that women receiving various types of HT, including ET or estrogen and progesterone replacement therapy, performed better than those not receiving therapy on tests for verbal learning and memory, including encoding, consolidation, and retrieval. In an earlier cross-sectional human female study, Resnick et al. (25) found that women who were current users of ET performed significantly better on a visual memory test, which has also been shown to predict
Estrogen Therapy for Prevention of Alzheimer’s Disease
141
the onset of AD than women who had never received ET. As part of this study, a subset of the women who never used hormone replacement was studied longitudinally. At the initial memory test, none of the women was using HT. During the intervening 6 years between the first and the second testing periods, a subset of these women began HT independent of the study. At the follow-up 6-year analysis, the women who did not use HT exhibited a significant decline in memory retention compared with their first test performance. In stark contrast, the women who independently began HT during the intervening 6 years showed no decline in memory function relative to their first test performance 6 years earlier. The disturbing aspect of this study was the observation that although the women who elected to receive HT between the first and the second testings were able to maintain memory function comparable to those exhibited at those first test (6 years prior) they were not able to match the performance of those women who had received HT earlier in their menopause. A recent review by Sherwin (36) provides a detailed analysis of studies of estrogen replacement and cognitive function in postmenopausal women. The analysis of randomized-controlled studies, which included studies conducted since the 1980s and which varied drastically in design, hormone-replacement composition, and route of administration, led Sherwin to tentatively conclude that short-term ET selectively maintains verbal memory in healthy, naturally, and surgically menopausal women. Analysis of cross-sectional studies of postmenopausal estrogen users and nonusers from the general population indicated that in all but one study, women over the age of 50 treated with estrogen performed significantly better than nonusers on at least one aspect of cognition, which were usually verbal memory and fluency. Analysis of data from longitudinal studies indicated a consistent result that use of estrogen is associated with better cognitive functioning and less cognitive deterioration in older women over time. The body of data taken as a whole indicates that ET can reverse certain cognitive deficits associated with menopause if begun soon after the initiation of menopause and that cognitive function is sustained over time in women using ET. Estrogen or hormone therapy intervention, initiated later in menopause, is associated with a preservation of memory function but not with a restoration to premenopausal level. 4.1.
Cellular Mechanisms Activated by Estrogen Leading to Regulation of Cognitive Function
Remarkably, one of the sites of estrogen action in neurons is perfectly poised to regulate the cognitive function associated with memory. Estrogen potentiates the action of the most broadly distributed memory system in the brain, the glutamate AMPA and NMDA receptors (37–40). Estrogen potentiation of NMDA receptor function is mediated by a complex signaling
142
Brinton
cascade involving the Src=MAP kinase signaling pathways (Fig. 3) (41,42). Potentiation of this receptor system leads to a rise in the intracellular messenger calcium and activation of a cascade that results in an increase in the morphological complexity of neurons, synapses, and ultimately new neural networks (Figs. 2 and 3) (37,43,44). Through this pathway, estrogen induces an advantage in activating mechanisms of cognition and memory over untreated neurons=brains.
Figure 2 Conceptual model of estrogen action in brain. Schematic conceptual framework of estrogen action in the brain that incorporates a temporal cascade, both dependent upon and regulating Ca2þ signaling. Our working model is a three-tiered temporal cascade composed of (1) an initiation mechanism that is Ca2þ dependent, (2) a propagation phase that enhances Ca2þ-signaling cascades, and (3) a third phase of proactive adaptation that protects against excesses in (27)j. The focus of our recent work is on the third phase, that of proactive adaptation. We use the term proactive adaptation to represent two conditions: (i) estrogen induction of a protected cellular state and (ii) a broad spectrum defense against multiple and seemingly unrelated toxic agents.
Estrogen Therapy for Prevention of Alzheimer’s Disease
143
Figure 3 Mechanism of estrogen advantage. (1) The estrogen signaling cascade leading to memory function and protection against degenerative diseases begins at the ER at the neuron cell membrane. (2) Estrogen binding to its membrane receptor leads to activation of a calcium channel that conducts calcium into the neuron. (3) Influx of calcium, very likely through L-type calcium channels, activates the Src kinase. (4) Src activates MAP kinase extracellular signal-regulated kinases (ERKl=2)=MAP kinases which then (5) phosphorylate the glutamate NMDA receptors, the major memory system of the brain, which (6) conduct even more calcium into the neuron. (7) Activated MAP kinase translocates to the nucleus where it activates the transcription factor CREB. (8) Calcium flowing into the neuron results in activation of calcium-dependent mechanisms, including activation of CREB and CREB-regulated synaptic plasticity genes that lead to generation of new synapses and long-term memory function. (9) CREB can also regulate expression of the genes for antiapoptotic proteins, Bcl-2 and Bcl-x. (10) Estrogen increases the sequestration of calcium into mitochondria thereby reducing the levels of free calcium within the cytoplasm protecting neurons against deleterious effects of excess calcium. Estrogen-inducible Bcl-2 protects mitochondria from potential deleterious effects of excesses in mitochondrial calcium and prevents the release of factors from mitochondria that will activate the apopotic cell suicide cascade. (11) These mechanisms in turn lead to estrogen-inducible neuroprotection and ultimately increased neuron survival in the face of factors that cause neurodegeneration. If estrogen deficiency occurs, insufficient estrogen is available to activate the cascade effectively. If the ER decreases, then there is insufficient receptor to transduce the estrogen signal. If the neuronal membrane is damaged or dysfunctional, then step 2 in the cascade is vulnerable to malfunction and the cascade that is required for estrogen potentiation of memory function and protection against toxic insults will not be activated. Alternatively, if the membrane is damaged and becomes leaky to extracellular calcium, then the estrogen-induced calcium signal required to activate step 3 in the cascade is not detectable relative to the large influx of calcium through leaky membranes. Dysfunction at any step in the cascade results in a loss of estrogen advantage.
144
4.2.
Brinton
Which ER Is Required to Promote Cognitive Function and Neuronal Survival?
Surprisingly, the cognitive and neuronal survival effects of estrogen are dependent upon a membrane-associated ER and not upon a direct regulation of gene transcription (37). Although the body of evidence thus far supports the idea that ERs at the membrane are essentially the same proteins required for nuclear actions of estrogen, it remains to be determined whether the ERa or ERb is responsible for the membrane actions of estrogen. To date, there are reports using ER knockout animals that indicate that both ERa and ERb can mediate the cognitive effects and neuronal survival effects of estrogen (37). In some model systems, there is a divergence in the actions of the a and b subtypes (45), whereas in other model systems, they appear to induce the same effects (41). A recent study by Yaffe et al. (46) also suggests that an isoform of the ER, namely ERp, is associated with an increased incidence of AD. The tools to determine which ER is required for the brain effects of estrogen, however, have only been recently available, and it will be years before the final answer is determined. The data thus far demonstrate that both ERa and ERb are present in brain, and suggest that both are required for the full array of estrogen action in the brain. Not yet conclusively determined is which ER is required for which estrogen-indueible action in brain. 4.3.
The Impact of Progestins on Estrogen-Inducible Responses in the Brain
Addition of progestins to ET has been the standard clinical practice, once the data indicated that women receiving unopposed ET had a 3-fold increase in the risk of uterine cancer (47). Progesterones can inhibit estrogen-stimulated proliferation in the uterus and thus are added as a standard approach to reduce hyperplasia of the endometrium and related tumorigenesis (48). In addition to prevention of endometrial hyperplasia, inclusion of progestins in HT was thought to reduce estrogen-associated increased risk of breast cancer (48). In contrast to this long held belief, a recent study found that an estrogen–progestin regimen increased cancer risk beyond that associated with estrogen alone (49). We sought to determine the impact of clinically relevant progestin agents on estrogen-inducible mechanisms of memory and neuron survival. Results of those studies revealed that not all progestins have the same impact on estrogen-inducible responses in brain (Table 2), Analyses of progestin regulation of estrogen-inducible memory mechanisms demonstrated that progesterone alone was a modest activator of memory mechanisms (27). The combination of estrogen and progesterone resulted in activation that was greater than control, but less than estrogen alone (27). In the presence of the progestin MPA (the most frequently prescribed progestin, and
Estrogen Therapy for Prevention of Alzheimer’s Disease
145
Table 2
Effect of Progestins on Neuroprotection and Estrogen InducibleMechanisms
Agent
Neuropro- Estrogen-inducible Estrogen-inducible memory tection neuroprotection mechanism
Progesterone Effective Synergizes with (26,27,40) estrogen 19-Norprogesterone Effective Synergizes with (26) estrogen Medroxyprogesterone Ineffective Completely acetate (26,27,40) antagonizes
Modestly attenuates Untested Completely antagonizes
the progestin in the formulation PremProÕ used in WHI), estrogen-inducible memory mechanisms were completely abolished. In the presence of progesterone and 19-norprogesterone, estrogeninducible Bcl-2 and neuronal survival were comparable to the effect of estrogen alone. In fact, there was a slight synergy in neuroprotection in the presence of both hormones (26,28). In stark contrast, MPA, a highly potent progestin with a long-half life that is highly efficacious in blocking estrogen proliferation in uterus, completely blocked the neuroprotective effects of estrogen when both hormones were present (26,28). Although decades of accumulated observational evidence point to a positive effect of estrogens on cognition and learning and memory (10), much of the recent public interest in estrogens has focused on a recent publication of results of the large-scale, long-term, prospective, placebo-controlled, and randomized clinical trial of HT in WHI studies of the National Institutes of Health (50). In this report, HT, which included a combination of CEEs and the progestin MPA, failed to show expected positive effects. On the contrary, with an average 5.2-year follow-up, overall health risks exceeded benefits from use of this particular combined CEE plus MPA regimen among 16,608 healthy postmenopausal U.S. women ranging from 50 to 79 years of age. In these women, CEE plus MPA increased the incidence of coronary heart disease, stroke, deep vein thrombosis, and pulmonary embolism with a trend towards an increased risk of invasive breast cancer while the benefits included significantly reducing hip and clinically observed vertebral fracture rates by one third and greatly reducing the incidence of colorectal cancer compared with the placebo group (50). On the basis of our data and those of the WHI HT clinical trial, the systemic use of the progestin, MPA, at the 2.5 mg=day dose should be re-evaluated. However, it is quite clear that the proliferative effects of estrogen in the uterus require the use of a progestin agent. Thus, two fundamental issues must be addressed. The first question is which progestin agent
146
Brinton
should be administered? The second question is how should this progestin agent be administered—locally or systemically? To address the first question, there are a host of effective progestins that are currently in clinical use either as part of HRTs or as part of oral contraceptive regimens (48). Our current data indicate that progesterone and 19-norprogesterone exert some benefit on their own, only modestly attenuate estrogen-inducible memory mechanisms, and synergize estrogen to promote neuroprotective mechanisms (28,51). The second issue is a route of administration issue. Are there organ systems that require progesterone for sustained optimal function? The success of ET in hysterectomized=oophorectomized women would suggest that women do not experience adverse affects when not replaced with progesterone. The beneficial effects of ET in hysterectomized=oophorectomized women would suggest that systemic administration of progesterone is not required and that local administration of the progestin to women with a uterus would provide an efficacious alternative to systemic administration. The technologies for delivering hormones to the uterus have been developed for contraception and fertility purposes and could be applied for HT (52–54).
5.
HOW MIGHT ESTROGEN-REPLACEMENT THERAPY REDUCE THE RISK OF AD?
Estrogen also imbues neurons with a survival advantage by increasing existing cellular defense and antiapoptotic strategies (Figs. 2 and 3). Numerous laboratories have demonstrated that pre-exposure to estrogen can promote the survival of neurons exposed to toxic insults. Pre-exposure of neurons to estrogen can significantly diminish the damage induced by b-amyloid, excitotoxic glutamate, and free radical generators like hydrogen peroxide (Fig. 1) (30,55,56). Estrogen pretreatment can also greatly attenuate neuron death resulting from ischemia (57). Each of these toxic insults can exert both unique and overlapping mechanisms of lethality. Because estrogen promotes neuronal survival in the face of differing mechanisms of damage, it suggests that estrogen induces a neuroprotective mechanism that is broad in scope and highly effective in promoting cell survival. One such mechanism is an increase in the antiapoptotic proteins, Bcl-2 and Bcl-x, and an increase in levels of their respective mRNAs (27,58). The complete repertoire of Bcl-2 effects is still under study; however, we do know that a principal target of Bcl-2 action is the mitochondria. We have recently discovered that estrogen increases calcium sequestration by mitochondria, thereby promoting the survival of neurons exposed to excitatory glutamate (51). Our current hypothesis is that estrogen-inducible Bcl-2 promotes the increased load capacity of mitochondria necessary to protect neurons against excesses in intracellular calcium.
Estrogen Therapy for Prevention of Alzheimer’s Disease
5.1.
147
Estrogen and b-Amyloid Generation and Deposition in Brain
Data emerging from both in vitro and in vivo model analyses have implicated estrogen as a significant regulator of b-amyloid processing and deposition. Specifically, estrogen has also been shown in vitro and in vivo to reduce the production of the principal constituent of plaques, the b-amyloid peptides (Ab) (59). In addition, estrogen can increase microglial uptake of Ab (60). Data from cultured cells exposed to estrogen provided an early indication that estrogen impacted amyloid precursor peptide (APP) processing to decrease soluble Ab levels while concomitantly increasing sAPP levels (61). In transgenic mouse models of b-amyloid over-expression, mice with a mutation in APP gene (Tg2576) and mice with a double mutation in the presenilin-1 gene and the APP gene PS=APP (PS1.5.1=Tg2576) show that estradiol deprivation due to ovariectomy significantly affects the level of Ab peptides in the brains of transgenic mice. The APP mice at 7 months of age do not have appreciable amounts of insoluble, aggregated Ab in the brain, and in this respect, they are similar to the guinea pig model used by Petanceska et al. (62). Both the guinea pig model and APP mice showed that soluble Ab accumulates in estrogendeprived animals, and this accumulation can be reversed by the administration of estrogen. As PS=APP mice age, they begin to accumulate aggregated Ab. Recently, Callahan et al. (63) have shown that aged female PS=APP mice have a significantly higher amyloid load than male littermates, suggesting that estrogen depletion, as a result of the equivalent of mouse menopause, impacts amyloid deposition in APP mice. Zheng et al. (64) have recently shown that mice with an accelerated amyloid phenotype (PS=APP) have significantly elevated Ab levels in response to ovariectomy-induced estrogen depletion, which can be the reversed estrogen replacement. Data from both the guinea pig study and the PS=APP mice show an increased response of Ab1-42 relative to Ab1-40, suggesting that the mechanisms involved in the generation or clearance of Ab1-42 are the preferential target for estrogen action. Data for estrogen supplementation clearly showed that in common with the guinea pig model, estrogen administration reduced the levels of Abl-42 in both the APP and the PS=APP mouse models. Mechanisms by which Ab accumulates in the estrogen-depleted PS=APP mice likely include effects on APP processing, but also could increase microglial clearance of amyloid, once deposition is initiated in vitro (60). 5.2.
Estrogen Therapy for the Treatment of AD
Although the existing clinical data on ET and prevention of AD remain encouraging (refer the analysis of HT mentioned earlier), the data on the treatment of women with existing disease are not. The association of AD with loss of basal forebrain neurons that synthesize acetylcholine, coupled
148
Brinton
with the findings of estrogen-induced increases in acetylcholine synthesis, motivated Fillit et al. (65) to investigate the impact of ET on cognitive function in women with AD. Results of this small clinical trial proved to be pivotal and led to multiple international clinical and epidemiological studies which, in turn, led to intense study of the impact of estrogen on mechanisms of memory in both animals and humans (37). Despite the positive benefits reported in earlier small, open label, clinical trials (13), two randomized, double-blind, placebo-controlled, parallel-group trials found that both short-term (66) and long-term (1 year) ET (67) did not improve symptoms of most women with AD. A small clinical trial by Asthana et al. (68) reported that ET improved memory function in women with AD; however, the cognitive effects were very modest at best. Thus, it would appear that estrogen maintains and sustains neuronal viability to prevent degenerative diseases but is ineffective in reversing the degenerative disease process. 5.3.
Why Is Estrogen Replacement Therapy Ineffective for Treating AD?
Several factors are reasonable candidates to explain why ET is ineffective in women with existing AD (69). The first of these is the age of the recipient. Like other organs, the brain’s responsiveness to estrogen changes with age (70–72). In the Alzheimer Disease Cooperative study, the mean age of the women enrolled in the study ranged from 56 to 91 years, with an average age of 75 years. In the Henderson et al. study, the mean age was 78 1.0 years years for the control and 77 1.4 years for the estrogen-replacement group. Experimental results indicated that age-related changes in estrogen responsiveness are associated with a decrease in ERa levels and a shift in its subcellular distribution away from the postsynaptic density (73). Thus, one could postulate that given the age of the women in the AD trials and the changes in ERs, a beneficial effect of ET would be minimal at best. Another factor is the duration of estrogen deficiency. In both AD and ET treatment studies, women were deficient in estrogen for nearly 20 years. To treat women with estrogen after more than two decades of estrogen deficiency, with the expectation of therapeutic benefit, assumes that the brain has not undergone some adaptation to the deficiency. This seems highly unlikely. We now know from studies in primates that ERs and their efficacy in promoting neuronal process outgrowth diminish significantly in brain with age (73). The changes in estrogen-induced spine formation in neurologically normal, aged primate brain are paralleled by the results of the Resnick study which showed that in neurologically normal, menopausal women who delayed receiving HT, the initiation of a hormonal replacement regime only partially reversed estrogen-inducible memory function (25). A crucial issue is that of optimal dose. For both studies, a single dose of ET, 0.625 mg CEEs, was used. Although this was certainly a reasonable
Estrogen Therapy for Prevention of Alzheimer’s Disease
149
dose based on the efficacy in neurologically normal menopausal women and clinical trials for a variety of healthy outcomes in neurologically normal women, this particular dose may have been suboptimal for therapeutic benefit. Optimization of dose and formulation for neurologically healthy women who are beginning ET at menopause are often challenging. In women with severely compromised neurological health, the variability in the response to estrogen is bound to be dramatic. Last, but certainly not least, there is the issue of activating mechanisms of estrogen action in neurons damaged from the disease process. To begin, the best evidence for an initiation mechanism of estrogen action thus far is an influx of intracellular calcium that activates the Src=MAP kinase signaling pathway (74). The initiation step for estrogen action is dependent upon a relatively small rise in intracellular calcium. In the presence of b-amyloid and other insults that damage neuronal membranes, calcium begins to leak into the neuron (75). In the case of AD, many neurons have or are in the process of degeneration. These degenerating neurons have compromised membrane integrity and therefore become leaky for many different ions, including calcium (75). If basal levels of intracellular calcium are already high (such as occurs in damaged neurons), then it is possible that a small rise in calcium is no longer detected as a meaningful signal relative to the already high calcium level—there has been a substantial loss in the signal-to-noise ratio. In these neurons, the estrogen effect would be negligible to ineffective depending upon the magnitude of membrane degeneration. Moreover, estrogen-inducible survival advantage relies heavily upon an increase in Bcl-2 (51). If the estrogen-induced signaling cascade required for Bcl-2 induction is dysfunctional, then it logically follows that estrogen inducible Bcl-2 will be compromised. ˆ system and, in particular, the hippocampus (19,76). The hippocamW pus is responsible for short-term memory function, which is the reason why the primary presenting symptom of AD is disabling short-term memory dysfunction. Although the exact magnitude of hippocampal degeneration required for symptomatology of AD to be evident remains unclear, results of both the Alzheimer Disease Cooperative Study and the Henderson et al. study indicate that even in the early stages of the disease, estrogen is ineffective in restoring cognitive function (10,67). Essentially, once the symptoms of the disease are manifested, the long-term beneficial effects of ET are negligible to nonexistent. 6.
SPECULATIONS ON WHEN TO INTERVENE WITH ESTROGEN-REPLACEMENT THERAPY
When is the optimal time to intervene with ET? For many years, it has been a standard clinical practice to begin ET when a women was diagnosable menopausal (elevated FSH) and when she had not experienced a menstrual cycle
150
Brinton
for 1 year (77). Following this algorithm provides a clearly defined set of criteria for clinical practice. However, by the time a woman is diagnosed as menopausal, the woman may also have lost some degree of estrogen responsiveness. Indeed, increasing evidence points to an early accumulation of degenerative insults that precede the onset of symptomatic disease by several decades (19,20). Although not yet proven, our working hypothesis is that perimenopause, characterized by sporadic peaks and troughs of estrogen, is a neuroendocrine strategy designed to disconnect the ER from its downstream effector mechanisms. A simpler but potentially irreversible strategy would be to decrease or cease production of ERs. The existing data suggest that the brain favors a strategy in which estrogen responsiveness is diminished but is not completely lost. From a species survival perspective, the diminution, but not complete loss of estrogen responsiveness, would allow a highly select population of females to continue to reproduce until late in life should some catastrophic event occur to the younger female population. Evolutionary biology aside, the clinical question that is also one for basic scientists is this. Can the dysregulation phenomenon be prevented, and if so, is there a strategy that might be effective in preventing the disconnection between the ER and its effector mechanism? The author emphasize that the disconnect hypothesis is just that, a hypothesis, and at this point there is very little evidence to support or disprove the disconnect hypothesis. If this hypothesis is true—which the extreme fluctuations in estrogen are the root cause of the hypothesized disconnect syndrome—then a strategy such as oral contraceptives, which provide a stable dose of estrogen while providing contraception during perimenopause, may be an approach to sustain the association between the ER and its downstream effector mechanisms, which might be crucial for both memory function and neuron survival. The preponderance of data on estrogen action in brain suggests a positive benefit of estrogen replacement in menopausal women. Despite the encouraging epidemiological data indicating a reduced risk of developing AD in women who have ever received ET (14), only 25% of eligible menopausal women elect to receive prescribed ET. Of that number, 50% discontinue use within the first year of therapy and >70% of those for whom it has been prescribed are not compliant (78). Moreover, only 20% of women who are prescribed for estrogen are still compliant 3 years later (79). The principal reason that women forego ET is the fear of developing breast cancer (78). Estrogen-replacement therapy provides 25% survival advantage for differentiated and nondifferentiated cells (i.e., tumor cells). The estrogen advantage provides a plausible explanation for why some women on ET will suffer an increased risk of developing breast cancer while also experiencing a better survival rate. If ET provides benefits in preventing AD, but women will not pursue this preventive strategy for fear of developing breast cancer, for example,
Estrogen Therapy for Prevention of Alzheimer’s Disease
151
Table 3
Impact of SERMs on Neuronal Markers of Memory and Neuroprotection
SERM
Memory correlate
Plasma membrane Marker of neuronal integrity viability
Tamoxifen (74,83) 4-hydroxy-tamoxifen (74,83) Raloxifene (74,83) Phytoestrogens (74,82)
Ineffective Ineffective
Modest protection Ineffective Modest protection Ineffective
Modest efficacya Ineffective
Modest protection Ineffective Modest protection Ineffective
a
Raloxifene induced neuronal markers of memory at a single concentration in a single brain region involved in memory function (83). However, the concentration required to induce the effect is unlikely to be achieved in brain.
then developing estrogen alternatives that exert estrogen agonist properties in brain, bone, and cardiovascular system, while exerting estrogen antagonist action in the breast and uterus, could be of enormous benefit. Several such compounds with a mixed ER agonist=antagonist profile have been developed, including tamoxifen and raloxifene (80,81), and many more compounds are in the pipeline (74). Thus far, our data predict that neither phytoestrogens nor the selective estrogen receptor modulators (SERMs), tamoxifen and raloxifene, promotes the full range of estrogen action in neurons (Table 3) (74,82,83). Our current efforts are focused on developing an effective NeuroSERM as an estrogen alternative for brain. Achievement of this goal will lead to a molecule that will promote estrogen-inducible mechanisms of memory and neuron survival, which may lead to a reduced risk of AD while also reducing the risk of estrogen-associated tumor growth.
7.
SUMMARY
The basic science and clinical data indicate that ET imbues healthy neurons with a survival advantage when challenged with a neurotoxic agent. The clinical data suggest that greater benefit will occur if ET is begun while neurons are still healthy and not biologically compromised. The data also indicate that deficiency of estrogen for an extended duration results in a diminution or loss of estrogen-inducible memory and survival advantage. The dependency on healthy neurons for estrogen-inducible advantage provides a plausible explanation for why ET is effective in women with AD for only a short time and for why it does not provide long-term benefit, and for why does not ET reverse the disease process. Because the efficacy of estrogen is dependent upon the biological integrity of neurons, variability in response to ET would be expected in the general population with varying
152
Brinton
degrees of neurological health. The future discovery and development of estrogen-like molecules that selectively promote estrogen-inducible mechanisms of memory and neuron survival in brain may lead to a reduced risk of AD without an increased risk of estrogen-associated tumor growth. ACKNOWLEDGMENTS This work was supported by the National Institute of Aging, the National Institute of Mental Health, The Kenneth T. and Eileen L. Norris Foundation, The L.K. Whittier Foundation, the Stanley Family Trust, and the John Douglas French Foundation. The author gratefully acknowledges the superb scientists who have contributed to the body of research from his laboratory, especially Drs. Jon Nilsen, Shuhua Chen, Liqin Zhao, Lixia Zhao, and JunMing Wang, graduate students TsuWei Wu, Kathleen O’Neill, and Michael Kim, and our collaborator Dr. Theodore W. Berger. REFERENCES 1.
Fillit HM, O’Connell AW, Brown WM, Altstiel LD, Anand R, Collins K, Ferris SH, Khachaturian ZS, Kinoshita J, Van Eldik L, Dewey CF. Barriers to drug discovery and development for Alzheimer disease. Alzheimer Dis Assoc Disord 2002; 16(suppl 1):S1–S8. 2. Alzheimer’s disease education & referal center. Alzheimer’s disease: unraveling the mystery [online]. URL: http:==www.alzheimers.Org=unraveling=unraveling. pdf. 3. Whitehouse PJ. Alzheimer’s disease: an international public health problem— clinical goals, strategies, and outcomes. In: Dekker MW, ed. Pharmacological Treatment of Alzheimer’s Disease. New York: Wiley-Liss, 1997:331–343. 4. Brookmeyer R, Gray S, Kawas C. Projections of Alzheimer’s disease in the United States and the public health impact of delaying disease onset. Am J Public Health 1998; 88(9):1337–1342. 5. Brinton RD. A women’s health issue: Alzheimer’s disease and strategies for maintaining cognitive health. Int J Ferril Womens Med 1999; 44(4):174–185. 6. Payami H, Montee K, Grimslid H, Shattuc S, Kaye J. Increased risk of familial late-onset Alzheimer’s disease in women. Neurology 1996; 46(1):126–129. 7. Gao S, Hendrie HC, Hall KS, Hui S. The relationships between age, sex, and the incidence of dementia and Alzheimer disease: a meta-analysis. Arch Gen Psych 1998; 55(9):809–815. 8. Lindsay J, Laurin D, Verreault R, Hebert R, Helliwell B, Hill GB, McDowell I. Risk factors for Alzheimer’s disease: a prospective analysis from the Canadian study of health and aging. Am J Epidemiol 2002; 156(5):445–453. 9. North American Menopause Society. Menopause core curriculum study guide. Basic overview, definitions & statistics [online]. URL: http:==www.menopause. org. 10. Henderson VW. Hormone therapy and the brain: a clinical perspective on the role of estrogen. New York: Parthenon Publishing, 2000.
Estrogen Therapy for Prevention of Alzheimer’s Disease 11. 12.
13. 14.
15. 16. 17.
18.
19.
20.
21.
22.
23.
24.
153
Holden C. New populations of old add to poor nations’ burdens [news]. Science 1996; 273:46–48. Tang MX, Jacobs D, Stern Y, Marder K, Schofield P, Gurland B, Andrews H, Mayeux R. Effect of oestrogen during menopause on risk and age at onset of Alzheimer’s disease. Lancet 1996; 348(9025):429–432. Birge SJ. The role of estrogen in the treatment of Alzheimer’s disease. Neurology 1997; 48(5 suppl 7):S36–S41. Yaffe K, Sawaya G, Lieberburg I, Grady D. Estrogen therapy in postmenopausal women: effects on cognitive function and dementia. JAMA 1998; 279(9):688–695. LeBlanc A. Estrogen and Alzheimer’s disease. Curr Opin Investig Drugs 2002; 3(5):768–773. Brinton RD, Yamazaki RS. Advances and challenges in the prevention and treatment of Alzheimer’s disease. Pharm Res 1998; 15:384–397. Gandy S, Almeida OP, Fonte J, Lim D, Waterrus A, Spry N, Flicker L, Martins RN. Chemical andropause and amyloid-beta peptide. JAMA 2001; 285(17):2195–2196. Zandi PP, Carlson MC, Plassman BL, Welsh-Bohmer KA, Mayer LS, Steffens DC, Breitner JC. Hormone replacement therapy and incidence of Alzheimer disease in older women: the Cache County Study. JAMA 2002; 288(17): 2123–2129. Troncoso JC, Martin LJ, Dal Forno G, Kawas CH. Neuropathology in controls and demented subjects from the Baltimore longitudinal study of aging. Neurobiol Aging 1996; 17(3):365–371. Troncoso JC, Cataldo AM, Nixon RA, Barnett JL, Lee MK, Checler F, Fowler DR, Smialek JE, Crain B, Martin LJ, Kawas CH. Neuropathology of preclinical and clinical late-onset Alzheimer’s disease. Ann Neurol 1998; 43(5):673–676. Shumaker SA, Legault C, Rapp SR, Thal L, Wallace RB, Ockene JK, Hendrix SL, Jones BN III, Assaf AR, Jackson RD, Kotchen JM, Wassertheil-Smoller S, Wactawski-Wende J. Estrogen plus progestin and the incidence of dementia and mild cognitive impairment in postmenopausal women: the Women’s Health Initiative Memory Study: a randomized controlled trial. JAMA 2003; 289(20):2651–2662. Wassertheil-Smoller S, Hendrix SL, Limacher M, Heiss G, Kooperberg C, Baird A, Kotchen T, Curb JD, Black H, Rossouw JE, Aragaki A, Safford M, Stein E, Laowattana S, Mysiw WJ. Effect of estrogen plus progestin on stroke in postmenopausal women: the Women’s Health Initiative: a randomized trial. JAMA 2003; 289(20):2673–2684. Chlebowski RT, Hendrix SL, Langer RD, Stefanick ML, Gass M, Lane D, Rodabough RJ, Gilligan MA, Cyr MG, Thomson CA, Khandekar J, Petrovitch H, McTiernan A. Influence of estrogen plus progestin on breast cancer and mammography in healthy postmenopausal women: the Women’s Health Initiative Randomized Trial. JAMA 2003; 289(24):3243–3253. Rapp SR, Espeland MA, Shumaker SA, Henderson VW, Brunner RL, Manson JE, Gass ML, Stefanick ML, Lane DS, Hays J, Johnson KC, Coker LH, Dailey M, Bowen D. Effect of estrogen plus progestin on global cognitive function in
154
25.
26.
27. 28.
29. 30.
31. 32.
33.
34. 35.
36. 37.
38.
39.
40.
Brinton postmenopausal women: the Women’s Health Initiative Memory Study: a randomized controlled trial. JAMA 2003; 289(20):2663–2672. Resnick SM, Metter EJ, Zonderman AB. Estrogen replacement therapy and longitudinal decline in visual memory. A possible protective effect? Neurology 1997; 49(6):1491–1497. Nilsen J, Brinton RD. Impact of progestins on estrogen-induced neuroprotection: synergy by progesterone and 19-norprogesterone and antagonism by medroxyprogesterone acetate. Endocrinology 2002; 143(1):205–212. Nilsen J, Brinton RD. Impact of progestins on estradiol potentiation of the glutamate calcium response. Neuroreport 2002; 13(6):825–830. Nilsen J, Brinton RD. Divergent impact of progesterone and medroxyprogesterone acetate (Provera) on nuclear mitogen-activated protein kinase signaling. Proc Natl Acad Sci USA 2003; 100(18):10506–10511. Yaffe K. Hormone therapy and the brain: deja vu all over again? JAMA 2003; 289(20):2717–2719. Brinton RD, Chen S, Montoya M, Hsieh D, Minaya J. The estrogen replacement therapy of the Women’s Health Initiative promotes the cellular mechanisms of memory and neuronal survival in neurons vulnerable to Alzheimer’s disease. Maturitas 2000; 34(suppl 2):S35–S52. Neugarten BL, Kraines RJ. ‘‘Menopausal Symptoms’’ in women of various ages. Psychosom Med 1965; 27:266–273. Sherwin BB. Impact of changing hormonal milieu on psychological functioning. In: Lobo RA, ed. Treatment of the Postmenopausal Woman: Basic and Clinical Aspects. Philadelphia: Lippincott Williams & Wilkins, 1999:179–188. Luine VN. Estradiol increases choline acetyltransferase activity in specific basal forebrain nuclei and projection areas of female rats. Exp Neurol 1985; 89(2):484–490. Sherwin BB. Oestrogen and cognitive function throughout the female lifespan. Novartis Found Symp 2000; 230:188–196; discussion 196–201. Maki PM, Zonderman AB, Resnick SM. Enhanced verbal memory in nondemented elderly women receiving hormone-replacement therapy. Am J Psych 2001; 158(2):227–233. Sherwin BB. Estrogen and cognitive aging in women. Trends Pharmacol Sci 2002; 23(11):527–534. Brinton RD. Cellular and molecular mechanisms of estrogen regulation of memory function and neuroprotection against Alzheimer’s disease: recent insights and remaining challenges. Learn Mem 2001; 8(3):121–133. Bi R, Broutman G, Foy MR, Thompson RF, Baudry M. The tyrosine kinase and mitogen-activated protein kinase pathways mediate multiple effects of estrogen in hippocampus. Proc Natl Acad Sci USA 2000; 97(7):3602–3607. Bi R, Foy MR, Vouimba RM, Thompson RF, Baudry M. Cyclic changes in estradiol regulate synaptic plasticity through the MAP kinase pathway. Proc Natl Acad Sci USA 2001; 98(23):13391–13395. Nilsen J, Chen S, Brinton RD. Dual action of estrogen on glutamate-induced calcium signaling: mechanisms requiring interaction between estrogen receptors and src=mitogen activated protein kinase pathway. Brain Res 2002; 930(1–2): 216–234.
Estrogen Therapy for Prevention of Alzheimer’s Disease 41.
42.
43.
44. 45.
46.
47.
48. 49.
50.
51.
52. 53. 54. 55.
56.
155
Wade CB, Robinson S, Shapiro RA, Dorsa DM. Estrogen receptor (ER)alpha and ERbeta exhibit unique pharmacologic properties when coupled to activation of the mitogen-activated protein kinase pathway. Endocrinology 2001; 142(6):2336–2342. Singh M, Setalo G Jr, Guan X, Frail DE, Toran-Allerand CD. Estrogeninduced activation of the mitogen-activated protein kinase cascade in the cerebral cortex of estrogen receptor-alpha knock-out mice. J Neurosci 2000; 20(5):1694–1700. McEwen BS, Alves SE, Bulloch K, Weiland NG. Ovarian steroids and the brain: implications for cognition and aging. Neurology 1997; 48(5 suppl 7): S8–S15. Woolley CS. Estrogen-mediated structural and functional synaptic plasticity in the female rat hippocampus. Horm Behav 1998; 34(2):140–148. Dubai DB, Zhu H, Yu J, Rau SW, Shughrue PJ, Merchenthaler I, Kindy MS, Wise PM. Estrogen receptor alpha, not beta, is a critical link in estradiolmediated protection against brain injury. Proc Natl Acad Sci USA 2001; 98(4):1952–1957. Yaffe K, Lui LY, Grady D, Stone K, Morin P. Estrogen receptor 1 polymorphisms and risk of cognitive impairment in older women. Biol Psych 2002; 51(8):677–682. Klein R, Berlin L. Benefits and risks of hormone replacement therapy. In: Pavlik EJ, ed. Estrogens, Progestins, and Their Antagonists. Boston: Birkhauser, 1996:3–50. Mattox JH, Shulman LP. Combined oral hormone replacement therapy formulations. Am J Obstet Gynecol 2001; 185(suppl 2):S38–S46. Schairer C, Lubin J, Troisi R, Sturgeon S, Brinton L, Hoover R. Menopausal estrogen and estrogen–progestin replacement therapy and breast cancer risk. JAMA 2000; 283(4):485–491. Writing group for the Women’s Health Initiative. I. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women’s Health Initiative randomized clinical trial. JAMA 2002; 288: 321–333. Nilsen J, Diaz Brinton R. Mechanism of estrogen-mediated neuroprotection: regulation of mitochondrial calcium and Bcl-2 expression. Proc Natl Acad Sci USA 2003; 100(5):2842–2847. Newton J. The current status of intra-uterine contraceptive devices and systems. Br J Fam Plann 2000; 26(1):14–15. Davis AJ. Advances in contraception. Obstet Gynecol Clin North Am 2000; 27(3):597– 610. Lahteenmaki P, Jukarainen H. Novel delivery systems in contraception. Br Med Bull 2000; 56(3):739–748. Behl C, Skutella T, Lezoualc’h F, Post A, Widmann M, Newton CJ, Holsboer F. Neuroprotection against oxidative stress by estrogens: structure-activity relationship. Mol Pharmacol 1997; 51(4):535–541. Green PS, Gridley KE, Simpkins JW. Nuclear estrogen receptor-independent neuroprotection by estratrienes: a novel interaction with glutathione. Neuroscience 1998; 84(1):7–10.
156
Brinton
57. Simpkins JW, Rajakumar G, Zhang YQ, Simpkins CE, Greenwald D, Yu CJ, Bodor N, Day AL. Estrogens may reduce mortality and ischemic damage caused by middle cerebral artery occlusion in the female rat. J Neurosurg 1997; 87(5):724–730. 58. Pike CJ. Estrogen modulates neuronal Bcl-xL expression and b-amyloidinduced apoptosis: relevance to Alzheimer’s disease. J Neurochem 2000; 72: 1552–1563. 59. Petanceska SS, Nagy V, Frail D, Gandy S. Ovariectomy and 17b-estradiol modulate the levels of Alzheimer’s amyloid b peptides in brain. Exp Gerontol 2000; 35(9–10):1317–1325. 60. Li R, Shen Y, Yang LB, Lue LF, Finch C, Rogers J. Estrogen enhances uptake of amyloid b-protein by microglia derived from the human cortex. J Neurochem 2000; 75(4):1447–1454. 61. Xu H, Gouras GK, Greenfield JP, Vincent B, Naslund J, Mazzarelli L, Fried G, Jovanovic JN, Seeger M, Relkin NR, Liao F, Checler F, Buxbaum JD, Chait BT, Thinakaran G, Sisodia SS, Wang R, Greengard P, Gandy S. Estrogen reduces neuronal generation of Alzheimer b-amyloid peptides. Nat Med 1998; 4(4):447–451. 62. Petanceska SS, Nagy V, Frail D, Gandy S. Ovariectomy and 17b-estradiol modulate the levels of Alzheimer’s amyloid b peptides in brain. Neurology 2000; 54(12):2212– 2217. 63. Callahan MJ, Lipinski WJ, Bian F, Durham RA, Pack A, Walker LC. Augmented senile plaque load in aged female b-amyloid precursor proteintransgenic mice. Am J Pathol 2001; 158(3):1173–1177. 64. Zheng H, Xu H, Uljon SN, Gross R, Hardy K, Gaynor J, Lafrancois J, Simpkins J, Refolo LM, Petanceska S, Wang R, Duff K. Modulation of A(b) peptides by estrogen in mouse models. J Neurochem 2002; 80(1):191–196. 65. Fillit H, Weinreb H, Cholst I, Luine V, McEwen B, Amador R, Zabriskie J. Observations in a preliminary open trial of estradiol therapy for senile dementia-Alzheimer’s type. Psychoneuroendocrinology 1986; 11(3):337–345. 66. Henderson VW, Paganini-Hill A, Miller BL, Elble RJ, Reyes PF, Shoupe D, McCleary CA, Klein RA, Hake AM, Farlow MR. Estrogen for Alzheimer’s disease in women: randomized, double-blind, placebo-controlled trial. Neurology 2000; 54(2):295–301. 67. Mulnard RA, Cotman CW, Kawas C, van Dyck CH, Sano M, Doody R, Koss E, Pfeiffer E, Jin S, Gamst A, Grundman M, Thomas R, Thal LJ. Estrogen replacement therapy for treatment of mild to moderate Alzheimer disease: a randomized controlled trial. Alzheimer’s Disease Cooperative Study. JAMA 2000; 283(8): 1007–1015. 68. Asthana S, Craft S, Baker LD, Raskind MA, Birnbaum RS, Lofgreen CP, Veith RC, Plymate SR. Cognitive and neuroendocrine response to transdermal estrogen in postmenopausal women with Alzheimer’s disease: results of a placebo-controlled, double-blind, pilot study. Psychoneuroendocrinology 1999; 24(6):657–677. 69. Marder K, Sano M. Estrogen to treat Alzheimer’s disease: too little, too late? So what’s a woman to do? Neurology 2000; 54(11):2035–2037. 70. Finch CE, Tanzi RE. Genetics of aging. Science 1997; 278(5337):407–411.
Estrogen Therapy for Prevention of Alzheimer’s Disease 71.
72.
73.
74.
75.
76.
77.
78. 79. 80.
81. 82.
83.
157
Adams MM, Shah RA, Janssen WG, Morrison JH. Different modes of hippocampal plasticity in response to estrogen in young and aged female rats. Proc Natl Acad Sci USA 2001; 98(14):8071–8076. Bi R, Foy MR, Thompson RF, Baudry M. Effects of estrogen, age, and calpain on MAP kinase and NMDA receptors in female rat brain. Neurobiol Aging 2003; 24(7):977–983. Adams MM, Fink SE, Shah RA, Janssen WG, Hayashi S, Milner TA, McEwen BS, Morrison JH. Estrogen and aging affect the subcellular distribution of estrogen receptor-a in the hippocampus of female rats. J Neurosci 2002; 22(9):3608–3614. Brinton RD. Selective estrogen receptor modulators (SERM) for the brain: advances and challenges in developing a neuroSERM. Drug Dev Res 2002; 15:1–13. Thibault O, Porter NM, Chen KC, Blalock EM, Kaminker PG, Clodfelter GV, Brewer LD, Landfield PW. Calcium dysregulation in neuronal aging and Alzheimer’s disease: history and new directions. Cell Calcium 1998; 24(5–6): 417–433. Solodkin A, Van Hoesen GW. Neuropathology and functional anatomy of Alzheimer’s disease. In: Brioni JD, Decker MW, eds. Pharmacological Treatment of Alzheimer’s Disease: Molecular and Neurobiological Foundations. New York: John Wiley & Sons, 1997:15–45. Rebar RW. Premature ovarian failure. In: Lobo RA, ed. Treatment of the Postmenopausal Woman: Basic and Clinical Aspects. Philadelphia: Lippincott Williams & Wilkins, 1999:13–22. Hammond CB. Women’s concerns with hormone replacement therapy— compliance issues. Fertil Steril 1994; 62(6 suppl 2):157S–160S. American Society for Reproductive Medicine. HRT use and compliance [online], URL: http:==www.asrm.com=Patients=BCHRT.html. Grese TA, Pennington LD, Sluka JP, Adrian MD, Cole HW, Fuson TR, Magee DE, Phillips DL, Rowley ER, Shetler PK, Short LL, Venugopalan M, Yang NN, Sato M, Glasebrook AL, Bryant HU. Synthesis and pharmacology of conformationally restricted raloxifene analogues: highly potent selective estrogen receptor modulators. J Med Chem 1998; 41(8):1272–1283. Gustafsson JA. Therapeutic potential of selective estrogen receptor modulators. Curr Opin ChemBiol 1998; 2(4):508–511. Zhao L, Chen Q, Diaz Brinton R. Neuroprotective and neurotrophic efficacy of phytoestrogens in cultured hippocampal neurons. Exp Biol Med (Maywood) 2002; 227(7):509–519. O’Neill K, Chen S, Brinton RD. Impact of the selective estrogen receptor modulator, raloxifene, on neuronal survival and outgrowth following toxic insults associated with aging and Alzheimer’s disease. Exp Neurol 2004; 185(1):63–80.
8 The Use of PET to Evaluate Neural Plasticity and Repair in Rodent Brain Harley I. Kornblum Department of Pharmacology, University of California at Los Angeles, School of Medicine, Los Angeles, California, U.S.A.
Keith J. Tatsukawa Department of Pediatrics, University of California at Los Angeles, School of Medicine, Los Angeles, California, U.S.A.
S. Thomas Carmichael Department of Neurology, David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A.
Simon R. Cherry Department of Biomedical Engineering, University of California Davis, Davis, California, U.S.A.
1.
INTRODUCTION
Neural plasticity and repair are the subjects of intense scientific investigation. Many basic neuroscience studies are now devoted to the investigation of the mechanisms underlying the induction of long-term changes in brain structure and function under normal or pathologic conditions. There is great hope that these studies will, in turn, lead to new treatments for disease and injury. Most of the current research in neural plasticity and repair uses rodent models. These types of studies require the neuroscientist to be able to discern changes in brain function or structure as a result of interventions, which may be pathological or therapeutic. Such information can often only 159
160
Kornblum et al.
be obtained by sacrifice of the animal and histological or biochemical observation of the brain or electrophysiological study of brain slices. Obviously, these approaches have their limitations. First, terminal studies preclude longitudinal evaluation of individual animals. Changes that may occur as a result of an intervention have to be inferred by examining groups of animals prior to and at different times following intervention. The groups then must be statistically compared to each other, requiring the use of large numbers of animals. Furthermore, this type of investigation underestimates the significance of variation between animals, potentially missing distinctions between groups of responders and nonresponders, for example. Ideally, one would like to have the ability to make repeated observations in individual animals in order to accurately assess interanimal variability and to allow for a direct analysis of change over time due to a given intervention. A key advantage of in vivo imaging is the ability to perform multiple kinds of studies in the same animal. For example, one might wish to examine behavioral, structural, and neurochemical observations as the result of an intervention or gene mutation. These types of correlative studies would gain an enormous degree of power if all observations were made in the same animal. Such an ability is particularly advantageous in cases where an enonnous amount of time, effort, and money are invested in individual animals, such as animals that have had lesions placed followed by therapeutic transplantation of stem cells or treatment with a novel drug. Nonconsumptive methods of study are particularly suitable for the investigation of knockout and transgenic mice where the numbers of animals may be severely limited and there is a high motivation to obtain the maximal amount of information from an individual animal. Positron emission tomography (PET) has been used for a number of years in the study of the function and neurochemistry in the human and nonhuman primate brain (1). It is a noninvasive, and extremely safe imaging modality that allows for repeated observations in an individual. The adaptation of PET to repeatedly and noninvasively image small animals represents a unique opportunity to explore neuroplasticity and neuropathology in living rodents. This chapter describes some of the basic physical principles underlying PET imaging, review some of the work that has successfully utilized PET in rodent brain imaging and will indicate future directions for rodent research with PET.
2.
PRINCIPLES OF PET
This section reviews some of the fundamental principals of PET, for more detailed descriptions, see Ref. 2, among many other reviews. Positron emission tomography utilizes radiolabeled nuclides to obtain images of where that tracer is concentrated within the living animal. These radionuclides
The Use of PET to Evaluate Neural Plasticity and Repair
161
emit positrons as they decay. Some of these radionuclides are isotopes of elements common to all biological systems, such as carbon, nitrogen, and oxygen, enabling labeled tracers of small organic molecules to be produced by direct isotopic substitution, for example, by replacing a carbon-12 atom with a positron-emitting carbon-11 atom. These radionuclides are produced in a biomedical cyclotron (3). After the radionuclide is produced, rapid chemical synthetic pathways are needed to produce the radiolabeled product and it must then be immediately injected into the subject, with the time between production and injection typically being no more than 4–6 halflives. With carbon-11, this amounts to around 2 hr from production of radionuclide to injection of radiotracer. Despite these difficulties, due to its chemical advantages, carbon-11 remains the radionuclide of choice in PET for labeling receptor ligands and many drugs (4). There also positron-emitting radionuclides that have a long enough half-life such that they can be synthesized elsewhere and then distributed from regional facilities eliminating the need for a cyclotron at the research site. The most commonly used radionuclides is fluorine-18 [18F] which has a 110-min half-life. [18F] can be used to substitute for hydrogen atoms or hydroxyl groups in small molecules, and its longer half-life permits fairly complex synthetic pathways to be pursued resulting in a wider range of PET tracers. However, fluorene-substituted analogs may differ from their parent compounds in their chemical and biological activities. Because the tracer [18F] fluorodeoxyglucose (FDG) is widely used for clinical PET studies of metabolism in oncology, neurology, and cardiology, fluorine-18 and FDG have become easily and relatively cheaply available in many parts of the world. Even longer-lived compounds, such as copper-64 (half-life 12.6 hr) and iodine-124 (half-life 4.2 days) exist and a growing number of centers are taking advantage of this (5). The wide range of positron-emitting nuclides allows for the creation of an unlimited number of biologically interesting compounds for use as PET tracers, which, in turn, have been used to investigate many important physiological and cellular processes, such as metabolism, blood flow, enzyme kinetics, receptor density and cellular proliferation, to name a few (6). Positron emission tomography is an extremely sensitive method due to the relatively high likelihood of a single radioactive decay being detected due to the efficiency of detection and high specific activity of the tracers. Therefore, experiments are generally performed using tracers in nanogram or microgram amounts, creating concentrations too small to interfere with normal biological processes, increasing their potential use and safety. An exception to this rule would be in cases of performing receptor saturation experiments in very small animals, such as mice. As stated above, the synthesis of an appropriate labeled compound is followed by injection of the compound. For rodents, this is typically via the tail vein, although indwelling catheters can be used. Typical injected doses
162
Kornblum et al.
are in the range of 102–103 mCuries (mCi) in the mouse and rat, respectively (5). The radioactive atom will decay, releasing a positron, which then loses kinetic energy by collision with nearby electrons in tissue. The positron then undergoes an annihilation interaction with an electron, in which the mass of the electron and positron are converted into high energy (511 keV) photons (gamma rays) that are emitted 180 apart (Fig. 1). Due to their high energy, the photons have a high probability of leaving the body to be sensed by the scanner. A PET scanner consists of rings of scintillation detectors placed around the subject (Fig. 1). A detector on one side of the ring is electronically linked to its partner on the opposite side of the ring, so that only those high-energy gamma rays that reach the two detectors in coincidence will be counted as real events. Statistical reconstruction methods, applying appropriate corrections are then used to create an image, in which a map of positron-emitting density is obtained. A range of such image reconstruction techniques are available, from simple Fourier based methods (e.g., filtered backprojection) to more complex statistically based iterative methods that can contain sophisticated models of the entire imaging system and the data formation process (7). Modern PET scanners may consist of as many as 10–20,000 individual scintillator detector elements, capable of producing images with a resolution
Figure 1 Principles of PET. Positron-emitting radionuclides are prepared in a cyclotron and injected into the body, usually via the intravenous route. The positrons travel through the tissue, until they strike an electron and emit high-energy gamma rays at opposite angles. The event is detected as being ‘‘real’’ if two detectors on the opposite side of the ring are simultaneously hit. Software is then used to estimate, in space, the origin of the positron.
The Use of PET to Evaluate Neural Plasticity and Repair
163
as high as 3–6 mm in the human brain and 1–2 mm in the mouse and rat brain, Temporal resolution can be as high as a few seconds in some instances, allowing for the evaluation of changing distribution of a radiolabeled tracer over a period of minutes to hours. This dynamic data can then be used to quantitatively measure the rates of specific biologic processes, or to assess binding characteristics of the tracer to the target of interest (8).
3.
DEDICATED LAB ANIMAL PET SCANNERS
Since the early 1990s, PET has been used to study nonhuman primates for multiple reasons, including the evaluation of new drugs and imaging tracers for the CNS (4,9), monitoring potential therapies for neurodegenerative diseases as well as fundamental studies of receptor distribution. While many of these types of studies have been carried out using conventional clinical PET scanners, several dedicated scanners have been developed to achieve higher spatial resolution and=or sensitivity for imaging of nonhuman primates by PET (10,11). The use of primates in neuroscience, however, is greatly limited due to their expense, difficulty in manipulation, slow breeding, and potential for being disease vectors. In addition, the evolutionary proximity of primates to humans has prompted ethical questions regarding the use of monkeys and apes in laboratory research. The majority of laboratory neuroscientists utilize small rodents; rats and mice. Rodents are cheaper, easy to breed and, particularly in the case of the mouse, relatively easily and commonly subjected to genetic manipulation. Furthermore, there is a huge body of knowledge regarding rodent brain plasticity and repair. Therefore, for an imaging modality to become widely used by laboratory neuroscientists, it will have to be effective at imaging the brain of rats and mice. Positron emission tomography imaging in smaller laboratory animals poses many technical challenges, especially in the area of spatial resolution. The rat brain has a volume of approximately 2.8 cc, while that of the mouse is 0.7 cc. This is about 2000 times smaller than the human brain. Despite these daunting challenges, there has been a large effort to develop PET systems for rodents. The first small animal PET scanner developed at the Hammersmith Hospital in London in the early 1990s was based on the same detector technology found in clinical PET systems (12). The resolution of this scanner was likely to be insufficient to analyze tracers with relatively universal uptake, such as FDG, but proved to be highly useful for imaging terminals in the highly localized nigrostriatal dopamine system, to examine, for example, the effects of fetal transplantation in a model of Parkinson’s disease (13). Subsequently, several groups across the world initiated studies to develop higher spatial resolution PET systems designed with the specific
164
Kornblum et al.
Figure 2 Glucose metabolism in the rat brain. [18F]fluorodeoxyglucose was injected into the tail vein of a rat. After 45 min, the rat was anesthetized and placed in the scanner for acquisition. Scans were reconstructed using the MAP algorhithm. Tracer activity (and, hence, glucose metabolism) is seen as darker signal.
intent for imaging rodents. In 2000, the first commercial small animal PET scanners became available and now several companies are offering such systems (10,14). The most widely distributed system is currently the micro PET system manufactured by Concorde Microsystems (Knoxville, TN) based on a prototype originally developed at UCLA (15). Examples of images in the rat brain obtained with the original micro PET prototype are shown in Fig. 2, utilizing FDG as a tracer. As can be seen in the figure, despite the relatively ubiquitous uptake of the tracer, large brain structures can be readily distinguished, including the neostriatum, cortex, and thalamus.
4.
STUDIES OF GLUCOSE UTILIZATION AND NEUROPLASTICITY
With the development of scanners with sufficient resolution, PET could then be used to assay neuroplasticity in vivo. As stated above, the most commonly used and widely available PET tracer is FDG. This tracer is taken
The Use of PET to Evaluate Neural Plasticity and Repair
165
up into cells and phosphorylated by hexokinase. However, the metabolism of FDG cannot proceed further along the glycolytic pathway and the tracer is trapped within the cell. The location of cells that retain FDG can be determined by PET after normal blood flow washes away the untrapped tracer. Cells that are highly metabolically active will retain more tracer, due to higher activity of hexokinase and, potentially, glucose transport. In the brain, this metabolic activity is directly proportional to local synaptic activity (16). High levels of FDG uptake are only observed in gray matter regions—the location of synapses. However, recent evidence suggests that local glia, rather than the firing neurons, themselves, are actually responsible for a significant proportion of FDG uptake. Fluorodreoxyglucose is an analogue of 2-deoxyglucose (DG). The DG autoradiographic method has long been used as a measure of neuroplasticity following injury or physiological manipulation in the brain (e.g., Refs. 17,18). In general, one places an indwelling arterial catheter into the animal then injects the glucose during the awake, but restrained state. The local metabolic rate can be quantitatively determined if arterial samples are taken at periodic intervals. Following uptake of DG, which is radioactively labeled with carbon-14, the animal is killed and the brain rapidly frozen, cryosectioned and brain sections exposed to x-ray films. The density of silver grains on the film is proportional to the synaptic activity in that particular region. The results obtained from one animal can be compared to those from another animal, due to the quantitative nature of the study. One example is the expansion of surrounding barrel receptive fields following whisker ablation (18), When DG is injected during single whisker stimulation, there is a reliable increase in neocortical signal corresponding to the receptive field of that whisker. If the whiskers around the one being studied are surgically removed, then this receptive field, as visualized by DG autoradiography expands. Of course, one cannot use this method to establish the change of receptive field size for a particular animal. Differences are inferred from group means. Another example of the use of DG autoradiography in the analysis of neuroplasticity is in the case of large cortical lesions. Suction ablation of the neocortex results in a remarkable rearrangement of connections between the remaining (contralateral) cortex and subcortical structures in young animals but not in adults. This has been studied in rats, cats, and nonhuman primates. The extent of connection rearrangement correlates with changes observed in glucose metabolism as observed in cats (17). In order to assess whether there were age-dependent changes in glucose metabolism following cortical ablations in the rat, we used the UCLA micro PET scanner (prototype of the Concorde system) to perform FDGPET in animals that had been lesioned on the 6th day of life (P6) or in adulthood (19). We were able to scan adult animals 3, 10, and 30 days after injury. Due to limitations of resolution, animals lesioned on P6 were only scanned 1 month following lesion. We assessed the relative uptake of
166
Kornblum et al.
FDG in the neostriatum and thalamus on the side ipsilateral to the lesion and compared it to that on the contralateral side. By using this relative quantification method, we were able to avoid the placement of chronic indwelling arterial catheters, which would have been necessary to establish fully quantitative measures. Of course, the use of relative measures would obscure potential global effects of the lesion. We found that both thalamic and striatal metabolism diminished following a cortical lesion in adults and then recovered, with less recovery occurring in the thalamus than the striatum. Animals lesioned on P6 appeared to have improved recovery of metabolism in the neostriatum, as compared to adults, but worse thalamic recovery. These studies demonstrated several principals. First, we were able to reliably and reproducibly assess changes in glucose metabolism following a lesion in groups of animals. Second, it was demonstrated that microPET could assess timedependent changes in glucose metabolism in serial studies of individual adult animals following a lesion. Additionally, we observed age-dependent differences in response to lesion, consistent with prior studies using conventional histological methods. Using the results of these studies as a foundation, current studies are now directed towards microPET as a platform to test potential therapies, such as neurotrophic factor or neural stem cell implantation. It is our hypothesis that effective therapies will result in improved glucose metabolism following injury. If such turns out to be the case, then future clinicalstudies will be able to utilize FDG-PET as an outcome measure. One example of such an ‘‘interventional’’ study consists of the administration of basic fibroblast growth factor (bFGF) via a localized source following cortical injury (Fig. 3). Animals underwent cortical lesions at 6 and 12 days of age and as adults. As predicted from the previous study, metabolism in the neostriatum and thalamus was depressed compared to the normal hemisphere at all ages, as exemplified on P12 in Fig. 3. However, treatment with 10 mg of bFGF impregnated in Gelfoam resulted in improved metabolism of the thalamus and neostriatum only in animals treated on P12 (as shown by the reduced percentage deficit). These data indicate that while neurotrophic factor therapy may have some efficacy, this efficacy is limited to special ‘‘vulnerable’’ periods of development. Further behavioral and anatomical studies are being performed on these animals. We have also used microPET to study metabolic depression, or diaschisis following a more clinically relevant cortical lesion, i.e., a small, focal stroke. It has long been known, using clinical PET studies, that stroke victims have depressed glucose metabolism in large areas of cortex, as well as connected subcortical structures following stroke (20). As shown in Fig. 4, we have found, using serial microPET studies in rats, that focal stroke produces a very large region of hypometabolism in surrounding cortical areas (Fig. 4A) as well as the neostriatum and thalamus, the size of which is reduced 1 week later (Fig. 4B). When compared to the deficit in blood flow observed 1 day after lesion (Fig. 4C) or to the ultimate size of the stroke
The Use of PET to Evaluate Neural Plasticity and Repair
167
Figure 3 Fibroblast growth factor treatment enhances glucose metabolism following cortical injury in P12 rats. A large lesion was made in the neocortex of P12 rats using a suction catheter, as described by Kornblum et al. Fibroblast growth factor or saline-coated gelfoam was placed in the lesion cavity. The rats then underwent FDGPET 1 month later (left) and the relative metabolism determined in the thalamus (black arrows) and the neostriatum (white arrows) and expressed as the percent deficit (right). A greater percent deficit indicates worse recovery. Note that the FGFtreated group had significantly less deficit. (From Ref. 19.)
Figure 4 Effects of a small cortical stroke on glucose metabolism. Rats underwent a small stroke in the barrel cortex, and then scanned using FDG 1 and 8 days later. Regional bloodflow was determined in a separate group of animals 1 day after stroke (C). The 1- and 8- day PET scan is shown for the same animal in (A) and (B) with the area of dramatic hypometabolism delineated by the black lines. The ultimate stroke is demonstrated using Cresyl violet in (D).
168
Kornblum et al.
(Fig. 4C), the region of hypometabolism is much larger. This region of hypometabolism corresponds to alterations in gene expression and cellular response, such that it may be a marker for poststroke neuroplasticity. Current studies will then test whether the region of hypometabolism will provide improved support for transplantation. Similarly, microPET has also been used to study responses to blunt trauma. Clinical studies have shown that PET performed within hours of severe trauma reveals a significant degree of hypermetabolism around the area of injury, while subsequent scans show hypometabolism (21). In animals, autoradiographic studies also demonstrate initial hypermetabolism followed by hypometabolism following blunt head trauma. Moore et al. (22) examined glucose metabolism in vivo using the microPET system following blunt head trauma at a time expected to reveal ipsilateral hypometabolism and, indeed, found that this was true. These studies, in which small animal PET results could be directly compared to autoradiography following a lesion, demonstrate the reliability of PET in small animal models of brain injury. Fluorodeoxyglucose as a measure of synaptic activity is not the only means by which pathological changes or neuroplasticity can be monitored by PET. As mentioned above, there are a virtually unlimited number of available and potential PET tracers that can be used to monitor neurochemical processes. One of the most frequent systems assayed by PET is the nigrostriatal dopamine system. Specific compounds exist to evaluate dopamine synthesis, reuptake sites, and receptors. Thus, alterations in the functional status in the nigrostriatal system can be assayed in vivo using PET. Clinically, the tracer 18F-DOPA has been used to distinguish idiopathic Parkinson’s disease from other disorders with similar phenotypic features.I8F-DOPA PET has also been used to monitor cell transplantation, with an improvement in uptake of I8F-DOPA generally correlating with functional improvement (23). In rodents, this system is also readily imaged using the currently available small animal scanner. The loss of dopamine following a unilateral 6-hydroxydopamine lesion can be detected using either 18 F-DOPA or compounds that label reuptake sites. Functional improvement following intervention, for example, stem cell transplantation, is mirrored by enhancement of the PET signal (24). Likewise, alterations in the function of other transmitter systems, using analogous compounds.
5.
OTHER POTENTIAL APPLICATIONS OF PET IN NEUROPLASTICITY
The broad-range compounds that are potentially useful as PET tracers allows for the assay of numerous biological processes, in addition to those described above. Pharmacological parameters such as receptor and
The Use of PET to Evaluate Neural Plasticity and Repair
169
transporter affinity, density, and subtype can be assayed (6). Neurochemical processes such as transmitter synthesis, enzyme concentration, and activity can be assayed. Additionally, compounds exist to measure cellular proliferation (25), although their use in normal brain have not been validated. Recently, PET has been used to image virally introduced genes. One example of this imaging of gene expression is the use of 18F-labeled ganciclovir analogues to visualized HSV thymidine kinase that is virally introduced into host animals as a ‘‘PET reporter gene’’ (26,27). Clinical studies, using this technology, are underway to utilize PET to monitor gene therapy. Additional reporter genes have been used to image reporter genes, including labeled spiperone to visualize transduced dopamine D2 receptors. Based on this wide variety of capabilities, one could theoretically envision the use of PET to assay numerous events in neuroplasticity. For example, one might use wild-type animals to image the induction of enzymes or receptors following induction of LTP. Additionally, one might use transgenic animals in which a PET reporter gene is transcriptionally regulated by a promoter of interest to examine gene induction in vivo under a variety of circumstances.
6.
CURRENT LIMITATIONS FOR THE USE OF PET IN STUDIES OF RODENT CNS PLASTICITY
One of the keys for the use of PET in small animal models is the accurate identification of brain structures. When tracers are taken up to a greater or lesser extent in all or most brain areas, relatively large structures can be easily identified and region of interest (ROI) analysis can be performed with some ease. However, for tracers which are taken up by one or a handful of brain structures, or in order to standardize the definition of structures, other strategies must be developed. One strategy for the identification of brain structures utilizes the relatively constant size of structures within the adult rodent brain. Stereotactic brain atlases have long been used accurately place electrodes, micropipettes, or surgical lesions into relatively small structures of the adult rodent brain. Several groups are developing analogous atlases to ascribe specified PET signal to a specific brain area. For more accurate identification of structural detail, PET can also be combined with other imaging modalities, such as CT or MRI to provide more precise anatomical localization of the biochemical processes visualized by PET. Currently, there are clinical scanners in use that combine PET with CT, In studies of mouse tumor models, CT and PET have been combined, using image registration protocols to align the images from different scanners (28). Small animal PET and MRI images have been aligned in studies of the effects of stem cell transplantation on the nigrostriatal dopamine system (24). For large structures, such as the neostriatum, currently between
170
Kornblum et al.
scanner registration protocols may be adequate. However, precise registration to smaller brain structures will likely require the use of an individual multimodal scanner. Regardless of the methodology to which PET and brain structure can be superimposed, accurate identification of small structures will be remained highly challenging. Since all PET scanners have and will have a limit of resolution that is much greater than the size of small nuclei, the PET images derived from specific uptake in that nucleus or brain area will be larger than the structure in question. Another way of looking at this problem is that the image signal superimposed over any one brain area will be the result of signals from that brain area, as well as from closely adjacent structures. Since the rodent brain contains very little white matter separating adjacent structures, this problem can be quite significant. For example, measurements made in the lateral part of the neostriatum will reflect values from the nearby neocortex, and vice versa. As animal scanners achieve greater resolution, such as in the microPET II (29), investigators will be able to accurately identify signal from smaller and smaller nuclei with confidence, and imaging in the mouse brain will be made feasible, even with ubiquitous tracers such as FDG. An additional, related limitation placed by the relatively limited resolution and sensitivity of rodent PET scanners is the ‘‘partial volume effect.’’ Signal from a very small brain area (less than 50% of the scanners linear resolution) may not be detectable, due to the averaging out of the signal over a larger brain area. An example of this is in the analysis of glucose metabolism in the hippocampus. The hippocampal layers are readily detectable when FDG is injected and the brain imaged by film autoradiography, indicating the relatively high metabolic rate of the neurons in this structure. However, because the neuronal layers of the hippocampus are densely packed and occupy a small volume compared to the resolution of the microPET scanner, the hippocampal signal is not detectable (Fig. 5A, arrow and Ref. 19). Some of this effect can be mitigated with a boost of signal from the structure. When the hippocampus is stimulated to a supraphysiological extent, as with seizure activity, the signal from the hippocampus can be detected (Fig. 5B, arrow and Ref. 19), although the principal hippocampal layers still cannot be resolved, due to partial volume effect. Since mice are rapidly becoming the laboratory animal model of choice in most neuroscience laboratories, functional imaging will need to be adapted to image the mouse brain, which is considerably smaller than the rat brain. The limited resolving capacity of even dedicated small animal PET scanners has made mouse brain imaging a challenging prospect. While highly specific tracers in relatively large structures, such as the neostriatum, can be readily observed, even large structures in FDG scans or smaller structures in scans utilizing other tracers cannot be resolved. Again, as newer, higher resolution
The Use of PET to Evaluate Neural Plasticity and Repair
171
Figure 5 Detection of hippocampal metabolic activity during a seizure. A rat underwent a baseline FDG-PET scan (A), demonstrating an absence of signal in the hippocampus (arrow). The same animal was then injected with FDG 10 min into a kainic acid-induced seizure. The seizure was stopped by ketamine–xylazine anesthesia and the animal placed in the scanner. Note the now easily detectable activity in the hippocampus (B; arrow). The color scale in (A) is the same for both images.
scanners become available, more imaging of mouse brain will be performed with greater accuracy (19). One drawback to the use of FDG specifically, is that the level of uptake reflects the total amount of neural activity during the entire uptake time, which is typically 20–45 min. Thus, minute-to-minute changes are not readily imaged. This is not a concern when there is a constant, strong stimulus or if global activity during the awake state is the subject of study. However, some studies may require the measurement of neural activity with a much greater time resolution. Potentially, this requirement can be met through the use of labeled water to measure blood flow, a reflection of local neuronal activity. However, rodent PET scanners, to date, have not had sufficient sensitivity to perform these studies. In addition to limitations imposed by resolution, an additional considerable limitation to the use of PET in rodents is currently expense. Small animal PET scanners are commercially available at costs of several hundred thousand dollars. Scanners must be installed in appropriately designed space to eliminate radiation exposure. Appropriately trained personnel must be available to operate the scanner. If compounds are being made or developed by the investigator, then there will need to be access to a cyclotron, and radiochemical facilities. Thus, currently, animal PET scanning is most efficiently performed in centers where PET is already performed.
7.
COMPARISON OF PET TO OTHER IN VIVO FUNCTIONAL IMAGING MODALITIES
Initial applications of PET in experimental neuroscience were for mapping brain activity in humans—ascribing functions to particular brain regions.
172
Kornblum et al.
While the signal obtained with PET in many instances is quite robust, causing PET to still be put to use for this function, PET has been largely replaced by functional magnetic resonance imaging (fMRI). Functional MRI has much greater time resolution than PET, even when shortlived tracers, such as 15O-labeled water are used, and allows for a greater range of tests in a given subject, without the need for radiation. Functional MRI has been used with some success in the rat to visualize ischemia and plasticity following ischemia (30). However, there are limitations for the use of fMRI in the rat brain as well, including the need to anesthetize or heavily restrain the animal for accurate imaging as well as the limited sensitivity and high expense of the instrument. However, it is likely that MRI will largely supplant PET in the study of neural activity in the rodent brain. Currently, MRI is not widely used to assay receptor-driven or many other biochemical processes. An additional method for the analysis of functional activation lies in optical imaging of blood flow (see Ref. 31). Changes in neuronal activity produce microscopic changes in blood flow, which can be imaged through windows in the bone. As with MRI, however, the animal needs to be kept still, usually through the use of anesthetics. Additionally, only activity near the cortical surface can be assayed using this method. Other methods for imaging a variety of compounds using light or fluorescence are also in use or being developed. These methods, while holding great promise currently lack the resolution of PET. Neurochemistry can be assayed in a noninvasive manner using MR spectroscopy (see Ref. 32). This method utilizes specific magnetic resonance spectra to evaluate the levels of biologically important compounds within specified brain areas. Examples include glutamate, aspartate, and GABA. Magnetic resonance spectroscopy can be performed using standard MRl scanners. As with other MR methods, the animal must be kept very still, which requires sedation during image acquisition. Additional limitations in MR spectroscopy include the number of potential molecules that can be reasonably detected in addition to relatively limited spatial resolution, given current technology. Thus, despite a wide array of useful imaging tools, rodent PET still has some advantages over other methods, in that virtually any biochemical process can be imaged with reasonable resolution, at least in the rat brain. Often the tracer uptake can be performed during the awake, unrestrained state. Pharmaceuticals can be directly tested in an efficient manner using PET in preclinical studies. Efficacy for in vivo receptor binding assays and other functions of a variety of molecules can be directly compared. Future rodent PET studies will likely focus less on metabolic studies and more on biochemical and pharmacological analyses.
The Use of PET to Evaluate Neural Plasticity and Repair
173
REFERENCES 1. 2.
3. 4. 5. 6.
7. 8.
9. 10. 11. 12.
13.
14. 15. 16. 17. 18. 19. 20. 21.
Cherry SR, Phelps ME. Brain Mapping: The Methods. In: Toga AW, Mazziotta JC. San Diego: Academic Press, 1996:191–221. Cherry SR, Phelps ME. In: Sandler MP, Patton JA, Coleman RE, Gottschalk A, Wackers FJT, Hoffer PB, eds. Diagnostic Nuclear Medicine. Vol.1. Baltimore: Williams & Wilkins, 1996:139–160. McCarthy TJ, Welch MJ. Semin Nucl Med 1998; 28:235–246. Fowler JS, Volkow ND, Wang GJ, Ding YS, Dewey SL. J Nucl Med 1999; 40:1154–1163. Cherry SR, Kornblum HI. In: van Bruggen N, Roberts T, eds. Biomedical Imaging in Experimental Neuroscience. New York: CRC Press, 2003:271–290. Fowler JS, Wolf AP. In: Phelps, M., Mazziotta, J, Schelbert H, eds. Positron Emission Tomography and Autoradiography. New York: Raven Press, 1986: 391–450. Qi J, Leahy RM, Cherry SR, Chatziioannou A, Farquhar TH. Phys Med Biol 1998; 43:1001–1013. Huang S-C, Phelps ME. Phelps M, Mazziotta J, Schelbert H, eds. Positron Emission Tomography and Autoradiography. New York: Raven Press, 1986:287–346. Burns HD, Hamill TG, Eng WS, Francis B, Fioravanti C, Gibson RE. Curr Opin Chem Biol 1999; 3:388–394. Tai Y-C, Chatziioannou A, Siegel S, Young J, Newport D, Goble RN, Nutt RE, Cherry SR. Phy Med Biol. In Press. Watanabe M, Okada H, Shimizu K, Omura T, Yoshikawa E, Kosugi T, Mori S, Yamashita T. IEEE Trans Nucl Sci 1997; 47:1277–1282. Bloomfield PM, Rajeswaran S, Spinks TJ, Hume SP, Myers R, Ashworth S, Clifford KM, Jones WF, Byars LG, Young J, Andreaco M, Williams CW, Lammertsma AA, Jones T. Phys Med Biol 1995; 40:1105–1126. Hume SP, Lammertsma AA, Myers R, Rajeswaran S, Bloomfield PM, Ashworth S, Fricker RA, Torres EM, Watson I, Jones T. J Neurosci Methods 1996; 67:103–112. Jeavons AP, Chandler RA, Dettmar CAR. IEEE Trans Nucl Sci 1999; 46: 468–473. Chatziioannou AF, Cherry SR, Shao Y, Silverman RW, Meadors K, Farquhar TH, Pedarsani M, Phelps ME. J Nucl Med 1999; 40:1164–1175. Bruehl C, Witte OW. Ann Neurol 1995; 38:414–420. Hovda DA, Villablanca JR, Chugani HT, Phelps ME. J Cereb Blood Flow Metab 1996; 16:134–146. Dietrich WD, Ginsberg MD, Busto R, Smith DW. J Neurosci 1985; 5:874–880. Kornblum HI, Araujo DM, Annala AJ, Tatsukawa KJ, Phelps ME, Cherry SR. Nat Biotechnol 2000; 18:655–660. Carmichael ST, Tatsukawa KJ, Katsman D, Tsuyuguchi N, Kornblum HI. Stroke. In Press. Bergsneider M, Hovda DA, Shalmon E, Kelly DF, Vespa PM, Martin NA, Phelps ME, McArthur DL, Caron MJ, Kraus JF, Becker DP. J Neurosurg 1997; 86:241–251.
174
Kornblum et al.
22. Moore TH, Osteen TL, Chatziioannou TF, Hovda DA, Cherry TR. J Cereb Blood Flow Metab 2000; 20:1492–1501. 23. Spencer DD, Robbins RJ, Naftolin F, Marek KL, Vollmer T, Leranth C, Roth RH, Price LH, Gjedde A, Bunney BS et al. N Engl J Med 1992; 327:1541–1548. 24. Bjorklund LM, Sanchez-Pernaute R, Chung S, Andersson T, Chen IY, McNaught KS, Brownell AL, Jenkins BG, Wahlestedt C, Kim KS, Isacson O. Proc Natl Acad Sci USA 2002; 99:2344–2349. 25. Barthel H, Cleij M, Collingridge D, Hutchinson O, Osman S, He Q, Luthra S, Brady F, Price P. Cancer Res 2002; 63:3791–3798. 26. Gambhir SS, Barrio JR, Phelps ME, Iyer M, Namavari M, Satyamurthy N, Wu L, Green LA, Bauer E, MacLaren DC, Nguyen K, Berk AJ, Cherry SR, Herschman HR. Proce Nat Acad Sci USA 1999; 96:2333–2338. 27. Gambhir SS, Herschman HR, Cherry SR, Barrio JR, Satyamurthy N, Toyokuni T, Phelps ME, Larson SM, Balatoni J, Finn R, Sadelain M, Tjuvajev J, Blasberg R. Neoplasia 2000; 2:118–136. 28. Goertzen A, Meadors A, Silverman R, Cherry S. Phys Med Biol 2002; 47: 4315–4328. 29. Tai Y, Chatziioanou A, Yang Y, Silverman R, Meadors K, Siegel S, Newport D, Stickel J, Cherry S. Phys Med Biol 2003; 48:1519–1537. 30. Lythgoe MF, Thomas DL, Calamante F. In: van Bruggen, N, Roberts T, eds. Biomedical Imaging in Experimental Neuroscience New York: CRC Press, 2003:21–54. 31. van Bruggen N, Roberts TP. In: van Bruggen N, Roberts T, eds. Biomedical Imaging in Experimental Neuroscience. New York: CRC Press, 2003:323–335. 32. Williams SR, Sibson NR. In: van Bruggen N, Roberts T, eds. Biomedical Imaging in Experimental Neuroscience. New York: CRC Press, 2003:237–270.
9 Use of Estrogens as Neuroprotectants in Stroke and Alzheimer’s Disease James W. Simpkins, Shao-Hua Yang, and Yi Wen Department of Pharmacology and Neuroscience, University of North Texas Health Science Center, Fort Worth, Texas, U.S.A.
1.
INTRODUCTION
In the United States alone, there are over 750,000 new strokes per year, making it the leading cause of neurological disability and a major cause of death (1,2). Despite these large personal and health care system costs of stroke, there is currently only a single therapy that attempts to reduce the damaging effects of stroke, the anticoagulant, t-PA (3–5). Desperately needed, but currently not available, are therapies that can prevent or reduce brain damage during cerebral ischemia, caused either by local interruption of brain blood flow or by global reduction in blood flow as a result of myocardial infarcts. Additionally, Alzheimer’s disease (AD) currently affects 4 million Americans and it is estimated that more than 14 million Americans will be affected by 2030 (6). There are currently no effective therapies to slow the progression or to prevent AD. In the present chapter, we will review the literature related to the potential use of estrogens in neuroprotection during and following cerebral ischemia, as well as in AD. This review is undertaken at this time, as there are only a few published reports on the use of estrogens in stroke protection in experimental models and there is limited clinical data on the subject. A similar situation exists for estrogen use and AD. However, there is now a compelling scientific rationale for the initiation of clinical trials of estrogens for the 175
176
Simpkins et al.
treatment of brain damage related to cerebral ischemia and there are currently ongoing trials for estrogen’s and AD prevention. We will review the basic and clinical literature indicating the potential of acute estrogen administration as a neuroprotective therapy for ischemia and as a potential treatment for AD. To achieve this aim, we will review the following: physiology and pharmacology of estrogens, current clinical use of estrogens, the discovery that chemical modification in estrogens can enhance their neuroprotective potencies while reducing their estrogenicity, describe published and unpublished data indicating that estrogens are efficacious against brain damage caused by cerebral ischemia, and finally, identify additional basic, translational and clinical studies that are needed to critically assess the potential for the use of estrogens as a neuroprotectant in stroke and AD. 2.
ESTROGEN PHYSIOLOGY AND PHARMACOLOGY
Estrogens were first desribed by Frank et al. (7) as a component of the ovary needed for reproductive function. Later, estradiol was isolated and crystallized by Doisy et al. (8). Because of the initial chemical identification of estradiol, more than 16,000 estrogen-related compounds have been synthesized and this class of compounds represents one of the most study groups of compounds known. In all mammals, estrogens serve a variety of functions related to reproductive development and function (9). Estrogens cause sexual differentiation of certain sexually dimorphic brain regions largely related to reproductive behavior and play a major role in the regulation of the cyclic secretion of gonadotropins and, in song birds, seasonal singing in males. Additionally, estrogens are essential for normal estrous=menstrual cycle secretion of gonadotropins, for female body physiques, in females for the closure of epiphyseal plates in long bones at puberty, for the growth of the uterine endometrium during the follicular phase of the menstrual cycle, and for the stimulation of ductal cell in mammary tissue. In addition to these well-known reproductive functions of estrogens, these molecules have a variety of other physiological effects that are less well known. Among these are effects on plasma lipids (10) and carbohydrates (11), maintenance of normal bone mineralization (12), effects on body temperature regulation (13), and effects on food consumption and food selection (14). 3.
CURRENT CLINICAL USES OF ESTROGENS
Pharmacologically, estrogens are used in therapies aimed at prevention of pregnancy as a component of birth control pills (9), in postmenopausal estrogen therapy (ET) or as an essential component of postmenopausal hormone therapy (HT) (9), and for the use in the treatment of a few kinds of otherwise untreatable cancers (9). As a component of birth control pills,
Estrogens as Neuroprotectants
177
ET and HT, estrogens are among the most prescribed compounds known, with literally billion of prescription written each year. Additionally, there are hundred of millions of women years of experience in the use of estrogens in postmenopausal ET and HT. As such, we have better knowledge of the pharmacology and toxicity of estrogen than any other class of drugs. Dozen of preparations of birth control pills are currently marketed in the United States. All of these preparations use synthetic estrogens and progestins, as naturally occurring estrogens and progestins are extensively metabolized in the first pass through the liver and therefore are not readily bioavailable. In the United States, the most commonly used ET is PremarinÕ , a conjugated equine estrogen preparation, and the most common HT is PremProÕ , a combination of Premarin and medoxyprogesterone acetate.
4.
EVIDENCE FOR EFFICACY OF ESTROGENS AND ESTROGEN ANALOGS IN STROKE NEUROPROTECTION
There is a growing interest in the actions of estrogens as neuroprotectants against neurodegenerative diseases and acute brain damage caused by stroke. A possible role of endogenous female hormones as neuroprotectants against global cerebral ischemia-reperfusion injury was suggested by studies demonstrating that intact adult female rodents sustain less neuronal damage when compared with age-matched males (15). Our laboratory first demonstrated that 17b-estradiol is a potent neuroprotectant in vitro (15) and is very effective against ischemia-induced brain damage (16). There is now abundant evidence for neuroprotection by estrogens both in vitro and in vivo. Protective effects of estrogen have been widely reported in different types of neuronal cell against different toxicities, which mimic cerebral ischemia in vitro, including serum deprivation, oxidative stress, and excitotoxicity (17). Multiple independently lethal mechanisms are involved in cerebral ischemia-induced neuronal death. Estrogens have been identified as multifaceted hormones and antagonize many aspects of cascade induced by cerebral ischemia. Antioxidant effects of the steroid (18) and attenuation of NMDA-receptor activation (19) have been implicated as mechanisms for the neuroprotective effects of estrogens. Two of the major signal pathways, ERK and PI3K-Akt, have been well characterized as being able to mediate inhibition of apoptosis and support neuronal survival. Both signal pathways have been demonstrated to be activated by estrogens (20,21). Mitochondria produce most of the cellular ATP by oxidative phosphorylation and generate most of the endogenous oxygen radicals as a toxic by-product. In addition, mitochondria are central in the regulation of apoptosis, calcium homeostasis, and cytoplasmic redox state. Several studies have also shown that estrogens may exert direct or indirect effects on mitochondria function. Estradiol can protect against ATP depletion, mitochondrial membrane
178
Simpkins et al.
potential decline, and the generation of reactive oxygen species, induced by the mitochondrial toxin, 3-nitropropionic acid (22), or the pro-oxidant, H2O2 (Wang and Simpkins, unpublished observations). 17b-Estradiol can modulated mitochondrial calcium influx and increase Bcl-2 expression (23). In in vivo studies, the neuroprotective effects of estrogens have been demonstrated in all acute cerebral ischemia scenarios. Cerebral ischemic stroke can be either transient (occlusion followed by reperfusion) or essentially permanent. Specific brain regions can be affected, as occurs during an arterial occlusion, or the entire brain can become globally ischemic, as occurs during a cardiac arrest. The neuroprotective effects of estrogens have been demonstrated in a variety of stroke models by different laboratories, including transient and permanent middle cerebral artery occlusion models (16,24,25), global forebrain ischemia models (26), photothrombotic focal ischemia models (27), and glutamate-induced focal cerebral ischemia models (28). The protective effects of estrogens have been documented in rats, mice, and gerbils (16,29,30). Estrogens’ neuroprotective effects have been demonstrated in adult female, middle-aged female, and reproductively senescent female rats (31). Further, these effects of estrogens have been shown despite the presence of diabetes and hypertension (32,33). The neuroprotective effects of estrogens have been demonstrated in a subarachnoid hemorrhage model, which is highly prevalent in females (34). Finally, the neuroprotective action of estrogen is not limited to female, as protective action of estrogens is also seen in males (35,36). Collectively, these results indicate that estrogens could be the valuable candidates for the treatment of acute stroke. A large concentration range of estrogens, from low physiological to high pharmacological concentrations, has been shown to produce protective effects in stroke models. No neuroprotection was afforded by physiological level of estradiol administered at the onset of ischemia (25) although neuroprotective effects of estrogens were clearly demonstrated by the acute treatment, even post-treatment, with pharmacological doses of estradiol (16,36,37). The therapeutic window of estrogens at the dose of 100 mg=kg could last up to 3 hr after insult (38). Recently, we demonstrated that the protective action of estrogens was dosedependent. The therapeutic window of estrogens could extend up to 6 hr after ischemic insult at the dose of 500–1000 mg=kg (Yang and Simpkins, unpublished observations). This is clinically relevant because the 6-hr therapeutic window could give physicians time sufficient for diagnosis of stroke and for the application of other therapeutic intervention, such as r-tPA. Additionally, the therapeutic window for estrogen neuroprotection could be insult severity-dependent and could be different between different species. However, it has been shown that the infarct penumbra, which can be protected, develops over a longer period in human subjects than in rodents (39). Therefore, it is reasonable to predicate that estrogens could have a longer therapeutic window in human than in rodents.
Estrogens as Neuroprotectants
179
Estrogen receptors (ERs) have been identified as nuclear receptors. Classical (genomic) actions of estrogens are mediated by the binding of the steroid to the nuclear ERs, and the binding of the steroid–receptor complex to the ER response element thereby activates transcriptional events. In the past two decades, growing evidence indicates nongenomic steroid action. In contrast to genomic action, nongenomic action is principally characterized by its rapid onset of action and insensitivity to inhibitors of transcription and protein synthesis (40). In as much as ERs are distributed throughout central nervous system (CNS), three possible mechanisms could contribute to the neuroprotective action of estrogens in vivo:(1) a genomic ER-mediated action, (2) a receptor-dependent non-genomic action such as prosurvival signaling activation, and=or (3) a receptor-independent nongenomic action. The different therapeutic window for the physiological and pharmacological dose of estrogens suggests that different dosages afford neuroprotective action through different mechanisms. Evidence from both in vitro and in vivo studies suggests that the neuroprotective effects of estrogens do not require ER-dependent gene transcription. It has become evident that estrogens, and non-feminizing estrogen analogs exert neuroprotective activities in neurons. 17b-E2 and nonfeminizing estrogen analogs, such as 17a-E2 and complete enantiomer of 17b-E2, can preserve mitochondrial function, cell viability, and ATP levels in human lens cells during oxidative stress, which did not block by ICI182,780, an ER antagonist. Surprisingly, ICI182,780 alone increases the cell survival (41). Several other synthesized estrogen analogs have also been reported to possess neuroprotective properties (42–44). Structure– activity relationship studies indicate that the most essential structural motif that elicits neuroprotective functions is the phenolic A ring of the steroid, rather than its ability to bind to ERs (42). Estrogen receptors, especially ERa, have been found to play an important role in the reproductive function in both female and males (45). Although ERs have been found throughout the CNS, no phenotypic change in CNS has been found in both ERa and ERb knockout mice, suggesting that ER-mediated effects of estrogens have relatively little role in CNS development (46). The rapid onset of neuroprotective action induced by estrogens indicates that the neuroprotective actions are unlikely mediated through genomic mechanisms. The neuroprotective actions against stroke have been demonstrated not to be limited to estrogens but are also seen with selective ER modulator, such as tamoxifen (47,48) and LY353381 (49). Further, neuroprotective actions of other nonfeminizing estrogen analogs, such as 17a-E2 (16), complete enantiomer of 17b-E2 (44), an adamantyl estrogen analog (50), have been also demonstrated in stroke models. The actions of these nonreceptor binding estrogen analogs indicated that ERs are not required for the neuroprotective effects of estrogens.
180
Simpkins et al.
A plethora of data supports a direct neuroprotective role against ischemic stroke for estrogens. Given the expected clinical safety of acutely administered estrogens, ET may be useful in treating acute cerebral ischemia. Further, the efficacy of nonfeminizing estrogen analogs suggests that these compounds may be clinically useful for treating neuronal death in men or women for whom ET is contraindicated. The wide therapeutic window of estrogens made them candidates for neuroprotection in stroke. Other neuroprotective agents have been developed and tested for almost all components of the ischemic cascade. However, the results of the neuroprotective clinical trials to date have been discouraging. It should be noted that most of the tested neuroprotective agents target a specific component of the ischemic cascade, which suggests that any single drug that interferes with a specific event in the ischemic cascade will not have a large clinical impact (51). Protection of neurons during ischemia=reperfusion will require a polypharmacy approach or use of a compound that have pleotrophic actions and appears to affect multiple neurotoxic processes. Encouragingly, estrogens have been proved to be multifacet hormones modulating many aspects of neuronal cascades induced by cerebral ischemia. The pathological mechanisms that are activated during stroke, such as oxidative stress, free radical activity, excitotoxicity, inflammatory response, mitochondrial dysfunction, and apoptosis, are antagonized by estrogens. The protective effects of estrogens during reperfusion made them the candidates of neuroprotectants against reperfusion injury induced by thrombolysis. It could be possible to prolong the therapeutic time window for successful thrombolysis by administration of estrogens before, during, or after the infusion of IV r-tPA. This strategy should be effective over the short term in producing pharmacotherapies that can reduce ischemic brain damage and the resulting neurological deficits. 5.
TRANSIENT CEREBRAL ISCHEMIA AS A MODEL FOR ALZHEIMER’S DISEASE NEUROPATHOLOGY
Epidemiological evidence indicated an increased prevalence of dementia among stroke patients. In patients over 60 years of age, the prevalence of dementia is ninefold higher than controls 3 months after an ischemic attack (52). Among patients who had their first cerebral infarct without previous dementia, the incidence of dementia in the first year is nine times greater than expected (53). Four years after a first lacunar infarct, 23% of patients develop dementia, 4–12 times more than controls (54). Brain damage from the initial infarct is considered as the direct cause of <50% of the dementia (52). Many of the poststroke dementia have a progressive onset, suggesting a degenerative process rather than a vascular event (53,55,56). Alzheimer’s disease is the most frequent cause of degenerative dementia (57). Interestingly, AD and stroke have many common genetic risk
Estrogens as Neuroprotectants
181
factors. The e 4 allele of the apolipoprotein E gene (APOE 4) is associated with a higher risk of ischemic stroke or coronary heart diseases (58,59) and dementia (59,60). The APOE e 4 allele has also been firmly established as a major risk factor for late-onset AD (58,61,62). It is also known that amyloid precursor protein accumulates in the regions of neurodegeneration following focal cerebral ischemia in the rat (63,64). On the other hand, Alzheimer-type pathology is often associated with cerebral amyloid angiopathy (65) and with infarcts (66). Alz-50-immunoreactive granules are found around cerebral infarction after a stroke (67). Therefore, the link between stroke and AD seems to be closer than expected by chance. Another cellular event, apoptosis, also links neurodegenerative diseases and stroke. Studies of the pathogenic mechanisms implicate apoptosis in agerelated neurological disorders including AD and Parkinson’s disease (68–70). The pathology involves synaptic degeneration, senile plaques, neurofibrillary tangles (NFTs), and death of neurons in limbic structures, including the hippocampus and the cerebral cortex. Substantial apoptosis was identified in AD and other neurodegenerative diseases in neurons containing NFTs and associated with amyloid deposits (71,72). Similarly, focal ischemic stroke
Figure 1 Immunohistochemical staining of Tg3 in the frontoparietal cortex of female rats, which received 1-hr MCA occlusion and 24-hr reperfusion. This antibody detects the NFT-like phosphorylation and conformational change in tau.
182
Simpkins et al.
results in brain damage with an ischemic core region and a surrounding ischemic penumbra (73). A number of proapoptotic factors appear to serve a similar role in promoting neuronal death, which include over-activation of glutamate receptors, calcium overload, and increased reactive oxygen species. In addition, complex cytokine cascades, involving microglial activation and the cerebrovasculature, are implied in both diseases. In our laboratory, we found that the transient focal ischemia induced a profound hyperphosphorylation and a conformation change in tau, both of which are similar to the tau hyperphosphorylation and NFTs seen in AD brain (Fig. 1). Additionally, the neuroprotective agent, 17-estradiol, substantially reduces a variety of tau phospho-epitopes in the ischemic cortex (Wen and Simpkins, unpublished observations). The relationship between ischemic stroke and the formation of NFTs is not clear, and potential mechanisms are only now being investigated. Emerging evidence from cell signaling transduction and biochemical studies suggests that a neuronal cdc2-like kinase, cyclin-dependent kinase 5 (cdk5), may bridge the gap between ischemic stroke and NFTs. Collectively, our data support the model for NFT formation following ischemia=reshown in Fig. 2. Ischemia=reperfusion causes a variety of proapoptotic events, including glutamate-induced excitotoxicity and severe oxidative stress.
Figure 2 Schematic depiction of the relationship between neurotoxic events and the formation of NFTs.
Estrogens as Neuroprotectants
183
These neurotoxic events lead to the disturbances in cellular Ca2þ homeostasis and results in elevated calpain activity. Activated calpain cleaves P35 to P25, which confers potent neurotoxicity in neurons and may lead to sustained deregulation of cdk5 in neurons (51,74,75). The resulting alteration in kinase activities ultimately leads to the hyperphosphorylation of tau and, eventually, formation of NFTs (76). This hypothesis is consistent with the observations that amyloid b peptides production and pro-oxidative events, such as ischemia=reperfusion, disrupt neuronal metabolic and ionic homeostasis and cause aberrant activation of kinases and=or inhibition of phosphatases. The resulting alteration in kinase and phosphatase activities ultimately leads to hyperphosphorylation of tau and formation of NFTs. Cyclin-dependent kinase 5 is a tau kinase that can be induced by amyloid b peptides or ischemic damage. Its deregulation may represent one of the signal transduction pathways that connect amyloid b toxicity to tau hyperphosphorylation. 6.
NEED FOR FURTHER STUDIES
Estrogens are potent, efficacious compounds in protecting the brain from the damaging effects of stroke and the neuropathology of AD, including neuronal death, APP expression, APP processing to Ab, hyperphosphorylation of tau, and aberrant neuronal mitosis. As such, additional research on these multifaceted compounds is warranted. This additional research should follow three tracts. First, extensive study is needed to determine the cellular targets and molecular events affected by estrogens to achieve their neuroprotective=anti-AD pathology activities. Second, applied animal studies are needed to determine the optimal estrogens, their formulation and route of administration, dosing schedule and optimal dosing time, and duration relative to an acute ischemic event or to the stage of development of AD neuropathology. Third, clinical studies of appropriate estrogen preparation for efficacy in stroke neuroprotection and AD are warranted, based upon the wealth of in vitro and animal data supporting the efficacy of estrogens for these conditions. ACKNOWLEDGMENTS This work was supported by NIH grants AG 10485 and 22550. REFERENCES 1.
Thorvaldsen P, Asplund K, Kuulasmaa K, Rajakangas AM, Schroll M. Stroke incidence, case fatality, and mortality in the WHO MONICA project. World Health Organization monitoring trends and determinants in cardiovascular disease. Stroke 1995; 26:361–367.
184 2.
3.
4. 5. 6. 7.
8. 9.
10. 11.
12. 13. 14. 15. 16.
17. 18.
19.
Simpkins et al. Stegmayr B, Asplund K, Kuulasmaa K, Rajakangas AM, Thorvaldsen P, Tuomilehto J. Stroke incidence and mortality correlated to stroke risk factors in the WHO MONICA Project. An ecological study of 18 populations. Stroke 1997; 28:1367–1374. Clark WM, Albers GW, Madden KP, Hamilton S. The rtPA (alteplase) 0–6-hour acute stroke trial, part A (A0276g): results of a double-blind, placebocontrolled, multicenter study. Thrombolytic therapy in acute ischemic stroke study investigators. Stroke 2000; 31:811–816. Hacke W, Brott T, Caplan L, et al. Thrombolysis in acute ischemic stroke: controlled trials and clinical experience. Neurology 1999; 53:S3–S14. Fisher M, Albers GW. Applications of diffusion–perfusion magnetic resonance imaging in acute ischemic stroke. Neurology 1999; 52:1750–1756. Ernst RL, Hay JW. The US economic and social costs of Alzheimer’s disease revisited. Am J Public Health 1994; 84:1261–1264. Frank RT, Frank ML, Gustavson RG, Weyerts WW. Demonstration of the female sex hormone in the circulating blood. I. Preliminary report. J Am Med Assoc 1925; 85:510. Edward A, Doisy CDV, Thayer S. The presentation of the crystalline ovarian hormone from the urine of pregnant women. J Biol Chem 1930; 86:499–509. Murad FaH RC Jr. Estrogens and progestins. In: Gilman AG, Goodman LS, Rall TW, Murad F, eds.The Pharmacological Basis of Therapeutics. 7th ed. New York, NY: Macmillan Publishing Company, 1985:1414–1439. Wahl P, Walden C, Knopp R, et al. Effect of estrogen=progestin potency on lipid=lipoprotein cholesterol. N Engl J Med 1983; 308:862–867. D’Eon T, Braun B. The roles of estrogen and progesterone in regulating carbohydrate and fat utilization at rest and during exercise. J Womens Health Gend Based Med 2002; 11:225–237. Marcus R. Post-menopausal osteoporosis. Best Pract Res Clin Obstet Gynaecol 2002; 16:309–327. Charkoudian N, Johnson JM. Female reproductive hormones and thermoregulatory control of skin blood flow. Exerc Sport Sci Rev 2000; 28:108–112. Wade GN. Sex steroids and energy balance: sites and mechanisms of action. Ann NY Acad Sci 1986; 474:389–399. Bishop J, Simpkins JW. Estradiol treatment increases viability of glioma and neuroblastoma cells in vitro. Mol Cell Neurosci 1994; 5:303–308. Simpkins JW, Rajakumar G, Zhang YQ, et al. Estrogens may reduce mortality and ischemic damage caused by middle cerebral artery occlusion in the female rat. J Neurosurg 1997; 87:724–730. Green PS, Simpkins JW. Neuroprotective effects of estrogens: potential mechanisms of action. Int J Dev Neurosci 2000; 18:347–358. Behl C, Skutella T, Lezoualc’h F, et al. Neuroprotection against oxidative stress by estrogens: structure–activity relationship. Mol Pharmacol 1997; 51:535–541. Weaver CE Jr, Park-Chung M, Gibbs TT, Farb DH. 17Beta-estradiol protects against NMDA-induced excitotoxicity by direct inhibition of NMDA receptors. Brain Res 1997; 761:338–341.
Estrogens as Neuroprotectants 20.
21.
22.
23.
24.
25. 26.
27.
28.
29.
30.
31.
32.
33.
185
Singh M, Setalo G Jr, Guan X, Frail DE, Toran-Allerand CD. Estrogeninduced activation of the mitogen-activated protein kinase cascade in the cerebral cortex of estrogen receptor-alpha knock-out mice. J Neurosci 2000; 20:1694–1700. Honda K, Sawada H, Kihara T, et al. Phosphatidylinositol 3-kinase mediates neuroprotection by estrogen in cultured cortical neurons. J Neurosci Res 2000; 60:321–327. Wang J, Green PS, Simpkins JW. Estradiol protects against ATP depletion, mitochondrial membrane potential decline and the generation of reactive oxygen species induced by 3-nitropropionic acid in SK-N-SH human neuroblastoma cells. J Neurochem 2001; 77:804–811. Nilsen J, Diaz Brinton R. Mechanism of estrogen-mediated neuroprotection: regulation of mitochondrial calcium and Bcl-2 expression. Proc Natl Acad Sci USA 2003; 100:2842–2847. Alkayed NJ, Harukuni I, Kimes AS, London ED, Traystman RJ, Hurn PD. Gender-linked brain injury in experimental stroke. Stroke 1998; 29:159–165, discussion 166. Dubal DB, Kashon ML, Pettigrew LC, et al. Estradiol protects against ischemic injury. J Cereb Blood Flow Metab 1998; 18:1253–1258. Sudo S, Wen TC, Desaki J, et al. Beta-estradiol protects hippocampal CA1 neurons against transient forebrain ischemia in gerbil. Neurosci Res 1997; 29:345–354. Fukuda K, Yao H, Ibayashi S, et al. Ovariectomy exacerbates and estrogen replacement attenuates photothrombotic focal ischemic brain injury in rats. Stroke 2000; 31:155–160. Mendelowitsch A, Ritz MF, Ros J, Langemann H, Gratzl O. 17Beta-estradiol reduces cortical lesion size in the glutamate excitotoxicity model by enhancing extracellular lactate: a new neuroprotective pathway. Brain Res 2001; 901: 230–236. Chen J, Xu W, Jiang H. 17 Beta-estradiol protects neurons from ischemic damage and attenuates accumulation of extracellular excitatory amino acids. Anesth Analg 2001; 92:1520–1523. Culmsee C, Vedder H, Ravati A, et al. Neuroprotection by estrogens in a mouse model of focal cerebral ischemia and in cultured neurons: evidence for a receptor-independent antioxidative mechanism. J Cereb Blood Flow Metab 1999; 19:1263–1269. Wise PM, Dubal DB, Wilson ME, Rau SW, Bottner M, Rosewell KL. Estradiol is a protective factor in the adult and aging brain: understanding of mechanisms derived from in vivo and in vitro studies. Brain Res Brain Res Rev 2001; 37:313–319. Toung TK, Hum PD, Traystman RJ, Sieber FE. Estrogen decreases infarct size after temporary focal ischemia in a genetic model of type 1 diabetes mellitus. Stroke 2000; 31:2701–2706. Carswell HV, Dominiczak AF, Macrae IM. Estrogen status affects sensitivity to focal cerebral ischemia in stroke-prone spontaneously hypertensive rats. Am J Physiol Heart Circ Physiol 2000; 278:H290–H294.
186
Simpkins et al.
34. Yang SH, He Z, Wu SS, et al. 17-Beta estradiol can reduce secondary ischemic damage and mortality of subarachnoid hemorrhage. J Cereb Blood Flow Metab 2001; 21:174–181. 35. Hawk T, Zhang YQ, Rajakumar G, Day AL, Simpkins JW. Testosterone increases and estradiol decreases middle cerebral artery occlusion lesion size in male rats. Brain Res 1998; 796:296–298. 36. Toung TJ, Traystman RJ, Hurn PD. Estrogen-mediated neuroprotection after experimental stroke in male rats. Stroke 1998; 29:1666–1670. 37. McCullough LD, Alkayed NJ, Traystman RJ, Williams MJ, Hurn PD. Postischemic estrogen reduces hypoperfusion and secondary ischemia after experimental stroke. Stroke 2001; 32:796–802. 38. Yang SH, Shi J, Day AL, Simpkins JW. Estradiol exerts neuroprotective effects when administered after ischemic insult. Stroke 2000; 31:745–749; discussion 749–750. 39. De Keyser J, Sulter G, Luiten PG. Clinical trials with neuroprotective drugs in acute ischaemic stroke: are we doing the right thing? Trends Neurosci 1999; 22:535–540. 40. Falkenstein E, Tillmann HC, Christ M, Feuring M, Wehling M. Multiple actions of steroid hormones—a focus on rapid, nongenomic effects. Pharmacol Rev 2000; 52:513–556. 41. Wang X, Simpkins JW, Dykens JA, Cammarata PR. Oxidative damage to human lens epithelial cells in culture: estrogen protection of mitochondrial potential, ATP, and cell viability. Invest Ophthalmol Vis Sci 2003; 44: 2067–2075. 42. Green PS, Gordon K, Simpkins JW. Phenolic A ring requirement for the neuroprotective effects of steroids. J Steroid Biochem Mol Biol 1997; 63:229–235. 43. Moosmann B, Behl C. The antioxidant neuroprotective effects of estrogens and phenolic compounds are independent from their estrogenic properties. Proc Natl Acad Sci USA 1999; 96:8867–8872. 44. Green PS, Yang SH, Nilsson KR, Kumar AS, Covey DF, Simpkins JW. The nonfeminizing enantiomer of 17Beta-estradiol exerts protective effects in neuronal cultures and a rat model of cerebral ischemia. Endocrinology 2001; 142:400–406. 45. Hewitt SC, Couse JF, Korach KS. Estrogen receptor transcription and transactivation: estrogen receptor knockout mice: what their phenotypes reveal about mechanisms of estrogen action. Breast Cancer Res 2000; 2:345–352. 46. Couse JF, Korach KS. Estrogen receptor null mice: what have we learned and where will they lead us? Endocr Rev 1999; 20:358–417. 47. Mehta SH, Dhandapani KM, De Sevilla LM, Webb RC, Mahesh VB, Brann DW. Tamoxifen, a selective estrogen receptor modulator, reduces ischemic damage caused by middle cerebral artery occlusion in the ovariectomized female rat. Neuroendocrinology 2003; 77:44–50. 48. Osuka K, Feustel PJ, Mongin AA, Tranmer BI, Kimelberg HK. Tamoxifen inhibits nitrotyrosine formation after reversible middle cerebral artery occlusion in the rat. J Neurochem 2001; 76:1842–1850.
Estrogens as Neuroprotectants 49.
50.
51. 52.
53. 54. 55. 56. 57.
58.
59.
60. 61.
62. 63.
64.
65. 66.
187
Rossberg MI, Murphy SJ, Traystman RJ, Hurn PD. LY353381.HC1, a selective estrogen receptor modulator, and experimental stroke. Stroke 2000; 31:3041–3046. Liu R, Yang SH, Perez E, et al. Neuroprotective effects of a novel non-receptor-binding estrogen analogue: in vitro and in vivo analysis. Stroke 2002; 33:2485–2491. Lee JM, Grabb MC, Zipfel GJ, Choi DW. Brain tissue responses to ischemia. J Clin Invest 2000; 106:723–731. Tatemichi TK, Desmond DW, Mayeux R, et al. Dementia after stroke: baseline frequency, risks, and clinical features in a hospitalized cohort. Neurology 1992; 42:1185–1193. Henon H, Durieu I, Lucas C, Godefroy O, Pasquier F, Leys D. Dementia in stroke. Neurology 1996; 47:852–853. Loeb C, Gandolfo C, Croce R, Conti M. Dementia associated with lacunar infarction. Stroke 1992; 23:1225–1229. Pasquier F, Leys D. Why are stroke patients prone to develop dementia? J Neurol 1997; 244:135–142. Tatemichi TK, Paik M, Bagiella E, et al. Risk of dementia after stroke in a hospitalized cohort: results of a longitudinal study. Neurology 1994; 44:1885–1891. Rocca WA, Bonaiuto S, Lippi A, et al. Prevalence of clinically diagnosed Alzheimer’s disease and other dementing disorders: a door-to-door survey in Appignano, Macerata Province, Italy. Neurology 1990; 40:626–631. Frisoni GB, Geroldi C, Bianchetti A, et al. Apolipoprotein E epsilon 4 allele frequency in vascular dementia and Alzheimer’s disease. Stroke 1994; 25:1703–1704. Lenzen HJ, Assmann G, Buchwalsky R, Schulte H. Association of apolipoprotein E polymorphism, low-density lipoprotein cholesterol, and coronary artery disease. Clin Chem 1986; 32:778–781. Shimano H, Ishibashi S, Murase T, et al. Plasma apolipoproteins in patients with multi-infarct dementia. Atherosclerosis 1989; 79:257–260. Pedro-Botet J, Senti M, Nogues X, et al. Lipoprotein and apolipoprotein profile in men with ischemic stroke. Role of lipoprotein(a), triglyceride-rich lipoproteins, and apolipoprotein E polymorphism. Stroke 1992; 23:1556–1562. Saunders AM, Roses AD. Apolipoprotein E4 allele frequency, ischemic cerebrovascular disease, and Alzheimer’s disease. Stroke 1993; 24:1416–1417. Stephenson DT, Rash K, Clemens JA. Amyloid precursor protein accumulates in regions of neurodegeneration following focal cerebral ischemia in the rat. Brain Res 1992; 593:128–135. Shi J, Yang SH, Stubley L, Day AL, Simpkins JW. Hypoperfusions induces overexpression of beta-amyloid precursor protein mRNA in a focal ischemic rodent model. Brain Res 2000; 853:1–4. Yoshimura M, Yamanouchi H, Kuzuhara S, et al. Dementia in cerebral amyloid angiopathy: a clinicopathological study. J Neural 1992; 239:441–450. Ince PG, McArthur FK, Bjertness E, Torvik A, Candy JM, Edwardson JA. Neuropathological diagnoses in elderly patients in Oslo: Alzheimer’s disease, Lewy body disease, vascular lesions. Dementia 1995; 6:162–168.
188
Simpkins et al.
67. Ikeda K, Akiyama H, Arai T, et al. Neurons containing Alz-50-immunoreactive granules around the cerebral infarction: evidence for the lysosomal degradation of altered tau in human brain? Neurosci Lett 2000; 284:187–189. 68. Guenette SY, Tanzi RE. Progress toward valid transgenic mouse models for Alzheimer’s disease. Neurobiol Aging 1999; 20:201–211. 69. Sathasivam K, Hobbs C, Mangiarini L, et al. Transgenic models of Huntington’s disease. Philos Trans R Soc Lond B Biol Sci 1999; 354:963–969. 70. Borchelt DR, Wong PC, Sisodia SS, Price DL. Transgenic mouse models of Alzheimer’s disease and amyotrophic lateral sclerosis. Brain Pathol 1998; 8:735–757. 71. Vila M, Przedborski S. Targeting programmed cell death in neurodegenerative diseases. Nat Rev Neurosci 2003; 4:365–375. 72. Mattson MP. Apoptosis in neurodegenerative disorders. Nat Rev Mol Cell Biol 2000; 1:120–129. 73. Graham SH, Chen J. Programmed cell death in cerebral ischemia. J Cereb Blood Flow Metab 2001; 21:99–109. 74. Tseng HC, Zhou Y, Shen Y, Tsai LH. A survey of Cdk5 activator p35 and p25 levels in Alzheimer’s disease brains. FEBS Lett 2002; 523:58–62. 75. Patzke H, Tsai LH. Calpain-mediated cleavage of the cyclin-dependent kinase-5 activator p39 to p29. J Biol Chem 2002; 277:8054–8060. 76. Lee MS, Tsai LH. Cdk5: one of the links between senile plaques and neurofibrillary tangles? J Alzheimers Dis 2003; 5:127–137.
10 Cell Death After Spinal Cord Injury: Basic Mechanisms and Potential Therapeutic Approaches to Promote Functional Recovery Steven S. Schreiber Department of Neurology, and Anatomy and Neurobiology, UCI College of Medicine, Irvine; and Neurology Section, VA Long Beach Healthcare System, Long Beach, California, U.S.A.
Despite recent advances, spinal cord injury (SCI) remains a leading cause of chronic disability with significant socioeconomic consequences. Accordingly, a large effort has been devoted to understanding and preventing the longterm effects of both the immediate primary and delayed secondary damage after SCI. Traumatic SCI is characterized by well-defined spatial and temporal patterns of tissue injury that evolve over time (1). The acute stage is characterized by hemorrhage and edema in and around the lesion epicenter (2). This is followed by a delayed stage with secondary injury due to oxidative stress, excitotoxicity, and inflammation, resulting in both neuronal and glial cell death (2). Pharmacological and=or cell-based approaches that mollify the immune response, reduce glial scarring, enhance axon regeneration, promote neuroprotection and restore irreparably damaged cells have thus far been among the most widely employed strategies for SCI therapy. Although these therapeutic approaches may significantly improve functional outcome in experimental animals, successful translation to the human condition remains to be proven. 189
190
Schreiber
Although progress has been made in defining cell death pathways in the central nervous system (CNS), a thorough understanding of the complex regulatory mechanisms and interactions between these pathways has yet to be achieved. Once available, this knowledge will doubtless lead to more efficacious ways to prevent cell death that, when combined with other therapeutic modalities, will enhance recovery of neurological function after SCI. This review will describe basic mechanisms of cell death, summarize current evidence for involvement of apoptosis in the pathophysiology of SCI and describe ways in which the apoptotic cascade may be attenuated to increase cell survival and promote functional recovery.
1.
APOPTOSIS VS. NECROSIS
Necrosis is a non-selective cell death process that occurs relatively rapidly following profound insult or injury to the CNS, usually resulting in extensive tissue loss and=or scarring (3). At the cellular level mitochondria and other intracellular organelles typically swell with eventual rupture of the cell membrane (Table 1). The extrusion of cytoplasmic contents into the extracellular space triggers an inflammatory response and subsequent scarring. In contrast, apoptosis is a more delayed, selective and energy-dependent process that targets specific cells for destruction and leaves the surrounding tissues relatively intact (4). Apoptotic cell death occurs through a cascade of molecular and biochemical events triggered by extracellular or intracellular stimuli. Morphologically, apoptosis is characterized by overall cell shrinkage with chromatin condensation, margination and fragmentation into specific oligonucleosome-sized fragments (Table 1). The fragments can be detected as a DNA ladder when analyzed by gel electrophoresis or identified in situ by techniques such as terminal deoxynucleotidyl-transferase mediated dUTP-biotin nick end labeling (TUNEL). Compared to necrosis, the appearance of intracellular organelles remains Table 1
Characteristic Features of Apoptosis and Necrosis Apoptosis
DNA Nucleus Membrane integrity Mitochondria Inflammation Pattern Cell volume Cell fragmentation
Necrosis
Internucleosomal cleavage (‘‘ladder’’) Random degradation Chromatin margination Pyknosis Persists until late Compromised early Appear normal Swollen, incr Ca2þ No Yes Individual cells affected Multiple cells affected Decreases Increases early Yes (apoptotic bodies) No (cell lysis)
Cell Death After Spinal Cord Injury
191
relatively unchanged. In the final stages of apoptosis portions of the cell pinch off into apoptotic bodies that are phagocytized in the absence of an inflammatory response. In addition to apoptotic and necrotic cell death, morphological and biochemical evidence indicates that there are alternative types of cell death with characteristic features (5). For example, paraptosis is a process that combines features of apoptosis (i.e., requirement for protein synthesis) and necrosis (i.e., cytoplasmic vacuolation) (3). Although ‘‘hybrid’’ forms of cell death have been identified, their clinical relevance remains poorly understood. The existence of cell death profiles that combine features of both apoptosis and necrosis have led many to speculate that apoptosis and necrosis lie at opposite ends of a cell death continuum (3,6). Both necrosis and apoptosis may occur side-by-side in various CNS disorders including stroke, neurodegenerative diseases, traumatic brain injury, and SCI (2,5,7). Accumulating evidence indicates that necrosis occurs through wellconserved pathways similar to apoptosis (3). However, progress in defining the mechanisms leading to necrotic cell death has been relatively slow when compared with apoptosis. A distinctive feature of cell death following SCI is the extensive amount of glial cell apoptosis that occurs (2). The delayed nature of apoptotic cell death may provide a convenient therapeutic window within which to initiate specific anti-apoptotic treatments. A strong rationale therefore exists to promote ongoing efforts to elucidate the cascade of events that define apoptosis after SCI. 2.
MOLECULAR AND CELLULAR MECHANISMS OF APOPTOSIS
During the last few years, a great deal of progress has been made in elucidating many diverse but inter-related molecular mechanisms that cause apoptotic cell death. It is now well-established that apoptosis can be activated by both extracellular stimuli and internal stressors that adversely affect the integrity of intracellular organelles (4). These and other mechanisms are described below. 2.1.
Extrinsic Pathway
Apoptosis may be initiated extracellularly through the so-called extrinsic pathway (4). Activation of the extrinsic pathway occurs upon binding of specific ligands to plasma membrane death receptors that are members of the tumor necrosis factor (TNF) family of receptors. Among the most prominent of these ligands are Fas=Apo-1=CD95 and TNF-related apoptosisinducing ligand (TRAIL) (8). Binding of cell membrane death receptors leads to clustering of intracellular receptor-associated death domains,
192
Schreiber
recruitment of the adaptor molecule Fas-associated death receptor (FADD) and activation of caspase-8 leading to caspase-3 cleavage and apoptosis (4,8). Caspases comprise a family of highly conserved cysteine proteases that are the primary initiators and effectors of apoptosis (9). Caspase-8 also cleaves the Bcl-2 family member, Bid, which then translocates to mitochondria and interacts with pro-apoptotic Bax=BclxL facilitating mitochondrial membrane permeabilization (MMP) (4,10). The p75 neurotrophin receptor (p75NTR) is a member of the TNF receptor superfamily that facilitates apoptosis during development and after various type of CNS injury (11). Overexpression of p75NTR in either neurons or glioma cells resulted in cytochrome c release, caspase activation and phosphorylation of the pro-apoptotic protein, Bad (12). Thus there are interactions between the extrinsic death receptor pathway and the intrinsic mitochondrial pathway (next section). This type of crosstalk likely serves to amplify apoptotic signaling in certain cells. 2.2.
Mitochondrial (Intrinsic) Pathway
In addition to their critical role in energy metabolism, mitochondria are major sources of reactive oxygen species (ROS) as well as key regulators of cell death. Accordingly, stressors such as excess mitochondrial calcium uptake increase the permeability of the mitochondrial membrane resulting in the release of cytochrome c into the cytoplasm (4,10). Cytochrome c interacts with apoptosis protease activating factor-1 (Apaf-1) and procaspase-9 to form a structure called the apoptosome. Activation of caspase-9 by the apoptosome results in activation of caspase-3, a major effector of apoptosis that cleaves many key intracellular substrates including ICAD (inhibitor of caspase-activated DNAase) leading to DNA fragmentation (13). The release of cytochrome c from mitochondria is a key event in the initiation of caspase-dependent apoptosis. In addition to excess calcium, interactions between pro-apoptotic members of the Bcl-2 family, i.e., Bax, Bak, Bad, and BH3-only proteins tBid and Bim, also result in MMP and release of cytochrome c (14). Various types of adverse stimuli induce translocation of cytosolic Bax to the mitochondrial outer membrane which is then breached by a poorly understood mechanism. Additionally, caspase-independent apoptosis may arise from other pro-apoptotic proteins released from damaged mitochondria such as apoptosis inducing factor (AIF), endonuclease G (EndoG), and SMAC (second mitochondriaderived activator of caspase, also known as Diablo) (15–17). SMAC=Diablo promotes apoptosis by preventing inhibitor of apoptosis protein (IAP)-mediated caspase inhibition (16,17). The family of IAP proteins prevents caspase activation by either directly interacting with and blocking caspase activity or enhancing ubiquitin-mediated caspase degradation (18).
Cell Death After Spinal Cord Injury
2.3.
193
DNA Damage
One of the most widely studied apoptotic pathways involves activation of the tumor suppressor protein, p53, by nuclear DNA damage (10). Following exposure to ionizing radiation or other genotoxic stressors, DNA damage sensors such as ATM (ataxia telangiectasia mutated) and DNA-dependent protein kinase (DNA-PK) phosphorylate specific residues within the amino terminus of the p53 protein. This leads to p53 accumulation and transcriptional activation of p53-regulated pro-apoptotic genes. In particular, p53 regulates the expression of Bax, Noxa, and PUMA, all of which are members of the Bcl-2 family and promote MMP (19,20). Additionally, p53 activation can induce proteins such as proline oxidase and p53AIPl leading to MMP (21,22), or enhance expression of Apaf-1 which facilitates caspase activation (23). Evidence supporting a more direct role for p53 in the intrinsic apoptosis pathway comes from the recent observation that p53 translocates to the mitochondrial membrane and alters its permeability (24). p53 activation may therefore transmit apoptotic signals from the nucleus to mitochondria through several different pathways. 2.4.
Organelle Stress
A growing body of evidence indicates that other intracellular organelles, when subjected to inordinate stress, may initiate apoptosis through the intrinsic pathway (10). Thus, accumulation of unfolded proteins in the endoplasmic reticulum (ER) may trigger a cascade of events resulting in inhibition of protein translation and increased expression of the anti-apoptotic gene Bcl-2 (25,26). As a major regulator of intracellular calcium concentration the ER may also trigger apoptosis through perturbations of steady-state calcium levels. In this regard, alterations of calcium release through inositol-l,4,5-triphosphate (InsP3R) and ryanodine receptors, or calcium uptake through the sarcoplasmic=endoplasmic calcium-ATPase (SERCA) may trigger apoptotic cell death (10,27). Apoptosis following ER stress has recently been linked to activation of the ER-specific caspase, caspase-12 (28). A recent study has also shown that cytochrome c binds to InsP3R and regulates the release of calcium from the ER, further enhancing cytochrome c release in a feed-forward manner that amplifies apoptotic signaling (29). In addition to the ER, apoptosis may be initiated by the Golgi apparatus. For example, studies have shown that Golgi membranes contain caspases and death receptors including TNF, Fas=CD95, and TRAIL receptors (10,30,31). However, compared to mitochondria- and ER-related pathways, relatively little information is currently available regarding mechanisms of Golgi-induced apoptosis. 2.5.
Protein Degradation
Apoptosis can also be initiated through abnormalities of protein degradation mediated by the ubiquitin–proteasome pathway (UPP) (32). Ubiquitin
194
Schreiber
(Ub) is a highly conserved 76 amino acid protein that is covalently bound to target proteins by the sequential actions of an El activating enzyme, an E2 conjugating enzyme that transfers ubiquitin and an E3 ligase that ubiquitinates target proteins (33). A polyubiquitin chain is subsequently formed by successive lysine–glycine isopeptide bonds between Ub monomers, i.e., K48 of one Ub molecule and G76 of another Ub molecule. Polyubiquitinated proteins are degraded by the 26S proteasome complex and free Ub is generated by Ub carboxyl terminal hydrolase (UCH=PGP9.5) and recycled (33,34). The ubiquitin–proteasome system is responsible for most of the non-lysosomal protein degradation in eukaryotic cells. Thus, many key apoptosis regulators including p53, members of the Bcl-2 family, and caspases are degraded by the UPP (32). Impaired protein ubiquitination and=or proteasome dysfunction may therefore lead to apoptosis through stabilization of the apoptotic machinery. Accordingly, downregulation of Ub leads to p53 accumulation and neuronal apoptosis (35,36). Conversely, UPP activation may facilitate apoptosis under certain conditions (37). Activation of the calcium-dependent cysteine protease, calpain, by a rise in intracellular free calcium plays a prominent role in cell death following various types of injury (38). In the CNS, increased calpain activation has been implicated in the pathophysiology of neurodegenerative diseases, cerebral ischemia, traumatic brain injury, and spinal cord injury (39). The broad substrate specificity of calpain includes cytoskeletal and membrane proteins, transcription factors, cytokines, protein kinases, and phosphatases (40). In this regard, calpain activation may occur directly upstream from caspase-3 and thereby regulate key proteins involved in the apoptotic cascade (39,41). 2.6.
Other Pathways: Oxidative Stress and Excitotoxicity
A large body of evidence has shown that the release of massive amounts of excitatory amino acids resulting in excessive calcium influx, and oxidative stress leading to the generation of ROS contribute to both necrosis and apoptosis following CNS injury (42–44). Additionally, excitotoxic and oxidative stress mechanisms may be coordinately regulated. In this regard, high concentrations of glutamate or N-methyl-d-aspartate induce production of the ROS generators nitric oxide (NO), superoxide, and hydrogen peroxide in cultured cells (42). Reactive oxygen species produced during oxidative stress may also trigger the release of cytochrome c from mitochondria and caspase-mediated apoptosis (43). Conversely, mice that lack the ability to generate neuronal NO exhibit reduced caspase activation and apoptotic cell death, elevated levels of Bcl-2 and smaller lesions after ischemic injury (45). In a human neuronal cell line, NO-induced apoptosis was shown to occur through increased MMP, caspase activation, and endonuclease acti-vation (46).
Cell Death After Spinal Cord Injury
195
Interestingly, caspase activation, downregulation of Bcl-2, and upregulation of Bax proteins occurred in cortical neurons cocultured with NO-producing astrocytes (47).
3.
EVIDENCE FOR APOPTOSIS AFTER SCI
Despite an intense level of activity and interest in elucidating mechanisms of apoptosis for more than a decade, the first studies demonstrating apoptosis after SCI were not performed until the late 1990s. These early observations, including human data, documented apoptotic features within cells including nuclear condensation and fragmentation, as well as the presence of nonrandomly cleaved DNA as shown by TUNEL staining in situ and DNA laddering by gel electrophoresis (48–50). Interestingly, the spatial and temporal distribution of apoptotic cell death after SCI has consistently been shown to be multiphasic (49,51–54). Thus, early after injury both apoptotic and necrotic cell death may occur within the primary region of injury. This is followed by a delayed stage characterized primarily by apoptotic cell death at sites distant from the lesion center. Between 4 and 8 hr following a moderately severe impact injury, Lui et al. (51) demonstrated an increased number of TUNEL-positive neurons within the lesion area. In the same study, an initial wave of TUNEL-positive glial cells was also observed in the lesion area that peaked by 24 hr after the injury. By seven days, a second wave of TUNEL-positive glial cells was observed in close proximity to degenerating axons located in fiber tracts far from the lesion. Doublelabeling confirmed that most of these cells were oligodendrocytes. The intensity of the injurious stimulus may influence the extent of apoptotic cell death, particularly at the lesion center where severe insults favor necrosis (2). Variations in lesion severity may also influence the type of cell involved. Accordingly, following a severe impact injury even apoptotic astrocytes may be found at the lesion center (52). The finding that the protein synthesis inhibitor, cycloheximide, reduces histological damage, and motor impairment following spinal cord contusion lends strong support to the idea that apoptosis plays a major role in the pathophysiology of SCI (51,55). It is commonly held that death of oligodendrocytes at sites far removed from the initial injury may contribute to chronic demyelination and functional impairment after SCI (2,56). This idea is supported by the remarkable overlap between the distribution of apoptotic cells and the pattern of degenerating axons in the white matter of both the injured rat and monkey spinal cord (2). Additionally, the relationship between dying oligodendrocytes and activated microglia may have functional relevance. Following moderate spinal cord contusion, there is a substantial increase in the number of microglial processes that are in direct contact with apoptotic oligodendrocytes (56). These findings suggest that the release of TNF
196
Schreiber
and other cytokines from activated microglia may cause oligodendrocyte cell death. 3.1.
Extrinsic Pathway
Activation of the extrinsic apoptosis pathway has been demonstrated after spinal cord contusion or ischemia. Accordingly, increased expression of TNF-alpha, Fas=CD-95=APO-l ligand (FasL), and Fas receptor as well as activation of caspase-8 were found primarily in neurons and oligodendroglia undergoing apoptotic cell death (57–61). Both TNF-alpha- and Fas=APO-1 -positive cells have been identified as early as 1 hr after SCI but can be detected up to 1 week or longer (60,62). A significant decrease in expression of Fas and Fas ligand occurred in white matter adjacent to the lesion epicenter within 24 hr of injury (57). In contrast, segments rostral and caudal to the lesion site exhibited increased Fas and FasL expression within several days that lasted for approximately 1 week (57). Both FADD and TRAIL were colocalized in motor neurons below the injury site following clip compression of the spinal cord (63). There is also compelling evidence for interactions between the extrinsic pathway and oxidative stress following SCI as TNF-induced apoptosis may be mediated, in part, through the production of NO (61,62). Recent findings have implicated the low affinity neurotrophin receptor, p75, in oligodendrocyte apoptosis following SCI. Upregulation of p75 occurs in both neurons and glia following SCI (2,64). In addition, oligodendrocyte apoptosis after SCI or following the administration of nerve growth factor (NGF) to cultured cells was significantly reduced in the absence of p75 (65). These findings support the idea that, under certain conditions, NGF released from microglia may be a paradoxical mediator of secondary damage after SCI (66). 3.2.
Intrinsic Pathway
Evidence for mitochondrial dysfunction and activation of the intrinsic apoptosis pathway has also been demonstrated after SCI. In this regard, decreased mitochondrial metabolic activity, generation of ROS, and increased lipid peroxidation occurred relatively early, while compensatory activation of the anti-oxidant enzymes, catalase, and glutathione occurred at later times after traumatic SCI (67). Importantly, caspase activation has been demonstrated after contusion, transection, and ischemic SCI (68–75). Evidence indicates that caspase-3 upregulation occurs at the transcriptional level (70). The caspase cascade is typically activated early in neurons around the site of injury and at later times in oligodendroglia at sites more distant from the injury. Both release of cytochrome c and activation of caspase-9 occurred within a relatively short time after SCI (68,76). Cleavage of the caspase-3 substrate, DNA fragmentation factor (DFF), was blocked by a caspase-3
Cell Death After Spinal Cord Injury
197
inhibitor (68). In addition to the caspase pathway, members of the Bcl-2 family of proteins undergo alterations after SCI, Decreased expression of the anti-apoptotic protein, Bcl-xL, was correlated with apoptosis after SCI (77). Transgenic mice overexpressing human Bcl-2 exhibited lower lesion volumes and greater functional recovery compared to control mice subjected to moderate contusion injury (78). The pro-apoptotic protein, BAD, is rapidly dephosphorylated by calcineurin after SCI (79). In Bax-deficient mice, oligodendrocyte apoptosis was attenuated following SCI in vivo and incubation of cultured cells with apoptosis-inducing drugs (80). 3.3.
DNA Damage
Activation of the p53 DNA damage response pathway has been demonstrated after SCI. In one study, p53-immunoreactive cells were found near the lesion site 30 min after spinal cord transection (81). There was a close correlation between p53 expression, DNA damage, and Bax induction. The number of p53-positive cells, primarily glia, increased with distance from the lesion over time. Three days after dorsal spinal cord transection, p53-Immunoreactive glia were observed in white matter tracts within 1 mm from the lesion epicenter (S. Schreiber, unpublished observations). p53-Immunoreactive motor neurons have also been reported following clip compression injury (63). 3.4.
Organelle Stress and Protein Degradation
In contrast to the mitochondrial pathway, there is currently little evidence to support a role for ER- or golgi-related activation of apoptosis following SCI. Similarly, overt dysfunction of the UPP has not been documented. Evidence indicates, however, that several components of the protein degradation system may be active. Swollen axons that stained positive for ubiquitin were first demonstrated in human spinal cord following surgery (82). Since then accumulations of either ubiquitin and=or PGP9.5 have been demonstrated in swollen axons in the injured rodent and equine spinal cord and after trauma in humans (83–88). The extent of both ubiquitin- and PGP9.5-like-immunoreactivity was found to vary directly with the degree of spinal cord injury (85,86). The accumulation of ubiquitin and PGP9.5 in injured axons may be due to impaired axonal transport and contribute to the UPP-mediated degradation of proteins following SCI. With advanced age, a decrease in spinal cord proteasome activity may sensitize cells to undergo apoptosis after exposure to various types of adverse stimuli (89). Activation of the calcium-dependent protease, calpain, has been convincingly demonstrated after SCI (39). In addition to degrading cytoskeletal and other cellular proteins, calpain activation has been associated with apoptotic features such as internucleosomal DNA fragmentation, elevated
198
Schreiber
Bax=Bcl-2 ratio, cytochrome c release, and cell death following contusion injury in rats (90–92). 3.5.
Oxidative Stress and Excitotoxicity
Impaired mitochondrial function can lead to increased production of ROS after SCI (67,93,94). Elevated expression of ROS-generating enzymes, and increased levels of NO and peroxynitrite have also been documented (95,96). As discussed above, excess generation of ROS may activate the caspase cascade. In this regard, capsase-3 activation was associated with elevated levels of peroxynitrite after SCI (97). There is substantial evidence for release of the excitatory amino acids glutamate and aspartate within a short period of time after SCI (2). However, the relative role of excitotoxicity in necrotic vs. apoptotic cell death after SCI remains unclear. The potential relevance of this pathway is underscored by the fact that glial cells express alpha-amino-3-hydroxy-5-ethyl-4isoxazoleproprionic acid (AMPA)=kainate-type glutamate receptors and AMPA receptor-mediated death of oligodendrocytes has been demonstrated in vitro (98,99). Further, infusion of the glutamate analog, kainic acid, into the uninjured rat spinal cord was associated with caspase-3 activation followed by the loss of oligodendrocytes (100). In addition, changes in excitatory amino acid transporter expression and a correlation between altered expression of AMPA receptors and glial cell apoptosis were demonstrated following SCI (99,101). 4. PREVENTION OF APOPTOSIS AFTER SCI 4.1. Extrinsic Pathway Inhibition Several studies have shown that the extrinsic pathway can be effectively mitigated by pharmacological approaches. Injection of either a neutralizing antibody against TNF-alpha, NOS inhibitors, or an NO scavenger into the lesion site after injury significantly reduced apoptotic cell death after SCI (61,62). Interestingly, TNF-1 receptor null mice had more apoptotic cells, larger lesions, and poorer functional recovery compared to wild-type controls after SCI (102). Additional work will be necessary to fully understand the complex role of TNF in the CNS. 4.2.
Intrinsic Pathway Inhibition
Caspase inhibitors have been shown to attenuate cell death, reduce lesion size, and promote functional recovery after SCI (69,75). Transgenic mice expressing a caspase dominant negative mutant demonstrated significant improvement in motor function and decreased lesion size compared with vehicle-treated mice (69). However, in another study, the caspase inhibitor,
Cell Death After Spinal Cord Injury
199
z-VAD, failed to prevent apoptosis or improve neurological deficits after moderate spinal cord contusion (103). Genetic manipulation has been used to inhibit the intrinsic pathway. Abrogation of Bax function in bax (=) mice protected oligodendrocytes from apoptotic cell death (80). Another anti-apoptotic approach is to increase the level of Bcl-2, which forms heterodimers with and blocks Bax function. In this regard, adenovirus-mediated transfer of Bcl-2 into the injured spinal cord resulted in a significant reduction in apoptotic cell death (104,105). Transgenic mice overexpressing the human Bcl-2 gene exhibited greater recovery of function and reduced lesion volume compared to control mice (78). In addition, retrograde cell death of neurons in Clarke’s Nucleus was attenuated by intraspinal injection of a DNA plasmid encoding the human Bcl-2 gene following thoracic hemisection (106). The expression and activity of IAP family members is altered after SCI (72). Manipulation of these endogenous caspase inhibitors may therefore provide another potential therapeutic approach. 4.3.
Antioxidant Approaches
Future SCI therapy may employ measures to reduce oxidative stress. In support of this approach, overexpression of Cu,Zn-superoxide dismutase (SOD1) decreased cytochrome c release, caspase-9 activation, and cell death following mild compression injury (76). Transgenic mice that were unable to generate NO exhibited delayed caspase activation, decreased cell death, and reduced infarct volumes following permanent middle cerebral artery occlusion, suggesting that similar strategies may be beneficial after SCI (45). To this end, anti-sense-mediated knockdown of inducible nitric oxide synthase (NOS) expression was reported to be neuroprotective after SCI (107). However, not all anti-oxidant approaches may be successful as the free radical scavenger, alpha-phenyl-N-tert-butyl nitrone (PBN), failed to prevent cell loss after spinal cord compression (108). 4.4.
Calpain Inhibitors
Reduction of calpain activity has proven to be an effective treatment for experimental SCI (39). As mentioned above, in addition to degrading a wide variety of substrate proteins, calpain may also interact with and regulate the caspase cascade. It is not surprising, therefore, that calpain inhibition attenuated apoptotic cell death within the lesion as well as in the surrounding region following SCI (55,90,91). 4.5.
Other Agents
Growth factors and other pharmacological agents that alter the relative levels of pro- and anti-apoptotic proteins, block cytochrome c release, and=or caspase activation may be cytoprotective. In this regard, the
200
Schreiber
anti-apoptotic actions of glutamate receptor antagonists, neurotrophins, chemokine, or interleukin receptor antagonists, insulin-like growth factor, erythropoietin and an antagonist of Rho GTPase offer promising therapeutic interventions for SCI (64,71,98,109–116). Additionally, downregulation of p75 neurotrophin receptor by the synthetic glucocorticoid, dexamethasone, reduced cell death and improved functional recovery after spinal cord contusion in rats (117,118). Interestingly, methylprednisolone, currently the only approved treatment for human SCI, was found to decrease tissue loss without affecting caspase activation following complete thoracic cord transection in rats (119). Other pharmacological approaches have been reported to attenuate cell death after SCI (Table 2). The mechanism of action of some of these agents may converge on apoptotic signaling cascades. 5.
CONCLUSION
The intricate biochemical and molecular cascades that underlie the pathophysiology of SCI are slowly being elucidated. Local tissue disruption, release of excitatory amino acids, oxidative stress, and inflammation are among the key processes resulting in cell loss, demyelination, glial scarring, and failed axonal regeneration. Cell loss after SCI occurs within the framework of well-defined temporal and spatial patterns associated with immediate primary and delayed secondary damage. Delayed apoptotic cell death of neurons and glia provides the most pragmatic target for cytoprotective and restorative therapies. Recent studies in experimental animals have provided the basis for effective anti-apoptotic treatments. It remains to be seen whether these types of therapeutic approaches can be successfully translated to humans. Anti-apoptotic treatments will no doubt need to be combined with other pharmacological and=or cell-based therapies to optimize functional recovery from a devastating and complex disease. ACKNOWLEDGMENT The author thanks Ms. Cassie Williams for assistance with preparation of the manuscript. REFERENCES 1.
2. 3.
Kostyk SK, Popovich PG, Stokes BT. Mechanisms and natural history of spinal cord injury. In: Lin VW, ed. Spinal Cord Medicine: Principles and Practice. New York: Demos, 2003:765–776. Beattie MS, Farooqui AA, Bresnahan JC. Review of current evidence for apoptosis after spinal cord injury. J Neurotrauma 2000; 17:915–925. Syntichaki P, Tavernarakis N. The biochemistry of neuronal necrosis: rogue biology? Nat Rev Neurosci 2003; 4:672–684.
Cell Death After Spinal Cord Injury 4. 5. 6. 7. 8. 9. 10. 11. 12.
13.
14.
15. 16.
17.
18. 19.
20.
21.
22.
201
Lawen A. Apoptosis—an introduction. BioEssays 2003; 25:888–896. Yuan J, Lipinski M, Degterev A. Diversity in the mechanisms of neuronal cell death. Neuron 2003; 40:401–413. Zeiss CJ. The apoptosis–necrosis continuum: insights from genetically altered mice. Vet Pathol 2003; 40:481–495. Behl C. Apoptosis and Alzheimer’s disease. J Neural Transm 2000; 107: 1325–1344. Wang S, El-Deiry WS. TRAIL and apoptosis induction by TNF-family death receptors. Oncogene 2003; 22:8628–8633. Hengartner MO. The biochemistry of apoptosis. Nature 2000; 407:770–776. Ferri KF, Kroemer G. Organelle-specific initiation of cell death pathways. Nat Cell Biol 2001; 3:E255–E263. Roux PP, Barker PA. Neurotrophin signaling through the p75 neurotrophin receptor. Prog Neurobiol 2002; 67:203–233. Bhakar AL, Howell JL, Paul CE, Salehi AH, Becker EB, Said F, Bonni A, Barker PA. Apoptosis induced by p75NTR overexpression requires Jun kinase-dependent phosphorylation of bad. J Neurosci 2003; 23:11373–11381. Enari M, Sakahira H, Yokoyama H, Okawa K, Iwamatsu A, Nagata S. A caspase-activated DNase that degrades DNA during apoptosis, and its inhibitor ICAD. Nature 1998; 391:43–50. Letai A, Bassik MC, Walensky LD, Sorcinelli MD, Weiler S, Korsmeyer SJ. Distinct BH3 domains either sensitize or activate mitochondrial apoptosis, serving as prototype cancer therapeutics. Cancer Cell 2002; 2:183–192. Li LY, Luo X, Wang X. Endonuclease G is an apoptotic DNase when released from mitochondria. Nature 2001; 412:95–99. Du C, Fang M, Li Y, Li L, Wang X. Smac, a mitochondrial protein that promotes cytochrome c-dependent caspase activation by eliminating IAP inhibition. Cell 2000; 102:33–42. Verhagen AM, Ekert PG, Pakusch M, Silke J, Connolly LM, Reid GE, Moritz RL, Simpson RJ, Vaux DL. Identification of DIABLO, a mammalian protein that promotes apoptosis by binding to and antagonizing IAP proteins. Cell 2000; 102:43–53. Deveraux QL, Reed JC. IAP family proteins—suppressors of apoptosis. Genes Dev 1999; 13:239–252. Oda E, Ohki R, Murasawa H, Nemoto J, Shibue T, Yamashita T, Tokino T, Taniguchi T, Tanaka N. Noxa, a BH3-only member of the Bcl-2 family and candidate mediator of p53- induced apoptosis. Science 2000; 288:1053–1058. Yu J, Wang Z, Kinzler KW, Vogelstein B, Zhang L. PUMA mediates the apoptotic response to p53 in colorectal cancer cells. Proc Natl Acad Sci USA 2003; 100:1931–1936. Donald SP, Sun XY, Hu CA, Yu J, Mei JM, Valle D, Phang JM. Proline oxidase, encoded by p53-induced gene-6, catalyzes the generation of proline-dependent reactive oxygen species. Cancer Res 2001; 61:1810–1815. Oda K, Arakawa H, Tanaka T, Matsuda K, Tanikawa C, Mori T, Nishimori H, Tamai K, Tokino T, Nakamura Y, Taya Y. p53AIPl, a potential mediator of p53-dependent apoptosis its regulation by Ser-46-phosphorylated p53. Cell 2000; 102:849–862.
202
Schreiber
23. Moroni MC, Hickman ES, Denchi EL, Caprara G, Colli E, Cecconi F, Muller H, Helin K. Apaf-1 is a transcriptional target for E2F and p53. Nat Cell Biol 2001; 3:552–558. 24. Marchenko ND, Zaika A, Moll UM. Death signal-induced localization of p53 protein to mitochondria. A potential role in apoptotic signaling. J Biol Chem 2000; 275:16202–16212. 25. Iwawaki T, Hosoda A, Okuda T, Kamigori Y, Nomura-Furuwatari C, Kimata Y, Tsuru A, Kohno K. Translational control by the ER transmembrane kinase=ribonuclease IRE1 under ER stress. Nat Cell Biol 2001; 3:158–164. 26. McCulIough KD, Martindale JL, Klotz LO, Aw TY, Holbrook NJ. Gadd l53 sensitizes cells to endoplasmic reticulum stress by down-regulating Bcl-2 and perturbing the cellular redox state. Mol Cell Biol 2001; 21:1249–1259. 27. Breckenridge DG, Germain M, Mathai JP, Nguyen M, Shore GC. Regulation of apoptosis by endoplasmic reticulum pathways. Oncogene 2003; 22: 8606–8618. 28. Morishima N, Nakanishi K, Takenouchi H, Shibata T, Yasuhiko Y. An endoplasmic reticulum stress-specific caspase cascade in apoptosis. Cytochrome c-independent activation of caspase-9 by caspase-12. J Biol Chem 2002; 277:34287–34294. 29. Boehning D, Patterson RL, Sedaghat L, Glebova NO, Kurosaki T, Snyder SH. Cytochrome c binds to inositol (1,4,5) triphosphate receptors, amplifying calcium-dependent apoptosis. Nat Cell Biol 2003; 5:1051–1061. 30. Mancini M, Machamer CE, Roy S, Nicholson DW, Thornberry NA, Casciola-Rosen LA, Rosen A. Caspase-2 is localized at the Golgi complex and cleaves golgin-160 during apoptosis. J Cell Biol 2000; 149:603–612. 31. Zhang XD, Franco AV, Nguyen T, Gray CP, Hersey P. Differential localization and regulation of death and decoy receptors for TNF-related apoptosisinducing Iigand (TRAIL) in human melanoma cells. J Immunol 2000; 164:3961–3970. 32. Yang Y, Yu X. Regulation of apoptosis: the ubiquitous way. FASEB J 2003; 17:790–799. 33. Hershko A, Ciechanover A. The ubiquitin system. Annu Rev Biochem 1998; 67:425–479. 34. Pickart CM. Targeting of substrates to the 26S proteasome. FASEB J 1997; 11:1055–1066. 35. Tan Z, Qu W, Tu W, Liu W, Baudry M, Schreiber SS. p53 accumulation due to down-regulation of ubiquitin: relevance for neuronal apoptosis. Cell Death Differ 2000; 7:675–681. 36. Tan Z, Tu W, Schreiber SS. Downregulation of free ubiquitin: a novel mechanism of p53 stabilization and neuronal cell death. Mol Brain Res 2001; 91:179–188. 37. Orlowski RZ. The role of the ubiquitin–proteasome pathway in apoptosis. Cell Death Differ 1999; 6:303–313. 38. Johnson DE. Noncaspase proteases in apoptosis. Leukemia 2000; 14: 1695–1703.
Cell Death After Spinal Cord Injury
203
39. Ray SK, Hogan EL, Banik NL. Calpain in the pathophysiology of spinal cord injury: neuroprotection with calpain inhibitors. Brain Res Rev 2003; 42: 169–185. 40. Chan SL, Mattson MP. Caspase and calpain substrates: roles in synaptic plasticity and cell death. J Neurosci Res 1999; 58:167–190. 41. Waterhouse NJ, Finucane DM, Green DR, Elce JS, Kumar S, Alnemri ES, Litwack G, Khanna K, Lavin MF, Watters DJ. Calpain activation is upstream of caspases in radiation-induced apoptosis. Cell Death Differ 1998; 5:1051–1061. 42. Mattson MP. Excitotoxic and excitoprotective mechanisms: abundant targets for the prevention and treatment of neurodegenerative disorders. Neuromol Med 2003; 3:65–94. 43. Annunziato L, Amoroso S, Pannaccione A, Cataldi M, Pignataro G, D’Alessio A, Sirabella R, Secondo A, Sibaud L, Di Renzo GF. Apoptosis induced in neuronal cells by oxidative stress: role played by caspases and intracellular calcium ions. Toxicol Lett 2003; 139:125–133. 44. Zipfel GJ, Babcock DJ, Lee JM, Choi DW. Neuronal apoptosis after CNS injury: the roles of glutamate and calcium. J Neurotrauma 2000; 17:857–869. 45. Elibol B, Soylemezoglu F, Unal I, Fujii M, Hirt L, Huang PL, Moskowitz MA, Dalkara T. Nitric oxide is involved in ischemia-induced apoptosis in brain: a study in neuronal nitric oxide synthase null mice. Neuroscience 2001; 105:79–86. 46. Moriya R, Uehara T, Nomura Y. Mechanism of nitric oxide-induced apoptosis in human neuroblastoma SH-SY5Y cells. FEBS Lett 2000; 484:253–260. 47. Tamatani M, Ogawa S, Niitsu Y, Tohyama M. Involvement of Bcl-2 family and caspase-3 like protease in NO-mediated neuronal apoptosis. J Neurochem 1998; 71:1588–1596. 48. Li GL, Brodin G, Farooque M, Funa K, Holtz A, Wang WL, OlssonY. Apoptosis and expression of Bcl-2 after compression trauma to rat spinal cord. J Neuropathol Exp Neurol 1996; 55:280–289. 49. Crowe MJ, Bresnahan JC, Shuman SL, Masters JN, Beattie MS. Apoptosis and delayed degeneration after spinal cord injury in rats and monkeys. Nat Med 1997; 3:73–76. 50. Emery E, Aldana P, Bunge MB, Puckett W, Srinivasan A, Keane RW, Bethea J, Levi AD. Apoptosis after traumatic human spinal cord injury. J Neurosurg 1998; 89:911–920. 51. Lui XZ, Xu XM, Hu R, Du C, Zhang SX, McDonald JW, Dong HX, Wu YJ, Fan GS, Jacquin MF, Hsu CY, Choi DW. Neuronal and glial apoptosis after traumatic spinal cord injury. J Neurosci 1997; 17:5395–5406. 52. Yong C, Arnold PM, Zuobine MN, Citron BA, Watanabe I, Berman NE, Festoff BW. Apoptosis in cellular compartments of rat spinal cord after severe contusion injury. J Neurotrauma 1998; 15:459–472. 53. Grossman SD, Rosenberg LJ, Wrathall JR. Temporal-spatial pattern of acute neuronal and glial loss after spinal cord contusion. Exp Neurol 2001; 168: 273–282. 54. Warden P, Bamber NI, Li H, Esposito A, Ahmad KA, Hsu CY, Xu XM. Delayed glial cell death following wallerian degeneration in white matter tracts
204
55.
56. 57.
58.
59.
60. 61.
62.
63.
64.
65.
66. 67.
68. 69.
Schreiber after spinal cord dorsal column cordotomy in adult rats. Exp Neurol 2001; 168:213–224. Ray SK, Matzelle DD, Wilford GG, Hogan EL, Banik NL. Cell death in spinal cord injury (SCI) requires de novo protein synthesis. Calpain inhibitor E-64-d provides neuroprotection in SCI lesion and penumbra. Ann NY Acad Sci 2001; 939:46–449. Shuman SL, Bresnahan JC, Beattie MS. Apoptosis of microglia and oligodendrocytes after spinal cord contusion in rats. 1997; 50:798–808. Li GL, Farooque M, Olsson Y. Changes of Fas and Fas ligand immunoreactivity after compression trauma to rat spinal cord. Acta Neuropathol 2000; 100:75–81. Matsushita K, Wu Y, Qui J, Lang-Lazdunski L, Hirt L, Waeber C, Hyman BT, Yuan J, Moskowitz MA. Fas receptor and neuronal cell death after spinal cord ischemia. J Neurosci 2000; 20:6879–6887. Casha S, Yu WR, Fehlings MG. Oligodendroglial apoptosis occurs along degenerating axons and is associated with FAS and p75 expression following spinal cord injury in the rat. Neuroscience 2001; 103:203–218. Zurita M, Vaquero J, Zurita I. Presence and significance of CD-95 (Fas=APOl) expression after spinal cord injury. J Neurosurg 2001; 94:257–264. Yune TY, Chang MJ, Kim SJ, Lee YB, Shin SW, Rhim H, Kim YC, Shin ML, Oh YJ, Han CT, Markelonis GJ, Oh TH. Increased production of tumor necrosis factor-alpha induces apoptosis after traumatic spinal cord injury in rats. J Neurotrauma 2003; 20:207–219. Lee YB, Yune TY, Baik SY, Shin YH, Du S, Rhim H, Lee EB, Kim YC, Shin ML, Markelonis GJ, Oh, TH. Role of tumor necrosis factor-alpha in neuronal and glial apoptosis after spinal cord injury. Exp Neurol 2000; 166:90–95. Panahian N, Maines MD. Site of injury-directed induction of heme oxygenase-1 and -2 in experimental spinal cord injury: differential functions in neuronal defense mechanisms? J Neurochem 2001; 76:539–554. Dubreuil Cl, Winton MJ, McKerracher L. Rho activation patterns after spinal cord injury and the role of activated Rho in apoptosis in the central nervous system. J Cell Biol 2003; 162:233–243. Beattie MS, Harrington AW, Lee R, Kim JY, Boyce SL, Longo FM, Bresnahan JC, Hempstead BL, Yoon SO. ProNGF induces p75-mediated death of oligodendrocytes following spinal cord injury. Neuron 2002; 36:375–386. Krenz NR, Weaver LJ. Nerve growth factor in glia and inflammatory cells of the injured rat spinal cord. J Neurochem 2000; 74:730–739. Azbill RD, Mu X, Bruce-Keller AJ, Mattson MP, Springer JE. Impaired mitochondrial function, oxidative stress and altered antioxidant enzyme activities following traumatic spinal cord injury. Brain Res 1997; 765:283–290. Springer JE, Azbill RD, Knapp PE. Activation of the caspase-3 apoptotic cascade in traumatic spinal cord injury. Nat Med 1999; 5:943–946. Li M, Ona VO, Chen M, Kaul M, Tenneti L, Zhang X, Stieg PE, Lipton SA, Friedlander RM. Functional role and therapeutic implications of neuronal caspase-1 and -3 in a mouse model of traumatic spinal cord injury. Neuroscience 2000; 99:333–342.
Cell Death After Spinal Cord Injury
205
70. Citron BA, Arnold PM, Sebastian C, Qin F, Malladi S, Ameenuddin S, Landis ME, Festoff BW. Rapid upregulation of caspase-3 in rat spinal cord after injury: mRNA, protein, and cellular localization correlates with apoptotic cell death. Exp Neurol 2000; 166:213–226. 71. Nesic O, Xu GY, McAdoo D, High KW, Hulsebosch C, Perez-Pol R. IL-1 receptor antagonist prevents apoptosis and caspase-3 activation after spinal cord injury. J Neurotrauma 2001; 18:947–956. 72. Keane RW, Kraydieh S, Lotocki G, Bethea JR, Krajewski S, Reed JC, Dietrich WD. Apoptotic and anti-apoptdtic mechanisms following spinal cord injury. J Neuropathol Exp Neurol 2001; 60:422–429. 73. Sakuri M, Nagata T, Abe K, Horinouchi T, Itoyama Y, Tabayashi K. Survival and death- promoting events after transient spinal cord ischemia in rabbits: induction of Akt and caspase 3 in motor neurons. J Thorac Cardiovasc Surg 2003; 125:370–377. 74. Takagi T, Takayasu M, Mizuno M, Yoshimoto M, Yoshida J. Caspase activation in neuronal and glial apoptosis following spinal cord injury in mice. Neurol Med Chir (Tokyo) 2003; 43:20–29. 75. McBride CB, McPhail LT, Vanderluit JL, Tezlaff W, Steeves JD. Caspase inhibition attenuates transection-induced oligodendrocyte apoptosis in the developing chick spinal cord. Mol Cell Neurosci 2003; 23:383–397. 76. Sugawara T, Lewen A, Gasche Y, Yu F, Chan PH. Overexpression of SOD1 protects vulnerable motor neurons after spinal cord injury by attenuating mitochondrial cytochrome c release. FASEB J 2002; 16:1997–1999. 77. Qui J, Nesic O, Ye Z, Rea H, Westlund KN, Xu GY, McAdoo D, Hulsebosch CE, Perez-Polo JR. Bcl-xL expression after contusion to the rat spinal cord. J Neurotrauma 2001; 18:1267–1278. 78. Seki T, Hida K, Tada M, Koyanagi I, Iwasaki Y. Role of the bcl-2 gene after contusive spinal cord injury in mice. Neurosurgery 2003; 53:192–198. 79. Springer JE, Azbill RD, Nottingham SA, Kennedy SE. Calcineurin-mediated BAD dephosphorylation activates the caspase-3 apoptotic cascade in traumatic spinal cord injury. J Neurosci 2000; 20:7246–7251. 80. Dong H, Fazzaro A, Xiang C, Korsmeyer SJ, Jacquin MF, McDonald JW. Enhanced oligodendrocyte survival after spinal cord injury in Bax-deficient mice and mice with delayed Wallerian degeneration. J Neurosci 2003; 23: 8682–8691. 81. Saito N, Yamamoto T, Watanabe T, Abe Y, Kumagai T. Implications of p53 protein expression in experimental spinal cord injury. J Neurotrauma 2000; 17:173–182. 82. Martin JE, Mather KS, Swash M, Garofalo O, Dale GE, Leigh PN, Anderton BH. Spinal cord trauma in man: studies of phosphorylated neurofilament and ubiquitin expression. Brain 1990; 113:1553–1562. 83. Ahlgren S, Li GL, Olsson Y. Accumulation of beta-amyloid precursor protein and ubiquitin in axons after spinal cord trauma in humans: immunohistochemical observations on autopsy material. Acta Neuropathol 1996; 92:49–55. 84. Jortner BS, Scarratt WK, Modransky PD, Walton A, Perkins SK. Ubiquitin expression in degenerating axons of equine cervical compressive myelopathy. Vet Pathol 1996; 33:356–359.
206
Schreiber
85. Li GL, Farooque M. Expression of ubiquitin-like immunoreactivity in axons after compression trauma to rat spinal cord. Acta Neuropathol 1996; 91: 155–160. 86. Li GL, Farooque M, Holtz A, Olsson Y. Expression of the ubiquitin carboxylterminal hydrolase PGP 9.5 in axons following spinal cord compression trauma. An immunohistochemical study in the rat. APMIS 1997; 105:384–390. 87. Yu WR, Olsson Y. Accumulation of immunoreactivity to ubiquitin carboxylterminal hydrolase PGP 9.5 in axons of human cases with spinal cord lesions. APMIS 1998; 106:1081–1087. 88. Westergren H, Yu WR, Farooque M, Holtz A, Olsson Y. Systemic hypothermia following spinal cord compression injury in the rat: axonal changes studied by beta-APP, ubiquitin, and PGP 9.5 immunohistochemistry. Spinal Cord 1999; 37:696–704. 89. Keller JN, Huang FF, Markesbery WR. Decreased levels of proteasome activity and proteasome expression in aging spinal cord. Neuroscience 2000; 98:149–156. 90. Ray SK, Matzelle DC, Wilford GG, Hogan EL, Banik NL. E-64-d prevents both calpain upregulation and apoptosis in the lesion and penumbra following spinal cord injury in rats. Brain Res 2000; 867:80–89. 91. Ray SK, Matzelle DD, Wilford GG, Hogan EL, Banik NL. Inhibition of calpain-mediated apoptosis by E-64-d reduced immediate early gene (IEG) expression and reactive astrogliosis in the lesion and penumbra following spinal cord injury in rats. Brain Res 2001; 916:115–126. 92. Wingrave JM, Schaecher KE, Sribnick EA, Wilford GG, Ray SK, HazenMartin DJ, Hogan EL, Banik NL. Early induction of secondary injury factors causing activation of calpain and mitochondria-mediated neuronal apoptosis following spinal cord injury in rats. J Neurosci Res 2003; 73:95–104. 93. Liu D, Sybert TE, Qian H, Liu J. Superoxide production after spinal injury detected by microperfusion of cytochrome c. Free Radic Biol Med 1998; 25:298–304. 94. Juurlink BH, Paterson PG. Review of oxidative stress in brain and spinal cord injury: suggestions for pharmacological and nutritional management strategies. J Spinal Cord Med 1998; 21:309–334. 95. Satake K, Matsuyama Y, Kamiya M, Kawakami H, Iwata H, Adachi K, Kiuchi K. Nitric oxide via macrophage iNOS induces apoptosis following traumatic spinal cord injury. Mol Brain Res 2000; 85:114–122. 96. Vaziri ND, Lee YS, Lin CY, Lin VW, Sindhu RK. NAD(P)H oxidase, superoxide dismutase, catalase, glutathione peroxidase and nitric oxide synthase expression in subacute spinal cord injury. Brain Res 2004; 995:76–83. 97. Bao F, Liu D. Peroxynitrite generated in the rat spinal cord induces apoptotic cell death and activates caspase-3. Neuroscience 2003; 116:59–70. 98. McDonald JW, Althomsons SP, Hyrc KL, Choi DW, Goldberg MP. Oligodendrocytes from forebrain are highly vulnerable to AMPA=kainate receptor-mediated excitotoxicity. Nat Med 1998; 4:291–297. 99. Park E, Liu Y, Fehlings MG. Changes in glial cell white matter AMPA receptor expression after spinal cord injury and relationship to apoptotic cell death. Exp Neurol 2003; 182:35–48.
Cell Death After Spinal Cord Injury 100.
101.
102.
103.
104. 105.
106.
107.
108.
109.
110.
111.
112. 113.
114.
207
Nottingham SA, Springer JE. Temporal and spatial distribution of activated caspase-3 after subdural kainic acid infusions in rat spinal cord. J Comp Neurol 2003; 464:463–471. Adamek D, Kaluza J, Jasinski A. Immunohistochemical investigations of excitotoxicity in experimental spinal cord trauma. Folia Histochem Cytobiol 2001; 39:175–176. Kim GM, Xu J, Xu J, Song SK, Yan P, Ku G, Xu XM, Hsu CY. Tumor necrosis factor receptor deletion reduces nuclear factor-kappaB activation, cellular inhibitor of apoptosis protein 2 expression, and functional recovery after traumatic spinal cord injury. J Neurosci 2001; 21:6617–6625. Ozawa H, Keane RW, Marcillo AE, Diaz PH, Dietrich WD. Therapeutic strategies targeting caspase inhibition following spinal cord injury in rats. Exp Neurol 2002; 177:306–313. Lou J, Lenke LG, Xu F, O’Brien M. In vivo Bcl-2 oncogene neuronal expression in the rat spinal cord. Spine 1998; 23:517–523. Yukawa Y, Lou J, Fukui N, Lenke LG. Optimal treatment timing to attenuate neuronal apoptosis via Bcl-2 gene transfer in vitro and in vivo. J Neurotrauma 2002; 19:1091–1103. Takahashi K, Schwarz E, Ljubetic C, Murray M, Tessler A, Saavedra RA. DNA plasmid that codes for human Bcl-2 gene preserves axotomized Clarke’s nucleus neurons and reduces atrophy after spinal cord hemisection in adult rats. J Comp Neuro 1999; 404:159–171. Zhang XY, Zhou CS, Jin AM, Guo AL, Tian J. Role of antisense oligodeoxynucleotide of inducible nitric oxide synthase in the apoptosis and p-p38 MAPK signal transduction in rat neurons after spinal cord injury. Di Yi Jun Yi Da Xue Xue Bao 2003; 23:578–580. Li GL, Farooque M, Holtz A, Olsson Y. Effects of alpha-phenyl-N-tert-butyl nitrone (PBN) on compression injury of rat spinal cord. Free Radic Res 1997; 27:187–196. Rosenberg LJ, Teng YD, Wrathall JR. 2,3-Dihydroxy-6-nitro-7-sulfamoylbenzo(f)quinoxaline reduces glial loss and acute white matter pathology after experimental spinal cord contusion. J Neurosci 1999; 19:464–475. Ghirnikar RS, Lee YL, Eng LF. Chemokine antagonist infusion promotes axonal sparing after spinal cord contusion injury in rat. J Neurosci Res 2001; 64:582–589. Nakao Y, Otani H, Yamamura T, Hattori R, Osako M, Imamura H. Insulinlike growth factor 1 prevents neuronal cell death and paraplegia in the rabbit model of spinal cord ischemia. J Thorac Cardiovasc Surg 2001; 122:136–143. Cao X, Tang C, Luo Y. Effect of nerve growth factor on neuronal apoptosis after spinal cord injury in rats. Chin J Traumatol 2002; 5:131–135. Koda M, Murakami M, Ino H, Yoshinaga K, Ikeda O, Hashimoto M, Yamazaki M, Nakayama C, Moriya H. Brain-derived neurotrophic factor suppresses delayed apoptosis in oligodendrocytes after spinal cord injury in rats. J Neurotrauma 2002; 19:777–785. Cheng H, Wu JP, Tzeng SF. Neuroprotection of glial cell line-derived neurotrophic factor in damaged spinal cords following contusive injury. J Neurosci Res 2002; 69:397–405.
208
Schreiber
115. Gorio A, Gokmen N, Erbayraktar S, Yilmaz O, Madaschi L, Cichetti C, Giulio AM, Vardar E, Cerami A, Brines M. Recombinant human erythropoietin counteracts secondary injury and markedly enhances recovery from experimental spinal cord trauma. Proc Natl Acad Sci USA 2002; 99:9450–9455. 116. Chong ZZ, Lin SH, Kang JQ, Maiese, K. Erythropoietin prevents early and late neuronal demise through modulation of Aktl and induction of caspase 1, 3, and 8. J Neurosci Res 2003; 71:659–669. 117. Brandoli C, Shi B, Pflug B, Andrews P, Wrathall JR, Mocchetti I. Dexamethasone reduces the expression of p75 neurotrophin receptor and apoptosis in contused spinal cord. Mol Brain Res 2001; 87:61–70. 118. Zurita M, Vaquero J, Oya S, Morales C. Effects of dexamethasone on apoptosis-related cell death after spinal cord injury. J Neurosurg 2002; 96 (1 suppl):83–89. 119. Li X, Oudega M, Dancausse HA, Levi AD. The effect of methylprednisolone on caspase-3 activation after rat spinal cord transection. Restor Neurol Neurosci 2000; 17:203–209. 120. Ibarra A, Correa D, Willms K, Merchant MT, Guizar-Sahagun G, Grijalva I, Madrazo I. Effects of cyclosporin-A on immune response, tissue protection and motor function of rats subjected to spinal cord injury. Brain Res 2003; 979:165–178. 121. Wells JE, Hurlbert RJ, Fehlings MG, Yong VW. Neuroprotection by minocycline facilitates significant recovery from spinal cord injury in mice. Brain 2003; 126:1628–1637. 122. Kaptanoglu E, Beskonakli E, Solaroglu I, Kilinc A, Taskin Y. Magnesium sulfate treatment in experimental spinal cord injury: emphasis on vascular changes and early clinical results. Neurosurg Rev 2003; 26:283–287. 123. Abe Y, Nakamura H, Yoshino O, Oya T, Kimura T. Decreased neural damage after spinal cord injury in tPA-deficient mice. J Neurotrauma 2003; 20:43–57. 124. Widenfalk J, Lipson A, Jubran M, Hofstetter C, Ebendal T, Cao Y, Olson L. Vascular endothelial growth factor improves functional outcome and decreases secondary degeneration in experimental spinal cord contusion injury. Neuroscience 2003; 120:951–960. 125. Rabchevsky AG, Sullivan PG, Fugaccia I, Scheff SW. Creatine diet supplement for spinal cord injury: influences on functional recovery and tissue sparing in rats. J Neurotrauma 2003; 20:659–669. 126. Hostettler ME, Carlson SL. PAF antagonist treatment reduces pro-inflammatory cytokine mRNA after spinal cord injury. Neuroreport 2002; 13:21–24. 127. Liao B, Newmark H, Zhou R. Neuroprotective effects of ginseng total saponin and ginsenosides Rbl and Rgl on spinal cord neurons in vitro. Exp Neurol 2002; 173:224–234. 128. Labombarda F, Gonzalez SL, Gonzalez DM, Guennoun R, Schumacher M, de Nicola AF. Cellular basis for progesterone neuroprotection in the injured spinal cord. J Neurotrauma 2002; 19:343–355. 129. Plunkett JA, Yu CG, Easton JM, Bethea JR, Yezierski RP. Effects of interleukin-10 (IL-10) on pain behavior and gene expression following excitotoxic spinal cord injury in the rat. Exp Neurol 2001; 168:144–154.
11 Motor Rehabilitation After Stroke: Novel Physical Treatment Strategies Stefan Hesse and Cordula Werner Klinik Berlin, Department of Neurological Rehabilitation, Free University Berlin, Berlin, Germany
1.
INTRODUCTION
Stroke is a leading cause of disability and handicap in the industrialized world. Each year 750,000 subjects suffer a stroke in the United States; the prevalence is 200–300 patients=100,000 inhabitants (1). Approximately, 90% of these subjects suffer from persisting neurological motor deficits leading to disability and handicap. Large outcome studies reported that only 5% of stroke survivors regain full arm function and 20% of them cannot use their arms at all (2). With respect to walking ability, approximately 75% of stroke sufferers regain limited walking ability within a period of 12 weeks and 25% remain wheelchair-bound (3). Best established prognostic factor with respect to recovery of motor function after stroke is the initial motor deficit accompanied by disabling focal upper and lower limb spasticity. The following article intends to provide a succinct, clinically oriented guide to the physical management of upper and lower limb rehabilitation after stroke including the neurolytic treatment of focal spasticity. The last decade has seen major developments in physical strategies in stroke rehabilitation fostered by the potential of brain plasticity. Conventional neurofacilitation techniques [e.g., NDT, Brunnstroem, PNF; none of them proved superior in controlled studies (4)] have been challenged by task-specific repetitive treatment approaches such as constrained-induced 209
210
Hesse and Werner
movement therapy, treadmill training with partial body weight support, and automated motor rehabilitation. Overall, a more functional less dogmatic view has gained more and more acceptance diminishing the former preponderance of general spasticity inhibition as regarded prerequisite for motor function recuperation. The treatment of spasticity nowadays follows functional considerations with Botulinum toxin enabling the precise treatment of focal spasticity. 2. GENERAL PRINCIPLES OF PHYSICAL THERAPY 2.1. Motor Rehabilitation Promotes Brain Plasticity Nudo et al. (5) studied the neural substrates of rehabilitative training on motor recovery after ischemic infarct in primates. A subtotal lesion, confined to a small portion of the representation of one hand, resulted in a further loss of hand territory in the adjacent, undamaged cortex of squirrel monkeys if no rehabilitation was applied. On the other hand, repetitive retraining of skilled hand use with the affected extremity resulted in prevention of loss of territory. In some instances, the hand representation even expanded after retraining indicating a reshaping of the cortical organization of the thumb. Potential mechanisms of brain plasticity were the modulation of existing pathways, the growth of new axonal processes, and suppression of diaschisis. 2.2.
A Task-Specific Repetitive Approach Is Most Promising
With respect to the content of rehabilitative training, modern concepts of motor learning favor a task-specific repetitive approach with the patients voluntarily activating the affected extremities (6). Skilful learning, for instance of sports or violin playing, requires several hundred thousand repetitions. For acute stroke patients, Kwakkel et al. (7) showed a positive correlation between therapy intensity and upper and lower limb motor function, i.e., patients should practice intensively (7). Secondly, Winstein et al. (8) reported that balance training while standing could improve balance but not gait symmetry in hemiparetic patients, i.e., patients should practice task specifically (8). Furthermore, a large Norwegian outcome study compared the Bobath (NDT) program and a task-specific motor relearning program [MRP, according to Carr and Shepherd (9)] in 61 acute stroke patients (10). The MRP group stayed fewer days in hospital and their improvement in general motor functions was significantly better than in the Bobath group. 3. SPECIAL ASPECTS OF UPPER LIMB REHABILITATION 3.1. Repetitive Training of Isolated Movements The potential of a repetitive training of isolated wrist movements was shown by Bu¨tefisch et al. (11) in 27 hemiparetic patients. The specific training consisted of repetitive hand and finger flexion and extensions against various
Motor Rehabilitation After Stroke
211
loads and was carried out twice daily for 15 min in addition to conventional therapy. In contrast to the 3-week baseline of conventional NDT therapy, grip strength, peak force of isometric hand extensions, and peak accelerations of isotonic wrist extension improved significantly during the training period. The value of repetitive sensorimotor stimulation of the paretic upper limb was further confirmed in a large (n ¼ 100) prospective study on severely affected subjects. The experimental group pushed a rocking chair with their paretic arm, fixed in inflatable splint, repetitively while in the control group the arm was rested on a cushion in front of the patient (12). Patients in the experimental group performed better on the Brunnstroem–Fugl–Meyer test than those in the control group throughout the study period, but differences were only significant at follow-up. 3.2.
Constrained-Induced (CI) Movement Therapy
CI movement therapy is gaining more and more acceptance nowadays. It is based on the concept of ‘‘learned non-use’’ saying that repeated disappointment in attempts to use the affected arm in the acute and subacute phase could lead to negative reinforcement of using the affected arm. Consequently, the patients do not use the affected extremity despite an existing motor potential. To overcome this learned non-use, Taub et al., introduced the CI movement therapy in arm rehabilitation after stroke following successful animal experiments. The patients wear a sling to immobilize the unaffected arm and are thus enforced to use the affected extremity repetitively and taskspecifically. In their original paper, Taub et al. (13) recommended to constrain the non-affected arm over 90% of the working day. The restraint devices were worn for 14 days. On each weekday during this period patients spent 7 hr at the rehabilitation center and were given a variety of tasks to be carried by the paretic upper extremity for 6 hr. Before treatment, patients should be able to extend at least 10 at the metacarpiand interphalangeal joints and 20 at wrists. Several controlled papers showed considerable improvements of hand function in chronic and acute hemiparetic patients following 14 days of CI therapy (for overview, see 12a). Also, brain-mapping studies revealed an enlargement of the cortical representation of the affected upper extremity following CI therapy (14). On the other hand, animal experiments in rats after unilateral sensorimotor cortex lesion showed that immobilization of the non-impaired forelimb for 15 days immediately following the lesion resulted in dramatic exaggeration of the neuronal injury, presumably attributable to forced overuse of the impaired limb. As clinical implication, the authors suggested a less aggressive ‘‘use-it-butdon’t-overuse-it’’ strategy for the impaired limb in the early recovery phase after stroke (15). Severely impaired patients suffering from attentional deficits and a lowered tolerance to physical and psychological stress may rarely practice as intensively, but these results definitely warrant further clinical studies to
212
Hesse and Werner
determine the optimum treatment amount in the early recovery period after stroke (15). 3.3.
Bilateral Practice
Bilateral practice is recommended for severely affected subjects, backed by recent f-MRI studies after stroke showing the significance of the ipsilateral hemisphere for motor recovery. It intends to use the facilitatory drive of the non-affected onto the affected extremity via intercallosal fibers. This phenomenon can be seen with many hemiparetic patients performing mirror movements when asked to activate their paretic muscles. Mudie and Matayas (16) studied the effects of bilateral practice in subchronic stroke patients, who were instructed to place a block, drink, from a glass (Fig. 1), or target a peg either unilaterally with the affected arm or bilaterally. The bilateral practice proved superior with regard to upper limb motor function improvement. In another open study, Whitall et al. (17) investigated bilateral arm training with rhythmic auditory cueing. The authors asked their chronic patients to practice bilaterally on a custom-designed machine (pushing two handles forward and backward) 20 min each workday for 6 weeks. The patients showed significant and durable increase of their arm function, isometric strength, and range of motion of the affected side.
Figure 1
A left hemiparetic subject drinking bilaterally.
Motor Rehabilitation After Stroke
3.4.
213
Electrical Stimulation
Electrical stimulation has a long and varied history in motor rehabilitation. A recent controlled study included 60 patients with acute poststroke hemiplegia (18). The patients were assigned randomly to one of two groups either receiving standard therapy or electrical stimulation (20 Hz, pulse width 0.2 msec, above motor threshold) of the wrist and finger extensors (3 times, 30 min for 8 weeks) in addition to standard therapy. The experimental group significantly scored better with respect to isometric wrist strength and grasping function after 8 weeks. Benefits of electrical stimulation were most apparent in those patients with some residual motor function at the wrist. A novel approach is electromyography-triggered electrical stimulation of the wrist and finger extensors. The patient is instructed to voluntarily activate the paretic extensor muscle; the corresponding EMG signal is sensed (Fig. 2). Once the muscle activity has reached a preset threshold, an external electrical stimulus elicits a full wrist and finger extension.
Figure 2 EMG-triggered biofeedback of the wrist extensor muscles, the patient voluntarily activates his muscles to a preset threshold, to be followed by an external electrical stimulation of the muscle.
214
Hesse and Werner
A randomized clinical study with 12 chronic (at least 1 year stroke interval) hemiparetic patients with a residual motor function followed a modified crossover treatment with one block of 12 sessions of electrical stimulation and another block of 12 treatment sessions where the patients were encouraged to attempt wrist and finger extension without external assistance (19). During the experimental phases, patients could significantly move more wooden blocks with their hand and displayed a higher isometric force of the wrist and finger extensors. Another study adopted a baseline treatment design showing that the additional daily use of this kind of electrical stimulation for 4 weeks further enhanced the motor recovery as compared to a 3-week baseline of conventional therapy (20). In daily practice, a therapist has to assist and encourage the patients to voluntarily activate their wrist and finger extensors, otherwise the patients tend to lower the threshold to a purely passive stimulation. 3.5.
Robot-Aided Simulation
Hogan et al. (21) introduced the ‘‘MIT-Manus’’ as a robotic device. The patient holding to a robotic arm could perform unrestricted shoulder and elbow movements in the horizontal plane (Fig. 3). The device guided the limb and provided a sensorimotor experience that responded quickly, just like a ‘‘hand-over-hand’’ therapy. For evaluation, Volpe et al. included 56 acute stroke patients who were either randomly assigned to the robot training for at least 25 hr or were exposed to the robotic device without training. At the end of treatment, the robot-trained group demonstrated
Figure 3 The MIT-Manus; a left hemiparetic subject is practicing unrestricted shoulder elbow movements in the horizontal plane.
Motor Rehabilitation After Stroke
215
Figure 4 The ‘‘Bi-Manu-Track’’; a left hemiparetic subjects is practicing bilateral passive and active forearm pro- and supination (left) and wrist extension=flexion (right).
improvement in motor outcome for the trained shoulder and elbow that did not generalize to untrained wrist and hand (22). Lum et al. (23) introduced the ‘‘MIME’’ robot enabling the bilateral practice of shoulder and elbow-movements in the horizontal plane following the master (non-affected) and slave (affected extremity) principle. A controlled study included 30 subacute stroke survivors, who either practiced with the robot or within a conventional training programme following the NDT concept for 22 sessions each. The robot group showed significantly better shoulder and elbow muscle strength and upper motor function at study end and at follow-up (23). Hesse et al. (24) introduced the ‘‘Bi-Manu-Track’’ (Fig. 4) robot enabling the bilateral passive and active practice of two distal movements: forearm pro-=supination and wrist flexion=extension. Bilateral practice, often seen in patients when asked to move their paretic upper extremity, intends to facilitate the paretic side via interacalosal fibers. The selection of a more distal approach was influenced by recent f-MRI studies describing a competition between proximal and distal body segments for plastic brain territories and the finding that regional anesthesia of the shoulder girdle improved the hand function in chronic stroke patients (25). A first open study in 15 severely affected chronic patients with no volitional control of the wrist and finger extensors found a relevant reduction of muscle tone of the wrist and fingers, motor functions also improved for a mean of þ85% on the Fugl–Meyer motor scale. Only one patient, however, could grasp and release a spheric object (a tennis ball) at the end of the study.
216
3.6.
Hesse and Werner
Neurolytic Treatment of Upper Limb Spasticity
Botulinum toxin A and more recently toxin B are nowadays accepted as the first treatment choice in the case of disabling upper limb focal spasticity (26,27). The i.m. injection of the toxin into muscles relevant for the spastic deformity resulted in their reversible paresis by blocking acetylcholine release in the terminal end plates. Muscle tone reduction, improved range of motion, pain relief, ease of hand hygiene, and dressing (e.g., putting an arm through a sleeve) are realistic therapeutic goals, to be defined within the therapeutic team. Motor function improvements, albeit often reported in open studies, have not yet been confirmed by controlled trials. The effects start 1 week after injection and last upto 4 months, side effects (e.g., dysphagia, bladder paresis) are minimal and reversible. The injection of the following muscles is generally recommended: M. pectoralis, M. subscapularis (shoulder spasticity), M. biceps brachii, M. brachioradialis (elbow spasticity), M. flexor carpi ulnaris, Mm. flexor digitorum profundus, and superficialis (wrist and finger spasticity). Of course, an individual adjustment of the injection sites may be necessary, a general rule says: prominent and tense muscles are good candidates. The technique (either blind or EMG-guided) depends on personal experience and equipment availability. Large muscles pose no problems whereas deep layer muscle in the forearm (or the mm. lumbricales in case of resistant finger flexor spasticity) requires EMG guidance. Most authors prefer one site=muscle injecting mid-belly; motor end plate charts of various muscles are currently being developed. Dosages vary from author to author; also conversion factors between the US (BOTOX2) and British product (Dysport2) vary from 1:3 to 1:5. Good practice is in one vial of either product per forearm or elbow or shoulder spasticity. For the B-toxin, a recent study reported on a total dosage of 10,000 U injected into five major muscles of the upper limb (27). To increase the effectiveness of the toxin and to reduce dosage, the reader should pay attention to the correlation of terminal nerve end activity and the toxin uptake (28,29). Since in clinical practice, spasticity is almost always accompanied by central paresis, the affected muscles are usually less active, on average, than normal or dystonic muscles. Correspondingly, the patients should use the injected extremities and receive intensive physiotherapy after injection. When not possible, electrical stimulation of the treated muscles can be used to artificially activate the terminal nerve ends. A viable protocol applies three strains of charge-balanced constant current pulses (20 Hz, 0.2 msec, 50–90 mA) for 30 min, five times=day, 3 days after the injection (29). Other discussed means of increasing the effectiveness of the costly toxin in daily practice are a larger dilution (3–5 ml=vial) and the combination with serial casting after injection to address the accompanying
Motor Rehabilitation After Stroke
217
changes of muscle properties by applying a tonic stretch. The best and safest method for consistently extracting the most Botulinum toxin from its vial was to use a long 21-gauge 2-in. needle attached to a 3-mL yringe (30). With respect to controlled studies, Simpson et al. (31) were the first to report a randomized, double-blind placebo-controlled trial using BOTOX in 39 individuals with spastic upper extremities. The author’s injected placebo or a total dosage of 75, 150, and 300 units BOTOX into three muscles: biceps brachii (65%), flexor carpi radialis (25%), and ulnaris (15% of total dosage). Only the treatment with the highest dose resulted in a statistical significant mean decrease of muscle tone 2, 4, and 6 weeks after injection. There were no adverse reactions. No significant differences were found between placebo and treatment groups for motor functions of the affected upper extremity, pain, caregiver dependency, and competence in daily activities. For the British product, Dysport, a large dose-finding randomized, placebo-controlled study in upper limb spasticity (n ¼ 83) reported a relevant muscle tone reduction at doses of 500, 1000, and 1500 units (32). Again, functional disability did not differ between treatment groups. Hesse et al. (33) studied the combined effect of Dysport and electrical stimulation with the help of a placebo-controlled study on upper limb spasticity with four treatment arms. Only the group receiving verum (1000 units Dysport) plus electrical stimulation showed a relevant muscle tone reduction and improvement for such tasks as putting the arm through a sleeve. In case of budget constraints, phenol 5% may be an alternative despite its risk of local inflammation and dysaesthesia in 20% of cases treated. We therefore only recommend the injection of the motor point of the N. musculocutaneus (a pure motor nerve) in case of elbow spasticity. Electrical stimulation via a mono-polar needle also suitable for injection helps to locate the optimum injection site. Only inject when you can elicit a strong muscle twitch at a stimulation intensity of less than 3 mA. Major advantage of phenol is its minimal cost as compared to BTX and the effects can last up to several months. Impairing side effects, however, are a local inflammation following the injection and the risk of painful dysaesthesias in up to 20% of patients. Clinical practice recommends local cooling and the instruction not to move the treated extremity for 24 hr following the injection. Another recently discussed option in the treatment of disabling upper and lower limb spasticity following stroke is the intrathecal application of Baclofen (ITB). Meythaler et al. either injected normal saline or a bolus of 50 mg Baclofen intrathecally in 21 chronic hemiparetic subjects. The Ashworth scores of the spastic upper and lower extremities diminished to a significantly larger extent in the verum group. A subsequent computer-controlled pump implantation for continuous ITB in 17 patients decreased the upper and lower
218
Hesse and Werner
Ashworth scores (0–5) for two levels over a follow-up period of 12 months. Relevant side effects did not occur; headaches due to CSF leak associated with catheter placement were transient. Functional benefits were an ease in personal hygiene and a better urinary voiding pattern (34).
4. SPECIAL ASPECTS OF GAIT REHABILITATION 4.1. Treadmill Training with Partial Body Weight Support In line with the task-specific repetitive approach, treadmill therapy with partial body weight (Fig. 5) support enables non-ambulatory hemiparetic subjects the repetitive practice of complex gait cycles at a very early stage (35). The harness substitutes for deficient equilibrium reflexes, the moving belt enforces locomotion, and part of the body weight is supported according to the extent of lower limb paresis. Initially two therapists assist the movement, setting the paretic limbs and controlling the trunk movements, to practice not only repetitively but also in a correct manner. Within one
Figure 5 A left hemiparetic subject walking on the treadmill with partial body weight support assisted by two therapists.
Motor Rehabilitation After Stroke
219
session of 20 min, patients can perform up to 1000 steps as compared to 50– 100 steps during a conventional physiotherapy session. Theoretical background is the activation of spinal and supraspinal gait pattern generators according to animal experiments (36). Biomechanical studies revealed that hemiparetic patients walked more dynamic, symmetric, and less spastic on the treadmill as compared to floor walking (37). Further, the amount of body weight relief negatively correlated with the activity of relevant weight-bearing muscles, accordingly one should not exceed 30% BWS and should reduce the support as soon as possible. A clinical criterion was the patients’ ability to carry their load adequately during the single stance phase of the paretic limb, i.e., without excessive knee flexion or swinging in the harness (37). Treadmill velocity ranged from 0.25 to 0.35 m=sec, daily sessions of at least 20 min net walking time are recommended for at least 3 weeks. Exclusion criteria are critical cardiovascular conditions, severe arthrosis of the major joints and contractures with more than 20 extension deficit of the hip and knee joints. A pusher syndrome (i.e., the active pushing to the affected side within a neglect syndrome) is no exclusion criteria. For initial clinical evaluation, Hesse et al. (35) conducted two single case design studies following an A–B–A design with 14 chronic, nonambulatory hemiparetic patients. During the two A-phases, the patients exclusively received the treadmill therapy for 3 weeks, and during the Bphase, conventional physiotherapy following a classic NDT approach was administered for another 3 weeks. Gait ability and walking velocity improved substantially during the first A-phase, then levelled during the B-phase, to increase again during the second A-phase. At the end of the study, all patients could walk independently at least with verbal support, the statistical analysis revealed that the task-specific approach was superior with regard to gait restoration and improvement of walking velocity. Another study of the same group in 28 chronic non-ambulatory hemiparetic patients showed that the combination of physiotherapy stressing gait practice on the floor and goal-oriented treadmill training resulted in a faster gait recuperation than treadmill training alone (38). Both kinds of therapies obviously added to each other and the double therapy intensity in the experimental group confirmed the postulated correlation between therapy intensity and gait outcome in stroke rehabilitation. Visintin et al. (39) compared treadmill therapy with and without BWS in 100 acute stroke survivors. The BWS-group scored significantly better in mobility outcomes and walking velocity, endurance, and balance after a 6-week training period. Three months later, the BWS group continued to have significantly higher scores for walking velocity and motor recovery (39). Limiting factors of the study were the lack of a statistical analysis taking into account the high drop-out rate of 21% and no blinded observers with respect to the group assignment of the patients.
220
Hesse and Werner
Two recent randomized controlled trials in 56 and 76 acute stroke survivors compared treadmill training with BWS and physiotherapy stressing the repetitive practice on the floor assisted by braces and orthoses (40,41). The studies failed to show any superiority, neither gait ability nor walking velocity differed among groups in the two studies. Merely a subgroup with major hemispheric stroke defined by the presence of hemiparesis, hemianopic visual deficit, and hemihypaesthesia who received more than 12 treatment sessions, showed significantly better overground endurance [90 34 vs. 44 10 m] and speed scores [12 4 vs. 8 2 m=min] for the treadmill compared with the physiotherapy group (42). Further, two smaller studies on acute stroke subjects compared treadmill training vs. a less aggressive physiotherapy approach with respect to gait practice on the floor. Da Cunha et al. (42) included 12 subjects in their study, the results revealed a better oxygen consumption during bicycle ergometry in the experimental group, while the general motor functions did not differ. Laufer et al. (43) studied 25 subjects, after 15 sessions of treadmill training the experimental group scored better with respect to gait ability, stride length, percentage of single stance duration, and gastrocnemius activity. Ambulatory hemiparetic subjects could improve their walking speed and endurance on the treadmill training with harness-support as compared to conventional therapy following 4 weeks of daily 20 min therapy (44). Speed and inclination of the treadmill were increased stepwise, so that heart rate reached a pulse of 190-age to reach aerobic training conditions. A preexercise test following the guidelines for cardiac patients (see American Heart Association) is recommended to minimize cardiovascular risks of speed training. Sullivan et al. (45) also showed that higher training speeds (2.0 mph) on the treadmill were favorable in ambulatory chronic patients to improve their self-adopted gait velocity on the floor.
4.2.
Automated Gait Rehabilitation
The major inconvenience of treadmill therapy (and physiotherapy) of nonambulatory subjects is the great physical effort required by two therapists to assist the patients’gait. Therefore, our group designed an electromechanical gait trainer ‘‘GT I’’ (46) and Colombo et al. (47) a computer-assisted version (‘‘Lokomat’’) with the patients wearing a powered lower limb exoskeleton. In the electromechanical gait trainer (Fig. 6), the harness-secured patient is positioned on two foot plates whose movements simulate natural walking, a servo-controlled motor supports the patients according to their abilities, and the vertical and horizontal trunk movements are controlled in a phasedependent manner by ropes attached to the harness. A randomized study in subacute, non-ambulatory subjects recently showed that 6 weeks of daily gait trainer therapy resulted in better gait outcome and less effort for the
Motor Rehabilitation After Stroke
Figure 6
221
A left hemiparetic subject practicing gait on the gait trainer.
therapists as compared to treadmill training with BWS (48). A preceding baseline treatment study in chronic patients had shown that 4 weeks of additional daily gait trainer therapy of 1000 steps per session resulted in a better gait ability and muscle activation as compared to a 3-week basline of conventional therapy (49). The shank muscle activity increased and the pattern of the thigh muscles became more physiological, similar to the finding in paraparetic subjects (50), A recent study compared the locomotor therapy of 30 subacute non-ambulatory stroke patients on the gait trainer and on the treadmill with the help of a randomized crossover design following an A–B–A or B–A–B design with A ¼ 2 weeks gait trainer and B ¼ 2 weeks treadmill therapy. The practice on the gait trainer required less assistance (one instead of two persons) and the A-group could significantly walk better at the end of the study (48). The ‘‘Lokomat’’ consists of a treadmill with body weight support; in addition the patients wear a powered exoskeleton where the programmable
222
Hesse and Werner
drives flex the hip and knee actively during the swing phase. A passive orthosis controls for the ankle movement. Two paraparetic subjects needed less therapeutic effort as during the manually assisted treadmill therapy while the muscle activation pattern of the lower limbs was comparable. 4.3.
Electrical and Rhythmic Auditory Stimulation
Since Liberson’s first description in 1961, functional electrical stimulation of the nervus peronaeus for the correction of a hemiparetic drop foot has seen many ups and downs. In recent years, British authors conducted a randomized trial, 32 chronic, ambulatory hemiparetic patients with a single drop-foot participated (51). They either received 10 physiotherapy or 10 treatment sessions with a common peroneal stimulator. At the end of the study and at follow-up, patients of the FES group walked significantly faster and more efficiently with the stimulator (þ20% respectively 24.9%) whereas the control group did not improve walking speed and efficiency. However, no improvement in these parameters was measured in the FES group when the stimulator was not used, i.e., a carry-over was not shown. Rhythmic auditory stimulation (RAS) of hemiparetic gait is a new kind of interactive biofeedback (52). The rhythmic stimulus consisted of music tapes played over headsets that were prerecorded on a synthesizer per sequencer module. The rhythm of the music (class, folk, country, jazz) was chosen to match the patients’ cadence. Therapeutic goals were an improvement of gait symmetry and gait velocity. Within a first study, 20 ambulatory stroke patients in the acute stage either received conventional therapy or RAS for 6 weeks each. Pre- vs. posttest measures revealed a statistically significant increase in velocity (164% vs. 107%), stride length (88% vs. 34%), and reduction in the EMG amplitude variability of the gastrocnemius muscle for the RAS-training group compared to the control group. 4.4.
Neurolytic Treatment of Focal Lower Limb Spasticity
In lower limb spasticity after stroke, Botulinum toxin type A (BTX-A) has become a preferred treatment option. For the spastic drop foot, several open studies showed that the intra-muscular injection of BTX-A into the plantar flexors reduced muscle tone, improved the mode of initial contact, ankle range of motion with better advancement of the body, gait symmetry, and walking velocity up to 4 months (Fig. 7). Longer lasting effects are rarely seen in adult hemiparetic subjects. Relevant side effects except a temporary bladder paresis did not occur with maximum dosages of up to 400 units BOTOX or 2000 units Dysport. Recommended doses are 200 units BOTOX or 1000 units Dysport for the treatment of spastic drop foot. Dynamic EMG recordings revealed a preferential diminution of the so-called premature activity of the plantar flexors originating already in the terminal swing
Motor Rehabilitation After Stroke
223
Figure 7 Trajectories of the force point of action in a right hemiparetic patient before and 2 weeks after the injection of Botulinum toxin A.
(53). This stretch-related activity promotes impairing plantar flexion during the terminal swing and has been described in spastic stroke patients and CP children suffering from an equinovarus deformity. Burbaud et al. (54) conducted a double-blind, placebo-controlled study in 24 individuals suffering from a spastic drop foot. The authors injected EMG-guided placebo or a total dosage of 1000 units Dysport into the soleus, gastrocnemius, tibialis posterior, and flexor digitorum longus muscles. Patients reported a clear subjective improvement in foot spasticity after BTX but not after placebo administration. Significant changes were noted in Ashworth scale values for ankle extensors and invertors, and for active ankle dorsiflexion upto 3 months after injection. Gait velocity was slightly but not significantly improved after the injection of BTX. To increase the effectiveness of the toxin and given the positive correlation of terminal nerve end activity and the toxin uptake (see Sect. 3.6), patients should walk vigorously after injection, staying idle will prevent a good result. When not possible, electrical stimulation of the treated muscles can be used to artificially activate the terminal nerve ends. A viable protocol applies trains of 3 sec of charge-balanced constant current pulses (20 Hz, 0.2 msec, 50–90 mA) for 30 min five times per day three days after the injection (29). Further, the exclusive injection of the gastrocnemius muscle bellies is not sufficient in most cases, the soleus, tibialis posterior, long toe flexor, and in case of a severe ankle inversion, the tibialis anterior muscle should be additionally treated (55). The combination with serial casting after injection (to address the accompanying changes of muscle properties by applying a tonic stretch) is also recommended to increase the effectiveness of the costly toxin (56).
224
Hesse and Werner
Major indications for phenol 5% are adductor spasticity (injection of nervus obturatorius) and equinovarus deformity. In the latter case, phenol 5% is either injected close to the motor point of the plantar flexors or close to the nervus tibialis in the fossa poplitea. Electrical stimulation via a monopolar needle also suitable for injection helps to locate the optimum injection site. Major advantage of phenol is its minimal cost as compared to BTX and the effects can last up to several months. Impairing side effects, however, are a local inflammation following the injection and the risk of painful dysaesthesias in up to 20% of patients. 5.
SUMMARY
Physical motor rehabilitation after stroke has seen major developments in the last decade. Motor therapy could promote brain plasticity in animals, and modern concepts of motor learning after stroke favor a task-specific repetitive approach inducing skill-acquisition relevant to the patient’s daily life. New physical strategies following this line have emerged in upper limb rehabilitation: constrained-induced movement therapy to successfully overcome a learned non-use of the affected extremity, EMG-triggered electrical stimulation of the wrist forcing the patients to voluntarily activate their muscles, robot assisted motor rehabilitation to increase therapy intensity and bilateral practice to facilitate the movement of the paretic extremity via intercallosal fibers. Lower limb rehabilitation has been enriched by Treadmill training with partial body weight support enabling the practice of up to 1000 steps per session, automated gait rehabilitation to relieve the strenuous effort from the therapists, and rhythmic auditory stimulation applying individually adjusted music to improve walking speed and symmetry. At this level of evidence, constrained-induced movement therapy in upper limb rehabilitation seems the most promising. For lower limb rehabilitation, treadmill training with partial body weight support may evolve as the future treatment option. For both techniques, controlled multicenter trials must be the next step. In the focal treatment of upper and lower limb spasticity, Botulinum toxin type A has become the major treatment option, the toxin proved effective and safe in several placebo-controlled trials. Currently toxin type B is evaluated as an alternative in case of nonresponders to toxin type A. Phenol injections are an inexpensive alternative but bear a larger risk of side effects and require specific technical skills. Intrathecal baclofen for spasticity treatment after stroke seems another promising treatment approach. REFERENCES 1.
Williams GR, Jiang JG, Matchar DB, Samsa GP. Incidence and occurrence of total (first-ever and recurrent) stroke. Stroke 1999; 30:2523–2528.
Motor Rehabilitation After Stroke 2.
3.
4. 5.
6. 7.
8.
9. 10.
11.
12.
12a. 13.
14. 15.
16. 17.
225
Nakayma H, Jorgensen HS, Raaschou HO, Olsen TS. Recovery of upper extremity function in stroke patients: the Copenhagen study. Arch Phys Med Rehabil 1994; 75:852–857. Jorgensen HS, Nakayma H, Raaschou HO, Olsen TS. Recovery of walking function in stroke patients: the Copenhagen stroke study. Arch Phys Med Rehabil 1995; 76:27–32. Ernst E. A review of stroke rehabilitation and physiotherapy. Stroke 1990; 21:1081–1085. Nudo RJ, Wise BM, SiFuentes F, Milliken GW. Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science 1996; 272:1791–1794. Asanuma H, Keller A. Neurobiological basis of motor learning and memory. Concepts Neurosci 1991; 2:l–30. Kwakkel G, Wagenaar RC, Twisk JWR, Lankhorst GJ, Koetsier JC. Intensity of leg and arm training after primary middle- cerebral-artery stroke: a randomised trial. Lancet 1996; 354:191–196. Winstein CJ, Gradner ER, McNeal DR, Barto PS, Nicholson DE. Standing balance training: effects on balance and locomotion in hemiparetic adults. Arch Phys Med Rehabil 1989; 70:755–762. Carr JH, Shepherd RB. A motor Relearing Programme. London: William Heinemann, 1987:1–25. Langhammer B, Stanghelle JK. Bobath or motor relearning programme? A comparison of two different approaches of pyhsiotherapy in stroke rehabilitation: a randomized controlled study. Clin Rehabil 2000; 14:361–369. Bu¨tefisch C, Hummelsheim H, Denzler P, Mauritz KH. Repetitive training of isolated movements improves the outcome of the centrally paretic hand. J Neurol Sci 1995; 130:59–68. Feys HM, De Weerdt WJ, Selz BE, Cox Steck GA, Spichiger R, Vereeck LE, Putman KD, Van Hoydonck GA. Effect of a therapeutic intervention for the hemiplegic upper limb in the acute phase after stroke. A single-blind, randomized, controlled multicenter trial. Stroke 1998; 29:785–792. Morris DM, Taub E. Constraint induced movement therapy approach to restoring function after neurological injury. Top Stroke Rehabil 2001; 8:16–30. Taub E, Miller NE, Novak TA, Cook EW, Fleming WC, Nepomuceno CS, Connell JS, Crago JE. Technique to improve chronic motor deficit after stroke. Arch Phys Med Rehabil 1993; 74:347–354. Liepert J, Bauder H, Miltner WHR, Taub E, Weiller C. Treatment-induced massive cortical reorganization after stroke in humans. Stroke 2000; 31:1210–1216. Kozlowski DA, James DC, Schallert T. Use-dependent exaggeration of neuronal injury after unilateral sensorimotor cortex lesion. J Neurosci 1996; 16:4776–4786. Mudie MH, Matayas TA. Upper extremity retraining following stroke: effects of bilateral practice. J Neuro Rehabil 1996; 10:167–184. Whitall J, McCombe S, Waller S, Silver KHC, Macko RF. Repetitive bilateral arm training with rhythmic auditory cueing improves motor function in chronic hemiparetic patients. Stroke 2000; 31:2390–2398.
226
Hesse and Werner
18. Powell J, Pandyan D, Granat M, Cameron M. Electrical stimulation of wrist extensors in poststroke hemiplegia. Stroke 1999; 30:1384–1389. 19. Caraugh J, Light K, Kim S, Thigpen M, Behrmann A. Chronic motor dysfunction after stroke. Recovering wrist and finger extension by electromyographytriggered neuromuscular stimulation. Stroke 2000; 31:1360–1364. 20. Hummelsheim H, Maier-Loth ML, Eickhof C. The functional value of electrical stimulation for the rehabilitation of the hand in stroke patients. Scand J Rehabil Med 1997; 29:3–10. 21. Hogan N, Krebs J, Charnnarong J, Sharon A. MIT-MANUS: a workstation for manual therapy and Training II. In: Proceedings of the SPIE Conference on Telemanipulator Technology 1992; 833:28. 22. Volpe BT, Krebs HI, Hogan N, Edelstein OTR L, Diels C, Aisen M. A novel approach to stroke rehabilitation. Neurology 2000; 54:1938–1944. 23. Lum PS, Burgar CG, Shor PC. Robot-assisted movement training compared with conventional therapy techniques for the rehabilitation of upper-limb motor function after stroke. Arch Phys Med Rehabil 2002; 83:952–959. 24. Hesse S, Schulte-Tigges G, Konrad M, Bardeleben A, Werner C. Robotassisted arm trainer for the passive and active practice of bilateral forearm and wrist movements in hemiparetic subjects. Arch Phys Med Rehabil 2003; 84:915–920. 25. Muelbacher W, Richards C, Ziermann U, Wittenberg G, Weltz D, Boroojerdi B, Cohen L, Hallett M. Improving hand function in chronic stroke. Arch Neurol 2002; 59:1278–1282. 26. Hesse S, Brandl-Hesse B, Bardeleben A, Werner C, Funk M. Botulinum toxin A treatment of adult upper and lower limb spasticity. Drugs Aging 2001; 18:255–262. 27. Brashear A, McAfee AL, Kuhn ER, Ambrosius WT. Treatment with Botulinum toxin type B for upper-limb spasticity. Arch Phys Med Rehabil 2003; 84:103–107. 28. Hughes R, Walker BC. Influence of nerve-ending activity and of drugs on therate of paralysis of rat-diaphragm preparations by Clostridium Botulinum type toxin A. J Physiol 1962; 160:221–233. 29. Hesse S, Jahnke MT, Luecke D, Mauritz KH. Short-term electrical stimulation enhances the effectiveness of Botulinum toxin in the treatment of lower limb spasticity in hemiparetic patients. Neurosci Lett 1995; 201:37–40. 30. Dykstra DD, Wieting J, McCuire J, Kowalkowski T. Maximizing extraction of Botulinum toxin type A from vials. Arch Phys Med Rehabil 2002; 83: 1638–1640. 31. Simpson DM, Alexander DN, O’Brian CF. Botulinum toxin type A in the treatment of upper extremity spasticity: a randomized, double-blind, placebocontrolled trial. Neurology 1996; 46:1306–1310. 32. Bakheit AM, Thilmann AF, Ward AB, Poewe W, Wissel J, Muller J, Benecke R, Collin C, Muller F, Ward CD, Neumann C. A randomized, double-blind, placebo-controlled, dose-ranging study to compare the efficacy and safety of three doses of Botulinum toxin type A (Dysport) with placebo in upper limb spasticity after stroke. Stroke 2000; 31:2402–2406.
Motor Rehabilitation After Stroke 33.
34. 35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
227
Hesse S, Reiter F, Konrad M, Jahnke M. Botulinum toxin type A and shortterm electrical stimulation in the treatment of upper limb flexor spasticity after stroke: a randomised, double-blind, placebo-controlled trial. Clin Rehabil 1998; 12:381–388. Meythaler JM, Guin-Renfroe S, Brunner R, Hadley MN. Intrathecal baclofen for spastic hypertonia from stroke. Stroke 2001; 32:2099–2109. Hesse S, Bertelt C, Jahnke MT, Schaffrin A, Baake P, Malezic M, Mauritz KH. Treadmill training with partial body weight support as compared to physiotherapy in non-ambulatory hemiparetic patients. Stroke 1995; 26:976–981. Lovely RG, Gregor RJ, Roy RR, Edgerton VR. Effects of training on the recovery of full weight-bearing stepping in the adult spinal cat. Exp Neurol 1986; 92:421–435. Hesse S, Konrad M, Uhlenbrock D. Treadmill walking with partial body weight support versus floor walking in hemiparetic subjects. Arch Phys Med Rehabil 1999; 80:421–427. Werner C, Bardeleben A, Mauritz KH, Kirker S, Hesse S. Treadmill training with partial body weight support and physiotherapy in stroke patients: a preliminary comparison. Eur J Neurol 2002; 9:l–6. Visintin M, Barbeau H, Korner-Bitensky N, Mayo NE. A new approach to retrain gait in stroke patients through body weight support and treadmill stimulation. Stroke 1998; 29:1122–1128. Kosak MC, Reding MJ. Comparison of partial body weight-supported treadmill gait training versus aggressive bracing assisted walking poststroke. Neurorehabil Neural Repair 2000; 14:13–19. Nilsson L, Carlsson J, Danielsson A, Fugl-Meyer A, Hellstro¨m K, Kristensen L. Walking training of patients with hemiparesis at an early stage after stroke: a comparison of walking training on a treadmill with body weight support and walking training on the ground. Clin Rehabil 2001; 15:515–527. Da Cunha Filho IT, Lim PAC, Qureshy H, Henson H, Monga T, Protas EJ. A comparison of regular rehabilitation and regular rehabilitation with supported treadmill ambulation training for acute stroke patients. J Rehabil Res Dev 2001; 38:87–97. Laufer Y, Dickstein R, Chefez Y, Marcovitz E. The effect of treadmill training on the ambulation of stroke survivors in the early stages of rehabilitation: a randomised study. J Rehabil Res Dev 2001; 38:69–78. Pohl M, Mehrholz J, Ritschel C, Ru¨ckriem S. Speed-dependent treadmill training in ambulatory hemiparetic stroke patients: a randomized controlled trial. Stroke 2002; 33:553–558. Sullivan KJ, Knowlton BJ, Dobkin BH. Step training with body weight support: effect of treadmill speed and practice paradigms on poststroke locomotor recovery. Arch Phys Med Rehabil 2002; 83:683–691. Hesse S, Uhlenbrock D, Werner C, Bardeleben A. A mechanized gait trainer for restoring gait in non-ambulatory subjects. Arch Phys Med Rehabil 2000; 81:1158–1161. Colombo G, Wirz M, Dietz V. Driven gait orthosis for improvement of locomotor training in paraplegic patients. Spinal Cord 2001; 39:252–255.
228
Hesse and Werner
48. Werner C, von Frankenberg S, Treig T, Bardeleben A, Hesse S. Treadmill training with partial body weight support and an electromechanical gait trainer for restoration of gait in subacute stroke patients: a randomised cross-over study. Stroke 2002; 33:111–118. 49. Hesse S, Werner C, Uhlenbrock D, von Frankenberg S, Bardelebn A, BrandlHesse B. An electromechanical gait trainer for restoration of gait in hemiparetic stroke patients: preliminary results. Neurorehabil Neural Repair 2001; 13: 157–165. 50. Dietz V, Colombo G, Jensen L. Locomotor activity in spinal man. Lancet 1994; 344:1260–1263. 51. Taylor P, Burridge J, Dunkerely A. Clinical use of the Oddstock dropped foot stimulator: its effect on the speed and effort of walking. Arch Phys Med Rehabil 1999; 80:1577–1583. 52. Thaut MH, Mclntosh GC, Rice RR. Rhythmic facilitation of gait training in hemiparetic stroke patients. J Neurol Sci 1997; 151:207–212. 53. Hesse S, Krajnik J, Luecke D, Jahnke MT, Gregoric M, Mauritz KH. Ankle muscle activity before and after botulinum toxin therapy for lower limb extensor spasticity in chronic hemiparetic patients. Stroke 1996; 27:455–460. 54. Burbaud P, Wiart L, Dubos JL, Gaujard E, Debeillex X, Joseph PA, Mazaux JM, Bioulac B, Barat M, Lagueny A. A randomised, double blind, placebo controlled trial of botulinum toxin in the treatment of spastic foot in hemiparetic patients. J Neurol Neurosurg Psychiatry 1996; 61:265–269. 55. Decq P, Cuny E, Filipetti P, Keravel Y. Role of the soleus muscle in spastic equinus foot. Lancet 1998; 352:118. 56. Reiter F, Danni M, Lagalla G, Ceravolo MG, Provinciali L. Low-dose botulinum toxin with ankle tapping for the treatment of spastic equinovarus foot after stroke. Arch Phys Med Rehabil 1998; 79:532–535.
12 Neural Bases for Rehabilitation After Stroke Randolph J. Nudo, Ann M. Stowe, Ines Eisner-Janowicz, and Numa Dancause Center on Aging and Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas, U.S.A.
Rapid developments in our understanding of neurobiological mechanisms of neuronal plasticity and recovery after injury, coupled with advances in behavioral theory and methodology for physiotherapeutic interventions, are now resulting in a fundamental change in treatment approach to neurological disorders. New intervention strategies are beginning to be based not only upon empirical clinical trial evidence, but also upon evidence-based neuroscientific models of plasticity and repair mechanisms, derived from animal and human studies. This recent change in approach is thought to be so integral to current thinking, and so rapid, that some have referred to an impending new paradigm shift in neurorehabilitation (1,2). In this chapter, we will review some of the more important neuroscientific data that have contributed to the development of new recovery models. Since the vast majority of recovery models employ injury to the primary sensorimotor cortex [either via middle cerebral artery occlusion (MCAo), occlusion of its end-arteries or destruction of the supplied tissue], and since the majority of clinical assessments after sensorimotor cortex injury are related to motor behavior, our review will focus on motor recovery. 229
230
1.
Nudo et al.
ANIMAL MODELS OF STROKE AND REHABILITATION
Several factors that affect the severity of the behavioral deficits and the degree and rate of recovery in animal models of cortical injury have been proposed, including the species of animal used, the model of lesion induction, the size of the lesion, and the age of the animal. A recent review by Traystman (3) summarizes currently popular models. Animal models of neurological deficits are essential for the assessment of new therapeutic options. Models of cerebral ischemia can be performed in both small and large animals, and each model has its advantages and disadvantages. For example, the development and proliferation of transgenic and knockout mouse mutants have provided new types of animals to study genomic and proteomic involvement in mechanisms of cell injury and neuroprotection from ischemia and reperfusion (3). Each animal model has its unique set of strengths and limitations. The choice of the most appropriate model should be based, not simply on cost and convenience, but on the scope and specific questions of the particular experiments, as well as on the anatomical and physiological characteristics of the species. Recent development of a greater variety of animal models of stroke and ischemia, and the increased focus on recovery models, are resulting in substantial progress towards bridging the gap between animal models and neurorehabilitation. Perhaps due to a large number of failed neuroprotection trials based on rodent models, it has been suggested that rats are not as appropriate as primates (or, alternatively, large animal models) for the symptomatic modeling of human disease. However, a large body of data suggests that many aspects of movement disorders after cortical injury are conserved across many mammalian orders. Analyses of movements in rats and primates show striking similarities, if not homologies, of many motor patterns across species. Rat models of hemiplegia, neglect, and tactile extinction are useful in assessing the outcome of ischemic or traumatic brain injury, and in monitoring the effects of therapeutic interventions. Studies in rodents that emphasize detailed behavioral analysis should continue to be developed as effective and inexpensive models that complement studies in primates (4). Nevertheless, there are important differences even within the same species that need to be taken into account. Duverger and MacKenzie (5) studied the influence of rat strain and arterial pressure (as well as other variables) on infarct volume after cerebral infarction. Of the normotensive rat strains (Wistar, Sprague–Dawley, Fischer-344), the most reproducible volume of infarcted tissue was seen in Fischer-344 rats. Also, the mean infarct volume did not vary from one series to another in this strain. Yet, the infarct of spontaneously hypertensive rat (SHR) and spontaneously hypertensive– stroke prone rat (SHRSP) strains was approximately 1.5 times that of the normotensive strains.
Neural Bases for Rehabilitation After Stroke
231
Although rodent species have been very valuable in the understanding of the cascade of events that follow acute stroke, and have begun to be utilized for studies of recovery mechanisms (6), the need for nonhuman primate models of stroke has received increasing attention. The primary advantage of nonhuman primate stroke models is that their vascular and cortical structures more closely resemble cognate human structures in their anatomy, morphology, cellular physiology, and biochemistry. Nonhuman primate models are thus uniquely useful in research on the cerebrovascular pathophysiology of focal ischemia resembling human ischemia (7). The advantages of one such nonhuman primate species, the squirrel monkey, are discussed more specifically in this chapter.
2.
EFFECTS OF MOTOR INJURY ON BEHAVIORAL FUNCTION: RECOVERY AND COMPENSATION
In animal as well as human studies of motor recovery after injury to sensorimotor cortex, the most severe deficits in the affected body parts are seen within the first few days (e.g., flaccid paralysis, movements limited to reflexive actions). However, over the course of the next several days to weeks, motor ability improves and eventually reaches a plateau. Further improvements, in the absence of intervention, are small, often resulting in permanent motor impairments (8,9). For example, in a squirrel monkey, we have replicated the symptoms often encountered in human stroke survivors. In this model, we have shown that manual skill, as assessed in a pellet retrieval task, results in severe deficits during the first 1–2 weeks after an ischemic infarct in the primary motor cortex (Ml) hand area (Fig. 1). Recovery proceeds rapidly during the ensuing weeks, and eventually plateaus by about 3 months postinfarct (8). Interpretation of the extent of motor recovery depends largely on the type of motor assessment one employs. For example, Fig. 1 depicts two measures of motor performance during the retrieval of food pellets by squirrel monkeys after an Ml hand area infarct. If success rate is measured (that is, the percentage of trials that the food pellet is actually retrieved and brought to the mouth), recovery can be said to be complete by about 3 months. This is a functional measure, evaluating whether the monkey can perform the task successfully and obtain a reward. However, if a more specific measure of motor skill is assessed (e.g., duration of retrieval or flexions per retrieval), one finds that a motor deficit persists well beyond 5 months. Similar persistent deficits, such as abnormal interjoint co-ordination and slower velocities, are also reported in human stroke survivors who are considered to be fully functional by traditional evaluation tools (e.g., Fugl-Meyer). This is an impairment measure, evaluating the difficulty of performing a specific motor task. The conclusion drawn from both the clinical
232
Nudo et al.
Figure 1 Time course of motor recovery after injury to primary motor cortex in squirrel monkeys. Dotted line indicates normal performance. ‘‘Percentage success’’ ¼ Percentage of trials in which a food pellet is retrieved from a small well and brought to the mouth. ‘‘1=duration’’ ¼ the inverse of the time required to extract the pellet from the well. Duration is proportional to the number of finger flexions required to retrieve a pellet. Rapid recovery occurs in the first few weeks after injury. Recovery decelerates and then plateaus by 3–4 months postinjury. While success (a functional measure) reaches normal, preinjury performance levels, a residual deficit in duration persists (an impairment measure).
and basic science literature is that much of recovery after brain injury is due to behavioral compensation, or the development of new sensorimotor strategies to circumvent long-lasting impairments (10–12). For example, behavioral compensation following stroke affecting skilled use of the hand may enable the individual to complete goal-directed tasks on a functional level through the recruitment of proximal limb and postural muscles not usually observed during ‘‘normal’’ movement (13). In a more extreme example, after unilateral lesions to sensorimotor cortex, rats develop a hyperreliance on the nonimpaired forelimb (10). These compensatory strategies often compound the assessment of ‘‘true’’ functional recovery following stroke. In humans, the presence of ‘‘abnormal’’ proximal compensatory strategies could potentially interfere with maximal behavioral recovery (14). Whishaw et al. (15) demonstrated a similar effect in rats in a study of the effects of cortical lesion size on motor performance in a reaching task. Although rats with larger lesions performed about as well as rats with small lesions, they showed that the recovery was accomplished in different ways. Rats with small lesions used movements identical to normal rats, whereas rats with large lesions used compensatory movements to achieve the goal. The debate between compensation and ‘‘true’’ recovery is quite complex
Neural Bases for Rehabilitation After Stroke
233
and well beyond the scope of this review. However, further progress in neurorehabilitation will require a critical examination of this issue from both behavioral and neural perspectives. 3.
NEUROPHYSIOLOGICAL CONSEQUENCES OF MOTOR CORTEX INJURY
Neurophysiological studies have shown repeatedly that there is a close correlation between the topography of somatosensory and motor maps and sensorimotor experience and skill (16–20). For example, in squirrel monkeys, the motor hand representation is larger in the hemisphere opposite the monkey’s preferred hand and is spatially more fractionated (21). As monkeys acquire skill on a manual dexterity task, the representation of the digits enlarges, paralleling results in humans (22,23). Representations of specific joint–movement combinations used in the task also enlarge, suggesting that motor skill learning drives the development of neural circuits in the motor cortex that are responsible for muscle and joint synergy (16). Thus, neurophysiological maps of motor topography can essentially be used as surrogate markers of recovery of skilled behavior in injury models. The study of topographic map plasticity allows for the testing of hypotheses regarding recovery mechanisms. In addition, neurophysiological techniques can assess the function of peri-infarct and remote cortical areas multiple times throughout the recovery process, allowing the timing of cortical plasticity to be determined. One popular technique for the derivation of functional representational maps in rodent and nonhuman primate models of stroke is intracortical microstimulation (ICMS). Typically, in a ketamine-anesthetized animal, a fine microelectrode is advanced into the cortical gray matter of the motor cortex until the tip reaches layer V, in the vicinity of corticospinal somata. Then, a short train burst of cathodal (or biphasic) current is delivered. The current is increased until a visible movement or electromyographic activity is evoked (typically less than 30 mA). By performing this stimulation procedure at multiple sites across motor cortex, the functional topography of the motor output map can be derived (Fig. 2). The invasive nature of the ICMS technique allows the derivation of functional maps at much higher resolution than can be achieved with noninvasive approaches, such as functional magnetic resonance imaging (fMRI) and PET. Higher resolution maps allow the detection of more subtle changes in functional topography. Additionally, animal models of cortical ischemia are advantageous since the size and location of the cortical injury can be precisely controlled. Human strokes are quite variable in extent and location and significant distortions of normal anatomy may be induced (24). Also, preinfarct data are rarely available. Because cortical motor maps display significant intersubject variability (25,26), subtle changes caused by the infarct are more difficult to detect.
234
Nudo et al.
Figure 2 Methods for defining functional organization of topographic maps in motor cortex. In anesthetized squirrel monkeys, the primary motor cortex is exposed. Then a microelectrode is driven into the cortical gray matter until the tip reaches layer V (approximately 1700–1800 mm in squirrel monkey). Movements evoked by nearthreshold current stimulation are defined sequentially at each site. In the figure above, each dot represents a microelectrode penetration site. The bold outline encloses sites where stimulation evoked movements of the hand (digit, wrist, and forearm). Calibration bar is 1 mm.
In our laboratory, we have been using the squirrel monkey for studies of cortical motor map plasticity for two decades and have compiled a large database of behavioral, physiological, and anatomical information regarding the cortical control of movement in this species (8). The squirrel monkey is a small New World primate species, possessing unique qualities that make it an attractive model for understanding the neurophysiological and neuroanatomical mechanisms that may underlie stroke recovery. The Ml hand area has been the focus of several experiments since the hand is often affected in human stroke survivors. Ischemic infarcts in this nonhuman primate model are more limited than those that occur in human stroke, typically encompassing solely the Ml hand representation, while sparing proximal upper extremity representations. The Ml hand area of the adult squirrel monkey is contained in a relatively flat, unfissured sector of the frontal cortex, allowing direct access for neurophysiological examination and infarct induction (Fig. 2). The hand dexterity and prehensile grip of the squirrel monkey allow for assessment of impairment and functional recovery outcome variables paralleling the effects of mild-to-moderate strokes in humans. Despite
Neural Bases for Rehabilitation After Stroke
235
numerous physiological and behavioral dissimilarities between squirrel monkey and man, the primary motor and premotor cortices of both species appear to be homologous (27). Following motor mapping procedures to precisely define functional boundaries in Ml, a small ischemic infarct is placed in all or part of the Ml hand area. In initial studies, we first sought to describe motor map plasticity in the spared tissue adjacent to the infarct (28). In these studies, the infarct was limited to 30% of the Ml hand representation. Then the animals were allowed to recover spontaneously (i.e., no behavioral intervention postinfarct). Several weeks postinfarct, the infarcted tissue, and thus a portion of the Ml hand representation, had become necrotic. However, in the adjacent portion of the Ml hand representation that had been spared by the infarct, the hand area was reduced a further 50% from preinfarct values (see ‘‘spontaneous recovery’’ in Fig. 3). Stimulation of the cortical sites that had previously evoked distal hand and wrist movements now evoked proximal shoulder and elbow movements. A large number of factors probably contribute to the reduction in hand representation in the spared, adjacent tissue. First, motor impairment caused by the injury may result in a disuse effect. That is, since the
Figure 3 Plasticity in Ml hand area after small ischemic infarct. The graph depicts the size of the area outside of the infarct relative to its preinfarct size (%). White bar indicates the relative size of the spared hand area after spontaneous recovery. Black bar indicates the relative size of the spared hand area after repetitive training with the impaired limb during the acute phase. To encourage use of the impaired limb, a jacket restricted the use of the less-affected limb. Solid line indicates the size of the spared hand area when repetitive training was not employed, but the restraint jacket was in place. Thus, the monkey was forced to use the impaired limb for some activities, but no specific repetitive practice was employed. These data indicate the importance of repetitive motor training in motor map plasticity after cortical injury. (Data were adapted from Ref. 57.)
236
Nudo et al.
unaffected (or less-affected) hand is used extensively for skilled reaching in the acute and subacute period following the infarct, the size of the hand area in the injured Ml may be further reduced due to disuse of the more affected limb. Even in uninjured monkeys, if movement of the dominant hand is restricted, the digit representations are reduced in size (Nudo and Milliken, unpublished observations). Likewise, in humans, disuse consequent to immobilization has been shown to induce a reduction of the cortical representational area of involved muscles (29). Second, it is possible that events set in motion during the initial postischemic period may result in long-lasting disruption of cortical circuitry in adjacent and remote tissue, as reflected in motor maps. Especially well characterized are the effects of an ischemic infarct on inhibitory neurotransmission in the cortex. The most ubiquitous inhibitory neurotransmitter is g-aminobutyric acid (GABA), predominantly associated with local circuit interneurons, though GABA projection neurons do exist. The rodent photothrombotic method for producing cortical ischemia has been used extensively to evaluate changes in GABAergic neurons. Inhibitory interneurons in the peri-infarct cortex undergo degeneration of dendrites and decline in neuronal numbers evident during the first days postinfarct (30,31). Schiene and colleagues showed a peri-infarct neuronal hyperexcitability evident at 1 day postinfarct and peaking 3–7 days postinfarct, coinciding with the loss of GABA activation (32). These experiments have also shown that GABAA downregulation and hyperexcitability may last for weeks. Furthermore, Redecker et al. (33) suggested that a postinfarct reduction in GABAergic neurons may be mediated through NMDA receptors. Application of NMDA-receptor antagonist MK-801 during lesion induction attenuated the characteristic loss of four GABAA receptor subunits in both ipsilesional and contralesional hemispheres. g–aminobutyric acid downregulation and hyperexcitability may have both maladaptive and adaptive effects. Hyperexcitability may be associated with increased excitotoxicity. Supporting a neuroprotective role for GABA is a study involving mutant mice lacking the Cav2.3 channel associated with GABAergic neurons. Following an MCAo, (=) Cav2.3 mice showed significant increases in infarct size at 24 hr postinfarct (34). But, if inhibitory neurons are neuroprotective, possibly through reduction of excitotoxic damage, then the question remains: why is the system experimentally downregulated following stroke? It has been suggested that GABA downregulation may serve a supportive role in plasticity of neuronal circuits after injury (30,35). Potentially, unmasking of latent intracortical connections can be accomplished via selective downregulation of GABA. Studies addressing this issue are still in their infancy. Alterations in functional topography following Ml injury are not limited to the immediately adjacent cortex. Mapping studies have also demonstrated changes in remote but functionally related areas of the ipsilesional
Neural Bases for Rehabilitation After Stroke
237
cortex following ischemic infarct. The ventral premotor cortex (PMv) hand area has been the focus of recent studies due to its proximity to the Ml hand area, allowing easy access in the same mapping procedure. Also, PMv has been implicated in the control of visually guided hand movements (36,37). In the squirrel monkey cortical ischemia model, the hand representation in PMv expanded several weeks following an infarct in the Ml hand representation and subsequent spontaneous recovery. This contrasts with the earlier results in adjacent Ml cortex (38). Further, it appears that substantial injury to the Ml hand area is necessary for the expansion to be demonstrated in PMv. An infarct destroying less than 50% of the Ml hand area resulted in very little change in PMv, while complete destruction resulted in a 50% increase in the PMv hand representation. While it is not yet clear what drives the functional changes in PMv, it is possible that the disruption of corticocortical connectivity between Ml and PMv may be critical (see below). Recent studies in human stroke survivors suggest that both the intact, peri-infarct zone and more remote cortical motor areas may play a role in neurological recovery (39–42). Using transcranial magnetic stimulation (TMS) after stroke, it has been shown that the excitability of motor cortex is reduced, and the cortical representation of the affected muscles is decreased (43,44). It is likely that these changes occur from a combination of diaschisis-like effects (45) and disuse of the affected limb (46). As a reduction in blood flow and excitotoxicity disrupts cortical function, excitatory input is lost to remote neurons interconnected with the infarcted region, thereby suppressing remote neuronal activity and metabolism. Perilesional changes in cortical activity have been shown using a variety of neuroimaging techniques (41,47–49). While the precise location of the injury is often unknown and=or uncontrolled, these human studies have consistently shown that functional changes occur in several cortical areas after stroke or other cortical damage, paralleling results from animal experiments (50,51). A common finding in neuroimaging studies after stroke is that other motor areas in the ipsilesional and contralesional hemisphere undergo functional changes. Seitz et al. (52) determined cortical activation patterns in the MCA distribution of stroke survivors. Regional cerebral blood flow (CBF) changed in both ipsilesional and contralesional premotor cortices during a finger tapping protocol. This result may support a role for these remote areas during clinical recovery. In another fMRI study, 14 stroke survivors were monitored in three imaging sessions over the first 6 months postinfarct. The research suggested two distinct processes in cortical activation following stroke. There is an initial ‘‘recruitment’’ of cortical areas not activated by hand movement in control cases, including ipsilesional and contralesional sensorimotor cortex (SMC), frontal premotor areas, and superior parietal areas. Then, in individuals without a disruption of Ml hand control, a second process of ‘‘focusing’’ occurred in which the activation was restricted to contralesional
238
Nudo et al.
sensorimotor cortex. Individuals with an Ml hand area lesion showed a persistent recruitment of ipsilesional premotor areas, echoing the results of PMv mapping studies in nonhuman primates. However, the relationship between neuroimaging findings and the degree of functional recovery after stroke is not yet clear. Some recent studies have found a negative relationship between the degree of task-related activation in motor areas and recovery, suggesting that further study of neuroimaging patterns after stroke is now critical (53).
4.
INFLUENCE OF POSTINJURY MOTOR EXPERIENCE ON REORGANIZATION OF MOTOR MAPS
Because it has long been suggested that physical therapeutic interventions might improve recovery after injury to motor cortex (54,55), the modulatory effects of postlesion motor training on motor map plasticity were also studied (56). In these experiments, monkeys were placed in restraint jackets that restricted the use of the unimpaired limb. Daily repetitive training procedures were employed to encourage improvement in manual skill as a form of rehabilitative training. After manual skill had returned to normal levels, the motor cortex was re-examined with ICMS techniques. In contrast to spontaneously recovering monkeys, these monkeys retained undamaged hand representations after postinjury behavioral training. On average, there was a net gain of approximately 10% in the total hand area adjacent to the lesion, in contrast to a nearly 50% decrease after spontaneous recovery (see ‘‘repetitive training’’ in Fig. 3). More recently, it has been shown that the retention of hand area adjacent to a microlesion in Ml requires repetitive behavioral training, since use of a restraint jacket alone resulted in little change in hand representations beyond that which was seen with spontaneous recovery (57). Hand representations in monkeys that wore the restraint jacket continuously for up to 1 year were only 80% of the prelesion area (see ‘‘jacket restraint—no training’’ in Fig. 3). It is interesting to note that studies in normal, uninjured brains parallel these results suggesting that repetitive practice of motor skills is necessary to drive neurophysiological changes. In both squirrel monkeys and rats, acquisition of motor skill results in enlarged representations of the trained limb. Conversely, repetitive exercise in the absence of skill acquisition does not result in areal changes in representation (58,59). If postinjury motor training is necessary to maximally drive adaptive plasticity, it is important to determine whether there is an optimal window for delivering therapy after stroke. Therefore, delayed rehabilitation in the squirrel monkey model of cortical ischemia has also been investigated. Barbay et al. (60) found that initiation of motor skill training 1 month postinfarct significantly reclaimed areal hand representation when compared to
Neural Bases for Rehabilitation After Stroke
239
spontaneous recovery controls (increase of 56%). However, postrecovery motor hand representations were smaller than those observed with early initiation of training. Thus, delayed rehabilitative training results were intermediate between spontaneous recovery and acute phase training. While it is positive that physiological, and perhaps concurrent anatomical, changes may be induced by therapy during chronic stroke, these data also suggest that the adaptability of the cerebral cortex declines with time after injury. Further evidence that reorganization of intact, adjacent cortical tissue is associated with functional recovery has come from similar motor mapping studies in rats. In these studies, after bilateral ablation of the forelimb area in rat motor cortex, motor performance was impaired. However, recovery occurred if electrical stimulation of the ventral tegmental nucleus was paired with forelimb movements (61). The stimulation is thought to have played a motivational role in encouraging forelimb use. When the motor cortex was re-examined following recovery, a novel forelimb representation appeared caudal and lateral to the ablated representation. The size of this representation was directly related to the behavioral performance of the recovered animals. Finally, ablation of the newly emerged forelimb representation resulted in reinstatement of the deficit. In studies in humans, after several weeks of rehabilitation, motor representations in the injured hemisphere are enlarged relative to the initial postinjury map (44). Also, constraint-induced (CI) movement therapy, in which the unimpaired hand is constrained to induce goal-directed movement with the impaired hand, produces a significant enlargement of the representation of the paretic limb closely paralleling results in nonhuman primates (56,62–65). 5.
ANATOMICAL CONSEQUENCES FOLLOWING MOTOR CORTEX INJURY
Neuroanatomical techniques can provide strong evidence that neuronal plasticity occurs after cortical injury. A large body of data has emerged from the animal literature, suggesting that several morphological features of cortical structure may be beneficial to functional recovery. While the vast majority of studies have focused on dendritic arborization, synaptogenesis, and axonal sprouting, recently the importance of other neuronal and nonneuronal forms of plasticity have been recognized (66). A variety of anatomical changes occur in a time-dependent fashion in the remaining, intact cortex after cortical injury. After injury to the forelimb representation in rats, Jones and Schallert (67) demonstrated increases in dendritic arborization at 18 days postinfarct. The overgrowth of dendrites corresponds to the required behavioral dependence on the ipsilesional (less-affected) forelimb. Dendrites then undergo subsequent branch pruning with concomitant increases in spine densities (68). The arborization and pruning is correlated with the return of behavioral function during the
240
Nudo et al.
subacute stage (69). By 30 days postinfarct, dendritic branching returns to control levels. Also, dendritic arborization and axonal outgrowth occur prior to the development of mature synapses (70). An interesting comparison of contralesional dendritic changes following lesions induced by either electrolytic or aspiration methods suggest that contralesional arborization may require the presence of the injured tissue (71). The time course of axonal sprouting closely echoes that of dendritic arborization. Early events include changes in axonal-related proteins. Transient rodent MCAo induced peri-infarct increases in mRNA associated with proteins for axonal growth (growth-associated protein-43, GAP-43) and myelination (myelin basic protein, MBP) (72). Growth-associated protein-43 and myelin basic protein mRNAs began upregulation at 24 hr postinfarct and peaked between 3 and 7 days postinfarct. Functional periinfarct oligodendrocytes begin to appear 1 week postinfarct. Another axonal protein, tau, is currently being investigated as a means to clinically monitor axonal degeneration following stroke (73). Tau protein is found in elevated levels in cerebrospinal fluid following neuronal=axonal injury in humans. Serum tau was detectable at 6 hr postinfarct and peaked between 3 and 5 days postinfarct. The serum data positively correlated with chronic loss of function and final infarct volume when assessed at 3 months postinfarct. Until recently, few studies had found direct evidence for axonal sprouting in adult cortical injury models. However, Carmichael and colleagues (74,75) have recently investigated axonal sprouting in the somatosensory barrel cortex of rats following ligation of distal branches of the MCA. At 21 days postinfarct, biotinylated dextran amine (BDA) was injected into peri-infarct cortex and allowed to retrogradely label neurons for another 7 days. The axons significantly shifted in orientation from control cortex and included sprouting from peri-infarct neurons, suggesting a functional salvage of tissue immediately adjacent to the infarct. Furthermore, there was a significant increase in the overall number of horizontal connections in ipsilesional vs. contralesional hemispheres. Thus, neurons in the peri-infarct cortex alter their axonal trajectories, a phenomenon potentially related to recovery. These events, however, are not solely confined to the peri-infarct regions. Secondary and tertiary events in structures connected with the injured cortical territory may occur over a long period of time during the recovery process. As expected, most cell death of neurons in the ischemic core is considered complete by 24–48 hr postinfarct. Axonal degeneration in remote areas connected to the injured zone may take weeks to months for cell death to be completed (76). Retrograde degeneration may occur through the elimination of retrogradely transported neurotrophins due to necrosis of target. Anterograde degeneration may be induced through excitotoxicity. In a rat model of MCAo, Nakane et al. (77) examined secondary
Neural Bases for Rehabilitation After Stroke
241
degeneration using MRI. Focusing on the thalamus and substantia nigra, as both share connections with the infarct, MR images showed shrinkage of both areas by 14 days. This supports the researchers’ earlier findings of neuronal necrosis, gliosis, and atrophy in the thalamus and substantia nigra beginning at 4 and 7 days postinfarct, respectively. These changes were not visible before 2 days postinfarct. Recent anatomical studies in nonhuman primates also support remote axonal sprouting. In a squirrel monkey model of cortical ischemia, an infarct was induced in the hand representation of Ml, as defined through ICMS techniques. Functional ablation of the Ml hand area resulted in new axonal and nerve terminal sprouting of the neurons in PMv (78). Secondary motor areas, such as PMv and supplementary motor area (SMA), have reciprocal connections with Ml and play a significant role in the control of voluntary movement (79). Dancause and colleagues found axonal projections from PMv to primary somatosensory cortex (probably areas 1 and 2), which were considerably more numerous than observed in control cases. As with dendritic changes in contralesional homotopic cortex, reorganization of axonal connections in remote but related ipsilesional areas may aid in the restitution of function following stroke (80). Anatomical changes in dendrites and axons culminate in increases in synapse number or synaptic strength. Stroemer et al. (81) investigated synaptogenesis following a combined distal MCAo in chronically hypertensive rats. Antibodies against synaptophysin, a protein found in nearly all nerve terminals, were used to characterize synapses. Synaptophysin reactivity reached significant levels in ipsilesional and contralesional cortices at 14, 30, and 60 days postinfarct. This late induction confirms the order of axonal and dendritic growth prior to synaptogenesis. The levels of synaptophysin also had a significant correlation to improved behavioral recovery in this model. Jones et al. (82) confirmed the contralesional synaptogenesis using electrolytic lesions in the forelimb motor and somatosensory representation in rats. At 30 days postinfarct, contralesional homotopic cortex showed a significant increase in the number of synapses per neuron in layer V. Once again, there was no significant increase in synapses at either 10 or 18 days postinfarct.
6.
INFLUENCE OF POSTINJURY MOTOR EXPERIENCE ON NEUROANATOMICAL PLASTICITY
Studies of motor training in normal, uninjured animals have contributed a wealth of information on neuroanatomical plasticity that is relevant to understanding similar changes after injury. In a rodent model of motor skill learning, the cerebellar effects of acrobatic training vs. exercise were determined (83). Motor skill training resulted in synaptogenesis with a subsequent increase in neuropil volume, whereas repetitive exercise, with little
242
Nudo et al.
to no skill required, did not. Both, however, induced angiogenesis, which may have occurred in response to plasticity-associated increases in metabolic demands. Further studies confirmed cerebellar synaptogenesis and revealed neocortical synaptogenesis in rats undergoing motor skill learning (84,85). Synaptogenesis in motor cortex was associated only with cortical tissue in areas that had undergone plasticity in representational maps. Therefore, anatomical changes coincide with physiological changes following exposure to motor learning tasks. A growing body of research has now investigated the specific effects of motor skill training following an infarct in animal models of stroke. For example, in a rodent MCAo study, rats experiencing an enriched environment coupled with motor skill training had significantly improved behavioral recovery at 4 and 9 weeks postinfarct vs. standard-housed spontaneous recovery controls (86). Dendritic arborization was also significantly increased in this group, though no reduction in infarct volume occurred. These studies are important to understanding recovery after stroke, since therapies employing repetitive massed practice are becoming increasingly popular. Evidence supports the hypothesis that neuroanatomical changes in the contralesional (intact) cortex are related to hyper-reliance on the ipsilateral limb. Binding of the ipsilateral forelimb in the first week, but not the second week postinfarct, greatly reduced dendritic arborization in contralesional homotopic cortex (67,87). Ipsilateral limb immobilization also significantly increased infarct volume vs. either contralesional limb binding or spontaneous recovery controls. It is possible that ipsilateral forelimb binding and subsequent use of the contralateral limb results in overexcitation of the vulnerable peri-infarct cortex, enhancing neuronal excitotoxicity and thus cell death (88). Administration of an NMDA-receptor antagonist during ipsilateral forelimb binding for 7 days postinfarct attenuated the increase in infarct volume (89). Larger and more enduring behavioral deficits coincided with these anatomical changes in the ipsilateral limb binding group (88). The enhanced behavioral deficits were still significant 8–14 days postinfarct and could also be attenuated with NMDA-receptor antagonist administration (87). Resolution of these behavioral changes (i.e., a return to symmetry) required that the infarct and peri-infarct regions remain intact. Chronic postural and motor impairments of the contralateral limb following an aspiration lesion may result from the loss of infarct and peri-infarct related structures (71). Furthermore, ipsilateral forelimb binding during the pruning stage did not affect pruning or the reintroduction of symmetrical limb use. These anatomical and behavioral results suggest not only a requirement for the infarcted tissue to be present, but also behavioral interactions to optimize plasticity. It can be suggested that early ipsilateral limb reliance and contralateral limb behavioral compensation may boost plasticity and recovery of function following cortical injury in this model.
Neural Bases for Rehabilitation After Stroke
7.
243
NEW APPROACHES TO MAXIMIZE NEUROPLASTICITY AND RECOVERY AFTER STROKE
Several new approaches to treatment in chronic stages after stroke have been initiated or proposed within the past several years. It may be possible to optimize interventions based on those strategies that maximize neuroplasticity mechanisms. A few of the more popular and novel strategies are briefly considered below. 7.1.
Pharmacotherapies
Modulation of recovery from brain injury, through the use of diverse pharmacological agents, has been demonstrated in several recent studies (90–95). Drugs targeting modulation of neurotransmitter systems have shown great potential. Particularly, drugs like d-amphetamine, which targets central noradrenergic activity, have received increasing attention. Amphetamine prolongs noradrenergic activation of postsynaptic receptors (for a review, see Ref. 96) and enhances long-term potentiation (97). Additionally, there is evidence that d-amphetamine facilitates neuroanatomical plasticity (e.g., synaptogenesis), especially when administration is combined with behavioral experience (98). In fact, the combined regimen of behavioral experience and amphetamine has been shown to be interactive, with a tight temporal coupling necessary for optimal recovery (99). Other factors contributing to optimal recovery include the frequency in which the combined behavioral and amphetamine treatments are implemented and the size of the dosages used (100). Several clinical studies in stroke survivors support a beneficial role of amphetamine when followed by physical therapy (101–105). Another pharmacological approach that aims at optimizing plasticity after stroke is administration of nerve growth factors (106). In rats, nerve growth factor, basic fibroblast growth factor, and osteogenic protein-1 have all been found to enhance recovery of sensorimotor function (107–110). Difficulties for the administration of these drugs, which do not cross the blood– brain barrier, have slowed this approach. Alternatively, mesenchymal stem and progenitor cells (MSC), which can pass through the blood–brain barrier, have been injected in the venous system of rats. Results indicate that enhanced recovery of function was associated with increased brain-derived neurotrophic factor and nerve growth factor and decreased apoptotic cells in the ischemic boundary of the lesion. Recent findings of substantial anatomical reorganization of connections following cortical injuries (70,75,78) provide support for strategies to modulate sprouting following strokes. 7.2.
Constraint-Induced Movement Therapy
Hypotheses that eventually led to the development of CI movement therapy (CIT) originated from fundamental experiments conducted in nonhuman
244
Nudo et al.
primates following deafferentation (111–116). In these experiments, disuse of the affected upper limb could be observed following the injury, and this maladaptive behavior would be sustained, perhaps permanently, without further intervention. However, the function of the deafferented limb could be greatly enhanced by forcing its use through restraint of the less-affected limb (115,116). These experiments led to the ‘‘learned nonuse’’ hypothesis stipulating that nonuse or less than maximal use of the deafferented limb results from negatively reinforced attempts to use the affected limb (117). Because of the phenomenon, the actual use of the affected limb remains well below its potential (117–119). The nonuse theory is not specific to stroke, but representative of a more general behavioral process that may underlie deficits in many neurological and non-neurological conditions (120). Constraint-induced movement therapy (121), as it was initially practiced, consisted of the use of a sling worn on the less-affected limb for 14 days. Use of this sling was combined with 6 hr of practice of functional movements with the impaired limb for 10 of those days. Later approaches utilized a mitt over the less-affected limb. The behavioral training used is based on movement repetition (122). In numerous experiments (117,118,123–126), stroke survivors have shown improvements beyond what had been reported with the sole use of the sling (127), which were maintained during a 2-year period of follow-up. The approach has shown efficacy in both the Motor Activity Log (MAL) and the Wolf Motor Function Test (WMFT) (128– 130). Most importantly, acquisitions made by the study participants have been shown to be transferable to other motor functions in their daily lives. A series of experiments to evaluate motor representation changes using fMRI and TMS following CIM paralleled the behavioral changes (46,62,129,131,132). These studies suggest behavioral therapies that employ repetitive practice of skilled motor activities support the hypothesis that postinjury motor recovery can be modulated through learning-dependent effects on neuroplasticity mechanisms. 7.3.
Treatments Based on Stimulation of the Cerebral Cortex
McKay et al. (133) recently developed a paired stimulation protocol using TMS. In this experiment, they showed that stimulation of the motor point of the dorsal interosseous muscle with TMS resulted in the expansion in muscle representation accompanied by large movement of the center of gravity (CoG). The changes, suggesting reorganization of the cortical representation zone, lasted for at least 2 days and up to 5 days without further stimulation. The progressive enlargement of the cortical representation area and the increase of motor evoked potential of these muscles are associated with functional recovery (134). Because of the similar nature of changes elicited by techniques such as paired stimulations (133) and those associated with functional recovery, it is possible to positively influence CNS
Neural Bases for Rehabilitation After Stroke
245
reorganization following diverse lesions with direct stimulation. Using this technique, it may be possible to enhance or facilitate cortical reorganization following brain injury, which may lead to improved functional outcomes. Recently, Muellbacher et al. (135) performed an original experiment in which they applied the principle following the plasticity of the cortical representational map in the cortex. The goal was to increase function of the upper limb by facilitating the increase of the hand and forearm cortical surface area over proximal body representations. To achieve this, they transiently deafferented the shoulder and upper arm cortical representation with regional anesthesia, while sparing the forearm and hand representation. This resulted in a constant state of competition between diverse body segments for a larger cortical area. Anesthesia was combined with hand motor practice. Regional anesthesia of the upper arm potentiated practice-induced improvement in hand motor function in chronic stroke survivors. In a novel treatment protocol, the application of subthreshold electrical stimulation using an indwelling surface electrode (Northstar Neuroscience) implanted over the peri-infarct cortex has recently been investigated. The hypothesis is that low-level electrical stimulation can influence physiological reorganization and behavioral recovery following ischemic cortical lesions. Following a cortical lesion in either nonhuman primates (136) or rats (137), behavioral training was combined with stimulation for about an hour each day for several days. The former study reported representational map expansion that was substantially greater than following spontaneous recovery with similar size infarcts (38). The latter study showed that animals with stimulation had faster and greater rates of behavioral improvement when compared to the animals that just underwent training following the stroke. 8.
SUMMARY
Interest in stroke recovery has increased rapidly in recent years, owing largely to the rapid accumulation of evidence for adaptive plasticity in the cerebral cortex after injury. These studies suggest that specific types of motor experience after stroke may augment the brain’s innate capacity for neural repair and compensation. Novel physiotherapeutic, pharmacotherapeutic, and stimulation-based techniques may provide significant new strategies to maximize the recovery process. Integration of theoretical models from neuroscience, motor control theory, and psychology are essential to rationally design interventions in the future. REFERENCES 1.
Nudo RJ, Nelson RJ. Introduction. Animal models of stroke and rehabilitation. ILAR J 2003; 44:81–82.
246
Nudo et al.
2.
Taub E, Uswatte G, Elbert T. New treatments in neurorehabilitation founded on basic research. Nat Rev Neurosci 2002; 3:228–236. Traystman RJ. Animal models of focal and global cerebral ischemia. ILAR J 2003; 44:85–95. Cenci MA, Whishaw IQ, Schallert T. Animal models of neurological deficits: how relevant is the rat? Nat Rev Neurosci 2002; 3:574–579. Duverger D, MacKenzie ET. The quantification of cerebral infarction following focal ischemia in the rat: influence of strain, arterial pressure, blood glucose concentration, and age. J Cereb Blood Flow Metab 1988; 8:449–461. Jones TA, Chu CJ, Grande LA, Gregory AD. Motor skills training enhances lesion-induced structural plasticity in the motor cortex of adult rats. J Neurosci 1999; 19:10153–10163. Fukuda S, del Zoppo GJ. Models of Focal Cerebral Ischemia in the Non-human Primate. ILAR, 2003. Nudo RJ, Larson D, Plautz EJ, Friel KM, Barbay S, Frost SB. A squirrel monkey model of poststroke motor recovery. ILAR J 2003; 44:161–174. Duncan PW, Lai SM. Stroke recovery. Topics in stroke rehabilitation 1997; 4:51–58. Jones TA, Bury SD, Adkins-Muir DL, Luke LM, Allred RP, Sakata JT. Importance of behavioral manipulations and measures in rat models of brain damage and brain repair. ILAR J 2003; 44:144–152. Friel KM, Nudo RJ. Recovery of motor function after focal cortical injury in primates: compensatory movement patterns used during rehabilitative training. Somatosens Mot Res 1998; 15:173–189. Whishaw IQ. Loss of the innate cortical engram for action patterns used in skilled reaching and the development of behavioral compensation following motor cortex lesions in the rat. Neuropharmacology 2000; 39:788–805. Nudo RJ, Plautz EJ, Frost SB. Role of adaptive plasticity in recovery of function after damage to motor cortex. Muscle Nerve 2001; 24:1000–1019. Michaelsen SM, Luta A, Roby-Brami A, Levin MF. Effect of trunk restraint on the recovery of reaching movements in hemiparetic patients. Stroke 2001; 32:1875–1883. Whishaw IQ, Pellis SM, Gorny BP, Pellis VC. The impairments in reaching and the movements of compensation in rats with motor cortex lesions: an endpoint, videorecording, and movement notation analysis. Behav Brain Res 1991; 42:77–91. Nudo RJ, Milliken GW, Jenkins WM, Merzenich MM. Use-dependent alterations of movement representations in primary motor cortex of adult squirrel monkeys. J Neurosci 1996; 16:785–807. Jenkins WM, Merzenich MM, Ochs MT, Allard T, Guic-Robles E. Functional reorganization of primary somatosensory cortex in adult owl monkeys after behaviorally controlled tactile stimulation. J Neurophysiol 1990; 63:82–104. Garraghty PE, Kaas JH, Florence SL. Plasticity of sensory and motor maps in adult and developing mammals. In: Shinkman PG, ed. Advances in Neural and Behavioral Development. Vol 4. Norwood, New Jersey: Ablex Publishing Corporation, 1995.
3. 4. 5.
6.
7. 8. 9. 10.
11.
12.
13. 14.
15.
16.
17.
18.
Neural Bases for Rehabilitation After Stroke
247
19. Xerri C, Merzenich MM, Jenkins W, Santucci S. Representational plasticity in cortical area 3b paralleling tactual–motor skill acquisition in adult monkeys. Cereb Cortex 1999; 9:264–276. 20. Xerri C, Merzenich MM, Peterson BE, Jenkins W. Plasticity of primary somatosensory cortex paralleling sensorimotor skill recovery from stroke in adult monkeys. J Neurophysiol 1998; 79:2119–2148. 21. Nudo RJ, Jenkins WM, Merzenich MM, Prejean T, Gedela R. Neurophysiological correlates of hand preference in primary motor cortex of squirrel monkeys. J Neurosci 1992; 12:2918–2947. 22. Classen J, Liepert J, Wise SP, Hallett M, Cohen LG. Rapid plasticity of human cortical movement representation induced by practice. J Neurophysiol 1998; 79:1117–1123. 23. Pascual-Leone A, Nguyet D, Cohen LG, Brasil-Neto JP, Cammarota A, Hallett M. Modulation of muscle responses evoked by transcranial magnetic stimulation during the acquisition of new fine motor skills. J Neurophysiol 1995; 74:1037–1045. 24. Frackowiak RS, Weiller C, Chollet F. The functional anatomy of recovery from brain injury. Ciba Found Symp 1991; 163:235–244. 25. Huntley GW, Jones EG. Relationship of intrinsic connections to forelimb movement representations in monkey motor cortex: a correlative anatomical and physiological study. J Neurophysiol 1991; 66:390–413. 26. Donoghue JP, Leibovic S, Sanes JN. Organization of the forelimb area in squirrel monkey motor cortex: representation of digit, wrist, and elbow muscles. Exp Brain Res 1992; 89:1–19. 27. Nudo RJ, Masterton RB. Descending pathways to the spinal cord, III: sites of origin of the corticospinal tract. J Comp Neurol 1990; 296:559–583. 28. Nudo RJ, Milliken GW. Reorganization of movement representations in primary motor cortex following focal ischemic infarcts in adult squirrel monkeys. J Neurophysiol 1996; 75:2144–2149. 29. Liepert J, Tegenthoff M, Malin JP. Changes of cortical motor area size during immobilization. Electroencephalogr Clin Neurophysiol 1995; 97:382–386. 30. Neumann-Haefelin T, Bosse F, Redecker C, Muller HW, Witte OW. Upregulation of GABAA-receptor alpha 1- and alpha2-subunit mRNAs following ischemic cortical lesions in rats. Brain Res 1999; 816:234–237. 31. Neumann-Haefelin T, Staiger JF, Redecker C, Zilles K, Fritschy JM, Mohler H, Witte OW. Immunohistochemical evidence for dysregulation of the GABAergic system ipsilateral to photochemically induced cortical infarcts in rats. Neuroscience 1998; 87:871–879. 32. Schiene K, Bruehl C, Zilles K, Qu M, Hagemann G, Kraemer M, Witte OW. Neuronal hyperexcitability and reduction of GABAA-receptor expression in the surround of cerebral photothrombosis. J Cereb Blood Row Metab 1996; 16:906–914. 33. Redecker C, Wang W, Fritschy JM, Witte OW. Widespread and long-lasting alterations in GABA(A)-receptor subtypes after focal cortical infarcts in rats: mediation by NMDA-dependent processes. J Cereb Blood Flow Metab 2002; 22:1463–1475.
248
Nudo et al.
34. Toriyama H, Wang L, Saegusa H, Zong S, Osanai M, Murakoshi T, Noda T, Ohno K, Tanabe T. Role of Ca(v) 2.3 (alphalE ) Ca2þ channel in ischemic neuronal injury. Neuroreport 2002; 13:261–265. 35. Witte OW, Stoll G. Delayed and remote effects of focal cortical infarctions: secondary damage and reactive cortical plasticity. In: Freund HJ, Sabel BA, Witte OW, eds. Brain Plasticity, Advances in Neurology. Vol. 73. 1997:207–227. 36. Fujii N, Mushiake H, Tanji J. An oculomotor representation area within the ventral premotor cortex. Proc Natl Acad Sci USA 1998; 95:12034–12037. 37. Mushiake H, Inase M, Tanji J. Neuronal activity in the primate premotor, supplementary and precentral motor cortex during visually guided and internally determined sequential movements. J Neurophysiol 1991; 66:705–718. 38. Frost SB, Barbay S, Friel KM, Plautz EJ, Nudo RJ. Reorganization of remote cortical regions after ischemic brain injury: a potential substrate for stroke recovery. J Neurophysiol 2003; 89:3205–3214. 39. Furlan M, Marchal G, Viader F, Derlon JM, Baron JC. Spontaneous neurological recovery after stroke and the fate of the ischemic penumbra. Ann Neurol 1996; 40:216–226. 40. Rapisarda G, Bastings E, de-Noordhout AM, Pennisi G, Delwaide PJ. Can motor recovery in stroke patients be predicted by early transcranial magnetic stimulation?. Stroke 1996; 27:2191–2196. 41. Cramer SC, Nelles G, Benson RR, Kaplan JD, Parker RA, Kwong KK, Kennedy DN, Finklestein SP, Rosen BR. A functional MRI study of subjects recovered from hemiparetic stroke. Stroke 1997; 28:2518–2527. 42. Nelles G, Spiekramann G, Jueptner M, Leonhardt G, Muller S, Gerhard H, Diener HC. Evolution of functional reorganization in hemiplegic stroke: a serial positron emission tomographic activation study. Ann Neurol 1999; 46:901–909. 43. Cicinelli P, Traversa R, Rossini PM. Post-stroke reorganization of brain motor output to the hand: a 2–4 month follow-up with focal magnetic transcranial stimulation. Electroencephalogr Clin Neurophysiol 1997; 105:438–450. 44. Traversa R, Cicinelli P, Bassi A, Rossini PM, Bernardi G. Mapping of motor cortical reorganization after stroke. A brain stimulation study with focal magnetic pulses. Stroke 1997; 28:110–117. 45. von Giesen HJ, Roick H, Benecke R. Inhibitory actions of motor cortex following unilateral brain lesions as studied by magnetic brain stimulation. Exp Brain Res 1994; 99:84–96. 46. Liepert J, Bauder H, Wolfgang HR, Miltner WH, Taub E, Weiller C. Treatment-induced cortical reorganization after stroke in humans. Stroke 2000; 31:1210–1216. 47. Seitz RJ, Huang Y, Knorr U, Tellmann L, Herzog H, Freund HJ. Large-scale plasticity of the human motor cortex. Neuroreport 1995; 6:742–744. 48. Classen J, Schnitzler A, Binkofski F, Werhahn KJ, Kim YS, Kessler KR, Benecke R. The motor syndrome associated with exaggerated inhibition within the primary motor cortex of patients with hemiparetic stroke. Brain 1997; 120:605–619. 49. Kamada K, Saguer M, Moller M, Wicklow K, Katenhauser M, Kober H, Vieth J. Functional and metabolic analysis of cerebral ischemia using
Neural Bases for Rehabilitation After Stroke
50.
51. 52.
53. 54. 55. 56.
57.
58.
59.
60.
61.
62.
63. 64. 65.
66.
249
magnetoencephalography and proton magnetic resonance spectroscopy. Ann Neural 1997; 42:554–563. Aizawa H, Inase M, Mushiake H, Shima K, Tanji J. Reorganization of activity in the supplementary motor area associated with motor learning and functional recovery. Exp Brain Res 1991; 84:668–671. Cramer SC, Bastings EP. Mapping clinically relevant plasticity after stroke. Neuropharmacology 2000; 39:842–851. Seitz RJ, Hoflich P, Binkofski F, Tellmann L, Herzog H, Freund HJ. Role of the premotor cortex in recovery from middle cerebral artery infarction. Arch Neurol 1998; 55:1081–1088. Ward NS, Brown MM, Thompson AJ, Frackowiak RS. Neural correlates of outcome after stroke: a cross-sectional fMRI study. Brain 2003; 126:1430–1448. Ogden R, Franz SI. On cerebral motor control: the recovery from experimentally produced hemiplegia. Psychobiology 1917; 1:33–50. Johansson BB. Environmental influence on outcome after experiment brain infarction. Acta Neurochir Suppl 1996; 66:63–67. Nudo RJ, Wise BM, SiFuentes F, Milliken GW. Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science 1996; 272:1791–1794. Friel KM, Heddings AA, Nudo RJ. Effects of postlesion experience on behavioral recovery and neurophysiologic reorganization after cortical injury in primates. Neurorehabil Neural Repair 2000; 14:187–198. Kleim JA, Barbay S, Cooper NR, Hogg TM, Reidel CN, Remple MS, Nudo RJ. Motor learning-dependent synaptogenesis is localized to functionally reorganized motor cortex. Neurobiol Learn Mem 2002; 77:63–77. Plautz EJ, Milliken GW, Nudo RJ. Effects of repetitive motor training on movement representations in adult squirrel monkeys: role of use vs. learning. Neurobiol Learn Mem 2000; 74:27–55. Barbay S, Plautz EJ, Friel KM, Frost SB, Stowe AM, Dancause NHW, Nudo RJ. Delayed rehabilitative training following a small ischemic infarct in primary motor cortex (Ml). Society for Neuroscience Abstracts 2001; 27: 931–934. Castro-Alamancos MA, Borrel J. Functional recovery of forelimb response capacity after forelimb primary motor cortex damage in the rat is due to the reorganization of adjacent areas of cortex. Neuroscience 1995; 68:793–805. Liepert J, Miltner WH, Bauder H, Sommer M, Dettmers C, Taub E, Weiller C. Motor cortex plasticity during constraint-induced movement therapy in stroke patients. Neurosci Lett 1998; 250:5–8. Liepert J, Storch P, Fritsch A, Weiller C. Motor cortex disinhibition in acute stroke. Clin Neurophysiol 2000; 111:671–676. Weiller C, Rijntjes M. Learning, plasticity, and recovery in the central nervous system. Exp Brain Res 1999; 128:134–138. Jang SH, Kim YH, Cho SH, Lee JH, Park JW, Kwon YH. Cortical reorganization induced by task-oriented training in chronic hemiplegic stroke patients. Neuroreport 2003; 14:137–141. Grossman AW, Churchill JD, Bates KE, Kleim JA, Greenough WT. A brain adaptation view of plasticity: is synaptic plasticity an overly limited concept? Prog Brain Res 2002; 138:91–108.
250
Nudo et al.
67. Jones TA, Schallert T. Use-dependent growth of pyramidal neurons after neocortical damage. J Neurosci 1994; 14:2140–2152. 68. Keyvani K, Schallert T. Plasticity-associated molecular and structural events in the injured brain. J Neuropathol Exp Neurol 2002; 61:831–840. 69. Kolb B. Overview of cortical plasticity and recovery from brain injury. Phys Med Rehabil Clin N Am 2003; 14:S7–S25. 70. Carmichael ST. Plasticity of cortical projections after stroke. Neuroscientist 2003; 9:64–75. 71. Voorhies AC, Jones TA. The behavioral and dendritic growth effects of focal sensorimotor cortical damage depend on the method of lesion induction. Behav Brain Res 2002; 133:237–246. 72. Gregersen R, Christensen T, Lehrmann E, Diemer NH, Finsen B. Focal cerebral ischemia induces increased myelin basic protein and growth-associated protein-43 gene transcription in peri-infarct areas in the rat brain. Exp Brain Res 2001; 138:384–392. 73. Bitsch A, Horn C, Kemmling Y, Seipelt M, Hellenbrand U, Stiefel M, Ciesielczyk B, Cepek L, Bahn E, Ratzka P, Prange H, Otto M. Serum tau protein level as a marker of axonal damage in acute ischemic stroke. Eur Neurol 2002; 47:45–51. 74. Carmichael ST, Chesselet MF. Synchronous neuronal activity is a signal for axonal sprouting after cortical lesions in the adult. J Neurosci 2002; 22: 6062–6070. 75. Carmichael ST, Wei L, Rovainen CM, Woolsey TA. New patterns of intracortical projections after focal cortical stroke. Neurobiol Dis 2001; 8:910–922. 76. Kawamata T, Speliotes EK, Finklestein SP. The role of polypeptide growth factors in recovery from stroke. Adv Neurol 1997; 73:377–382. 77. Nakane M, Tamura A, Nagaoka T, Hirakawa K. MR detection of secondary changes remote from ischemia: preliminary observations after occlusion of the middle cerebral artery in rats. AJNR Am J Neuroradiol 1997; 18:945–950. 78. Dancause N, Barbay HS, Frost SB, Plautz EJ, Friel KM, Stowe AM, Zoubina EV, Nudo RJ. Redistribution of premotor cortical connections after an ischemic lesion in primary motor cortex. Soc Neurosci Abst 2002; 28:262–269. 79. Dum RP, Strick PL. Motor areas in the frontal lobe of the primate. Physiol Behav 2002; 77:677–682. 80. Stein D. Brain injury and theories of recovery. In: Goldstein L, ed. Restorative Neurology: Advances in Pharmacotherapy for Recovery After Stroke. Armonk, NY: Futura Publishing Co., Inc, 1998:1–34. 81. Stroemer RP, Kent TA, Hulsebosch CE. Neocortical neural sprouting, synaptogenesis, and behavioral recovery after neocortical infarction in rats. Stroke 1995; 26:2135–2144. 82. Jones TA, Kleim JA, Greenough WT. Synaptogenesis and dendritic growth in the cortex opposite unilateral sensorimotor cortex damage in adult rats: a quantitative electron microscopic examination. Brain Res 1996; 733:142–148. 83. Isaacs KR, Anderson BJ, Alcantara AA, Black JE, Greenough WT. Exercise and the brain: angiogenesis in the adult rat cerebellum after vigorous physical activity and motor skill learning. J Cereb Blood Flow Metab 1992; 12: 110–119.
Neural Bases for Rehabilitation After Stroke
251
84. Kleim JA, Swain RA, Armstrong KA, Napper RM, Jones TA, Greenough WT. Selective synaptic plasticity within the cerebellar cortex following complex motor skill learning. Neurobiol Learn Mem 1998; 69:274–289. 85. Kleim JA, Lussnig E, Schwarz ER, Comery TA, Greenough WT. Synaptogenesis and Fos expression in the motor cortex of the adult rat after motor skill learning. J Neurosci 1996; 16:4529–4535. 86. Biernaskie J, Corbett D. Enriched rehabilitative training promotes improved forelimb motor function and enhanced dendritic growth after focal ischemic injury. J Neurosci 2001; 21:5272–5280. 87. Humm JL, Kozlowski DA, James DC, Gotts JE, Schallert T. Use-dependent exacerbation of brain damage occurs during an early post-lesion vulnerable period. Brain Res 1998; 783:286–292. 88. Kozlowski DA, James DC, Schallert T. Use-dependent exaggeration of neuronal injury after unilateral sensorimotor cortex lesions. J Neurosci 1996; 16:4776–4786. 89. Humm JL, Kozlowski DA, Bland ST, James DC, Schallert T. Use-dependent exaggeration of brain injury: is glutamate involved? Exp Neurol 1999; 157:349–358. 90. Goldstein L. Restorative neurology. In: Rengachary S, ed. Neurosurgery. New York: McGraw-Hill, 1996:459–470. 91. Lee JM, Zipfel GJ, Choi DW. The changing landscape of ischaemic brain injury mechanisms. Nature 1999; 399:A7–A14. 92. Fisher M, Finklestein S. Pharmacological approaches to stroke recovery. Cerebrovasc Dis 1999; 9(suppl 5):29–32. 93. Staines WR, Mcllroy WE, Graham SJ, Black SE. Bilateral movement enhances ipsilesional cortical activity in acute stroke: a pilot functional MRI study. Neurology 2001; 56:401–404. 94. Goldstein LB. Influence of common drugs and related factors on stroke outcome. Curr Opin Neurol 1997; 10:52–57. 95. Goldstein LB. Potential effects of common drugs on stroke recovery. Arch Neurol 1998; 55:454–456. 96. Feeney DM. Mechanisms of noradrenergic modulation of physical therapy: effects on functional recovery after cortical injury. In: Goldstein LB, ed. Restorative Neurology: Advances in Pharmacotherapy for Recovery after Stroke. New York: Futura, 1998:35–78. 97. Hasselmo ME, Linster C. Neuromodulation and memory function. In: Kats PS, ed. Beyond Neurotransmission: Neuromodulation and its Importance for Information Processing. New York: Oxford University Press, 1999:318–348. 98. Stroemer RP, Kent TA, Hulsebosch CE. Enhanced neocortical neural sprouting, synaptogenesis, and behavioral recovery with d-amphetamine therapy after neocortical infarction in rats. Stroke 1998; 29:2381–2393, discussion 2393–2385. 99. Feeney DM, Gonzalez A, Law WA. Amphetamine, haloperidol, and experience interact to affect rate of recovery after motor cortex injury. Science 1982; 217:855–857. 100. Goldstein LB. Pharmacology of recovery after stroke. Stroke 1990; 21: III139–III142.
252
Nudo et al.
101. Clark AN, Mankikar GD. d-Amphetamine in elderly patients refractory to rehabilitation procedures. J Am Geriatr Soc 1979; 27:174–177. 102. Crisostomo EA, Duncan PW, Propst M, Dawson DV, Davis JN. Evidence that amphetamine with physical therapy promotes recovery of motor function in stroke patients. Ann Neurol 1988; 23:94–97. 103. Walker-Batson D, Unwin H, Curtis S. Use of amphetamine in the treatment of aphasia. Restor Neurol Neurosci 1992; 4:47–50. 104. Walker-Batson D, Smith P, Curtis S, Unwin H, Greenlee R. Amphetamine paired with physical therapy accelerates motor recovery after stroke. Further evidence. Stroke 1995; 26:2254–2259. 105. Hornstein A, Lennihan L, Seliger G, Lichtman S, Schroeder K. Amphetamine in recovery from brain injury. Brain Inj 1996; 10:145–148. 106. Finklestein S. The potential use of neuro trophic growth factors in the treatment of cerebral ischemia. In: Wieloch T, ed. Cellular and Molecular Mechanisms of Ischemic Brain Damage. Vol. 71. Philadelphia: Lippincott-Raven, 1996:413–418. 107. Ren J, Kaplan PL, Charette MF, Speller H, Finklestein SP. Time window of intracistemal osteogenic protein-1 in enhancing functional recovery after stroke. Neuropharmacology 2000; 39:860–865. 108. Kawamata T, Dietrich WD, Schallert T, Gotts JE, Cocke RR, Benowitz LI, Finklestein SP. Intracistemal basic fibroblast growth factor enhances functional recovery and up-regulates the expression of a molecular marker of neuronal sprouting following focal cerebral infarction. Proc Natl Acad Sci USA 1997; 94:8179–8184. 109. Kawamata T, Ren J, Chan TC, Charette M, Finklestein SP. Intracisternal osteogenic protein-1 enhances functional recovery following focal stroke. Neuroreport 1998; 9:1441–1445. 110. Kolb B, Cote S, Ribeiro-da-Silva A, Cuello AC. Nerve growth factor treatment prevents dendritic atrophy and promotes recovery of function after cortical injury. Neuroscience 1997; 76:1139–1151. 111. Taub E, Perrella P, Barro G. Behavioral development after forelimb deafferentation on day of birth in monkeys with and without blinding. Science 1973; 181:959–960. 112. Taub E, Goldberg IA, Taub P. Deafferentation in monkeys: pointing at a target without visual feedback. Exp Neurol 1975; 46:178–186. 113. Taub E. Movement in Nonhuman Primates Deprived of Somatosensory Feedback. Exercises and Sports Science Reviews. Santa Barbara: Journal Publishing Affiliates, 1977:335–374. 114. Taub E. Somatosensory deafferentation research with monkeys. Implications for rehabilitation medicine. In: Ince LP, ed. Behavioral Psychology in Rehabilitation Medicine: Clinical Implication. New York: Williams and Wilkins, 1980:371–401. 115. Taub E, Berman AJ. Movement and learning in the absence of sensory feedback. Freedman SJ, ed. The Neuropsychology of Spatially Oriented Behavior. Homewood, IL: Dorsey Press, 1968:173–192. 116. Knapp HD, Taub E, Berman AJ. Movements in monkeys with deafferented limbs. ExpNeurol 1963; 7:305–315.
Neural Bases for Rehabilitation After Stroke 117.
118. 119. 120. 121.
122.
123.
124.
125.
126. 127.
128. 129.
130.
131.
132.
133.
253
Taub E, Uswatte G, Pidikiti R. Constraint-induced movement therapy: a new family of techniques with broad application to physical rehabilitation—a clinical review. J Rehabil Res Dev 1999; 36:237–251. Taub E, Wolf SL. Constraint-induced (CI) movement techniques to facilitate upper extremity use in stroke patients. Top Stroke Rehabil 1997; 3:38–61. Andrews K, Stewart J. Stroke recovery: he can but does he? Rheumatol Rehabil 1979; 18:43–48. Catania AC. Learning. Upper Saddle River, NJ: Prentice Hall, 1998. Taub E, Miller NE, Novack TA, Cook EW III, Fleming WC, Nepomuceno CS, Connell JS, Crago JE. Technique to improve chronic motor deficit after stroke. Arch Phys Med Rehabil 1993; 74:347–354. Taub E, Crago JE, Burgio LD, Groomes TE, Cook EW III, DeLuca SC, Miller NE. An operant approach to rehabilitation medicine: overcoming learned nonuse by shaping. J Exp Anal Behav 1994; 61:281–293. Taub E, Crago JE, Uswatte G. Constraint-Induced (CI) Therapy: a new approach to treatment in physical rehabilitation. Rehabil Psychol 1998; 43:152–170. Morris D, Crago JE, DeLuca SC, Pidikiti R, Taub E. Constraint-Induced (CI) Movement Therapy for motor recovery after stroke. Neurorehabilitation 1997; 9:29–43. Kunkel A, Kopp B, Muller G, Villringer K, Villringer A, Taub E, Flor H. Constraint-induced movement therapy for motor recovery in chronic stroke patients. Arch Phys Med Rehabil 1999; 80:624–628. Morris D, Taub E. Constraint-induced therapy approach to restoring function after neurological injury. Top Stroke Rehabil 2001; 8:16–30. Wolf SL, Lecraw DE, Barton LA, Jann BB. Forced use of hemiplegic upper extremities to reverse the effect of learned nonuse among chronic stroke and head-injured patients. Exp Neurol 1989; 104:125–132. Cohen. Statistical power analysis for the behavioral sciences. New York, NY: Academic Press, 1977. Levy CE, Nichols DS, Schmalbrock PM, Keller P, Chakeres DW. Functional MRI evidence of cortical reorganization in upper-limb stroke hemiplegia treated with constraint-induced movement therapy. Am J Phys Med Rehabil 2001; 80:4–12. Miltner WH, Bauder H, Sommer M, Dettmers C, Taub E. Effects of constraint-induced movement therapy on patients with chronic motor deficits after stroke: a replication. Stroke 1999; 30:586–592. Liepert J, Uhde I, Graf S, Leidner O, Weiller C. Motor cortex plasticity during forced-use therapy in stroke patients: a preliminary study. J Neurol 2001; 248:315–321. Kopp B, Kunkel A, Muhlnickel W, Villringer K, Taub E, Flor H. Plasticity in the motor system related to therapy-induced improvement of movement after stroke. Neuroreport 1999; 10:807–810. McKay DR, Ridding MC, Thompson PD, Miles TS. Induction of persistent changes in the organisation of the human motor cortex. Exp Brain Res 2002; 143:342–349.
254
Nudo et al.
134. Cicineili P, Traversa R, Rossini PM. Post-stroke reorganization of brain motor output to the hand: a 2–4 month follow-up with focal magnetic transcranial stimulation. Electroencephalogr Clin Neurophysiol 1997; 105: 438–450. 135. Muellbacher W, Richards C, Ziemann U, Wittenberg G, Weltz D, Boroojerdi B, Cohen L, Hallett M. Improving hand function in chronic stroke. Arch Neurol 2002; 59:1278–1282. 136. Plautz EJ, Barbay S, Frost SB, Friel KM, Dancause N, Zoubina EV, Stowe A, Quaney BM, Nudo RJ. Post-infarct cortical plasticity and behavioral recovery using concurrent cortical stimulation and rehabilitative training: a feasibility study in primates. Neurol Res 2003 Dec; 25(8):801:810. 137. Adkins DL, Jones TA. Cortical electrical stimulation combined with rehabilitative training: enhanced functional recovery and dendritic plasticity following focal cortical ischemia in rats. Neurol Res 2003, 25(8):780–788.
13 Plasticity of Language Networks Cynthia K. Thompson Department of Communication Sciences and Disorders, and Neurology, Cognitive Neurology and Alzheimer’s Disease Center, Northwestern University, Evanston, Illinois, U.S.A.
There is substantial evidence that individuals with language impairments resulting from brain damage, such as stroke-induced aphasia, show recovery of language function despite sustained damage to left perisylvian language areas (1). A long-held view is that language recovery following left hemisphere injury depends on right hemisphere mechanisms. Lesion studies support this claim, i.e., language recovers to some degree in individuals with large left hemisphere lesions (2,3), and patients with left hemisphere lesions show disruption of residual language after transient anesthesia of the right hemisphere with sodium amobarbital (4). Further, patients, who show recovery following left hemisphere lesions, show deterioration of language after a second lesion in the right hemisphere (5). Studies of language development and restitution in left-hemispherectomized individuals also provide compelling evidence that the right hemisphere has the capacity to subserve language (6,7). However, there also is clinical evidence that recovery of function involves expanded cortical language regions in the left hemisphere (8). For example, individuals with aphasia often show further language decline following a second stroke in the left hemisphere (2), suggesting that cortical regions in the left hemisphere are recruited into the language network when brain damage occurs. Findings from lesion studies, thus, indicate that the primary neural mechanisms instantiated to sustain language recovery include areas in the right hemisphere, and=or spared brain tissue in the damaged (usually left) cerebral hemisphere. More recently, neuroimaging studies 255
256
Thompson
examining the mechanisms underlying recovery of language in aphasia have been reported with evidence largely supporting these patterns, but providing additional detail regarding the tissue recruited in recovery and highlighting variability in recovery patterns. In this chapter, neuroimaging studies of brain-damaged individuals are reviewed and factors related to neuroplastic processes in aphasia are discussed. 1. NEUROIMAGING STUDIES 1.1. Studies in Lesioned vs. Non-lesioned Patients Neuroimaging has been used to study the neural networks engaged as humans undertake complex, cognitive tasks, and thus can be used to study neuroplasticity non-invasively in patients with brain damage. These methods are particularly useful for comparing activation in lesioned vs. nonlesioned individuals and for following the mechanisms of recovery over time. Studies using single photon emission computed tomography (SPECT) or positron emission tomography (PET) show increased glucose metabolism or regional cerebral blood flow (rCBF) in the right hemisphere as well as in undamaged portions of the language network in the left hemisphere (9–13). Weiller et al. (13), for example, examined rCBF in six normal subjects and six individuals who had putatively recovered from Wernicke’s aphasia. Subjects repeated pseudowords and performed a verb generation task. Results showed that, compared to the normal participants, the aphasic patients evidenced greater rCBF in the right hemisphere homologues of both Wernicke’s and Broca’s areas, i.e., in the superior temporal gyrus and the inferior premotor and lateral prefrontal cortices (see Fig. 1). Buckner et al. (14) showed evidence of right hemisphere recruitment in a patient (LF1) with a left frontal lobe lesion and mild non-fluent aphasia
Figure 1 Activation patterns found by Weiller et al. in a PET study examining verb generation (top) and pseudoword repetition (bottom) in normal participants and six patients after recovery from Wernicke’s aphasia. Lateral left and right hemispheres shown for both tasks for both participant groups. (From Ref. 13.)
Plasticity of Language Networks
257
resulting from stroke. At six months poststroke, neuropsychological testing showed impairment in verbal fluency across a number of tasks; however, the patient showed good performance on a word-stem completion task. This task, used in several PET and fMRI studies (15–17), required generation of a complete word when presented with letter strings comprising the first few letters of target words. For example, ‘‘str —’’ and ‘‘cou—’’ were presented and the subject was expected to produce the words ‘‘string’’ and ‘‘courage.’’ To determine what parts of the brain were recruited to support this task, a PET study was performed. Results showed that, while normal participants showed robust activation in the left frontal area—precisely the area lesioned in LF1—the patient showed activation in the right frontal area (see Fig. 2). Using the same task, Biasi et al. (18) studied eight patients with stroke-related lesions in the left inferior frontal cortex near or within Broca’s area and 14 age-matched control subjects. Prior to scanning, the patients underwent learning trials showing an improvement in the letter string task from 59% correct to 73% correct. Scanning results using fMRI showed stronger activation in the right frontal cortex (middle frontal gyrus) in the patients as compared to the normal participants. Similar findings have been reported by others. Using fMRI to examine lexical-semantic processing in patients who were at least five months poststroke, Cao et al. (19) found significantly greater right hemisphere activation in the patients as compared to the normal subjects. Gold and Kertesz (20), examining visual semantic processing of words in an aphasic individual with a large lesion in left hemisphere language areas, found right hemisphere activation in areas homologous to left brain language areas. We also found this pattern in a recent study examining verb processing using a lexical decision task in a patient with a left anterior stroke and concomitant Broca’s aphasia (21). While normal participants (n ¼ 10) recruited the left inferior frontal and posterior superior temporal gyri, our patient recruited the right
Figure 2 PET activation seen in normal participants (left image) and in LF1, a stroke patient with a left anterior lesion (middle image) during a speech production (word-stem completion) task (right image). (From Ref. 14.)
258
Thompson
hemisphere homologue of Broca’s area. Data such as these suggest that homologous right hemisphere regions assume the function of damaged left hemisphere regions. Several studies also have found activation of a left hemisphere network associated with language recovery (13,19,22–24). In the aforementioned Weiller et al. (13) report, the Wernicke’s patients with temporal lobe lesions recruited neural tissue in undamaged areas of the left hemisphere in addition to greater right hemisphere activation than normal, i.e., left frontal areas in and around Broca’s area, extending to the prefrontal cortex were activated more so in the aphasic patients than in the normal participants. Alsop et al. (22) reported a similar pattern in patients with temporal lobe lesions (one patient with an arteriovenous malformation and one with a glioma). However, these patients’ recruitment was constrained to left hemisphere perilesional areas, beyond that seen in normal subjects, for language processing. Similarly, Zahn et al. (24) studied two patients with transcortical sensory aphasia (TSA). One patient with a left middle cerebral artery infarction affecting both prefrontal and posterior parietal parts of the language network (RC) showed recovery of language comprehension associated with activation of left perilesional pre-frontal areas as well as posterior (Wernicke’s) area. Patients with lesions in Broca’s area also have shown perilesional activation. Miura et al. (25) undertook an fMRI study with a 45-year-old woman who developed Broca’s aphasia after infarction to the left frontal lobe. The patient was scanned three times while performing verbal tasks. On the first scan, two weeks following stroke, fMRI results showed no signal in the left frontal lobe. However, an increase in activation was noted in left frontal areas on scans administered four weeks later, and on the last scan undertaken seven months later when she had regained ability to speak. On the final scan the patient showed left frontal signal activity similar to that of normal subjects. 1.2.
An Unresolved Issue: Right vs. Left Hemisphere Contributions to Recovery of Language
The neural mechanisms underlying language recovery are incompletely understood, although the data thus far indicate that both the right hemisphere—largely regions homologous to left hemisphere language processing areas—and left hemisphere areas—largely perilesional regions—can be recruited to support language. It is important to keep in mind, however, that normal participants show both right hemisphere activation as well as recruitment of left hemisphere regions beyond those traditionally considered crucial for language processing while performing language tasks under fMRI or PET scanning, i.e., inferior temporal regions (26,27). Right brain and=or perilesional (or other left brain) activation seen in studies with
Plasticity of Language Networks
259
aphasic patients, therefore, may reflect, at least in part, the large-scale neural network that subserves language under normal conditions (28–30). Unfortunately, for the most part, neuroimaging studies of aphasia have not included scans correlating behavioral performance with activation patterns over time (i.e., scans prior to, during, and=or following recovery). Without repeat scans, the extent to which activation patterns noted in recovery states reflect recruitment of new neural tissue or premorbid language processing routines is unknown. Normal subjects also show variability in activation patterns, reflecting at least in part, processing routines that are unique to individual language learning experience. Further, brain-damaged patients are heterogeneous with regard to lesion site and extent (see below for further discussion), thus the degree of right hemisphere or perilesional tissue recruitment will likely be influenced by the extent of cell loss (lesion size) in the left hemisphere. Issues also have been raised with regard to the competence of the right hemisphere when it is recruited to perform language tasks. Some researchers have suggested that right brain take-over of function may reflect inefficient language processing (31). Belin et al. (32), for example, investigated rCBF patterns in seven aphasic patients who received melodic intonation therapy (MIT), a treatment program focused on teaching patients to produce words using songlike prosody (33,34). Following treatment, the patients underwent PET studies in which they were required to repeat words under two conditions: (a) when the words were presented with normal intonation, a ‘‘usual word’’ condition, and (b) when the word were ‘‘MIT-loaded.’’ Results showed that under the usual word condition, the focus of PET activation was in the right perisylvian region. However, under the ‘‘MIT-loaded’’ condition, activation shifted to the left anterior brain. This and other observations led Selnes (31) to conclude that ‘‘ . . . recruitment of right-hemisphere structures for language recovery is a last-resort type of strategy and one that yields a less than satisfactory overall degree of language recovery in most instances. The preferred strategy, and one that yields the greatest functional degree of recovery, is to integrate other left hemisphere structures with remaining functional language areas in the dominant left hemisphere’’ (p. 419). Zahn et al. (24) further suggest that the left hemisphere is a more likely candidate for subsuming recovery of language because regions located in close approximation to the language processing network have inherent ‘‘redundancy’’ of function, i.e., these areas have a latent capacity for language processing and=or they normally perform language related functions and are thus available for language recovery. Zahn et al. (24) suggest that such ‘‘redundancy recovery’’ is more likely than that which recruits areas of the brain normally involved in processing routines unrelated to language (‘‘vicarious functioning’’). Naesser and colleagues (35) have recently begun to investigate this hypothesis using repetitive transcortical magnetic stimulation (rTMS). Patients with left brain lesions and chronic aphasia are
260
Thompson
provided naming treatment while they receive rTMS in right hemisphere areas homologous to Broca’s area. Preliminary findings using this method indicate that naming improves under rTMS conditions, suggesting left hemisphere support for recovered naming. Grafman (36) also suggests that homologous right brain adaptation is most likely to occur when lesions completely destroy cortical regions that serve a particular function. Transfer of function is less likely to occur when damage is incomplete because homologous sites are inhibited under normal conditions by connections from contralateral regions. When damage is incomplete, inhibitory input is retained, thereby precluding transfer of function. While these theories about the role of the right hemisphere in recovery of function are interesting, there are data suggesting that they may not be completely correct. For example, Thompson et al. (37) and Heiss et al. (12) showed that patients with lesions constrained to the left frontal region show recruitment of Wernicke’s area on the right, which correlate with improved language function. 2.
FACTORS RELATED TO NEUROPLASTIC PROCESSES
There are several factors that may influence the mechanisms recruited to support language processing in aphasic individuals, including neurophysiological processes that are at work during spontaneous recovery and thereafter, such as neuronal regeneration and sprouting, changes in neurotransmitter release, and return to pre-stroke levels of blood flow (38–40). Factors related to the neural insult itself, such as the site and extent of lesion also impact recovery mechanisms as do subject variables, including age, education, motivation, and other related factors (see Ref. 41). 2.1.
Size and Size of Lesion, and Time Poststroke
While little research has focused on the influence of these variables, a few studies highlight the impact of site and size of lesion and time poststroke. Thulborn et al. (42) examined recovery in two patients with different left hemisphere lesion sites. One patient evinced a lesion in Broca’s area, and the other had a lesion in Wernicke’s area. Recovery of simple sentence (reading) comprehension was tested over time, with results showing a rightward shift in activation in both patients. However, the specific neural tissue recruited differed in the two patients. The Broca’s area lesioned patient progressed to strong right hemisphere activation of Broca’s area, while he retained the normal pattern of left hemisphere activation in Wernicke’s area (at six months poststroke). Conversely, the Wernicke’s patient showed (at nine months poststroke) strong activation in the right hemisphere homologue of Wernicke’s area and a notable, albeit weaker, increase in activation in Broca’s area in the left hemisphere. These data suggest that recruitment of
Plasticity of Language Networks
261
neural tissue for language processing following brain damage is influenced by lesion site. The size of the lesion also will likely impact the neural tissue engaged to support recovery of function. While neuroimaging studies with aphasic patients have not directly examined lesion size, there are animal data suggesting that it is important to consider. Kolb and Wishaw (43), for example, have shown that rats trained to use their forepaws to retrieve small pieces of food loose and then regain this ability with small lesions to the motor cortex. However, rats with larger lesions show less improvement, and those with lesions encompassing most of the sensorimotor cortex show little, if any, recovery. Time poststroke also is important, since activation patterns noted during spontaneous recovery may differ, often quite drastically, from those seen in later stages of recovery. Heiss et al. (12), for example, studied patients at two and eight weeks poststroke, and reported remarkable changes in activation patterns from one test period to the other. Thulborn et al. (42) also reported considerable change in activation patterns during early periods of recovery, i.e., they scanned patients at 76 hr poststroke and several (up to nine) months later and showed shifts in activation patterns over time. It also is important to emphasize that different patterns of neural activity result from different linguistic tasks. For example, a number of studies have shown different patterns of activation for normal participants when they perform word production-type tasks vs. semantic processing tasks. Similarly, aphasic patients will likely show different patterns of neural recruitment depending on the task that they are required to perform. Perani et al. (44) showed that task requirements as well as task performance influence neural recruitment. They studied five patients with aphasia resulting from stroke and normal control subjects using a letter-fluency and semantic-fluency task. In the letter task, subjects were required to covertly generate words beginning with a letter cue; in the semantic task they covertly generated related words. Results showed that for the letter task, patients with good performance showed activation in either left Broca’s area or in the right-side homologue of Broca’s area. Conversely, in the case of semantic fluency, extensive recruitment of bilateral frontal areas was seen in patients who showed poor performance. These findings suggested that performance in letter fluency depends on the integrity of the left inferior frontal cortex, with participation of the homologous right hemispheric region when the left frontal cortex is damaged. Semantic fluency, which engaged more extensive patterns of cerebral activation, reflected retrieval effort rather than retrieval success. In a study by Calvert et al. (45) task-specific differences in fMRI activation were found in a 28-year-old woman after partial recovery from a left hemisphere stroke. Performance of a verbal semantic decision task recruited a network of brain areas that excluded the inferior frontal gyrus (in either
262
Thompson
hemisphere). In contrast, a rhyming task—thought to place a greater demand on production processes and which activated the left inferior frontal gyrus strongly in normal controls—resulted in prominent activation in the right hemisphere homologue of Broca’s area. These findings indicate that a lack of recruitment of certain brain tissue does not mean that this tissue is not part of the poststroke language network. Rather, they suggest that this tissue was not crucial to the task performed. Neuroimaging studies of recovery from aphasia may, therefore, benefit from inclusion of several tasks, tapping language performance in several domains, in order to get a full picture of language recovery. Researchers also should consider including non-language, control tasks to confirm normal, reliable activation under these conditions, in contrast to the language tasks selected. 2.2.
Effects of Treatment on Neural Recruitment
One very important, and often overlooked, variable relevant to recovery concerns the poststroke linguistic experiences of the patient. It is now a well known fact that neural plasticity extends into adulthood and that there are strong effects of the environment on neural organization and reorganization of function. Studies have shown, for example, that motor learning and motorically enriched environments, tactile stimulation, and auditory stimulation strongly influence neural organization of the primary motor, somatosensory, and auditory cortex, respectively (46–50). For example, Nudo et al. (48) found plastic changes in the functional topography of the primary motor cortex in adult squirrel monkeys following motor learning tasks, and Jenkins et al. (47) reported enlargements of somatosensory areas associated with controlled tactile stimulation in adult owl monkeys. Studies also have shown that rehabilitative training after injury results in enhancement of representational plasticity (48,51). Nudo et al. (48) trained monkeys to retrieve pellets from small wells, an activity that requires skilled digital use. Following training, lesions in the motor cortex were induced, following which the monkeys were once again trained to perform the task. Comparison of intracortical micro-stimulation maps of the motor cortex derived before and after lesion revealed substantial rearrangement of representations. Areas of cortical digital representation were expanded, while wrist and forearm representations were contracted. These findings indicate that experience directly shapes physiological reorganization following brain damage. Thus, it is likely that treatment provided for aphasia influences the extent and manner of reorganizational processes. Although research examining the neural bases of treatment-induced recovery from aphasia is limited, the studies available indicate that treatment results can be mapped onto the brain. Further, there is some indication that different patterns of brain recruitment may be seen when different treatments are provided. One study examining the effects of short-term auditory
Plasticity of Language Networks
263
comprehension treatment on brain activation was that of Musso et al. (52). Four patients with Wernicke’s aphasia secondary to lesions in the left temporoparietal area underwent a series of 12 consecutive PET scans. During each scan, subjects were required to follow commands to either ‘‘point to’’ or ‘‘take’’ certain objects. Between each of the 12 scans (12 min intervals), the patients’ comprehension was tested using a shortened form of the Token Test (the sTT), part of the Aachen Aphasia Bedside Test (53) and treatment focused on language comprehension was provided. Group results showed activation in two brain sites correlating with improved sTT performance. These included the right hemisphere homologue of Wernicke’s area and the posterior part of the precuneus in the left hemisphere. We also undertook an fMRI study to determine neural patterns associated with recovery (37). The aim was to examine the neural correlates of treatment-induced improvements in sentence comprehension in patients with agrammatic, Broca’s aphasia. Six patients and eight normal individuals participated in the study. All patients showed reduced sentence length and grammaticality in their spontaneous speech; they produced more open-class than close-class words, and more nouns than verbs. Language comprehension was largely intact, although all had impaired ability to comprehend complex, non-canonical sentences. Three of the six patients and all normals were tested under fMRI conditions, following which three of the patients were provided with treatment. One of these patients (OJ) underwent repeat behavioral testing and fMRI scans at a five-month interval prior to treatment. The fMRI task required listening to sentences during active conditions, both simple subject-clefts and complex object-clefts, or single animate nouns during the control condition, and viewing pictures presented in the scanner. When a sentence (or word) matched a picture, they pressed a button. The treatment focused on production and comprehension of noncanonical sentences using a linguistic-specific treatment, i.e., treatment of underlying forms (54). This treatment uses a series of steps to train: (a) verb and verb argument comprehension and production, and (b) the movement operations required to form the surface structure of non-caononical sentences. Following treatment, fMRI scans were once again completed for the treated subjects, and all language measures were readministered to all subjects. Results showed significant changes in behavioral tests administered pre- and post-treatment for the treated subjects only. Concomitant changes also were noted in fMRI activation patterns for the treated subjects, while no change in activation patterns was seen on repeat pre-treatment fMRI scans for OJ (see Fig. 3). When we compared the post-treatment activation patterns of our aphasic patients to normal participants, we found some compelling similarities. As shown in Fig. 4, our normal participants showed bilateral activation of Wernicke’s area and left hemisphere Broca’s area recruitment for
264
Thompson
Figure 3 FMRI activation (SPM images) found on repeat fMRI studies with a non-fluent aphasic patient studied by Thompson et al. Shown are significant activations found during sentence vs. single word processing. Pre-treatment (top and middle images) activation was constrained to small superior parietal and tempopareital foci; on post-treatment, bilateral activation of the middle temporal cortex, and the right hemisphere homologue of Broca’s area as well as dorsolateral prefrontal and intraparietal areas were noted. (From Ref. 21.)
sentences as compared to words. In addition, they showed bilateral dosolateral prefrontal and intraparietal activation, likely reflecting visual attention as well as working memory demands. Comparing the patient’s recruitment to that of the normals, they recruited bilateral temporal lobe regions (BA 22) similar to that of the normal participants (see circled areas in Fig. 4). However, instead of left recruitment of Broca’s area, they showed activation of BA 44 in the right hemisphere (two of three patients). Finally, they showed activation of dosolateral prefrontal (BA 9) and intraparietal areas (BA 7) similar to that of the normals, but constrained to the right hemisphere.
Plasticity of Language Networks
265
Figure 4 SPM image (axials) showing areas of significant activation for eight normal participants during sentence as compared to single word conditions in an fMRI study by Thompson et al. Slices are 3 mm thick, going from Z ¼ 60 (top left slice) to Z ¼ 9 (bottom right slice). Areas activated by patients indicated by yellow circles, include bilateral middle temporal gyri (mTG), and right hemisphere Broca’s area (iFG), dorsolateral prefrontal cortex (DLPC), and the intraparietal sulcus (iPS). (From Ref. 21.)
These findings indicate that treatment gains can be mapped onto the brain and suggest that the type of treatment provided may influence neural recruitment patterns. Musso et al. (52) trained patients to comprehend simple auditory commands and found recruitment of right temporal areas, whereas Thompson et al. trained complex sentence production and comprehension which resulted in recruitment of portions of a larger functional language network. In summary, there are many factors that have potential to influence the neural mechanisms that support recovery of language in aphasia. The precise influence of these factors on recovery, however, is presently unclear and research is needed to clarify the role that they play. 2.3.
Interpretation of Results
One last comment relative to neuroimaging studies with aphasic patients pertains to what the particular activation patterns might mean. Intuitively,
266
Thompson
one might expect increases in activation to show up with recovery, e.g., a patient may show little or no activation of a particular area prior to recovery, but once the function returns, activation increases. Cardebat et al. (55) showed this pattern in a study of eight aphasic patients who received two PET scans with a 1-year interval between them. Patients performed a word generation task. Results showed that language performance improved from scan 1 to scan 2 and was associated with increased perisylvan activation. However, other observations suggest that increases in activation are associated with inefficient language processing. For example, abnormally large areas of activation have been noted in patients with dementia. Sonty et al. (56), for example, recently showed this pattern in patients with primary progressive aphasia (PPA) in two tasks, a homonyms task and a synonyms task as is shown in Fig. 5. Decreases in activation also have been associated with improved performance. In one of our patients, Thompson et al. (37), for example, who showed improved sentence production and comprehension
Figure 5 FMRI results from Sonty et al. Areas of significant activation for PPA > control in (a) phonology (HOM) and (b) semantics (SYN). Areas include intra-parietal sulcus (iPS), fusiform gyrus (Fus), precentral gyrus (prCG), and thalamus (not pictured). No areas of significant activation for control > PPA were present for either task. (From Ref. 56.)
Plasticity of Language Networks
267
performance when provided with treatment showed decreased right hemisphere perisylvian activation from pre- to post-treatment scans. 3.
CONCLUSION
This chapter has provided an overview of studies using neuroimaging techniques for studying the recovery of language processing in persons with aphasia. While neuroimaging research is difficult, and there remain many unresolved issues, particularly pertaining to studies with brain-damaged patients, it will help to inform theories of neuroplasticity and assist with our understanding of the mechanisms that support language recovery. REFERENCES 1. 2. 3. 4. 5. 6.
7. 8.
9. 10. 11. 12.
13.
14.
Holland AL, Fromm V, DeRuyter F. Treatment efficacy for aphasia. J Speech Hearing Res 1996; 39:S27–S36. Basso A, Gardelli M, Grassi MP, Mariotti M. The role of the right hemisphere in recovery from aphasia: two case studies. Cortex 1989; 25:555–556. Willme K, Poeck K. To what extent can aphasic syndrome be localized? Brain 1993; 116:1527–1540 Kinsbourne M. The minor hemisphere as a source of aphasic speech. Trans Am Neurolog Assoc 1971; 96:141–145. Growers WR. A Manual of Diseases of the Nervous System 1893. London: J & A Churchill. Dennis M, Whitaker HA. Language acquisition following hemidecortication: linguistic superiority of the left over the right hemisphere. Brain Lang 1976; 3:404–433. Wilson PJE. Cerebral hemispherectomy for infantile hemiplegia: a report of 50 cases. Brain 1970; 93:147–180. Kertesz A. What do we learn from recovery from aphasia. In: Waxman SG, ed. Functional Recovery in Neurological Disease. Vol. 47. New York: Raven Press, 1988:897–903. Cappa SF, Perani D, Grassi F. A PET follow-up study of recovery after stroke in acute aphasics. Brain Lang 1997; 56:55–67. Demeurisse G, Capon A. Language recovery in aphasic stroke patients: clinical, CT and CBF studies. Aphasiology 1987; 1:301–315. Heiss WD, Kessler J, Karbe H. Cerebral glucose metabolism as a predictor of recovery from aphasia in ischemic stroke. Arch Neurol 1993; 50:958–964. Heiss WD, Kessler J, Thiel A, Ghaemi M, Karbe H. Differential capacity of left and right hemispheric areas for compensation of poststroke aphasia. Ann Neurol 1999; 45:430–438. Weiller C, Isensee C, Rijntjes R, Huber W, Muller S, Bier D, Dutschka K, Wood RP, Noth J, Diener HC. Recovery from Wernicke’s aphasia: a positron emission tomographic study. Ann Neurol 1995; 37:723–732. Buckner RL, Corbetta M, Schatz J, Raichle ME, Petersen SE. Preserved speech abilities and compensation following prefrontal damage. Proc Nat Acad Sci USA 1996; 93:274–277.
268
Thompson
15. Buckner RL, Raichle ME, Peterson SE. Dissociation of human prefrontal cortical areas across different speech production tasks and gender groups. J Neurophysiol 1995; 74:2163–2173. 16. Ojemann JG, Bucker RL, Akbudaak E, Snyder AZ, Ollinger JM, McKinstry RC, Rosen BR, Petersen SE, Raichle ME, Conturo TE. Functional MRI studies of word stem completion: reliability across laboratories and comparison to blood flow imaging with PET. Human Brain Mapping 1998; 6:203–215. 17. Squire LR, Ojemann JG, Miezin FM, Petersen SE, Videen TO, Raichle ME. Activation of the hippocampus in normal humans: a functional anatomical study of memory. Proc Nat Acad Sci USA 1992; 89:1837–1341. 18. Biasi V, Young AC, Tansy AP, Petersen SE, Synder AZ, Corbetta M. Word retrieval learning modulates right frontal cortex in patients with left frontal damage. Neuron 2002; 36:159–170. 19. Cao Y, Vikingstad E, George K, Johnson A, Welch K. Cortical language activation in stroke patients recovering from aphasia with functional MRI. Stroke 1999; 30:2331–2340. 20. Gold BT, Kertesz A. Right hemisphere semantic processing of visual words in an aphasic patient: an fMRI study. Brain Lang 2000; 73:456–465. 21. Thompson CK, Bonakdarpour B, Blumenfeld H, Fix S, Parrish T, Gitelman DR, Mesulam M-M. Neural correlates of word class processing: an fMRI study. J Cognitive Neurosci Suppl 2004; 106. 22. Alsop DC, Detre JA, D’Esposito M, Howard RS, Maldjian JA, Grossman M, Listerud J, Flamm ES, Judy KD, Atlas SW. Functional activation during an auditory comprehension task in patients with temporal lobe lesions. Neuroimage 1996; 4:55–59. 23. Heiss WD, Karber H, Weber-Luxenburger G, Herholz K, Kessler J, Pietrzyk U, Pawlik G. Speech-induced cerebral metabolic activation reflects recovery from aphasia. J Neurolog Sci 1997; 145:213–217. 24. Zahn R, Huber W, Drews E, Specht K, Kemeny S, Reith W, Willmes K, Schwarz M. Recovery of semantic word processing in transcortical sensory aphasia: a functional magnetic resonance imaging study. Neurocase 2002; 8: 376–386. 25. Miura K, Nakamura Y, Miura F, Yamada I, Yoshikawa A, Mizobata T. Functional magnetic resonance imaging to word generation task in a patient with Broca’s aphasia. J Neurol 1999; 246:939–942. 26. Damasio H, Grawbowski TJ, Trannel D, Hichwa RD, Damasio AR. A neural basis for lexical retrieval. Nature 1996; 380:499–505. 27. Martin A, Wiggs CL, Ungerleider LG, Haxby JV. Neural correlates of category-specific knowledge. Nature 3 1996; 79:649–652. 28. Mesulam M-M. Large-scale neurocognitive networks and selectively distributed processing for attention, language, and memory. Ann Neurol 1990; 8: 597–613. 29. Mesulam M-M. Neurocognitive networks and selectively distributed processing. Revue Neurologique 1994; 150:564–569. 30. Mesulam M-M. From sensation to cognition. Brain 1998; 121:1013–1052. 31. Selnes OA. Recovery from aphasia: activating the ‘‘right’’ hemisphere. Ann Neurol 1999; 45:419–420.
Plasticity of Language Networks 32.
33. 34. 35.
36. 37.
38.
39.
40. 41.
42. 43.
44. 45.
46.
47.
48.
269
Belin P, Van Eeckhkout PM, Zilbovicius M, Remy P, Francois C, Guillaume S, Chain F, Rancurel G, Samson Y. Recovery from nonfluent aphasia after melodic intonation therapy: a PET study. Neurology 1996; 47:1504–1511. Spark R, Helm NA, Albert ML. Aphasia rehabilitation resulting from melodic intonation therapy. Cortex 1874; 10:303–316. Sparks R, Holland AL. Method: melodic intonation therapy for aphasia. J Speech Hearing Disord 1976; 41:287–297. Martin PI, Naesser MA, Theoret H, Tomos M, Nicholas M, Kurland J, Fregni F, Seekins H, Doron K, Pascual-Leone A. Transcranial magnetic stimulation as a complementary treatment for aphasia. Semin Speech Lang 2004, 181–191. Grafman J. Conceptualizing functional neuroplasticity. J Commun Disord 2000; 33:345–355. Thompson CK, Fix SC, Gitelman DR, Parrish TR, Mesulam M-M. FMRI studies of agrammatic sentence comprehension before and after treatment: comparison with normal subjects. Brain Lang 2000. Arboix A, Gracia-Eroles L, Comes E, Oliveres M, Balcells M, Pacheco, G, Tonga A. Predicting spontaneous early neurological recovery after acute ischemic stroke. Eur J Neurol 2003; 10:429–435. Ergenzinger ER, Pons TP. Growth of new connections and adult reorganizational plasticity in the somatosensory system. In: Levin HS, Grafman J, eds. Cerebral Reorganization of Function after Brain Damage. New York: Oxford University Press, 2000:68–84. Fawcett JW, Rosser AE, Dunnet SB. Brain Damage, Brain Repair. Oxford University Press. D’Esposito M, Zarahn E, Aguirre GK, Rypma B. The effect of normal aging on the coupling of neural activity in to the Bold hemodynamic response. NeuroImage 1999; 10:6–14. Thulborn KR, Carpenter PA, Just MA. Plasticity of language-related brain function during recovery from stroke. Stroke 1999; 30:749–754. Kolb B, Wishaw IQ. Reorganization of function after cortical lesions in rodents. In: Levin HS, Grafman J, eds. Cerebral Reorganization after Brain Damage. Oxford: Oxford University Press, 2000:109–129. Perani D, Cappa SF, Tettamanti M, Rosa M, Scifo P, Miozzo A, Basso A, Fazio F. A fMRI study of word retrieval in aphasia. Brain Lang 2003, 357–368. Calvert GA, Brammer MJ, Morris RG, Williams SCR, King N, Matthews PM. Using fMRI to study recovery from acquired dysphasia. Brain Lang 2000; 71:391–199. Greenough WT, Larson JR, Withers GS. Effects of unilateral and bilateral training in a reaching task on dendritic branching of neurons in the rat motor sensory forelimb cortex. Behav Neural Biol 1985; 44:301–314. Jenkins WM, Merzenich MM, Och MT, Allard T, Guic-Robles E. Functional reorganization of primary somatosensory cortex in adult owl monkeys after behaviorally controlled tactile stimulation. J Neurophysiol 1990; 63:82–104. Nudo, RJ, Milliken GW, Jenkins WM, Merzenich MM. Neural substrates for the effects of rehabilitate training on motor recovery after ischemic infarct. Science 1996; 171(5269):1791–1794.
270
Thompson
49. Recanzone GH, Schreiner CE, Merzenich MM. Plasticity in the frequency representation of the primary auditory cortex following discrimination training in adult owl monkeys. J Neurosci 1993; 13:87–103. 50. Van Praag H, Kempermann G, Gage FH. Running increases cell proliferation and neurogenesis in the adult mouse dentate gyrus. Nat Neurosci 1999; 2: 266–270. 51. Xerri C, Merzenich MM, Peterson BE, Jenkins W. Plasticity of primary somatosensory cortex paralleling sensorimotor skill recovery from stroke in adult monkeys. J Neurophysiol 1998; 79:2119–2148. 52. Musso M, Weiller C, Kiebel S, Muller S, Bulau P, Rijntjes M. Training-induced brain plasticity in aphasia. Brain 1999; 122:1781–1790. 53. Biniek R, Huber W, Glindemann R, Willmes K, Klumm, H. The Aachen Aphasia Bedside Test. Criteria for validity of pscyologic tests [German]. Nervenarzt 1992; 63:473–479. 54. Thompson CK. Treatment of underlying forms: A linguistic specific approach for sentence production deficits in agrammatic aphasia. In: Chapey R ed. Language Intervention Strategies in Adult Aphasia. 4th ed. Baltimore: Williams & Wilkins, 2001:605–628. 55. Cardebat D, Demonet JF, de Boisssezon X, Marie N, Marie R-M, Lambert J, Baron J-C, Puel M. Behavioral and neurofunctional changes over time in healthy and aphasic subjects. A PET language activation study. Stroke 2003; 34:2900–2907. 56. Sonty SP, Mesulam M-M, Thompson CK, Johnson NA, Weintraub S, Parrish TB, Gitelman DR. Anatomical and functional analysis of language changes in primary progressive aphasia. Ann Neurol 2003; 53:35–49.
14 Locomotor Training-Induced Plasticity Bruce H. Dobkin Department of Neurology, Geffen School of Medicine, Neurologic Rehabilitation and Research Program, University of California Los Angeles, Los Angeles, California, U.S.A.
1.
INTRODUCTION
Effective interventions to rehabilitate patients with neurological impairments and disabilities depend upon the ability of clinicians to drive adaptations within the spared nodes of neural networks. For the important skill of walking, studies in lower vertebrates and mammals, as well as in humans, are defining the interactions among interconnected neuronal assemblies, their neurotransmitters, ion channels, and genes, the effects of sensory feedback on their outputs, and molecular mechanisms of activity-dependent plasticity (1,2). This chapter examines several of these interactions as they relate to the recovery of walking after hemiplegic stroke and spinal cord injury (SCI). Intrinsic mechanisms of neuronal and network plasticity such as modifications in phosphorylation states, in spines and dendrites, and other aspects of synaptic plasticity at all levels of the neuroaxis are developed in other chapters.
2.
THE DISTRIBUTED LOCOMOTOR SYSTEM
For rehabilitation, the most prominent features of the locomotor system include its flexible and highly adaptable motor programs and inherent ability to learn new skilled movements, called procedural learning. Although theories of motor control are a work in progress, the system allows for rapid automatic, 271
272
Dobkin
corrective compensatory, predictive anticipatory, and willed movements for postural control and mobility. The brain has information about limb biomechanics, muscle elasticity, and other properties of the peripheral motor system and the environment, which are incorporated into planning and controlling movements. The interactive, distributed locomotor system permits quadripedal and bipedal walking and running across environmental challenges, and offers substitutive substrates that can be trained after injury. 2.1.
Cerebral Sensorimotor System
The cortex in humans especially contributes to the selection and initiation of walking-related motor control. The corticospinal tract contains the most direct descending excitatory influences on the sensory and motor pools of the cord for selective trunk and lower extremity movements. The macaque’s L-6-S-1 neurons, which activate hindlimb stepping movements in a similar fashion to the human leg, receive descending corticospinal tract projections from about 24,000 neurons in the primary motor cortex (Ml), 6000 from the supplementary motor area (SMA), 6200 from dorsal and ventral cingulate, 5000 from dorsal premotor, and 10 from ventral premotor cortices (3). Descending projections to the dorsal horns from the primary sensory cortex (S1) also descend in the corticospinal tract to modulate segmental spinal inputs. Each of these cortical regions interacts with visual, vestibular, aural, proprioceptive, cutaneous and other inputs to help plan, select, initiate, and maintain unilateral and bilateral skilled movements. Although the corticospinal tract is considered the most direct output for firing spinal motoneurons, it also plays a role in the modulation of seemingly simpler intrinsic spinal pathways, such as the H-reflex (4). Uninjured cortical descending tracts from crossed and uncrossed projections may play a greater role when patients relearn to walk after stroke and SCI. Functional neuroimaging studies reveal experience-dependent remodeling of these cortical regions for selective ankle and leg movements over the course of locomotor rehabilitation, much as is found in studies of representational plasticity for the upper extremity with training (5). The Ml, SMA, dorsal and ventral premotor, and rostral cingulate cortices project to the basal ganglia and spinal cord. The corticostriatal neurons are distinct from those within the corticospinal tract. The former respond especially to sensory inputs associated with directional movements and feed back to Ml parameters about the direction and force of movement. Premotor neurons participate in the internal guidance and sequence of movement. Thus, cues from therapists to evoke cognitive strategies that bring these regions into greater play during training may increase the level of descending drive on motoneurons of the spinal cord. Dorsiflexion of the ankle is a distinctly human aspect of gait, required during heel strike at the onset of the stance phase and during swing to help
Locomotor Training-Induced Plasticity
273
clear the foot. Functional magnetic resonance imaging reveals cortical, thalamic, and cerebellar activity during isolated passive or voluntary ankle dorsiflexion (6). Serial studies of rehabilitation interventions for walking may demonstrate adaptations within these nodes as locomotor control improves. Such changes in the volume, location, or lateralization of fMRI signals would be associated with mechanisms of learning, such as long-term potentiation (7). 2.2.
Brain Stem Nodes
The basal ganglia project to brainstem locomotor regions. The diencephalic locomotor center within the zona incerta and the mesopontine locomotor center with its cholinergic and glutaminergic cells activate the lumbar spinal central pattern generators (CPGs) when stimulated electrically or with certain drugs. These brain stem regions project to reticulospinal nuclei that are glutaminergic. The cerebellum monitors the outcome of every movement from proprioceptive inputs from the dorsal and ventral spinocerebellar tract. These inputs are also copied to the thalamus and the brain stem locomotor centers. The timing of coordinated movement sequences, as well as computations on the position, velocity, acceleration, and inherent viscous forces of the moving limbs, is partly orchestrated by the cerebellar nuclei and Purkinje cells. The great interest in these afferent signals suggests that locomotor training should aim to optimize kinematic and kinetic inputs that are associated with normal walking. 2.3.
Spinal Cord Systems
The spinal cord contains automatic and flexible modulators for walking. Central pattern generators for locomotion are neural circuits that produce oscillating patterns of behavior independently of sensory input or supraspinal commands. Elemental CPGs may control different muscles around each joint, but all of these are interlocked by their connections and responses to segmental afferents and descending command centers. Evidence for pattern generation comes from experiments in vertebrates, including nonhuman primates, in which the spinal cord is transected in the low thoracic region and deafferented from all dorsal root inputs below that level. Electrical stimulation and monamines placed on the isolated lumbar cord produce alternating electrical activity in the ventral roots of limb flexors and extensors. When cats and rats undergo spinal transection, their paraplegic hindlimbs lose the ability to step on a treadmill. With practice, they regain alternating stepping movements, though the paw does not readily clear the surface. Noradrenergic agents may help initiate stepping and the training has lasting effects (8,9). The kinematics and electromyographic activity of the hindlimbs elicited after training at near normal walking
274
Dobkin
speeds closely resembles that of the normal animal. The animals generally cannot walk over ground, however. Training requires some truncal support and stepping is best induced by loading the hindlimbs by pulling down on the tail. Animals trained to stand rather than to step cannot step well, pointing to the specificity of the type of practice (10). These findings suggest that a CPG is at work. Evidence for a CPG in humans has been found in both the spontaneous rhythmic movements made by some patients after SCI (11) and from the evolution of EMG activity in the legs of patients with complete SCI who are manually stepped on a treadmill (12) or undergo electrical stimulation of the dorsal horns at particular frequencies at the L-2 level (13). In the lamprey, modulation of the CPG involves several neurotransmitters (14). Glutaminergic reticulospinal neurons excite ipsilateral spinal motoneurons and interneurons and contralateral glycinergic inhibitory neurons. Glycinergic neurons also have axons that cross to the other half-center of the CPG. Metabotropic receptors for serotonin (14), gammaaminobutyric acid (GABA), and glutamate are also activated within the CPG network during locomotion. Other amines and peptides also modulate the initiation, maintenance, and termination of cell bursts. The interactions of these excitatory and inhibitory ipsilateral and contralateral inputs to the motor pools produce alternating flexion and extension patterns of neuromuscular firing. These systems seem to be conserved in cats and rats (15) and are presumably conserved in humans. Another intrinsic system that may contribute to the flexibility of spinal control for walking has been suggested by studies of microstimulation of regions of the cord in frogs and rodents. A small set of modules appear to store the movements that carry out components of flexion and extension synergistic tasks within the typical workspace of the lower extremity (16). The motor primitives can be activated in chains to achieve functional movements. This synaptic organization could share connections with the CPG, as well as respond to segmental sensory inputs and descending controllers of kinematics for walking (17). In addition, interconnections among columns of motor pools via propriospinal pathways aid postural adjustments during ambulation, along with spinal reflex pathways and vestibulospinal inputs. Sensory inputs provide a powerful source of functional modulation of the CPG, as well as for positive and negative force feedback during walking (18). In the lamprey, rhythmic activation of stretch receptor neurons by alternating locomotor-like activity will entrain the CPG’s locomotor rhythm. In studies of higher vertebrates and human subjects, cutaneous inputs from the sole and Ia and Ib inputs to the hips and ankles are especially important drives for walking. For example, the spinal cord responds to variable levels of loading the legs in patients with clinically complete SCI (19). Studies of cats and humans reveal the impact of the timing of hip extension at the end of stance for one leg and simultaneous loading of the opposite leg for successful initiation of the swing phase (20,21). These
Locomotor Training-Induced Plasticity
275
sensory inputs presumably contribute to the internal models the brain possesses about the properties of limbs and limb mechanics (22). 3.
INTERVENTIONS TO DRIVE ACTIVITY-DEPENDENT PLASTICITY 3.1. Training Sensory inputs generated by locomotor-related movements make a large contribution to the regulation of the stance and swing phases, as well as to ongoing adaptations within cortical representations for movements. Rehabilitation techniques that provide sensory inputs that are recognized by the locomotor system as being typical of walking-related inputs would seem most likely to drive activity-dependent plasticity and behavioral gains. At the same time, cognitive motor strategies to improve motor control will refine the gait pattern and contribute to procedural learning. Investigators have used an approach for patients that simulates the treadmill training used in the studies of the cat after spinal transection. This technique for locomotor retraining, called body weight-supported treadmill training (BWSTT), partially supports the weight of a patient by a parachute-type harness attached at the shoulders to an overhead lift. Initial weight support prevents the hemiparetic leg or paraparetic legs from buckling at the knees. The lift allows vertical displacement during stepping and supports up to 50% of the subject’s weight. Therapists then systematically train patients to walk on the treadmill at increasingly faster speeds and reduce the amount of weight support when feasible. The therapists aim to optimize the kinematic, kinetic, and temporal components of gait that are tied to the stance and swing phases of walking. The upper extremities are assisted to swing rhythmically. This training aims to facilitate walkingrelated sensory inputs for a reciprocal gait pattern, without the immediate requirements of good postural stability and full weight bearing. The training also carries over facilitative cues and an emphasis on sensory inputs relevant to stance and swing for walking over ground and in the community. Robotic steppers (23,24) and functional electrical stimulation (25) may supplement the physically demanding assistance provided by the trainers, if the technique proves to be efficacious. Clinical trials of BWSTT have produced suggestive evidence for the value of this approach for improving walking-related functions in disabled subjects (26,27). Studies of very disabled subjects suggested that the approach would be useful (28). One controlled randomized clinical trial found statistically significant, if modest, increases in walking speed and balance in patients who began their assigned therapy by 70 days after a stroke (29). Other trials that began within 30 days of stroke did not reveal significant gains (30,31). Uncontrolled trials for patients with subacute and chronic SCI also suggest efficacy (32). A randomized, single-blinded clinical
276
Dobkin
trial of patients with incomplete SCI entered 150 subjects during their inpatient rehabilitation, within 8 weeks of onset, to either conventional mobility training or to an equal amount of time using BWSTT and overground training (33). Unlike prior trials, this study aimed to maximize the walking speed of subjects when on the treadmill, rather than using the very slow speeds of former studies, which were often less than 1 mph. Practice at faster speeds leads to greater overground walking velocities (34,35). Walking speed is a fair approximation for the overall pattern of the gait and velocities that approach 2 mph correlate with the ability to walk in the community. Progress toward improved walking after stroke, during BWSTT, may be associated with reorganization of the cortical representation for ankle dorsiflexion (36). One of the most important features of BWSTT is that it allows for massed practice of stepping, unlike usual therapies for patients who cannot walk or walk poorly. Voluntary adjustments of the pattern of gait and manual assistance to engrain best kinematics offer a standardized approach for rehabilitation of walking for highly impaired subjects. Such practice approaches may enable future biologic interventions, such as methods for axonal regeneration, to be brought into the functional reorganization of the brain and spinal cord. 3.2.
Pharmacotherapy
Any of the neurotransmitters discussed could be used to excite, inhibit, or modulate the locomotor pathways in patients with a spinal or supraspinal lesion, even at the level of the CPG (37). Motor skills learning, for example, that produces adaptations in cortical sensorimotor representations may require acetylcholine (38). Cholinergic agents are available to treat Alzheimer’s disease, so experimentation with these agents along with practice of locomotor skills can be put to clinical trials. Case reports describe locomotor changes induced by monamines (39). Training effects and variations synaptic reorganization at each level of the neuroaxis will lead to differential effects of agents. For example, the glycine agonist strychnine enabled stepping in spinal transected cats that were trained to stand, but had no effect on the performance of cats that had been trained to step on a treadmill (40). Clinicians often employ antispasticity agents to reduce spasms and stiffness, but measures of hypertonicity in relation to functional improvements in gait are difficult to ascertain. A specific pharmacotherapy to augment the reacquisition of motor skills during training will require more basic knowledge about manipulating the molecules associated with procedural learning and synaptic plasticity, followed by carefully designed clinical trials. 4.
SUMMARY
At every level of the neuroaxis, experiential practice of behaviorally important activities leads to synaptic adaptations associated with rehabilitation
Locomotor Training-Induced Plasticity
277
training and behavioral changes. The intensity and duration of practice and the essential components of what is practiced to improve walking after a brain or spinal cord injury require continued clinical research. It also remains a challenge to relate specific cellular events, such as production of neurotrophins with exercise (41), to behavioral gains. Understanding the neurobiology of rehabilitation, however, offers clinicians a powerful tool for developing best-of-possible interventions for patients. ACKNOWLEDGMENTS The Larry L Hillblom Foundation; NIH=NICHD R01 #€HD39629; NIH=NICHD R01 HD37439. REFERENCES 1. 2. 3.
4. 5.
6.
7. 8.
9.
10. 11.
Dobkin B. The Clinical Science of Neurologic Rehabilitation. New York: Oxford University Press, 2003. Orlovsky G, Deliagina T, Grillner S. . Neuronal Control of Locomotion: From Mollusc to Man. Oxford, England: Oxford University Press, 1999. Cheney P, Hill-Karrer J, Belhaj-Saif A, McKiernan B, Park M, Marcario J. Cortical motor areas and their properties: implications for neuroprosthetics. In: Seil F, ed. Neural Plasticity and Regeneration. Amsterdam: Elsevier, 2000:136–160. Wolpaw J, Tennissen A. Activity-dependent spinal cord plasticity in health and disease. Ann Rev Neurosci 2001; 24:807–843. Johansen-Berg H, Dawes H, Guy C, Smith S, Wade D, Matthews P. Correlation between motor improvements and altered fMRI activity after rehabilitative therapy. Brain 2002; 125:2731–2742. Dobkin B. Spinal and supraspinal plasticity after incomplete spinal cord injury: correlations between functional magnetic resonance imaging and engaged locomotor networks. In: Seil F, ed. Progress in Brain Research. In: Amsterdam: Elsevier, 2000 Vol.128. 99–111. Dobkin B. Activity-dependent learning contributes to motor recovery. Ann Neurol 1998; 44:158–160. de Leon R, Hodgson J, Roy R, Edgerton V. Locomotor capacity attributable to step training versus spontaneous recovery after spinalization in adult cats. J Neurophysiol 1998; 79:1329–1340. Rossignol S, Chau C, Brustein E, et al. Pharmacological activation and modulation of the central pattern generator for locomotion in the cat. Ann NY Acad Sci 1998; 860:346-359. Edgerton V, Roy R. Paralysis recovery in human and model systems. Curr Opin Neurobiol 2002; 12:658–667. Calancie B, Needham-Shropshire B, Green B, et al. Involuntary stepping after chronic spinal cord injury. Brain 1994; 117:1143–1159.
278
Dobkin
12. Edgerton V, Harkema S, Dobkin B. Retraining the human spinal cord. In: Lin V, ed. Spinal Cord Medicine: Principles and Practice. New York: Demos Medical Publishing, 2003:817–828. 13. Dimitrijevic M, Gerasimenko Y, Pinter M. Evidence for a spinal central pattern generator in humans. Ann NY Acad Sci 1998; 860:360–376. 14. Grillner S, Wallen P. Cellular bases of a vertebrate locomotor system-steering, intersegmental and segmental co-ordination and sensory control. Brain Res Brain Res Rev 2002; 40:92–106. 15. Tillakaratne N, de Leon R, Hoang T, Roy R, Edgerton V, Tobin A. Use- dependent modulation of inhibitory capacity in the feline lumbar spinal cord. J Neurosci 2002; 22:3130–3143. 16. Bizzi E, Tresch M, Saltiel P, d’Avella A. New perspectives on spinal motor systems. Nat Rev=Neurosci 2000; 1:101–108. 17. Poppele R, Bosco G. Sophisticated spinal contributions to motor control. Trends Neurosci 2003; 26:269–276. 18. Dietz V. Spinal cord pattern generators for locomotion. Clin Neurophysiol 2003; 114:1379–89. 19. Harkema S, Hurley S, Patel U, Dobkin B, Edgerton V. Human lumbosacral spinal cord interprets loading during stepping. J Neurophysiol 1997; 77: 797–811. 20. Duysens J, Clarac F, Cruse H. Load-regulating mechanisms in gait and posture: comparative aspects. Physiol Rev 2000; 80:83–133. 21. Dietz V, Muller R, Colombo G. Locomotor activity in spinal man: significance of afferent input from joint and load receptors. Brain 2002; 125:2626–2634. 22. Scott S. Optimal strategies for movement: success with variability. Nat Neurosci 2002; 5:1110–1111. 23. Hesse S, Uhlenbrock D, Werner C, Bardeleben A. A mechanized gait trainer for restoring gait in nonambulatory subjects. Arch Phys Med Rehabil 2000; 81:1158–1161. 24. Colombo G, Joerg M, Schreier R, Dietz V. Treadmill training of paraplegic patients using a robotic orthosis. J Rehabil Res Develop 2000; 37:693–700. 25. Barbeau H, Ladouceur M, Mirbagheri M, Kearney R. The effect of locomotor training combined with functional electrical stimulation in chronic spinal cord injured subjects: walking and reflex studies. Brain Res Rev 2002; 40:274–291. 26. Barbeau H. Locomotor training in neurorehabilitation: emerging rehabilitation concepts. Neurorehabil Neural Repair 2003; 17:3–11. 27. Dobkin B. Overview of treadmill locomotor training with partial body weight support: a neurophysiologically sound approach whose time has come for randomized clinical trials. Neurorehabil Neural Repair 1999; 13:157–165. 28. Hesse S, Bertelt C, Jahnke M, Baake P, Mauritz K. Treadmill training with partial body weight support compared with physiotherapy in nonambulatory hemiparetic patients. Stroke 1995; 26:976–981. 29. Visintin M, Barbeau H, Korner-Bitensky N, Mayo N. A new approach to retrain gait in stroke patients through body weight support and treadmill stimulation. Stroke 1998; 29:1122–1128.
Locomotor Training-Induced Plasticity 30.
31.
32.
33.
34.
35. 36. 37. 38.
39.
40.
41.
279
Kosak M, Reding M. Comparison of partial body weight-supported treadmill gait training versus aggressive bracing assisted walking post stroke. Neurorehabil Neural Repair 2000; 14:13–19. Nilsson L, Carlsson J, Danielsson A, et al. Walking training of patients with hemiparesis at an early stage after stroke: a comparison of walking training on a treadmill with body weight support and walking training on the ground. Clin Rehabil 2002; 15:515–527. Wernig A, Nanassy A, Miller S. Maintenance of locomotor abilities following Laufband (treadmill) therapy in para- and tetraplegic persons: follow-up studies. Spinal Cord 1998; 36:744–749. Dobkin B, Apple D, Barbeau H, et al. Methods for a randomized trial of weight-supported treadmill training versus conventional training for walking during inpatient rehabilitation after incomplete traumatic spinal cord injury. Neurorehabil Neural Repair. NNR 2003; 17:153–167. Sullivan K, Knowlton B, Dobkin B. Step training with body weight support: effect of treadmill speed and practice paradigms on post-stroke locomotor recovery. Arch Phys Med Rehabil 2002; 83:683–691. Pohl M, Mehrholz J, Ritschel C, Ruckriem S. Speed-dependent treadmill training in ambulatory hemiparetic stroke patients. Stroke 2002; 33:553–558. Dobkin B. Functional MRI: a potential physiologic indicator for stroke rehabilitation interventions. Stroke 2003; 34:e23–e24. Pearson K. Neural adaptation in the generation of rhythmic behavior. Annu Rev Physiol 2000; 62:723–753. Conner J, Culberson A, Packowski C, Chiba A, Tuszynski M. Lesions of the basal forebrain cholinergic system impair task acquisition and abolish cortical plasticity associated with motor skill learning. Neuron 2003; 38:819–829. Barbeau H, Norman K, Fung J, Visintin M, Ladouceur M. Does neurorehabilitation play a role in the recovery of walking in neurological populations? Ann N Y Acad Sci 1998; 860:377–382. de Leon R, Tamaki H, Hodgson J, Roy R, Edgerton VR. Hindlimb locomotor and postural training modulates glycinergic inhibition in the spinal cord of the adult spinal cat. J Neurophysiol 1999; 82:359–369. Gomez-Pinilla F, Ying Z, Roy R, Molteni R, Edgerton V. Voluntary exercise induces a BDNF-mediated mechanism that promotes neuroplasticity. J Neurophysiol 2002; 88:2187–2195.
15 Constraint-Induced Therapy for Functional Recovery After Brain Injury: Unraveling the Key Ingredients and Mechanisms C. J. Winstein and M. G. Prettyman Department of Biokinescology Physical Therapy, University of Southern California, Los Angeles, California, U.S.A.
1.
INTRODUCTION
Stroke is the major cause of neurological disability in the United States. The 2003 American Heart Association stroke statistics update placed the annual costs of stroke at $51.2 billion, of which $31 billion were direct medical costs and $20.2 billion were indirect costs due to lost productivity (1). A 1998 projection indicated that the number of strokes may be dramatically higher than previously estimated, reaching in excess of 730,000 every year. Moreover, more than half of these individuals are left with sensorimotor disability and two-thirds of these survivors have persistent disability 5 years later, 37% mildly so and 29% moderately or severely so (2). It is likely that the number of stroke survivors will increase greatly as the population progressively ages over the next 50 years; indeed, a recent projection is that the prevalence of stroke will more than double during this period (3). The high prevalence of stroke and its major economic costs makes the reduction of stroke-related disability a national health care priority. In the past decade, considerable effort has been directed toward the development of novel and innovative approaches to reduce the impairments 281
282
Winstein and Prettyman
and the compromised ability for patients to function in the real world more effectively—to use their paretic limbs to walk, manipulate, and control their environments (4–6). Among the neurological rehabilitation community, this effort has been fueled, to a large extent, by the desire to harness the plastic, adaptive properties of neuronal networks in the adult brain through therapies that engage the patient in active voluntary use of the impaired limbs. One approach for which a literature is rapidly expanding is forced-use or constraint-induced movement therapy (CI therapy) (7–9). Forced-use, in the classic sense, emphasizes perpetual use of the impaired limb during regular activities only by means of restricting movement of the better limb (8). The CI therapy involves the same physical restraint (e.g., mitt, sling, or other devices) but with the addition of a high dose of structured task-specific training (7,9–12). In most cases, CI therapy is administered by a skilled practitioner using principles of training derived from behavioral psychology (13), and includes the operant conditioning method of shaping. In some laboratories and clinics, the training techniques appear similar to those advocated by the CI therapy research group (E. Taub, Director, CI Therapy Research Group), however, they derive, instead, directly from the neural and behavioral domains concerned with the acquisition of skilled movements and motor learning (14–16). This chapter highlights the fundamental differences between these two training approaches in terms of theory and evidence. Two separate detailed reviews of CI therapy have recently appeared and provide an excellent starting point for this chapter (16,17). In addition to these recent reviews, impending outcomes, from a currently running multisite randomized trial (extremity constraint induced therapy evaluation—EXCITE) investigating the effects of CI therapy for improving upper extremity function among adults in the subacute recovery stage following a cerebrovascular stroke, are likely to provide valuable information about the details of efficacy in relation to baseline characteristics, compliance, and specific training protocols (18). Indeed, the impetus for the EXCITE trial was that up to the time of its formulation; CI therapy had only been subjected to feasibility studies in relatively small-scale, single-site projects and limited to the chronic postduration period after stroke (e.g., see (7,8). The reasoning for targeting the subacute population was that if the sensorimotor deficit could be reduced earlier after brain insult or prevented entirely, participants with stroke could avoid experiencing more severe impairments and longer periods of disability than is necessary. Since EXCITE was proposed, a preliminary-randomized clinical trial (19) and two case reports (20,21) appeared suggesting that CI therapy might be helpful for patients who have recently experienced a stroke (19) or who are in the subacute phase of recovery (20,21). The need for such studies is also important, because, depending on the mechanism(s), the nervous system may possess greater potential for functional reorganization sooner after stroke if interventions having
Therapy for Functional Recovery: Key Ingredients and Mechanisms
283
favorable impact can be demonstrated. Indeed, a recent report suggests that the efficacy of rehabilitative experience declines with time after focal ischemic brain injury in a rat model (22). This chapter is divided into two major sections: In Sec. 1, we introduce an evidence-based approach to understanding the essentials of CI therapy. We do this by posing a series of pointed questions that concern the fundamental and foundational assumptions for this therapy. In some (really most) cases, there are no direct answers to these questions or there is not enough evidence at this time to provide definitive conclusions. However, for each issue, we provide an evidence-based discussion including the most relevant data where possible and, when not, we speculate an answer in the hopes that our speculation will provide an appropriate stimulus for future research. We begin by posing a question that requires us to reflect back to the implicit perspective among the clinical rehabilitation community prevalent before the ‘‘decade of the brain’’ (1990s) and, in some circles, this mind-set remains prevalent today (i.e., policy makers for healthcare and rehabilitation). Why was the rehabilitation community initially surprised and even a bit skeptical in response to the early reports of efficacy of CI therapy in cases with chronic hemiparesis? After setting the stage with a discussion of the initial cognitive dissonance that emerged from these early reports, we move into more specific questions related to: inclusion criteria, efficacy and effectiveness, sensitivity of outcome measures, responsiveness to the intervention, and predictors of responsiveness. Specifically, the first section of our review is organized around six fundamental questions: (1) How much of the benefit from CI therapy is due to a reversal of learned nonuse? (2) Does the effector (i.e., arm, hand, leg, mouth) influence responsiveness and outcome? (3) What percentage of the hemiparetic disabled population might benefit from this therapy and why? (4) Is there a fundamental difference between CI therapy and a simple manipulation of the intensity and dose of task-specific training? In other words, has the rehabilitation community rediscovered the well-known power law of learning (practice) under a new and ‘‘improved’’ form termed CI therapy? (5) How critical is the specified CI protocol for optimal success? Finally, (6) what are the best predictors of success with CI therapy and why? Section 2 of this chapter is shorter and provides a review of the evidence for two complimentary mechanisms thought to mediate the effects of CI therapy, one of which ties this work into the primary theme of this book, neural plasticity, and the other, learned nonuse, provides a behavioral theory for the compensatory strategy thought to be triggered after brain damage and reversed with CI therapy. We propose that if the target of CI therapy is the reversal of learned nonuse, there should be further exploration of this theoretical construct (23). In conclusion, we propose a general neurobehavioral model for the activity-dependent plasticity response that occurs following neurological insult, in which there are two divergent
284
Winstein and Prettyman
behavioral paths termed ‘‘optimal function’’ and ‘‘compensatory function’’ (24). We suggest that CI therapy is but one limited example of a larger class of task-relevant behavioral therapeutic interventions, including electromyography (EMG) triggered-functional electrical muscle stimulation, bilateral upper extremity training, robotic therapy, and Body-weight assisted treadmill training that alone or combined with certain pharmacological agents (e.g., neural growth factors, NMDA, and acetylcholinesterase inhibitor) can be used to drive the system toward optimal functional recovery by integrating rewired motor representations (procedural skills) into the reorganized neural circuits. 2.
ESSENTIALS OF CONSTRAINT-INDUCED THERAPY: WHAT IS THE EVIDENCE?
First coined ‘‘forced-use’’ therapy developed from the basic animal model work of Taub and others (25–27) with observations in primates that, after unilateral forelimb deafferentation, the monkey did not use the affected limb in the free environment (27). However, in these cases of unilateral deafferentation, the monkey could be induced to use the deafferented extremity, though somewhat clumsily, if the intact limb was restricted from use. Recent descriptions of forced-use therapy propose a generic term for this approach as ‘‘constraint-induced’’ (CI) therapy. Taub and Wolf (11) describe CI as a group of therapies with one important type characterized as forced-use. In 1989, following the first case report (28), Wolf conducted the first feasibility study of ‘‘forced-use’’ in humans with a chronic hemiparesis of the upper limb (8). The inclusion criteria were a minimum motor ability to voluntarily extend (lift the hand) the wrist 20 from a fully flexed posture and, to extend, the fingers approximately 10 . There was no specified sensory criterion. These motor criteria were derived from a series of previous EMG biofeedback studies conducted to predict which patients with chronic stroke would gain independent use of the hemiparetic upper extremity (29). This feasibility study included 21 patients, 16 stroke and 5 traumatic brain injury, all greater than 1 year from onset who were ‘‘forced’’ to use the involved arm by wearing a sling (worn during waking hours, but removed for sleep and 30 min for exercise) to immobilize the less-involved (relatively intact) arm over a period of 2 weeks. The results from this 1989 feasibility study showed significant changes on 19 of the 21 items from a laboratory-based timed task battery (i.e., patients moved faster at the posttest compared to the pretest), most of which persisted at the 1-year follow-up point. Further, there were no differences between the stroke and traumatic brain injury patients or right and left-hemiparetic responders.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
285
Following this landmark feasibility study of the application of a forced-use protocol in humans with upper extremity hemiparesis, Taub et al. (7) conducted a second small-scale (n ¼ 9) feasibility study in 1993, and used a randomized-trial design (attention control group compared with a forced-use experimental group) and added a supervised-task practice component to the forced-use protocol. Inclusion criteria were similar to those used by Wolf et al. (8), for the minimum motor criterion, but included only right-handed, right-hemisphere-affected patients post-stroke who were between 1 and 18 years from onset. This hand-dominance and hemisphere selection criterion is interesting because it is not predicated on the prior primate deafferentation work but rather, some suspicion that hand= hemisphere dominance could be important in the translation to human clinical trials or similarly, that the language deficit associated with left hemisphere stroke might influence participation in the intervention. The other significant difference between this and the other prior trial (at that time) conducted in 1989 (8) concerns the intervention itself. Here, on each of the 8 weekdays during a 12-day restraint period (restraint consisted of a resting hand splint and sling), the patients spent 7 hr at the rehabilitation center where they practiced a variety of functional tasks with the paretic upper extremity (e.g., eating a meal, throwing a ball, playing dominoes, writing, using purdue dexterity board) for 6 hr=day. The patients in the forced-use plus task practice group (n ¼ 4) did not receive any kind of therapy, but simply an opportunity to practice functional tasks in a supervised and friendly environment. Taub et al. did not define the supervised-task practice as a form of ‘‘therapy’’ in this case, in large part, because the personnel who supervised the task practice had no specific professional training to administer physical therapy. The attention control group was provided with instructions to focus attention at home on using the affected extremity for as many activities as possible and this was augmented with two physical therapy sessions that emphasized passive movement of the involved limb or other nonpurposive movements of the affected side, and self-range-ofmovement exercises to be done 15 min=day at home. The results of this pilot study were encouraging, specifically improved performance times on the Emory timed task (later renamed, Wolf Motor Function Test (WMFT), and the Arm Motor Activity Test (AMAT) for the restraint plus task practice group, compared to the attention control group. In addition, the experimental group increased their ability to use their upper extremity in a variety of daily activities as evidenced by improved scores on the motor activity log interview (self-ratings of amount of use (AOU) for 14 functional, daily activities) compared to the attention control group. Most importantly, the changes for these four experimental subjects were maintained at the 2-year follow-up point.
286
2.1.
Winstein and Prettyman
Why Was the Rehabilitation Community Initially Surprised and Even a Bit Skeptical in Response to These Early Reports of Efficacy of Forced-Use=CI Therapy in Cases of Chronic Hemiparesis?
Among neurorehabilitation therapists, CI therapy provides one of the first examples of a clinical therapeutic intervention that evolved directly from nonhuman primate work. Having said that, we acknowledge that a parallel development, also from nonhuman animal experiments (e.g., spinalized cat and rat), characterizes the clinical therapeutic intervention of body-weightassisted treadmill training (BWATT) to promote recovery of locomotor capability in patients recovering from an incomplete spinal cord injury or stroke-hemiparesis (see Ref. 30 for a recent review). Prior to publication of the early efficacy studies of CI therapy in humans, much of the therapeutic approach in neurorehabilitation was based on theoretical models (e.g., systems model, hierarchical control model) and expert opinion that together characterized the ‘‘neurofacilitation’’ approach (31). Although there was scattered evidence that focused functional task practice protocols and positional feedback with electrical stimulation applied even in the chronic stage of recovery (i.e., when recovery was thought to have reached a plateau) could result in significant improvements in performance (32–36), these few reports went relatively unnoticed, particularly as they were inconsistent with the traditional view that further physiological recovery was not possible at this chronic stage after brain damage. However recently, a major paradigm shift in rehabilitation medicine has come about primarily due to the recognition and application of the principles of learning and adaptation associated with the acquisition of motor skills to treatments aimed at the recovery and rehabilitation of movement following disease or injury (16,24,37–39). This quiet revolution has evolved slowly as the traditional views in neurobiology of ‘‘relatively permanent’’ and fixed impairments have given way to more recent views that embrace the inherent adaptive plasticity of the brain and spinal cord systems through ‘‘activity-, use-, and learning-dependent’’ processes (14,40–42). While a passive awareness of therapy-triggered adaptation in the damaged nervous system is not particularly new, its acceptance, as evidenced by the active development of innovative therapies to exploit (drive the system) this capacity, has evolved progressively over the last decade. A 1915 preliminary report written by the well-known physician-scientist Franz et al. (32) was titled: The possibility of recovery in long-standing hemiplegia. This report along with the early CI therapy results in patients with chronic stroke (7,8,28) was virtually ignored by some, or viewed with skepticism by others. These reports provided contrary findings to the prevalent view of the injured nervous system that attributed the recovery process primarily to the early physiological responses (e.g., diaschesis) and not the later behavioural
Therapy for Functional Recovery: Key Ingredients and Mechanisms
287
adaptation (e.g., motor learning) that only recently has become the focus of attention in neurorehabilitation practice. On the other hand, reasons for some of the past and current skepticism toward reports of efficacy of CI therapy are not entirely unwarranted (43). Unfortunately, there have been recent claims of efficacy from CI therapy (i.e., ‘‘95% effective’’) that appear to go beyond the available evidence. Indeed, to date, there have been only four published randomized-clinical trials (RCT) of CI therapy (7,9,19,44). In all four studies, the authors reported positive results, yet the effect sizes calculated without covariates yielded no statistically significant differences (23). In the largest of these four RCTs, Van der Lee et al. (9) found a differential effect for patients with sensory disorders and hemineglect that lead to the hypothesis that learned nonuse may be associated with an afferent impairment (similar to the deafferented primate). This line of investigation requires further exploration and is the focus of one of our guiding questions. We apply a systematic series of questions in the sections that follow as a means to review the evidence surrounding the efficacy of CI therapy in clinical practice, and, in doing so, place this therapeutic approach within a broader category of interventions designed to promote functional recovery (45). 2.2.
Attempts to Answer a Number of Key Questions 2.2.1. Question 1: How Much of the Benefit from CI Therapy Is Due to the Reversal of Learned Nonuse?
One mechanism that has been suggested to explain the effects of CI therapy has, as its premise, the phenomenon of ‘‘learned nonuse.’’ In fact, CI treatment is intended to help patients with central paresis overcome ‘‘learned nonuse’’ of the paretic limb by discouraging use of the less affected limb in combination with intensive task-specific training of the paretic limb. The learned nonuse hypothesis is a conceptualization based on the observations of the behavior of monkeys after somatosensory deafferentation (12,27,46). The hypothesis is that following unilateral deafferentation (or central damage and during the CNS shock period) attempts to use the limb are met with failure (punishment, negative reinforcement), and the monkey or human soon ‘‘learns’’ through operant conditioning to use the other arm for which attempts are met with some success (reward, positive reinforcement). Successful accomplishment of tasks with the mostly intact extremities further reinforces the ‘‘learned nonuse’’ of the impaired side, so the argument goes. This ‘‘learning’’ phenomenon is thought to mask the natural recovery that occurs following the insult and results in a discrepancy between what ‘‘can’’ be done with the impaired limb (motor capability) and what ‘‘is’’ done with the impaired limb (47,48). In unilateral deafferented monkeys, application of a restraint to the intact upper limb, with or without conditioning of the deafferented limb,
288
Winstein and Prettyman
produced long-term motoric use (though somewhat clumsy use) of the deafferented limb for normal tasks. In contrast, following the recovery period from bilateral deafferentation, monkeys would automatically apply the remaining motor capability to meaningful mobility and feeding behaviors without the use of a restraint. This suggests that learned nonuse might be one of a class of compensatory strategies that is exclusive to asymmetrical impairment conditions. Finally, if the monkey is restrained from attempting to use the limbs (both deafferented and intact) for 3 months (unilateral deafferentation procedure was performed in utereo, 3 months prior to birth), functional motor use was observed after birth in two of the three monkeys. One interpretation of this experiment is that the restricted use of both limbs during the recovery period (in utereo) prevented the development of learned nonuse. Together the observations of motor behaviors following unilateral, bilateral, and in utereo restraint experiments with monkeys led to the development of the learned nonuse hypothesis. While the results of these few experiments with monkeys support the notion that ‘‘nonuse’’ is a learned behavior following a systematic afferent lesion at the dorsal root (deafferentation), the degree to which this learning phenomenon can be applied to the human with central paresis, that includes various combinations of motor and=or sensory impairments, is not altogether obvious. We suspect the presence and evolution of learned nonuse is much more complicated and that it varies substantially between patients and conditions. There is considerable need for further systematic research in this area and into the development of valid and reliable instruments for measuring and monitoring the evolution of nonuse in humans that begins immediately after a strokehemiparesis. For comparison purposes, it would be useful to develop a database or classification system of how and for what natural tasks (e.g., unimanual, bimanual) humans use their arms and hands during everyday activities. This could then be applied to the development of a functional instrument to capture use data after stroke and other disabling conditions of upper extremity function. Learned nonuse of the upper extremity in humans following CNS damage has been described by Taub (27) and Wolf et al. (8). Cerebrovascular accident (CVA) or stroke and brain injury populations may undergo operant learning processes similar to that described for the deafferented monkey. As learned nonuse evolves, a mismatch develops between movement capability (i.e., minimal motor criteria) and the patient’s reported use of the affected limb for functional activities. In humans, limb use has been captured by a structured interview instrument, the motor activity log (MAL) that was developed by Taub et al. (7). Since the original description of the MAL as a self-rating (6 point scale) of affected extremity use (quantity) for 14 daily tasks, the number of tasks assessed has varied in the literature with examples of 20 items (48), 25 items (9), and 30 task items (18), and the instrument was expanded to include a quality (how well) rating. Criteria
Therapy for Functional Recovery: Key Ingredients and Mechanisms
289
for what constitutes learned nonuse in the chronic stroke population include two characteristics: (1) affected arm achieves the minimum motor criteria, and (2) MAL score for affected arm use is not greater than 2.5 (mean score using a range of 0–5); 2.5 is between ‘‘rarely’’ (2.0) and ‘‘half as much’’ (3.0) use compared to prestroke use levels for that task (10,49). Sterr et al. (48) are the only research group to have systematically tested the learned nonuse hypothesis with a group (n ¼ 21) of young (mean 18 years, range 5–26 years) childhood brain injury participants who exhibited some degree of upper limb hemiparesis. Learned nonuse was operationally defined as a conditioned suppression of affected limb movement that results in a clear discrepancy between residual motor function (motor capability) and daily activity use; performance was compared with a healthy age-matched control group (n ¼ 21). This specific age group was chosen with an expectation of an enhanced recovery potential, a quicker learning capacity, with perhaps more pronounced learned nonuse or, on the contrary, a lessoned learned nonuse dampened by an enhanced and natural exploratory capacity found in younger compared to older participants. All subjects were right-hand dominant and had a mean age of 18 years with a diverse time postonset of between 3 and 122 months (mean 27 months). Inclusion criteria for the patients were similar to that of adult CI therapy studies and included affected arm residual motor capacity, cognitive function, and minimal muscle tone (minor spasm on Ashworth Scale). The hemiparetic patient and normal control groups were tested with two versions each of the MAL and actual amount of use test (AAUT) during a single test session (2 tests 2 versions=conditions). Each test was administered in the ‘‘spontaneous’’ version and the ‘‘forced’’ version. The 14-item AAUT consisted of asking the participants to perform different tasks without specifying how or with which hand (spontaneous, sAAUT) and the forced version (fAAUT) in which participants were asked to perform the same tasks, but specifically to perform with the affected arm. Similarly, the MAL (20 tasks) consisted of a subjective judgment by the patient= control (0–5 rating) of the amount (sAOU) and quality (sQOM) of arm use, and the actual (forced) ability to perform (aAOU). Comparison of the affected limb in spontaneous (sAAUT) and subjective (sAOU and sQOM) conditions with that in the actual and forced conditions (fAAUT, aAOU) was used as a measure of learned nonuse. A mean difference of 75.8% between the spontaneous use and forceduse of the affected arm in the hemiparetic group on the AAUT revealed a significant arm nonuse despite motor capability. Of 21 hemiparetic participants, 19 had the capability in the forced-use condition (fAAUT) to use the affected arm for 100% of the tasks. The discrepancy between spontaneous and actual use was evident for both dominant (50%) and nondominant (55.6%) affected arms. On the MAL, the hemiparetic group rated the amount (mean 2.3) and quality (mean 2.2) of movement significantly lower
290
Winstein and Prettyman
than the normal controls (mean 4.9 and 4.9, respectively). The subjective (sQOM) and actual performance (aQOM) ratings of quality on the MAL were not significantly different in the controls, but did differ significantly in the hemiparetic group regardless of whether the right or left arm was affected. Differences were twice as large when the dominant arm was affected indicating a potentially important factor for the development of learned nonuse. These initial findings specific to a young population offer support for the hypothesis that learned nonuse represents a mismatch between actual use and residual motor capability. Additionally, the mean sAOU score of 2.3 in the young hemiparetic group is close to the mean MAL amount criterion score (2.5) previously used to establish the presence of learned nonuse (10,49). Using the learned nonuse criterion and in the presence of minimal motor impairment, we describe the case of a right-hand dominant 19 year-old male (MC) to illustrate the magnitude of change that can be elicited with the MAL questionnaire and motor behavior following a bout of CI therapy. Nine months prior to receiving CI therapy, MC sustained cerebral damage following a gamma knife procedure to correct an arteriovenous malformation thought to be associated with a seizure disorder foci in the left cerebral hemisphere (Fig. 1). Sensory (12=12) and motor (60.5=66) impairment were evaluated with the upper extremity portion of the Fugl-Meyer assessment (FMA) (50); this revealed intact sensation and nearly normal movement capability of the affected right upper extremity (Fig. 2, left side). Functionally, MC was able to perform various natural tasks and could perform all 17 tasks on the WMFT (13 of the 15 timed tasks were performed more slowly with the affected dominant right arm than with the nondominant arm). Despite this movement capability, his subjective report was ‘‘rare’’ use (MAL, mean ‘‘amount of use’’ score 1.7) of the right extremity for 30 different daily tasks. A family member concurred with the subjective report of ‘‘rare’’ (caregiver MAL mean AOU score 1.2) use of the affected extremity even though movement capability was present. Following a 12day CI intervention period (10 days of task practice training in the lab and 12 days of mitt=restraint), MC’s MAL AOU score changed from ‘‘rare’’ to ‘‘almost as much as before’’ (MAL mean score 4.2), during daily tasks. In contrast, the motor behavior tests showed only a small change; the WMFT average movement time change (faster performance) across the 15 timed tasks was 2.1 sec. This case provides an illustration of the characteristic mismatch between motor capability and functional use that exemplifies the phenomenon of nonuse. On the other hand, it is inconsistent with the idea that learned nonuse may be a phenomenon more associated with afferent disorders (23,46); MC tested with intact sensation, at least in response to our crude clinical tests. However, on closer exam and during administration of the MAL, he did report subtle differences in feel (‘‘slight numbness’’)
Therapy for Functional Recovery: Key Ingredients and Mechanisms
291
Figure 1 T2-weighted MRI brain images for case MC showing the small subcortical lesion as the bright white area in the left hemisphere (right side) adjacent to the lateral ventricle and in the next two inferior slices all including a portion of the left corona radiata projections.
when he used his right hand for certain tasks (e.g., using the TV remote; touch tone on phone). To more closely track the change in use during this bout of CI therapy, we administered the MAL at the midpoint after the first week of training. Already, there was a marked increase in reported use with an average MAL amount score of 4.4. For 19 of the 30 tasks, MC rated his AOU the same as before his gamma knife surgery. This suggests that simple exposure to various functional tasks during the training and with attempted right dominant arm use, MC developed a perception that he could perform these tasks (self-efficacy) relatively well. This exposure allowed him to incorporate these activities into his daily repertoire and be reflected in his responses to the MAL interview by the end of the first week. Although difficult to ascertain with one case, an important cofactor in the relatively rapid reversal of learned nonuse in this case may have been that the affected limb
292
Winstein and Prettyman
Figure 2 Data from two cases (MC, BD). FMA, upper extremity impairment measure of motor ability (50). MAL structured interview of subjective ‘‘use’’ of upper extremity (30 daily activity tasks, rated for how much and how well).
was also the dominant limb. There is no evidence that hemispheric asymmetry or hemispheric dominance are factors in either the incidence of learned nonuse or its reversal, however, to our knowledge, only one study has investigated this question (48). For the case of MC, it appears that much of the short-term change in performance (use of the affected arm and hand) could be explained as a reversal of learned nonuse and not the learning of any novel skills or relearning of the original motor skills. With the affected right side, he was able to perform 15=15 of the timed tasks on the WMFT at the pretest, but had a slightly longer movement time than the left hand on 13=15 of these tasks. Likely, it would be a different case, if motor capability was also impaired. We provide a second example (see section for clinical Question 3) for comparison that describes a case (BD) with a much greater initial impairment level and associated brain lesion. To date, the work of Sterr et al. (48) provides the only published systematic, objective investigation of the phenomenon of learned nonuse. Although this is an excellent beginning, one study cannot validate this complex phenomenon. Caution is warranted in generalizing these findings for three reasons: (1) the study group consisted of young brain injured participants; (2) established clinametric properties for the MAL and AAUT are lacking for the standard instruments, and with the modifications of Sterr et al., and (3) no other related patient impairments (e.g., motor, sensation, neglect, or apraxia) were reported, making it difficult to understand the full clinical characteristics of this group. Recently Van der Lee (23) echoed our
Therapy for Functional Recovery: Key Ingredients and Mechanisms
293
concerns and suggested, ‘‘there is no validated measure to diagnosis the presence or severity of learned nonuse’’ . . . learned nonuse ‘‘is only based on clinical impressions, and may be associated with sensory disorders or hemineglect’’ (p. 41). Further research that includes other relevant dimensions of the nonuse phenomenon, such as sensation (afferent perception), hand-dominance, and reward–punishment elements is needed to allow a more precise characterization and better methods to quantify the role of operant conditioning in the behavioral development of learned nonuse. Recent reports of rapid modulation of GABA and rapid cortical reorganization in the sensorimotor cortex after acute deafferentation might provide a physiological explanation for the emergence of learned nonuse (51). Longitudinal measurements of GABA modulation and cortical reorganization with parallel behavioral observations of limb use that is documented from the time immediately following cortical injury to the time of full, limited or absent recovery, would allow a more complete understanding of the behavioral phenomenon of nonuse, its neural correlates, and tests for interventions that might prevent or reverse its development. It may be that nonuse is not learned at all, but the result of a rapid reorganization of the sensorimotor representation of the limb. 2.2.2. Question 2: Does the Effector (i.e., Arm-Hand, Leg, Mouth) Make a Difference? Effectors, as body parts receiving neural input for behavioral actions or movements, may become impaired as a result of injury or disease causing a decline in function and disability. The CI therapy targets the impaired effector with constraint of the less impaired body part (effector) in order to induce increased behavioral experience (practice) with the impaired effector. For the most part, CI therapy has been applied to the upper extremity for those with unilateral brain injury (stroke or brain trauma). However, there are reports of CI therapy effectiveness for other effectors impaired by chronic stroke (12,52,53) and other pathologies (12,54,55). These brief reports are summarized in the following section to provide a critical appraisal of the similarities and differences of these applications with current systematic studies and with standard clinical practice. Repetitive and intense task practice, identified as a component of CI therapy, have been applied to the impaired lower extremity with a small number of patients whose disability was from chronic stroke, spinal cord injury, or hip fracture (12,52). For 16 chronic stroke patients ranging from nonambulatory to moderately impaired coordination, a 3-week, 7-hr=day task practice of lower limb activities with body weight support harness was reported to be effective for all participants. Application of the body weight support harness was not specified in its frequency, duration, or percent body weight; alternatively, lower limb tasks were described as treadmill and over-ground walking; sit-to-stand and lie-to-sit transfers; step climbing;
294
Winstein and Prettyman
and various balance and support exercises. Specific measures and outcome data were not reported, but four minimally ambulatory patients became independent and 12 moderately impaired patients reportedly improved. Despite the lack of any quantitative data, a power analysis, or presentation of baseline patient characteristics, a large effect size (1.6) was reported. Further, the proposed mechanism for this large effect was described as the overcoming of learned misuse (bad habits of coordination, or maladaptive behavioral compensation). In the same report (12), two patients with incomplete spinal cord injury and one recovering from a total hip joint replacement surgery participated in a similar 3-week lower extremity CI therapy program. In all cases, there was improved gait velocity and independence during gait. Constraint-induced ‘‘facilitation’’ is the mechanism described by Birbaumer and Taub (55) in an unpublished case report of lower limb treatment following spinal cord injury. ‘‘Constraint’’ in this manuscript was defined as the withdrawal of crutches, walker, and mirror observation of gait. Treatment consisted of gait training while holding grocery bags in a safe environment (path surrounded by foam) for 2-hr sessions on three consecutive days. Immediate outcome after the third treatment was independent ambulation without an assistive device at a slow speed and with visual attention directed to the lower extremities. Taub et al. (54) reported on five patients with spinal cord injury who received lower extremity CI therapy with concentrated training (shaping) as described in 1999 and the addition of limb load monitor feedback and a lower limb electrogoniometer for 6 hr=day for 15 weekdays. Again, despite a paucity of data, reported outcomes for the intervention group were better than a placebo or reduced treatment group with effect sizes above 1.5. What is particularly confusing about these reports is that there is no obvious ‘‘constraint’’ for the CI therapy protocols with the lower extremity. The constraint is a consistent component of the signature CI therapy protocol, but for these lower extremity CI therapy protocols, the constraint is embedded into the way that tasks are practiced (i.e., favoring the affected extremity). It is not obvious to us how removal of an assistive device, or preclusion of vision during gait is analogous to the constraint of the unaffected limb. Instead, these procedures are very similar to those routinely used by skilled clinicians who manipulate the environment and tasks to foster less reliance on the unaffected limb, reduce maladaptive strategies, and promote functional recovery. Another important factor that is lost with such an approach across effectors is an appreciation for the fundamental neural and behavioral differences between the motor control underlying locomotion and leg function compared to that for arm and hand function for reaching– grasping actions. We are confident that a better understanding of the similarities and differences between the neural mechanisms responsible for gait (locomotion) and lower extremity compensation after cortical injury, and
Therapy for Functional Recovery: Key Ingredients and Mechanisms
295
that associated with upper extremity hand and finger dependence on intact corticospinal tract integrity, and its compensation through unimanual function, will lead eventually to better hypothesis-driven approaches to foster functional recovery (56–59). Indeed, recent evidence with animal and human studies suggests that spinal cord networks participate in locomotion-like behavior and that the use of BWATT (4,30,60) and task-specific practice (24,61,62) is effective for improving functional outcomes. It is therefore surprising and somewhat confusing that such therapeutic techniques as BWATT have been subsumed under the guise of the signature CI therapy protocol when they do not share a critical element of that signature—the constraint! Linguistic function, presumably with oral musculature, in aphasic patients forms a nonextremity effector group that has received treatment under the auspices of CI therapy. In a single report, chronic CVA (n ¼ 17) patients with expressive aphasia due to left hemispheric damage were randomized to an experimental CI aphasia therapy group (n ¼ 10) or conventional treatment group (n ¼ 7). The experimental group had gestures and written communication constrained (suppressed) along three dimensions: (1) difficulty of material; (2) rules of the game and shaping; and (3) imposed reinforcement contingencies while playing a card game in a group game activity for 3–4 hr=day for 10 days (53). The syndrome-specific standard approach received by the conventional therapy control group included naming, repetition, sentence completion, following directions, and conversation over a 3–5-week period for equating total treatment hours between groups, but not frequency or number of treatment hours per day. Language function was assessed with the subsets of the Aachen Aphasia Battery and a Communication Activity Log (CAL) questionnaire developed for the project. Significant positive outcomes one day following the end of treatment were found for the experimental group in the token test, naming test, and language comprehension of the Aachen Aphasia Battery, a 17% change in the CI aphasia group and only 2% change (1 subtest improvement) in the control group. Group and time were found to have a significant interaction with aphasia test scores. A subjective increase (30%) in the amount of communication (CAL) in everyday life was found for the experimental group but no improvement was found for the conventional therapy group. Blinded raters of amount of communication in everyday life reported a 10% improvement for the CI therapy group. The authors concluded: ‘‘(1) that massed-practice CI aphasia therapy appears to be efficient for improving language performance of patients with chronic aphasia within a short period of 10 days and (2) that a better outcome was obtained by use of a concentrated CI aphasia therapy regimen instead of the same amount of conventional therapy spread out over a longer period (53) (p. 1625)’’. Both of these conclusions suggest that the intensity or dose (duration and frequency of practice) of therapy, rather than a novel intervention technique, may be the critical factor for improving language function in a chronic aphasia population.
296
Winstein and Prettyman
Evidence for functional benefits following CI therapy for the upper extremity as the effector includes four randomized-controlled trials and the signature constraint principle with constraint of the less impaired upper extremity, an unambiguous part of the intervention (23). In contrast, what has been described as CI therapy for the lower extremity and oral–motor structures consists primarily of small group designs without any clear application of a constraint device. Baseline characteristics, control groups with equivalent treatment dose, and specific outcome data are lacking in the publications of CI therapy for effectors other than the upper extremity. The descriptions of CI therapy components for gait and verbal-speech do not provide a unique therapeutic intervention, other than the extended duration and frequency (dose) of therapy that is different from standard care. This factor alone clouds the issues and raises doubts about the nature of CI therapy. Indeed, a recent review (56) raised the question ‘‘Is CI therapy different or simply more of the same?’’ The standard dose of out-patient therapy in the chronic stage post stroke ranges from zero to the more common dose of a 1-hr session 2–3 times per week for up to 6 weeks duration. We will discuss this further in response to question 5 below. In other words, are the positive outcomes found for gait and verbal function unique to CI therapy, or simply a more intensified application of common training principles with a protocol of enhanced duration, frequency, and feedback and using creative means to prevent reliance on common compensatory task solutions (use of assistive device, nonverbal communication)? 2.2.3. Question 3: What Percentage of the Hemiparetic Stroke and Head Injury Population Might Benefit From This Therapy and Why? The CVA or stroke may affect many areas of the nervous system and result in a variety of impairments that contribute to a lack of activity (functional limitation) and participation in society (previously termed handicap) (58, 63–65). While it is not the purpose of this chapter to review the extensive stroke outcomes literature, it is noteworthy that this population is quite heterogeneous with regard to: medical outcomes (death to full recovery with premorbid status), lesion characteristics, and clinical assessment measures (58,63,65,66). Stroke severity classifications and outcome prediction models have been correlated with lesion volume, recovery of motor skills, and functional gains (66–69). Recovery of arm function following stroke is plagued with problems of definition (e.g., what constitutes ‘‘recovery’’, what constitutes ‘‘function’’?) and measurement (57,58,65,70). The majority of longitudinal studies specific to upper extremity recovery after stroke suggest that prediction is possible within the first month and functional improvement generally reaches a plateau 3–6 months after onset (57,58,67,69). Outcome studies using arm-specific measures indicate full recovery for 11–14%, partial recovery
Therapy for Functional Recovery: Key Ingredients and Mechanisms
297
for 25–38% and absent functional recovery for approximately 62% of stroke survivors (57,69). Factors proven to have a predictive value for recovery of upper extremity function following stroke include: initial motor impairment (arm and leg), presence of grip strength, loss of position sense, and cognition (57,68,69). The use of established stroke severity classifications or predictive factors in the CI therapy literature is lacking making comparison difficult. Despite this problem, claims have been made that CI therapy is applicable in 75% of the chronic stroke population (12). Perhaps the predominance of CI therapy in the chronic stroke population has precluded the need to classify subjects along a severity continuum. Nevertheless, we will attempt to integrate the upper extremity CI therapy inclusion characteristics with specific outcome findings from the literature to offer a revised hypothesis for what percentage of the stroke population we predict could benefit from a CI therapy protocol. For this classification, we consider motor impairment, sensory impairment, and other indicators including cognitive function. Motor Impairment: Based on longitudinal study outcomes, 30–40% of stroke patients appear to have potential for some degree of recovery of arm function contralateral to a unilateral hemispheric lesion (57,58,67– 69). Initial (within 2–3 weeks of onset) motor deficit of the arm is a predictor of the 6-month recovery. Specific measures of initial motor impairment with grip dynamometry, the FMA (upper extremity motor portion) (50) and Motricity Index have been significant for outcomes on the Action Research Arm (ARA) test, Frenchy Arm Test and a hierarchical scale of graded arm tasks (57,58,68,69). The CI therapy inclusion criteria include the capability for active voluntary extension of the affected wrist and fingers, (minimal motor criteria, high level: wrist and all fingers; low level: wrist two fingers and thumb) to ensure a minimum level of voluntary movement control. Two recent CI therapy clinical trials reported having used the National Institutes of Health (NIH) Stroke Scale, Motor Assessment Scale, and ARA test for inclusion criteria and to exclude those who might have reached full recovery (9,19). Together these inclusion criteria indicate that the populations enrolled and found to benefit from CI therapy had an initial motor impairment with distal movement capability in the affected arm, but whose affected upper extremity was not fully recovered. Indeed, a comparison of early (1–4 weeks following stroke) outcome predictors and the minimal motor inclusion criteria, typical for the chronic population, yields a strong potential for partial to full recovery (68) of arm function that has been demonstrated by the majority of characteristic CI therapy patients studied thus far. As such, the findings for enhanced recovery are not particularly surprising, given that improvement potential is present based on the apparent motor capability. Grip strength, a well-known predictor of upper extremity outcome (58,68), has not been used as an inclusion criteria in CI therapy-controlled
298
Winstein and Prettyman
trials (7,9,19), but the ability to grasp and release an object has been used as an inclusion criterion for several more severe patients who did not qualify under the standard minimal motor criteria, but for whom, CI therapy was administered (12,71). An untested and speculative assumption posits that the wrist and finger extension minimal motor criterion is correlated in some way to the generation of active finger flexion necessary for functional grip strength. Grip strength has been shown to be a good predictor of upper extremity recovery and is included as an item in the WMFT; we suggest that a retrospective analysis of grip strength and its association with CI therapy outcomes would provide concurrence between standard outcomes in the literature and those used for CI therapy (8,9,18,44). Several of the published CI therapy trial outcomes provide baseline data for affected arm FMA and ARA test in acute (19), subacute (44), and chronic (9) patients; these data could be used to predict functional arm recovery. The upper extremity baseline FMA score (max 66) is reported in two studies (9,44) and provides data for comparison with other published stroke recovery outcome data. A 94% probability of recovering-affected extremity dexterity was determined if FMA motor score was 19 at 4 weeks poststroke (69). Application of this FMA criteria predicts that 100% of the stroke patients in Van der Lee et al. (9) (mean FMA ¼ 51) and Page et al. (44) (mean FMA ¼ 47.5) would be expected to recover hand dexterity and functional arm use. Kwakkel et al. (69) found that 38% of the patients with an ARA test score 10 gained dexterity. Translating this to baseline test results for CI therapy outcomes is speculative. However, three CI therapy publications reported baseline scores (mean ARA ¼ 32.1, 25.9, and 48.5, respectively) (9,19,44). These scores are above the ARA score 10 found in 38% of patient who recovered hand dexterity (69). Assuming other patients enrolled in CI therapy with the ‘‘standard’’ (voluntary wrist and finger extension) minimal motor criteria have similar baseline FMA and ARA scores to those found in Kwakkel et al., (69), improvement would not be surprising, indeed, it would be expected. In addition, consistent use of valid and reliable instruments including the FMA and ARA would provide a more objective and clear definition of ‘‘In addition, consistent use of valid and reliable instrument including the FMA and ARA would provide a more objective and clear definition of residual motor deficit that is often used to characterize the chronic stroke group thought to benefit from CI therapy [17]’’ that is often used to characterize the chronic stroke group thought to benefit from CI therapy (17) Together the predictor variables for motor impairment from the literature and the minimal motor criteria used exclusively for CI therapy inclusion, we estimate that approximately 40% of all stroke survivors demonstrate potential for functional arm recovery (2,57,69). Of the 40% who are typically those characterized with ‘‘mild disability,’’ we suggest that between 70% and 80% would be expected to benefit from CI therapy or its equivalent.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
299
Sensory Impairments: Sensory perception, a second early predictor of recovery of arm function after stroke (57), poses an additional challenge for estimating benefit from CI therapy. Impairments of sensation include multiple modalities (tactile, proprioceptive, visual, and auditory) and have not consistently been reported in the hemiparetic arm outcome literature. Impaired sensation (nonspecified modality) was found to limit recovery on the nine hole peg test (68), and loss of position sense was found to be a significant predictor of poor upper extremity outcome (57,69). Only one of the CI therapy RCTs stratified subjects by sensory loss in the affected upper extremity; after recruitment, 54% of the intervention group demonstrated some sensory impairment (9). Of particular interest, the subjects with sensory impairment who were randomized to the CI therapy group demonstrated a lasting (1 year) improvement in arm function that was greater than those randomized to the bimanual therapy group. Van der Lee et al. (9) suggested that there was a clinically meaningful (persistent improvement) change for the CI therapy patients with sensory disorders and that perhaps patients without sensory impairment had already reached an upper limit of function (without learned nonuse). While others have demonstrated limited recovery for subjects with sensory impairments, especially of the hand and digits, (58,69) it may be that CI therapy benefits this group through the dynamic and intense sensorimotor retraining that is inherent in functional task practice; evidence for this suggestion is limited to a single study (9). To date, the frequency of specific types of upper extremity sensory impairments and the relationship of these impairments to recovery of arm function is not well established for the general stroke population or for a CI therapy eligible population. Visual impairments following stroke may vary and pose an evaluation challenge especially with concomitant cognitive impairments. Hemispatial neglect and visual inattention are controversial in the literature and among clinicians (72). Despite the controversy, visual disturbances of gaze, homonymous hemianopia, and visual inattention within 2 weeks of stroke onset were found to significantly deter recovery of affected arm dexterity at 6 months in a general stroke population (69). Visual inattention, classified as a cognitive deficit, was an exclusion criterion for the 1993 CI therapy trial (7). Only Van der Lee et al. (9) considered hemineglect (with line bisection and letter cancelation) as a potential factor in CI therapy responsiveness. There was a significant interaction between the presence of hemineglect (n ¼ 3) and treatment type (CI therapy or bimanual training), showing an initial greater benefit with CI therapy for those with hemineglect, but the benefit was not maintained. The inconsistencies in definition, measurement, and inclusion of patients with hemineglect or other visual impairments preclude any estimation of the percentage of stroke survivors who may benefit from CI therapy. Two Case Examples: Recently, Bonifer and Anderson (71) reported a single case of a chronic hemiplegic stroke patient who did not achieve the
300
Winstein and Prettyman
minimal motor criteria of wrist and finger extension, but was able to grossly grasp and release a washcloth. The signature CI therapy protocol was used for 15 days during a 3-week period (21 days). Outcomes for this case were not surprising—the patient demonstrated no changes in movement time on the graded WMFT the transient improvement in the amount and quality of movement as indexed with the MAL was not retained. At the 1-month follow-up, subjective MAL ratings were ‘‘rare use’’ for amount (mean 2.0) and ‘‘poor’’ (mean 1.68) for movement quality; at 6-month follow-up, each rating declined to less than ‘‘very rare’’ (mean 0.27) and ‘‘very poor’’ (mean 0.47), respectively, for affected arm use in daily life activities. For this severely impaired patient, there was no benefit from participation in the signature CI therapy protocol. Here, we describe a complimentary case example to highlight the importance of establishing a clinically meaningful change in response to CI therapy in chronic hemiplegia patients who lack the minimal motor capabilities. We illustrate the lesion size and location from the MRI scans to illustrate the importance of residual motor capability. BD was a 25-year-old right-hand dominant male who suffered a left intercerebral hemorrhage with right hemiparesis 11 months prior to screening for CI therapy (Fig. 3—MRI scan showing lesion). At the time of screening, the patient could actively extend his wrist 10 and partially extend the metacarpal and interphalangeal joints, but did not achieve the minimum 10 of active extension for at least two fingers and thumb (previously described as the lower functioning criteria) (12). Despite qualification for exclusion based solely on inadequate voluntary motor control and the unlikely potential benefits of CI therapy in this case, the patient and family chose to participate in a 2-week trial of the signature CI therapy protocol. However, implementation of task-specific training proved to be a challenge for both the patient and therapist. Typical tasks consisted of gross grasp movements, whole arm pushing with assisted placement on surfaces, instead of the usual complex functional hand manipulation skills typical of CI therapy practice. The patient attempted to use the constraint for 90% of waking hours, but family members were needed to assist with personal care; even basic function tasks including eating with a utensil placed in the affected hand were not possible. Baseline pre- and post-CI therapy data are plotted in Fig. 1 (right side). Immediately following the intervention the amount of arm use (MAL) had improved from between ‘very rarely’ and ‘rarely’ (mean 1.64) to just over ‘rarely’ (mean 2.28). At the 6-month followup, the self-rated arm use was between ‘rarely’ and ‘half as much’ as pre-stroke (mean 2.47). Motor impairment and functional outcomes showed similar gains. The FMA score increased 4.5 points (from 17.5 to 22) immediately following CI therapy and continued to increase at the 6-month follow-up (motor score 26). If the mean WMFT movement time is considered the patient slowed movement by 11 sec following intervention, however, the slower mean movement time (pre 6.78 sec, post 17.805 sec) was likely due to an increase in
Therapy for Functional Recovery: Key Ingredients and Mechanisms
301
Figure 3 T1-weighted MRI brain images for case BD showing the large subcortical lesion as the darkened area extending through six slices from superior to inferior. This lesion is including a majority of the corona radiata and internal capsule. (Continued on next page.)
ability to perform three additional items (five items pre, eight items post) attempted at the pretest but timed out beyond the 2 min limit. At 6 months, seven items were possible with a mean movement time of 11.77 sec, indicating retention of movement capability. Taken as a whole, these immediate and 6-month changes indicate some persistent benefit from CI therapy. By simple examination of these change scores, irrespective of the level of performance, the data look impressive, but they are misleading. Instead, by asking if
302
Fgure 3
Winstein and Prettyman
(Continued.)
these changes are clinically meaningful, the answer is very different. It would be difficult to argue that slightly more than rare use (MAL mean 2.47) of the limb for functional tasks in combination with a persistent level of motor impairment (26=66, 40% FMA motor score) and ability to use the limb slowly for a limited number (7=15 WMFT tasks) of tasks represents a clinically meaningful change in arm function. We suggest that these small changes represent the acquisition of specific strategies for the affected limb that were not specific to the CI therapy intervention itself. For example, BD learned how to use his nonaffected hand to wrap his affected hand around a bag of coffee beans, hold it there with his nonaffected hand, and pour from the bag using both hands wrapped around the bag. In essence, BD learned a strategy for assisting his affected limb for this task by using his unaffected limb. This strategy is not consistent with the essential goal of lifting the learned suppression of behavior. Like the first case described above, this case suggests that there is nothing particularly magic about CI therapy; motor and sensory impairment levels are useful gauges for predicting clinically meaningful changes from CI therapy. Once we better understand the electrical, cellular, and neurochemical mechanisms that drive positive plasticity in the injured brain, it might be that CI therapy or its analog combined with a pharmacologic agent or direct cortical stimulation (FDA approved investigational multicenter clinical trial currently underway and sponsored by Northstar Neuroscience, Inc.) could provide the additional boost for a meaningful therapeutic effect in these more severely involved patients, but for now this severely impaired group represents the limit of neural machinery that is required for behavioral engines such as CI therapy.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
303
2.2.4. Question 4: Is There a Fundamental Difference Between CI Therapy and a Simple Manipulation of the Intensity and Dose of Task-Specific Training? In Other Words, Has the Rehabilitation Community Rediscovered the Well-Known Power Law of Learning Under a New and Improved Name of CI Therapy? The intensity and duration of the signature protocol of CI therapy includes 60 hr of one-on-one coached task-specific training in the laboratory (6 hr=day), delivered over a 14-day period, with outside laboratory constraint-induced use encouraged (restraint device worn on other limb for 90% waking hours and according to the behavioral contract) by home task practice (homework assignments) and caregiver buy-in (behavioral contract for patient and caregiver). This degree of intensity and duration represents a dosage of therapy significantly greater than anything that is available through standard therapy protocols delivered either in inpatient or outpatient settings (73–75). The choice of the control treatment becomes a key factor for understanding the locus of any benefit. In one recent review, Van der Lee (56) highlighted this issue by asking, ‘‘Is CI therapy different or simply more of the same?’’ Numerous intervention studies using a varied set of therapeutic media applied across different time points after stroke have generally documented improvements in function (7–9,35,76–83). These interventions vary in dosage and contents that range from electromyography-triggered neuromuscular electrical stimulation (81) to repetitive force production against resistance (79). More clinical research is needed to determine if there are any differences in outcomes from these various approaches (e.g., forced-use, EMG-triggered electrical stimulation, resistance strength training, bilateral arm training). A task-oriented approach may represent the common element between CI therapy and many of these other effective interventions. For example, Kwakkel et al. (84) found that an additional 30 min of task oriented therapy daily for 5 days a week for up to 20 weeks can make a substantial difference in functional recovery. Similar benefits of a task-oriented approach have been shown in studies focused on mobility (78) and upper extremity function applied during the immediate postacute phase (62). In her 1997 review, Duncan (85) suggested that the CI therapy intervention was simply a proxy for more task-specific training. The question we raise here is twofold. First, if any of these other effective interventions were applied at the same dosage (intensity and duration) as the signature CI therapy protocol, would the outcome be any different? Answering this particular question would allow a determination of whether any treatment effect was specific to CI therapy or a less specific treatment effect that simply validates the time-honored principle of practice illustrated by the well-known power law of learning (86). Indeed, Taub and Uswatte
304
Winstein and Prettyman
(87) have recently conferred that ‘‘the primary difference between CI therapy and conventional PT is in the duration and intensity of the treatment . . . . It is not the constraint in CI therapy that is important; it is the intensive practice of correct use of the more affected extremity.’’ (p. 34,37). Second, if functional task specificity is the critical ingredient that is shared between CI therapy and many of the other effective therapeutic interventions, would the outcome be different for nonfunctional task practice protocols? In a recent commentary, Page (88) suggested that therapy that was more task-specific than usual care, while provided within the typical contact time parameters (30–45 min sessions), could be more efficacious than more traditional motor rehabilitative approaches. Recent findings from a pilot phase II randomized-clinical trial support the importance of functional task-specific training for more efficacious outcomes in stroke. We found that only 20 hr of upper extremity-specific therapy (1 hr=day) delivered over a 4–6-week period, significantly affected functional outcomes. Although the immediate benefit of a therapy protocol specific for functional tasks was similar to that of a strength and motor control protocol, and both were superior to standard care immediately post-treatment, the functionaltask practice protocol retained more substantive benefits 9 months after stroke (62). 2.2.5. Question 5: How Critical Is the Specified Protocol for Success? The signature protocol for CI therapy of upper extremity paresis in stroke rehabilitation is being used for the EXCITE multisite RCT (18). It consists of a 14-day period (two full weeks) in which the participant attends an intense daily training session (6 hr=day) during the weekdays in an outpatient clinic setting. In addition to the 10 days of daily training, the participant is instructed to wear a constraint device (soft mitt) on the less affected side for a target of 90% of waking hours. The mitt can be removed during any water activities (e.g., bathing, washing) or for safety reasons, but must be worn during the daily training session in the evenings after training, in the mornings before training, and on weekend days during the 14-day period. The interventional training has been described elsewhere in detail (10); here, we highlight the two distinct training procedures that have been employed in previous research. Shaping or adaptive task practice (ATP) and standard task practice are used with each participant as functional activities (e.g., writing, turning pages, rotation of rolodex file) are repeated. ‘Shaping’ derives its name from the behavioral training technique developed by Skinner (89) and others (90). With this Skinnerian framework, the learner is relatively passive while performance is progressively ‘‘shaped’’ as the behavioral objective (goal) is approached in small steps through reinforcement or reward (positive feedback). Using this method, a motor
Therapy for Functional Recovery: Key Ingredients and Mechanisms
305
objective (task goal) is approached in small steps by successive approximations (i.e., parts of the task), the task is made more difficult in accordance with the participant’s motor capabilities, or the speed of performance is progressively increased. Each functional activity is practiced for a set of 10 trials, and explicit feedback is provided regarding the participant’s performance with each trial (18). An open question remains a point of debate concerning the degree to which ‘‘shaping’’ clearly describes the fundamental training procedure that is used with CI therapy. The debate is focused on defining the goal of the training. For the behavioral psychologists, the goal is to modify behavior; by contrast, for the rehabilitation practitioner, the goal is to improve skill in performing functional motor tasks with the affected limb. Indeed, Morris et al. (10) acknowledge that ‘‘shaping is very similar to training techniques commonly used by physical and occupational therapists giving patients task practice. The main difference is that with shaping, patients or subjects are given explicit feedback concerning even small improvements in performance,’’ (p. 39) and errors are ignored. Specific training protocols have evolved from earlier descriptions that emphasized an operant conditioning approach to training (13) to more recent descriptions, ‘‘CI therapy is typically administered on a one-on-one basis by a therapist who continuously monitors patient performance, provides positive reinforcement to the patient, and shapes the difficulty of the task upwards’’ (p. 38) (87). The fundamental difference in approach that depends on the goal concerns the nature and schedule of feedback. If the goal is to modify behavior, feedback is more reward-like and less informational in content; it is provided frequently and immediately, and it does not include error information. In sharp contrast, if the goal is to enhance skill learning, feedback is more informational than reward-like in content; it is provided less frequently with practice and with some delay, and it provides information on errors to enhance motor learning (37). To our knowledge, there has not been any systematic direct comparison of the relative effectiveness of these two fundamentally different approaches for training–feedback protocols applied to stroke rehabilitation. The other form of practice described for CI therapy protocols—standard ‘‘task practice’’—is less structured than ‘‘shaping’’ for example, the tasks are not set up to be carried out as individual trials of discrete movements; but involve functionally based activities performed continuously for a period of 15–20 min (e.g., setting the table for lunch). In successive periods of task practice, the spatial requirements of the activity or other parameters (such as duration) can be changed to require more demanding control of limb segments for task completion. Feedback about overall performance is provided at the end of the 15–20 min period. Several investigators have tested modified regimens of the signature protocol that include changing the intensity, the duration, the setting,
306
Winstein and Prettyman
and even the specific components of CI therapy (9,19,20,44,91,92). Sterr et al. (91) directly compared the effectiveness of CI therapy that used ‘‘shaping training’’ delivered 3 hr=day for 14 consecutive days to 6 hr=day for the same training period. They found that the 3-hr regimen of CI therapy significantly improved motor function in chronic hemiparesis, but it was less effective than the 6-hr training schedule. These results are limited, however, by a small sample (7–8=group) and a relatively short follow-up period (1 month). In general, for most of the studies that used a modified regimen of CI therapy, some benefit was achieved (9,20,44,91,92). For example, 18 patients participated in a traditional outpatient neurological rehabilitation program that included a program of home forced-use. Each patient (17 patients with chronic stroke, 1 with subacute stroke) completed an individualized program comprised of seven 2-hr treatment sessions (1 hr of occupational therapy and 1 hr of physical therapy, approximately every other day) over a 2–3-week period and included instruction on the use of a restraining mitt at home and in the outpatient setting during functional activities (92). Significant improvements in impairment and function from baseline to posttreatment were reported for 12 of the 17 items from the Wolf motor function test. However, without a direct comparison of the effectiveness of these modified regimens to that of the signature protocol, it is difficult to ascertain just how critical the precise intensity, duration, and target constraint elements are to the success of the CI therapy intervention. The ‘‘massed practice’’ element was targeted as the critical ingredient that could explain the relatively small treatment effect reported by Van der Lee et al. (9) who compared the effects of CI therapy with bimanual neurodevelopmental treatment (NDT) (49). A much larger treatment effect was reported by Taub et al. (7) who compared the magnitude of change (pre to post) between treatment and control groups. One plausible explanation for the smaller between groups relative change found by Van der Lee et al. (compared to Taub et al. (7)), could be due to differences in the intensity of training administered in each study (49). Another equally plausible explanation for the smaller between group differences (effect size) is associated with differences in the comparison group that was used in each study (bimanual NDT vs. attention control), and the starting point defined as the initial degree of learned nonuse (MAL upper cut-off 2.5 vs. 3.0). In essence, if the comparison group also improves and the patients enrolled in the study do not suffer importantly from learned nonuse (higher MAL), the treatment effect size (comparison of change scores between groups) for the intervention will be smaller than if the comparison group does not change and the potential range of change on the MAL is larger (i.e., low pretreatment MAL). Thus, effect size alone can be misleading unless you know the baseline level for each group, and the degree to which the control or comparison group also changes. In the Van der Lee et al. case, the bimanual NDT group
Therapy for Functional Recovery: Key Ingredients and Mechanisms
307
received the same intensity of training as the CI therapy group, and they demonstrated a substantial change in their performance between the pre- and posttests. This is an important finding because it suggests that these two apparently disparate therapeutic protocols might share some common elements that are beneficial for recovery of upper extremity function in the chronic stage after stroke. Altogether, it is difficult to determine just how critical the exact protocol is for success without the more controlled and proper comparison studies. There is an emerging literature (see earlier Question 5 discussion) which suggests that the critical ingredient of training regimens to increase the use and function of the more affected limbs of stroke patients may be task specificity and not necessarily intensity of training or even the use of a constraint (45,62,88). Intensity (training epoch=duration) that is defined with the goal to train functional skills may be quite different from that typically used to modify behaviors. The signature CI therapy protocol clearly geminates from the desire to modify behavior and may not be optimal for skill acquisition. 2.2.6. Question 6: What Are the Best Predictors (Inclusion=Exclusion Criteria) of Success with CI Therapy and Why? Popular news media reports and radio broadcasts report the effectiveness of CI therapy as a ‘‘new breakthrough’’ but rarely provide the caveat that this approach is not appropriate for every stroke, or traumatic brain injured victim who has a paretic limb. With the exception of only one published case report (71), upper extremity CI therapy trials have enrolled patients with very specific motor capabilities of the affected extremity (7–9,19). Twenty degrees of active wrist extension and 10 of finger extension (metacarpal and interphalangeal) were the inclusion criteria used in two phase II randomized-clinical trials with chronic stroke patients (7,9). Likewise, the first trial of affected arm ‘‘forced-use’’ with constraint of the less affected arm (8) and clinical reports without control groups included active wrist and finger extension criteria (93–95). Thus to date, for subjects with upper extremity paresis following cerebral injury, the ability to actively recruit muscles for wrist and finger extension movements constitutes an important predictor for successful outcomes with CI therapy. A second and corollary inclusion criteria and presumptive predictor of success are the upper cut-off for intake of 2.5 (less than half as much use of the more impaired extremity as before the stroke) on the AOU scale (range ¼ 0–5) of MAL (7). This together with the above movement capability clearly delineates a population of subjects who have movement capability (active voluntary motion of the wrist and finger), but are not using that capability in an automatic and functional manner. Indeed, Taub argued that ‘‘patients with scores greater than 2.5 cannot be said to suffer importantly
308
Winstein and Prettyman
from ‘learned nonuse’’’ (49). In our view, this population is one with adequate physiological recovery to respond very well to a protocol of taskspecific practice, and at the same time, they lack the intrinsic drive or knowledge that they could make functional gains with practice (see case below). This discrepancy between ‘‘can do’’ and ‘‘actually does’’ is what characterizes the ‘‘learned nonuse’’ syndrome (47,48). A third predictor is cognitive or mental function; this criteria is generally not mentioned in the popular news media reports, but it is described in the scientific literature as an important inclusion criterion. Chronic and subacute patients are routinely screened for cognitive ability and apparently understand consent procedures for clinical trial participation. Previous RCT investigations have used the mini mental state exam (7,9,18,91), NIH Stroke Scale (19), or unspecified cognitive performance measures (93,96) to exclude patients with significant cognitive deficit from participation in CI therapy trials. Based on the cognitive prerequisite for participation, executive function is another likely and important predictor of responsiveness to CI therapy, particularly given the level of active participation, compliance, and accountability required for the protocol. In addition to the cognitive measures, patients enrolled in CI therapy clinical trials must undergo the informed consent procedure. Those involved in research with human subjects are acutely aware of consent requirements for answering questions and providing explanations of trial purpose, procedures, and risks and benefits. Included in the explanation of procedures are instructions for application of the constraint device to the less affected extremity, intensity of training, measurements, and risks of frustration, falls, and fatigue. Cognitive appraisal and motivation are inherent characteristics of the consent process that may be important predictors for CI therapy effectiveness. In contrast to cognitive function measures, formal behavioral assessments including motivation, self-efficacy, and expected outcomes are not routinely included in CI therapy trials; however, they may provide important adjunct or corollary information associated with predictors of success (97). Given that a critical inclusion criterion and predictor of success is some degree of physiological recovery of wrist and finger extension movement, it follows that the neural substrate that mediates that voluntary movement (i.e., corticospinal tract and cortical sensorimotor hand area) should also predict responsiveness to CI therapy. For this reason, we summarize the studies from humans with stroke that have examined the relationship between lesion characteristics (size and location), functional anatomy, and behavioral outcomes specific to upper limb voluntary movement recovery (98–110). Generally, these studies differed widely with respect to critical dimensions including patient chronicity (acute, chronic), imaging method (e.g., MR, PET, TMS), outcome measures (e.g., paresis, coordination, dexterity, and self-assessments), and design (cross-section, longitudinal).
Therapy for Functional Recovery: Key Ingredients and Mechanisms
309
However, in spite of these differences, several common elements emerged that provide some guidance regarding the physiological integrity of the voluntary motor system and its relationship to the minimum motor criteria for CI therapy protocols. To date, no study has directly examined the lesion characteristics (location and size) of subjects who have participated in a trial of CI therapy to determine if there is a correlation between lesion characteristics and efficacy suggestive of a valid predictor. Several studies have documented the functional changes in activation for specific motor tasks using fMRI (see Sec. 2 for discussion) and some have documented the lesion characteristics for subject demographic information, but none have directly correlated the lesion size or location with the reorganized patterns of activity associated with CI therapy, e.g., Ref. 111. The presumption is that for the most part, if the subjects meet the clinical criteria for inclusion, they have a lesion that is relatively small, and largely spares the corticospinal projections (i.e., posterior limb of the internal capsule) critical for upper limb and hand function. For example, one study examined the correlation between motor improvements and altered fMRI activity after 2 weeks of home-based CI therapy in seven participants whose lesion volumes did not include the hand area of the primary motor cortex or the region of dorsal premotor cortex anterior to the primary hand area (111). Not surprisingly, many investigators have suggested that lesion volume, alone, is a poor predictor of upper limb or hand recovery and can be misleading, particularly if considered in isolation (99,100,103,112,113). Instead, multivariate approaches that use logistic regression or multiple regression analyses with several variables have been generally more informative. For example, Feys et al. (101) used multiple regression with a sample of 100 stroke patients who had obvious motor deficits of the upper limb to identify which clinical variables assessed at different time points (2–5-weeks poststroke, 2, 6, and 12 months post) were predictive of motor recovery. Unfortunately, no measures of lesion characteristics were included with this large sample. However, consistent with the minimum motor criteria, in the multiple regression analyses, upper limb motor performance was retained as the predictive factor with the highest R-square (101). In a second study, this same group showed that neurophysiological measures including somatosensory- and motor-evoked potentials from TMS together with clinical variables improved the predictive accuracy of arm recovery after stroke (102). In a comprehensive study of recovery in 52 hemiparetic patients that used clinical and magnetic resonance-morphometric measures, lesion size was not correlated with outcome, but the amount of spared residual function appeared as a major determinant for the capacity for motor recovery (100). In an earlier study, Binkofski et al. (99) studied remote metabolic depressions and pyramidal tract involvement in 23 patients with acute ischemic stroke. Using measures of regional cerebral glucose metabolism (rCMRGlu) with PET, lesion location and spatial extent with MRI and
310
Winstein and Prettyman
measures of motor impairment during the acute and chronic states with motor scores and magnetic-evoked motor potentials, they found that neither lesion size nor spatial extent of the lesion and remote rCMRGlu depressions outside the thalamus correlated with improved motor score. Instead, the preservation of parts of the pyramidal tract (from MRI analysis) and thalamic circuitry were found to be major determinants for the quality of hand motor recovery. With a newer method of three-dimensional anisotrophy contrast (3D-AC), Wallerian degeneration in the corticospinal tracts of patients who sustained a stroke can be detected within 2–3 weeks after stroke onset. Using this method, Watanabe et al. (107) correlated the patients’ motor functions using the Brunnstrom criteria at 12 weeks after stroke to the magnitude of Wallerian degeneration detected with 3D-AC. Patients with poor recovery (Brunnstrom stage I–IV) showed Wallerian degeneration, while the patients with good recovery (Brunnstrom stages V–VI) showed no evidence of Wallerian degeneration. Consistent with the results of this more advanced method, an earlier study employed an imaging method that used CT images and correlated recovery of voluntary movement in chronic hemiparetic patients (n ¼ 89) with degenerative shrinkage of the cerebral peduncles. They found that recovery was incomplete in patients whose cerebral peduncles were less than 60% normal size; in contrast, when more than 60% of the cerebral peduncles were spared, the degree of recovery from motor weakness was positively correlated with the quantity of intact fibers spared (106). Taken together and using a diverse set of methods, the majority of the studies reviewed show convergent findings suggesting that the degree of preservation of the corticospinal tract is a good predictor of arm and hand recovery. Studies that have assessed the effects of stroke involvement of primary and secondary cortical motor areas (primary motor cortex, premotor areas, supplementary motor area, corona radiata, internal capsule, basal ganglia and thalamus) on the probability of recovery of isolated upper limb movements generally agree that measures of the posterior limb of the internal capsule, representing the convergence of corticofugal motor efferents, provide the strongest predictor (98,99,108,110,112,113).
3.
TWO COMPLEMENTARY MECHANISMS UNDERLYING CI THERAPY: WHAT IS THE EVIDENCE?
Our second major section provides a review of the evidence for two complementary mechanisms for the effects of CI therapy, one of which ties this work into the primary theme of this book, neural plasticity, and the other, learned nonuse, provides a behavioral explanation for the compensatory strategy thought to be triggered after brain damage. We provide a brief synopsis of the evidence for each of these mechanisms while highlighting the potential interaction between the two.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
3.1.
311
Role of Neuroplasticity in Recovery of Function After Brain Injury
The ability of neurons to adapt and form new synaptic circuits in response to changing environmental cues and behavioral experience is a common feature of both the developing and adult nervous systems (114,115). This adaptive process, generally referred to as neural plasticity, is most obvious in the developing brain and is essential to the formation of the wide network of limbic, striatal, and cortical circuits that regulate the complex array of human behaviors. Subsequent experiences during adulthood continue to modify these circuits as new skills and memories are acquired, and these adaptive mechanisms are considered essential to compensate for the agerelated cell loss that occurs in some regions of the brain (116). Likewise, when the brain is damaged by trauma or disease, these compensatory repair mechanisms are brought into play and are critical for determining the limits of functional recovery that can be expected after brain injury (41,117,118). In particular, previous studies suggest that cortical neurons either adjacent or contralateral to the damaged area can compensate and form new synaptic contacts that can take over the function of the necrotic area (105,119–123). In addition, it is well known that these adaptive processes are not random, but are governed by local neurotrophin and intracellular signaling pathways that can be manipulated by various experimental paradigms including exercise (124,125), motor skills training (126,127) complex social environments (128), and pharmacological treatment with monoamine affecting drugs (129–132). 3.2.
Influence of Behavioral Experience on Neural Plasticity and Recovery of Function
The influence of behavioral experience on neural plasticity and recovery of function in the adult mammalian brain has received increased attention in recent years. Previous studies have found that neocortical damage results in significant morphological and physiological changes in remaining regions of the cortex and underlying striatum of both hemispheres, and, that compared with the intact brain, the damaged brain may make connected cortical regions especially sensitive to behavioral interventions that can influence recovery of function. In animal models, unilateral damage to the forelimb area of the rat sensorimotor cortex leads to chronic deficits in somatosensory function and preferential use of the forelimb ipsilateral to the damage for movements that can be easily quantified by behavioral testing (133–136). In addition, changes in motor function are accompanied by compensatory changes in dendritic morphology and synapse density in pyramidal neurons of the contralateral sensorimotor cortex that are dependent, in part, on forelimb behavior. Previous studies have found that motor skills learning, acrobatic training, and enriched environments can enhance dendritic
312
Winstein and Prettyman
arborization and synaptogenesis after cortical injury (137,138), while immobilization or forced-overuse of the unaffected forelimb during the period of structural change (but not afterwards) can exaggerate the primary injury, inhibit compensatory changes normally seen in the contralateral homotopic cortex (134), and lead to more severe and enduring deficits in motor function (139). Although the cellular mechanisms that underlie the influence of behavior on neural plasticity are still unclear, previous studies have found that ischemic damage to the cortex leads to the upregulation of a variety of neuronal and glial genes and proteins that have been linked to the compensatory plasticity responsible for functional recovery. These include: immediate early genes and transcription factors, growth-associated proteins, neurotrophic factors, neurotransmitter receptors, glial proteins, and adhesion molecules (140,141). In addition, since most of these molecules have a well-documented function in learning processes and in plastic changes that occur in a use-dependent manner, it is reasonable to suggest that the basic mechanisms involved in lesion-induced cortical plasticity are similar to those found in the experience-induced plasticity in the noninjured brain. If so, the use of task-specific behavioral training provides a unique opportunity to shape the compensatory changes that underlie the degree and rate of functional recovery that can be achieved after brain injury (135). However, controlled studies examining how the timing, type, and intensity of such treatments affect the effectiveness of task-specific training are clearly needed before the optimal rehabilitation schedule for stroke survivors can be determined. 3.3.
Use-Dependent Cortical Reorganization and CI Therapy
During the past decade, considerable effort has been made toward establishing novel approaches to overcome the compromised ability of stroke patients to use their upper extremities to carryout activities of daily living (5,6). Within the neurological rehabilitation community, this effort has been fueled, largely, by the desire to harness the plastic, adaptive properties of neuronal networks in the adult brain through therapies that engage the patient in active voluntary use of the impaired limbs. With the rapid development of neuroimaging techniques available for human investigations, hypotheses based on basic neurobiological research in clinical practice can be tested. Reorganization of the motor system after stroke as assessed by various neural imaging methods including TMS, PET, fMRI, and MRI techniques is an intriguing but challenging new area of research (142–144). The signature CI therapy protocol emphasizes repetitive use of the impaired limb using task-specific training of functional movements, while restricting movement of the better limb. The idea is that the intensive task practice drives an adaptive response that manifests similarly to that invoked with motor learning and shown in humans (145–148) using fMRI and TMS
Therapy for Functional Recovery: Key Ingredients and Mechanisms
313
methodologies. However, after a brain injury, these practice-dependent changes associated with motor skill learning proceed in the context of the brain injury-induced adaptive response. Nudo et al. (38) describe the interactive effect of the injury and the practice-induced cortical reorganization. Several investigators have shown that without task-specific practice applied to the affected limb, the injury-induced plasticity alone results in maladaptive or suboptimal behaviors including nonuse, misuse (poor coordination), and self-imposed impoverishment (40,134,149). These behavioral manifestations are thought to be the direct result of a combination of effects including: the neuronal degeneration at the site of injury, disruption of remote connections from the injury site, and the self-imposed impoverishment that derives from the adoption of compensatory behaviors. Together, the changes induced through practice-induced skill acquisition and the neurological injury can affect the use of the affected limb that has a pronounced effect on the resultant reorganization of the undamaged neuronal networks for movement (14,40). It is the interaction between the injury-induced reorganization and the therapy-induced reorganization that is thought to provide for a more optimal and functional recovery (see Fig. 4). One hypothesis that derives from this perspective is that the extent of ‘‘functional recovery’’ from stroke is tightly linked to the extent and nature of cortical remodeling that occurs after brain injury (105,111,119,150–157). More importantly, variable approaches to motor training of the paretic arm are thought to stimulate
Figure 4 General neurobehavioral model of postneural lesion activity-dependent brain plasticity with two divergent paths to recovery: compensatory (negativemaladaptive) or functional (positive-adaptive).
314
Winstein and Prettyman
use-dependent cortical reorganization essential for determining the amount of functional recovery that can be expected in stroke survivors. The evidence from human studies in support of this hypothesis is limited at this time, but, nonetheless, promising (94–96,111,158–164). For example, Johansen-Berg et al. (111) conducted a longitudinal study with repeated fMRI to determine the changes in brain activity that are associated with improved function in a group of seven patients whose chronicity ranged between 6- and 84-month poststroke. The group consisted of patients with mild to moderate hemiparesis, whose stroke lesions were mostly cortical (one subcortical) and ranged in size from less than 0.1 cm3 to 230 cm3. Subjects were scanned while performing a hand flexion–extension movement before and after a 2-week home-based therapy program combining restraint of the unaffected limb with progressive exercises for the affected limb that focused on improved hand and arm movement skill. Patients were instructed to complete a 30-min graded exercise program with the affected arm twice daily for the 2-week period. This protocol is considerably less intense than the signature CI therapy protocol, but, nonetheless, a majority of the subjects showed improved performance on the upper limb section of the motricity index, Jebsen arm test, and grip strength. As expected, there was considerable variability between subjects in the degree of performance improvement. For example, the percent grip strength change for the affected hand ranged from –23.1% to 88.7% change (six of the seven subjects showed greater post-training grip strength compared with pretraining scores). Of particular interest, were the therapy-related changes in brain activity for the affected hand movements. The areas where changes in grip strength ratio after therapy correlated with increased activation during the fMRI were the cerebellum (crux I and lobule VI, bilaterally), secondary somatosensory cortex (parietal), and the premotor cortex. There are clearly methodological concerns with this kind of research in which serial imaging is performed to derive the therapy-related changes in cortical reorganization. A critical control is task performance. If task performance is well controlled and reliable through consistent rate and force parameters, it allows training-induced changes to be identified separately from the well-known task-performance effects such as force magnitude (153). In spite of these constraints, the fMRI methodology allows an assessment of changes across the whole brain and with a spatial resolution that is not possible with TMS or EEG techniques. Several studies have used TMS mapping to identify changes in cortical representations following CI therapy (94,96,158,162). Generally, these studies have reported changes in the excitable surface area derived from TMS motor maps (i.e., an expansion of the excitable surface area after therapy). In a recent review, Wolf et al. (16) provide an alternative explanation for the observed changes in map area; ‘‘changes in the surface area may represent increased excitability to current spread, without reflecting a true expansion or contraction of the cortical
Therapy for Functional Recovery: Key Ingredients and Mechanisms
315
representation area (pp 334).’’ A recent review of the TMS methodology to assess cortical plasticity for stroke rehabilitation suggests that the data derived from these studies may be profoundly over-interpreted (165). A similar concern exists for the interpretation of changes in brain activation with fMRI methods. Are the changes in cortical activation after therapy associated with the newly acquired skill (motor learning) or with other more general cognitive processes such as the focus of attention? (166). We suggest that through the combined use of fMRI and TMS, advances can be made in our understanding of the mechanisms underlying therapy-induced cortical reorganization associated with functional recovery. Hypotheses derived from fMRI findings for the putative role of specific cortical regions or networks can be tested using TMS methods (e.g., virtual lesion) to provide insight into the physiological mechanisms mediated by motor training (167,168). Taken together, the evidence is limited that use-dependent cortical reorganization is provoked by the CI therapy protocol. Rather, there is compelling evidence to suggest that such changes in cortical reorganization accompany most forms of motor learning and particularly the learning that is induced through functional task-specific practice both with and without attention (implicit procedural learning without attention). Thus, to conclude that use-dependent cortical reorganization (following CI therapy) provides a mechanism underlying the effects of CI therapy may be missing a more overarching and important concept. The CI therapy is designed to harness the important ingredients that are known to promote plasticity and functional change—voluntary, repetitive, challenging, and functional task practice. Friel et al. (169) demonstrated that the critical postlesion experience associated with areal changes in the hand area of primary motor cortex was the daily progressive and repetitive training after the infarct in primates and not simply the restraint of the unimpaired hand. In some cases, the constraint may lead to detrimental effects, especially in the early stages after injury (170). 3.4.
The Reversal of Learned Nonuse as a Mechanism for CI Therapy
The reversal of ‘‘learned nonuse’’ has been suggested as an explanation for at least part of the functional improvement reported with CI therapy in humans and animal models (7,48,170–172). The motivation to use the impaired limb may be reduced initially because of early failures or fatigue in that limb and a compensatory increased reliance on the ‘‘working’’ extremity. As these maladaptive compensatory strategies emerge, a self-selected failure to engage the available but underused sensorimotor systems becomes more likely. The stroke-induced motor impairments of hemiparesis may contribute to self-imposed behavioral impoverishment, leading to a reduced or suboptimal plasticity response (141).
316
Winstein and Prettyman
The behavioral mechanism underlying the development of learned nonuse and its reversal with therapy has been described in the context of operant conditioning i.e., punishment and reward (13,173,174). However, an alternative explanation that is consistent with current models of injuryinduced functional outcomes from stroke (135) posits that the self-imposed compensatory strategy that favors the ‘‘good’’ limb emerges from an interaction between the injury-induced cortical remodeling and stroke-induced motor impairments of hemiparesis. An intriguing report describes a putative area in the cingulate motor area (CMAr) buried in the ventral bank of the cingulate sulcus, just below the presupplementary motor area that is engaged in motor intention of the contralateral upper limb in humans (175). In this single-subject clinical case, direct, low intensity stimulation of the right CMAr incited the patient to act, and resulted in an irresistible urge to grasp something. After the patient had visually localized a potential target, her left hand moved toward the object to grasp it, as if mimicking a spontaneous movement. This irrepressible need started and ended with the stimulation, and the patient was unable to control it. Therefore, a potential neurophysiological explanation for nonuse might involve a reduced output from the intention to move area, CMAr. It is possible that with a cortical or subcortical capsular stroke, the CMArs are functionally disconnected from the putative motor systems for voluntary actions. Such a network malfunction might increase the probability for the development of nonuse of the contralateral limb. There is considerable variability in the incidence of learned nonuse among the human stroke population. Fifteen percent of the patient contacts for the EXCITE clinical trial scored greater than 2.5 on the MAL AOU scale within 9 months of stroke onset, demonstrating that they exceeded the criteria for ‘‘learned nonuse’’ (18). Most likely ‘‘learned nonuse’’ is a multivariate phenomenon, with several suspect mechanisms including: operant conditioning, injury-induced neuromodulation of neurotrophic factors, sensory or neglect-related processes, and the interaction between post-traumatic neural events and behavioral experience. Given that plasticity mechanisms are particularly active during time-windows early after focal cortical damage, longitudinal studies of the physiological changes in cortical function (e.g., TMS, fMRI, and PET) beginning within the first few days and weeks of stroke onset will likely provide information on the relationship between the early behavioral suppressions of arm use associated with the dynamic neuroplasticity response. Critical timing of effective therapeutic interventions that address the expectations for functional use during the early ‘‘cortical shock’’ period should be tested. There are several interventions described in the literature that could be used during this early postacute phase such as progressive mental practice of specific and familiar tasks through guided imagery (176–178) or customized virtual environments; EMG-triggered neuromuscular electrical stimulation
Therapy for Functional Recovery: Key Ingredients and Mechanisms
317
(81), bilateral arm training (83), or functional task practice (62,88), to name a few. These approaches to enhance movement capability might be useful as a cost effective means to prevent long-term suboptimal behavioral compensation and to prevent or retard the detrimental effects of disuse. There is really nothing in the rehabilitation literature to suggest that CI therapy contains unique elements that are better suited to shape the appropriate neuroplastic changes associated with functional improvements than a varied set of novel or even standard interventions; all of which aim to promote practice of task-specific skills using progressive amounts of voluntary control in a challenging, motivating, and meaningful context, and with at least a moderate degree of intensity. Because neural events in the injured hemisphere can be affected by behavior differently than the neural events in the noninjured hemisphere, different therapeutic strategies might well be used on opposing limbs at different times after unilateral sensorimotor cortex injury to optimize recovery. From this perspective, avoiding long durations of constraint of the nonimpaired limb, especially in the early stages postinjury, is recommended (170). We conclude by proposing a general model (Fig. 4) of postneural damage activity-dependent plasticity that describes two divergent behavioral paths termed ‘‘optimal function’’ and ‘‘compensatory function’’ for the injury-induced plasticity response. We suggest that CI therapy is but one example of a larger class of therapeutic interventions that, when used alone or in combination with certain pharmacological agents or even direct cortical stimulation, may have the potential to drive the system toward the ‘‘functional’’ path of recovery by integrating the rewired motor representations (procedural skills) into the reorganized neural circuits. REFERENCES 1.
2.
3.
4.
5.
American Heart Association and American Stroke Association, Heart Disease and Stroke Statistics. 2003 Update. 2003, American Heart Association and American Stroke Association. Broderick J, Brott T, Kothari R, Miller R, Khoury J, Pancioli A. The greater Cincinnati=Northern Kentucky stroke study: preliminary first-ever and total incidence rates of stroke among blacks. Stroke 1998; 29:415–421. Davis P, Taylor TN, Torner JC. Projected number of strokes in the year 2050 in the United States. In: American Heart Association’s 23rd International Joint Conference on Stroke and Cerebral Circulation. Davis, CA: Stroke, 1998. Barbeau H, Norman K, Fung J, Visintin M, Ladouceur M. Does rehabilitation play a role in the recovery of walking in neurological popluations? Ann NY Acad Sci 1998; 860:377–392 Woldang H, Hummelsheim H. Evidence-based physiotherapeutic concepts for improving arm and hand function in stroke patients. J Neurol 2002; 249: 518–528.
318
Winstein and Prettyman
6.
Van der Lee J, Snels IAK, Beckerman H, Lankhorst GJ, Wagenaar RC, Bouter LM. Exercise therapy for arm function in stroke patients: a systematic review of randomized controlled trials. Clin Rehabil. 2001; 15:20–31. Taub E, Miller NE, Novack TA, Cook EW III, Fleming WC, Nepomuceno CS, Connell JS, Crago JE. Technique to improve chronic motor deficit after stroke. Arch Phys Med Rehabil 1993; 74:347–354. Wolf S, Lecraw DE, Barton LA, Jann BB. Forced use of hemiplegic upper extremities to reverse the effect of learned nonuse among chronic stroke and head-injured patients. Exp Neurol 1989; 104:104–132. Van der Lee J, Wagenaar RC, Lankhorst G, Vogelaar TW, Deville WL, Bouter LM. Forced use of the upper extremity in chronic stroke patients: results from a single-blind randomized clinical trial. Stroke 1999; 30:2369–2375. Morris D, Crago J, Deluca S, Pidikiti R, Taub E. Constraint-induced (CI) movement therapy for motor recovery after stroke. NeuroRehabilitation 1997; 9:29–43. Taub E, Wolf SL. Constraint-induced (CI) movement techniques to facilitate upper extremity use in stroke patients. Top Stroke Rehabil 1997; 3:38–61. Taub E, Uswatte G, Pidikiti RD. Contstraint-induced (CI) movement therapy: a new family of techniques with broad appliation to physical rehabilitation — a clinical review. J Rehabil Res Dev 1999; 36:237–251. Taub E, Crago JE, Burgio LD, Groomes TE, Cook EW. An operant approach to overcoming learned nonuse after CNS damage in monkeys and man: the role of shaping. J Exp Anal Beh 1994; 61:281–293. Plautz E, Milliken GW, Nudo RJ. Effects of repetitive motor training on movement representations in adult squirrel monkeys: role of use versus learning. Neurobiol Learn Mem 2000; 74:27–55. Winstein C, Wing AM, Whitall J. Motor control and learning principles for rehabilitation of upper limb movements after brain injury. In: Grafman J, Robertson I, eds. Handbook of Neurospychology. Amsterdam: Elsevier Science BV, 2003:77–137. Wolf S, Blanton S, Baer H, Breshears J, Butler AJ. Repetitive task practice: a critical review of constraint induced therapy in stroke. The Neurologist 2002; 8:325–338. Taub E, Uswatte G, Morris DM. Improved motor recovery after stroke and massive cortical reorganization following constraint-induced movement therapy. Phys Med Rehabil Clin North Am 2003; 14:S77–S91. Winstein C, Miller JP, Blanton S, Morris D, Uswatte G, Taub E, Nichols D, Wolf S. Methods for a multi-site ransomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke. Neurorehabil Neural Repair 2003; 17:137–152. Dromerick A, Edwards DF, Hahn M. Does the application of constraintinduced movement therapy during acute rehabilitation arm impairment after ischemic stroke? Stroke 2000; 31:2984–2988 Page S, Sisto S, Johnston MV, Levine P, Hughes M. Modified constraintinduced therapy in subacute stroke: a case report. Arch Phys Med Rehabil 2002; 83:286–290.
7.
8.
9.
10.
11. 12.
13.
14.
15.
16.
17.
18.
19.
20.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
319
21. Blanton S, Wolf SL. An application of upper-extremity constraint-induced movement therapy in a patient with subacute stroke. Phys Ther 1999; 79:847–853. 22. Biernaskie J, Chernenko G, Corbett D. Efficacy of rehabilitative experience declines with time after focal ischemic brain injury. J Neurosci 2004; 24(5):1245–1254. 23. Van der Lee J. Constraint-induced therapy: some thoughts about theories and evidence. J Rehabil Med 2003; 41(suppl):41–45. 24. Fisher B, Sullivan KJ. Activity-dependent factors affecting poststroke functional outcomes. Top Stroke Rehabil 2001; 8:31–44. 25. Mott PW, Sherrington CS. Experiments upon the influence of sensory nerves upon movement and nutrition of the limbs. Proc R Soc London 1895; 57: 481–488. 26. Tower S. Pyramidal lesion in the monkey. Brain 1940; 63:36–90. 27. Taub E. Somatosensory deafferentation research with monkeys: implication for rehabilitation medicine. In: Ince LP, ed. Behavioral Psychology in Rehabilitation Medicine: Clinical Applications. Baltimore: Williams and Wilkens, 1980:371–401. 28. Ostendorf C, Wolf SL. Affect of forced use of the upper extremity of a hemiplegic patient on changes in function: a single-case design. Phys Ther 1981; 61:1022–1028. 29. Wolf S, Binder-Macleod SA. Electromyographic biodfeedback application to the hemiplegic patient: changes in upper extremity neuromuscular and functional status. Phys Ther 1983; 63:1393–1403. 30. Barbeau H, Fung J. The role of rehabilitation of walking in the neurological population. Curr Opin Neurol 2001; 12(6):735–740. 31. Gordon J. Assumptions underlying physical therapy interventions: theoretical and historical perspectives. In: Carr J, Shepard RD, eds. Movement Science: Foundation for Physical Therapy in Rehabilitation. 2nd ed. Gaithersburg, MD: Aspen, 2000:1–31. 32. Franz S, Scheetz ME, Wilson AA. The possibility of recovery of motor function in long-standing hemiplegia. J Am Med Assoc 1915; 65:2150–2154. 33. Baker L, Yeh C, Wilson D, Waters RL. Electrical stimulation of wrist and fingers for hemiplegic patients. Phys Ther 1979; 59(12):1495–1499. 34. Bowman B, Baker LL, Waters RL. Positional feedback and electrical stimulation: an automated treatment for the hemiplegic wrist. Arch Phys Med Rehabil 1979; 60:497–502. 35. Tangeman P, Banaitis DA, Williams AK. Rehabilitation of chronic stroke patients: changes in functional performance. Arch Phys Med Rehabil 1990; 1: 876–880. 36. Bach-y-Rita P. Recovery from brain damage. J Neurorehabil 1993; 6:191–199. 37. Winstein C, Knecht HG. Movement science and its relevance to physical therapy. Phys Ther 1990; 70(12):759–762. 38. Nudo RJ, Barbay S, Kleim JA. Role of Neuroplasticity in Functional Recovery after Stroke. Cerebral Reorganization of Function After Brain Damage. In: Harvey JG, Levin S, eds. New York: Oxford University Press, Inc., 2000.
320
Winstein and Prettyman
39. Fuhrer M, Keith RA. Facilitating patient learning during medical rehabilitation, a research agenda. Am J Phys Med Rehabil 1998; 77:557–561. 40. Nudo R, Wise BM, SiFuentes F, Milliken GW. Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science 1996; 272:1791–1794. 41. Johansson B. Brain plasticity and stroke rehabilitation. The Willis lecture. Stroke 2000; 31:223–230. 42. Tillerson J, Cohen AD, Philhower J, Miller GW, Zigmond MJ, Shallert T. Forced limb-use effects on the behavioural and neurochemical effects of 6-hydrozydompamine. J Neurosci 2001; 21:4427–4435. 43. Siegert RJL, Porter SK. Constraint-induced movement therapy: time for a little restraint?. Clin Rehabil 2004; 18(1):110–114. 44. Page S, Sisto S, Levine P, Johnston MV, Hughes M. Modified constraint induced therapy : a randomized feasibility and efficacy study. J Rehabil Res Develop 2001; 38(5):583–590. 45. Dromerick A. Evidence-based rehabilitation: the case for and against constraint-induced movement therapy. J Rehabil Res Dev 2003; 40(6):vii–ix. 46. Taub E. Movement in nonhuman primates deprived of somatosensory feedback. Exer Sport Sci Rev 1977; 4:335–374. 47. Andrews K, Stewart J. Stroke Recovery: he can but does he? Rheumatol Rehabil 1979; 18:43–48 48. Sterr A, Freivogel S, Schmalohr D. Neurobehavioral aspects of recovery: an assessment of the learned nonuse phenomenon in hemiparetic adolescents. Arch Phys Med Rehabil 2002; 83:1736–1731. 49. Taub E, Uswatte G. Constraint-induced movement therapy and massed practice [Letter to the Editor]. Stroke 2000; 31:983. 50. Fugl-Meyer A, Jaasko L, Leyman I, Olsson S, Steglind S. The post-stroke hemiplegic patient. 1. A method for evaluation of physical performance. Scand J Rehabil Med 1975; 7(1):13–31. 51. Levy L, Ziemann U, Chen R, Cohen LG. Rapid Modulation of GABA in sensorimotor cortex induced by acute deafferentation. Ann Neurol 2002; 52: 755–761. 52. Taub E, Pidikiti RD, Chatterice A, Uswatte G, King DK, Byrson C, Willcutt C, Jannett S, Yakley S, Spear M. CI therapy extended from upper to lower extremity in stroke patients. Neurosci Abstracts 1999; 25:320. 53. Pulvermuller F, Nieninger B, Elbert T, Mohr B, Rockstroh B, Koebbel P, Taub E. Constraint-induced therapy of chronic aphasia following stroke. Stroke 2001; 32:1621–1626. 54. Taub E, Uswatte G, Mark V, Willcutt C, Pearson T, Jannett T, King DK. CI therapy extended from stroke to spinal cord injured patients [abstr]. New Orleans: Society for Neuroscience, October 2000. 55. Birbaumer N, Taub E. Constraint-induced facilitation to overcome impaired movement; Application in treatment of a spinal cord-injured patient (Unpublished manuscript). Birmingham: University of Alabama, AL, 1999. 56. Van der Lee J. Constraint-induced therapy for stroke: more of the same or something completely different? Curr Opin Neurol 2001; 14:741–744.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
321
57. Wade D, Langton-Hewer R, Wood VA, Skilbeck CE, Ismail HM. The hemiplegic arm after stroke: measurement and recovery. J Neuro Neurosurg Psych 1983; 46:521–524. 58. Heller A, Wade DT, Wood VA, Sunderland A, Langton-Hewer R, Ward E. Arm function after stroke: measurement and recovery over the first three months. J Neuro Neurosurg Psych 1987; 50:714–719. 59. Wade D. Rehabilitation therapy after stroke. Lancet 1999; 354:176–177. 60. Sullivan K, Knowlton BJ, Dobkin BH. Step training with body weight support: effect of treadmill speed and practice paradigms on poststroke locomotor recovery. Arch Phys Med Rehabil 2002; 83:683–691. 61. Winstein C, Gardner ER, McNeal DR, Nicholson DE. Standing balance training: effect on balance and locomotion in hemiparetic adults. Arch Phys Med Rehabil 1989; 70(10):755–762. 62. Winstein C, Rose DK, Tan SM, Lewthwaithe R, Chui HC, Azen SP. A randomized controlled comparison of upper extremity rehabilitation strategies in acute stroke: a pilot study of immediate and lont-term outcomes. Arch Phys Med Rehabil 2004; 85:620–628. 63. Johnston K, Connors AF, Wagner DP, Knaus WA, Wang XQ, Haley EC. A predictive model for outcomes of ischemic stroke. Stroke 2000; 31:448–455. 64. Wade D, de Jong BA. Recent advances in rehabilitation. Brit Med J 2000; 320:1385–1388. 65. Lai S, Studenski S, Duncan PW, Perera S. Persisting consequences of stroke measured by the stroke impact scale. Stroke 2002; 33:1840–1844. 66. Steinman M, Maislin G, Fiedler RC, Granger CV. A prediction model for functional recovery in stroke. Stroke 1997; 28:550–556. 67. Nakayama H, Jorgensen HS, Raaschou HO, Olsen TS. Recovery of upper extremity function in stroke patients: the Copenhagen stroke study. Arch Phys Med Rehabil 1994; 75:394–398. 68. Sunderland A, Tinson D, Bradley L, Langton-Hewer R. Arm function after stroke: an evaluation of grip strength as a measure of recovery and a prognostic indicator. J Neuro Neurosurg Psych 1989; 52:1267–1272. 69. Kwakkel G, Kollen B, van der Grond J, Prevo AJH. Probability of regaining dexterity in the flaccid upper limb: impact of severity of paresis and time since onset in acute stroke. Stroke 2003; 34:2181–2186. 70. Van der Lee J, Beckerman H, Lankhorst GJ, Bouter LM. The responsiveness of the action research arm test and the Fugl-Meyer assessment scale in chronic stroke patients. J Rehabil Med 2001; 33:110–113. 71. Bonifer N, Anderson KM. Application of constraint-induced movement therapy for an individual with severe chronic upper-extremity hemiplegia. Phys Ther 2003; 83:384–398. 72. Parton A, Malhptra P, Husain M. Hemispatial neglect. J Neurol Neurosurg Psych 2004; 75:13–21. 73. Mackay F, Ada L, Heard R, Adams R. Stroke rehabilitation: are highly structured units more conducive to physical activity than less structured units? Arch Phys Med Rehabil 1996; 77:1066–1070.
322
Winstein and Prettyman
74. Newall J, Wood VA, Hewer RL, Tinson DJ. Development of a neurological rehabilitation environment: an observational study. Clin Rehabil 1997; 11: 146–155. 75. Keith R, Cowell KS. Time use of stroke patients in three rehabilitation hospitals. Soc Sci Med 1987; 24:529–533. 76. Kraft G, Fitt SS, Hammond MC. Techniques to improve function of the arm and hand in chronic hemiplegia. Arch Phys Med Rehabil 1992; 73(3):220–227. 77. Sunderland A, Tinson D, Bradley L, Fletcher D, Langton-Hewer R, Wade DT. Enhanced physical therapy improves recovery of arm function after stroke. A randomized controlled trial. J Neuro Neurosurg Psych 1992; 55: 530–535. 78. Dean C, Shepherd RB. Task-related training improves the performance of seated reaching tasks after stroke. A randomised controlled trial. Stroke 1997; 28:722–728. 79. Butefisch C, Hummelsheim H, Denzler P, Mauritz KH. Repetitive training of isolated movements improves the outcome of motor rehabilitation of the centrally paretic hand. J Neurolog Sci 1995; 130:59–68. 80. Kunkel A, Kopp B, Muller G, Villringer K, Villringer A, Taub E. Constraintinduced movement therapy for motor recovery in chronic stroke patients. Arch Phys Med Rehabil 1999; 80:624–628. 81. Cauraugh J, Light K, Kim S, Thigpen M, Behrman A. Chronic motor dysfunction after stroke: recovering wrist and finger extension by electromyography-triggered neuromuscular stimulation. Stroke 2001; 31(6):1360–1364. 82. Lum P, Burgar CG, Shor PC, Majmunder M, van der Loos M. Robotassistend movement training compared with conventional therapy techniques for the rehabilitation of upper-limb motor function after stroke. Arch Phys Med Rehabil 2002; 83:624–628. 83. Whitall J, McCombe-Waller S, Silver KH, Macko RF. Repetitive bilateral arm training with rhythmic auditory cueing improves motor function in chronic hemiparetic stroke. Stroke 2000; 31:2390–2395. 84. Kwakkel G, Wagenaar RC, Twisk JW, Lankhorst GJ, Koetsicr JC. Intensity of leg and arm training after primary middle-cerebral artery stroke: a randomized trial. Lancet 1999; 354:191–196. 85. Duncan P. Synthesis of intervention trials to improve motor recovery following stroke. Top Stroke Rehabil 1997; 3:1–20. 86. Newell A, Rosenbloom PS. Mechanisms of skill acquisition and the law of practice. In: Anderson JE, ed. Cognitive Skills and their Acquisition. Hillsdale, New Jersey: Erlbaum, 1981:1–55. 87. Taub E, Uswatte G. Constraint-induced movement therapy: bridging from primate laboratory to the stroke rehabilitation laboratory. J Rehabil Med 2003; 419(suppl):34–40. 88. Page SJ. Intensity versus task-specificity after stroke: how important is intensity. Am J Phys Med Rehabil 2003; 82(9):730–732. 89. Skinner B. The Technology of Teaching. New York: Appleton-CenturyCrofts, 1968. 90. Morgan MG. The shaping game: a teaching technique. Behav Ther 1974; 5: 271–272.
Therapy for Functional Recovery: Key Ingredients and Mechanisms
323
91. Sterr A, Elbert T, Berthold I, Kolbel S, Rockstroh B, Taub E. Longer versus shorter daily constraint-induced movement therapy of chronic hemiparesis: an exploratory study. Arch Phys Med Rehabil 2002; 83:1374–1377. 92. Pierce S, Gallagher KG, Schaumburg SW, Gershkoff AM, Gaughan JP, Shutter L. Home forced use in an outpatient rehabilitation program for adults with hemiplegia: a pilot study. Neurorehabil Neural Repair 2003; 17:214–219. 93. Miltner W, Bauder H, Sommer M, Dettmers C, Taub E. Effects of constraintinduced movement therapy on patients with chronic deficits after stroke: a replication. Stroke 1999; 30:586–592. 94. Liepert J, Miltner WHR, Bauder H, Sommer M, Dettmers C, Taub E, Weiller C. Motor cortex plasticity during constraint-induced movement therapy in stroke patients. Neurosci Lett 1998; 250:5–8. 95. Levy C, Nichols DS, Schmalbrock PM, Keller P, Chakeres DW. Functional MRI evidence of cortical reorganization in upper-limb stroke hemiplegia treated with constraint-induced movement therapy. Am J Phys Med Rehabil 2001; 80:4–12. 96. Liepert J, Bauder H, Miltner WHR, Tuab E, Weiller C. Treatment-induced cortical reorganization after stroke in humans. Stroke 2000; 31:1210–1216. 97. Gillot AJ, Holder-Wallis A, Kurtz JR, Varley NC. Perceptions and experiences of two survivors of stroke who participated in constraint-induced movement therapy home programs. Am J Occup Ther 2003; 57(2):168–176. 98. Shelton F, Volpe BT, Reding M. Motor impairment as a predictor of functional recovery and guide to rehabilitation treatment after stroke. Neurorehabil Neural Repair 2001; 15:229–237. 99. Binkofski F, Seitz RJ, Arnold S, Classen J, Benecke R, Freund HJ. Thalamic metabolism and corticospinal tract integrity determine motor recovery in stroke. Ann Neurol 1996; 39(4):460–470. 100. Binkofski F, Seitz RJ, Hacklander T, Pawelec D, Mau J, Freund HJ. Recovery of motor function following hemiparetic stroke: a clincial and magnetic resonance-morphometric study. Cerebrovasc Dis 2001; 11:273–281. 101. Feys H, De Weerdt W, Nuyens G, van de Winckel A, Selz B, Kiekens C. Predicting motor recovery of the upper limb after stroke rehabilitation: value of a clinical examination. Physiother Res Int 2000; 5(1):1–18. 102. Feys H, Van Hees J, Bruyninckx F, Mercelis R, De Weerdt W. Value of somatosensory and motor evoked potentials in predicting arm recovery after a stroke. J Neurol Neurosurg Psych 2000; 68:323–331. 103. Crafton K, Mark AN, Cramer SC. Improved understanding of cortical injury by incorporating meausures of functional anatomy. Brain 2003; 126: 1650–1659. 104. Kunesch E, Binkofski F, Steinmetz H, Freund HJ. The pattern of motor deficits in relation to the site of stroke lesions. Eur Neurol 1995; 35(1):20–26. 105. Feyday A, Carlier R, Roby-Brami A, Bussel B, Cazalis F, Pierot L, Burnod Y, Maier MA. Longitudinal study of motor recovery after stroke: recruitment and focusing of brain activation. Stroke 2002; 33:1610–1617. 106. Warabi T, Inoue K, Noda H, Murakami S. Recovery of voluntary movement in hemiplegic patients. Correlation with degenerative shrinkage of the cerebral peduncles in CT images. Brain 1990; 113(pt 1):177–189.
324
Winstein and Prettyman
107. Watanabe T, Honda Y, Fujii Y, Koyama M, Matsuzawa H, Tanaka R. Three-dimensional anisotripy contrast magnetic resonance axonography to predict the prognosis for motor function in patients suffering from stroke. J Neurosurg 2001; 94(6):955–960. 108. Hendricks H, Hageman G, van Limbeck J. Prediction of recovery from upper extremity paralysis after stroke by measuring evoked potentials. Scan J Rehabil Med 1997; 29(3):155–159. 109. Hendricks H, van Limbeck J, Geurts AC, Zwarts MJ. Motor recovery after stroke: a systematic review of the literature. Arch Phys Med Rehabil 2002; 83(11):1629–1637. 110. Hendricks H, Pasman JW, Merx JL, van Limbeck J, Zwarts MJ. Analysis of recovery processes after stroke by means of transcranial magnetic stimulation. J Clin Neurophysiol 2003; 20(3):188–195. 111. Johansen-Berg H, Dawes H, Guy C, Smith SM, Wade DT, Matthews PM. Correlation between motor improvements and altered fMRI activity after rehabilitative therapy. Brain 2002; 125:2731–2742. 112. Seitz R. Recovery of executive motor functions. In: Boller F, Grafman J, eds. Handbook of Neuropsychology. Elsevier Science BV., 1997. 113. Pineiro R, Pendlebury ST, Smith S, Flitney D, Blamire AM, Styles P, Matthews P. Relating MRI changes to motor deficit after ischemic stroke by segmentation of functional motor pathways. Stroke 2000; 31(3):672–679. 114. Purves D, Lichtman JW. Principles of Neural Development. Sunderland, MA: Sinauer Associates Inc., 1985. 115. Cotman C. Axon sprouting and regeneration. In: Siegel G, Sgranoff BW, Albers RW, Fisher SK, Uhler MD, eds. Basic Neurochemistry: Molecular, Cellular and Medical Aspects. 6th ed. PA: Lippincott-Raven, 1999:589–612. 116. Flood D, Coleman PD. Hippocampal plasticity in normal aging and decreased plasticity in Alzheimer’s disease. Prog Brain Res 1990; 83:435–443. 117. Kolb B. Overview of cortical plasticity and recovery from brain injury. Phys Med Rehabil Clin North Am 2003; 14:7–26. 118. Nudo R, Plautz EJ, Frost SB. Role of adaptive plasticity in recovery of function after damage to motor cortex. Muscle Nerve 2001; 24:1000–1019. 119. Chollet F, DiPiero V, Wise RJ, Brooks DJ, Dolan RJ, Frackowiak RS. The functional anatomy of motor recovery after stroke in humans: a study with positron emission tomography. Ann Neurol 1991; 29:63–71. 120. Weiller C, Ramsay SC, Wise RJ, Friston KJ, Frackowiak RS. Individual patterns of functional reorganization in the human cerebral cortex after capsular infarction. Ann Neurol 1993; 33:181–189. 121. Cramer S, Nelles G, Benson RR, Kaplan JD, Parker RA, Kwong KK, Kennedy DN, Finklestein SP, Rosen BR. A functional MRI study of subjects recovered from hemiparetic stroke. Stroke 1997; 28:2518–2527. 122. Pariente J, Loubinoux I, Carel C, Albucher JF, Leger A, Manelfe C, Rascol O, Schollet F. Fluoxetine modulates motor performance and cerebral activation of patients recovering from stroke. Ann Neurol 2001; 50:718–729. 123. Small S, Hlustik P, Noll DC, Genovese C, Solodkin A. Cerebellar hemispheric activation ipsilateral to the paretic hand correlates with functional recovery after stroke. Brain 2002; 125:1544–1557.
Therapy for Functional Recovery: Key Ingredients and Mechanisms 124. 125.
126. 127.
128. 129.
130. 131.
132.
133. 134. 135.
136.
137. 138.
139.
140.
325
Cotman C, Berchtold NC. Exercise: a behavioral intervention to enhance brain health and plasticity. TINS 2002; 25:295–301. Gomez-Pinilla F, Ying Z, Roy RR, Molteni R, Edgerton VR. Voluntary exercise induces a BDNF-mediated mechanism that promotes neuroplasticity. J Neurophysiol 2002; 88:2187–2195. Nudo R. Adaptive plasticity in motor cortex: implication of rehabiliation after brain injury. J Rehabil Med 2003; 41(suppl):7–10. Bury S, Jones TA. Unilateral sensorimotor cortex lesions in adult rats facilitate motor skill learning with the ‘‘unaffected’’ forelimb and training-induced dendritric structural plasticity in the motor cortex. J Neurosci 2002; 22: 8597–8606. Van Praag H, Kempermann G, Gage FH. Neural consequence of environmental enrichment. Nature Rev Neurosci 2000; 1:191–198. Feeney D. Mechanisms of noradrenergic modulation of physical therapy: effect on functional recover after cortical injury. In: Goldstein L, ed. Restorative Neurology: Advances in Pharmacotherapy for Recovery after Stroke. New York: Futura, 1998:35–78. Goldstein L. Amphetamines and related drugs in motor recovery after stroke. Phys Med Rehabil Clin North Am 2003; 14:125–134. Russo-Neustadt A. Brain-derived neurotropic factor, behavior, and new directions for the treatment of mental disorders. Sem Clin Neuropsych 2003; 8: 109–118. Stroemer R, Kent TA, Hulsebosch CE. Enhanced neocortical neural sprouting, synaptogenesis, and behavioral recovery with d-amphetamine therapy after neocortical infarction in rats. Stroke 1998; 29:2381–2395. Jones T, Schallert T. Overgrowth and pruning of dendrites in adult rats recovering from neocortical damage. Brain Res 1992; 581:156–160. Jones T, Schallert T. Use dependent growth of pyramidal neurons after neocortical damage. J Neuroscience 1994; 14(4):2140–2152. Schallert T, Leasure JL, Kolb B. Experience-associated structural events, subependymal cellular proliferative activity, and functional recovery after injury to the central nervous system. J Cereb Blood Flow Metab 2000; 20: 1513–1528. Schallert T, Fleming S, Leasure JL, Tillerson J, Bland S. CNS plasticity and assessment of forelimb sensorimotor outcome in unilateral models of stoke, cortical ablation, parkinsonism and spinal cord injury. Neuropharmacology 2000; 39:777–787. Jones T. Multiple synapse formation in the motor cortex opposite unilateral sensorimotor cortex lesions in adult rats. J Comp Neurol 1999; 414:57–66. Jones T, Chu CJ, Grande LA, Gregory AD. Motor skills training enhances lesion-induced structural plasticity in the motor cortex of adult rats. J Neurosci 1999; 19:10153–10163. Kozlowski D, James DC, Schallert T. Use-dependent exaggeration of neuronal injury follwing unilateral sensorimotor cortex lesions. J Neurosci 1996; 16:4776–4786. Keyvani K, Schallert T. Plasticity-associated molecular and structural events in the injured brain. J Neuropathol 2002; 61:831–840.
326
Winstein and Prettyman
141. Kleim J, Jones TA, Schallert T. Motor enrichment and the induction of plasticity before or after brain injury. Neurochem Res 2003; 28(11):1757–1769. 142. Rossini P, Pauri F. Neuromagnetic integrated methods tracking human brain mechanisms of sensorimotor areas ‘plastic’ reorganization. Brain Res Rev 2000; 33(2–3):131–154. 143. Frackowiak R. Functional imaging of recovery from stroke: a review of personal experience. Cerebrovasc Dis 1999; 9(suppl 5):23–28. 144. Azari N, Steitz RJ. Brain Plasticity and recovery from stroke. Am Sci 2000; 88:426–431. 145. Karni A, Meyer G, Jezzard P, Adams MM, Turner R, Ungerleider LB. Functional MRI evidence for adult motor cortex plasticity during motor skill learning. Nature 1995; 377:155–158. 146. Ungerleider LG, Doyon J, Karni A. Imaging brain plasticity during motor skill learning. Neurobiol Learn Mem 2002; 78(3):553–564. 147. Classen J, Liepert J, Wise SP, Hallett M, Cohen LG. Rapid plasticity of human cortical movement representation induced by practice. J Neurophysiol 1998; 79:1117–1123. 148. Pascual-Leon A, Dang M, Cohen LG, Brasil-Neto P, Cammoroto A, Hallett M. Modulation of muscle responses evoked by transcranial magnetic stimulation during the acquisition of new fine motor skills. J Neurophysiol 1995; 74:1037–1045. 149. Pons TP, Garraghty PE, Ommaya AK, Kaas JH, Taub E, Mishkin M. Massive cortical reorganization after sensory deafferentation in adult macaques. Science 1991; 252:1857–1860. 150. Ward N, Brown MM, Thompson AJ, Frackowiak RSJ. Neural correlates of outcome after stroke: a cross-sectional fMRI study. Brain 2003; 126: 1430–1448. 151. Werhahn K, Conforto AB, Kadom, Hallet M, Cohen LG. Contribution of the ipsilateral motor cortex to recovery after chronic stroke. Ann Neurol 2003; 54(4):464–472. 152. Calautti C, Baron JC. Functional neuroimaging studies of motor recovery after stroke in adults. A review. Stroke 2003; 34:1553–1566. 153. Cramer S, Weisskoff RM, Schaechter JD, Nelles G, Foley M, Finklestein SP, Rosen BR. Motor cortex activation is related to force of squeezing. Hum Brain Map 2002; 16(4):197–205. 154. Cicinelli P, Traversa R, Rossini PM. Post-stroke reorganization of brain motor output to the hand: a 2–4 month follow-up with focal magnetic transcranial stimulation. Electroenchepalogr Clin Neurophysiol 1997; 105(6): 438–450. 155. Cao Y, D’Olhaberriague L, Vikingstad EM, Levine SR, Welch KM. Pilot study of functional MRI to assess cerebral activation of motor function after poststroke hemiparesis. Stroke 1998; 29:112–122. 156. Dettmers C, Stephan KM, Lemon RN, Frackowiak RSJ. Reorganization of the executive motor system after stroke. Cerebrovasc Dis 1997; 7:187–200. 157. Jang SH, Kim Y-H, Cho S-H, Chang Y, Lee ZI, Ha JS. Cortical reorganization associated with motor recovery in hemiparetic stroke patients. Neuroreport 2003; 14(10):1305–1310.
Therapy for Functional Recovery: Key Ingredients and Mechanisms 158.
159.
160.
161.
162.
163.
164.
165.
166.
167.
168.
169.
170.
171. 172.
327
Liepert J, Uhde I, Graf S, Leidner O, Weiller C. Motor cortex plasticity during forced-use therapy in stroke patients: a preliminary study. J Neurol 2001; 248:315–321. Kopp B, Kunkel A, Muhlnickel W, Villringer K, Taub E, Flor H. Plasticity in the motor system related to therapy-induced improvement of movement after stroke. Neuroreport 1999; 10:807–810. Nelles G, Jentzen W, Jueptner M, Muller S, Deiner HC. Arm training induced brain plasticity in stroke studied with serial positron tomography. NeuroImage 2001; 13:1146–1154. Carey J, Kimberly TJ, Lewis SM, Auerback EJ, Dorsey L, Rundquist P, Ugurbil K. Analysis of fMRI and finger tracking training in subjects with chronic stroke. Brain 2002; 125:773–788. Wittenberg G, Chen R, Ishii K, Bushara KO, Eckloff S, Croarkin E, Taub E, Gerber LH, Hallet M, Cohen LG. Constraint-induced therapy in stroke: magnetic-stimulation motor maps and cerebral activation. Neurorehab Neural Repair 2003; 14(1):48–57. Schaechter J, Kraft E, Hilliard TS, Dijkhuizen RM, Benner T, Finklestein SP, Rosen BR, Cramer SC. Motor recovery and cortical reorganization after constrain-induced movement therapy in stroke patients: a preliminary study. Neurorehab Neural Repair 2002; 16(4):326–338. Kimberly T, Lewis SM, Auerbach EJ. Electrical stimulation driving functional improvements and cortical changes in subjects with stroke. Exper Brain Res 2004; 154(4):450–460. Butler A, Wolf SL. Transcranial magnetic stimulation to assess cortical plasticity: a critical prespective for stroke rehabilitation. J Rehabil Med 2003; 41(suppl):20–26. Binkofski F, Fink GR, Geyer S, Buccino G, Gruber O, Shah NJ, Taylor JG, Seitz RJ, Zilles K, Freund H-J. Neural activity in human primary motor cortex areas 4a and 4p is modulated differently by attention to action. J Neurophysiol 2002; 88:514–519. Curra AM, Inghilleri M, Manfredi M, Hallett M, Bererdelli A. Transcranial magnetic stimulation techniques in clinical investigation. Neurology 2002; 59(2):1851–1859. Pascual-Leon A, Walsh V, Rothwell J. Transcranial magnetic stimulation in cognitive neuroscience —virtual lesion, chronometry, and functional connectivity. Curr Opin Neurobiol 2000; 10:232–237. Friel K, Heddings AA, Nudo RJ. Effects of postlesion experience on behavioral recovery and neurophysiologic reorganization after cortical injury in primates. Neurorehab Neural Repair 2000; 14(3):187–198. Schallert T, Fleming, SM, Woodlee MT. Should the injured and intact hemispheres be treated differently during the early phases of physical restorative therapy in experimental stroke or parkinsonism? Phys Med Rehabil Clin N Am 2003; 14(suppl 1):S27–S46. Schallert T, Kozlowski DA, Humm JL, Cocke RR. Use-dependent structural events in recovery of function. Adv Neurol 1997; 73:229–238. Denny-Brown D. Disintegration of motor function resulting from cerebral lesions. J Nerv Ment Disord 1950; 112:1–45.
328
Winstein and Prettyman
173. Ince LP. Escape and avoidance conditioning of response in the plegic arm of stroke patients: a preliminary study. Psychonom Sci 1969; 16:49–50. 174. Halberstam JL, Zaretsky HH, Brucker BS, Guttman A. Avoidance conditioning of motor responses in elderly brain-damaged patients. Arch Phys Med Rehabil 1971; 52:318–327. 175. Kremer S, Chassagnon S, Hoffmann D, Benabid AL, Kahane P. The cingulate hidden hand [Letter to the Editor]. J Neuro Psych 2001; 70:264–268. 176. Page SJ. Mental practice: a promising restorative technique in stroke rehabilitation. Top Stroke Rehabil 2001; 8(3):54–63. 177. Page SJ, Sisto SA, Johnston MV. Mental practice combined with physical practice for upper-limb motor deficit in subacute stroke. Phys Ther 2001; 81(8):1455–1462. 178. Lafleur MF, Jackson PL, Malouin F, Richards CL, Evans AC, Doyon J. Motor learning produces parallel dynamic functional changes during the execution and imagination of sequential foot movements. NeuroImage 2002; 16:142–157.
16 Stem Cell Grafting and Spinal Cord Injury Richard L. Benton, Gaby U. Enzmann, and Scott R. Whittemore Kentucky Spinal Cord Injury Research Center (KSCIRC), Department of Neurology, University of Louisville School of Medicine, Louisville, Kentucky, U.S.A.
1.
INTRODUCTION
From the earliest observations of the mammalian central nervous system (CNS), it was believed that in the post developmental CNS, regeneration is significantly decreased in response to insult and injury (1). This poses a significant challenge for researchers in their search for a cure to spinal cord injury (SCI). The idea of a static posts developmental CNS is now challenged by the determination that functionally significant levels of neurogenesis occurs in the adult hippocampus and subventricular zone (SVZ) (2–4) and thus, contrary to initial dogma related to this topic, it appears that localized regeneration of CNS tissue is at least theoretically possible. Indeed, despite these challenges, early success was obtained with fetal tissue and=or fetal cell suspension grafts in improving function and augmenting regeneration following SCI (for review see Refs. (5–9). These pioneering studies provided the foundation for the hope that one day, with the appropriate intervention, regeneration of the injured adult spinal cord would be possible. Despite this optimism, issues regarding ethical and sustainable sources and availability of fetal tissue for transplantation have tempered advances in these areas of treatment. An exciting possibility for the avoidance of such ethical problems lies in the field of stem cell biology, which has been rapidly facilitated by
329
330
Benton et al.
advances in molecular neurobiology, which now allow transplantation neurobiologists to expand potential therapeutic horizons of stem cell-centric therapies. The use of stem cells in SCI includes not only engraftment of cells for gray matter replacement (neurons), but also the use of glial cell transplants for both trophic support of injured host tissue and providing a permissive substrate for axonal regrowth as well as remyelinating transplants. Not only is it feasible to use stem cells for cellular replacement strategies and=or direct replacement of lost spinal cord parenchyma, stem cells can now be reliably genetically modified into cellular ‘‘minipumps.’’ Such approaches make it possible to deliver neurotropic=neurotrophic molecules to the host spinal parenchyma by expressing transgenes encoding cytokines, neurotrophins, and=or transcription factors. Moreover, it may be possible to control the differentiation potential of the graft via autocrine=paracrine mechanisms. Neural stem cells (NSCs), more restricted neural precursor cells, immortalized NSCs and neural precursor cells, as well as non-CNS stem cells have been grafted into the spinal cord with five main goals in mind, neuronal replacement, remyelination, provision of a permissive substrate for axonal regeneration, delivery of neurotrophic factors or other proteins that facilitate regeneration, and=or providing factors that spare spinal tissue from secondary cell death. Often, the desired outcome of a particular engrafting strategy encompasses more than one of these therapeutic goals. The studies detailed below address a number of these approaches individually, but often there is substantial overlap.
2.
NEURAL STEM CELL TRANSPLANTATION IN THE SPINAL CORD 2.1. Neuronal Replacement by Engraftment of NSCs or NeuralRestricted Precursor Cells into the Injured Spinal Cord Cervical and lumbar injuries result in a loss of motor neurons essential for forelimb and hindlimb locomotor activity, respectively. While motor neuron loss is not a significant component of thoracic SCI, it is becoming increasingly clear that intrasegmental propriospinal neurons are critical for coordination between forelimb and hindlimb activity and provide crucial information to central pattern generator circuits [see Ref. (10) and references therein]. Thus, neuronal replacement has a key role in the therapeutic treatment of SCI. As is evident from Table 1, stem and precursor cells that were derived from multiple CNS areas have been grafted into the normal and injured spinal cord. The common property that all of these NSCs share is that they are pluripotent in vitro, having the capacity to differentiate into neurons, astrocytes, and oligodendrocytes. The relative proportion of these cells after
Adult rat=T10 contusion
Adult rat=motor neuron depleted
Adultrat=spinal hemisection Adult rat=C4=5 contusion Adult rat=T10 contusion Adult rat=L2 kainate lesion Adult rat=intact spinal cord
Cell type=source
ES cell derived NSCs ROSA26=D3 mouse ES
EG cell derived NSCs Human primary EGs
Embryo derived NSCs E14 rat spinal cord
E14 rat spinal cord
E14 rat cortex E14 rat cortex E15 rat cortex
Stem Cell Transplantation in the Spinal Cord
Host age=model grafted
Table 1
Astrocytes > oligodendrocytes > neurons (16 weeks) Astrocytes >> neurons > oligodendrocytes (5 weeks) Astrocytes >> oligodendrocytes (8 weeks) Undifferentiated (12 weeks) Undifferentiated (2 weeks)
Astrocytes >> neurons > oligodendrocytes (24 weeks)
Oligodendrocytes > > astrocytes > neurons (5 weeks)
Differentiated phenotype (Survival time)
18 27
No improvement Not assessed
(Continued)
13
17
12
56
16
Reference
Not assessed
Coordinated motor score improvement
Not assessed
Locomotor improvement
Locomotor improvement
Therapeutic outcome
Stem Cell Grafting and Spinal Cord Injury 331
Adult human cortex
Adult rat striatum
Adult rat SVZ
Adult derived NSCs Adult rat spinal cord
P1 porcine SVZ
E58-P11 canine striatum Neonatal derived NSCs P1 rat striatum
E14–15 rat cortex E14–15 mouse cortex E16 rat hippocampus E13 rat spinal cord E16 rat hippocampus
P6-8 rat=myelindeficient Adult rat=focal demyelinated
Adult rat=intact spinal cord Adult rat=EAE spinal lesion
Astrocytes ¼ oligodendrocytes (6 weeks) Oligodendrocytes > neurons > > astrocytes (4 weeks) Oligodendrocytes (2 weeks) Schwann cells (3 weeks)
Astrocytes > oligodendroNot Assessed cytes (2 weeks) Astrocytes >> oligodendro- Not assessed cytes (3 weeks)
Electrophysiologic improvement
Behavioral improvement electrophysiologic improvement Not assessed
Not assessed
31
35
33
14
36
32
34
28
Not assessed Not assessed
15
Hammang et al., 1997
Reference
Not assessed
Adult rat=EAE spinal lesion Adult rat=focal demyelination
Not assessed
Therapeutic outcome
Oligodendrocytes (1 week) Neurons >>> astrocytes (32 weeks) Oligodendrocytes (2 weeks)
Oligodendrocytes (2 weeks)
Differentiated phenotype (Survival time)
myelin-deficient P7 rat Intact spinal cord Adult rat=T10 contusion P6-8 rat=myelindeficient
P6-8 rat
Host age=model grafted
Stem Cell Transplantation in the Spinal Cord (Continued )
Cell type=source
Table 1
332 Benton et al.
Adult mouse L4 spinal cord ALS (SOD) model Adult mouse L4 spinal cord ALS (SOD) model
hNT
hNT
Adult rat=T10 contusion
Juvenile nude mouse intact spinal cord
P7 rat=myelin-deficient
P7 mouse myelindeficient Adult rat=mouse demyelinated myelin deficient
Adult rat=T10 contusion Adult rat=intact spinal cord
hNT
Neoplastic=clonal immortalized stem cells hNT
Embryo derived GRPs E13.5 rat spinal cord
ROSA26=D3 murine ES
ES cell derived GRPs Murine ES cells
E13.5 rat spinal cord
Embryo derived NRPs E13.5 rat spinal cord
Morphologically mature neurons (20 weeks)
NF-H-, GAD, synaptophysin-IR (< 15 months) Morphologically mature neurons (8 weeks) Morphologically mature neurons (20 weeks)
Oligodendrocytes (2 weeks)
Oligodendrocytes > astrocytes (2 weeks) Oligodendrocytes (4 weeks)
Immature neurons (4 weeks) Mature neurons (4 weeks)
67 Locomotor improvement, electrophysiologic improvement Locomotor improvement, increased animal survival Locomotor improvement delayed onset of motor dysfunction
(Continued)
68
69
66
Not assessed
50
41
Not assessed
Not assessed
40
26
11
Not assessed
Not assessed
Not assessed
Stem Cell Grafting and Spinal Cord Injury 333
Adult rat=L1 spinal cord Sciatic nerve constriction
Adult rat L1 spinal cord Sciatic nerve constriction
Adult rat=L1 spinal cord Sciatic nerve constriction
RN33B (Galanin expressing)
RN33B ( BDNF expressing)
RN46A (BDNF expressing)
Morphologically differentiated (8 weeks)
Morphologically differentiated (7 weeks)
Morphologically differentiated (7 weeks)
79
82
85
Reversed pathologic allodynia
Reversed pathologic allodynia
Reversed pathologic allodynia increased host tissue sparing
79
Adult rat=L1 spinal cord=sciatic nerve constriction
RN33B (GABA expressing)
Reversed pathologic allodynia
76
Not assessed
P3-Adult rat=spinal transection Morphologically undifferentiated Kainic acid lesion (7 weeks) Motor neuron depleted
RN33B
Morphologically differentiated (7 weeks)
71
Not assessed
Reference
Morphologically undifferentiated (2 weeks)
Therapeutic outcome
Adult rat=intact spinal cord
Differentiated phenotype (Survival time)
RN33B
Host age=model grafted
Stem Cell Transplantation in the Spinal Cord (Continued )
Cell type=source
Table 1
334 Benton et al.
Rat BM
Murine CD34þ cells
Murine BM
Murine BM
Hematopoietic stem cells Murine BM
Adult rat=focal demyelination Adult rat=focal demyelination
Adult mouse=lethally irradiated Adult mouse=sub-lethally irradiated Adult rat=focal demyelination
No trans-differentiation (1,2, and 5 months) Microglia > astrocytes > neurons (3 months) Schwann cells, oligodendrocytes (3 weeks) No trans-differentiation (3 weeks) Schwann cells, oligodendrocytes
Adult rat intact spinal cord Adult rat=T9 funiculotomy Adult rat=intact spinal cord
Neurons > astrocytes ¼ oligodendrocytes (8 weeks) Neurons > astrocytes (8 weeks) Astrocytes ¼ oligodendrocytes (4 weeks)
Adult rat=C3 dorsal column lesion
C17.2 (naı¨ve and NT-3 expressing) C17.2 (NT-3 expressing)
C17.2 (NT-3 expressing) GRIP-1
Undifferentiated (2 weeks)
Adult rat=T10 spinal hemisection
C17.2
Morphologically differentiated (6 weeks) Undifferentiated (10 weeks)
Adult rat=T13 spinal hemisection
RN46A (BDNF expressing)
105
105
Not assessed
Not assessed
(Continued)
106
104
Not assessed
Locomotor improvement
103
99
95
93
98
97
83
Not assessed
Increased host neuronal survival Not assessed
Locomotor improvement reversed pathologic allodynia Locomotor improvement increased host tissue sparing Increased host tissue sparing augmented host axonogenesis Not assessed
Stem Cell Grafting and Spinal Cord Injury 335
No trans-differentiation (1–3 weeks)
Umbilical cord blood (UCB) Human UCB Adult rat=T10 clip compression
Murine stromal cells
Rat stromal cells
Rat stromal cells
<< NeuN-IR cells (5 weeks) No trans-differentiation (1–4 weeks) NeuN-IR cells (5 weeks) Schwann cells (3 weeks)
(3 weeks)
Differentiated phenotype (Survival time)
Adult rat= T10 spinal contusion Juvenile rat= T10 spinal contusion Adult rat= T10 spinal contusion Adult rat=focal demyelination
Marrow stromal cells Rat stromal cells
Cell type=source
Host age=model grafted
Table 1 Stem Cell Transplantation in the Spinal Cord (Continued )
Locomotor improvement
Locomotor improvement
Locomotor improvement
Locomotor improvement
Locomotor improvement
Therapeutic outcome
114
113
112
111
109
Reference
336 Benton et al.
Stem Cell Grafting and Spinal Cord Injury
337
constitutive differentiation secondary to mitogen withdrawal differs between the various NSC populations and the extent of lineage-specific differentiation in vitro that can be enhanced by pharmacological treatment of the cells with specific exogenous factors also varies. Data from numerous grafting studies where similar stem and precursor cells have been grafted into other CNS regions, demonstrate that these cells are pluripotent in vivo when grafted into neurogenic CNS regions such as the CA1, CA3, and the dentate gyrus of the hippocampus or the rostral migratory stream (RMS), but show predominantly glial differentiation when grafted into non-neurogenic regions (11). The important conclusion that can be drawn from all of these studies is that the grafted NSCs have the capacity to respond appropriately to local signals in the developing and adult CNS, even when grafted ectopically, with morphological characteristics that are often indistinguishable from the host cells. Thus, the engrafted NSCs are highly plastic and endogenous signals that direct CNS differentiation are still present in the adult CNS. Collectively, these studies suggest that it is the local environment rather than the intrinsic properties of NSCs that is the predominant determinant of the differentiated fate of engrafted cells. However, in contrast to the inductive signals in neurogenic CNS regions, the epigenetic factors that control engrafted stem cell differentiation in the spinal cord do not favor neuronal differentiation. After grafting into the uninjured adult spinal cord, pluripotent NSCs derived from rat E14 spinal cord (12), rat E14 cortex (13), adult rat spinal cord (14), rat E16 hippocampus, or E13 spinal cord (15) differentiated predominantly into astrocytes and oligodendrocytes, with very few, if any, neurons. This is despite the fact that all of these engrafted cell populations were shown to be tripotent in vitro. Moreover, both adult spinal cord NSCs (14) and E14 cortical NSCs (13) showed extensive neuronal differentiation when grafted into the adult hippocampus. These data indicate that the adult spinal cord either lacks endogenous signals that induce neuronal differentiation or express factor(s) that actively repress neuronal and=or drive glial differentiation. Data discussed below support the latter hypothesis. Such restrictive cues, which remain to be elucidated, appear to be augmented in the injured and=or diseased spinal cord. Experiments in which NSCs were engrafted into the injured spinal cord have provided somewhat conflicting results. Three studies have grafted NSCs into the contused thoracic adult spinal cord. Using cells originally derived from mouse embryonic stem (ES) cells, McDonald et al. (16) found that the engrafted cells differentiated predominantly into oligodendrocytes, with some astrocytes, and few neurons. In contrast, grafts of E14 rat cortical NSCs differentiated exclusively into astrocytes, with many cells remaining undifferentiated, expressing only the stem cell marker nestin (13). Ogawa et al. (17) engrafted E14.5 rat spinal cord NSCs into a C4=5 contused spinal cord. They observed that the majority of the cells (32.6%)
338
Benton et al.
were astrocytes, with fewer numbers of oligodendrocytes (4.4%) and neurons (5.9%) with the phenotype of the remaining cells undetermined. Importantly, some functional recovery was seen with the ES cell grafts and the E14.5 rat spinal cord NSC grafts, although the exact mechanism(s) through which this occurred are unknown. After grafting E14 rat spinal cord NSCs into a spinal hemisection injury, Chow et al. (12) observed many astrocytes with fewer oligodendrocytes and even fewer neurons. Even more striking were results from Magnuson et al. (18), who observed that after grafting E14 rat cortical NSCs into the spinal cord depleted of neurons by injection of kainic acid, the cells survived but failed to differentiate along any neural lineage. They expressed only the NSC marker nestin. The discrepancy as to why no neurons were seen in some of these studies and few neurons in others may reflect the different stem cell populations used or differences in species and=or lesion model. However, what is important to recognize is that in no instance were large numbers of neurons seen. This is contrary to what is seen with NSC differentiation in vitro and suggests that successful neuronal replacement may require transplanting cells more committed to a neuronal lineage and=or modifying the host environment or the response of the engrafted cells to those local signals. Both pharmacological and genetic approaches have been taken to induce neuronal lineage restriction of NSCs. Exogenous factors such as PDGF, NT-3, and retinoic acid enhance neuronal differentiation in vitro (19,20). Genetic expression of transcription factors integral for normal neuronal development has been only partially successful in enhancing neuronal differentiation in vitro (21). However, these strategies are insufficient to produce pure neuronal populations of cells in vitro or enhance neuronal differentiation of engrafted NSCs (Cao and Whittemore, unpublished data). Thus, one alternative approach for neuronal replacement therapy may be the use of neuronal-restricted precursors (NRPs). Neuronal-restricted precursors differ from NSCs in that they are committed to the neuronal lineage at the time of isolation. They have been isolated from embryonic CNS tissue, ES cells, and proliferating multipotential NSCs (22). They give rise solely to neurons in vitro, even in the presence of glial-promoting differentiation factors (23). Following transplantation of NRPs into the SVZ of developing rat, the grafted cells migrated broadly into a variety of areas including, but not restricted to, the olfactory bulb and along the RMS. More important, the grafted NRPs adopted mature neurotransmitter phenotypes identical to the host neurons of the engraftment site (24). In contrast to results obtained with NSCs, NRP grafts do not differentiate into a glial phenotype. When syngeneic NRPs are engrafted into the normal, uninjured adult rat spinal cord, they differentiate with neuronal morphology and express neurotransmitters appropriate for the site of engraftment (25,26). Glial differentiation was not observed. However, when NRPs are engrafted into the contused adult spinal cord, neuronal
Stem Cell Grafting and Spinal Cord Injury
339
maturation of the grafted cells is significantly retarded (25). Thus, while transplantation of neuronal precursors might be a sound approach toward the goal of neuronal replacement therapy for specific CNS disorders, additional modification of the grafts and=or the host environment will be needed for mature neuronal differentiation. Consistent with this suggestion, Castellanos et al. (27) genetically modified E14 rat cortical NSCs to overexpress trkC. These cells were treated with NT3 prior to syngeneic engraftment into the uninjured adult spinal cord. They observed enhanced survival and reduced astrocytic differentiation of the engrafted trkC–NSCs. The most convincing demonstration of the need to induce lineage-specific differentiation of NSCs prior to spinal cord engraftment was shown by Wu et al. (28). They treated fetal human NSCs with a combination of FGF2, heparin, and laminin prior to engraftment into the uninjured spinal cord. Untreated NSCs differentiated into astrocytes with no neuronal differentiation observed, except in neurogenic areas such as the hippocampus. In contrast, the primed cells differentiated almost exclusively (>95%) into neurons with only small numbers of astrocytes. Importantly, over half of these neurons were cholinergic with the remainder showing GABAergic or glutamatergic phenotypes. These data highlight the fact that modulating survival and restricting some lineage choices while inducing other cell fate(s) are all likely to be important components of successful neuronal replacement therapies. Although many studies have documented successful engraftment of NSCs and NRPs into the host CNS in various transplantation paradigms, no irrefutable data exist regarding the physiological function of these cellular grafts. While apparently expressing normal neuronal markers, such as Tuj1, Map2, and=or NeuN, the demonstration of synaptogenesis or target directed axonogenesis by engrafted neurons has been conclusively shown in only one study (29). Some NSC grafting experiments have utilized measures of behavioral improvement to assess successful functional engraftment. As discussed above, transplantation of NSCs derived from ES cells into contused thoracic adult rat spinal cord has resulted in some locomotor function improvement (16) and embryonic rat hippocampal and spinal cord NSCs grafted into the contused cervical spinal cord also enhanced forelimb function (17). However, it is unclear whether this improvement was the result of functional generation of mature CNS cells and if so which one or due to neurotrophic or tissue sparing action of the graft on the host tissue. Thus, in the absence of sound in vivo functional determination of cellular grafts, even behavioral results remain enigmatic. 2.2.
Remyelination by Transplantation of NSCs or Precursors
Neural stem cells derived from embryonic rat SVZ have been grafted into the spinal cord of the myelin-deficient rat and differentiate into mature
340
Benton et al.
oligodendrocytes and remyelinate axons (17,30). Similarly, embryonic human NSCs remyelinate chemically demyelinated axons in the adult rat spinal cord (31). In experimental autoimmune encephalomyelitis, intraspinal engraftment of newborn striatal NSCs (32) or intravenous injection of adult SVZ NSCs (33) resulted in remyelination of axons in the spinal cord. In the latter study, there was functional and electrophysiological improvement. However, this study is an exception in that in the majority of the studies to date, in that the extent of remyelination by the engrafted cells and its functional in that consequences were not documented. When NSCs are transplanted into the contused rat spinal cord, the vast majority of these cells terminally differentiate into only astrocytes, with no oligodendrocytes observed (13). Thus, to get larger numbers of oligodendrocytes from the grafted NSCs, it may be necessary to initiate oligodendrocyte lineage commitment in vitro prior to transplantation. Duncan and colleagues expanded oligospheres, clusters of progenitor cells expressing immature oligodendrocyte markers, in culture using medium conditioned by B104 neuroblastoma cells. Transplantation of these oligospheres into the spinal cord of the myelin-deficient rat results in significantly larger areas of myelination as compared to that obtained from untreated neurosphere transplants (34,35). Similarly, Blakemore and colleague showed that in vitro induction of NSCs to an oligodendrocyte lineage prior to transplantation is necessary to achieve significant remyelination in the demyelinated adult CNS (36). Such approaches will be essential to obtain sufficient numbers of oligodendrocyte precursors to ensure that remyelination by engrafted cells will be therapeutically successful. Other potential sources of remyelinating NSCs are those derived from ES cells. Since undifferentiated ES cells engrafted into ectopic locations do not respond to environmental cues and result in heterotopias or tumors (37–39), transplantation of glial precursor cell populations purified from in vitro predifferentiated ES cells either by immunopanning or FACS enrichment is essential. Following transplantation into developing CNS of the myelin-deficient shiverer mice, ES-derived glial precursor cells migrate widely in the host CNS. Some cells terminally differentiate into mature oligodendrocytes, which form the myelin and ensheathed the demyelinating axons (40,41). Neural stem cells derived from rat E13.5 spinal cord or E16 hippocampus similarly remyelinated the shiverer spinal cord (15). Here again, the extent of remyelination and any potential functional improvement was unclear. More committed glial precursors have also been transplanted to remyelinate the injured and diseased CNS. The best-characterized precursor is the oligodendrocyte-type-2-astrocyte (O-2A) precursor, which can differentiate into oligodendrocytes and type 2 astrocytes in vitro but only oligodendrocytes following transplantation into brain or spinal cord [for reviews, see Refs. (42,43)]. These cells were originally isolated from
Stem Cell Grafting and Spinal Cord Injury
341
the postnatal optic nerve but have also found in other areas of the developing and adult CNS (44,45). Despite the ability of O-2A progenitors to remyelinate the demyelinated spinal cord (42,43), there are disadvantages in using O-2A cells for transplantation. These cells proliferate slowly due to what appears to be complex regulation of their capacity of self-renewal (46). Therefore, obtaining sufficient numbers of O-2A precursors for transplantation has proved problematic. Secondly, due to the paucity of terminal astrocyte differentiation from the grafted O-2A cells, hypomyelination has been observed. It appears that astrocytes serve a critical role in promoting oligodendrocyte remyelination (47–49), perhaps by activating host oligodendrocytes=precursors (47). In addition, because astrocytes also have other functions, such as providing substrates for regeneration and trophic support for surviving neurons, reconstitution of an appropriate glial environment including astrocytes and oligodendrocytes may be essential for promoting functional axonal regeneration and remyelination. Engrafting glial-restricted precursors (GRPs) may be more appropriate as a cellular therapy for remyelination as they generate both oligodendrocytes and astrocytes following transplantation into the adult CNS (50). Glial-restricted precursors can be purified from embryonic spinal cord, NSCs, and ES cells (22,51,52), and can terminally differentiate not only into mature astrocytes and oligodendrocytes, but also generate other glialrestricted precursors like the O-2A cell (53). It has been hypothesized that GRPs may represent a ‘‘master’’ glial stem cell of glial lineage commitment with the capacity to terminally differentiate into all mature glial cell types found in the mammalian CNS (53,54). In spite of their promise as a suitable remyelinating transplant, the efficacy of GRPs for functional remyelination in vivo following CNS injury has yet to be evaluated. But these cells can clearly differentiate into myelin producing cells in the normal and injured spinal cord (Cao et al., unpublished observations). While each of these distinct glial precursors and NSCs can promote myelination, it remains to be determined which will prove to be therapeutically optimal. Engrafting additional cells that are similar to endogenous precursors may in some cases be of greater benefit therapeutically. However, in other cases, providing a novel cell type, which may be more resistant to ongoing damage or less responsive to the host microenvironmental signals may be a better choice. Thus, the remyelinating efficiency of the respective glial precursor cells needs to be empirically determined for each specific injury. Thus far, quantitative data about the extent of remyelination achieved by transplantation are limited. Also, whether the myelin derived from the grafted cells functions in vivo still remains an unanswered question in most studies. Conduction velocities of the remyelinated axons showed partial recovery as evidenced by in vitro electrophysiological recording in the isolated spinal cord (31). However, no electrophysiological recovery in live animals has been demonstrated and in no studies has behavioral
342
Benton et al.
recovery been shown to result after glial precursor cell grafting into the demyelinated CNS. Moreover, most glial precursor cell transplantation experiments were done in acutely demyelinated models and how these distinct glial precursors behave in a more clinically relevant, chronic demyelinating injuries needs to be elucidated. The environment in the chronic injury is quite different from that in the acute injury. All of these issues await resolution before any reproducibly effective remyelinating transplant therapy can be initiated. 2.3.
NSC Grafts as ‘‘Cellular Minipumps’’
Two recent reports document the fact that NSC grafts may have beneficial effects on restoring function in the injured spinal cord that are distinct from cell replacement. Lin et al. (55) intrathecally engrafted adult rat spinal cord NSCs into adult rats with chronic constriction injury of the sciatic nerve, a condition that results in allodynia and enhanced sensitivity to thermal stimuli. The grafted cells survived and these animals showed significant increases in thermal withdrawal latencies compared to the control animals. The authors attribute their therapeutic effects to their GABAergic phenotype of these NSCs and microdialysis studies showed that glycine levels in the CSF were concomitantly higher in the grafted animals, consistent with its role in the inhibition of nociceptive transmission. Using a model of motor neuronopathy that utilizes injection of a neuroadapted Sindbis virus (NSV) that specifically targets motor neurons, Kerr et al. (56) infused embryoid bodies derived from an embryonic human germ cell line into the CSF of paralyzed animals. These cells survived, some cells integrated into spinal cord parenchyma, and some differentiated with neuronal characteristics, including the expression of ChAT. The grafted animals partially recovered motor function up to 24 weeks postgrafting. The effects of these grafts were not attributed to their neuronal differentiation. Rather, these cells were shown to secrete brain-derived neurotrophic factor BDNF and TGF-a, which was likely neuroprotective for the endogenous motor neurons. It is expected that additional such approaches with NSCs and precursors will facilitate recovery from SCI. 3.
IMMORTALIZED CELL TRANSPLANTATION IN THE SPINAL CORD 3.1. Spontaneously Arising=Neoplastic Cell Lines Spontaneously arising or neoplastic neural cell lines have been isolated from a variety of sources. Those derived from neuro-glioblastomas, while demonstrating reliable source lineage restriction in vitro, have exhibited higher rates of oncogenecity in the injured spinal cord, making them unsuitable for therapeutic transplantation (57). In contrast, embryonal carcinoma
Stem Cell Grafting and Spinal Cord Injury
343
(EC) cells, which are derived from testicular germ cell tumors (58), contain a population of totipotent cells, which have demonstrated the capacity to produce multiple CNS cell types including neurons. The mouse EC-derived PCC7-S-aza-R-1009 cell line has been transplanted in the injured adult rat thalamus and gives rise to apparent mature neuronal phenotypes (59). Also, P19 cells give rise to all CNS cell types following retinoic acid induction (60) and exhibit mature electrophysiological properties in vitro (61). Following transplantation into the lesioned rat striatum, they express a range of mature neuronal markers (62). Despite these successes in transplantation into other areas of the CNS, and the absence of significant tumorigenicity, studies have not shown success in spinal cord transplantation paradigms with these cell lines. In contrast, much more is known regarding the use of the human teratoma-derived cell line NTera-2 for transplantation into various CNS regions, including spinal cord. This enthusiasm and subsequent flurry of experimental approaches using this cell line is likely related to the fact that these cells were the first human neuronal cell line to undergo U.S. Food and Drug Administration (FDA)-regulated preclinical and clinical safety testing (63). Like other neoplastic cell lines, Ntera-2 cells, when appropriately induced in vitro, give rise to cultures of largely pure, postmitotic neurons (transduced nomenclature hNT neuron-like cells) (64). Further, these cells have been shown to engraft, survive, and retain regionally appropriate neuronal morphologies chronically, with no apparent tumorigenecity (65). Recent studies have attempted to expand the feasibility of the use of hNT cells for therapeutic transplantation in SCI. Hartley et al. (66) performed transplantation of hNT cells into uninjured spinal cords of athymic nude mice. Transplants were done in animals at variable developmental stages, neonatal, juvenile (21–24 days old), and adult. Grafts were evaluated at various time points following transplantation for up to 15 months posttransplantation. Results show that with increasing survival times, cells increasingly expressed mature neurochemical, structural, and synaptic markers. While not directly demonstrating functional synaptogenesis by the grafted cells, morphological assessment suggested that the engrafted cells resembled functional neuronal phenotypes, and were comparable to those phenotypes observed in the host parenchyma. The migration of engrafted hNT cells was greater than 2 cm in host spinal parenchyma, extending into both white and gray matter regions. Further, neuronal processes emanating from engrafted cells appeared to become myelinated by host oligodendrocytes. As an extension of these results, Saporta et al. (67) performed transplantation of hNT cells into adult male Sprague–Dawley rats that had previously received a modified T9 contusion injury. To test the hypothesis that epigenetic signals resulting from SCI may differentially affect graft survival, differentiation, and therapeutic efficacy, transplants were performed immediately (2 days) or chronically (2 weeks) postinjury. Results
344
Benton et al.
from these studies showed moderate improvement in electrophysiologic function as assessed by motor-evoked potential recording, with the largest improvements observed in the delayed transplant group. Despite this positive outcome, no significant changes in post-SCI ambulatory status were observed. With respect to immunotyping of engrafted hNT cells, no immunohistochemistry for non-neuronal markers was observed and cells exhibited morphologies representative of neuronal cell types, sending axons into host white matter. The hNT cell line has been transplanted into transgenic SOD-1 mutation G93A mouse models of amyotrophic lateral sclerosis (ALS) with beneficial therapeutic results (68,69). Willing and colleagues transplanted hNT neurons into the L4-L5 gray matter where active motor neuron degeneration occurs and ultimately leads to locomotor dysfunction (70). In this study, cells survived for up to 16 weeks in areas of neurodegeneration and transplanted mice exhibited significantly delayed onset of motor dysfunction and significantly increased life expectancy. Garbuzova-Davis et al. (69) replicated these results in a follow-up study where transplanted hNT cells were found to survive in recipient mice for up to 19 weeks. Despite these positive results, no information regarding functional engraftment or mature neuronal phenotypes within grafts was shown in either study. However, in the latter study, G93A mice receiving transplants demonstrated improved survival of motor neurons colocalized with the xenografts. Thus, it was concluded that one potential reason for functional improvement was likely relayed to a graft-induced neurotrophic effect rather than functional integration of the transplanted cells. The issue of immune suppression of the host required to prevent xenograft rejection complicates translation of the results from these two studies to therapeutic transplantation into the injured spinal cord. The aggressive immunosuppression employed to ensure significant graft survival remains a controversial issue related to this type of cell replacement therapy. 3.2.
Clonal=Conditionally Immortalized Stem Cell Lines
To date, several cell lines, all derived from embryonic or postnatal CNS tissue, have been thoroughly investigated as candidate cells for the therapeutic treatment of SCI, with variable degrees of success. These cells are favorable to neoplastic cell lines for CNS transplantation in that they exhibit low levels of tumorigenicity and limited immune rejection (71,72). Another beneficial attribute of these stem cell lines relates to their amenability to genetic manipulation, making it possible to insert transgenes encoding for therapeutic factors and=or mitogens to facilitate autocrine control of lineage differentiation in transplantation paradigms following SCI or other CNS injuries [see Ref. (73) and citations therein]. Thus, perhaps the most effective use of immortalized stem cells in the context
Stem Cell Grafting and Spinal Cord Injury
345
of SCI therapy has proven to be their use for ex vivo delivery of gene products to enhance regeneration, provide neuroprotection, as well as curtail the some of the most debilitating sequelae associated with chronic SCI. From a theoretical perspective, the use of these cell lines as cellular minipumps for ex vivo gene therapy could not only provide benefit as a result of the delivery of gene products, but could also differentiate into cell types, which would replace injured spinal parenchyma, and thus provide more cellular substrate for recovery. The neural cell lines most widely used for intraspinal grafting are RN33B, RN46A, and C17.2. The RN33B and RN46A cell lines were generated by the immortalization of embryonic rat raphe´ nuclei-derived neurons by retroviral infection of the temperature-sensitive allele of SV40 large T-antigen (74,75). At permissive temperature (33 C), cells are controlled by the ts-Tag and remain in a proliferative, undifferentiated state. When cultured at physiologic temperature ( >37 C), the cells differentiate over the course of 4–7 days, along a restricted neuronal phenotype and exhibit glutamatergic and serotonergic properties, respectively. Results from grafting experiments (71,76) demonstrate that following RN33B transplants into the normal adult rat lumbar spinal cord, as well as into other areas of the adult CNS, cells survive for up to 2 weeks following transplantation in spinal cord, but remain largely undifferentiated while cells engrafted into neurogenic areas (i.e., hippocampus) exhibited complex, multipolar, neuronal morphologies. Further, this lack of maturation was more pronounced in grafts following transplantation into neuronal depleting spinal lesions and a transection SCI (76). Similar observations were made in parallel with identical cell transplants in areas of the injured CNS, which were determined to favor morphological, and neurochemical maturation of this cell line (77,78). In total, these results were among the first to suggest that host, CNS region-specific, epigenetic factors may exert significant control of lineage fate in neurogenic stem cell grafts, which appear to be augmented by CNS trauma. Eaton et al. (79) investigated the feasibility of ex vivo gene therapy using RN33B and RN46A cells genetically modified to secrete the inhibitory neurotransmitter GABA, or galanin (80), or BDNF (81–88) for the treatment of intractable pain following spinal trauma. Further, promising results related to increased tissue sparing following SCI have also been obtained with transplantation of RN33B engineered to express the BDNF gene (82). Promising results have been obtained from these studies including findings of stabilization of spinal sensory gray matter in injured animals as well as improvement of overground locomotor capacity, decreased central mediated hyperalgesia, restoration of spinal serotonin and BDNF levels, and improved tissue sparing in animals receiving BDNF secreting RN46A transplants following SCI (83–85). In total, these results provide perhaps the most promising functional improvements resulting from the therapeutic
346
Benton et al.
transplantation of conditionally immortalized stem cell–lines in the injured spinal cord. In contrast to the clonal cell–lines discussed above, the C17 cell–line (and subclones i.e., C17.2, C27-3) were derived from neonatal mouse cerebellum and exhibit tripotentiality both in vitro and in vivo, forming the three major CNS cell types (neurons, oligodendrocytes, and astrocytes) (89–92). In initial transplantation experiments, C17.2 cells were infected with virus expressing the neurotrophic factor NT-3. In vitro characterization of NT-3 secreting C17.2 cells prior to transplantation demonstrated that the majority of the cells did not express mature markers for neurons (Map2) or astrocytes (GFAP), but expressed the immature but lineage committed marker for oligodendrocytes, O1 (93). Upon transplantation into the normal rat spinal cord, engrafted cells were found to integrate and survive in the host spinal parenchyma for up to 2 months post-transplantation and exhibited extensive migratory activity. Results of both morphological assessment and immunotyping of engrafted cells showed limited neuronal differentiation. However, the majority of engrafted cells analyzed expressed no markers of differentiation (GFAP, RIP, and Map2), consistent with previous results demonstrating that these cells remain quiescent and uncommitted following transplantation into the adult rat spinal cord (93). Interestingly, NT-3 transgene expression in these grafts appeared to exert a positive neurotropic action on host tissue, with extensive ingrowth of endogenous neurofilaments observed within the substance of the grafts. Further, NGF- and NT-3-secreting C17 cells have been shown to rescue Clarke’s nucleus neurons from retrograde degeneration following axotomy (94,95). More recently, the NT-3-secreting C17.2 cells were demonstrated to support host neurofilament regeneration into the lesion cervical dorsal columns (96). Subsequent studies using naive C17.2 cells seeded on polymeric scaffolds by Teng et al. (97) have recently demonstrated this therapeutic approach to yield functional improvements following engraftment into the hemisected spinal cord. Recently, Lu et al. (98) showed that NT-3producing C17.2 cells promote host axonal regeneration in the dorsal columns after a C3 hemisection injury (98). In these experiments, while host axons regenerated, very few engrafted cells differentiated into neurons, further illustrating the common theme of injury-induced epigenetic restriction of multipotent clonal stem cell grafts within the spinal cord. Collectively, these results hold promise for the use of this cell line in ex vivo gene therapeutic strategies for a combinatorial approach for the treatment of SCI. In a recent report, Wu et al. (99) characterized a conditionally immortalized GRP clone termed GRIP-1. When transplanted into normal adult rat white matter, GRIP-1 cells survived for up to 1 month post-transplantation, migrated several segments both rostral and caudal to the injection site, and expressed markers of both mature astrocytes and oligodendrocytes, demonstrating their retention of bipotentiality in vivo (99). While these
Stem Cell Grafting and Spinal Cord Injury
347
results hold promise for the development of a novel cell line for remyelination following SCI, these studies are yet to be completed.
4.
BONE MARROW AND STROMAL STEM CELL TRANSPLANTATION IN THE SPINAL CORD
Whole bone marrow (BM) consists of hematopoietic stem cells (HSC), marrow stromal cells (MSC), and a variety of differentiated cells as hematopoietic cells, osteoblasts, fibroblasts, and endothelial cells. Bone marrow is relatively easy to obtain, propagate, and modify ex vivo. In light of findings that BM can overcome its germ layer restriction and can give rise to astrocytes, microglia and neurons after transplantation into the brain of BM ablated and genetically altered mice (100–102), several groups are pursuing the potential of HSC and MSC in SCI. A number of different BM cell preparations have been engrafted into the normal or injured spinal cord. As described in Table 1 and in more detail below, unfractionated BM, CD34þ BM cells, side population BM cells isolated by FACS, density gradient-enriched mononuclear cells, BM stromal cells, and umbilical cord blood stem cells have all been used as donor cells in a number of different injury paradigms. While the results from these studies are rarely consistent between laboratories, these differences may reflect the fact that different species, cell preparations, and injury models were used. Additionally, all these studies lack any kind of cellular transplantation control making it difficult to interpret reported functional changes. What emerges from these studies is the intriguing possibility that BM cells may have a unique potential to transdifferentiate into cells of the neural lineage. If those studies can be independently replicated, it raises an enormous therapeutic potential for their use as autologous cell grafts in a number of traumatic and=or neurodegenerative CNS disorders. Recently, Castro et al. (103) reported the failure of BM to differentiate into neural cells in vivo. Lethally irradiated mice were systemically transplanted with either unfractionated BM or side-population cells derived from BM of ROSA26 mice. Immunohistochemical analysis revealed that the engrafted cells did not differentiate into cells with a neural phenotype. To potentially induce transdifferentiation, animals underwent additional cortical stab or cortical contusion injury 4 months after transplantation. In neither neural injury paradigm were b-Gal-expressing cells observed in the cervical spinal cord, except for a very few cells associated with blood vessels. Consistent with these data, 3 months after transplantation into sublethally irradiated mice, Corti et al. (104) found GFP-positive cells that had migrated into the spinal cord and were found mainly in and around vessels, at leptomeningeal sites and sparsely in the parenchyma, with a majority of the cells coexpressing microglial markers, consistent with their original lineage.
348
Benton et al.
However, a very small number of the grafted cells, found preferentially in the white matter, were positive for GFAP and both immature and differentiated neuronal markers, respectively. In contrast, Sasaki et al. (105) transplanted both acutely density gradient isolated BM and CD34þ cells derived from LacZ transgenic mice into an X-irradiated and chemically demyelinated rat dorsal column of immunosuppressed animals. They observed relatively extensive remyelination 3 weeks after transplantation of BM. The majority of the cells featured large nuclear and cytoplasmic regions and a basement membrane, consistent with the phenotype of peripheral myelin-forming Schwann cells. Unlike Schwann cells, however, the transplanted cells often engaged more than one axon, and differed in other histological aspects. Additionally, some axons were remyelinated by oligodendrocytes. CD34þ cells survived but did not myelinate the demyelinated axons. Although there were no details about the cell type that developed the myelinating phenotype, the authors concluded that crude BM contains a cell fraction capable of myelin repair in injured spinal cord. It is possible that the lethally irradiated animals used in the studies above would not support neural differentiation, but the demyelination paradigm used in these studies also involved high dose X-irradiation in combination with injection of a glial toxin. These respective studies are difficult to reconcile. After identical X-irradiation and chemical demyelination of the dorsal columns, Akiyama et al. (106) intravenously administered the mononuclear layer of freshly isolated unlabeled allogeneic BM. Three weeks after infusion, eight out of 15 animals featured myelinated axons with both Schwann cell-like and oligodendrocyte-like myelin. In vitro electrophysiological studies confirmed that the conduction velocity of the BM-injected group was significantly faster (6.2 m=sec) than the demyelinated axons (0.8 m=sec). Surprisingly, animals that received LacZ-transfected BM cells showed b-Gal reaction product in less than 10% of the myelin-forming cells. Most of the reaction product was restricted to macrophages and some cellular debris within the lesion area. While these data suggest that BM can elicit myelin repair, the specific cell type capable of that repair remains unknown. It is likely that the engrafted allogeneic cells elicited an endogenous response, likely inflammatory in nature, that facilitated remyelination by endogenous cells. Inflammation can be both detrimental to as well as facilitate functional recovery after SCI depending on the timing and specifics of the inflammatory reaction (107,108). It should be noted that both of these studies that observed remyelination of the demyelinated spinal cord came from the same laboratory and while the data are quite compelling, they need to be independently confirmed. In contrast to the well-characterized hematopoietic BM cell populations used in the above studies, mesenchymal (or stromal) stem cells (MSC) are poorly characterized and more heterogeneous. This results from the fact that there are no specific sets of markers to characterize these cells and they are initially isolated on the basis of their ability to adhere to
Stem Cell Grafting and Spinal Cord Injury
349
uncoated tissue culture plastic. It is time-consuming to obtain pure MSC cultures, noncontaminated with residual hematopoietic cells and there is great diversity in the immunotype of the cells used in the various grafting experiments. Chopp et al. (109) transplanted allogeneic, in vitro-expanded, BrdU-labeled MSC directly into a moderately contused thoracic rat spinal cord. Analysis of open field locomotor activity showed significant improvement in functional outcome in transplanted animals 4 weeks after injury. Histological examination revealed that the cells survived, migrated, and a very small number of them expressed NeuN, a marker for mature neurons. While these data are suggestive of transdifferentiation, the authors did not do the rigorous analyses needed to unequivocally demonstrate double labeling (110). Moreover, the number of surviving cells declined over the course of the experiment. The authors hypothesize that MSC placed at an appropriate time following SCI may promote repair or activate endogenous compensatory or inflammatory mechanisms by inducing cytokines or growth factor release. Similar data have been independently observed in recent experiments, done by Wu et al. (111) and Hofstetter et al. (112). Both groups transplanted cultured MSC into contused thoracic rat spinal cord. The former group saw some behavioral recovery when the cells were grafted immediately after injury while the latter study saw marginal, but statistically significant, functional improvement when the cells were grafted 1 week after injury but not if the cells were grafted immediately. Again, these data cannot be easily reconciled. In both cases, the grafted cells survived and appeared to be connected with the host tissue leaving smaller cavities than in the control. The host tissue appeared to be pulled towards the graft that was embedded in a matrix of collagen fibers. The number of transplanted cells declined gradually over the time of the experiment from 1 to 4 weeks. The former study found no differentiation of MSC into neuronal (with bIIItubulin the only marker examined) or glial cells, while the latter found extensive numbers of weakly NeuN-expressing grafted cells with none expressing other neuronal markers neurofilament protein, microtubulinassociated protein 2, or PGP-9.5. The lack of the neurofilament and PGP-9.5 in NeuN-positive cells is curious, as all should be expressed in mature neurons. Both studies concluded that the engrafted MSC indirectly promote tissue repair by stimulating endogenous processes and=or providing surfaces permissive for axonal growth. Akiyama et al. (113) transplanted mouse GFP expressing MSC into the X-irradiated and chemically demyelinated dorsal funiculus of immunosuppressed rats. Prior to transplantation, the MSC were kept in culture for some weeks and were immunopositive for CD44, collagen type I, and fibronectin. An almost complete remyelination of the lesion center and a partial remyelination at the lateral margin were observed 3 weeks after transplantation. Colocalization of both MBP and P0 with GFP-expressing cells was detected in the lesion area. Most of the remyelinated fibers featured the characteristic of peripheral myelin.
350
Benton et al.
Electrophysiological studies in vitro revealed that the conduction velocity of the transplanted group (6.03 m/sec) was slower than the control (10.76 m/ sec) but significantly faster than the demyelinated axons (0.89 m=sec). These results indicate that MSC have the potential to differentiate into myelinating cells when transplanted into a lesioned spinal cord and can improve the conduction velocity of remyelinated axons. The possibility that injected cells might recruit or facilitate an endogenous repair mechanism requires further investigation. The conflicting results between this study and those referenced above may reflect the fact that the former used rat MSC and the latter mouse MSC with immunosuppression and the lesion models were different. Moreover, only the Akiyama group (113) stained for peripheral myelin and it may have been present in those other studies. Again, independent replication is essential to sort out these conflicting results. Most recently, investigations of the therapeutic potential of umbilical cord blood derived stem cells following SCI have yielded promising results. Saporta et al. (114) infused immunosuppressed spinal cord contused rats with human umbilical cord blood stem cells. They applied prelabeled cells at various days after injury intravenously and observed significant improvement of hindlimb locomotor deficits 1–3 weeks after transplantation. Even though the engrafted cells were distributed sparsely in the cord they were found throughout the lesioned area. The authors conclude that blood cord-derived stem cells can access the lesioned area and ameliorate some of the behavioral deficits after SCI. These data are again consistent with engrafted cells initiating host responses that are secondarily responsible for the observed behavioral effects. Collectively, these studies raise more questions than they answer. If somatic stem cells indeed can give rise to cells different from their origin, autologous BM would be the ideal source to replace cells that are damaged in a variety of neurological diseases. Where most of these reports agree is that partial functional recovery after engrafting BM or MSC into the injured spinal cord can be mediated to varying degrees by graft-induced host responses. These may, in some instances, be inflammatory responses but in the cases of the immuno suppressed animals may be noninflammatory as well. To date, the nature of these responses remains poorly understood. Where these reports are irreconcilable is on the issue of transdifferentiation of engrafted BM and MSC. In a recent commentary, it was suggested that a more rigorous standard be used to document transdifferentiation of engrafted cells (115). These authors suggested that ‘‘The donor population should be isolated and grafted without expansion in cell culture, the engrafted cells should result in extensive regeneration of host tissue, evidence of functional integration and not just anatomical cell-specific marker expression should be demonstrated, the frequency of conversion needs to be determined, and lastly evidence of natural transdifferentiation in chimeric animals in the absence of injury should be demonstrated.’’ While these
Stem Cell Grafting and Spinal Cord Injury
351
are high standards, they are clearly warranted. The work from the Kocsis laboratory (105,106,113) comes the closest to meeting these criteria, but the data presented here raise one further crucial issue. In the face of conflicting data, positive results need to be independently replicated. 5.
CONCLUSIONS
The results outlined above demonstrate the complexity of the problem regarding successful stem cell treatment of SCI. Further, they indicate that despite significant progress in stem cell biology and mechanisms of SCI, we still lack sufficient knowledge to successfully engineer a therapeutic strategy. While many of these results are promising, no single approach to date has proven to be the one definitive approach to significantly improve function and=or regeneration following SCI. What is now obvious is that combinatorial strategies are the only way in which successful repair of the injured spinal cord will be achieved. Functional cell replacement, be it neurons for circuit reconstruction, oligodendrocytes and=or their precursors for remyelination, or astrocytes for providing permissive substrates for axonal regeneration, is likely to be an important component of that successful therapeutic strategy. At present, the cell autonomous and endogenous epigenetic factor(s) that control cell lineage of engrafted stem and precursor cells in specific target regions in vivo are unknown. As such, it is very difficult to design rational approaches to repair the damaged spinal cord. However, as these control mechanisms are elucidated, cellular therapies in conjunction with other approaches such as limiting the extent of secondary cell death, reducing inflammation, supplying exogenous trophic support, enhancing regeneration, limiting glial scar formation, and taking aggressive rehabilitation approaches will enable more functional repair after SCI.
REFERENCES 1. 2.
3.
4.
5.
Ramon y, Cajal S. Degeneration and Regeneration of the Nerve System. New York: Hafner, 1928. Lois C, Alvarez-Buylla A. Proliferating subventricular zone cells in the adult mammalian forebrain can differentiate into neurons and glia. Proc Natl Acad Sci USA 1993; 90:2074–2077. Eriksson PS, Perfilieva E, Bjork-Eriksson T, Alborn AM, Nordborg C, Peterson DA et al. Neurogenesis in the adult human hippocampus. Nat Med 1998; 4:1313–1317. Gould E, Reeves AJ, Fallah M, Tanapat P, Gross CG, Fuchs E. Hippocampal neurogenesis in adult Old World primates. Proc Natl Acad Sci USA 1999; 96:5263–5267. Bregman BS, Reier PJ. Neural tissue transplants rescue axotomized rubrospinal cells from retrograde death. J Comp Neurol 1986; 244:86–95.
352
Benton et al.
6.
Bregman BS, Kunkel-Bagden E. Effect of target and non-target transplants on neuronal survival and axonal elongation after injury to the developing spinal cord. Prog Brain Res 1988; 78:205–211. Falci S, Holtz A, Akesson E, Azizi M, Ertzgaard P, Hultling C, et al. Obliteration of a posttraumatic spinal cord cyst with solid human embryonic spinal cord grafts: first clinical attempt. J Neurotrauma 1997; 14:875–884. Reier PJ, Anderson DK, Thompson FJ, Stokes BT. Neural tissue transplantation and CNS trauma: anatomical and functional repair of the injured spinal cord. J Neurotrauma 1992; 9(suppl 1):S223–S248. Zompa EA, Cain LD, Everhart AW, Moyer MP, Hulsebosch CE. Transplant therapy: recovery of function after spinal cord injury. J Neurotrauma 1997; 14:479–506. Jordan LM, Schmidt BJ. Propriospinal neurons involved in the control of locomotion: potential targets for repair strategies? Prog Brain Res 2002; 137:125–139. Cao Q, Benton RL, Whittemore SR. Stem cell repair of central nervous system injury. J Neurosci Res 2002; 68:501–510. Chow SY, Moul J, Tobias CA, Himes BT, Liu Y, Obrocka M, et al. Characterization and intraspinal grafting of EGF=bFGF-dependent neurospheres derived from embryonic rat spinal cord. Brain Res 2000; 874:87–106. Cao QL, Zhang YP, Howard RM, Walters WM, Tsoulfas P, Whittemore SR. Pluripotent stem cells engrafted into the normal or lesioned adult rat spinal cord are restricted to a glial lineage. Exp Neurol 2001; 167:48–58. Shihabuddin LS, Horner PJ, Ray J, Gage FH. Adult spinal cord stem cells generate neurons after transplantation in the adult dentate gyrus. J Neurosci 2000; 20:8727–8735. Enomoto M, Shinomiya K, Okabe S. Migration and differentiation of neural progenitor cells from two different regions of embryonic central nervous system after transplantation into the intact spinal cord. Eur J Neurosci 2003; 17:1223–1232. McDonald JW, Liu XZ, Qu Y, Liu S, Mickey SK, Turetsky D et al. Transplanted embryonic stem cells survive, differentiate and promote recovery in injured rat spinal cord. Nat Med 1999; 5:1410–1412. Ogawa Y, Sawamoto K, Miyata T, Miyao S, Watanabe M, Nakamura M, et al. Transplantation of in vitro -expanded fetal neural progenitor cells results in neurogenesis and functional recovery after spinal cord contusion injury in adult rats. J Neurosci Res 2002; 69:925–933. Magnuson DS, Zhang YP, Cao QL, Han Y, Burke DA, Whittemore SR. Embryonic brain precursors transplanted into kainate lesioned rat spinal cord. Neuroreport 2001; 12:1015–1019. Johe KK, Hazel TG, Muller T, Dugich-Djordjevic MM, McKay RD. Single factors direct the differentiation of stem cells from the fetal and adult central nervous system. Genes Dev 1996; 10:3129–3140. Whittemore SR, Morassutti DJ, Walters WM, Liu RH, Magnuson DS. Mitogen and substrate differentially affect the lineage restriction of adult rat subventricular zone neural precursor cell populations. Exp Cell Res 1999; 252: 75–95.
7.
8.
9.
10.
11. 12.
13.
14.
15.
16.
17.
18.
19.
20.
Stem Cell Grafting and Spinal Cord Injury
353
21. Yamamoto S, Nagao M, Sugimori M, Kosako H, Nakatomi H, Yamamoto N et al. Transcription factor expression and Notch-dependent regulation of neural progenitors in the adult rat spinal cord. J Neurosci 2001; 21: 9814–9823. 22. Mujtaba T, Piper DR, Kalyani A, Groves AK, Lucero MT, Rao MS. Lineage-restricted neural precursors can be isolated from both the mouse neural tube and cultured ES cells. Dev Biol 1999; 214:113–127. 23. Mayer-Proschel M, Kalyani AJ, Mujtaba T, Rao MS. Isolation of lineagerestricted neuronal precursors from multipotent neuroepithelial stem cells. Neuron 1997; 19:773–785. 24. Yang H, Mujtaba T, Venkatraman G, Wu YY, Rao MS, Luskin MB. Regionspecific differentiation of neural tube-derived neuronal restricted progenitor cells after heterotopic transplantation. Proc Natl Acad Sci USA 2000; 97:13366–13371. 25. Cao QL, Howard RM, Dennison JB, Whittemore SR. Differentiation of engrafted neuronal-restricted precursor cells is inhibited in the traumatically injured spinal cord. Exp Neurol 2002; 177:349–359. 26. Han SS, Kang DY, Mujtaba T, Rao MS, Fischer I. Grafted lineage-restricted precursors differentiate exclusively into neurons in the adult spinal cord. Exp Neurol 2002; 177:360–375. 27. Castellanos DA, Tsoulfas P, Frydel BR, Gajavelli S, Bes JC, Sagen J. TrkC overexpression enhances survival and migration of neural stem cell transplants in the rat spinal cord. Cell Transplant 2002; 11:297–307. 28. Wu S, Suzuki Y, Noda T, Bai H, Kitada M, Kataoka K et al. Immunohistochemical and electron microscopic study of invasion and differentiation in spinal cord lesion of neural stem cells grafted through cerebrospinal fluid in rat. J Neurosci Res 2002; 69:940–945. 29. Auerbach JM, Eiden MV, McKay RD. Transplanted CNS stem cells form functional synapses in vivo. Eur J Neurosci 2000; 12:1696–1704. 30. Ader M, Schachner M, Bartsch U. Transplantation of neural precursor cells into the dysmyelinated CNS of mutant mice deficient in the myelin-associated glycoprotein and Fyn tyrosine kinase. Eur J Neurosci 2001; 14:561–566. 31. Akiyama Y, Honmou O, Kato T, Uede T, Hashi K, Kocsis JD. Transplantation of clonal neural precursor cells derived from adult human brain establishes functional peripheral myelin in the rat spinal cord. Exp Neurol 2001; 167:27–39. 32. Ben Hur T, Einstein O, Mizrachi-Kol R, Ben Menachem O, Reinhartz E, Karussis D et al. Transplanted multipotential neural precursor cells migrate into the inflamed white matter in response to experimental autoimmune encephalomyelitis. Glia 2003; 41:73–80. 33. Pluchino S, Quattrini A, Brambilla E, Gritti A, Salani G, Dina G et al. Injection of adult neurospheres induces recovery in a chronic model of multiple sclerosis. Nature 2003; 422:688–694. 34. Zhang SC, Lipsitz D, Duncan ID. Self-renewing canine oligodendroglial progenitor expanded as oligospheres. J Neurosci Res 1998; 54:181–190. 35. Zhang SC, Ge B, Duncan ID. Adult brain retains the potential to generate oligodendroglial progenitors with extensive myelination capacity. Proc Natl Acad Sci USA 1999; 96:4089–4094.
354
Benton et al.
36. Smith PM, Blakemore WF. Porcine neural progenitors require commitment to the oligodendrocyte lineage prior to transplantation in order to achieve significant remyelination of demyelinated lesions in the adult CNS. Eur J Neurosci 2000; 12:2414–2424. 37. Thomson JA, Itskovitz-Eldor J, Shapiro SS, Waknitz MA, Swiergiel JJ, Marshall VS et al. Embryonic stem cell lines derived from human blastocysts. Science 1998; 282:1145–1147. 38. Reubinoff BE, Pera MF, Fong CY, Trounson A, Bongso A. Embryonic stem cell lines from human blastocysts: somatic differentiation in vitro . Nat Biotechnol 2000; 18:399–404. 39. Bjorklund A. Cell replacement strategies for neurodegenerative disorders. Novartis Found Symp 2000; 231:7–15. 40. Brustle O, Jones KN, Learish RD, Karram K, Choudhary K, Wiestler OD, et al. Embryonic stem cell-derived glial precursors: a source of myelinating transplants. Science 1999; 285:754–756. 41. Liu S, Qu Y, Stewart TJ, Howard MJ, Chakrabortty S, Holekamp TF et al. Embryonic stem cells differentiate into oligodendrocytes and myelinate in culture and after spinal cord transplantation. Proc Natl Acad Sci USA 2000; 97:6126–6131. 42. Franklin RJ, Blakemore WF. Transplanting oligodendrocyte progenitors into the adult CNS. J Anat 1997; 190:23–33. 43. Franklin RJ, Blakemore WF. Glial-cell transplantation and plasticity in the O-2A lineage—implications for CNS repair. Trends Neurosci 1995; 18: 151–156. 44. Engel U, Wolswijk G. Oligodendrocyte-type-2 astrocyte (O-2A) progenitor cells derived from adult rat spinal cord: in vitro characteristics and response to PDGF, bFGF and NT-3. Glia 1996; 16:16–26. 45. Shi J, Marinovich A, Barres BA. Purification and characterization of adult oligodendrocyte precursor cells from the rat optic nerve. J Neurosci 1998; 18:4627–4636. 46. Yakovlev AY, Boucher K, Mayer-Proschel M, Noble M. Quantitative insight into proliferation and differentiation of oligodendrocyte type 2 astrocyte progenitor cells in vitro. Proc Natl Acad Sci USA 1998; 95:14164–14167. 47. Franklin RJ, Crang AJ, Blakemore WF. Transplanted type-1 astrocytes facilitate repair of demyelinating lesions by host oligodendrocytes in adult rat spinal cord. J Neurocytol 1991; 20:420–430. 48. Franklin RJ, Crang AJ, Blakemore WF. The role of astrocytes in the remyelination of glia-free areas of demyelination. Adv Neurol 1993; 59:125–133. 49. Blakemore WF. Transplanted cultured type-1 astrocytes can be used to reconstitute the glia limitans of the CNS: the structure which prevents Schwann cells from myelinating CNS axons. Neuropathol Appl Neurobiol 1992; 18: 460–466. 50. Herrera J, Yang H, Zhang SC, Proschel C, Tresco P, Duncan ID et al. Embryonic-derived glial-restricted precursor cells (GRP cells) can differentiate into astrocytes and oligodendrocytes in vivo. Exp Neurol 2001; 171:11–21. 51. Rao MS, Mayer-Proschel M. Glial-restricted precursors are derived from multipotent neuroepithelial stem cells. Dev Biol 1997; 188:48–63.
Stem Cell Grafting and Spinal Cord Injury
355
52. Rao MS, Noble M, Mayer-Proschel M. A tripotential glial precursor cell is present in the developing spinal cord. Proc Natl Acad Sci USA 1998; 95: 3996–4001. 53. Rao MS, Mayer-Proschel M. Precursor cells for transplantation. Prog Brain Res 2000; 128:273–292. 54. Lee JC, Mayer-Proschel M, Rao MS. Gliogenesis in the central nervous system. Glia 2000; 30:105–121. 55. Lin CR, Wu PC, Shih HC, Cheng JT, Lu CY, Chou AK et al. Intrathecal spinal progenitor cell transplantation for the treatment of neuropathic pain. Cell Transplant 2002; 11:17–24. 56. Kerr DA, Llado J, Shamblott MJ, Maragakis NJ, Irani DN, Crawford TO, et al. Human embryonic germ cell derivatives facilitate motor recovery of rats with diffuse motor neuron injury. J Neurosci 2003; 23:5131–5140. 57. Hasegawa K, Grumet M. Trauma-induced tumorigenesis of cells implanted into the rat spinal cord. J Neurosurg 2003; 98:1065–1071. 58. Andrews PW. Teratocarcinomas and human embryology: pluripotent human EC cell lines. Apmis 1998; 106:158–167. 59. Wojcik BE, Nothias F, Lazar M, Jouin H, Nicolas JF, Peschanski M. Catecholaminergic neurons result from intracerebral implantation of embryonal carcinoma-cells. Proc Natl Acad Sci U S A 1993; 90:1305–1309. 60. Staines WA, Morassutti DJ, Reuhl KR, Ally AI, Mcburney MW. Neurons derived from P19 embryonal carcinoma-cells have varied morphologies and neurotransmitters. Neuroscience 1994; 58:735–751. 61. Magnuson DS, Morassutti DJ, Staines WA, Mcburney MW, Marshall KC. In vivo electrophysiological maturation of neurons derived from a multipotent precursor (embryonal carcinoma) cell line. Brain Res Dev Brain Res 1995; 84:130–141. 62. Morassutti DJ, Staines WA, Magnuson DS, Marshall KC, Mcburney MW. Murine embryonal carcinoma-derived neurons survive and mature following transplantation into adult rat striatum. Neuroscience 1994; 58:753–763. 63. Zurn AD, Henry H, Schluep M, Aubert V, Winkel L, Eilers B et al. Evaluation of an intrathecal immune response in amyotrophic lateral sclerosis patients implanted with encapsulated genetically engineered xenogeneic cells. Cell Transplant 2000; 9:471–484. 64. Andrews PW. Retinoic acid induces neuronal differentiation of a cloned human embryonal carcinoma cell line in vitro. Dev Biol 1984; 103:285–293. 65. Trojanowski JQ, Kleppner SR, Hartley RS, Miyazono M, Fraser NW, Kesari S et al. Transfectable and transplantable postmitotic human neurons: a potential ‘‘platform’’ for gene therapy of nervous system diseases. Exp Neurol 1997; 144:92–97. 66. Hartley RS, Trojanowski JQ, Lee VM. Differential effects of spinal cord gray and white matter on process outgrowth from grafted human NTERA2 neurons (NT2N, hNT). J Comp Neurol 1999; 415:404–418. 67. Saporta S, Makoui AS, Willing AE, Daadi M, Cahill DW, Sanberg PR. Functional recovery after complete contusion injury to the spinal cord and transplantation of human neuroteratocarcinoma neurons in rats. J Neurosurg 2002; 97:63–68.
356
Benton et al.
68. Willing AE, Garbuzova-Davis S, Saporta S, Milliken M, Cahill DW, Sanberg PR. hNT neurons delay onset of motor deficits in a model of amyotrophic lateral sclerosis. Brain Res Bull 2001; 56:525–530. 69. Garbuzova-Davis S, Willing AE, Milliken M, Saporta S, Zigova T, Cahill DW et al. Positive effect of transplantation of hNT neurons (NTera 2=D1 cell-line) in a model of familial amyotrophic lateral sclerosis. Exp Neurol 2002; 174:169–180. 70. Gurney ME. Transgenic-mouse model of amyotrophic lateral sclerosis. N Engl J Med 1994; 331:1721–1722. 71. Onifer SM, Whittemore SR, Holets VR. Variable morphological differentiation of a raphe-derived neuronal cell line following transplantation into the adult rat CNS. Exp Neurol 1993; 122:130–142. 72. Whittemore SR, Holets VR, Keane RW, Levy DJ, McKay RD. Transplantation of a temperature-sensitive, nerve growth factor-secreting, neuroblastoma cell line into adult rats with fimbria-fornix lesions rescues cholinergic septal neurons. J Neurosci Res 1991; 28:156–170. 73. Whittemore SR, Onifer SM. Immortalized neural cell lines for CNS transplantation. Prog Brain Res 2000; 127:49–65. 74. Whittemore SR, White LA. Target regulation of neuronal differentiation in a temperature-sensitive cell line derived from medullary raphe. Brain Res 1993; 615:27–40. 75. White LA, Eaton MJ, Castro MC, Klose KJ, Globus MY, Shaw G, et al. Distinct regulatory pathways control neurofilament expression and neurotransmitter synthesis in immortalized serotonergic neurons. J Neurosci 1994; 14: 6744–6753. 76. Onifer SM, Cannon AB, Whittemore SR. Altered differentiation of CNS neural progenitor cells after transplantation into the injured adult rat spinal cord. Cell Transplant 1997; 6:327–338. 77. Shihabuddin LS, Hertz JA, Holets VR, Whittemore SR. The adult CNS retains the potential to direct region-specific differentiation of a transplanted neuronal precursor cell line. J Neurosci 1995; 15:6666–6678. 78. Shihabuddin LS, Brunschwig JP, Holets VR, Bunge MB, Whittemore SR. Induction of mature neuronal properties in immortalized neuronal precursor cells following grafting into the neonatal CNS. J Neurocytol 1996; 25:101–111. 79. Eaton MJ, Plunkett JA, Martinez MA, Lopez T, Karmally S, Cejas P, et al. Transplants of neuronal cells bioengineered to synthesize GABA alleviate chronic neuropathic pain. Cell Transplant 1999; 8:87–101. 80. Eaton MJ, Karmally S, Martinez MA, Plunkett JA, Lopez T, Cejas PJ. Lumbar transplant of neurons genetically modified to secrete galanin reverse pain-like behaviors after partial sciatic nerve injury. J Peripher Nerv Syst 1999; 4:245–257. 81. Hains BC, Yucra JA, Eaton MJ, Hulsebosch CE. Intralesion transplantation of serotonergic precursors enhances locomotor recovery but has no effect on development of chronic central pain following hemisection injury in rats. Neurosci Lett 2002; 324:222–226. 82. Cejas PJ, Martinez M, Karmally S, McKillop M, McKillop J, Plunkett JA, et al. Lumbar transplant of neurons genetically modified to secrete
Stem Cell Grafting and Spinal Cord Injury
83.
84.
85.
86.
87.
88.
89.
90.
91. 92. 93.
94.
95.
96.
357
brain-derived neurotrophic factor attenuates allodynia and hyperalgesia after sciatic nerve constriction. Pain 2000; 86:195–210. Hains BC, Johnson KM, Mcadoo DJ, Eaton MJ, Hulsebosch CE. Engraftment of serotonergic precursors enhances locomotor function and attenuates chronic central pain behavior following spinal hemisection injury in the rat. Exp Neurol 2001; 171:361–378. Hains BC, Fullwood SD, Eaton MJ, Hulsebosch CE. Subdural engraftment of serotonergic neurons following spinal hemisection restores spinal serotonin, downregulates serotonin transporter, and increases BDNF tissue content in rat. Brain Res 2001; 913:35–46. Eaton MJ, Plunkett JA, Karmally S, Martinez MA, Montanez K. Changes in GAD- and GABA- immunoreactivity in the spinal dorsal horn after peripheral nerve injury and promotion of recovery by lumbar transplant of immortalized serotonergic precursors. J Chem Neuroanat 1998; 16:57–72. Hains BC, Johnson KM, Eaton MJ, Willis WD, Hulsebosch CE. Serotonergic neural precursor cell grafts attenuate bilateral hyperexcitability of dorsal horn neurons after spinal hemisection in rat. Neuroscience 2003; 116: 1097–1110. Eaton MJ, Santiago DI, Dancausse HA, Whittemore SR. Lumbar transplants of immortalized serotonergic neurons alleviate chronic neuropathic pain. Pain 1997; 72:59–69. Eaton MJ, Whittemore SR. Autocrine BDNF secretion enhances the survival and serotonergic differentiation of Raphe neuronal precursor cells grafted into the adult rat CNS. Exp Neurol 1996; 140:105–114. Shimohama S, Rosenberg MB, Fagan AM, Wolff JA, Short MP, Breakefield XO, et al. Grafting genetically modified cells into the rat brain: characteristics of E. coli beta-galactosidase as a reporter gene. Brain Res Mol Brain Res 1989; 5:271–278. Ryder EF, Snyder EY, Cepko CL. Establishment and characterization of multipotent neural cell lines using retrovirus vector-mediated oncogene transfer. J Neurobiol 1990; 21:356–375. Kitchens DL, Snyder EY, Gottlieb DI. FGF and EGF are mitogens for immortalized neural progenitors. J Neurobiol 1994; 25:797–807. Whittemore SR, Snyder EY. Physiological relevance and functional potential of central nervous system-derived cell lines. Mol Neurobiol 1996; 12:13–38. Liu Y, Himes BT, Solowska J, Moul J, Chow SY, Park KI et al. Intraspinal delivery of neurotrophin-3 using neural stem cells genetically modified by recombinant retrovirus. Exp Neurol 1999; 158:9–26. Himes BT, Goldberger ME, Tessler A. Grafts of fetal central nervous system tissue rescue axotomized Clarke’s nucleus neurons in adult and neonatal operates. J Comp Neurol 1994; 339:117–131. Himes BT, Liu Y, Solowska JM, Snyder EY, Fischer I, Tessler A. Transplants of cells genetically modified to express neurotrophin-3 rescue axotomized Clarke’s nucleus neurons after spinal cord hemisection in adult rats. J Neurosci Res 2001; 65:549–564. Blesch A, Lu P, Tuszynski MH. Neurotrophic factors, gene therapy, and neural stem cells for spinal cord repair. Brain Res Bull 2002; 57:833–838.
358
Benton et al.
97. Teng YD, Lavik EB, Qu X, Park KI, Ourednik J, Zurakowski D et al. Functional recovery following traumatic spinal cord injury mediated by a unique polymer scaffold seeded with neural stem cells. Proc Natl Acad Sci USA 2002; 99:3024–3029. 98. Lu P, Jones LL, Snyder EY, Tuszynski MH. Neural stem cells constitutively secrete neurotrophic factors and promote extensive host axonal growth after spinal cord injury. Exp Neurol 2003; 181:115–129. 99. Wu YY, Mujtaba T, Han SS, Fischer I, Rao MS. Isolation of a glial-restricted tripotential cell line from embryonic spinal cord cultures. Glia 2002; 38:65–79. 100. Eglitis MA, Mezey E. Hematopoietic cells differentiate into both microglia and macroglia in the brains of adult mice. Proc Natl Acad Sci USA 1997; 94:4080–4085. 101. Mezey E, Chandross KJ, Harta G, Maki RA, McKercher SR. Turning blood into brain: cells bearing neuronal antigens generated in vivo from bone marrow. Science 2000; 290:1779–1782. 102. Brazelton TR, Rossi FMV, Keshet GI, Blau HM. From marrow to brain: expression of neuronal phenotypes in adult mice. Science 2000; 290:1775–1779. 103. Castro RF, Jackson KA, Goodell MA, Robertson CS, Liu H, Shine HD. Failure of bone marrow cells to transdifferentiate into neural cells in vivo. Science 2002; 297:1299. 104. Corti S, Locatelli F, Donadoni C, Strazzer S, Salani S, Del Bo R, et al. Neuroectodermal and microglial differentiation of bone marrow cells in the mouse spinal cord and sensory ganglia. J Neurosci Res 2002; 70:721–733. 105. Sasaki M, Honmou O, Akiyama Y, Uede T, Hashi K, Kocsis JD. Transplantation of an acutely isolated bone marrow fraction repairs demyelinated adult rat spinal cord axons. Glia 2001; 35:26–34. 106. Akiyama Y, Radtke C, Honmou O, Kocsis JD. Remyelination of the spinal cord following intravenous delivery of bone marrow cells. Glia 2002; 39:229–236. 107. Popovich PG, Jones TB. Manipulating neuroinflammatory reactions in the injured spinal cord: back to basics. Trends Pharmacol Sci 2003; 24:13–17. 108. Schwartz M. Macrophages and microglia in central nervous system injury: are they helpful or harmful? J Cereb Blood Flow Metab 2003; 23:385–394. 109. Chopp M, Zhang XH, Li Y, Wang L, Chen JL, Lu DY, et al. Spinal cord injury in rat: treatment with bone marrow stromal cell transplantation. Neuroreport 2000; 11:3001–3005. 110. Anderson DJ, Gage FH, Weissman IL. Can stem cells cross lineage boundaries? Nat Med 2001; 7:393–395. 111. Wu SF, Suzuki Y, Ejiri Y, Noda T, Bai HL, Kitada M et al. Bone marrow stromal cells enhance differentiation of cocultured neurosphere cells and promote regeneration of injured spinal. J Neurosci Res 2003; 72:343–351. 112. Hofstetter CP, Schwarz EJ, Hess D, Widenfalk J, El Manira A, Prockop DJ, et al. Marrow stromal cells form guiding strands in the injured spinal cord and promote recovery. Proc Natl Acad Sci USA 2002; 99:2199–2204. 113. Akiyama Y, Radtke C, Kocsis JD. Remyelination of the rat spinal cord by transplantation of identified bone marrow stromal cells. J Neurosci 2002; 22:6623–6630.
Stem Cell Grafting and Spinal Cord Injury 114.
115.
359
Saporta S, Kim JJ, Willing AE, Fu ES, Davis CD, Sanberg PR. Human umbilical cord blood stem cells infusion in spinal cord injury: Engraftment and beneficial influence on behavior. J Hematother Stem Cell Res 2003; 12:271–278. Weissman IL, Anderson DJ, Gage F. Stem and progenitor cells: origins, phenotypes, lineage commitments, and transdifferentiations. Annu Rev Cell Dev Biol 2001; 17:387–403.
17 Stem Cells and Huntington’s Disease N. Popovic, A˚. Peterse´n, J.-Y. Li, and P. Brundin Section for Neuronal Survival, Wallenberg Neuroscience Center, Department of Physiological Sciences, Lund University, Lund, Sweden
1. FEATURES OF HUNTINGTON’S DISEASE 1.1. Neurobiology of Huntington’s Disease Choreic movement was recognized as a neurological entity in the 17th century by the English physician Thomas Sydenham. A hereditary pattern of chorea was observed by Charles Waters in 1842 in families in New York state. In 1872, George Sumner Huntington was the first to publish a description of the symptoms of chorea and dementia in families in Long Island and, as a result, the disease came to bear his name. Huntington’s disease (HD) is an autosomal dominant disorder. Typically, the disease initially manifests itself with subtle psychiatric symptoms such as personality changes and depression. These can last for many years before chorea begins. Dementia then follows and death inevitably occurs 15–20 years after the onset of motor symptoms. The eventual cause of death is usually aspiration pneumonia and other complications of immobility. There is widespread death of neurons in the striatum and in the cerebral cortex, and some neurons also form cytoplasmic and intranuclear protein inclusions suggesting that their functions may be impaired. No satisfactory treatment is available. The restoration of the dysfunctional or dead neurons by transplantation of normal neurons is a potentially interesting therapeutic approach. In this review, we discuss the attempts to restore striatal function in animal models of HD by cell transplantation or augmentation of adult neurogenesis (Fig. 1). We also describe recent clinical neural grafting trials in HD. 361
362
Popovic et al.
Figure 1 Cell replacement and growth factor therapy for HD. In this review, different approaches to restore function of the diseased striatum in HD are discussed. One possibility involves grafting of the striatal anlage (LGE and MGE) in either the form of cell suspension or the whole tissue pieces, a strategy that has already been applied in clinical trials. A future approach, due to the low availability of donor embryos, may be transplantation of genetically modified striatal stem cells. Another direction may be to manipulate the endogenous restorative capacity, i.e., neurogenesis, via delivery of growth factors.
Stem Cells and Huntington’s Disease
1.2.
363
Huntington’s Disease: Mutation, Pathology, and Pathogenesis
Twenty years ago, the HD mutation was identified to be associated to the short arm of chromosome 4 (1). Ten years later, The Huntington’s Disease Collaborative Group (consisting of six research groups) (Huntington’s disease collaborative research group) (2) identified the HD gene that was found to contain an expanded CAG trinucleotide repeat. In normal subjects, the gene typically contains 9–34 CAG repeats (3) and individuals carrying the HD mutation have more than 37 repeats (4). The prevalence of clinically affected carriers is 10 in 100,000. There is an inverse correlation between the number of CAG repeats and the age of onset of HD and severity of expressed symptoms of HD (5,6). Disease onset is midlife in carriers of 40–50 CAG repeats, whereas individuals with a repeat length >70 exhibit juvenile onset. The HD gene encodes a protein of around 349 kDa, called huntingtin, which is widely expressed throughout the body. Normal huntingtin is localized in the cytoplasm, whereas the mutant form is present in both the cytoplasm and the nucleus (7–9). The functions of normal and mutant huntingtin are not fully understood. Association with synaptic proteins, vesicles, and microtubules suggest that huntingtin may be involved in vesicle trafficking and synaptic function (7,10,11 see Ref. 12 for a review). The brain appears to bear the brunt of the pathology. In the brain, huntingtin is mainly expressed in neurons, less so in glia (8,13). The striatum, the structure most severely affected in HD, expresses low to intermediate level of huntingtin mRNA, compared with the olfactory bulbs, hippocampus, and cerebellum, which paradoxically display less cell death in HD. However, inclusions rich in mutant huntingtin are found both in the striatum and in the cerebral cortex (14), which, together with hypothalamus, display progressive neuronal loss (15–17), with consequently developed astrogliosis. In advanced disease, the reduction in total brain weight can reach 30% with the striatum and neocortex, exhibiting 60% and 20% reductions in volume, respectively (14). The pathogenetic mechanisms underlying HD are important to consider in the context of stem-cell therapies for the disease. Here, we only briefly discuss the molecular events that may contribute to the development of HD neuropathology and finally relate their relevance to stem-cell therapy. Mutant huntingtin is primarily believed to cause a toxic gain-offunction, as one mutant allele is sufficient to cause dominant disease. The mechanism of cellular toxicity is not known. Proteolytic cleavage of the mutant huntingtin by caspases or calpains may be a key event in the pathogenetic process. The cleavage product containing the polyglutamine tract can be translocated to the nucleus. Both intranuclear and cytoplasmic mutant huntingtins have a propensity to form insoluble protein aggregates.
364
Popovic et al.
The protein aggregation can occur spontaneously, but the cytoplasmic aggregation is partly the consequence of an active transportation orchestrated by the centrosomes. It is not clear whether the aggregates are an efficient way for the cell to sequester toxic huntingtin species or if this process contributes to the eventual failure of the neurons. The aggregates can recruit normal huntingtin (transcribed from the wild-type allele) (18–20) and it has been suggested that the aggregates thereby deplete the cell of huntingtin and also cause a loss of normal protein function (21). In addition to mutant and wild-type huntingtin ending up in the aggregates, a large range of other cellular proteins appear to colocalize with the aggregates. These include molecular chaperones (e.g., heat shock proteins); members of the ubiquitin proteasome pathway, synaptic proteins (12), and transcription factors. As a consequence, the aggregates may trigger a loss of many vital cell functions. For example, there is evidence of transcriptional dysregulation, reduced efficacy of proteolytic cleavage by the ubiquitin proteasome system, and failure of synaptic function (22). Currently, the prevailing view is that the neurons that die in HD do so following an intrinsic cellular process, i.e., the cell ‘‘suicide’’ scenario. Although not completely excluded, there is currently less support for the idea that the mutation leads to extracellular changes, e.g., excessive release of excitotoxins from certain subsets of neurons, which cause death of nearby neurons, i.e., the concept of cell ‘‘murder.’’ This is important when considering a cell-replacement strategy. A cell ‘‘suicide’’ model implies that neurons transplanted in order to replace those lost to HD would survive rather than succumb to the disease process. On the other hand, the evidence presented earlier that neurons can malfunction in response to mutant huntingtin long before they die implies that a cell transplant may also have to interact with a markedly disturbed circuitry in HD. Moreover, as will be discussed in more detail later in this chapter, the presence of mutant huntingtin could also disturb endogenous repair mechanisms at the cellular level to the extent that stem-cell therapies based on adult neurogenesis (for example) could be difficult to perform. Thus, as mutant huntingtin may well disrupt a function in cells that do not necessarily die, one or more of the sensitive components of a successful adult neurogenesis, i.e., proliferation, migration, and differentiation of cells, might fail in HD.
2.
SYMPTOMS OF HD
The classical signs of HD are progressive chorea, rigidity, and dementia. The symptoms begin gradually and are preceded by a phase of mild psychotic and behavioral symptoms. Psychiatric disturbances in HD vary and are sometimes hard to distinguish from normal behavior. Many HD patients express nonspecific changes in personality and mood, such as
Stem Cells and Huntington’s Disease
365
irritability, impulsiveness, social withdrawal, apathy, or disinhibition. Although depression is the most common psychiatric problem, some HD patients might become manic or display symptoms of obsessive–compulsive disorder (23–25). The most prominent cognitive impairments in HD involve the ‘‘executive functions’’ such as abilities of organization, regulation, and perception (26–28). Patients with HD always display motor dysfunctions, which usually start by the irregular and spasmodic action of certain muscles, as of the face and arms. Sometimes, they may seem like fidgetiness or a nervous restlessness. Gradually, they evolve into chorea (29) and affect limbs, as well as proximal and distal muscles of the trunk. In advanced disease, there is bradykinesia and dystonia, and eventually patients reach an akinetic stage. In juvenile onset of disease, there is rapid development of rigidity, tremor, and often seizures. The HD patients also suffer from progressive weight loss for an unknown reason. Increased muscle activity, reduced capacity to swallow, impaired uptake from the intestine, loss of appetite, and an increased metabolic rate have been suggested as possible explanations. The weight-loss, together with the finding of an increased prevalence of diabetes in HD patients (10%) compared with the general population (1%), suggests that metabolic and=or endocrine dysfunctions occur (30,31). The eventual cause of death is usually aspiration pneumonia and other complications of immobility. 3.
STRATEGIES IN THE TREATMENT OF HD
There is still no successful therapy for HD. Symptomatic treatment typically involved neuroleptics to reduce involuntary movement and psychiatric symptoms. Recently, different pharmacological treatments have been tested with the aim of retarding neurodegeneration early in the disease (32–35). Identification of gene carriers before the onset of neural degeneration opens the possibility of developing treatments that delay the onset of clinical symptoms. A different approach is to replace damaged and lost neural cells by intracerebral neural transplantation. The grafted cells are not stem cell, as they are neither pluripotent nor have the ability to selfreplicate indefinitely. However, neural grafting is usually discussed in the context of stem-cell therapy, as it represents a cell-replacement strategy similar to those discussed for stem cells. Consequently, in this chapter, we will describe the experimental data underlying the neural grafting approach in HD and summarize the clinical findings obtained so far. 3.1.
Neural Transplantation in Animal Models of HD
Studies of neural transplantation in animal models of HD were initiated in the middle of 1980s by implantation of cell suspension or tissue pieces
366
Popovic et al.
dissected from embryonic rodent striatum into the excitotoxically damaged neostriatum of adult rats (36,37). Further experiments in the last two decades have generated important information about striatal development and function, as well as the potential to use embryonic striatal cells for clinical grafting in HD. This section will describe those experiments in detail. 3.2.
Fetal Tissue 3.2.1. Age of Donor Embryo, Tissue Dissection, and Graft Preparation
The striatum is formed from two transient protrusions into the lateral ventricle; the lateral ganglionic eminence (LGE) and medial ganglionic eminence (MGE) (38). The LGE is the main source of striatal 32 kDa dopamine- and cAMP-regulated phosphoprotein (DARPP-32)- and somatostatin-positive projection neurons (39–43), whereas cholinergic and parvalbumin-containing neurons are generated predominantly from the MGE (42,43). In rat, striatal neurogenesis is protracted and takes place between embryonic day (E) 12 and postnatal day (P) 2 with a peak around E15 (44,45). Recent evidence indicates that there is even a low degree of striatal neurogenesis in adult rodents (46,47). Grafting studies using telencephalic ganglionic eminences as donor tissue indicate that the LGE is the predominant source of striatal neurons (40,41). Commonly, neurons expressing striatal markers (called patch or P-zones) occupy approximately one-third of the volume of grafts derived from a mixture of rat LGE and MGE at El4–15 (48,49). In contrast, grafts of rat LGE tissue alone are composed of 40% (39) to 80% (40,42) of DARPP-32-immunopositive neurons located in Pzones, whereas MGE grafts contain only a low number of DARPP-32-neurons (4%) (42). In addition to the dissected structure, the donor age of graft tissue also influences the proportion of P-zones within the grafts. Grafts of tissue from E14 old donors, can be made up of 40% P-zones, whereas grafts generated from tissue of E18 old donors have been reported to contain only 15% P-zones (50). Immunohistochemical or in situ hybridization studies show that, in addition to DARPP-32, the neurons in P-zones express many of the normal striatal neuropeptides and neurotransmitters, such as GABA, GAD, choline acetyltransferase, metenkephalin, somatostatin, NPY, and substance P (51–59). In addition to that, using ligand-binding autoradiography, it has been demonstrated that striatal transplants express neurotransmitter receptors: Dl and D2 receptors, muscarinic acetylcholine receptors, and opiate receptors (52,58,60–63). Neurons in LGE grafts are capable of extending fibers over long distances into the globus pallidus. In contrast, MGE grafts exhibit less projections and rarely reach the globus pallidus (42). Two principally different protocols for striatal graft tissue preparation exist. Either the striatal primordial is implanted into the rodent striatum as
Stem Cells and Huntington’s Disease
367
a solid piece of tissue (typically not greater than 4 mm3 in size). Alternatively, the tissue is subjected to enzymatic digestion and is mechanically dissociated before being stereotactically injected in the form of a cell suspension. In the vast majority of cases, the grafts are placed into a striatum that has been damaged one or more weeks earlier by a local injection of an excitotoxin such as ibotenic acid or quinolinic acid (QA). Striatal grafts implanted into the lesioned striatum grow 5–8-fold in volume over the first 3 weeks. Despite the large expansion in volume of the grafts, only in the range of 10–40%, the dissected striatal neurons survive the grafting procedure (39,64–67). The low cell-survival rate has been attributed to the exposure of embryonic tissue to hypoxia and hypoglycemia, as well as to mechanical trauma during tissue preparation. Most probably, the majority of the cells are damaged or dead even before they are injected into the adult host brain. However, the lack of appropriate trophic support in the adult striatal host tissue may further reduce graft survival. Studies, comparing the survival of striatal grafts placed in an intact striatum (nonfavorable environment) and into a striatum which has sustained prior excitotoxic damage (favorable trophic environment), indicate that host milieu can significantly affect the survival of transplanted striatal neurons (68–70). Indeed, striatal grafts survive and grow significantly better in a damaged than in an intact host striatum (Fig. 2). Providing support to the view that the tissue dissection and preparation are particularly detrimental to the striatal neurons, experi-
Figure 2 Transplantation of the LGE into a rat striatum. Representative photograph of a rat-brain section with bilateral LGE grafts showing the influence of the host striatum on graft survival. The right striatum was subjected to a QA (120 nmol) injection 2 weeks prior to grafting. The brains were retrieved and processed for acetylcholinesterase and counterstained with the Nissl stain cresyl violet 4 weeks postgrafting. Arrow heads indicate the needle tract and arrows indicate the grafts. Scale bar: 2 mm.
368
Popovic et al.
ments have shown that treatment with a high concentration of the caspase inhibitor, Ac-YVAD-cmk, can increase striatal graft volume to 200% and the number of DARPP-32-positive cells in grafts to 320% of control values (67). 3.2.2. Grafts of Immortalized Cells that Deliver Trophic Factors An alternative approach to cell therapy in the striatum has been to implant cells with a capacity to protect the striatal neurons from exogenous insults such as the excitotoxin QA or the mitochondrial inhibitor 3-nitropropionic acid (3-NP). Initially, retrovirally-transfected fibroblasts were used to deliver nerve growth factor (NGF) in rats receiving subsequent injections of QA or 3-NP. Grafts of NGF-secreting cells dramatically reduced the size of the lesion and restricted neural degeneration to the region surrounding the needle track (71–76). Transplantation of glial fibrillary acidic protein -positive stem cells transfected with the human NGF gene resulted in similar protection of striatal neurons (76). Implants of brain-derived neurotrophic factor (BDNF)-producing neural stem cells are also effective, although less so than NGF-producing cells, in preventing excitotoxin-induced cell loss in the striatum (74). Immortalized fibroblasts, engineered to overexpress glial cell line-derived neurotrophic factor, can also protect against the damage induced by subsequent injections of QA (77). Encapsulated cells that have been genetically engineered to produce ciliary neurotrophic factor (CNTF) have also been implanted into animals prior to injections of excitotoxins. The encapsulation procedure allows diffusion, through a porous polymer membrane, of CNTF produced by the cells into the brain parenchyma. At the same time, the implanted cells are protected from a cell-mediated host immune response. Such implants of encapsulated CNTF-producing cells protect against QA-induced damage in both rat and primate striatum (78–80), and these animal experiments have stimulated clinical trials with a similar approach as will be discussed further in what follows. 3.2.3. Neural Stem Cells Neural stem cells have stem or progenitor-like properties: i.e., they can selfrenew and be expanded in large numbers in culture. After grafting into the rodent striatum, some forms of rodent and human neural stem cells have been shown to survive, integrate, and differentiate into both neurons and glia (81–83,83a,83b). 3.3.
Assessment of Grafted Neurons 3.3.1. Importance of the Patch Component for Function of Striatal Grafts
As mentioned earlier, graft tissue derived from the embryonic ganglionic eminences is only partly made up of striatal neurons that are located in so-called patch or P-zones. Those regions of the implants that consist of
Stem Cells and Huntington’s Disease
369
nonstriatal cells are called nonpatch zones or ‘‘NP-zones.’’ Studies of rat intrastriatal allograft transplantation have shown a positive correlation between the size of the P-zones and the improvement in skilled motor performance in animals with unilateral excitotoxic lesions of the striatum (39). It has been suggested that the graft should contain more than 25–50% of the P-zones to induce significant functional recovery (39). However, in nonhuman primates, the percentage of the P-zones in striatal allografts tends to be low (19–38%) and does not clearly correlate to the degree of functional recovery (84). 3.3.2. Integration of Striatal Grafts with the Host Brain Striatal grafts innervate the host brain and receive afferent connections from the host brain. Catecholamine histofluorescence, tyrosine hydroxylase (TH) immunohistochemical, and retrograde-tracing studies have demonstrated dopaminergic innervation of the P-zones of intrastriatal grafts derived from the host substantia nigra (58,85–89). Serotonergic raphe-derived fibers also innervate the grafts, but in a diffuse pattern, in contrast to the patchy dopaminergic fibers (87,88). Evidence for input from the frontal cortex has been demonstrated using both anterograde and retrograde tracers (85,88) and has been confirmed using electron microscopy (88,90,91). There are both afferent and efferent contacts between thalamus and striatal grafts (85,87,88). The most prominent efferent connections of striatal grafts are those with the host globus pallidus and, to a lesser extent, with the substantia nigra pars reticulata (85,92,93). Ultrastructural studies have revealed that the fibers invading the host globus pallidus form synaptic contacts (94). The transplanted striatal neurons projecting to the globus pallidus are GABAergic and can coexpress enkephalin and substance P (63). 3.3.3. Functional Properties of Striatal Grafts Pharmacological and electrophysiological studies have demonstrated that grafted striatal neurons, localized within the P-zones, exhibit several similarities to normal striatal neurons (for detailed reviews, see Refs. 64,95). Systemic administration of dopamine-releasing drugs, such as amphetamine and cocaine, induces expression of c-Fos in the grafted striatal neurons (58,96). Electrical stimulation of the cerebral cortex can also activate grafted striatal neurons and enhance their expression of c-Fos (90,97,98). As would be predicted from studies on normal brain, the removal of the host dopaminergic afferents to grafts results in an inability of dopamine-releasing drugs to induce c-Fos expression in the striatal grafts. Treatment with apomorphine, an agonist of both dopamine, Dl and D2 receptors, at a dose which has no effect on the normal striatal neurons, can increase c-Fos expression in dopamine-deafferented striatal grafts, indicating that their dopamine receptors are supersensitive (58,96,99).
370
Popovic et al.
Studies on changes in cerebral glucose metabolism have provided evidence that grafted striatal neurons can normalize changes in neuronal activity induced by excitotoxic striatal lesions. Thus, lesion-induced increases in 14C deoxyglucose uptake in the globus pallidus and subthalamic nuclei, due to loss of innervation from the striatum, are mitigated following intrastriatal implantation of striatal tissue (37,100,101). Electrophysiological data supports the idea that striatal grafts can reinstate an inhibitory GABAergic tone in the globus pallidus and has revealed a reversal of the increase discharge rate in pallidal neurons that is induced by striatal lesions (101). Microdialysis in vivo studies show that striatal grafts release GABA and that dopaminergic and glutaminergic inputs from the host control GABA overflow (102). However, there is not a complete restoration of normal neuronal activity in the basal ganglia, as increased glucose uptake in the entopeduncular nucleus and substantia nigra have so far not been found to be affected by the grafts (100). 3.4.
Amelioration of Behavioral Deficits by Intrastriatal Grafts
Assessment of the ameliorative effect of intrastriatal neural grafting in rodent models of HD has been performed by examining a variety of motor and cognitive tests that are described in detail in the following sections (see Ref. 64 for a detailed review). 3.4.1. Locomotor Hyperactivity and Drug-Induced Circling There is now ample evidence that striatal grafts can reverse simple motor deficits in rats induced by lesions of the striatum. The first ameliorative effects of striatal graft on locomotor impairments induced by excitotoxic lesions of the striatum were reported by Deckel et al. (36,60,103–105) and by Isacson et al. (37,106). In one series of studies, solid grafts of El7–E18 whole ganglionic eminence implanted bilaterally into the previously lesioned adult striatum, reversed spontaneous, but not amphetamine-induced, locomotor hyperactivity 3–5 months after transplantation (60,103–105). Autoradiography revealed low levels of D2-receptor binding within the graft. Cell-suspension grafts, also composed of both MGE and LGE, exhibited similar effects on nocturnal locomotor hyperactivity after implantation in rats with either unilateral (37) or bilateral (106) striatal lesions. Later studies have reported reduced amphetamine-induced hyperlocomotion (107–112). Striatal grafts have been found to reduce the rotational asymmetry induced by dopamine agonists or releasing drugs in rats with unilateral excitotoxic lesions of the striatum. Grafts, consisting of both LGE and MGE, can ameliorate dopamine-active drug-induced circling behavior by 50–70% (113–115). Grafting only LGE tissue can have similar consequences, whereas grafts of MGE tissue alone do not have any beneficial effects (100).
Stem Cells and Huntington’s Disease
371
3.4.2. Complex Motor Deficits Skilled forelimb use in rats depends on the integrity of the sensorimotor area of the neocortex, intrinsic striatal neurons, and the nigrostriatal dopaminergic system (116). Grafts generated from the mixture of LGE and MGE, implanted into the unilaterally ibotenate-lesioned striatum, can improve impairment of reaching abilities of both paws up to normal levels (115– 117). Unilateral implantation of striatal tissue in rats with bilateral kainic acid lesions has been reported to induce a significant shift of the paw preference toward the contralateral side to the graft (118). Later studies have shown that improvement in paw reaching deficits is highly correlated with volume of P-zones (39,119) and binding densities of both Dl and D2 receptor ligands within the graft (120), indicating that it is the component of striatal neurons in the implants that are crucial in eliciting the effects on this complex motor behavior. Furthermore, nonspecific motor training and cell transplantation can contribute to recovery of lost function after striatal damage. It seems that these two factors interact in a task-specific manner (121). 3.4.3. Conditioned Behavior Deficits Intrastriatal striatal grafts can exhibit beneficial effects on conditioned behavior impairments. Rats with bilateral striatal lesions were trained for acquisition of a delayed alteration in a T-maze task after having implanted cell suspension (El4) grafts into the lesioned loci. Graft recipients achieved around 90% correct alterations in comparison with 50% correct responses only in lesioned rats (106). However, time necessary to acquire this task was still longer in graft recipient rats than in normal animals. In contrast to that, grafts localized in GP were less effective in induction of improvement of T-maze test performance. Transplantation of solid striatal grafts (El8) into the lesioned striatum of pretrained rats was less successful (103,104). Striatal grafts can also ameliorate impairments in visual-choicereaction time task (122). Animals with implanted striatal grafts, even initially as deficient in performing the visual choice reaction task, showed improvement if they were repeatedly tested. Results of that experiment indicated that postoperative training could significantly contribute to recovery of functions induced by neural implantation (122). 3.5.
Neural Transplantation in a Transgenic Mouse Model of HD
The R6 line of transgenic mice expressing exon 1 of the mutant HD gene exhibits progressive behavioral symptoms, weight loss, and premature death, although very little neuronal cell loss has been reported. The neuropathological characteristics include neuronal atrophy (123) and widespread presence of intranuclear protein inclusions (124). Several reports have shown a dysfunction of striatal, cortical, and dopaminergic neurons
372
Popovic et al.
(125–127). This transgenic mouse has recently been used in two neural grafting studies (115,128). In one study, striatal tissue was dissected from E14 wild-type mouse embryos and implanted bilaterally into the R6=2 striatum. The grafts showed a good survival but failed to ameliorate the behavioral symptoms or the weight loss (115). It could be argued that the lack of effect was due to the grafts being implanted too late in the development of the disease phenotype and because R6=2 mice have a short lifespan, the grafts did not have time to develop their full potential. Another study using the same line of transgenic mice studied the effects of wild-type cortical tissue transplanted into the anterior cingulate cortex (128), with the view that this region in itself is crucial for the neurological dysfunction or may indirectly affect development of the disease phenotype. The grafts only ameliorated the development of an abnormal paw clasping but had no effect on fine motor co-ordination making the utility of this grafting approach questionable (128).
4.
STRIATAL NEURAL TRANSPLANTATION IN NONHUMAN PRIMATES
Intrastriatal grafting has been applied in either of two primate models of striatal lesions (84,129,130): (1) unilateral excitotoxic lesions of the putamen that induces behavioral alterations associated with motor hyperactivity in HD patients or (2) 3-NP-induced bilateral dorsolateral caudate–putamen lesions accompanied with behavioral impairments similar to an end-stage, motor hypoactive phase of HD. The first intrastriatal transplantation of striatal tissue in primates was performed by Isacson et al. (131). They demonstrated a partial amelioration of drug-induced dyskinesias following implantation of rat tissue xenografts in baboons previously given excitotoxic lesions of the caudate and putamen. Histological analysis revealed grafts with acetylcholinesterase-positive patches in one grafted baboon. A more detailed study was performed with multiple lesion and rat xenograft implants in the caudate and putamen on six baboons immunosuppressed with cyclosporin A (132). Administration of l-DOPA or apomorphine elicited abnormal movements in lesioned monkeys, including chorea and dyskinesias (133,134), which declined gradually in graft recipients. Withdrawal of cyclosporine in two baboons induced increases of dyskinesia, suggesting that the transplants were directly responsible for the functional recovery. Moreover, functional recovery was only achieved after transplantation of appropriate striatal tissue while implantation of brain stem tissue in one baboon was ineffective. The volume of the striatal grafts was typically 15–30 mm3, but in two baboons from which immunosuppression was withdrawn, there were signs of ongoing rejection 15 weeks later (132). Further studies have shown that the implantation of allogeneic striatal ganglionic
Stem Cells and Huntington’s Disease
373
eminence (in most cases, LGE and MGE combined) in marmosets, 4–5 weeks after unilateral excitotoxic lesion of putamen, can produce a partial amelioration of the reaching deficits in a paw-reaching task (84). The transplantation procedure did not have any effect in apomorphine-induced rotation in the lesioned marmosets. Positron emission tomography (PET), with the D2 receptor antagonist,11C-raclopride, in grafted monkeys, which exhibited behavioral recovery, detected a clear signal on the implanted side (84,135,136). Postmortem analysis showed that surviving graft tissue had the characteristic patchy appearance in acetylcholinesterase and DARPP-32 staining, with the mean percentage of the patch within each graft being between 16% and 30%. Although there was a trend for the primates with the largest patch component to exhibit the greatest degree of recovery, direct correlations between graft volume and patch percentages with functional recovery were not found. An increase of the percentage of P-zone was not achieved by grafting selectively the tissue from the LGE (136). Chronic treatment of primates with daily systemic injections of 3-NP produces a marked, bilateral, neuronal cell loss restricted to the caudate and the putamen (137). In addition, these lesions are accompanied by progressive development of a variety of dyskinetic movements and dystonic postures. Macaques treated with 3-NP have been stereotactically grafted with cell suspensions derived from the LGE of macaque embryos at multiple sites in the caudate and putamen. Three of six grafted macaques displayed complete recovery in certain cognitive tasks at 2 months post grafting. Dystonia was significantly improved at 4 and 5 months following surgery (136). Finally, histological analysis of the striatal allografts revealed the characteristic P-zones rich in calbindin, DARPP-32, acetylcholinesterase, and TH. In summary, grafting studies in primate models of HD have shown that striatal implants can survive and exert functional effects in both cognitive and motor tasks. The histological appearance of the grafts is similar to that observed in rodents. However, the expense and difficulties involved in performing trials in primate models of neurological disease have precluded the collection of extensive data where key experimental parameters have been systematically varied. Therefore, it is not surprising that although the existence of successful studies in primates have been essential for the development of clinical grafting programs in HD, these nonhuman primate experiments have had limited influence on the design of the clinical trials. Thus, experiments performed in monkeys have not been able to clearly define the optimal embryonic developmental stage of primate striatal donor tissue. Moreover, it is not known if the P-zone component is crucial for functional effects, as has been suggested from rat studies (39,50) and it is not clear what locations within the caudate and putamen are most crucial to support functional effects. As is apparent from the following sections, the lack of this information has not precluded the advancement of the field into clinical cell transplantation trials in HD patients.
374
Popovic et al.
5. CLINICAL TRANSPLANTATION TRIALS IN HD 5.1. Initial Studies The first organized discussions on the possibility of grafting as a treatment of HD were initiated at the meeting of the Network of European CNS Transplantation And Restoration (NECTAR), in November 1992. Soon after that, the European Network for Striatal Transplantation in Huntington’s Disease (NEST-HD) was established in 1994, with the aim to advance experimental work and to develop methodology for clinical trials in HD. This working group took the initiative to establish the Core Assessment Program for Intracerebral Transplantation in Huntington’s Disease (CAPITHD) in 1994 (138). The CAPIT-HD defines criteria that assist in the selection of HD patients suitable for surgery and includes a detailed protocol for the assessment regarding motor and cognitive functions and brain imaging, prior to and following transplantation surgery. As will be apparent in the following section, the age of the donor tissue, the precise region dissected, and the methods used to prepare and implant have varied between the different clinical trial centers. Prior to initiation of these clinical trials, there was a series of experiments investigating morphological features of human embryonic striatal tissue in histological sections, cell culture, and human-to-rat xenograft experiments. The first such study focused on grafting human embryonic telencephalic precursors to the striatum of rats that had previously received injections of excitotoxin into the striatum (139). The main observation in this study was that immature human neuroblasts can exhibit a remarkable capacity to grow axons, which could be traced using a human-specific marker for neurofilament, over long distances in the adult rat CNS. Thus, human axons were seen to grow from the striatum to the substantia nigra and even along the corticospinal tract into the cervical spinal cord. This study preceded the subsequent understanding of selective subdissections of the ganglionic eminences, and therefore, the precise origin of the human telencephalic neuroblasts with remarkable growth capacity was not determined. It is likely that they constituted a mixture of cortical and striatal precursors. It is only about a decade ago that some groups attempted to define what type of human embryonic tissue would be most suitable for grafting into patients with HD. These studies indicated that, in analogy to findings made in rodents, in human embryos, the vast majority of DARPP-32-positive striatal precursors appear to be located in the LGE, both according to histological studies (140) and if cell cultures from the LGE and MGE are directly compared (141). In agreement with this, tissue derived from the LGE developed more P-zones than MGE tissue when grafted to the lesioned striatum of immunosuppressed rats, but typically no more than 30% of the grafts were ever made up of P-zones (141). Histological features of the LGE grafts indicated that they were relatively immature even several months after implantation into rats
Stem Cells and Huntington’s Disease
375
(140,142). It was speculated that the MGE may be essential to promote the maturation of striatal neurons in the LGE during the protracted development of the human striatum, but combined MGE and LGE grafts did not result in an increased maturation of DARPP-32 neurons (142). Notwithstanding the low proportion of the grafts exhibiting features of P-zones (140–143), human embryonic striatal grafts have still been reported to reduce locomotor deficits in rats with striatal lesions (144). The combination of knowledge from rat-to-rat transplantation studies and experiments in nonhuman primates laid the foundation for proceeding with clinical trials in HD (145). At the time, there was also concern that the level of understanding on how one can achieve human grafts rich in striatal neurons was not sufficient to warrant the transfer into clinical trials (146). This concern did not significantly delay the field entering the clinical era. On the basis of the available literature, intrastriatal transplantation of human embryonic tissue has been reported to date in two patients in Mexico, five patients in Cre´teil, France, three patients in Los Angeles, U.S.A., four patients in Cambridge, U.K., seven patients in Tampa, U.S.A., and several patients in Cuba and Czech Republic (Table 1). In addition, 12 HD patients have received xenografts of fetal porcine neuronal cells in a multicenter trial in the U.S.A. (147). A recently initiated European transplantation trial program might involve more than 100 patients in several clinical centers in U.K., France, Switzerland, and Italy (148). However, the first systematic account of the results from such a trial should not be expected to be available until 4–5 years from now. The following section will briefly summarize some of the methodological aspects of the surgery and the main features of the results from the trials that so far have been reported in detail.
Table 1
Type of Graft and Surgical Parameters for Intrastriatal Neural Transplantation in Several Clinical Trials
Group
Number of patients
France
5
Los Angeles
3
Tampa
7
U.K.
4
Type of graft Solid grafts WGE Solid grafts LGE Solid grafts LGE Cell suspension WGE
Target and number of implantation tracks per side Bilateral caudate and putamen Four to five implantation tracks Bilateral caudate and putamen Five implantation tracks Bilateral caudate and putamen Two to eight implantation tracks Unilateral caudate and putamen Four to six implantation tracks
Note: The above table describes the transplantation approach of the four main clinical trials of intrastriatal grafting in HD patients reported to this date. Similarities and differences in the type and number of grafts, as well as targets for the grafts, are shown.
376
5.2.
Popovic et al.
Clinical Transplantation Trials in HD: Methodology and Main Results
Not all of the eight programs that have reported clinical grafting trials in HD have provided sufficient information in the literature to make a scientific evaluation of the procedure and its outcome. For this reason, the patients operated in Mexico, Cuba, and Czech Republic will not be described further in this review. In addition, the pig-to-human xenotransplantation trial in HD constitutes a special case; in that, it was driven by a company in the biotech industry. Unilateral intrastriatal xenotransplantation of cell suspensions derived from LGE of 35–38 embryonic days-old porcine embryos was performed into caudate and putamen in several HD patients treated with cyclosporine A. Currently, there is only a relatively brief account of these patients in the literature, and it states that no improvement was observed after grafting (147). It is not unlikely that immune rejection of the porcine xenografts took place in the majority of these HD patients. Three patients operated in Los Angeles have been reported in the literature. They received a combination of LGE and MGE from five to eight donors bilaterally into the caudate and putamen and were given cyclosporine immunosuppression following the procedure (149). The investigators suggest that magnetic resonance imaging provides evidence of surviving grafts. However, after 1 year, there was no evidence of functional benefits of the transplants. Patients grafted in Cambridge, U.K., received stereotactic implants of a mixed LGE and MGE cell suspension, derived from dissected human embryo, into the nondominant, right hemisphere (150). A second transplantation in the contralateral side in these four patients was planned at the time of the initial report being published. Patients were initially treated with immunosuppressants before operation and continuously postoperatively with cyclosporine A, azathioprine, and prednisolone for at least 6 months. Magnetic resonance imaging revealed a reduction in the volume of the implanted area 3 months after transplantation compared with immediately after surgery, probably due to the decrease in postoperative edema (150). However, by 6 postoperative months, a modest increase in volume of the presumed graft tissue had taken place in some of the patients, suggestive of graft growth and development (150). The functional outcome has only been followed in the short-term, i.e., 6 months, posttransplantation and indicates a trend of motor improvement in these patients. However, as this initial report on the U.K. patients was focused on the safety aspects of the procedure, the investigators quite correctly do not make any major claims for functional benefits occurring. In contrast to surgical procedure performed in the U.K., where dissociated tissue from both LGE and MGE was implanted, patients in Tampa, Florida, received bilateral implants of solid pieces of LGE into putamen and
Stem Cells and Huntington’s Disease
377
caudate nucleus (151). In accordance with the interpretation that most striatal precursors are born in the lateral aspect of the human embryonic LGE (152), this specific part was harvested for grafting from a total of two to eight embryos per side of the patient’s brain. The number of graft sites was unusually high with the graft tissue being distributed over two to eight stereotactic injections on each side of the brain. There was a 1–14-week interval between the surgery in the two hemispheres, and patients were given cyclosporine immunosuppression starting 1–2 weeks before the first operation and up to 2 weeks following the second surgery (151). There were subdural hemorrhages in three of the seven patients, possibly related to the high number of cannula penetrations during the stereotactic surgery. After surgery, maintenance of metabolic activity was also observed by PET without changes in putaminal dopamine receptor binding, as assessed using raclopride as a ligand. Importantly, neurological examinations still showed slight motor improvements 1 year after transplantation (151). Unfortunately, one patient from Tampa died 18 months after transplantation from cardiovascular disease. Interestingly, postmortem histological analysis revealed surviving transplanted cells with characteristic morphology of the developing striatum (153). The graft occupied only 8–10% of the patient’s striatal volume. There was no clear histological evidence of an immune rejection. Neuronal protein aggregates of mutant huntingtin were not found within the transplanted embryonic tissue. There were several grafted neurons exhibiting immunoreactivity for DARPP-32, calretinin, calbindin, enkephalin, and substance P, and areas of the grafts were also stained by acetylcholinesterase histochemistry. The P-zones positive for these striatal neuronal markers occupied 36–56% of the surviving transplants (153), which is in the upper range of what had been observed with human LGE grafts placed in rats (141). As predicted from earlier animal studies, the transplants were innervated by TH-immunoreactive fibers derived from the host. Perhaps, the greatest attention in the field of clinical neural transplantation has been given to the study coming from a French group based in Creteil because it provides tentative evidence for graft function coupled to transplant survival. They implanted small pieces of LGE and WGE tissue bilaterally into the putamen. The majority of the patients received tissue first into the nondominant hemisphere and 1 year later into the putamen in the contralateral hemisphere (154). Immunosuppressive treatment with cyclosporine A was given postoperatively for at least 6 months with the addition of prednisolone and azathioprine for 1 year (154–156). Four patients out of the five who received transplants demonstrated motor and cognitive improvements. These functional benefits appeared about 1 year after the first neural implantation and persisted even 3 years postgrafting. The PET-scan evaluations showed increased metabolic activity in the striatum in three of five graft recipients (156). Contrary to that, in the two patients with no functional motor and cognitive improvements by the grafts, striatal and cortical
378
Popovic et al.
hypometabolism progressed over the 2-year follow-up (157). In one patient, the disappearance of signal in the magnetic resonance imaging ascribed to surviving graft tissue coincided in time with deterioration of a functional improvement that had lasted around 7 months after surgery (154–156). 6.
RISK FACTORS OF NEURAL TRANSPLANTATION
Implantation of primary human fetal tissue is accompanied with several risks that can be deleterious for the final result of the transplantation. (1) Graft recipients are exposed to the possibility of being infected with pathological viruses, bacteria, fungi, or prions transferred from donor tissue. Donors at risk of being infected with sexually transmitted viruses should be excluded from transplantation program. In addition, careful viral screening of maternal donor serum, including HIV and hepatitis A, B, and C, as well as maternal vaginal flora for bacterial and fungal infection, should be performed in the procedure of identifying adequate donors. (2) Donor tissue can be contaminated with inappropriate cells due to inaccurate dissection and handling of donor tissue. In addition to that, there is a possibility that the graft can expand so much that it would cause further neurological dysfunction. (3) Transplantation is accompanied with a risk for rejection of grafted tissue that could induce further neurological and overall deteriorations of the recipients. In addition to that, immunosuppressive therapy, which is necessary to enable graft survival against recipient immune response, can induce toxic side effects with severe consequences on the health conditions of the recipient. (4) Surgical operation involves risk of different general complications for patient’s health and life, including cerebral hemorrhage after stereotaxic manipulations. Although groups in Los Angeles, France, and U.K. did not report complications in the total of 12 operated patients, a group in Florida found that three among seven graft recipients had subdural hematomas. (5) Psychiatric disturbances, such as depression, anxiety, and aggressivity, are common in HD patients. These alterations can be augmented after transplantation procedure, especially in the first postoperative period where the lack on improvement can be expected. Irritability and impulsiveness can induce further problems in family and close relatives. 7. NEUROGENESIS IN HD 7.1. Neurogenesis in Patients Neurogenesis, i.e., proliferation and maturation of new neurons, has been shown to occur at a significant rate in specific regions of the adult brain; i.e., the olfactory bulb, the subependymal layer (SEL), and the dentate gyrus of the hippocampus (for review, see Refs. 158–160). Although the full functional impact of neurogenesis is not yet clarified, it opens up possibilities of both new pathogenetic mechanisms and potential therapies for
Stem Cells and Huntington’s Disease
379
neurodegenerative disorders. Neurogenesis has been found to be a highly dynamic process with a number of controlling factors. It can be upregulated by enriched environment, exercise, learning, estrogen, antidepressant drugs, electroconvulsive therapy, and growth factors such as BDNF (161) and insulin growth factor (IGF-1). It is downregulated by stress, glucocorticoids, inflammation, excitotoxicity, opiate intake, and during aging. A number of pharmacological approaches, including growth factor delivery, could therefore potentially augment the proliferative capacity of endogenous neural progenitor cells, which may slow disease progression or even induce clinical recovery. One of the advantages of this approach is that interventions can start prior to the neuronal death and thereby prolong the life of the affected neuronal populations. However, a recent study has indicated that there is an enhanced neurogenesis in the postmortem analyzed brain from HD patients. A significant increase in the number of proliferating cells was found in the SEL adjacent to the caudate nucleus compared with age-matched control brains. Among the proliferating cells were both immature neurons and glial cells (162). It is not known why there is an enhanced neurogenesis in the SEL of the HD patients. It has been shown that pathological conditions such as electrical stimulation (epileptic seizures) and cerebral ischemia (163–167) enhance growth factor levels. However, BDNF levels have been reported to be reduced in HD brains (168), rendering this explanation less likely. Further studies are needed to explore to what extent neurogenesis is altered in the human HD brain and to elucidate the underlying mechanisms, which may show to be important in the overall pathogenetic process. 7.2.
Neurogenesis in a Transgenic Mouse Model of HD
We have recently found that there is reduced neurogenesis in the R6 line of HD mice (169,170). These results were obtained by labeling proliferating cells with a 12-hr cycle of daily injections of BrdU lasting for 3–6 days in these mice (Fig. 3). We found that there was a reduction in neurogenesis in the dentate gyrus, which commenced already at a presymptomatic stage and then progressed as the disease advanced. This area has never been studied with regard to neurogenesis in HD patients. In contrast to findings in the postmortem HD patient brains, the neurogenesis in the subventricular zone was at a symptomatic stage similar in the R6 mice compared to their wild-type littermates. The underlying mechanism that causes the reduced neurogenesis in the HD mice is still under investigation, but it is possible that a reduced level of BDNF may play a role (Popovic et al., unpublished observations). 8.
CONCLUSION
Stem cells may contribute to the therapy in HD in the future. The two main strategical options are cell transplantation and manipulation of neurogenesis
380
Popovic et al.
Figure 3 Decreased neurogenesis in R6=2 HD transgenic mice. Photographs of brain sections from the dentate gyrus in the hippocampus of 12-week-old R6=2 and control wild-type littermates. The sections are processed for BrdU in wild-type (A) and R6=2 (B) mice, demonstrating a substantial decrease in the number of labeled cells in the R6=2 mice. Adjacent sections were stained for the neuronal marker NeuN in wild-type (C) and R6=2 (D) mice to show the structure of the dentate gyrus.
in the adult brain. Clinical trials and experimental studies indicate that transplantation of human embryonic striatal cells is a feasible procedure and there is convincing histological evidence for survival of transplanted cells. However, there are still many remaining obstacles related to the further implementation of a grafting protocol for HD. Many of these obstacles would be easier to overcome if there was a stem-cell alternative source for donor tissue. These cells should be available in large numbers and reproducibly express the most important phenotypic features of medium-sized spiny striatal neurons. If such cells could be differentiated from stem cells, improvement of the surgical procedure with assessment of proper number and location of the fetal striatal implants would increase the success of the final outcome. Appropriate immunosuppressive therapy and potential cotreatment with trophic factors, antioxidants, or caspase inhibitors might also be critical for the survival of transplanted grafts, as has been found to be the case for transplants in Parkinsons disease (171). Regarding the option of using neurogenesis to promote repair in the HD brain, this is an area that requires a great deal of further investigation before clinical trials can be considered. First, it is necessary to clarify the partially contradictory findings between a decrease in hippocampal neurogenesis in a transgenic mouse HD model and the one published study reporting increased striatal neurogenesis in
Stem Cells and Huntington’s Disease
381
postmortem samples from patients. Second, factors that regulate the proliferation, migration, and differentiation of stem cells in the brain need a deeper understanding. It is not until basic science has reached this stage that we can even consider translation of the exciting putative therapeutic strategy of manipulating neurogenesis into the first clinical trials.
REFERENCES 1.
Gusella JF, Wexler NS, Conneally PM, Naylor SL, Anderson MA, Tanzi RE, Watkins PC, Ottina K, Wallace MR, Sakaguchi AY, et al. A polymorphic DNA marker genetically linked to Huntington’s disease. Nature 1983; 306:234–238. 2. Group THsDCR. A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington’s disease chromosomes. The Huntington’s Disease Collaborative Research Group. Cell 1993; 72:971–983. 3. Snell RG, MacMillan JC, Cheadle JP, Fenton I, Lazarou LP, Davies P, MacDonald ME, Gusella JF, Harper PS, Shaw DJ. Relationship between trinucleotide repeat expansion and phenotypic variation in Huntington’s disease. Nat Genet 1993; 4:393–397. 4. Gusella JF, MacDonald ME, Ambrose CM, Duyao MP. Molecular genetics of Huntington’s disease. Arch Neurol 1993; 50:1157–1163. 5. Brandt J, Bylsma FW, Gross R, Stine OC, Ranen N, Ross CA. Trinucleotide repeat length and clinical progression in Huntington’s disease. Neurology 1996; 46:527–531. 6. Trottier Y, Biancalana V, Mandel JL. Instability of CAG repeats in Huntington’s disease: relation to parental transmission and age of onset. J Med Genet 1994; 31:377–382. 7. DiFiglia M, Sapp E, Chase K, Schwarz C, Meloni A, Young C, Martin E, Vonsattel JP, Carraway R, Reeves SA, et al. Huntingtin is a cytoplasmic protein associated with vesicles in human and rat brain neurons. Neuron 1995; 14:1075–1081. 8. Sharp AH, Loev SJ, Schilling G, Li SH, Li XJ, Bao J, Wagster MV, Kotzuk JA, Steiner JP, Lo A, et al. Widespread expression of Huntington’s disease gene (IT15) protein product. Neuron 1995; 14:1065–1074. 9. Trottier Y, Devys D, Imbert G, Saudou F, An I, Lutz Y, Weber C, Agid Y, Hirsch EC, Mandel JL. Cellular localization of the Huntington’s disease protein and discrimination of the normal and mutated form. Nat Genet 1995; 10:104–110. 10. Gutekunst CA, Levey AI, Heilman CJ, Whaley WL, Yi H, Nash NR, Rees HD, Madden JJ, Hersch SM. Identification and localization of huntingtin in brain and human lymphoblastoid cell lines with anti-fusion protein antibodies. Proc Natl Acad Sci USA 1995; 92:8710–8714. 11. Tukamoto T, Nukina N, Ide K, Kanazawa I. Huntington’s disease gene product, huntingtin, associates with microtubules in vitro. Brain Res Mol Brain Res 1997; 51:8–14.
382
Popovic et al.
12. Li J-Y, Plomann M, Brundin P. Huntington’s disease: a synaptopathy? Trends Mol Med 2003; 9:414–420 13. Li SH, Schilling G, Young WS III, Li XJ, Margolis RL, Stine OC, Wagster MV, Abbott MH, Franz ML, Ranen NG, et al. Huntington’s disease gene (IT15) is widely expressed in human and rat tissues. Neuron 1993; 11:985–993. 14. DiFiglia M, Sapp E, Chase KO, Davies SW, Bates GP, Vonsattel JP, Aronin N. Aggregation of huntingtin in neuronal intranuclear inclusions and dystrophic neurites in brain. Science 1997; 277:1990–1993. 15. Vonsattel JP, Myers RH, Stevens TJ, Ferrante RJ, Bird ED, Richardson EP Jr. Neuropathological classification of Huntington’s disease. J Neuropathol Exp Neurol 1985; 44:559–577. 16. Kremer HP, Roos RA, Dingjan G, Marani E, Bots GT. Atrophy of the hypothalamic lateral tuberal nucleus in Huntington’s disease. J Neuropathol Exp Neurol 1990; 49:371–382. 17. Kremer HP, Roos RA, Dingjan GM, Bots GT, Bruyn GW, Hofman MA. The hypothalamic lateral tuberal nucleus and the characteristics of neuronal loss in Huntington’s disease. Neurosci Lett 1991; 132:101–104. 18. Huang CC, Faber PW, Persichetti F, Mittal V, Vonsattel JP, MacDonald ME, Gusella JF. Amyloid formation by mutant huntingtin: threshold, progressivity and recruitment of normal polyglutamine proteins. Somat Cell Mol Genet 1998; 24:217–233. 19. Kazantsev A, Preisinger E, Dranovsky A, Goldgaber D, Housman D. Insoluble detergent-resistant aggregates form between pathological and nonpathological lengths of polyglutamine in mammalian cells. Proc Natl Acad Sci USA 1999; 96:11404–11409. 20. Wheeler VC, White JK, Gutekunst CA, Vrbanac V, Weaver M, Li XJ, Li SH, Yi H, Vonsattel JP, Gusella JF, Hersch S, Auerbach W, Joyner AL, MacDonald ME. Long glutamine tracts cause nuclear localization of a novel form of huntingtin in medium spiny striatal neurons in HdhQ92 and HdhQl11 knock-in mice. Hum Mol Genet 2000; 9:503–513. 21. Cattaneo E, Rigamonti D, Goffredo D, Zuccato C, Squitieri F, Sipione S. Loss of normal huntingtin function: new developments in Huntington’s disease research. Trends Neurosci 2001; 24:182–188. 22. Ciechanover A, Brundin P. The ubiquitin proteasome system in neurodegenerative diseases: sometimes the chicken, sometimes the egg. Neuron 2003; 40:427–446. 23. Boll TJ, Heaton R, Reitan RM. Neuropsychological and emotional correlates of Huntington’s chorea. J Nerv Ment Dis 1974; 158:61–69. 24. Josiassen RC, Curry LM, Mancall EL. Development of neuropsychological deficits in Huntington’s disease. Arch Neurol 1983; 40:791–796. 25. Bylsma FW, Brandt J, Strauss ME. Personal and extrapersonal orientation in Huntington’s disease patients and those at risk. Cortex 1992; 28:113–122. 26. Albert MS, Butters N, Brandt J. Development of remote memory loss in patients with Huntington’s disease. J Clin Neuropsychol 1981; 3:1–12. 27. Rosenberg NK, Sorensen SA, Christensen AL. Neuropsychological characteristics of Huntington’s disease carriers: a double blind study. J Med Genet 1995; 32:600–604.
Stem Cells and Huntington’s Disease
383
28. Bamford KA, Caine ED, Kido DK, Cox C, Shoulson I. A prospective evaluation of cognitive decline in early Huntington’s disease: functional and radiographic correlates. Neurology 1995; 45:1867–1873. 29. Quarrell N, Harper P. The clinical neurology of Huntington’s disease. In: Harper P, ed. Huntington’s Disease. Cambridge: WB Saunders Company LTD, University Press, 1996:31–72. 30. Sanberg PR, Fibiger HC, Mark RF. Body weight and dietary factors in Huntington’s disease patients compared with matched controls. Med J Aust 1981; 1:407–409. 31. Pratley RE, Salbe AD, Ravussin E, Caviness JN. Higher sedentary energy expenditure in patients with Huntington’s disease. Ann Neurol 2000; 47:64–70. 32. Hughes RE, Olson JM. Therapeutic opportunities in polyglutamine disease. Nat Med 2001; 7:419–423. 33. Dedeoglu A, Kubilus JK, Yang L, Ferrante KL, Hersch SM, Beal MF, Ferrante RJ. Creatine therapy provides neuroprotection after onset of clinical symptoms in Huntington’s disease transgenic mice. J Neurochem 2003; 85: 1359–1367. 34. Verbessem P, Lemiere J, Eijnde BO, Swinnen S, Vanhees L, Van Leemputte M, Hespel P, Dom R. Creatine supplementation in Huntington’s disease: a placebo–controlled pilot trial. Neurology 2003; 61:925–930. 35. Group HS. A randomized, placebo-controlled trial of coenzyme Q10 and remacemide in Huntington’s disease. Neurology 2001; 57:397–404. 36. Deckel AW, Robinson RG, Coyle JT, Sanberg PR. Reversal of long-term locomotor abnormalities in the kainic acid model of Huntington’s disease by 18 fetal striatal implants. Eur J Pharmacol 1983; 93:287–288. 37. Isacson O, Brundin P, Kelly PA, Gage FH, Bjo¨rklund A. Functional neuronal replacement by grafted striatal neurones in the ibotenic acid-lesioned rat striatum. Nature 1984; 311:458–460. ¨ berg 38. Smart IHM, Sturrock RR. Ontogeny of the neostriatum. In: Divac I, O RGE., ed. The Neostriatum. Oxford: Pergamon Press, 1979:127–146. 39. Nakao N, Grasbon-Frodl EM, Widner H, Brundin P. DARPP-32-rich zones in grafts of lateral ganglionic eminence govern the extent of functional recovery in skilled paw reaching in an animal model of Huntington’s disease. Neuroscience 1996; 74:959–970. 40. Pakzaban P, Deacon TW, Burns LH, Isacson O. Increased proportion of acetylcholinesterase-rich zones and improved morphological integration in host striatum of fetal grafts derived from the lateral but not the medial ganglionic eminence. Exp Brain Res 1993; 97:13–22. 41. Deacon TW, Pakzaban P, Isacson O. The lateral ganglionic eminence is the origin of cells committed to striatal phenotypes: neural transplantation and developmental evidence. Brain Res 1994; 668:211–219. 42. Olsson M, Campbell K, Wictorin K, Bjo¨rklund A. Projection neurons in fetal striatal transplants are predominantly derived from the lateral ganglionic eminence. Neuroscience 1995; 69:1169–1182. 43. Olsson M, Bjo¨rklund A, Campbell K. Early specification of striatal projection neurons and interneuronal subtypes in the lateral and medial ganglionic eminence. Neuroscience 1998; 84:867–876.
384
Popovic et al.
44. Bayer SA. Neurogenesis in the rat neostriatum. Int J Dev Neurosci 1984; 2:163–175. 45. Marchand R, Lajoie L. Histogenesis of the striopallidal system in the rat. Neurogenesis of its neurons. Neuroscience 1986; 17:573–590. 46. Parent J, Vexler ZS, Gong C, Derugin N, Ferriero DM. Rat forebrain neurogenesis and striatal neuron replacement after focal stroke. Ann Neurol 2002; 52:802–813. 47. Stenman J, Toresson H, Campbell K. Identification of two distinct progenitor populations in the lateral ganglionic eminence: implications for striatal and olfactory bulb neurogenesis. J Neurosci 2003; 23:167–174. 48. Bjo¨rklund A, Campbell K, Sirinathsinghji DJS, Fricker RA, Dunnett SB. Functional capacity of striatal transplants in the rat Huntington model. In: Dunnett SB, Bjo¨rklund A, eds. Functional Neural Transplantation. New York: Raven Press, 1994:157–195. 49. Wictorin K. Anatomy and connectivity of intrastriatal striatal transplants. Prog Neurobiol 1992; 38:611–639. 50. Fricker RA, Torres EM, Dunnett SB. The effects of donor stage on the survival and function of embryonic striatal grafts in the adult rat brain. I. Morphological characteristics. Neuroscience 1997; 79:695–710. 51. Isacson O, Brundin P, Gage FH, Bjo¨rklund A. Neural grafting in a rat model of Huntington’s disease: progressive neurochemical changes after neostriatal ibotenate lesions and striatal tissue grafting. Neuroscience 1985; 16:799–817. 52. Isacson O, Dawbarn D, Brundin P, Gage FH, Emson PC, Bjo¨rklund A. Neural grafting in a rat model of Huntington’s disease: striosomal-like organization of striatal grafts as revealed by acetylcholinesterase histochemistry, immunocytochemistry and receptor autoradiography. Neuroscience 1987; 22:481–497. 53. Roberts RC, DiFiglia M. Localization of immunoreactive GABA and enkephalin and NADPH-diaphorase-positive neurons in fetal striatal grafts in the quinolinic-acid-lesioned rat neostriatum. J Comp Neurol 1988; 274:406–421. 54. Roberts RC, DiFiglia, M. Long-term survival of GABA-, enkephalin-, NADPH-diaphorase and calbindin-D28k-containing neurons. Brain Res 1990; 532:151–159. 55. Graybiel AM, Liu FC, Dunnett SB. Intrastriatal grafts derived from fetal striatal primordia. I. Phenotype and modular organization. J Neurosci 1989; 9:3250–3271. 56. Zhou FC, Buchwald N. Connectivities of the striatal grafts in adult rat brain: a rich afference and scant striatonigral efference. Brain Res 1989; 504:15–30. 57. Giordano M, Ford LM, Shipley MT, Sanberg PR. Neural grafts and pharmacological intervention in a model of Huntington’s disease. Brain Res Bull 1990; 25:453–465. 58. Liu FC, Graybiel AM, Dunnett SB, Baughman RW. Intrastriatal grafts derived from fetal striatal primordia: II. Reconstitution of cholinergic and dopaminergic systems. J Comp Neurol 1990; 295:1–14. 59. Sirinathsinghji DJ, Wisden W, Northrop A, Hunt SP, Dunnett SB, Morris BJ. Cellular localisation of neurotransmitter mRNAs in striatal grafts. Prog Brain Res 1990; 82:433–439.
Stem Cells and Huntington’s Disease
385
60. Deckel AW, Moran TH, Robinson RG. Receptor characteristics and recovery of function following kainic acid lesions and fetal transplants of the striatum. II. Dopaminergic systems. Brain Res 1988; 474:39–47. 61. Mayer E, Heavens RP, Sirinathsinghji DJ. Autoradiographic localisation of Dl and D2 dopamine receptors in primordial striatal tissue grafts in rats. Neurosci Lett 1990; 109:271–276. 62. Helm GA, Robertson MW, Jallo GI, Simmons N, Bennett JP Jr. Development of Dl and D2 dopamine receptors and associated second messenger systems in fetal striatal transplants. Exp Neurol 1991; 111:181–189. 63. Campbell K, Wictorin K, Bjo¨rklund A. Neurotransmitter-related gene expression in intrastriatal striatal transplants—I. Phenotypical characterization of striatal and non-striatal graft regions. Neuroscience 1995; 64:17–33. 64. Nakao N, Itakura T. Fetal tissue transplants in animal models of Huntington’s disease: the effects on damaged neuronal circuitry and behavioral deficits. Prog Neurobiol 2000; 61:313–338. 65. Watts C, Dunnett SB. Towards a protocol for the preparation and delivery of striatal tissue for clinical trials of transplantation in Huntington’s disease. Cell Transplant 2000; 9:223–234. 66. Watts C, McNamara IR, Dunnett SB. Volume and differentiation of striatal grafts in rats: relationship to the number of cells implanted. Cell Transplant 2000; 9:65–72. ˚ , Emga˚rd M, Dunnett SB, Brundin P. Caspase 67. Mundt-Petersen U, Petersen A inhibition increases embryonic striatal graft survival. Exp Neurol 2000; 164:112–120. 68. DiFiglia M, Schiff L, Deckel AW. Neuronal organization of fetal striatal grafts in kainate- and sham-lesioned rat caudate nucleus: light- and electron-microscopic observations. J Neurosci 1988; 8:1112–1130. 69. Watts C, Dunnett SB. Effects of severity of host striatal damage on the morphological development of intrastriatal transplants in a rodent model of Huntington’s disease: implications for timing of surgical intervention. J Neurosurg 1998; 89:267–274. 70. Petersen A, Brundin P. Effects of ciliary neurotrophic factor on excitotoxicity and calcium-ionophore A23187-induced cell death in cultured embryonic striatal neurons. Exp Neurol 1999; 160:402–412. 71. Frim DM, Short MP, Rosenberg WS, Simpson J, Breakefield XO, Isacson O. Local protective effects of nerve growth factor-secreting fibroblasts against excitotoxic lesions in the rat striatum. J Neurosurg 1993; 78:267–273. 72. Frim DM, Uhler TA, Short MP, Ezzedine ZD, Klagsbrun M, Breakefield XO, Isacson O. Effects of biologically delivered NGF, BDNF and bFGF on striatal excitotoxic lesions. Neuroreport 1993; 4:367–370. 73. Emerich DF, Hammang JP, Baetge EE, Winn SR. Implantation of polymerencapsulated human nerve growth factor-secreting fibroblasts attenuates the behavioral and neuropathological consequences of quinolinic acid injections into rodent striatum. Exp Neurol 1994; 130:141–150. 74. Martinez-Serrano A, Bjo¨rklund A. Protection of the neostriatum against excitotoxic damage by neurotrophin-producing, genetically modified neural stem cells. J Neurosci 1996; 16:4604–4616.
386
Popovic et al.
75. Kordower JH, Chen EY, Mufson EJ, Winn SR, Emerich DF. Intrastriatal implants of polymer encapsulated cells genetically modified to secrete human nerve growth factor: trophic effects upon cholinergic and noncholinergic striatal neurons. Neuroscience 1996; 72:63–77. 76. Kordower JH, Chen EY, Winkler C, Fricker R, Charles V, Messing A, Mufson EJ, Wong SC, Rosenstein JM, Bjo¨rklund A, Emerich DF, Hammang J, Carpenter MK. Grafts of EGF-responsive neural stem cells derived from GFAP-hNGF transgenic mice: trophic and tropic effects in a rodent model of Huntington’s disease. J Comp Neurol 1997; 387:96–113. 77. Perez-Navarro E, Arenas E, Marco S, Alberch J. Intrastriatal grafting of a GDNF-producing cell line protects striatonigral neurons from quinolinic acid excitotoxicity in vivo. Eur J Neurosci 1999; 11:241–249. 78. Emerich DF, Lindner MD, Winn SR, Chen EY, Frydel BR, Kordower JH. Implants of encapsulated human CNTF-producing fibroblasts prevent behavioral deficits and striatal degeneration in a rodent model of Huntington’s disease. J Neurosci 1996; 16:5168–5181. 79. Emerich DF, Cain CK, Greco C, Saydoff JA, Hu ZY, Liu H, Lindner MD. Cellular delivery of human CNTF prevents motor and cognitive dysfunction in a rodent model of Huntington’s disease. Cell Transplant 1997; 6: 249–266. 80. Emerich DF, Winn SR, Hantraye PM, Peschanski M, Chen EY, Chu Y, McDermott P, Baetge EE, Kordower JH. Protective effect of encapsulated cells producing neurotrophic factor CNTF in a monkey model of Huntington’s disease. Nature 1997; 386:395–399. 81. Martinez-Serrano A, Bjo¨rklund A. Immortalized neural progenitor cells for CNS gene transfer and repair. Trends Neurosci 1997; 20:530–538. 82. Fricker RA, Carpenter MK, Winkler C, Greco C, Gates MA, Bjo¨rklund A. Site-specific migration and neuronal differentiation of human neural progenitor cells after transplantation in the adult rat brain. J Neurosci 1999; 19: 5990–6005. 83. Englund U, Fricker-Gates RA, Lundberg C, Bjo¨rklund A, Wictorin K. Transplantation of human neural progenitor cells into the neonatal rat brain: extensive migration and differentiation with long-distance axonal projections. Exp Neurol 2002; 173:1–21. 83a. Bosch M, Pineda JR, Sunol C, Petriz J, Cattaneo E, Alberch J, Canais JM. Induction of GABAergic phenotype in a neural stem cell line for transplantation in an excitotoxic model of Huntington’s disease. Exp Neurol 2004; 190:42–58. 83b. McBride JL, Behrstock SP, Chen EY, Jakel RJ, Siegel I, Svendsen CN, Kordower JH. Human neural stem cell transplants improve motor function in a rat model of Huntington’s disease. J Comp Neurol 2004; 475:211–219. 84. Kendall AL, Rayment FD, Torres EM, Baker HF, Ridley RM, Dunnett SB. Functional integration of striatal allografts in a primate model of Huntington’s disease. Nat Med 1998; 4:727–729. 85. Pritzel M, Isacson O, Brundin P, Wiklund L, Bjo¨rklund A. Afferent and efferent connections of striatal grafts implanted into the ibotenic acid lesioned neostriatum in adult rats. Exp Brain Res 1986; 65:112–126.
Stem Cells and Huntington’s Disease
387
86. Clarke DJ, Dunnett SB, Isacson O, Sirinathsinghji DJ, Bjo¨rklund A. Striatal grafts in rats with unilateral neostriatal lesions—I. Ultrastructural evidence of afferent synaptic inputs from the host nigrostriatal pathway. Neuroscience 1988; 24:791–801. 87. Wictorin K, Isacson O, Fisher W, Nothias FH, Peschanski M, Bjo¨rklund A. Connectivity of striatal grafts implanted into the ibotenic acid-lesioned striatum—I. Subcortical afferents. Neuroscience 1988; 27:547–562. 88. Wictorin K, Bjo¨rklund A. Connectivity of striatal grafts implanted into the ibotenic acid-lesioned striatum—II. Cortical afferents. Neuroscience 1989; 30: 297–311. 89. Sanberg PR, Zubrycki E, Ragozzino ME, Giordano M, Shipley MT. Tyrosine hydroxylase-positive fibers and neurons in transplanted striatal tissue in rats with quinolinic acid lesions of the striatum. Brain Res Bull 1990; 25:889–894. 90. Xu ZC, Wilson CJ, Emson PC. Restoration of the corticostriatal projection in rat neostriatal grafts: electron microscopic analysis. Neuroscience 1989; 29: 539–550. 91. Clarke DJ, Dunnett SB. Synaptic relationships between cortical and dopaminergic inputs and intrinsic GABAergic systems within intrastriatal striatal grafts. J. Chem. Neuroanat 1993; 6:147–158. 92. Wictorin K, Clarke DJ, Bolam JP, Bjo¨rklund A. Host corticostriatal fibres establish synaptic connections with grafted striatal neurons in the ibotenic acid lesioned striatum. Eur J Neurosci 1989; 1:189–195. 93. Wictorin K, Ouimet CC, Bjo¨rklund A. Intrinsic organization and connectivity of intrastriatal striatal transplants in rats as revealed by DARPP-32 immunohistochemistry: specificity of connections with the lesioned host brain. Eur J Neurosci 1989; 1:690–701. 94. Wictorin K, Clarke DJ, Bolam JP, Bjo¨rklund A. Fetal striatal neurons grafted into the ibotenate lesioned adult striatum: efferent projections and synaptic contacts in the host globus pallidus. Neuroscience 1990; 37:301–315. 95. Dunnett SB, Svendsen CN. Huntington’s disease: animal models and transplantation repair. Curr Opin Neurobiol 1993; 3:790–796. 96. Mandel RJ, Wictorin K, Cenci MA, Bjo¨rklund A. Fos expression in intrastriatal striatal grafts: regulation by host dopaminergic afferents. Brain Res 1992; 583:207–215. 97. Rutherford A, Garcia-Munoz M, Dunnett SB, Arbuthnott GW. Electrophysiological demonstration of host cortical inputs to striatal grafts. Neurosci Lett 1987; 83:275–281. 98. Labandeira-Garcia JL, Guerra MJ. Cortical stimulation induces fos expression in intrastriatal striatal grafts. Brain Res 1994; 652:87–97. 99. Cenci MA, Kalen P, Mandel RJ, Wictorin K, Bjo¨rklund A. Dopaminergic transplants normalize amphetamine- and apomorphine-induced Fos expression in the 6-hydroxydopamine-lesioned striatum. Neuroscience 1992; 46: 943–957. 100. Nakao N, Nakai K, Itakura T. Changes in the mitochondrial enzyme activity in striatal projection areas after unilateral excitotoxic striatal lesions: partial restoration by embryonic striatal transplants. Exp Neurol 1998; 153: 268–276.
388
Popovic et al.
101. Nakao N, Ogura M, Nakai K, Itakura T. Embryonic striatal grafts restore neuronal activity of the globus pallidus in a rodent model of Huntington’s disease. Neuroscience 1999; 88:469–477. 102. Campbell K, Kalen P, Wictorin K, Lundberg C, Mandel RJ, Bjo¨rklund A. Characterization of GABA release from intrastriatal striatal transplants: dependence on host-derived afferents. Neuroscience 1993; 54:403–415. 103. Deckel AW, Moran TH, Coyle JT, Sanberg PR, Robinson RG. Anatomical predictors of behavioral recovery following fetal striatal transplants. Brain Res 1986; 365:249–258. 104. Deckel AW, Moran TH, Robinson RG. Behavioral recovery following kainic acid lesions and fetal implants of the striatum occurs independent of dopaminergic mechanisms. Brain Res 1986; 363:383–385. 105. Deckel AW, Moran TH, Robinson RG. Receptor characteristics and recovery of function following kainic acid lesions and fetal transplants of the striatum. I. Cholinergic systems. Brain Res 1988; 474:27–38. 106. Isacson O, Dunnett SB, Bjo¨rklund A. Graft-induced behavioral recovery in an animal model of Huntington disease. Proc Natl Acad Sci USA 1986; 83: 2728–2732. 107. Sanberg PR, Henault MA, Deckel AW. Locomotor hyperactivity: effects of multiple striatal transplants in an animal model of Huntington’s disease. Pharmacol Biochem Behav 1986; 25:297–300. 108. Sanberg PR, Calderon SF, Garver DL, Norman AB. Brain tissue transplants in an animal model of Huntington’s disease. Psychopharmacol Bull 1987; 23: 476–482. 109. Sanberg PR, Henault MA, Hagenmeyer-Houser SH, Giordano M, Russell KH. Multiple transplants of fetal striatal tissue in the kainic acid model of Huntington’s disease. Behavioral recovery may not be related to acetylcholinesterase. Ann NY Acad Sci 1987; 495:781–785. 110. Sanberg PR, Giordano M, Henault MA, Nash DR, Ragozzino ME, Hagenmeyer-Houser SH. Intraparenchymal striatal transplants required for maintenance of behavioral recovery in an animal model of Huntington’s disease. J Neural Transplant 1989; 1:23–31. 111. Giordano M, Hagenmeyer-Houser SH, Sanberg PR. Intraparenchymal fetal striatal transplants and recovery in kainic acid lesioned rats. Brain Res 1988; 446:183–188. 112. Emerich DF, Zubricki EM, Shipley MT, Norman AB, Sanberg PR. Female rats are more sensitive to the locomotor alterations following quinolinic acid-induced striatal lesions: effects of striatal transplants. Exp Neurol 1991; 111:369–378. 113. Norman AB, Calderon SF, Giordano M, Sanberg PR. Striatal tissue transplants attenuate apomorphine-induced rotational behavior in rats with unilateral kainic acid lesions. Neuropharmacology 1988; 27:333–336. 114. Norman AB, Giordano M Sanberg PR. Fetal striatal tissue grafts into excitotoxin-lesioned striatum: pharmacological and behavioral aspects. Pharmacol Biochem Behav 1989; 34:139–147. 115. Dunnett SB, Isacson O, Sirinathsinghji DJS, Clarl DJ, Bjo¨rklund A. Striatal grafts in rats with unilateral neostriatal lesions. Part III. Recovery from dopa-
Stem Cells and Huntington’s Disease
116.
117. 118.
119.
120.
121. 122.
123.
124.
125.
126.
127.
128.
389
mine-dependent motor asymmetry and deficits in skilled paw reaching. Neuroscience 1998; 24:813–820. Whishaw IQ, O’Connor WT, Dunnett SB. The contributions of motor cortex, nigrostriatal dopamine and caudate–putamen to skilled forelimb use in the rat. Brain 1986; 109:805–843. Montoya CP, Astell S, Dunnett SB. Effects of nigral and striatal grafts on skilled forelimb use in the rat. Prog Brain Res 1990; 82:459–466. Valouskova V, Bracha V, Bures J, Hernandez-Mesa N, Macias-Gonzales R, Mazurova Y, Nemecek S. Unilateral striatal grafts induce behavioral and electrophysiological asymmetry in rats with bilateral kainate lesions of the caudate nucleus. Behav Neurosci 1990; 104:671–680. Nakao N, Grasbon-Frodl EM, Widner H, Brundin P. Antioxidant treatment protects striatal neurons against excitotoxic insults. Neuroscience 1996; 73: 185–200. Fricker RA, Torres EM, Hume SP, Myers R, Opacka-Juffrey J, Ashworth S, Brooks DJ, Dunnett SB. The effects of donor stage on the survival and function of embryonic striatal grafts in the adult rat brain. II. Correlation between positron emission tomography and reaching behaviour. Neuroscience 1997; 79:711–721. Dobrossy MD, Dunnett SB. Motor training effects on recovery of function after striatal lesions and striatal grafts. Exp Neurol 2003; 184:274–284. Mayer E, Brown VJ, Dunnett SB, Robbins TW. Striatal grafts-associated recovery of a lesion-induced performance deficit in the rat requires learning to use the transplant. Eur J Neurosci 1992; 4:119–126. Hansson O, Castilho RF, Korhonen L, Lindholm D, Bates GP, Brundin P. Partial resistance to malonate-induced striatal cell death in transgenic mouse models of Huntington’s disease is dependent on age and CAG repeat length. J Neurochem 2001; 78:694–703. Davies SW, Turmaine M, Cozens BA, DiFiglia M, Sharp AH, Ross CA, Scherzinger E, Wanker EE, Mangiarini L, Bates GP. Formation of neuronal intranuclear inclusions underlies the neurological dysfunction in mice transgenic for the HD mutation. Cell 1997; 90:537–548. Petersen A, Larsen KE, Behr GG, Romero N, Przedborski S, Brundin P, Sulzer D. Expanded CAG repeats in exon 1 of the Huntington’s disease gene stimulate dopamine-mediated striatal neuron autophagy and degeneration. Hum Mol Genet 2001; 10:1243–1254. Petersen A, Puschban Z, Lotharius J, NicNiocaill B, Wiekop P, O’Connor WT, Brundin P. Evidence for dysfunction of the nigrostriatal pathway in the R6=1 line of transgenic Huntington’s disease mice. Neurobiol Dis 2002; 11: 134–146. Behrens PF, Franz P, Woodman B, Lindenberg KS, Landwehrmeyer GB. Impaired glutamate transport and glutamate-glutamine cycling: downstream effects of the Huntington mutation. Brain 2002; 125:1908–1922. van Dellen A, Deacon R, York D, Blakemore C, Hannan AJ. Anterior cingulate cortical transplantation in transgenic Huntington’s disease mice. Brain Res Bull 2001; 56:313–318.
390
Popovic et al.
129. Emerich DF, Sanberg PR. Animal models of Huntington’s disease. In: Boulton A, Baker G, Butterworth R, eds. Neuromethods. Animal Models of Neurological Disease. Vol. 2. Clifton, NJ: Human Press, 1992:65–134. 130. Beal MF, Brouillet E, Jenkins BG, Ferrante RJ, Kowall NW, Miller JM, Storey E, Srivastava R, Rosen BR, Hyman BT. Neurochemical and histologic characterization of striatal excitotoxic lesions produced by the mitochondrial toxin 3-nitropropionic acid. J Neurosci 1993; 13:4181–4192. 131. Isacson O, Riche D, Hantraye P, Sofroniew MV, Maziere M. A primate model of Huntington’s disease: cross-species implantation of striatal precursor cells to the excitotoxically lesioned baboon caudate–putamen. Exp Brain Res 1989; 75:213–220. 132. Hantraye P, Riche D, Maziere M, Isacson O. Intrastriatal transplantation of cross-species fetal striatal cells reduces abnormal movements in a primate model of Huntington disease. Proc Natl Acad Sci USA 1992; 89:4187–4191. 133. Hantraye P, Riche D, Maziere M, Isacson O. A primate model of Huntington’s disease: behavioral and anatomical studies of unilateral excitotoxic lesions of the caudate–putamen in the baboon. Exp Neurol 1990; 108: 91–104. 134. Kanazawa I, Kimura M, Murata M, Tanaka Y, Cho F. Choreic movements in the macaque monkey induced by kainic acid lesions of the striatum combined with l-dopa. Pharmacological, biochemical, and physiological studies on neural mechanisms. Brain 1990; 113(Pt 2):509–535. 135. Besret L, Kendal AL, Dunnett SB. Aspects of PET imaging relevant to the assessment of striatal transplantation in Huntington’s disease. J Anat 1999; 196:597–607. 136. Kendall AL, Hantraye P, Palfi S. Striatal tissue transplantation in non-human primates. Prog Brain Res 2000; 127:381–404. 137. Brouillet E, Hantraye P, Ferrante RJ, Dolan R, Leroy-Willig A, Kowall NW, Beal MF. Chronic mitochondrial energy impairment produces selective striatal degeneration and abnormal choreiform movements in primates. Proc Natl Acad Sci USA 1995; 92:7105–7109. 138. Boer GJ. The Network of European CNS Transplantation and Restoration (NECTAR): an introduction on the occasion of its tenth meeting. Cell Transplant 2000; 9:133–137. 139. Wictorin K, Brundin P, Gustavii B, Lindvall O, Bjo¨rklund A. Reformation of long axon pathways in adult rat central nervous system by human forebrain neuroblasts. Nature 1990; 347:556–558. 140. Naimi S, Jeny R, Hantraye P, Peschanski M, Riche D. Ontogeny of human striatal DARPP-32 neurons in fetuses and following xenografting to the adult rat brain. Exp Neurol 1996; 137:15–25. 141. Grasbon-Frodl EM, Nakao N, Lindvall O, Brundin P. Phenotypic development of the human embryonic striatal primordium: a study of cultured and grafted neurons from the lateral and medial ganglionic eminences. Neuroscience 1996; 73:171–183. 142. Grasbon-Frodl EM, Nakao N, Lindvall O, Brundin P. Developmental features of human striatal tissue transplanted in a rat model of Huntington’s disease. Neurobiol Dis 1997; 3:299–311.
Stem Cells and Huntington’s Disease 143.
144.
145.
146.
147.
148. 149.
150.
151. 152.
153.
154.
155. 156.
391
Pundt LL, Kondoh T, Conrad JA, Low WC. Transplantation of human striatal tissue into a rodent model of Huntington’s disease: phenotypic expression of transplanted neurons and host-to-graft innervation. Brain Res Bull 1996; 39:23–32. Pundt LL, Kondoh T, Conrad JA, Low WC. Transplantation of human fetal striatum into a rodent model of Huntington’s disease ameliorates locomotor deficits. Neurosci Res 1996; 24:415–420. Peschanski M, Cesaro P, Hantraye P. Rationale for intrastriatal grafting of striatal neuroblasts in patients with Huntington’s disease. Neuroscience 1995; 68:273–285. Brundin P, Fricker RA, Nakao N. Paucity of P-zones in striatal grafts prohibit commencement of clinical trials in Huntington’s disease. Neuroscience 1996; 71:895–897. Fink JS, Schumacher JM, Ellias SL, Palmer EP, Saint-Hilaire M, Shannon K, Penn R, Starr P, VanHorne C, Kott HS, Dempsey PK, Fischman AJ, Raineri R, Manhart C, Dinsmore J, Isacson O. Porcine xenografts in Parkinson’s disease and Huntington’s disease patients: preliminary results. Cell Transplant 2000; 9:273–278. Peschanski M, Dunnett SB. Cell therapy for Huntington’s disease, the next step forward. Lancet Neurol 2002; 1:81. Kopyov OV, Jacques S, Lieberman A, Duma CM, Eagle KS. Safety of intrastriatal neurotransplantation for Huntington’s disease patients. Exp Neurol 1998; 149:97–108. Rosser AE, Barker RA, Harrower T, Watts C, Farrington M, Ho AK, Bumstein RM, Menon DK, Gillard JH, Pickard J, Dunnett SB. Unilateral transplantation of human primary fetal tissue in four patients with Huntington’s disease: NEST-UK safety report ISRCTN no 36485475. J Neurol Neurosurg Psychiatry 2002; 73:678–685. Hauser RA, Sandberg PR, Freeman TB, Stoessl AJ. Bilateral human fetal striatal transplantation in Huntington’s disease. Neurology 2002; 58:1704. Freeman TB, Sanberg PR, Isacson O. Development of the human striatum: implications for fetal striatal transplantation in the treatment of Huntington’s disease. Cell Transplant 1995; 4:539–545. Freeman TB, Cicchetti F, Hauser RA, Deacon TW, Li XJ, Hersch SM, Nauert GM, Sanberg PR, Kordower JH, Saporta S, Isacson O. Transplanted fetal striatum in Huntington’s disease: phenotypic development and lack of pathology. Proc Natl Acad Sci USA 2000; 97:13877–13882. Bachoud-Levi A, Bourdet C, Brugieres P, Nguyen JP, Grandmougin T, Haddad B, Jeny R, Bartolomeo P, Boisse MF, Barba GD, Degos JD, Ergis AM, Lefaucheur JP, Lisovoski F, Pailhous E, Remy P, Palfi S, Defer GL, Cesaro P, Hantraye P, Peschanski M. Safety and tolerability assessment of intrastriatal neural allografts in five patients with Huntington’s disease. Exp Neurol 2000; 161:194–202. Bachoud-Levi AC, Hantraye P, Peschanski M. Fetal neural grafts for Huntington’s disease: a prospective view. Mov Disord 2002; 17:439–444. Bachoud-Levi AC, Remy P, Nguyen JP, Brugieres P, Lefaucheur JP, Bourdet C, Baudic S, Gaura V, Maison P, Haddad B, Boisse MF, Grandmougin T,
392
157.
158.
159. 160. 161.
162.
163.
164.
165. 166.
167.
168.
169.
Popovic et al. Jeny R, Bartolomeo P, Dalla Barba G, Degos JD, Lisovoski F, Ergis AM, Pailhous E, Cesaro P, Hantraye P, Peschanski M. Motor and cognitive improvements in patients with Huntington’s disease after neural transplantation. Lancet 2000; 356:1975–1979. Gaura V, Bachoud-Levi AC, Ribeiro MJ, Nguyen JP, Frouin V, Baudic S, Brugieres P, Mangin JF, Boisse MF, Palfi S, Cesaro P, Samson Y, Hantraye P, Peschanski M, Remy P. Striatal neural grafting improves cortical metabolism in Huntington’s patients. Brain 2004; 127:65–72. Kuhn HG, Palmer TD, Fuchs E. Adult neurogenesis: a compensatory mechanism for neuronal damage. Eur Arch Psychiatry Clin Neurosci 2001; 251:152–158. Duman RS, Nakagawa S, Malberg J. Regulation of adult neurogenesis by antidepressant treatment. Neuropsychopharmacology 2001; 25:836–844. Kempermann G. Regulation of adult hippocampal neurogenesis—implications for novel theories of major depression. Bipolar Disord 2002; 4:17–33. Pencea V, Bingman KD, Wiegand SJ, Luskin MB. Infusion of brain-derived neurotrophic factor into the lateral ventricle of the adult rat leads to new neurons in the parenchyma of the striatum, septum, thalamus, and hypothalamus. J Neurosci 2001; 21:6706–6717. Curtis MA, Penney EB, Pearson AG, van Roon-Mom WM, Butterworth NJ, Dragunow M, Connor B, Faull RL. Increased cell proliferation and neurogenesis in the adult human Huntington’s disease brain. Proc Natl Acad Sci USA 2003; 100:9023–9027. Kokaia Z, Kelly ME, Elmer E, Kokaia M, Molntyre DC, Lindvall O. Seizureinduced differential expression of messenger RNAs for neurotrophins and their receptors in genetically fast and slow kindling rats. Neuroscience 1996; 75:197–207. Kokaia Z, Nawa H, Uchino H, Elmer E, Kokaia M, Carnahan J, Smith ML, Siesjo BK, Lindvall O. Regional brain-derived neurotrophic factor mRNA and protein levels following transient forebrain ischemia in the rat. Brain Res Mol Brain Res 1996; 38:139–444. Lindvall O, Kokaia Z, Bengzon J, Elmer E, Kokaia M. Neurotrophins and brain insults. Trends Neurosci 1994; 17:490–496. Arvidsson A, Kokaia Z, Lindvall O. N-methyl-d-aspartate receptor-mediated increase of neurogenesis in adult rat dentate gyrus following stroke. Eur J Neurosci 2001; 14:10–18. Arvidsson A, Colin T, Kirik D, Kokaia Z, Lindvall O. Neuronal replacement from endogenous precursors in the adult brain after stroke. Nat Med 2002; 8:963–970. Zuccato C, Ciammola A, Rigamonti D, Leavitt BR, Goffredo D, Conti L, MacDonald ME, Friedlander RM, Silani V, Hayden MR, Timmusk T, Sipione S, Cattaneo E. Loss of huntingtin-mediated BDNF gene transcription in Huntington’s disease. Science 2001; 293:493–498. ˚ . Asialoerythropoietin is Gil JM, Leist M, Popovic N, Brundin P, Peterse´n A not effective in the R6=2 line of Huntington’s disease mice. BMC Neurosci 2004; 5:17.
Stem Cells and Huntington’s Disease 170.
171.
393
Lazic SE, Grote H, Armstrong RJ, Blakemore C, Hannan AJ, van Dellen A, Barker RA. Decreased hippocampal cell proliferation in R6=1 Huntington’s mice. Neuroreport 2004; 15:811–813. ˚, Brundin P, Karlsson J, Emgard M, Schierle GS, Hansson O, Peterse´n A Castilho RF. Improving the survival of grafted dopaminergic neurons: a review over current approaches. Cell Transplant 2000; 9(2):179–195.
18 Plasticity, Cell Transplantation, and Brain Repair Maˇte´ D. Do¨bro¨ssy and Stephen B. Dunnett The Brain Repair Group, Cardiff School of Biosciences, Cardiff University, Cardiff, Wales, U.K.
Once development was ended, the founts of growth and regeneration of axons and dendrites dried up irrevocably. In adult centres, the nerve paths are something fixed and immutable, everything may die, nothing may be regenerated.
Santiago Ramon y Cajal, In Degeneration and Regeneration of the Nervous System (1928).
1.
INTRODUCTION
The concept that the central nervous system (CNS) loses its plastic characteristics after development and becomes a rigid network of permanently fixed neurones has been surrendering credibility progressively throughout the twentieth century. Hebb (1949) introduced the idea that behavioral adaptation throughout life might rely on activity-dependent modification of the configuration and strength of connections between neurones. Further experiments demonstrated that the brain’s chemistry and anatomy are indeed sculpted by the environment and experience suggesting that the adult brain 395
396
Do¨bro¨ssy and Dunnett
retains some of the plasticity it has widely been believed to have lost following the end of development and in the postmitotic state. However, ‘‘plasticity’’ must be defined in the context in which it is being used, because its presence or absence can be investigated at multiple levels of organization from the molecular level to changes in behavior of the whole organism. In this chapter, we examine the plasticity of the adult brain, damaged as a result of disease or injury, and we discuss the degree to which brain repair can be achieved by intracerebral transplantation of neurones from an external source into this altered and extraordinary state. By plasticity in this context, we refer to adaptive events affecting the morphology, anatomical connections, and the functions subserved by neurones. Particular attention is paid to the relationship between host and transplanted cells, the anatomical integration of the grafted cells, and how circuit reconstruction within the host brain might promote functional recovery. Furthermore, we examine the role of the environment and the experience of the individual suffering from brain damage or disease. Together, these data indicate that rehabilitative training has the potential to modify the functional plasticity both of the affected brain and of the transplants. The different mechanisms by which cells from different sources, and following different types of brain damage, can affect host function will also determine their dependence on training and experience for maximizing therapeutic outcome. During the period immediately following injury and during the early stages of disease, the brain is at its most plastic and privileged state, offering a window of opportunity for therapeutic intervention. In contrast to a pessimistic interpretation of Cajal’s dictum, cited above, spontaneous recovery following injury shows that the adult brain does retain the capacity at some level to reorganize itself functionally. In order to improve the outcome following brain damage, cell replacement therapy must both make use of the endogenous potential for recovery of the host and optimize the external circumstances associated with any intervention, such as the source and treatment of the cells, the timing of the intervention, and the experience and the environment of the recipient. The role of these factors in neural plasticity and transplant-induced brain repair is examined in more detail in the following sections.
2. CELL TRANSPLANTATION AND PLASTICITY 2.1. Cell Replacement Therapy The rigid nature of the adult brain, a requirement for long-term behavioral stability, is the price paid for the brain’s limited capacity to regenerate or to repair itself following injury. The non-regenerative properties are not intrinsic to the neurones of the CNS, which can regrow axons over long distances in a peripheral environment (1), but are due to a variety of growth-
Plasticity, Cell Transplantation, and Brain Repair
397
inhibiting mechanisms present in the brain (2–4). There is presently considerable optimism that the recent identification of key molecules blocking regeneration—such as nogo—might lead to a functional therapy to promote axonal regeneration. However, such treatments might be limited to injuries involving fiber tract damage and not be relevant to people suffering from diseases or injury where critical populations of neurones are themselves irrevocably lost. Cell replacement therapy has a self-explanatory objective: to repair the nervous system by introducing new cells that can replace the function of the compromised or lost cells. In particular, the alternative potential source(s) of cells to be used is an area of active exploration, particularly with the growing interest in stem cells. However, the current choice of cells used in the clinic is primary embryonic neurons, which have been used in the majority of the fundamental research into grafting. Neural tissue dissected from the developing brain can survive transplantation into the adult mammalian brain and spinal cord, integrate and connect with the host nervous system, and influence the host’s behavior. Indeed, several clinical trials have now shown that embryonic tissue grafts can alleviate symptoms—at least partially—in Parkinson’s disease (PD)(5,6), and related strategies are under evaluation for Huntington’s disease (HD), spinal cord injury, stroke, and other CNS disorders (7–12). 2.2.
Cellular and Biochemical Plasticity of the Adult Host and the Embryonic Graft
Stroke, brain traumas, or neurodegeneration are commonly modeled in the laboratory by using blood flow occlusion, percussion, and stab wound injuries, or toxin-induced cell death, respectively. These techniques result in damage and eventual loss of axonal projections and disruption of the behavioral functions that the affected axons subserved. All damage stimulates inherent mechanisms of readjustment—i.e., plasticity—at the cellular level, which are in balance with factors that counter growth and regeneration (2,3,13). The neuronal adaptations observed in the damaged brain are collateral and regenerative sprouting, involving the repopulation of lost synapses by neighboring intact neurones and the reinnervation of proximal denervated targets by damaged axons, respectively. These two types of sprouting are the most common examples of spontaneous plasticity, and can lead to either maladaptive connections or to functional recovery, dependent upon circumstances. However, any recovery or reorganization that occurs is limited by the cellular and molecular capacities of affected neurons—as well as of nearby intact neurons—to interpret external growth and inhibitory cues. This lies behind the different regenerative powers seen following damage both to the developing embryonic or neonatal brain and to the mature adult brain. The complement of receptors and intracellular
398
Do¨bro¨ssy and Dunnett
factors change with age, giving more stability but less regenerative potential to the adult CNS. Furthermore, the sprouting observed in the adult is confined to the immediate vicinity of the injury, and no reconnectivity is seen if the damaged axon is remote from the denervated target. The primary neural tissue used for intracerebral transplantation is usually collected at a different developmental stage to its host. For maximum growth potential, the tissue is typically harvested from embryonic brain, an environment much more permissive and less inhibitory than the adult brain. This could be the key as to why the embryonic grafts could be more sensitive to the growth-promoting and less sensitive to the inhibitory factors released by the damaged neuronal and non-neuronal cells of the adult brain. The cellular machinery of the embryonic neurones is programed for growth, particularly with respect to adhesion molecules and the absence of receptors for inhibitory molecules. Damage to the adult brain can result in the return to developmental expression patterns of some growth-associated molecules in axons, but the examples are limited and their expression is short-lasting. For example, GAP-43, an important constituent of the growth cone, is down-regulated after development in most neurones, but is upregulated following injury (14,15). However, during development GAP-43 is highly expressed and experimental data show that GAP-43 mRNA is abundantly present in embryonic striatal grafts as well during the period crucial for the establishment of synaptic connections. Interestingly, after about 4week post-transplantation, the GAP-43 expression profile of the graft assumes that of the adult host’s (16). Other intracellular proteins responsible for axonal growth and stability, like the cytoskeletal proteins of the microtubule family (e.g., b-tubulin III), or the proteins associated with the microtubule-based structures (e.g., MAP2ab), are present in different forms in embryonic and adult neurones and directly influence the neurone’s plasticity (17–20). The expression profiles of several other molecules that have important roles in the interaction of the axon with the environment also differ between the adult and the embryonic neurones. Integrins, communicating with the extracellular matrix, and neural cell adhesion molecules (NCAM), mediating cell-to-cell interaction, both play important roles in axonal growth in the embryo but have different, less growth conducive, subtypes present on adult neurones. The embryonic version of NCAM carries a polysialic acid link, which affects the affinity of the molecule. Although PSA-NCAM is generally down-regulated in the adult, it remains high in areas of neurogenesis and plasticity like in hippocampus or olfactory bulbs (21). However, in rat striatum, PSA-NCAM becomes undetectable after postnatal day 25, which coincides with the end of corticostriatal synaptogenesis. Cortical lesions provoke axonal sprouting in the denervated striatum resulting in prolonged PSA-NCAM expression and extended period of plasticity (22).
Plasticity, Cell Transplantation, and Brain Repair
399
Following transplantation, the embryonic cells are separated from their normal afferent and efferent projections and vascular connections. Moreover, they have different metabolic and nutritional requirements than the adult cells. There is pressure on the grafted cells to adapt to the new environment and in order to survive the cells must go through many changes quickly (23). Apart from the GAP-43 results, data are inconclusive on how the expression patterns of molecules associated with plasticity in the embryonic tissue change, if at all, as a consequence of transplantation of the cells into the adult host. Anatomical studies confirm that embryonicstriatal grafts integrate extensively and appropriately into the host’s circuitry, receiving and sending out correct afferent and efferent projections (24– 27). Thus, the results support the idea that even following transplantation, at least during the period required for reconnectivity, the embryonic grafts retain their inherent plasticity even in an otherwise relatively nonpermissive adult environment. In recent years, electrophysiological analyses of neuronal activity in the striatum have become increasingly sophisticated (28), allowing a similar approach to the evaluation of mechanisms of plasticity in striatal grafts, as had first been undertaken within the hippocampus (29). Grafted-striatal neurones have been shown to produce appropriate monosynaptic responses to cortical stimulation (30,31). Studying the integration of the graft and the host using electrophysiological methods can address the larger issue of whether use-dependent plasticity in the intact adult striatum (32,33) shares common features with the apparently use-dependent integration of embryonic tissue into the damaged striatum. Importantly, the technique will allow the study of models of use-dependent changes in synaptic function, such as long-term potentiation (LTP) and long-term depression (LTD), in the transplant, and the neuronal circuitry that includes the transplant. Electrophysiological approach may also help address whether grafted neurones exhibit mechanisms of plasticity associated with their location within the host brain or of their site of origin. Whether proteins that modify synaptic plasticity, such as ERK1 or RasGRFl (34–36), also play a role in graft integration is in the process of investigation by transplanting tissue from genetically modified donors into normal hosts.
3.
ENHANCING GRAFT FUNCTION
In the last two decades, research has validated intracerebral transplantation as a potential therapy in neurodegenerative disease, stroke, and acute brain injury (37). However, there are realistic concerns that current transplantation procedures could be considerably improved. Further in-roads must be made to enhance the survival and to control the integration of implanted embryonic tissues, and to identify alternative sources of suitable cells or tis-
400
Do¨bro¨ssy and Dunnett
sues for transplantation that can be expanded and selected in vitro to yield precisely specified cells on demand (4). Advances have been made in the last decade in understanding the causes of cell death in transplanted tissues and developing strategies to improve cell survival (38), differentiation (39,40), and long-distance axonal growth (41) within the host brain. The positive changes that have been made to the protocols, which produce an improved outcome, suggest that the cells used are sensitive to manipulation and treatment prior to implantation and show properties of plasticity. What is more pertinent to this chapter is the question of graft plasticity following transplantation, which leads to functional brain repair. The factors responsible for enabling an anatomically well-integrated graft to provide effective and efficient recovery of function have been less well studied. Clearly, one such factor is that the grafts need to establish afferent and efferent connections with the host brain, which are appropriate to the particular function that needs to be restored (42–44). It is becoming increasingly apparent, though, that circuit reconstruction at an appropriate level may not be sufficient for full functional recovery. During normal development, the growing animal must not only acquire the necessary anatomical structures and systems, but also obtain and refine appropriate patterns of behavior through learning and experience. Even if the right cellular network is present, without the specific experience there cannot be appropriate behavior. It is through a mechanism of anatomical plasticity stimulated by motor or complex learning that the established neuronal pathways progress from simply transmitting a signal representing a ‘‘perception’’ to conducting a message that subserves a learned function with a ‘‘meaning.’’ Similarly, to promote spontaneous recovery following an injury or after surgical repair of damage to the adult brain, patients need appropriate rehabilitation and can benefit through experience, practice, and retraining in reacquiring the lost function. For example, there is ample and convincing evidence in the literature that recovery from stroke can be improved by well-designed movement therapy that can act at both the anatomical level, resulting in cortical reorganization, and at the behavioral level, through the reversal of learned nonuse of the affected limbs or other body system (45–49). Although the capacity of the environment and experience to influence indices of brain function is well established, the degree to which these changes are mirrored in neural grafts placed within the brain is not known. Environmental enrichment, behavioral experience, and cell transplantation can each influence neuronal plasticity and recovery of function after brain damage, and each research topic has been extensively investigated in its own right. However, the degree to which housing conditions or behavioral training can modify the survival, integration, or function of transplanted tissues is less well established. In the following sections, we examine this topic, and suggest that this factor should be considered and integrated into the
Plasticity, Cell Transplantation, and Brain Repair
401
postoperative care following clinical application of neural transplantation (50). This question is essential because it probes the problem of functional reconnectivity of the grafts. More importantly, the extent to which a reconstructed graft–host system can respond normally to environmental and behavioral manipulation impinges directly upon the issue of whether the grafted tissues do become integrated within the neuronal circuitry of the host brain, which could be taken as a definition of true repair. Experimental evidence is accumulating in support of the hypothesis that experience and training play important parts in the plasticity of, and functional recovery provided by, neural grafts. 3.1.
The Influence of Environmental Enrichment on Graft Function
Housing animals in an enriched environment enhance their behavioral experiences by allowing greater opportunities for activity, exploration; standard, conventional housing provides play and social interaction than. Rosenzweig and Bennett (51) pioneered studies in the 1960s demonstrating that such experiences produce changes in the structural anatomy of the brain, including the thickness of the cortex, and morphological changes, such as the size of cell soma, branching of dendrites, and density of synaptic spines of cortical neurones. Recent evidence shows that environmental enrichment can also influence neurogenesis within the small population of resting stem cells in the adult brain (52). The effects of enriched environment described in the literature are associated with alterations in brain neurochemistry, physiology, and behavior, including enhanced performance on diverse cognitive and motor tasks (53–57). Moreover, environmental enrichment has been shown to improve recovery from brain injury at both the structural (58) and the functional levels (59,60). The following sections will examine the influence of the environment in the structural and functional plasticity of neural transplants in a variety of rodent models of brain damage and degeneration. 3.2.
Cholinergic Grafts in Cortex and Hippocampus
Cholinergic neurone-rich transplants derived from the embryonic-basal forebrain can reverse a variety of cognitive deficits associated with cholinergic depletion in the rat cortex and hippocampus caused by aging or explicit lesions (61,62). In an early model system investigating structural plasticity, cholinergic fiber outgrowth from basal forebrain tissues implanted as dissociated cell suspension grafts into the dorsal neocortex was greater in rats housed in an enriched environment than in rats kept in standard cages (63). The difference was particularly marked at 4 weeks after grafting, but was less important by 10 weeks, suggesting that the greatest effect is on the initial stimulation of neurite growth rather than asymptotic expansion of the grafts.
Do¨bro¨ssy and Dunnett
402
Other, more behaviorally orientated studies have analyzed the effects of enriched or standard housing on the ability of basal forebrain grafts to alleviate maze learning deficits in rats with hippocampal cholinergic lesions (64,65). In the first of a series of studies by Christian Kelche et al. (64), only rats that were housed for 10 months in an enriched environment after receiving grafts exhibited significant attenuation of the profound lesion-induced deficit in relearning a ‘Hebb-Williams’ maze task. Neither the transplant alone nor enrichment alone was effective. By contrast, a shorter period of enriched housing over 2 months was not sufficient to benefit even the grafted rats. In a second study, rats experienced standard housing alone, standard housing with daily handling and training in a water maze spatial navigation task, or environmental enrichment, before being retested in the Hebb-Williams mazes (65). Again, neither enrichment alone nor grafts alone were sufficient to yield recovery over a shorter, 5-month postsurgery period, whereas the specific handling and spatial training did produce selective recovery in the grafted animals. As in the earlier cortical study (63), in both these graft studies, asymptotic cholinergic outgrowth from the basal forebrain grafts was similar in both the enriched and the standard housed animals (64). This suggests that enrichment alone is not the sole component of plasticity, and agrees with other structural–functional correlations showing that cholinergic reinnervation is necessary but not sufficient for graft-derived recovery (61). In each case, grafts were effective in alleviating deficits in the ability to learn complex spatial mazes only when the grafted animals received additional training or experience over a protracted period. 3.3.
Cortical Grafts in Stroke and Lesion Models
The observation in human stroke victims of spontaneous functional recovery over several months following the incident offers strong evidence for brain plasticity (66). This suggests that the poststroke cortical environment is conducive to cortical reorganization. In rodents, the functional integration of cortical tissue has been widely studied in a controlled model of ischemia induced by middle cerebral artery occlusion (MCAO). Cortical grafts implanted into the infarcted area of the adult rat receive extensive afferent host connections from somatosensory and other pathways (67), which are functional, as demonstrated by an increased glucose uptake in the graft in response to stimulation of the contralateral vibrissae (68). Projections into the host brain are relatively sparse when the grafts are implanted into adult hosts, in contrast to the rich graft–host connections that are seen to form in the more permissive neonatal host brain (69). The absence of reciprocal anatomical integration partially explains why in this model the transplants may have little influence on host neuronal function (70–72). Nevertheless, when grafted animals have been given the added benefit of housing in enriched environments, functional repair has been seen, in particular, on simple tests
Plasticity, Cell Transplantation, and Brain Repair
403
of motor asymmetry such as rotation, postural position, and balance on a rotating rod (73,74). However, recovery was not seen on tests of skilled motor coordination, such as the skilled forelimb reaching task, where functional effects of grafts appear to depend upon the re-establishment of cortical and basal ganglia connections such as the skilled forelimb reaching task (44,74,75). This suggests that the behavioral benefits that were provided by a combination of cortical grafts and environmental enrichment in the ischemic cortex are attributable to a secondary protection against thalamic atrophy rather than to a primary cortical reconstruction (73). Cortical grafts have also been reported to yield behavioral benefit in a model involving aspirative lesions of the somatosensory cortex (76). In this study, housing the animals in enriched conditions enhanced recovery on a beam-walking task in all groups. Whereas grafts implanted into frontal sensorimotor cortex promoted early recovery, they had little additional long-term benefit, whether or not they were combined with environmental enrichment. Thus, this study suggests two separate mechanisms of plasticity-mediated repair: an early trophic effect of the grafts on spontaneous recovery processes (76,77) and an independent effect of enrichment-promoting behavioral compensation in the foot-slip tests on the beams, rather than reconstruction (76,78,79). 3.4.
Nigral Grafts in the Striatum
Grafting dopamine-rich cells into parkinsonian animals has been perhaps the most frequently used experimental model of the past 20 years investigating the validity of cell replacement therapy. However, the influence of the environment on the outcome of this model has been neglected. A study by one of us examined the influences of environmental manipulation on intrastriatal dopaminergic grafts in the rodent parkinsonian model under two extreme conditions: collective and enriched housing with extensive training on various behavioral tasks, and individual housing with limited behavioral testing (80). Although the enriched animals were not spared from the behavioral deficits induced by unilateral 6-hydroxydopamine lesions, a significantly higher number of grafted-dopaminergic neurones survived under these conditions (80). Since most cell death in nigral grafts occurs during the first week after implantation (81), the enhanced graft survival in enriched animals is likely to be attributable to induced-neuroprotective changes in the host cellular environment. For example, the upregulation of trophic factors and=or corticosterone levels can influence neuronal survival, and such systemic changes are known to be influenced by the housing conditions (58,82,83). The lesion and graft experiments discussed used examples from diverse experimental brain damage models emphasize that both the graft and the host can exhibit plasticity and that the maximal graft effect will be determined not
Do¨bro¨ssy and Dunnett
404
just by the technical conditions of implantation. Rather, graft survival, growth, integration, and functional responses are themselves all modulated by extrinsic influences of the experience and behavior of the host animal. 4.
THE INFLUENCE OF BEHAVIORAL EXPERIENCE AND TRAINING ON GRAFT FUNCTION
Environmental enrichment has been demonstrated to have an effect on brain plasticity, including spontaneous and graft-mediated brain repair. However, ‘‘enrichment’’ is a nebulous term that has distinct components, such as increased social interaction and exploration. Neurogenesis studies have shown that exercise and movement can stimulate plasticity and proliferation of the pool of progenitor cells in the adult dentate gyrus (84,85). Here we examine, using examples from different models of transplantation into experimentally damaged brain, how behavioral experience and training can affect graft function and repair. 4.1.
Nigral Grafts
The effects of experience on nigral graft function have been investigated in the context of the interaction between conditioning and the expression of motor asymmetries in grafted rats. When rats with unilateral nigrostriatal lesions are stimulated either pharmacologically (with direct or indirect dopamine agonists such as apomorphine or amphetamine, respectively) or by a stressor (e.g., tail pinch), the spontaneous motor asymmetry induced by the lesions is manifested in a quantifiable and vigorous head-to-tail turning response, known as ‘‘rotation’’ (86). Dopaminergic cell suspension grafts from the embryonic substantia nigra implanted into the denervated dorsal striatum can reverse the rotation leading to an over compensatory response (87). However, the repeated experience of drug treatment on its own can lead to behavior plasticity (88). Recurring association of amphetamine treatment with the apparatus yields a conditioned response such that if the animal is subsequently injected with saline in the same location it now rotates in the same direction as had been elicited by the drug (89). Previous exposure to drug treatments can similarly influence the expression of rotational behaviors in grafted animals by a process that appears to involve both pharmacological sensitization and behavioral conditioning. Thus, if lesion or grafted rats receive a series of amphetamine injections, the rotation response increases from test to test (90). Moreover, if they receive repeated doses of amphetamine in one environment, and control injections of saline in a second environment, then a subsequent injection of saline will induce strong rotation (and in the same direction, ipsilateral for lesion rats, contralateral for grafted rats) in the environment that was associated with amphetamine treatment. By contrast, administering the
Plasticity, Cell Transplantation, and Brain Repair
405
drug in the environment previously paired with the control injection has no effects (90). Thus, the rate and direction of conditioned rotation are not due to previous treatment with amphetamine per se, which was matched between the groups, but is determined by specific place conditioning to establish an association between the environmental cues associated with a particular location and a specific response. The results suggest that previously learned associations and experiences can mimic the actions of drug treatment by activating similar neuronal pathways in the damaged brain. The development of learned nonuse of limbs following brain injury leading to suppression of particular behaviors is another example of how, in the absence of suitable training and reinforcement, the injured organism can develop inappropriate response patterns. However, as will be discussed in the following section, appropriate training can help reverse the deficits and promote transplant-mediated brain repair. 4.2.
Learning to Use the Transplant
The phenomenon of ‘‘learning to use the graft’’ has emerged from the transplantation literature over the past decade. It is an idea that is based on the functional plasticity of the graft, most likely subserved by morphological changes, and has been demonstrated in complex behavioral tasks. ‘‘Learning to use the graft’’ is a specific aspect not only of the effects of, but also the absolute requirement for, conditioning of the graft function. The term refers to recovery from brain damage where it is not sufficient for the transplanted tissues simply to survive and be well established anatomically, as applies, for example, in the restitution of cholinergic activation of cortical or hippocampal circuits (64,65) or dopaminergic activation of striatal motor routines (91,92). Rather, the animal must undergo specific training for the graft to exert a functional influence, as applies in situations where the graft provides essential transduction of sensory experience, the interpretation of which is dependent upon conditioning, or it actually constitutes the neural substrate for the conditioning itself. This phenomenon has been investigated in two main model systems—retinotectal grafts and striatal grafts. 4.3.
Retinotectal Grafts
Grafts of retina implanted into the midbrain of enucleated rats not only survive and connect anatomically with the host brain, but also can transduce a light stimulus (by direct illumination of the graft laying on the tectal surface) to drive a pupillary reflex (93). But does the grafted rat actually use the transplant to ‘‘see?’’ Does the perception of the light confer a meaning, or if lost due to brain damage is that meaning something that needs to be relearned? Coffey and Lund (94,95) tested rats in two paradigms: a simple open field avoidance chamber and an operant conditioned suppression test. The open field comprized a circular arena, divided into three equal segments
Do¨bro¨ssy and Dunnett
406
with opaque, transparent, and open ceilings. A bright light was positioned overhead. When allowed to explore, a normal rat spends most of the time under the shade of the opaque roof, whereas a sightless rat does not distinguish between the zones. The experimental rats had their normal vision obscured by eye patches and were prepared surgically with a Perspex window in the skull over the retinal grafts. When first exposed to the open field, they showed no differences in the time spent in each quadrant. However, it appeared that the failure was not in their ability to detect the light, but that they did not attach meaning to this novel channel of sensory input. This was confirmed by providing specific training in which light was paired with a foot-shock in the training phase of a conditioned suppression task. Effective learning was shown by a progressive reduction in lever pressing for food reward during light (but not control tone) stimulation, indicating that the rats could detect the visual stimulus. More importantly, when the rats were placed back into the open arena, they avoided the bright segments when viewing the world exclusively through the transplanted retina, indicating that they had learned to use the stimulus to control behavior (94,95). This bears intriguing similarity to the long-established principles of visual development where animals (and people) have to be proactive in learning to interpret the visual world. A parallel exists also with respect to the rare cases of restoration of sight in the congenitally blind where learning to interpret the signals derived from the new channel of input in adulthood has proved remarkably difficult (96). It requires extensive training for these patients to learn to ascribe meaning to the rich and complex—but initially uninterpretable—visual world with which they are suddenly confronted for the first time in adulthood. An important step in the retinotectal grafting experiments is still to demonstrate that the effects are specific. Since the operant training used aversive conditioning to train the animals to avoid the light (94), it could be that the animals simply learned to avoid the novel arbitrary stimulus input rather than learning to interpret it as a new visual channel. The latter, specific, interpretation would be suggested if rats still avoided the bright segments in the open field after a different form of operant training in which the light illumination was conditioned as a positive signal to be approached rather than an aversive signal to be avoided. This has apparently been found (P.J. Coffey, personal communication) but unfortunately never published. 4.4.
Striatal Grafts
‘‘Learning to use a graft’’ has also been investigated in the context of striatal transplantation in rats with striatal damage. In the quinolinic acid lesion model of HD, striatal grafts survive, establish afferent and efferent connections with the host brain, and alleviate deficits in a variety of motor and cognitive tests (44,97). Functionally, the striatum has been considered as the
Plasticity, Cell Transplantation, and Brain Repair
407
neural substrate for habit formation and motor learning (28,98). All experiments employed a ‘‘9-hole box’’ apparatus that allows precise presentation of stimuli and registration of the animals’ responses in multiple spatial locations provided by an array of 9 holes in the back wall of the apparatus. In the ‘‘Carli’’ task (99), animals are trained to hold a nose poke into the central hole in the array and then to respond rapidly to brief light stimuli on the left or right sides. Animals were trained to respond to either the same side as the light stimulus or the opposite side. Unilateral lesions of either the dopamine afferents to the striatum or intrinsic striatal neurones disrupted the initiation of responses on the contralateral side. The lesions specifically affect initiation without affecting the animals’ ability to detect or attend to the eliciting stimulus (99–101). Similarly, when unilateral striatal lesions and striatal grafts are made in rats that have been trained to respond on the opposite side to the stimulus, both groups (lesioned and grafted) exhibit profound deficits in responding on the contralateral side when returned to the test 4 months later. However, whereas the lesioned rats cannot relearn the task, the grafted animals do relearn with training (102–104). The relearning takes place over a similar period to that required by naive rats to learn the task de novo. Thus, similar to the phenomenon described by Coffey and Lund (94,95), the reformation of appropriate connections by the striatal grafts was not by itself sufficient for functional recovery, rather, the rats had to be retrained to use the reconstructed graft–host circuitry to perform the relevant stimulus–response associations. The need for relearning to use the transplant implies not only functional plasticity, but also a structural representation of that process within the reconstructed graft-host circuit. Motor skills and habits are acquired and refined throughout development and a lifetime of experience. To the extent that motor learning (whether expressed in the language of habit formation, procedural, or stimulus–response associative learning) is mediated by the striatum, it is not surprising that striatal lesions generate such lasting impairments in a variety of skilled motor tasks. More remarkable is the fact that recovery can be achieved by a process as apparently crude as embryonic-striatal cell transplantation, and that the implanted cells not only organize themselves and project appropriately (44,105), but also retain the physiological plasticity to act as a substrate for new learning (106). The first challenge is to determine whether more direct evidence can be obtained that the relearning is actually subserved within the striatal circuitry, before considering further what the cellular mechanisms involved might be. The fact that striatal grafts transduce afferent information arriving via cortical and pallidal inputs and project to primary outputs in the globus pallidus has been shown using a variety of methods including induction of immediate early gene expression (107–109), the electrophysiological identification of monosynaptic projections (30,31,110), and in vivo monitoring of
408
Do¨bro¨ssy and Dunnett
changes in neurotransmitter turnover in the grafts and their targets (111– 113). Brasted and colleagues undertook a series of experimental manipulations to explore the limits of the relearning effect at the behavioral level. First, since host cortical and other afferents slowly retract after excitoxic removal of their striatal targets, we reasoned that delaying the lesion–graft interval would reduce the opportunities for the grafts to establish afferent innervation from the host brain, and found that extending the interval from 9 days to 10 weeks abolished the ability of the grafted rats to relearn the Carli task (104). Moreover, in a related task designed to allow separate training on the ipsilateral and contralateral sides on alternate days (101), training the animals on both sides prior to surgery and then providing extensive postoperative training on the ipsilateral side did not transfer to the critical tests conducted on the animals’ abilities to respond on the contralateral side (103). This indicates that the relearning involves specific stimulus–response associations mediated by the transplanted striatum, and is not achieved simply by general training in task performance (103,114). 4.5.
Motor Training and Recovery of Function After Lesion and Graft
The phenomenon of ‘‘learning to use the graft’’ suggests that the functions that had been established in the intrinsic circuits of the brain through a lifetime’s training and experience prior to lesion need to be re-established— through relearning—using the graft circuitry (94,95,102–104,115). We tested whether a generic motor training protocol had a proactive effect on the performance of a related behavioral task (116). Specifically, we have pretrained animals on a skilled lateralized-reaching task in the staircase test followed by unilateral striatal lesions and striatal grafts that have previously been shown to disrupt and restore, respectively, performance on the contralateral side (117,118). During the period 3 months after grafting, some of the animals were given extensive additional training in a bilateral manual dexterity test and were tested explicitly whether this bilateral motor training transferred to promoting recovery on the paw-reaching performance of either (or both) lesion or graft groups. The transplant partially reversed the contralateral deficit on the skilled reaching task, but in this task the generic motor training did not confer any additional benefits onto the baseline improvement mediated by the grafts. Although they did not approach the performance level of either the control or the grafted groups, the additional motor training promoted a modest—but still significant—degree of recovery in the lesioned animals. Thus, both transplantation and motor training were seen to be beneficial, but, in this particular test, the effects were not additive. Embryonic-striatal grafts are not homogenous in their make-up. Rather, the grafts have a typically patchy appearance comprizing a mixture
Plasticity, Cell Transplantation, and Brain Repair
409
of both striatal-like cells and nonstriatal (including cortical and pallidal) cells (119). This admixture is due to the presence of precursors of multiple cell types in the embryonic-ganglionic eminence, which cannot be separated simply by differential dissection at the time of preparation of the graft tissues for implantation. However, it is the striatal-like cells within the graft that are critical for functional recovery, e.g., on the skilled reaching task (117,119). Thus, among the grafted animals, behavioral recovery was shown to be correlated both with the total volume of the grafts and with the volume of the patches. However, in the trained, but not in the nontrained group, an additional factor that influenced the patch volumes was the volume of the lesion. This suggest that the severity of the damage had an influence on the development of the striatal-like tissue within the graft, but only under the condition that the animals were trained. The mechanism seems to have been specific toward the striatal-like tissue within the graft because a similar relationship was not observed between the lesion and total graft volume. A common and relevant link between the extent of the lesion—the motor training or exercise—and the striatal projection neurones is the brain-derived neurotrophic factor (BDNF). The BDNF has been demonstrated to be upregulated following a lesion or as a consequence of exercise, and has been shown to have a role in the development and plasticity of striatal neurones (120–126). Thus, we have hypothesized that endogenous levels of BDNF or similar trophic factors may be manipulated to influence the anatomical growth and integration of grafted-striatal tissues and to maximize the graft-induced benefits on functional recovery. These questions directly relate to the plasticity of the brain and of the transplants following brain damage and need to be explored experimentally. The hypothesis that additional generic motor training in the bilateral reaching task would transfer skill from one task to another, leading to a gain of proficiency in the staircase task, was not corroborated by this experiment. The training task that was used did provide extensive practice in general motor skills, and the two tasks do share the vital element of identifying and collecting reward pellets with either forelimbs, however, in the generic task the rats were not constrained to use the impaired limb. There is now growing evidence, in particular in the rehabilitation field, that both experimental animals and patients must be forced to use the affected limb if recovery is to be promoted (127,128). This suggests that the motor demands of the generic motor training—offering the freedom of choice of which paw to use—did not provide the specificity of retraining required for recovery in reaching by the paw under the control of the grafts in the staircase task. That, in turn, suggests that the training necessary to influence graft integration and function needs to be quite specific, attesting to the specificity of the motor habits that require retraining within the graft circuitry (50,103).
Do¨bro¨ssy and Dunnett
410
5.
THE CHOICE OF CELLS
Selecting the type of cells to be used for transplantation is influenced by multiple factors, including availability, safety, ethics, the disposition of the damage, and the level of repair that is required. Whether or not a particular graft is able to reverse any functional impairment will depend on the characteristics of the cell, including its plastic properties, and its mechanism of action. How a graft works, though, is very often regarded as a secondary question to whether it functions at all, and it is rarely pursued to great depths. However, mechanisms of graft function must be considered when thinking about the use of alternative cell sources in order to use the appropriate cells to achieve maximum recovery from specific functional deficits. Moreover, understanding the actual mechanisms of action by which a graft functions in a particular type of lesion may promote more effective and efficient development of new cell therapies based on principles of rational design to complement trial-and-error searches for effective parameters. Transplantation is an intrusive neurosurgical procedure involving the introduction of cells into the brain using a metallic or glass canulla. Any effects, either negative or positive, might arise from this procedure through nonspecific mechanisms. The physical presence of the canulla has been shown to stimulate the release of excitatory amino acids and transmitters that could accentuate, attenuate, or in some other way modify the function of the grafted cells both short and long term (129). Furthermore, in clinical application of transplantation, the patient’s anticipation of functional improvements could serve as the basis of a placebo effect, and could last up to 6–18 month postsurgery (130–132). However, graft-mediated functional recovery appears—at least in most cases—to involve specific mechanisms borne out of the interaction of the graft with its new cellular environment. The various types of association involve various degrees of anatomical integration, from none at all to full circuit reconstruction, and this will in turn determine the degree and the quality of recovery. Without any functional anatomical reconnection, cells placed at a strategic site can still act as a source of neurohormone, neurotransmitter, or neurotrophic factor replacement. In this scenario, the grafts mediate recovery through secretion of a molecules, which either are absent or only present in the host brain at levels below physiological activity. Grafted cells which mediate their function in this way have been shown to be effective in some neuroendocrine disturbances in the hypothalamic-pituitary axis. For example, embryonic-hypothalamic grafts secreting gonadotrophin releasing hormone (GnRH) have been shown to reverse the catalogue of developmental deficits in GnRH deficient transgenic mice (133). Nigral grafts implanted ectopically into the dopamine-depleted striatum, as is done in the animal models of parkinsonism and in the clinical application to treat Parkinson
Plasticity, Cell Transplantation, and Brain Repair
411
disease, may also function in a similar fashion, acting as a pharmacological minipump releasing the absent neurotransmitter (4), although a variety of lines of evidence suggest that embryonic-nigral neurones do in fact more fully that this implies. There are growing number of candidate cells other than the traditional primary tissue that can be used at this level of dopamine replacement. Cells can be cultured, or particular cell lines obtained with required characteristics, such as PC12 cells that secrete high levels of dopamine. However, an outstanding problem with grafting cultured cells, in particular ones that have been immortalized using proto-oncogenes, is the possibility that the cells can continue proliferating within the CNS and form tumors. As a consequence, considerable energy is being put into exploring methods of conditional immortalization or encapsulation of grafted cells in order to protect the host from uncontrolled cell division (134–139). Whether the grafted cells release a hormone, a neurotransmitter, or a neuroprotective factor, without functional anatomical integration, the secretion will not be regulated. In some cases, this might be sufficient to redress a deficit but, in complex behavior, signal transduction must be temporally and spatially controlled. Descriptions of retinotectal and striatal grafts in the previous sections are good examples of situations where anatomical integration of the graft is a prerequisite of functional recovery. Currently, the best source of graft tissue to obtain recovery on complex behavioral tasks comes from dissected embryos due to their propensity for growth and establishment of afferent and efferent connections. However, there are major obstacles in the use of primary embryonic tissue for grafting in the clinic, including screening, storing, safety, and standardization of cells obtained separately on each occasion, as well as the moral, ethical, and practical concerns associated with collection of the products of elective abortion. An increasing effort is therefore being put into investigating the use of stem cells for cell replacement therapies. Stem cells are predifferentiation, self-renewing, multipotent cells that can be found both during development and in the adult. Their unique properties have opened up at least two new research avenues that could help improve on current transplantation methods. Firstly, because they have the capacity to proliferate, stem cells can in theory be expanded indefinitely in vitro and grown in unlimited quantities to supply cells for transplantation, on demand. Secondly, because they have the potential to differentiate into all phenotypes of the CNS, appropriate stem cells may provide a common pool for application in a variety of both common and rare diseases, if we can identify the specific sequence of signals necessary to determine the distinct phenotype of specific-therapeutic interest. The challenge now is to determine the molecular cues that commit the multipotent stem cells into the desired phenotype. Transplanted stem cells have shown to be exceptionally sensitive to the environment and their plasticity is manifest in their ability in certain circumstances to adopt the phenotype of the region they are placed into (138,140). Furthermore, in animal
412
Do¨bro¨ssy and Dunnett
models of stroke, grafted stem cells have been shown to migrate great distances into the areas of the infarct (141,142). Therefore, we may eventually be able to recruit the interaction between the cells intrinsic plasticity and the signals provided by the damaged host nervous system to guide appropriate differentiation, rather than having to control that differentiation entirely in vitro before implantation. However, the technology is not yet well developed, and functional recovery following the transplantation of stem cells has so far not been very convincing. An alternative avenue of research made possible by our increased understanding of stem cells relates to their presence in the mature CNS. Cells and regions with neurogenic potential have been identified in the subventricular zone (SVZ) of the lateral ventricle (143,144), the dentate gyrus of the hippocampus (145), and recently in the substantia nigra (146). The new interneurones generated in the SVZ migrate to the olfactory bulb, but little is currently known about the function of neurogenesis in the dentate gyrus (147). Nevertheless, the knowledge of these regions raises the exciting possibility that the mature adult brain has a dormant capacity, much more than previously thought, for regeneration. The prospect of manipulating the neurogenic environment and using endogenous stem cells to promote regeneration and self-repair after neuronal damage (148) has recently become another major topic of investigation, not least because such an approach may avoid the need to source graft tissues from embryonic or fetal donors.
6.
CONCLUSION: PLASTICITY AND THE CLINICAL APPLICATION OF CELL REPLACEMENT THERAPY
Cell replacement therapy has to date been tested clinically in PD, HD, epilepsy, and stroke (37). The clinical outcomes have been variable, due mainly to the differing levels of preclinical, basic experimental evidence that was available prior to the respective clinical trials. The most promising results to date have been seen in the PD trials, but early results from current application of cell replacement therapy in HD are also encouraging (7). A common feature of the trials is that they have concentrated on the biological and technical aspects of transplantation without presupposing that the outcome of the therapy might be influenced by events after the surgery. The growing evidence of plasticity demonstrated by the brain and grafts in response to environmental and training stimuli has hitherto been almost totally neglected throughout the clinical application of cell therapy (149). Physiotherapy and rehabilitation have typically been considered as a quite separate process, based on assisting patients to find new strategies to cope with and compensate for disability. The data discussed here indicate that a different approach might be required to maximize recovery: postoperative experiences, including physiotherapy and explicit behavioral retraining, can have
Plasticity, Cell Transplantation, and Brain Repair
413
marked direct, as well as secondary effects, on the integration and function of grafted cells in the host neural system. Following on from demonstrations that enriched housing conditions or activity training regimes promote not only functional recovery from brain damage, but also the intrinsic plasticity of the brain itself (52,57,85) preliminary but convincing evidence is now becoming available that similar factors influence the integration and function of neural grafts (50). The benefits of forced limb use, constraintinduced movement therapy, and training on recovery from stroke have been clinically validated showing substantial improvements in motor performance even in chronic stroke patients (30,31,110,150) or after spinal cord injury (151). The knowledge gained about brain plasticity following brain damage needs to be linked with what we know about promoting intrinsic recovery processes and how this can boost neurobiological and surgical strategies for repair at the clinical level. With proof of principle now established, a rich area for innovative research with profound therapeutic application is now wide open for investigation. ACKNOWLEDGMENTS We thank the Medical Research Council of the United Kingdom for funding their studies referred to in this manuscript. Earlier versions of particular discussions included in this chapter have previously appeared in part in a published review written by us (50,103). REFERENCES 1. 2. 3. 4. 5. 6. 7.
8.
Richardson PM, McGuinness UM, Aguayo AJ. Axons from CNS neurones regenerate into PNS grafts. Nature 1980; 284:264–265. Horner PJ, Gage FH. Regenerating the damaged central nervous system. Nature 2000; 407(6807):963–970. Horner PJ, Gage FH. Regeneration in the adult and aging brain. Arch Neurol 2002; 59(11):1737–1720. Fawcett JW, Rosser AE, Dunnett SB. Brain Damage, Brain Repair. Oxford: Oxford University Press, 2001. Dunnett SB, Bjorklund A, Lindvall O. Cell therapy in Parkinson’s disease— stop or go? Nat Rev Neurosci 2001; 2(5):365–369. Lindvall O, Hagell P. Clinical observations after neural transplantation in Parkinson’s disease. Prog Brain Res 2000; 127:299–320. Rosser AE, Barker RA, Harrower T, Watts C, Farrington M, Ho AK, Burnstein RM, Menon DK, Gillard JH, Pickard J, Dunnett SB. Unilateral transplantation of human primary fetal tissue in four patients with Huntington’s disease: NEST-UK safety report ISRCTN no 36485475. J Neurol Neurosurg Psychiatry 2002; 73(6):678–685. Bachoud-Levi AC, Remy P, Nguyen JP, Brugieres P, Lefaucheur JP, Bourdet C, Baudic S, Gaura V, Maison P, Haddad B, Boisse MF, Grandmougin T,
414
9.
10.
11.
12.
13. 14. 15.
16.
17.
18.
19.
20.
21.
Do¨bro¨ssy and Dunnett Jeny R, Bartolomeo P, Dalla BG, Degos JD, Lisovoski F, Ergis AM, Pailhous E, Cesaro P, Hantraye P, Peschanski M. Motor and cognitive improvements in patients with Huntington’s disease after neural transplantation. Lancet 2000; 356(9246):1975–1979. Falci S, Holtz A, Akesson E, Azizi M, Ertzgaard P, Hultling C, Kjaeldgaard A, Levi R, Ringden O, Westgren M, Lammertse D, Seiger A. Obliteration of a posttraumatic spinal cord cyst with solid human embryonic spinal cord grafts: first clinical attempt. J Neurotrauma 1997; 14(11):875–884. Kondziolka D, Wechsler L, Goldstein S, Meltzer C, Thulborn KR, Gebel J, Jannetta P, DeCesare S, Elder EM, McGrogan M, Reitman MA, Bynum L. Transplantation of cultured human neuronal cells for patients with stroke. Neurology 2000; 55(4):565–569. Franklin RJ, french-Constant C. Transplantation and repair in multiple sclerosis. In: Russell WC, ed. Molecular Biology of Multiple Sclerosis. London: Wiley, 1996:231–242. Wenning GK, Tison F, Scherfler C, Puschban Z, Waldner R, Granata R, Ghorayeb I, Poewe W. Towards neurotransplantation in multiple system atrophy: clinical rationale, pathophysiological basis, and preliminary experimental evidence. Cell Transplant 2000; 9(2):279–288. Fawcett JW, Asher RA. The glial scar and central nervous system repair. Brain Res Bull 1999; 49(6):377–391. Woolf CJ. Turbocharging neurons for growth: accelerating regeneration in the adult CNS. Nat Neurosci 2001; 4(1):7–9. Bomze HM, Bulsara KR, Iskandar BJ, Caroni P, Skene JH, Spinal axon regeneration evoked by replacing two growth cone proteins in adult neurons. Nat Neurosci 2001; 4(1):38–43. Sirinathsinghji DJ, Mayer E, Stam R, Fernandez JM, Dunnett SB. The expression of GAP-43 mRNA in developing embryonic striatal tissue grafts. Neuroreport 1993; 4(2):175–178. Akita J, Takahashi M, Hojo M, Nishida A, Haruta M, Honda Y. Neuronal differentiation of adult rat hippocampus-derived neural stem cells transplanted into embryonic rat explanted retinas with retinoic acid pretreatment. Brain Res 2002; 954(2):286–293. Przyborski SA, Cambray-Deakin MA. Developmental regulation of MAP2 variants during neuronal differentiation in vitro. Brain Res Dev Brain Res 1995; 89(2):187–201. Sanchez-Ramos JR, Song S, Kamath SG, Zigova T, Willing A, CardozoPelaez F, Stedeford T, Chopp M, Sanberg PR. Expression of neural markers in human umbilical cord blood. Exp Neurol 2001; 171(1):109–115. Poltavtseva RA, Marey MV, Aleksandrova MA, Revishchin AV, Korochkin LI, Sukhikh GT. Evaluation of progenitor cell cultures from human embryos for neurotransplantation. Brain Res Dev Brain Res 2002; 134(l–2):149–154. Cremer H, Chazal G, Lledo PM, Rougon G, Montaron MF, Mayo W, Le Moal M, Abrous DN. PSA- NCAM: an important regulator of hippocampal plasticity. Int J Dev Neurosci 2000; 18(2–3):213–220.
Plasticity, Cell Transplantation, and Brain Repair
415
22. Butler AK, Uryu K, Morehouse V, Rougon G, Chesselet MF. Regulation of the polysialylated form of the neural cell adhesion molecule in the developing striatum: effects of cortical lesions. J Comp Neural 1997; 389(2):289–308. 23. Rosenstein JM. Why do neural transplants survive? An examination of some metabolic and pathophysiological considerations in neural transplantation. Exp Neurol 1995; 133(1):1–6. 24. Wictorin K, Bjorklund A. Connectivity of striatal grafts implanted into the ibotenic acid-lesioned striatum. II. Cortical afferents. Neuroscience 1989; 30(2):297–311. 25. Wictorin K, Simerly RB, Isacson O, Swanson LW, Bjorklund A. Connectivity of striatal grafts implanted into the ibotenic acid-lesioned striatum. III. Efferent projecting graft neurons and their relation to host afferents within the grafts. Neuroscience 1989; 30(2):313–330. 26. Wictorin K. Anatomy and connectivity of intrastriatal striatal transplants. Prog Neurobiol 1992; 38(6):611–639. 27. Clarke DJ, Dunnett SB. Ultrastructural organization within intrastriatal striatal grafts. Prog Brain Res 1990; 82:407–415. 28. Jog MS, Kubota Y, Connolly CI, Hillegaart V, Graybiel AM. Building neural representations of habits. Science 1999; 286(5445):1745–1749. 29. Buzsaki G, Wiesner J, Henriksen SJ, Gage FH. Long-term potentiation of evoked and spontaneous neuronal activity in the grafted hippocampus. Exp Brain Res 1989; 76(2):401–408. 30. Rutherford A, Garcia-Munoz M, Dunnett SB, Arbuthnott GW. Electrophysiological demonstration of host cortical inputs to striatal grafts. Neurosci Lett 1987; 83(3):275–281. 31. Xu ZC, Wilson CJ, Emson PC. Synaptic potentials evoked in spiny neurons in rat neostriatal grafts by cortical and thalamic stimulation. J Neurophysiol 1991; 65(3):477–493. 32. Calabresi P, Centonze D, Gubellini P, Marfia GA, Bernardi G. Glutamatetriggered events inducing corticostriatal long-term depression. J Neurosci 1999; 19(14):6102–6110. 33. Centonze D, Picconi B, Gubellini P, Bernardi G, Calabresi P. Dopaminergic control of synaptic plasticity in the dorsal striatum. Eur J Neurosci 2001; 13(6):1071–1077. 34. Mazzucchelli C, Vantaggiato C, Ciamei A, Fasano S, Pakhotin P, Krezel W, Welzl H, Wolfer DP, Pages G, Valverde O, Marowsky A, Porrazzo A, Orban PC, Maldonado R, Ehrengruber MU, Cestari V, Lipp HP, Chapman PF, Pouyssegur J, Brambilla R. Knockout of ERK1 MAP kinase enhances synaptic plasticity in the striatum and facilitates striatal-mediated learning and memory. Neuron 2002; 34(5):807–820. 35. Giese KP, Friedman E, Telliez JB, Fedorov NB, Wines M, Feig LA, Silva AJ. Hippocampus-dependent learning and memory is impaired in mice lacking the Ras-guanine-nucleotide releasing factor 1 (Ras-GRF1). Neuropharmacology 2001; 41(6):791–800. 36. Chapman PF. Giving drugs to knockout mice: can they do that? Trends Neurosci 2002; 25(6):277–279.
416
Do¨bro¨ssy and Dunnett
37. Bjorklund A, Lindvall O. Cell replacement therapies for central nervous system disorders. Nat Neurosci 2000; 3(6):537–544. 38. Brundin P, Karlsson J, Emgard M, Schierle GS, Hansson O, Petersen A, Castilho RF. Improving the survival of grafted dopaminergic neurons: a review over current approaches. Cell Transplant 2000; 9(2):179–195. 39. Lundberg C, Martinez-Serrano A, Cattaneo E, McKay RD, Bjorklund A. Survival, integration, and differentiation of neural stem cell lines after transplantation to the adult rat striatum. Exp Neurol 1997; 145(2 Pt 1):342–360. 40. Studer L, Tabar V, McKay RD. Transplantation of expanded mesencephalic precursors leads to recovery in parkinsonian rats. Nat Neurosci 1998; 1(4):290–295. 41. Brecknell JE, Fawcett JW. Axonal regeneration. Biol Rev Camb Philos Soc 1996; 71(2):227–255. 42. Dunnett S, Bjorklund A. In: Dunnett S, Bjorklund A, eds. Functional Neural Transplantation. New York: Raven, 1994:531–657. 43. Bregman BS. In: Dunnett S, Bjorklund A, eds. Functional Neural Transplantation. New York: Raven, 1994:489–529. 44. Dunnett SB. Functional repair of striatal systems by neural transplants: evidence for circuit reconstruction. Behav Brain Res 1995; 66(1–2):133–142. 45. Turbes CC. Repair, reconstruction, regeneration and rehabilitation strategies to spinal cord injury. Biomed Sci Instrum 1997; 34:351–356. 46. Robertson IH, Murre JM. Rehabilitation of brain damage: brain plasticity and principles of guided recovery. Psychol Bull 1999; 125(5):544–575. 47. Wilson BA. Recovery of cognitive functions following nonprogressive brain injury. Curr Opin Neurobiol 1998; 8(2):281–287. 48. Taub E, Uswatte G, Elbert T. New treatments in neurorehabilitation founded on basic research. Nat Rev Neurosci 2002; 3(3):228–236. 49. Taub E, Morris DM. Constraint-induced movement therapy to enhance recovery after stroke. Curr Atheroscler Rep 2001; 3(4):279–286. 50. Dobrossy MD, Dunnett SB. The influence of environment and experience on neural grafts. Nat Rev Neurosci 2001; 2(12):871–879. 51. Rosenzweig MR, Bennett EL. Psychobiology of plasticity: effects of training and experience on brain and behavior. Behav Brain Res 1996; 78(l):57–65. 52. Kempermann G, Kuhn HG, Gage FH. More hippocampal neurons in adult mice living in an enriched environment. Nature 1997; 386(6624):493–495. 53. Renner MJ, Rosenzweig MR. The golden-mantled ground squirrel (Spermophilus lateralis) as a model for the effects of environmental enrichment in solitary animals. Dev Psychobiol 1987; 20(l):19–24. 54. Tees RC. The influences of rearing environment and neonatal choline dietary supplementation on spatial learning and memory in adult rats. Behav Brain Res 1999; 105(2):173–188. 55. Fernandez-Teruel A, Escorihuela RM, Castellano B, Gonzalez B, Tobena A. Neonatal handling and environmental enrichment effects on emotionality, novelty=reward seeking, and age-related cognitive and hippocampal impairments: focus on the Roman rat lines. Behav Genet 1997; 27(6):513–526. 56. Xerri C, Coq JO, Merzenich MM, Jenkins WM. Experience-induced plasticity of cutaneous maps in the primary somatosensory cortex of adult monkeys and rats. J Physiol Paris 1996; 90(3–4):277–287.
Plasticity, Cell Transplantation, and Brain Repair
417
57. van Praag H, Kempermann G, Gage FH. Neural consequences of environmental enrichment. Nat Rev Neurosci 2000; 1(3):191–198. 58. Young D, Lawlor PA, Leone P, Dragunow M, During MJ. Environmental enrichment inhibits spontaneous apoptosis, prevents seizures and is neuroprotective. Nat Med 1999; 5(4):448–453. 59. Ohlsson AL, Johansson BB. Environment influences functional outcome of cerebral infarction in rats. Stroke 1995; 26(4):644–649. 60. Rose FD, al Khamees K, Davey MJ, Attree EA. Environmental enrichment following brain damage: an aid to recovery or compensation? Behav Brain Res 1993; 56(1):93–100. 61. Dunnett SB, Low WC, Iversen SD, Stenevi U, Bjorklund A. Septal transplants restore maze learning in rats with fornix-fimbria lesions. Brain Res 1982; 251(2):335–348. 62. Dunnett S. Cholinergic grafts, memory and ageing. Trends Neurosci 1991; 14(8):371–376. 63. Dunnett SB, Whishaw IQ, Bunch ST, Fine A. Acetylcholine-rich neuronal grafts in the forebrain of rats: effects of environmental enrichment, neonatal noradrenaline depletion, host transplantation site and regional source of embryonic donor cells on graft size and acetylcholinesterase-positive fibre outgrowth. Brain Res 1986; 378(2):357–373. 64. Kelche C, Dalrymple-Alford JC, Will B. Housing conditions modulate the effects of intracerebral grafts in rats with brain lesions. Behav Brain Res 1988; 28(3):287–295. 65. Kelche C, Roeser C, Jeltsch H, Cassel JC, Will B. The effects of intrahippocampal grafts, training, and postoperative housing on behavioral recovery after septohippocampal damage in the rat. Neurobiol Learn Mem 1995; 63(2):155–166. 66. Johansson BB. Brain plasticity and stroke rehabilitation. The Willis lecture. Stroke 2000; 31(1):223–230. 67. Grabowski M, Brundin P, Johansson BB. Fetal neocortical grafts implanted in adult hypertensive rats with cortical infarcts following a middle cerebral artery occlusion: ingrowth of afferent fibers from the host brain. Exp Neurol 1992; 116(2):105–121. 68. Grabowski M, Brundin P, Johansson BB. Functional integration of cortical grafts placed in brain infarcts of rats. Ann Neurol 1993; 34(3):362–368. 69. Sorensen JC, Castro AJ, Klausen B, Zimmer J. Projections from fetal neocortical transplants placed in the frontal neocortex of newborn rats. A Phaseolus vulgaris-leucoagglutinin tracing study. Exp Brain Res 1992; 92(2):299–309. 70. Grabowski M, Johansson BB, Brundin P. Neocortical grafts placed in the infarcted brain of adult rats: few or no efferent fibers grow from transplant to host. Exp Neurol 1995; 134(2):273–276. 71. Schulz MK, Hogan TP, Castro AJ. Connectivity of fetal neocortical block transplants in the excitotoxically ablated cortex of adult rats. Exp Brain Res 1993; 96(3):480–486. 72. Schulz MK, McNulty JA, Handa RJ, Hogan TP, Tillotson GL, Shaw PL, Zimmer J, Castro AJ. Fetal neocortical transplants grafted into neocortical
418
73.
74.
75. 76.
77.
78.
79.
80.
81.
82.
83. 84. 85.
86.
87.
Do¨bro¨ssy and Dunnett lesion cavities made in newborn rats: an analysis of transplant integration with the host brain. Cell Transplant 1995; 4(1):123–132. Mattsson B, Sorensen JC, Zimmer J, Johansson BB. Neural grafting to experimental neocortical infarcts improves behavioral outcome and reduces thalamic atrophy in rats housed in enriched but not in standard environments. Stroke 1997; 28(6):1225–1231. Grabowski M, Sorensen JC, Mattsson B, Zimmer J, Johansson BB. Influence of an enriched environment and cortical grafting on functional outcome in brain infarcts of adult rats. Exp Neurol 1995; 133(l):96–102. Montoya CP, Astell S, Dunnett SB. Effects of nigral and striatal grafts on skilled forelimb use in the rat. Prog Brain Res 1990; 82:459–466. Christie MA, Dalrymple-Alford JC. Behavioural consequences of frontal cortex grafts and enriched environments after sensorimotor cortex lesions. J Neural Transplant Plast 1995; 5(4):199–210. Kesslak JP, Brown L, Steichen C, Cotman CW. Adult and embryonic frontal cortex transplants after frontal cortex ablation enhance recovery on a reinforced alternation task. Exp Neurol 1986; 94(3):615–626. Jones TA, Hawrylak N, Klintsova AY, Greenough WT. Brain damage, behavior, rehabilitation, recovery, and brain plasticity. Mental Retard Dev Disabil Res Rev 1998; 4:231–237. Rose FD, Davey MJ, Love S, Dell PA. Environmental enrichment and recovery from contralateral sensory neglect in rats with large unilateral neocortical lesions. Behav Brain Res 1987; 24(3):195–202. Dobrossy MD, Le Moal M, Montaron MF, Abrous N. Influence of environment on the efficacy of intrastriatal dopaminergic grafts. Exp Neurol 2000; 165(1):172–183. Barker RA, Dunnett SB, Faissner A, Fawcett JW. The time course of loss of dopaminergic neurons and the gliotic reaction surrounding grafts of embryonic mesencephalon to the striatum. Exp Neurol 1996; 141(l):79–93. Greco AM, Gambardella P, Sticchi R, D’Aponte D, de Franciscis P. Circadian rhythms of hypothalamic norepinephrine and of some circulating substances in individually housed adult rats. Physiol Behav 1992; 52(6):1167–1172. McEwen BS, Sapolsky RM. Stress and cognitive function. Curr Opin Neurobiol 1995; 5(2):205–216. Kempermann G, Kuhn HG, Gage FH. Experience-induced neurogenesis in the senescent dentate gyrus. J Neurosci 1998; 18(9):3206–3212. van Praag H, Kempermann G, Gage FH. Running increases cell proliferation and neurogenesis in the adult mouse dentate gyrus. Nat Neurosci 1999; 2(3):266–270. Ungerstedt U, Arbuthnott GW. Quantitative recording of rotational behavior in rats after 6-hydroxydopamine lesions of the nigrostriatal dopamine system. Brain Res 1970; 24(3):485–493. Dunnett SB, Bjorklund A, Schmidt RH, Stenevi U, Iversen SD. Intracerebral grafting of neuronal cell suspensions IV. Behavioural recovery in rats with unilateral 6-OHDA lesions following implantation of nigral cell suspensions in different forebrain sites. Acta Physiol Scand Suppl 1983; 522:29–37.
Plasticity, Cell Transplantation, and Brain Repair
419
88. Robinson TE. Behavioral sensitization: characterization of enduring changes in rotational behavior produced by intermittent injections of amphetamine in male and female rats. Psychopharmacology (Berl) 1984; 84(4):466–475. 89. Carey RJ. Conditioned rotational behavior in rats with unilateral 6-hydroxydopamine lesions of the substantia nigra. Brain Res 1986; 365(2):379–382. 90. Annett LE, Reading PJ, Tharumaratnam D, Abrous DN, Torres EM, Dunnett SB. Conditioning vs. priming of dopaminergic grafts by amphetamine. Exp Brain Res 1993; 93(l):46–54. 91. Bjorklund A, Stenevi U, Dunnett SB, Iversen SD. Functional reactivation of the deafferented neostriatum by nigral transplants. Nature 1981; 289(5797):497–499. 92. Bjorklund A, Lindvall O, Isacson O, Brundin P, Wictorin K, Strecker R, Clarke D, Dunnett S. Mechanisms of action of intracerebral neural implants: studies on nigral and striatal grafts to the lesioned striatum. TINS 1987; 10(12):509–516. 93. Klassen H, Lund RD. Retinal transplants can drive a pupillary reflex in host rat brains. Proc Natl Acad Sci USA 1987; 84(19):6958–6960. 94. Coffey PJ, Lund RD, Rawlins JN. Retinal transplant-mediated learning in a conditioned suppression task in rats. Proc Natl Acad Sci USA 1989; 86(18):7248–7249. 95. Coffey PJ, Lund RD, Rawlins JN. Detecting the world through a retina! implant. Prog Brain Res 1990; 82:269–275. 96. Gregory RL. Eye and Brain: The Psychology of Seeing. Oxford: Oxford University Press, 1990. 97. Dunnett SB, Nathwani F, Bjorklund A. The integration and function of striatal grafts. Prog Brain Res 2000; 127:345–380. 98. Mishkin M, Malamut B, Bachevalier J. Neurobiology of Learning and Memory. New York: Guilford, 1984. 99. Carli M, Evenden JL, Robbins TW. Depletion of unilateral striatal dopamine impairs initiation of contralateral actions and not sensory attention. Nature 1985; 313(6004):679–682. 100. Mittleman G, Brown VJ, Robbins TW. Intentional neglect following unilateral ibotenic acid lesions of the striatum. Neurosci Res Commun 1988; 2:1–8. 101. Brasted PJ, Humby T, Dunnett SB, Robbins TW. Unilateral lesions of the dorsal striatum in rats disrupt responding in egocentric space. J Neurosci 1997; 17(22):8919–8926. 102. Mayer E, Brown VJ, Dunnett SB, Robbins TW. Striatal graft-associated recovery of a lesion-induced performance deficit in the rat requires learning to use the transplant. Eur J Neurosci 1992; 4:119–126. 103. Brasted PJ, Watts C, Robbins TW, Dunnett SB. Associative plasticity in striatal transplants. Proc Natl Acad Sci USA 1999; 96(18):10524–10529. 104. Brasted PJ, Watts C, Torres EM, Robbins TW, Dunnett SB. Behavioral recovery after transplantation into a rat model of Huntington’s disease: dependence on anatomical connectivity and extensive postoperative training. Behav Neurosci 2000; 114(2):431–436. 105. Bjorklund A, Campbell K, Sirinathsinghji DJ, Fricker RA, Dunnett S. Functional capacity of striatal transplants in the rat Huntington Model. In:
420
106.
107.
108. 109.
110.
111.
112.
113.
114. 115.
116.
117.
118.
119.
Do¨bro¨ssy and Dunnett Dunnett S, Bjorklund A, eds. Functional Neuronal Transplantation. New York: Raven, 1994:157–195. Mayer E, Brown VJ, Dunnett SB, Robbins TW. Striatal graft-associated recovery of a lesion-induced performance deficit in the rat requires learning to use the transplant. Eur J Neurosci 1992; 4(2):119–126. Labandeira-Garcia JL, Tobio JP, Guerra MJ. Comparison between normal developing striatum and developing striatal grafts using drug-induced Fos expression and neuron-specific enolase immunohistochemistry. Neuroscience 1994; 60(2):399–415. Labandeira-Garcia JL, Guerra MJ. Cortical stimulation induces fos expression in intrastriatal striatal grafts. Brain Res 1994; 652(1):87–97. Mandel RJ, Wictorin K, Cenci MA, Bjorklund A. Fos expression in intrastriatal striatal grafts: regulation by host dopaminergic afferents. Brain Res 1992; 583(1–2):207–215. Nakao N, Nakai K, Itakura T. Fetal striatal transplants reinstate the electrophysiological response of pallidal neurons to systemic apomorphine challenge in rats with excitotoxic striatal lesions. Eur J Neurosci 2000; 12(9): 3426–3432. Campbell K, Kalen P, Wictorin K, Lundberg C, Mandel RJ, Bjorklund A. Characterization of GABA release from intrastriatal striatal transplants: dependence on host-derived afferents. Neuroscience 1993; 53(2):403–415. Sirinathsinghji DJ, Dunnett SB, Isacson O, Clarke DJ, Kendrick K, Bjorklund A. Striatal grafts in rats with unilateral neostriatal lesions. II. In vivo monitoring of GABA release in globus pallidus and substantia nigra. Neuroscience 1988; 24(3):803–811. Sirinathsinghji DJ, Heavens RP, Torres EM, Dunnett SB. Cholecystokinindependent regulation of host dopamine inputs to striatal grafts. Neuroscience 1993; 53(3):651–663. Marshall JF. The education of a brain transplant. Proc Natl Acad Sci USA 1999; 96(18):9976–9978. Brasted PJ, Watts C, Torres EM, Robbins TW, Dunnett SB. Behavioral recovery following striatal transplantation: effects of postoperative training and P-zone volume. Exp Brain Res 1999; 128(4):535–538. Dobrossy MD, Dunnett SB. Motor training effects on recovery of function after striatal lesions and striatal grafts. Exp Neurol 2003; doi:10.1016=S0014– 4886(03)00028–l. Fricker RA, Torres EM, Hume SP, Myers R, Opacka-Juffrey J, Ashworth S, Brooks DJ, Dunnett SB. The effects of donor stage on the survival and function of embryonic striatal grafts in the adult rat brain II. Correlation between positron emission tomography and reaching behaviour. Neuroscience 1997; 79(3):711–721. Watts C, Brasted PJ, Dunnett SB. Embryonic donor age and dissection influences striatal graft development and functional integration in a rodent model of Huntington’s disease. Exp Neurol 2000; 163(l):85–97. Graybiel AM, Liu FC, Dunnett SB. Intrastriatal grafts derived from fetal striatal primordia. I. Phenotype and modular organization. J Neurosci 1989; 9(9):3250–3271.
Plasticity, Cell Transplantation, and Brain Repair 120. 121.
122.
123. 124. 125.
126. 127. 128.
129.
130.
131.
132. 133.
134.
135.
421
Cotman CW, Berchtold NC. Exercise: a behavioral intervention to enhance brain health and plasticity. Trends Neurosci 2002; 25(6):295–301. Ivkovic S, Polonskaia O, Farinas I, Ehrlich ME. Brain-derived neurotrophic factor regulates maturation of the DARPP-32 phenotype in striatal medium spiny neurons: studies in vivo and in vitro. Neuroscience 1997; 79(2):509–516. Ivkovic S, Ehrlich ME. Expression of the striatal DARPP-32=ARPP-21 phenotype in GABAergic neurons requires neurotrophins in vivo and in vitro. J Neurosci 1999; 19(13):5409–5419. Mizuno K, Carnahan J, Nawa H. Brain-derived neurotrophic factor promotes differentiation of striatal GABAergic neurons. Dev Biol 1994; 165(l):243–256. Neeper SA, Gomez-Pinilla F, Choi J, Cotman C. Exercise and brain neurotrophins. Nature 1995; 373(6510):109. Ventimiglia R, Mather PE, Jones BE, Lindsay RM. The neurotrophins BDNF, NT-3 and NT-4=5 promote survival and morphological and biochemical differentiation of striatal neurons in vitro. Eur J Neurosci 1995; 7(2): 213–222. Yurek DM, Fletcher-Turner A. Lesion-induced increase of BDNF is greater in the striatum of young vs. old rat brain. Exp Neurol 2000; 161(l):392–396. Schallert T, Kozlowski DA, Humm JL, Cocke RR. Use-dependent structural events in recovery of function. Adv Neurol 1997; 73:229–238. Schallert T, Leasure JL, Kolb B. Experience-associated structural events, subependymal cellular proliferative activity, and functional recovery after injury to the central nervous system. J Cereb Blood Flow Metab 2000; 20(11): 1513–1528. Zietlow R, Sinclair SR, Schwiening CJ, Dunnett SB, Fawcettt JW. The release of excitatory amino acids, dopamine, and potassium following transplantation of embryonic mesencephalic dopaminergic grafts to the rat striatum, and their effects on dopaminergic neuronal survival in vitro. Cell Transplant 2002; 11(7):637–652. Freeman TB, Vawter DE, Leaverton PE, Godbold JH, Hauser RA, Goetz CG, Olanow CW. Use of placebo surgery in controlled trials of a cellular-based therapy for Parkinson’s disease. N Engl J Med 1999; 341(13):988–992. Greenamyre JT, Shoulson I. We need something better, and we need it now: fetal striatal transplantation in Huntington’s disease? Neurology 2002; 58(5):675–676. Peschanski M, Dunnett S. Cell therapy for Huntington’s disease, the next step forward. Lancet Neurol 2002; 1(2):81. Wood MJA, Charlton HM. Hypothalamic grafts and neuroendocrine function. In: Dunnett S, Bjorklund A, eds. Functional Neuronal Transplantation. New York: Raven Press, 1994:451–466. Aebischer P, Winn SR, Tresco PA, Jaeger CB, Greene LA. Transplantation of polymer encapsulated neurotransmitter secreting cells: effect of the encapsulation technique. J Biomech Eng 1991; 113(2):178–183. Emerich DF, Lindner MD, Winn SR, Chen EY, Frydel BR, Kordower JH. Implants of encapsulated human CNTF-producing fibroblasts prevent behavioral deficits and striatal degeneration in a rodent model of Huntington’s disease. J Neurosci 1996; 16(16):5168–5181.
422
Do¨bro¨ssy and Dunnett
136. Emerich DF, Winn SR, Harper J, Hammang JP, Baetge EE, Kordower JH. Implants of polymer-encapsulated human NGF-secreting cells in the nonhuman primate: rescue and sprouting of degenerating cholinergic basal forebrain neurons. J Comp Neurol 1994; 349(1):148–164. 137. During MJ, Ashenden LM. Towards gene therapy for the central nervous system. Mol Med Today 1998; 4(11):485–493. 138. Renfranz PJ, Cunningham MG, McKay RD. Region-specific differentiation of the hippocampal stem cell line HiB5 upon implantation into the developing mammalian brain. Cell 1991; 66(4):713–729. 139. McKay R, Renfranz P, Cunningham M. Immortalized stem cells from the central nervous system. C R Acad Sci III 1993; 316(12):1452–1457. 140. Vicario-Abejon C, Cunningham MG, McKay RD. Cerebellar precursors transplanted to the neonatal dentate gyrus express features characteristic of hippocampal neurons. J Neurosci 1995; 15(10):6351–6363. 141. Hoehn M, Kustermann E, Blunk J, Wiedermann D, Trapp T, Wecker S, Focking M, Arnold H, Hescheler J, Fleischmann BK, Schwindt W, Buhrle C. Monitoring of implanted stem cell migration in vivo: a highly resolved in vivo magnetic resonance imaging investigation of experimental stroke in rat. Proc Natl Acad Sci USA 2002; 99(25):16267–16272. 142. Savitz SI, Rosenbaum DM, Dinsmore JH, Wechsler LR, Caplan LR. Cell transplantation for stroke. Ann Neurol 2002; 52(3):266–275. 143. Conover JC, Allen RL. The sub ventricular zone: new molecular and cellular developments. Cell Mol Life Sci 2002; 59(12):2128–2135. 144. Bedard A, Levesque M, Bernier PJ, Parent A. The rostral migratory stream in adult squirrel monkeys: contribution of new neurons to the olfactory tubercle and involvement of the antiapoptotic protein Bcl-2. Eur J Neurosci 2002; 16(10):1917–1924. 145. Gage FH. Neurogenesis in the adult brain. J Neurosci 2002; 22(3):612–613. 146. Lie DC, Dziewczapolski G, Willhoite AR, Kaspar BK, Shults CW, Gage FH. The adult substantia nigra contains progenitor cells with neurogenic potential. J Neurosci 2002; 22(15):6639–6649. 147. van Praag H, Schinder AF, Christie BR, Toni N, Palmer TD, Gage FH. Functional neurogenesis in the adult hippocampus. Nature 2002; 415(6875): 1030–1034. 148. Bjorklund A, Lindvall O. Self-repair in the brain. Nature 2000; 405(6789): 892–895. 149. Polgar S, Borlongan CV, Koutouzis TK, Todd SL, Cahill DW, Sanberg PR. Implications of neurological rehabilitation for advancing intracerebral transplantation. Brain Res Bull 1997; 44(3):229–232. 150. Liepert J, Miltner WH, Bauder H, Sommer M, Dettmers C, Taub E, Weiller C. Motor cortex plasticity during constraint-induced movement therapy in stroke patients. Neurosci Lett 1998; 250(l):5–8. 151. Field-Fote EC. Combined use of body weight support, functional electric stimulation, and treadmill training to improve walking ability in individuals with chronic incomplete spinal cord injury. Arch Phys Med Rehabil 2001; 82(6):818–824.
19 Brain-Implantable Biomimetic Electronics as Neural Prostheses to Restore Lost Cognitive Function Theodore W. Berger Department of Biomedical Engineering, Center for Neural Engineering, and Program in Neuroscience, University of Southern California, Los Angeles, California, U.S.A.
Ashish Ahuja and Patrick Nasiatka Department of Electrical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A.
Spiros H. Courellis, Ghassan Gholmieh, Min Chi Hsaio, and Dong Song Department of Biomedical Engineering and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A.
Gopal Erinjippurath Department of Biomedical Engineering, University of Southern California, Los Angeles, California, U.S.A.
John J. Granacki Departments of Biomedical Engineering and Electrical Engineering, Center for Neural Engineering, and Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A.
Jeffrey LaCoss and Vijayaraghavan Srinivasan Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A.
423
424
Berger et al.
Vasilis Z. Marmarelis Departments of Biomedical Engineering and Electrical Engineering, and Center for Neural Engineering, University of Southern California, Los Angeles, California, U.S.A.
Armand R. Tanguay, Jr. Departments of Electrical Engineering, Biomedical Engineering, and Materials Science; Center for Neural Engineering; and Program in Neuroscience, University of Southern California, Los Angeles, California, U.S.A.
Jack Wills Department of Electrical Engineering and Information Sciences Institute, University of Southern California, Los Angeles, California, U.S.A.
1.
INTRODUCTION
One of the frontiers in the biomedical sciences is developing prostheses for the central nervous system (CNS) to replace higher thought processes that have been lost due to damage or disease. Prosthetic systems that interact with the CNS are currently being developed by several groups (1), though virtually all other CNS prostheses focus on sensory or motor system dysfunction and not on restoring cognitive loss resulting from damage to central brain regions. Systems designed to compensate for loss of sensory input attempt to replace the transduction of physical energy from the environment into electrical stimulation of sensory nerve fibers (e.g., cochlear implant or artificial retina), or sensory cortex (2–4). Systems designed to compensate for loss of motor control do so through functional electrical stimulation (FES), in which preprogrammed stimulation protocols are used to activate muscular movement (5,6), or by ‘‘decoding’’ premotor=motor cortical commands for control of robotic systems (7–9). The type of neural prosthesis that performs or assists a cognitive function is qualitatively different than those of the cochlear implant, artificial retina, or FES. We consider here a protheses device that functions in a biomimetic manner to replace information transmission between cortical brain regions (10,11). In such a prosthesis, damaged CNS neurons would be replaced with a biomimetic system composed of silicon neurons. The replacement silicon neurons would have functional properties specific to those of the damaged neurons, and would both receive as inputs and send as outputs electrical activity to regions of the brain with which the damaged region previously communicated (Fig. 1). Thus, the prosthesis being proposed is one that would replace the computational function of damaged brain regions,
Biomimetic Electronics as Neural Prostheses
425
Figure 1 Schematic diagram for the general case of replacing a damaged central brain region with a very large-scale integrated (VLSI) circuit implementation of a biomimetic model, and connecting the inputs of the VLSI-based model to the afferents of the damaged region and the outputs of the VLSI-based model to the efferents of the damaged region.
and restore the transmission of that computational result to other regions of the nervous system. Such a new generation of neural prostheses would have a profound impact on the quality of life throughout society, as it would offer a biomedical remedy for the cognitive and memory loss that accompanies Alzheimer’s disease, the speech and language deficits that result from stroke, and the impaired ability to execute skilled movements following trauma to brain regions responsible for motor control. 2.
THE SYSTEM: HIPPOCAMPUS
We are in the process of developing such a cognitive prosthesis for the hippocampus, a region of the brain involved in the formation of new long-term memories. The hippocampus lies beneath the phylogenetically more recent neocortex, and comprises several different subsystems that form a closed feedback loop (Fig. 2), with input from the neocortex entering via the entorhinal cortex, propagating through the intrinsic subregions of hippocampus, and then returning to neocortex. The intrinsic pathways consist of a cascade of excitatory connections organized roughly transverse to the longitudinal axis of the hippocampus. As such, the hippocampus can be conceived of as a set of interconnected, parallel circuits (12,13). The significance of this organizational feature is that, after removing the hippocampus from the brain, transverse ‘‘slices’’ (approximately 500 mm thick) of the structure may be maintained in vitro that preserve a substantial portion of the intrinsic circuitry, and thus allow detailed experimental study of its principal neurons in their open-loop condition (14,15).
426
Berger et al.
Figure 2 Left panel: Diagrammatic representation of the rat brain (lower left), showing the relative location of the hippocampal formation on the left side of the brain (white); diagrammatic representation of the left hippocampus after isolation from the brain (center), and slices of the hippocampus for sections transverse to the longitudinal axis. Right panel: Diagrammatic representation of one transverse slice of hippocampus, illustrating its intrinsic organization: fibers from the entorhinal cortex (ENTO) project through perforant path (pp) to the dentate gyrus (DG); granule cells of the DG project to the CA3 region, which in turn projects to the CAl region; CAl cells project to the subiculum (SUB), which in the intact brain then projects back to the entorhinal cortex. In a slice preparation, return connections from CAl and the subiculum are transected, creating an open-loop condition for experimental study of hippocampal neurons.
The hippocampus is responsible for what have been termed long-term ‘‘declarative’’ or ‘‘recognition’’ memories (16,17): the formation of mnemonic labels that identify a unifying collection of features (e.g., those comprising a person’s face), and the formation of relations among multiple collections of features (e.g., associating the visual features of a face with the auditory features of the name for that face). In lower species not having verbal capacity, an analogous hippocampal function is evidenced by an ability, for example, to learn and remember a sequence of environmentally cued motor movements (18–23). Major inputs to the hippocampus arise from virtually all other cortical brain regions, and transmit to hippocampus highlevel features extracted by each of the sensory systems subserved by these cortical areas. Thus, the hippocampus processes both unimodal and multimodal features for virtually all classes of sensory input, and modifies these neural representations so that they can be associated (e.g., forming a link between a face and a name) and stored in long-term memory in a manner
Biomimetic Electronics as Neural Prostheses
427
that allows appropriate additional associations with previously learned information (the same face may have context-dependent names, e.g., first name basis in an informal, social setting vs. position title in a formal, business setting), and that minimizes interference (the same name may be associated with several faces). After processing by the hippocampal system, new representations for important patterns are transmitted back to other cortical regions for long-term storage; thus, long-term memories are not stored in hippocampus, but propagation of neural representations through its intrinsic circuitry, and the transformations in those representations consequent to that propagation, are required for a re-encoding essential for long-term memory. It is well established that damage to the hippocampus results in a loss of ability to form new long-term memories (16,17). It is the degeneration and malformation of hippocampal neurons that is the underlying cause of the memory disorders associated with Alzheimer’s disease. Similarly, hippocampal pyramidal cells, particularly those in region CA1, are highly susceptible to even brief periods of anoxia, such as those that accompany stroke. Even blunt head trauma has been shown to be associated with preferential loss of hippocampal neurons in the hilus of the dentate gyrus (DG). Finally, there is a long history of association between hippocampal dysfunction (particularly in region CA3) and epileptiform activity. Thus, there is a wide array of neural damage and neurodegenerative disease conditions for which a hippocampal prosthesis would be clinically relevant.
3.
GENERAL STRATEGY AND SYSTEM REQUIREMENTS
Our strategy for achieving a hippocampal neural prosthesis is based on several system requirements for replacing any CNS region with a biomimetic device that interacts bidirectionally with the rest of the undamaged brain (i.e., sensing and communicating input to the biomimetic device from afferents of the damaged region; communicating and electrically stimulating outputs from the biomimetic device to efferents of the damaged region) (10,11). First and foremost is the nature of the biomimetic model that constitutes the core of the protheses system. Information in the hippocampus and all other parts of the brain is coded in terms of variation in the sequence of all-or-none, point-process (spike) events, or temporal pattern (for multiple neurons, variation in the spatiotemporal pattern). The essential signal processing capability of a neuron is derived from its capacity to change an input sequence of interspike intervals into a different, output sequence of interspike intervals. In all brain areas, the resulting input=output transformations are strongly nonlinear, due to the nonlinear dynamics inherent in the cellular=molecular mechanisms characteristic of neurons and their synaptic connections (15,24–29). As a consequence, the outputs of virtually all
428
Berger et al.
neurons in the brain are highly dependent on temporal properties of the inputs. The spatiotemporal patterns of activity expressed by neurons in the hippocampus are the only ‘‘features’’ that the neocortex has to work within constructing representations for long-term memory. Thus, identifying the nonlinear input=output properties of hippocampal neurons, and the composite input=output transformations of hippocampal circuitry, is essential to mimicking the short-term to long-term memory re-encoding process of the hippocampal system. It is this fundamental functionality of hippocampal neurons that must be captured by any mathematical model designed to replace damaged hippocampal tissue. Attempting to accomplish this modeling goal with compartmental neuron models (30,31) based on cable theory is simply not feasible. The number of parameters required to represent complex dendritic structures and the number and variety of ligand- and voltage-dependent conductances common to hippocampal neurons is simply too large to include in a multineuron network model that is sufficiently compact for a microchip or even a multichip module. Although simplifications of compartmental neuron models are an option, the trade-offs between complexity of the model and representation of neuron and network dynamics are not yet fully understood (26). For this reason, we are using a nonlinear systems analytic approach to modeling hippocampal neurons (14,15,32–35). In this approach, neurons and circuits or networks to be modeled are first characterized experimentally using a ‘‘broad-band’’ stimulus, e.g., a series of impulses (typically 1000–2000 total) in which the interimpulse intervals vary according to a random (Poisson) process. Because the distribution of interimpulse intervals is exponential, the mean frequency can remain relatively low (2 Hz), and thus be physiological, yet the range of intervals can be wide (10–5000 msec). Such a stimulation protocol ensures that the majority, if not all, of synaptic and cellular mechanisms are activated, and as a consequence, contribute to neuron, circuit, or network outputs that are measured (e.g., electrophysiologically). The modeling effort then becomes focused on estimating linear and nonlinear components of the mapping of the known input to the experimentally measured output. Given accurate estimation methods (36–38), the result is a model that is ‘‘compact’’ (many fewer terms than a compartmental neuron network model), predictive for virtually any temporal pattern, and that incorporates at least the majority of known and unknown mechanisms (thus, not requiring modification and optimization for each new discovery in the future). A biomimetic model of a hippocampal network of neurons must be reduced to a hardware implementation in microcircuitry (very large–scale integrated (VLSI) circuit) for at least three reasons. First, for a neural protheses device used clinically, it will be necessary to simulate multiple neurons and neural circuits in parallel; hardware implementations provide the most efficient means for realizing parallel processing. Second, by definition, the device in question will be required to interact with the intact brain in
Biomimetic Electronics as Neural Prostheses
429
real-time if it is to substitute for lost neural function and at least partially reinstate normal levels of cognition and behavior. Again, the rapid operational rates of modern VLSI technology provide the best means for achieving the goal of real-time signal processing and sufficient computational speed to support ongoing interaction of a patient with the environment. Finally, there is the practical consideration of integration of the protheses system with the patient, i.e., it must be miniaturized sufficiently that it can be easily carried ‘‘on board’’ with the patient. A biomimetic device for central brain regions must interact bidirectionally with the undamaged brain to support cognitive function and influence behavior. As discussed above, one of the indisputable characteristics of the mammalian brain is that information (a recognized object, object relations, a motor target or plan, etc.) is represented in the activity of multiple neurons, i.e., population or ‘‘ensemble’’ coding. Anatomical connections between multiple neurons in one brain region and multiple neurons in a second brain region are not random—there is typically an identifiable topography in which neurons from one subregion of a given brain region project primarily to a localized subset of neurons in the target structure. Moreover, virtually all brain areas have region-specific cytoarchitectures, e.g., varying degrees of cellular and=or dendritic layering, geometries of curvature, etc., within which topographical connections are embedded. As a consequence of these considerations, bidirectional communication between a biomimetic device and the brain must be accomplished with one or more multisite electrode arrays, with the spatial distribution and density of recording= stimulating electrodes designed to match (be ‘‘conformal’’ with) the cytoarchitecture and topography-density of the brain regions providing the inputs and outputs of the device. Issues of power and data transmission can be substantial given that, at present, it is not known how many neuron models or what level of network complexity is required to achieve a clinically significant improvement in quality of life when attempting to restore cognitive brain function. In-depth treatment of these issues must await results of preclinical studies in behaving animal models (which we have scheduled as the next step in the development of a hippocampal prosthesis). Issues of long-term biocompatibility also are relevant to chronic use of implanted recording=stimulating electrodes to interface with the brain. Although we will not discuss those issues here, we have dealt with them elsewhere and continue to research this fundamental problem (39). 4.
PROOF-OF-CONCEPT: REPLACEMENT OF THE CA3 REGION OF THE HIPPOCAMPAL SLICE WITH A BIOMIMETIC DEVICE
We have developed a multistage plan for achieving a neural prosthesis for the hippocampus. The first stage will be described in this chapter, and
430
Berger et al.
involves a ‘‘proof-of-concept’’ in which we develop a replacement biomimetic model of the CA3 subregion of a hippocampal in vitro slice. We have chosen to realize our first-generation prosthesis in the context of a hippocampal slice for several reasons. Among them is that the 500 mm thickness of the slice essentially allows us to reduce the problems of modeling the three-dimensional function of the hippocampus, and of interfacing with its complex, three-dimensional structure, to a more tractable two dimensions. This allows us to develop the initial stages of experimental strategies, modeling methodologies, hardware designs, and interfacing technologies within the context of a more simplified and controlled set of conditions. The major intrinsic circuitry of the hippocampus consists of an excitatory cascade of the dentate, CA3, and CA1 subregions (dentate ! CA3 ! CAl; Fig. 3A), and is maintained in a slice preparation. Our stage 1 protheses demonstration consists of: (i) surgically eliminating dentate input to the CA3 subregion, (ii) replacing the biological CA3 with an VLSI-based model of the nonlinear dynamics of CA3 (Fig. 3B and C), and (iii) through a specially designed multisite electrode array, transmitting dentate output to the VLSI model, and VLSI model output to the inputs of CA1 (Fig. 3C). The definition of a successful implementation of the protheses demonstration is the propagation of spatiotemporal patterns of activity from dentate through VLSI model to CA1 that reproduces those observed experimentally in the biological dentate ! CAS ! CAl circuit.
5.
EXPERIMENTAL CHARACTERIZATION OF NONLINEAR PROPERTIES OF THE HIPPOCAMPAL TRISYNAPTIC PATHWAY
The first step in achieving the defined ‘‘proof-of-concept’’ is to experimentally characterize the nonlinear input=output properties both of field CA3 and of the entire trisynaptic pathway, i.e., the combined nonlinearities due to propagation through the dentate ! CA3 ! CAl fields. The model of CA3 will be used to develop the biomimetic replacement device to substitute for CA3 dynamics after the biological CA3 has been removed from the slice. The input=output properties of the trisynaptic pathway will be used to evaluate the extent to which hippocampal circuit dynamics have been restored after substituting the biomimetic device for field CA3. Completing such an experimental characterization requires a stable in vitro slice preparation in which trisynaptic responses (simultaneous recordings from dentate granule cells, CA3 pyramidal neurons, and CA1 pyramidal neurons) to random impulse train (RIT) stimulation of perforant path afferents to the DG can be recorded reproducibly, thus providing the data required for the modeling effort. After experimenting with slices prepared from several different points along the septotemporal axis of the
Biomimetic Electronics as Neural Prostheses
431
Figure 3 Strategy for replacing the CA3 region of hippocampus with a VLSI model of its nonlinear dynamics, and interfacing the VLSI biomimetic device with the remaining, active slice through a conformal, multi electrode (cMEA) array, thus restoring whole-circuit dynamics. (A) Diagrammatic representation of the trisynaptic circuit of the hippocampus. (B) Conceptual representation of replacing the CA3 field with a VLSI-based model. (C) Hippocampal slice in which the CA3 field has been removed. Overlaid is an integrated system in which impulse stimulation from an external source is used activate dentate granule cells and is delivered through one component of a conformal multi electrode array. A second component of the electrode array senses the responses of dentate granule cells and transmits the responses to the VLSIbased model. The VLSI device performs the same nonlinear input=output transformations as biological CA3 neurons, and transmits the output through the multisite electrode array to the dendrites of CA1 neurons, thus activating the last component of the trisynaptic pathway.
hippocampus, varying the concentration of the chloride channel blocker, picrotoxin, which was used to facilitate trans-synaptic propagation through the trisynaptic circuit, and systematically sampling from different subregions of the dentate, CA3, and CA1 fields, we successfully defined an optimal preparation for characterizing hippocampal nonlinearities. Studies have been conducted using both conventional glass electrode, extracellular recording (three different electrodes in the three fields), and glasssubstrate-based multisite electrode arrays (see Sec. 8 below). It is important to note that input in the form of RIT stimulation is applied only to the perforant path afferents to dentate. The nonlinearities
432
Berger et al.
of the dentate then determine the actual input to CA3; in turn, the nonlinearities of CA3 determine the input to CA1. In this stage 1 prosthesis development, field potentials are used as the measure of output from each of the three hippocampal regions: population spikes of dentate granule cells, population spikes of CA3 pyramidal cells, and population excitatory postsynaptic potentials (EPSPs) of CA1 pyramidal cells. This is important because it means that for both fields CA3 and CAl, not only does the continuous input of the random train vary in terms of interimpulse interval, but because of nonlinearities ‘‘upstream,’’ the input also varies in terms of the number of active afferents. As will become clear below, this places a major constraint on modeling CA3 input=output properties, i.e., the model must be capable of predicting CA3 output as a function of both the temporal pattern and the amplitude of dentate input. A small segment of one random train and the evoked dentate, CA3, and CA1 field potential responses is shown in Fig. 4. The top trace represents the impulse stimulations used to activate perforant path (pp in the upper panel) afferents to DG. The strong nonlinearities of the dentate are evident in the second trace: the amplitudes of the negative-going population spikes vary considerably as a function of the interimpulse interval. Likewise, CA3 population spikes and CA1 population EPSPs also exhibit strong variation in amplitude as a function of the temporal intervals of the train. It is the dependence of CA3 output on interimpulse interval and dentate population amplitude that must be captured by the model. 6.
NONLINEAR DYNAMIC MODELING OF CA3 INPUT/OUTPUT PROPERTIES
Replacement of the CA3 hippocampal region requires that a sufficiently accurate quantitative representation of the CA3 functional mapping be developed, leading to a scalable implementation with reduced complexity. Considering that such a quantitative representation should be developed relying solely on the input=output datasets generated experimentally, the Volterra–Poisson modeling approach was chosen, appropriately adapted to our problem. The Volterra modeling approach is a mathematically rigorous, scalable method (40,41) with predictive capabilities that models nonlinear dynamics with an arbitrarily chosen level of accuracy, based on input=output data, and has been successfully applied in modeling biological systems (42). The nonlinear dynamic input=output characteristics of the modeled system are quantitatively captured by the Volterra kernels. These kernels are system descriptors that remain invariant irrespective of the type or the power of the stimulus. The experimental datasets used to estimate the Volterra–Poisson model of CA3 were population spike sequences recorded at the granule cell layer (input) and the corresponding population spike sequences recorded at
Biomimetic Electronics as Neural Prostheses
433
Figure 4 Top: Diagrammatic representation of a hippocampal slice (from Andersen et al. (12)). First trace: Segment of a series of impulses with randomly varying interimpulse intervals delivered in an in vitro hippocampal slice experiment to perforant path (pp) fibers originating from the entorhinal cortex (ENTO) and terminating in the DG. Second trace: Field potential responses recorded from the DG. The narrow, biphasic deflections preceding the field potential responses are stimulation artifacts, and thus, correspond to the occurrence of stimulation impulses in the first trace. The large negative-going deflection in each response is the ‘‘population spike.’’ The amplitude of the population spike correlates positively with the number of granule cells reaching threshold and generating an action potential. The amplitude of the population spike is used as a measure of the output from both DG and CA3. Third trace: Field potential responses recorded from CA3. The stimulation artifacts are much smaller in amplitude. Note the occurrence of multiple population spikes, which are presumed to be due to excitatory, recurrent collaterals unique to the CA3 field. Bottom trace: Field potential responses recorded from CAl. These responses are recorded from the dendritic region of CAl and thus reflect population excitatory postsynaptic potential responses (EPSPs) of CAl neurons rather than population action potentials (‘‘population spikes’’). Note the progressively longer delay in the onset of the response following the stimulation artifact for fields DG, CA3, and CAl, reflecting the propagation of activity through the hippocampal trisynaptic circuit.
434
Berger et al.
the pyramidal cell layer of CA3 (output). These population spike sequences were evoked by fixed amplitude, Poisson distributed (2 Hz mean rate) RIT stimulation applied at the perforant path. The model estimation was carried out using only the amplitude and the interspike intervals of the population spikes in the input and output sequences (Fig. 5). Given the sequence of events in the input and the output datasets, quantitative determination of the nonlinear dynamic properties of CA3 was achieved by computing the Volterra–Poisson kernels. One of the key advantages of Volterra–Poisson kernels is their invariance with respect to the variability of the amplitude of the spikes in the stimulus sequence. The computation of Wiener–Poisson kernels (33,43) and Volterra– Poisson kernels (38) has been addressed in the literature before, when the amplitude of the impulses in the input sequence was fixed. However, in our case, besides variability of the interimpulse interval, there was variability in the amplitude of the impulses at the input sequence. As a result, the equation representing the single input=single output, third order Volterra– Poisson model employed to capture the CA3 nonlinear dynamic properties was adapted as follows: X yðni Þ ¼ Ai k1 þ Ai Aj k2 ðni nj Þ ni m
þ Ai
X
X
Aj1 Aj2 k3 ðni nj1 ; ni nj2 Þ
ð1Þ
ni m
where Ai, Aj represent the varying amplitudes of the population spikes recorded at the granule cell layer (input), y(ni) represents the amplitudes of the population spikes recorded at CA3 (output), k1, k2, and k3 are the first, second, and third order kernels, respectively, ni is the time of occurrence of the current impulse in the input=output sequence, and nj is the time of occurrence of the jth impulse prior to the present impulse within the kernel’s memory window m.
Figure 5 The nonlinear dynamic model of the CA3 hippocampal region receives input x(n) from the dentate, a Poisson point process with events (electrical impulses) of variable amplitude (Ai) and randomly varying interimpulse intervals, and generates a series y(n) of variable amplitude point-process events synchronous to the input.
Biomimetic Electronics as Neural Prostheses
435
The estimation of the kernels was facilitated by expanding them on the orthonormal basis of the associated Laguerre functions fL1 ðnÞg(36) as follows: k2 ðni nj Þ ¼
L1 X
al Ll ðni nj Þ
l¼0
k3 ðni nj1 ; ni nj2 Þ ¼
L1 X L1 X
bl1 ;l2 Ll1 ðni nj1 ÞLl2 ðni nj2 Þ
ð2Þ
l1 ¼0 l2 ¼0
Combining Eqs. (1) and (2), we obtain: yðni Þ ¼ Ai k1 þ Ai
L1 X al l¼0
þ Ai
L1 X L1 X l1 ¼0 l2 ¼0
X
ni m
bl1 l2
Aj2 L1 ðni nj Þ
X
X
Aj1 Aj2 Ll1 ðni nj1 ÞLl2 ðni nj2 Þ
ni m
ð3Þ
which can be solved via least squares to yield estimates of the expansion coefficients falg, fblll2g. The computed expansion coefficients are used to reconstruct the kernel estimates and are the quantities entered in the field-programmable gate array (FPGA) along with kl to implement the model on the VLSI chip for the functional replacement of CA3, as described in the following section. The accuracy of the model and the reliability of the kernels were evaluated using the prediction normalized mean square error (NMSE) mathematically defined by the following equation: P ½ymodel ðni Þ ydata ðni Þ2 NMSE ¼ i ð4Þ P ½ydata ðni Þ2 i
where ymodel is the model response obtained by substituting the computed kernels into Eq. (1) and ydata is the sequence of amplitudes of the population spikes recorded at CA3. Small NMSE values indicate that the kernels reliably capture the nonlinear dynamics of the system they model. Larger NMSE values suggest that higher order terms are needed or that the data are noisy. Data collected from several hippocampal slices were analyzed and the associated Volterra–Poisson kernels were computed. A third-order model was selected as it provided a consistent improvement in NMSE between 4% and 9% compared to a second-order model. The class of computed kernels with consistently low NMSE in the neighborhood of 6% exhibited behavior similar to the representative case shown in Fig. 6. In particular, the second-order kernels exhibited a fast depressive phase in the beginning
436
Berger et al.
Figure 6 Representative second-order kernel (A) and third-order kernel (B) of the CA3 nonlinear dynamic mapping.
followed by a brief facilitatory phase and a slow, shallow depressive phase before returning to the zero line. The third-order kernels exhibited a fast facilitatory phase in the beginning followed by slower and shallower depressive and facilitatory phases before returning to the zero plane. One of the most important properties of a Volterra–Poisson model is its predictive capability for arbitrary input patterns. This property
Biomimetic Electronics as Neural Prostheses
437
is necessary for the CA3 model to function as a CA3 replacement, since it is not bound to a specific input sequence. Each value predicted by the Volterra–Poisson model is the resultant of the nonlinear functional terms of the corresponding Volterra–Poisson series. The nth term of the series contributes to the computation of the predicted output with the effect of nth-order interactions among the input impulses weighted by the nth-order kernel across its memory window. Thus, the second-order term represents the effect on the predicted output of the interaction between pairs of input impulses weighted by the second-order kernel across the memory window of about 1000 msec. Similarly, the third-order term represents the effect on the predicted output of the interaction among triplets of input impulses weighted by the third-order kernel across the memory window of the kernel. An example of how each term of the Volterra–Poisson model of the CA3 contributes to the computation of predicted output of the model is shown in Fig. 7. The amplitude of the third spike of the CA3 response (Fig. 7—top panel; top waveform) is predicted using the computed Volterra–Poisson model of CA3. The amplitude of the three population spikes (Fig. 7—top panel; bottom waveform) measured at the dentate (CA3 input) are processed by the model of Eq. (1) to produce the model response for the third CA3 spike amplitude. In this case, Eq. (1) gives: yðni Þ ¼ Ai k1 þ First-order term ðright-most positive gray arrowÞ þ Ai ½Ai1 k2 ðDn1 Þ þ Ai2 k2 ðDn2 Þ þ Second-order term ðright negative black arrowÞ þ Ai ½Ai1 Ai1 k3 ðDn1 ; Dn1 Þ þ Ai2 Ai2 k3 ðDn2 ; Dn2 Þ þ 2Ai1 Ai2 k3 ðDn1 ; Dn2 Þ Third-order term ðgray arrow to the left of the 2nd-order termÞ ð5Þ The term of each order in Eq. (5) contributes to the value predicted by the model with the effect of input interactions of the same order weighted by the corresponding kernel. The predictive capability of the Volterra–Poisson model enabled us to evaluate quantitatively the accuracy of the model representing the CA3 functional mapping and the reliability of the computed kernels that reflect the nonlinear dynamic input=output properties of CA3. The measure we used for this evaluation was the prediction NMSE defined by Eq. (4). The small NMSE values (on the order of 6%) observed during the analysis for the selected class of models indicate that the kernels computed from the experimental datasets reliably captured the CA3 nonlinear dynamics and that the prediction accuracy of the resulting CA3 model is high. An example of model prediction using the Volterra–Poisson model developed for CA3 using the slice experimental data is shown in Fig. 8.
438
Berger et al.
Figure 7 (Caption on facing page.)
Biomimetic Electronics as Neural Prostheses
439
7.
MICROCIRCUITRY IMPLEMENTATION OF THE CA3 INPUT/OUTPUT MODEL 7.1. System Overview The goal of the hardware implementation phase of the hippocampal protheses project is to build a highly efficient, low power system on a microchip for replacing damaged portions of the hippocampus. The system as depicted in Fig. 9 accepts analog signals from the biological tissue, buffers the signals (amplifies), and then converts these signals into digital form; the analog-todigital conversion (ADC) is controlled by a finite–state machine (FSM) Once the signals are in binary form, the spike detection circuitry extracts the spike amplitude and passes the result to the digital circuitry that calculates the nonlinear response. The digital nonlinear response is converted back to an analog signal (by means of the digital-to-analog converter, or DAC) to form a ‘‘biphasic output’’ compatible with typical neuronal output potentials. This signal is then injected through the conformal multisite electrodes connected to the damaged tissue. 7.2.
Real-Time Population Spike Amplitude Extraction
Real-time population spike amplitude extraction is performed in the digital domain. Digital circuitry prototyped on an FPGA is used to determine the population spike amplitude by a process of filtering, differentiation, and intelligent integration. The original algorithm for real-time population spike extraction was developed in test programs written in the C language. The programs were tested against previously recorded extracellular population field potential waveforms. The programs were optimized to provide
Figure 7 (Facing page) An example of model prediction. (A) Bottom waveform (measured CA3 input): three consecutive dentate population spikes and the corresponding population spike amplitudes (gray); top waveform (measured CA3 output): three consecutive CA3 population spikes and the corresponding population spike amplitudes (black arrows). (B) Detail of the third population spike of the top waveform in panel (A) recorded at CA3, its measured amplitude (positive, black arrow starting the peak of the population spike), and its model predicted counterpart (positive, black arrow to the right of the measured amplitude) which is the sum of the component attributed to the first order interactions (rightmost, positive, gray arrow), the component attributed to the second order interactions (negative, black arrow to the left of the first-order term), and the component attributed to the third- order interactions (positive, gray arrow to the left of the second-order term). (C) Second-order kernel with the gray dots marking its weight to the contribution of the second-order interactions of the input population spike amplitudes shown on the bottom waveform in panel (A). (D) Third-order kernel with the gray dots marking its weight to the contribution of the third-order interactions among the input population spike amplitudes shown on the bottom waveform in panel (A).
440
Berger et al.
Figure 8 CA3 model prediction (NMSE ¼ 8.21%). (A) Predicted population spike amplitudes (black) and their differences from the corresponding measured population spike amplitudes (gray). (B) Segment of the CA3 model prediction. (C) The corresponding segment of the measured CA3 population spike amplitudes. The shaded rectangles highlight two areas for comparison between model predicted values and recorded values.
Figure 9 A top-level system diagram showing the major functional blocks and the signal flow, including the relationship of the real time spike detection, the response generator, and the output waveform generator, which are all in the digital domain. This function has subsequently been implemented in a cmos VLSI device.
Biomimetic Electronics as Neural Prostheses
441
population spike amplitudes essentially identical to classical population spike extraction methods. The C programs were then used to create signal flow diagrams and a text description of the real-time spike extraction algorithm. Once the algorithm was properly specified, we recoded the algorithm in Very High Speed Integrated Circuit (VHSIC) Hardware Description Language (VHDL). The VHDL code has two uses; (1) it can be used to program the FPGA, and (2) it can also be used to synthesize the VLSI implementation. By using the FPGA, both the algorithm and the VHDL code can be validated in an experimental setting, before the integrated circuit is fabricated. This provides a significant level of risk reduction when we design and fabricate the integrated circuit. The FPGA we selected is configured as a plug-in card for a conventional Intel=AMD style personal computer. The computer is used to compile the VHDL code and then to download the VHDL code into the FPGA. The computer also allows data logging while the FPGA is operating, In this fashion, we record all the data samples from the analog-to-digital converter, as well as the spike amplitudes measured by the FPGA. This allows us to verify the quality of the spike extraction following an experiment. We designed and built a chassis for the interface electronics that includes: an instrumentation amplifier providing very high input impedance, a variable gain amplifier (used to set the operating range of the analog-todigital converter), a high-pass filter at 0.5 Hz, a low-pass filter at 30 kHz and a level shifter. The filtered signal is then applied to an analog-to-digital converter (ADC). The A=D converter is a 16 bit device with an input range from 0 to 5 V. The converter is clocked at 10 kHz, controlled by the FPGA. Communications between the interface electronics and the FPGA are handled by low-voltage differential signaling (LVDS). A considerable effort was made to optimize the VHDL code so that we create a minimum size digital circuit. The spike extraction algorithm uses the FPGA (or VLSI chip) for filtering, differentiation, and integration of signals above threshold. Spectral analysis of EPSP waveforms showed little energy above 500 Hz, so this was chosen as the initial filter bandwidth. The filter design was optimized to use the smallest number of gates. The optimal choice was a single section, infinite impulse response (IIR) low-pass filter, with a 3 dB frequency of 452 Hz. This filter is described by the recursive equation: yiþ1 ¼ 0:75yi þ 0:25xi ; which can be rewritten in terms of the filter outputs ðyi Þ and inputs ðxi Þ as yiþ1 ¼ yi þ
ðxi yi Þ : 22
442
Berger et al.
This equation shows that the filter can be implemented with one subtraction, one addition, and a right shift of 2 bits (which is just a wiring connection). Differentiation is performed using first-order differences. d xi xi xi1 dt The integration algorithm just consists of accumulating the absolute value of the derivatives, as long as they exceed a predetermined threshold. At present, we are using a threshold value of 8.333 mV=sec. In Fig. 10, we show the dentate population spike (detector input signal) and the filtered difference values used to compute the size of the spike. The integration begins and ends when the difference value crosses software-settable thresholds (shown here as y ¼ 0). Threshold values were determined that deliver good accuracy but are relatively insensitive to the noise amplified by the differencing operation. Future implementations of this logic will retain the capability to adjust the threshold values for experimental flexibility. The population-spike-amplitude measurement algorithm operates with very high accuracy. Figure 11 depicts the comparison of spike data from three experiments. The first two panels show near-perfect agreement between the floating-point software models and fixed-integer software models computed using hardware. The third panel shows a higher NMSE due to the loss of a single sample in the logging file: because the values do
Figure 10 In this figure, we see the input waveforms: the DG response and the DG population spike and the results of the intermediate processing: the first difference and the integration areas that are gated by the stimulus artifact.
Biomimetic Electronics as Neural Prostheses
443
Figure 11 These panels depict the results of three separate experiments and show the close agreement of the actual experimental measurements to the software and hardware models. The higher error in the third panel was caused by an input sample that was not detected correctly.
not correlate in time, the point appears as an off-axis error. The detected spike agrees with the software model when the sample logging is padded by one time step. We have completed the initial design and construction of the interface electronics chassis as described earlier in this section. The interface electronics receive analog population spike waveforms from the existing experimental apparatus, which is either a multielement array built by Multi Channel Systems (57) or a multi-microelectrode plate system built by Gunter Gross (58), or a conformal multielectrode array (cMEA) as described in Sect. 8. The complete system has been repeatedly tested with live slices of hippocampal tissue to verify the algorithms and hardware implementation. 7.3.
Hardware Implementation of Nonlinear Modeling
We have previously explored several different hybrid analog=digital designs for hardware implementation of hippocampal nonlinear input=output properties (44–48). However, due to difficulties in controlling offset currents and the lack of scalability of capacitors we have opted for totally digital designs (49,50). As described in the previous section, nonlinearities of the hippocampal system are to be expressed by Volterra–Poisson kernel functions. Each kernel function is expressed as a summation of basis functions that are chosen to be generalized Laguerre functions. By appropriate choice of the Laguerre parameter, a, we obtain an excellent fit to experimental data with a small set of basis functions. Laguerre functions can be evaluated either directly by summing a power series, or indirectly, by a recursive tableau. We have written computer programs implementing both methods to evaluate complexity. The recursive
444
Berger et al.
form of the Laguerre calculations is shown below. 8 > < i ¼ 0; n 0 Li ðnÞ ¼ i > 0; n ¼ 0 > : i > 0; n > 0
pffiffiffiffiffi an ð1 aÞ pffiffiffi aLi1 ð0Þ pffiffiffi a½Li ðn 1Þ þ Li1 ðnÞ Li1 ðn 1Þ
Recursion is desirable for the hardware implementation for several reasons. First, only one time step of memory is required. We can overwrite each memory location as ‘‘stale’’ results are no longer needed. Second, the operating speed of the circuit can be greatly reduced: only values required for the current time step need be calculated. Finally, the calculation never changes: hardware required for direct computation is much less complicated than a fully programmable processor. The three cases of the calculation (initialization or time-step zero, polynomial order zero, and all others) are handled by multiplexing the boundary values into the circuit. The tableau in Fig. 12 shows the calculation of three orders of Laguerre polynomial over three time steps. A major concern in calculation is numerical accuracy. The general requirement that implants use minimal power forces calculation with fixed-point integer arithmetic. Numerical experimentation determined that 20-bit fixed point calculation is adequate for the Laguerre recurrence. We use signed integers, and round all calculations to the closest integer. It is not necessary to use a full 20-bit multiplier; by factoring the recurrence relation we find we can subtract quantities of nearly equal magnitude, obtaining a much smaller magnitude that significantly reduces the multiplier size. It is possible to compute the Laguerre polynomials one time only, and store the values for all orders and all significant time lags. The remainder of the calculations could then be based on table lookup. Typical values are orders 0,1, . . . 8 and time lags 0,1, . . . ,1000, as experience has shown that
Figure 12 Tableau showing the calculation of three orders of Laguerre polynomial over three time steps and how the values are reused across time steps; ai is a fixedpoint representation proportional to the current spike amplitude Ai.
Biomimetic Electronics as Neural Prostheses
445
Figure 13 Simplified ‘‘register transfer’’ level diagram of the recursive function used to compute the Laguerre polynomials. Note that only two adders and one multiplier are required in this ‘‘compact’’ implementation.
the polynomials decay to zero within 1000 time steps (lags). The logic initiates a new Laguerre recursion each time a pulse arrives, and (instead of saving all values) the recursion relation is used to obtain the current values of the Laguerre functions. Thus, 10 (or fewer) recursions are running simultaneously, with each recursion requiring storage of only approximately 16 numbers (the current values of the Laguerre functions of order 0 through 7 as well as the values of the Laguerre functions at the previous time step). When the Laguerre functions reach zero computation continues, but yields a zero result. Figure 13 depicts a simplified block diagram of the recursive hardware showing the ‘‘compactness’’ of this implementation. The logic for the CA3 response calculation can be implemented in approximately 20,000 gates, a small fraction of an FPGA on a commercial board available from Dini Group in San Diego, CA. To ensure real-time response on every sample, the state of the Laguerre polynomials must be updated, the presence of a CA3 population spike detected, and a response generated if a spike is detected. Figure 14 shows a ‘‘timing diagram’’ illustrating how these functions are related in time. A C program was written to demonstrate the computation when performed with integer arithmetic. Integer arithmetic is highly desirable for hardware implementations as it is very compact when compared to floating-point processors. The results of the C program were compared to values derived by mathematical modeling to ensure accuracy and consistency. The C program was also used as a platform for studying the numerical accuracy required for the recurrent computation of the Laguerre polynomials used as
446
Berger et al.
Figure 14 ‘‘Timing diagram’’ showing the time sequence within the 10 msec sample period of the calculations performed after a spike is detected.
basis functions in the mathematical model. It was determined that 20-bit signed arithmetic would produce stable and accurate results over a recursion of 1000 iterations. The elements were modeled using FPGA design tools, which provide an integrated design management interface for the user. A portion of this design is shown in Fig. 15, which depicts the main state machine controlling the update and response calculations. Once specified, all elements were described with the VHDL and simulated. The results of the simulation were compared to the results of the C program executions to verify that the circuit would function correctly. Once the simulation was verified, the VHDL description was used to generate a configuration file for the FPGA.
Figure 15 Finite-state machine that controls all of the update and response calculation sequencing.
Biomimetic Electronics as Neural Prostheses
447
The configured FPGA was evaluated in several ways to determine proper operation, using a data file captured during a brain-slice experiment to generate an input waveform. 7.4.
Output Waveform Generation
Finally, the output amplitude as calculated from the Laguerre nonlinear model is scaled by a user-selected factor and is input to the DAC, which produces a biphasic analog pulse to stimulate the tissue. The duration and duty cycle of the biphasic output are also user-selected to allow maximum experimental flexibility.
8.
CONFORMAL, MULTISITE ELECTRODE ARRAY INTERFACE
In this section, we describe in detail the design, fabrication, and testing of novel planar conformal multielectrode arrays (cMEAs) that comprise topographically mapped, spatially distributed electrodes that conform to the local cytoarchitecture of the brain regions stimulated and recorded from in a given experiment. These conformally mapped arrays allow the spatiotemporal relationships of neuronal activity to be recorded without the need for external electrodes. Such site-specific electrical simulation and recording enables efficient data collection for complete nonlinear systems modeling of the hippocampal trisynaptic pathway, as described above. Two specific types of cMEAs have been implemented: (1) a multisite stimulation=recording multielectrode array capable of sampling the entire trisynaptic pathway, including the dentate gyms, CA1, and CA3 regions of rat hippocampus, and (2) a similar stimulation=recording multielectrode array capable of stimulating at the input to the dentate, recording the output from dentate, interfacing with an external electronic or computer system that is capable of simulating the nonlinear input=output characteristics of the CA3 hippocampal region, stimulating at the input to CA1, recording the output from CA1, and interfacing once again with an external electronic or computer system. CA3 replacement demonstration experiments to date have been based on both computer and FPGA interfaces, pointing toward full integration with an application-specific integrated circuit (ASIC) implemented in a silicon-based VLSI circuit. Electrode pad layouts were devised to provide full coverage of the three critical regions of rat hippocampal slices for use in extracting nonlinear dynamical models, and also to provide an interface between the neural tissue and the FPGA=VLSI hardware during acute hippocampal slice testing. In order to find the optimal electrode pad placement, we conducted numerous experiments with conventional sharp glass electrodes designed to locate regions of high neuronal density, both for stimulation and recording.
448
Berger et al.
Figure 16 Optical photomicrograph of a cMEA incorporating the trisynaptic electrode pad layout that is proximity-coupled to a rat hippocampal slice. The nine sets of seven linearly spaced 28 mm diameter pads conform to the DG, CA3, and CA1 regions.
Based on the exact coordinates provided by these external probe experiments, we designed two sets of cMEAs, one for full coverage of the trisynaptic pathway (so-called ‘‘trisynaptic cMEAs’’) and one for the CA3 replacement experiments (so-called ‘‘CA3 replacement cMEAs’’). The latter arrays incorporate optimally placed sets of recording pads in the CA3 region, and also have optimally placed stimulating and recording electrodes in both DG and CAl. As such, these conformal multielectrode arrays have allowed for direct interfacing with the FPGA=VLSI hardware that implements the functional nonlinear dynamical model of CA3. The trisynaptic cMEA design, as shown in Fig. 16, includes nine sets of seven linearly-spaced 28 mm diameter pads with a 50 mm center-to-center spacing, each set spanning one of the key input=output regions of the DG, CA3, and CAl regions of rat hippocampus, thereby allowing for a complete diagnostic assessment of the nonlinear dynamics of the trisynaptic hippocampal circuitry. The CA3 replacement cMEA design, as shown in Fig. 17, includes two different circular pad sizes: (1) 28 mm diameter pads with a 50 mm center-tocenter spacing are grouped in series to form sets of stimulating pads in DG (three at a time) and CAl (two at a time), and (2) 36 mm diameter pads also with a 50 mm center-to-center spacing are used for recording in DG, CA3, and CAl. By grouping sets of stimulating pads in series, we are able to achieve significantly larger pad surface areas and correspondingly larger
Biomimetic Electronics as Neural Prostheses
449
Figure 17 Optical photomicrograph of the central region of a new conformal multielectrode array specifically designed for CA3 replacement demonstration. Stimulating and recording electrodes are arranged in both the DG (lower left) and CA1 (upper middle) regions, while two sets of recording electrodes are provided in the CA3 (lower right) region for pre-replacement characterization. The array pin-out is designed to be compatible with the Multi Channel Systems MEA-60 apparatus.
total stimulating currents than are achievable with single pads, while still maintaining essential conformality to cytoarchitecturally relevant features. These stimulating pads have been placed only in DG and CAl in order to interface with the FPGA=VLSI hardware that replaces the CA3 region entirely, and to thereby support the envisioned CA3 replacement experiment protocol. These cMEAs have provided a durable electrical interface between the neural tissue and the FPGA=VLSI hardware in initial CA3 replacement experiments, as described in more detail in other sections. The general microfabrication procedure employed for these cMEA’s is as follows (see Fig. 18 below). A set of indium tin oxide (ITO) electrical leads is defined on an ITO-coated glass substrate (51) of dimensions 49 mm 49 mm 1.1 mm by the first photomask exposure and subsequent acid-bath etching. A silicon nitride insulation layer is then deposited by means of plasma-enhanced chemical vapor deposition (PECVD). This layer electrically insulates individual ITO lines from each other and from the saline solution that the cMEA will be exposed to during acute slice testing. A second photomask is then used to define the vias that will form the conformal metal (Cr=Au) electrode tips. The Cr=Au metal layer is deposited by electron-beam evaporation. After deposition of a second insulating layer composed of SU-8 photoresist, a final photolithographic exposure is performed to pattern vias through the insulation layer to the previously
450
Berger et al.
Figure 18 Conformal multielectrode array micro-fabrication process. The threephotomask lithographic fabrication procedure allows for submicron resolution to be achieved, and allows for the design of many different electrode pad layouts based on the photomask design.
deposited gold electrodes. The thick (1.5 mm) epoxy-based SU-8 photoresist provides for decreased shunt capacitance of the cMEA as a whole, thereby enabling higher amplitude neural recordings. The array pinout includes 60 signal channels and is designed to be compatible with the Multi Channel Systems MEA-60 apparatus. The development of such high-density cMEAs enables not only the CA3 replacement experiment described herein as a specific example, but also the high precision extraction of nonlinear dynamical models from other brain regions. The combination of high-density electrode arrays with conformal mapping also anticipates the development of functional cortical neural protheses devices. By way of comparison, it should be noted
Biomimetic Electronics as Neural Prostheses
451
that many research groups have employed conventional multielectrode arrays (based on geometrically regular grid patterns) for purposes of neurophysiological characterization (52–56). Currently, several such conventional arrays are commercially available (e.g., Multi Channel Systems, Ref. 57, and the Center for Network Neuroscience at the University of North Texas, Ref. 58). These arrays typically feature large pad sizes (on the order of 150–250 mm) as well as electrode placements that allow for large area coverage but not for conformal topoiogical mapping with high-density electrode placements in the region(s) of interest. The electrode array pads we have employed herein are considerably smaller in size, allowing for higher density placement, and have been accurately placed so that they conform to the specific cytoarchitecture of the brain region of interest (e.g., hippocampus).
9.
SYSTEM INTEGRATION: CURRENT STATUS OF THE HIPPOCAMPAL SLICE PROSTHETIC
As stated earlier, the goal of this multidisciplinary effort is to achieve the first stage of a protheses system for the hippocampal slice by integrating the components described above. More specifically, we seek to functionally replace the biological CA3 subregion of the hippocampal slice with an FPGA=VLSI-based model of the nonlinear dynamics of CA3, such that the propagation of temporal patterns of activity from dentate ! VLSI model ! CAl reproduces that observed experimentally in the biological dentate ! CA3 ! CAl circuit. We have successfully performed the first integration of an FPGA-based nonlinear model of the CA3 hippocampal region with the hippocampal slice, by interconnecting the FPGA and the living slice through a conformal, planar, multisite electrode array. At the time of this review, we have succeeded in supporting real-time communication of stimulation-induced responses of dentate granule cells from the slice to the FPGA, generation of CA3-like outputs from the FPGA device, FPGA-triggered stimulation of CA1, and recording of electrophysiological output from CA1. The preparation we are using to test our first stage protheses system is a hippocampal slice with severed CA3 afferents (Fig. 19; the transection cannot be seen visually). Electrical stimulation of inputs to the DG still evokes excitatory responses from dentate granule cells in this preparation, but excitation cannot propagate through the remainder of the trisynaptic pathway, i.e., to CA3 and from CA3 to CA1. To reinstate the function of CA3 and thus reinstate propagation through the trisynaptic circuit, we have interconnected our FPGA model of CA3 through the specially designed, conformal multisite electrode array (black dots in the slice photomicrograph) described in the previous section, which allows recording electrical activity from, and stimulating electrical activity in, the slice.
452
Berger et al.
Figure 19 System integration of an FPGA-based nonlinear model of the CA3 hippocampal region with a hippocampal slice in which the output from the DG (mossy fibers in the hilus) has been transected, thus eliminating the normal propagation of activity from dentate ! CA3 ! CAl. The FPGA-based model bidirectionally communicates with the living slice through a conformal, planar, multisite electrode array. This system supports real-time communication of stimulation-induced responses of dentate granule cells from the slice to the FPGA, generation of CA3-like outputs from the FPGA device, FPGA-triggered stimulation of CAl, and recording of electrophysiological output from CAl.
As inputs to the transected slice preparation, we have used only a subset of possible stimulation patterns, and more specifically, four-impulse trains with variable interimpulse intervals. Although limited in terms of number of impulses, the interimpulse intervals were chosen to evoke large-magnitude nonlinearities in the responses of dentate granule cells. Thus, stimulation of inputs to the dentate with a variable-interval four-impulse train (Fig. 19, panel 1) generates variable-amplitude output responses from dentate granule cells (panel 2). The output responses are recorded by the multisite array and transmitted to the FPGA model of CA3 (panel 3). The FPGA model A=D converts the extracellular population waveform, identifies the population spike component of that waveform, and calculates its amplitude. Based on the amplitude and history of interimpulse intervals, the FPGA-based nonlinear model of CA3 computes the appropriate CA3 output responses in terms of biphasic stimulation pulses (panel 4)— the magnitude of which are equivalent to the population spike amplitude that would have been generated in the CA3 by that particular output. The biphasic output pulses generated by the FPGA are transmitted through the multisite array and used to electrically stimulate inputs to CA1 (panel 5). The resulting variable-amplitude output responses are recorded from CA1
Biomimetic Electronics as Neural Prostheses
453
(panel 6), demonstrating the functional reinstatement of the dentate ! CA3 ! CAl circuit. These results demonstrate that the various components we have proposed for our protheses device can be developed and effectively integrated into a working system. We are currently in the process of evaluating the output responses generated in the CA1 region by this first-generation CA3 substitution system, i.e., comparing CA1 output in response to the FPGA input with the output of CA1 in response to the biological CA3. The initial results of this evaluation are promising, and further quantitative analysis will be used both to validate and refine the CA3 substitution protocol. Successful demonstration of the ability to substitute external control circuitry for a biological cortical region is a critical first step toward functional cortical prostheses. The FPGA hardware described herein must be replaced by a VLSI circuit that implements the same function. Robust interfaces between cortical tissue and conformal microelectrode arrays capable of chronic implantation and operation must be developed. The FPGA=VLSI models must be generalized to include multiple inputs and multiple outputs from closely associated but functionally differentiable brain regions. Means for accessing deep regions of the brain such as hippocampus from subdural or ebidural regions must be developed, for example by means of deep penetrating probe arrays that are made to be conformal in all three coordinates (dimensions). Finally, retraining protocols must be developed to allow patients to relearn lost functionality by using cortical protheses devices such as those described herein. ACKNOWLEDGMENTS This research was supported by the Brain Restoration Foundation, DARPA DSO (HAND Program), NSF (ERC in Biomimetic MicroElectronic Systems), NSF (BITS Program), ONR (Adaptive Neural Systems Program), ONR (AINS Program), and NIH=NIBIB (USC Biomedical Simulations Resource). REFERENCES 1.
2.
3. 4.
Berger TW, Glanzman DL. Toward Replacement Parts for the Brain: Implantable Biomimetic Electronics as the Next Era in Neural Prosthetics. Cambridge, MA: MIT Press. In press. Humayun MS, de Juan E Jr, Weiland JD, Dagnelie G, Katona S, Greenberg RJ, Suzuki S. Pattern electrical stimulation of the human retina. Vis Res 1999; 39:2569–2576. Loeb GE. Cochlear prosthethics. Ann Rev Neurosci 1990; 13:357–371. Normann RA, Warren DJ, Ammermuller J, Fernandez E, Guillory S. Highresolution spatio-temporal mapping of visual pathways using multi-electrode arrays. Vis Res 2001; 41:1261–1275.
454 5. 6.
7. 8. 9.
10.
11.
12. 13. 14.
15. 16.
17. 18.
19. 20. 21.
Berger et al. Loeb GE, Peck RA, Moore WH, Hood K. BION2 system for distributed neural prothesis interfaces. Med Eng Phys 2001; 23:9–18. Mauritz KH, Peckham HP. Restoration of grasping functions in quadriplegic patients by Functional Electrical Stimulation (FES). Int J Rehabil Res 1987; 10: 57–61. Donoghue JP. Connecting cortex to machines: Recent advances in brain interfaces. Nat Neurosci 2002; 5(suppl):1085–1088. Nicolelis MA. Brain–machine interfaces to restore motor function and probe neural circuits. Nat Rev Neurosci 2003; 4:417–422. Shenoy KV, Meeker D, Cao S, Kureshi SA, Pesaran B, Buneo CA, Batista AP, Mitra PP, Burdick JW, Andersen RA. Neural prothesis control signals from plan activity. Neuroreport 2003; 14:591–596. Berger TW, Baudry M, Brinton RD, Liaw J-S, Marmarelis VZ, Park Y, Sheu BJ, Tanguay AR Jr. Brain-implantable biomimetic electronics as the next era in neural prothesis. Proceedings of the IEEE 2001; 89:993–1012. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Berger TW, Brinton, RB, Marmarelis VZ, Sheu BJ, Tanguay AR Jr. VLSI implementations of biologically realistic hippocampal neural network models. In: Berger TW, Glanzman DL, eds. Toward Replacement Parts for the Brain: Implantable Biomimetic Electronics as the Next Era in Neural Prosthetics. Cambridge, MA: MIT Press. In press. Andersen P, Bliss TVP, Skrede KK. Lamellar organization of hippocampal excitatory pathways. Exp Brain Res 1971; 13:222–238. Amaral DG, Witter MP. The three-dimensional organization of the hippocampal formation: a review of anatomical data. Neuroscience 1989; 31:571–591. Berger TW, Harty TP, Xie X, Barrionuevo G, Sclabassi RJ. Modeling of neuronal networks through experimental decomposition. Proceedings of the IEEE 34th Midwest Symposium on Circuits and Systems 1992:91–97. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Berger TW, Chauvet G, Sclabassi RJ. A biologically based model of functional properties of the hippocampus. Neural Netw 1994; 7:1031–1064. Milner B. Memory and the medial temporal regions of the brain. In: Pribram KH, Broadbent DE, eds. Biology of Memory. New York: Academic Press, 1970:29–50. Squire LR, Zola SM. Episodic memory, semantic memory, and amnesia. Hippocampus 1998; 8:205–211. Berger TW, Bassett, JL. System properties of the hippocampus. In: Gormezano I, Wasserman EA, eds. Learning and Memory: The Biological Substrates. Hillsdale, NJ: Lawrence Erlbaum, 1992:275–320. Eichenbaum H. The hippocampus and mechanisms of declarative memory. Behav Brain Res 1999; 3:123–133. Shapiro ML, Eichenbaum H. Hippocampus as a memory map: synaptic plasticity and memory encoding by hippocampal neurons. Hippocampus 1999; 9:365–384. O’Keefe J, Nadel L. The Hippocampus as a Cognitive Map. Oxford University Press, 1978.
Biomimetic Electronics as Neural Prostheses 22.
23. 24.
25.
26.
27.
28.
29.
30.
31.
32. 33.
34.
35.
36.
455
Deadwyler SA, Bunn T, Hampson RE. Hippocampal ensemble activity during spatial delayed-nonmatch-to-sample performance in rats. J Neurosci 1996; 16: 354–372. Hampson RE, Simeral JD, Deadwyler SA. Distribution of spatial and nonspatial information in dorsal hippocampus. Nature 1999; 402:610–614. Magee J, Hoffman D, Colbert C, Johnston D. Electrical and calcium signaling in dendrites of hippocampal pyramidal neurons. Ann Rev Physiol 1998; 60: 327–346. Dalai SS, Marmarelis VZ, Berger TW. A nonlinear positive feedback model of glutamatergic synaptic transmission in dentate gyrus. Proceedings of the 4th Annual Joint Symposium on Neural Comput 1997; 7:68–75. La Jolla, CA: Instutite for Neural Computation, University of California San Diego. Song D, Wang Z, Marmarelis VZ, Berger TW. Non-parametric interpretation and validation of parametric models of short-term plasticity. Proceedings of the IEEE EMBS Conference 2003; 1901–1904. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Wang Z, Xie X, Song D, Berger TW. Probabilistic transformation of temporal information at individual synapses. Proceedings of the IEEE EMBS Conference 2003; 1909–1912. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Gholmieh G, Courellis SH, Song D, Wang Z, Marmarelis VZ, Berger TW. Characterization of short-term plasticity of the dentate gyrus-CA3 system using nonlinear systems analysis. Proceedings of the IEEE EMBS Conference 2003; 1929–1932. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Dimoka A, Courellis SH, Song D, Marmarelis VZ, Berger TW. Identification of lateral and medial perforant path using single- and dual-input random impulse train stimulation. Proceedings of the IEEE EMBS Conference 2003; 1933–1936. Piscataway, NJ: Institute of Electrical and Electronics Engineers. Wang Z, Song D, Berger TW. Contribution of NMDA receptor channels to the expression of LTP in the hippocampal dentate gyrus. Hippocampus 2002; 12: 680–688. Song D, Wang Z, Berger TW. Contribution of T-type VDCCs to TEA-induced long-term synaptic modification in hippocampal CA1 and dentate gyrus. Hippocampus 2002; 12:689–697. Sclabassi RJ, Krieger DN, Berger TW. A systems theoretic approach to the study of CNS function. Ann Biomed Eng 1988; 16:17–34. Sclabassi RJ, Eriksson JL, Port R, Robinson G, Berger TW. Nonlinear systems analysis of the hippocampal perforant path-dentate projection I. Theoretical and interpretational considerations. J Neurophysiol 1988; 60:1066–1076. Sclabassi RJ, Eriksson JL, Port R, Robinson G, Berger TW. Nonlinear systems analysis of the hippocampal perforant path-dentate projection II. Effects of random train stimulation. J Neurophysiol 1988; 60:1077–1094. Sclabassi RJ, Eriksson JL, Port R, Robinson G, Berger TW. Nonlinear systems analysis of the hippocampal perforant path-dentate projection III. Comparison of random train and paired impulse analyses. J Neurophysiol 1988; 60:1095–1109. Marmarelis VZ. Identification of nonlinear biological systems using Laguerre expansions of kernels. Ann Biomed Eng 1993; 21:573–589.
456
Berger et al.
37. Alataris K, Berger TW, Marmarelis VZ. A novel network for nonlinear modeling of neural systems with arbitrary point-process inputs. Neural Netw 2000; 13:255–266. 38. Gholmieh G, Courellis SH, Marmarelis VZ, Berger TW. A novel method for modeling short-term synaptic plasticity. J Neurosci Methods 2002; 21(2):111–127. 39. Brinton RB, Soussou W, Baudry M, Thompson M, Berger TW. The biotic= abiotic interface: Achievements and foreseeable challenges. In: Berger TW, Glanzman DL. Toward Replacement Parts for the Brain: Implantable Biomimetic Electronics as the Next Era in Neural Prosthetics. Cambridge, MA: MIT Press. In press. 40. Wiener N. Nonlinear Problems in Random Theory. New York: The Technology Press of MIT and Wiley, 1958. 41. Volterra V. Theory of Functions and of Integral and Integro-differential Equations. New York: Dover Publications Inc., 1930. 42. Marmarelis PZ, Marmarelis VZ. Analysis of Physiological Systems: The WhiteNoise Approach. New York: Plenum Press, 1978. Russian translation: Mir Press, Moscow, 1981. Chinese translation: Academy of Sciences Press, Beijing, 1990. 43. Krausz HI, Friesen WO. Identification of nonlinear systems using random impulse train inputs. Biol Cybern 1975; 19:217–230. 44. Sheu BJ, Berger TW, Wu TH, Tsai RH. VLSI neural network implementation of a hippocampal model. Proceedings of the IEEE International Symposium on Circuits and Systems 1995; 3:1664–1667. Piscataway, NJ: Institute of Electrical and Electronics Engineers. 45. Berger TW, Sheu BJ, Tsai RH-J. Analog VLSI implementation of biological neural network models. Proceedings of the IEEE International Conference on Neural Networks 1996;1:12–17. Piscataway, NJ: Institute of Electrical and Electronics Engineers. 46. Tsai R, Sheu B, Berger TW. Design of a programmable pulse-coded neural processor for the hippocampus. Proceedings of the IEEE International Conference on Neural Networks 1998; 1:784–789. Piscataway, NJ: Institute of Electrical and Electronics Engineers. 47. Park Y, Liaw J-S, Sheu BJ, Berger TW. Compact VLSI neural network circuit with high-capacity dynamic synapses. Proceedings of the IEEE International Conference on Neural Networks 2000; 4:214–218. Piscataway, NJ: Institute of Electrical and Electronics Engineers. 48. Tsai RH, Sheu BJ, Berger TW. A VLSI neural network processor based on a model of the hippocampus. Analog Integrated Circuits Signal Process 1998; 15: 201–213. 49. Berger TW, Granacki JJ, Marmarelis VZ, Sheu BJ, Tanguay AR Jr. Brainimplantable biomimetic electronics as neural prothesis. Proceedings of the 1st International IEEE EMBS Conference on Neural Engineering 2003; 108–111. Piscataway, NJ: Institute of Electrical and Electronics Engineers. 50. Berger TW, Granacki JJ, Marmarelis VZ, Sheu BJ, Tanguay AR Jr. Brainimplantable biomimetic electronics as neural prothesis. Proceedings of the IEEE EMBS Conference 2003; 1956–1959. Piscataway, NJ: Institute of Electrical and Electronics Engineers.
Biomimetic Electronics as Neural Prostheses 51. 52.
53. 54.
55.
56.
57. 58.
457
Delta Technologies, Ltd., Stillwater, Maine; NPL-49=11, 15 to 25 O=square sheet resistivity. Novak JL, Wheeler BC. Multisite hippocampal slice recording and stimulation using a 32 element microelectrode array. J Neurosci Methods 1988; 23(2): 149–159. Gross GW, Schwalm FU. A closed flow chamber for long-term multichannel recording and optical monitoring. J Neurosci Methods 1994; 52(1):73–85. Gross GW, Wen WY, Lin JW. Transparent indium-tin oxide electrode patterns for extracellular, multisite recording in neuronal cultures. J Neurosci Methods 1985; 15:243–252. Stoppini L, Duport S, Correges P. New extracellular multirecording system for electrophysiological studies: Applications to hippocampal organotypic cultures. J Neurosci Methods 1997; 72:23–33. Borkholder DA, Bao J, Maluf NI, Perl ER, Kovacs GTA. Microelectrode arrays for stimulation of neural slice preparations. J Neurosci Methods 1997; 77:61–66. Multi Channel Systems GmbH, Reutlingen, Germany; http:==www.multichannelsystems.com=. Center for Network Neuroscience, Department of Biological Sciences, University of North Texas, Denton, Texas; http:==www.cnns.org=.
20 Behaviorally Relevant Neural Codes in Hippocampal Ensembles: Detection on Single Trials John D. Simeral, Robert E. Hampson, and Sam A. Deadwyler Department of Physiology and Pharmacology, Wake Forest University School of Medicine, Winston-Salem, North Carolina, U.S.A.
1.
INTRODUCTION
The approaching era of expanded bioengineering will deal with many issues that relate to the improvement, recovery, and even replacement of neural functions. An obvious requirement of such ‘‘neurobehavioral engineering’’ is to determine how populations of neurons represent information relevant to task performance. A commonly applied method for quantifying information representation in neural ensembles is to determine how different behavioral conditions can be inferred from simultaneously recorded neural firing (1–8). In rat, position and trajectory representations have been decoded from ensembles of hippocampal neurons (9–12). Although several of these investigations applied nonlinear decoding methods including Bayesian algorithms and artificial neural networks, linear approaches were also successful in reconstructing behavioral contingencies from firing activity in small ensembles of simultaneously recorded neurons (1,3,4,7,8,11,13). Several studies have also investigated the reliability of detected neural representations on single trials (3,14,15), often in real time (4,5,7). Successful efforts to extract sensory, motor, and spatial information from neural activity support the conclusion that the identified neural representations are present on individual trials across sessions. 459
460
Simeral et al.
Similar analyses have not been applied to putative ‘‘memory’’ codes. In delayed nonmatch-to-sample (DNMS) tests of short-term memory for odors and spatial locations, task-relevant information encoding has been demonstrated in averaged, event-locked firing patterns of ensembles of rat hippocampal neurons (16–26). The strength or specificity of neural representations of spatial position, as indicated by increased firing rates, correlated with high performance during the session, while weak (low rates) or incorrect sample codes were associated with trials on which errors occurred (25,27). However, one paradox in these findings was that the encoded sample information did not (and should not) persist in the hippocampus through delay (memory retention) intervals, so it remains unclear exactly how the nonmatch representations evolve during the trial to provide the basis for accurate performance. Characterization of the reliability and time course of hippocampal neural representations on a trial-to-trial basis is therefore a necessary prerequisite for modeling and constructing hippocampalprosthetic devices. We demonstrate here the use of linear statistical pattern recognition and classification algorithms to examine the reliability and dynamic evolution of hippocampal neural representations of task-relevant position and phase (sample or nonmatch) information on single trials in the DNMS task. Reliability was quantified by how consistently response position and phase could be inferred from single trial hippocampal ensemble activity. Dynamic changes in the ensemble codes were examined by applying a derived linear transformation to ongoing neural activity and tracking the strength of the position and phase ‘‘codes’’ throughout each trial.
2. RECORDING AND ANALYSIS METHODS 2.1. Recording During the DNMS Task The DNMS behavioral paradigm and recording methodology were described in detail previously (21,25,27,28) and are illustrated in Fig. 1. Briefly, male Long–Evans rats (n ¼ 5) rats were trained to criterion on a two-lever DNMS memory task with random trial-to-trial variable delay intervals (Fig. 1A). On each trial, the left or right lever (random) extended to begin the sample phase. A response on the extended (‘‘sample’’) lever retracted it and illuminated a cue light on the opposite wall over the nosepoke device. The animal traversed the box and nosepoked into a photobeam device, initiating starting the delay interval (1–30 sec, Fig. 1B). At termination of the delay, the next nosepoke turned off the cue light and extended both levers. A lever press opposite the original sample position (‘‘nonmatch’’) was reinforced with a water reward. Errors (match responses) produced no reward and turned out the houselights for 5 sec. New trials commence after 10 sec (intertrial interval). All animals were trained to
Neural Codes in Hippocampal Ensembles
461
Figure 1 Multielectrode recording in hippocampus subfields CA3 and CA1 during the DNMS task. (A) Three phases of the spatial delayed nonmatch-to-sample task. (B) Delay-dependent DNMS task performance. Percentage correct trials for different delay lengths (5 sec bins) averaged across animals (mean SEM). (C) Microelectrode array implanted in CA3 and CA1 allowed simultaneous recording from up to eight electrodes along the longitudinal axis of each region (200 mm spacing within each row, 800 mm between rows).
criterion performance (90% correct responses on trials with 1–5 sec delays) prior to surgery. 2.2.
Surgery and Recording
Microelectrode arrays were implanted in hippocampal subfields CA3 and CA1 as described previously (21,25). Each array (NB Labs, Denison, TX) consisted of 16 fixed-position 40 mm microwires arranged in two rows of eight (200 mm within-row longitudinal spacing, 800 mm between rows) as illustrated in Fig. 1C. Electrode length was constructed such that CA3 and CA1 pyramidal cell layers in dorsal hippocampus could be recorded simultaneously. All animal procedures conformed to National Institutes of Health, USDA, and AAALAC guidelines. After recovery of preoperative performance, animals resumed daily 100 trial sessions in which action potentials from up to four neurons per electrode were isolated using waveform criteria (Spike Sorter and Sort
462
Simeral et al.
Client software, Plexon, Dallas, TX). Behavioral events and extracellular action potentials (spikes) were synchronously time-stamped with 200 msec resolution for offline analysis. Single neuron waveforms (800–1200 msec) were digitized (40 kHz) and stored continually throughout each session. Offline analysis using interspike interval histograms, auto and crosscorrelograms, perievent histograms, and waveform shapes were used to confirm the specificity of single neuron isolation. Individual units could be recorded and maintained for days or weeks. Only putative pyramidal cells with daily mean firing rates between 0.5 and 2 Hz, that were stable across multiple daily sessions, were included in analyses. 2.3.
Derivation of Discriminant Function Coefficients Based on Observed Behavior
A linear statistical pattern recognition technique was used to detect task-relevant information within ensembles of 8–12 hippocampal neurons on single trials. All analyses were performed with custom software scripts in Matlab (The Mathworks, Natick, MA). The linear discriminant methodology was chosen to capture significant event-related firing features in multiple single neurons as well as cross neuron temporal relations. For each animal, four to five consecutive behavioral sessions (350 correct trials) were identified in which 8–12 neurons maintained consistent firing properties across all trials in the sessions. For all correctly performed trials, each lever press, p, was categorized by position (left or right) and task phase (sample or nonmatch) according to the observed behavior. The ensemble firing pattern y(tp) associated with lever press event p at time t was estimated by N parallel vectors xk of binned spike counts, acquired from N neurons in a small time window W centered on time tp: Yðtp Þ ¼ ½xk ðtp þ 1=2W Þ; xk ðtp þ 1=2W TÞ; xk ðtp þ 1=2W 2TÞ . . . xk ðtp þ 1=2W BTÞ where k ¼ 1–N neurons, T ¼ bin size, W ¼ length of the analysis window, and B ¼ W=T ¼ number of time bins. Hence, each neural pattern y(tp) was a vector of N B variables (binned spike counts) that estimated the pattern of neural activity observed during a single lever press event (i.e., on a single trial). For each ensemble of hippocampal neurons recorded from a particular animal, these pattern parameters were varied to yield optimal classifications (8 N 12, 250 ms T 500 ms, 2000 ms W 3000 ms). To reveal consistent task-related firing patterns, a discriminant analysis (29) was first applied to event-related firing patterns y(tp) sorted by left, [y(tL)], and right, [y(tR)], lever positions. This procedure implemented a multivariate eigenvector decomposition to derive the linear combination of the N B variables that had the highest possible correlation with the observed response position. This yielded a vector of N B raw discriminant function
Neural Codes in Hippocampal Ensembles
463
coefficients, Co, that specified the relative contribution of each variable to the unique features that distinguished left from right firing patterns. This same statistical procedure, applied to the same firing patterns sorted by task phase (sample and nonmatch), yielded coefficients Ch that maximized firing pattern features differentiating sample [y(ts)] from nonmatch [y(tN)] events. Thus, firing activity observed during each individual lever press was evaluated in terms of both position and phase of the task in which it occurred. Together, the coefficient vectors Co and Ch specified a simple linear model which could be used to infer either variable at any time t from the firing activity y(t) according to the appropriate scalar discriminant scores: So ðtÞ ¼ yðtÞ Co0 position Sn ðtÞ ¼ yðtÞ Ch0 phase where C 0 denotes the transpose of coefficient vector C and denotes the vector dot product. S0 and Sh were z-normalized through the use of ‘‘raw’’ coefficients (29) to establish zero mean and unity standard deviation across all position and phase scores.
2.4.
Inference of Position and Phase from Instantaneous Firing Patterns
Discriminant scores provided a quantitative measure of the strength of mnemonic representations in hippocampal neural firing at any time t. We quantified the reliability of instantaneous hippocampal mnemonic representations by attempting to infer behavioral position and phase using a blind classification of events based on discriminant scores of ensemble firing patterns. In general, such a classification algorithm must account for the prior probabilities of each event to be classified. However, limiting analyses to correct trials provided equal numbers of left and right, sample and nonmatch lever presses, ensuring that class prior probabilities were always equal (0.5). Because So and Sh, were z-normalized, the optimal classification and linear discriminant functions (30) reduced to a linear separation boundary at S ¼ 0 which yielded the following classification criteria: So > 0; classification ¼ right So < 0; classification ¼ left Sh > 0; classification ¼ sample Sh < 0; classification ¼ nonmatch: Scores near 0 indicated indiscriminate firing patterns lacking all class specificity while scores near 3 reflected patterns with 99.7% probability of correct classification.
464
Simeral et al.
Behavioral inference testing applied a two-stage classification protocol in which randomly selected fractions (80%) of the available events were first used to derive the discriminant coefficients Co and Ch (i.e., train the model). Then testing was performed by applying the derived coefficients to the remaining naı¨ve cases (20%) to yield event scores and classifications. Classification success was quantified through three metrics: hit rate, information content, and the t-value testing of the means of the distributions of event discriminant scores. Each hit rate Hc measured the proportion of events (scores) correctly assigned to class c. This estimates the probability that an ideal observer could predict the behavioral condition c given only the neuronal ensemble data. For example, the model’s performance at inferring phase of the DNMS task was quantified in terms of a sample hit rate, HS , and a nonmatch hit rate, HN, for each rat: Number of events correctly classified as sample phase Total number of events classified as sample phase Number of events correctly classified as nonmatch phase HN ¼ Total number of events classified as nonmatch phase HS ¼
The use of these two hit rate measures had the advantage of revealing whether detection sensitivity was balanced between the two classes. Information content was calculated to provide a single metric of how well ensemble activity predicted phase or position (31,32). Information content in the two-class case with equal prior probabilities exhibits a maximum of one (bit) when all events are correctly classified and a minimum of 0 when half are correctly classified (chance level). While hit rate and information content provide indicators of classification success and ensemble representation, they are less informative regarding the distribution of discriminant scores. In the present study, the distributions of within-animal left and right scores, as quantified by a two-class t-test, indicated the uniqueness and consistency of ensemble representation of left and right positions. To avoid comparisons across unequal numbers of event observations when comparing t-value metrics of resubstitution and naı¨ve data scores, scores were randomly selected from the resubstitution population to provide an equal number of scores for comparison with naı¨ve data, thus permitting equal degrees of freedom for the analysis. Because the training–test procedure randomly assigned events to training and test groups, data sampling dependencies were minimized by repeating the entire protocol 4 or 5 times and averaging results. An additional estimate of classification success was obtained using a leave-one-out test in which model coefficients Co and Ch were derived using all but one of the p available event exemplars and then applied to test the remaining naı¨ve case. For each set of coefficients, the training and classification testing procedure was repeated with a different event left out of the training set each
Neural Codes in Hippocampal Ensembles
465
time until p unique classification attempts (600) had been made. Performance metrics were averaged from the p repetitions. 2.5.
Dynamic Evolution of Ensemble Firing Patterns
Using the discriminant pattern recognition algorithm, we examined the dynamic evolution of position and phase firing patterns during each DNMS trial. Applying the discriminant coefficients derived above, scores were calculated for all activity recorded during 100-trial DNMS sessions in 250 ms time steps to provide an ongoing measure of instantaneous position and phase information within the ensemble. This approach applied a static linear model to examine the time course of task-related changes in position and phase representations across single trials. 3. REPRESENTATION OF PHASE AND POSITION 3.1. Ensemble Representation of Phase and Position on Single Trials The linear discriminant models were successful in correctly identifying the respective behavioral events on 82–99% of trials used to derive the models (i.e., resubstitution). When applied to naı¨ve data (not used to derive the model), the success rates ranged from 76% to 99%. The phase and position scores were calculated for individual sessions and animals, and used to calculate classification success and information content for the ensemble. Tables 1 and 2 (sample and nonmatch events) show the distribution of position and phase scores (means þ SD), classification (hit rate), information content (I), and statistical significance of the score separation (t-test, t) for the single ensemble best (Table 1) and worst (Table 2) case classification sessions. In the best case session, task phase was correctly inferred from 91% of naı¨ve ensemble firing patterns, with resubstitution suggesting an upper limit of over 97%. Sample and nonmatch phase classification were equally successful, with a distinct separation of discriminant scores (Table 1, mean Table 1
Bert Case Classification Results Resubstitution
Event
Score (mean std) Hit rate
Naı¨ve data t
I
Score Hit rate (mean std)
t
I
Sample Nonmatch
2.3 1.2 2.3 0.7
99.0% 97.5%
56.9 0.88
2.2 1.5 2.1 0.9
98.6% 91.3%
40.1 0.76
Left Right
1.9 0.8 1.8 1.1
96.4% 98.3%
45.5 0.78
1.6 0.9 1.7 1.3
95.4% 96.2%
35.1 0.75
466
Table 2
Simeral et al. Worst Case Classification Results Resubstitution
Event
Score (mean std) Hit rate
Naı¨ve data t
I
Score (mean std) Hit rate
t
I
Sample Nonmatch
0.9 1.0 1.0 1.0
85.70% 21.9 0.37 82.6%
0.9 1.1 0.9 1.1
78.7% 80.5%
18.3 0.27
Left Right
0.9 1.2 1.0 0.8
84.0% 82.0%
0.9 1.3 0.9 0.8
85.0% 76.6%
18.5 0.28
21.3 0.34
Figure 2 Phase score distributions for best and worst case classification. (A) Best case score distributions from the animal and session with the highest phase classification hit rate for resubstitution and naı¨ve data (see Table 1). Sample events (dark bars) with scores greater than 0 were correctly classified while those with negative scores were misclassified ( ). Similarly, all nonmatch events (light bars) with negative scores were correctly classified while those with positive scores ( ) were misclassified. (B) Worst case score distributions from the animal and session with the lowest phase classification hit rates (Table 3). Bins ¼ 0.25 sec.
Neural Codes in Hippocampal Ensembles
Table 3
467
Mean Classification Results Across Animals (n ¼ 5) Resubstitution Score (mean std) Hit rate
Event
Naı¨ve data t
I
Score (mean std) Hit rate
t
I
Sample Nonmatch
1.2 1.0 1.3 1.0
93.1% 92.2%
26.3 0.62
1.3 1.0 1.2 1.1
89.4% 90.2%
25.7 0.52
Left Right
1.2 0.9 1.2 1.1
89.0% 87.4%
23.7 0.47
1.1 1.0 1.2 1.1
85.1% 83.0%
21.6 0.37
SD) and significant (t(4) > 40.1, p < 0.001, Table 1) separation of scores for both the naı¨ve and resubstitution classification tests (see also Fig. 2A). Mean scores across sessions for the same ensemble (i.e., same animal) were similar, with significant (t(4) > 38.6, p < 0.001) separation of scores and classification hit rates ranging between 90.9% and 98.6%. Even for the worst case session (Table 2), mean phase classification hit rates were still near 80% (range 78.7–85.7%), which indicated distinct and stable ensemble representation (all t(4) > 18.3, p < 0.001). The lower classification success was reflected in the large degree of overlap in the discriminant scores (see Fig. 2B) which led to event ‘‘misclassification’’ as shown previously (21,27). Scores were similar, but slightly more separated (all t(4) > 25.8, p < 0.001) and hit rates were higher (range 82.2–91.2%) across all sessions for this ensemble. Since the sessions with best and worst case classification were from different animals, the degree of classification success was specific to the ensemble, with only minor variance due to different sessions (33). Over all five animals (ensembles) tested, classification hit rate was 89–93% (Table 3), with similar ranges of scores and separations (all t(4) > 25.7, p < 0.001). Figure 2 shows the best and worst case data as distributions of scores for sample vs. nonmatch classification. The best case score distributions in Fig. 2A (also Table 1) show a distinct separation of nonmatch and sample scores corresponding to high discrimination within the ensemble. The worst case score distribution for phase in Fig. 2B (Table 2) shows many nonmatches (blue bars) with positive scores, as well as samples (red bars) with negative scores, leading to reduced success in ensemble encoding. Ensemble firing in either case reliably provided on average 88% correct identification of individual lever press events with respect to phase of the task. 3.2.
Position Encoding
Position reclassification on single trials was nearly as reliable as that for phase encoding (Fig. 3). Tables 1 and 2 (left and right events) show the scores,
468
Simeral et al.
Figure 3 Position score distributions from rats with the best and worst classification hit rates. (A) For the animal and session with the highest classification hit rate (see Table 1), score distributions from resubstitution (left, 96.4%; right, 98.3%; t ¼ 45.5) and naı¨ve data (left, 89.3%; right, 94.8%; t ¼ 34.1). (B) Score distributions for the lowest classification hit rate (see Table 3) using resubstitution (left, 84.0 %; right, 82.0%; t ¼ 23.6) and naı¨ve data (left, 85.0%; right, 76.6%; t ¼ 20.9). Colors indicate the actual response position for each event (light bars, left press; dark bars, right press). Score bins ¼ 0.25.
separation, and classification hit rates for the same best and worst case sessions as above. Fig. 3 shows the best and worst case sessions as distributions of scores for left vs. right position classification. The average across all sessions, indicated that classification of position (range 76–98%) was as successful as classification of phase (range 78–99%). One difference however was that the distribution of position scores showed more overlap than phase scores for the best case (Fig. 3A) and there was less overlap in position (vs. phase) scores in the worst case (Fig 3B). Thus, the distinctly separated scores (all t(4) > 21.6, p < 0.001) and successful classification rates across ensembles (mean 86%, Table 3) revealed that position information can be extracted on single trials as reliably as phase information using the same present methodology. For each animal, the phase and position coefficients were derived from separate linear models (discriminant functions) using the same neural data. Single leverpress event results for each animal were simultaneously scored as a function of both the position (left–right, bottom axis), and phase (sample– nonmatch, left axis) models and plotted independently (Fig. 4). Behavioral
Neural Codes in Hippocampal Ensembles
469
Figure 4 Simultaneous inference of position and phase for one animal. Position and phase scores were calculated for the ensemble firing pattern associated with 1000 naı¨ve correct lever presses in a leave-one-out analysis. Each point on the plot indicates the phase and position score calculated for a single leverpress, colors indicate the actual behavior of each event (squares, left sample; circles, right sample; diamonds, left nonmatch; triangles, right nonmatch). Clustered points of each color reveal that different behavioral contingencies were consistently associated with distinct simultaneous ensemble representations of position and phase.
events are indicated by color and symbol (black diamond—left sample, green square—right sample, blue triangle—left nonmatch, red circle—right nonmatch). Since the majority of symbols fall in the appropriate quadrant, Fig. 4 graphically confirms that nearly 90% of angle trials were appropriately classified on the basis of the derived position and phase coefficients from their respective linear models. Hence, as indicated by the separation of clusters of scores in Fig. 4, both mnemonic representations (position and phase) coexisted within ensembles on single trials. 3.3.
Dynamic Evolution of Ensemble Firing Patterns
It was important to determine when these codes emerged during the trial. Trials were segregated by correct and error, sample position, and duration of delay, and position and phase coefficients were derived from perievent ensemble activity throughout the session by sliding the discriminant functions along the single trial ensemble spike trains in successive 250 msec
470
Simeral et al.
increments. The sliding phase scores acted as a ‘‘filter’’ for position scores so that only intervals within the trial with a significant phase score were examined for position codes within that same time period. The contingent position scores were then plotted to on a trial-by-trial basis to reveal taskrelated dynamic changes in ensemble representations across the session. An example of a single trial assessment for one session is shown in the chromorasters for left (L) and right (R) sample trials in Fig. 5. Examples of
Figure 5 Dynamic ensemble representation of position within DNMS trials. (A) Example of multineuron (ensemble) activity for left (upper) and right (lower) single trials represented as strip charts spanning 5 sec before the sample (S) to 5 sec after the nonmatch (NM) responses. Respective left (L) and right (R) response positions are indicated. Note phase and position specific firing by individual neurons from which discriminant functions are derived. (B) Correct trials for one animal (single session) have been sorted by trial type and length of delay, and single trial position scores plotted as color intensity rasters (chromorasters). Ensemble phase and position scores were derived by ‘‘sliding’’ the discriminant coefficients in 250 ms increments on single trials from 5 sec before the sample, through the sample, delay and nonmatch phases, ending 5 sec after the nonmatch response. Position scores were filtered such that scores were plotted only when significant phase scores occurred simultaneously. Trials starting with left lever presses (upper plot) or right lever presses (lower plot) were sorted by the length of the delay. White space within trials indicates jposition scoresj < 0.5. Perievent firing rate position scores on each trial, yielded 96% and 98% classification hit rates for left and right positions, respectively for these data.
Neural Codes in Hippocampal Ensembles
471
real-time firing of the entire ensemble for each trial type and its relation to the successive discriminant scores plotted in the chromoraster are shown in Fig. 5A. It is clear that different neurons in the ensemble are active at different times during the trial and on different types of trials (i.e., L vs. R). The ‘‘chromorasters’’ in Fig. 5B show successive (top to bottom) single trial discriminant scores as differential color codes (red–brown ¼ right trial, blue–yellow ¼ left trial). In general, trials that began with a left sample lever press (Fig. 5B, top) demonstrated reliable left position codes in the sample phase. Those patterns stabilized briefly (approximately 3.0 sec) then devolved into a ‘‘random’’ pattern during the initial portion (1–10 sec) of the delay interval. At 10–15 sec into the delay interval, ensemble activity exhibited a steady shift to a right lever position code (brown), culminating in maximal representation of that position (yellow) at the time of the right nonmatch response. On trials beginning with the opposite lever (Fig. 5B, bottom) the right position code clearly appeared at the time of the sample response and persisted in the delay interval for 5–10 sec before transitioning to a progressively stronger left lever response code to finish out the trial (Fig. 5B, blue). The clustering of position scores along the sample and nonmatch marker lines in the chromorasters (Fig. 5B) confirms the high classification hit rates for position in the ensemble from this animal (Table 1), and independently confirmed that the firing pattern consistently represented the animal’s response. 4. DETECTION AND DYNAMICS OF REPRESENTATIONS 4.1. Reliability of Instantaneous Firing Pattern Detection The discriminant score algorithm proved effective in identifying task-relevant neural representations in individual hippocampal ensembles examined at the time of each event. The stability and reliability of pattern detection was quantified by classification success measures similar to those used by others to describe how neuronal responses vary across different behavioral conditions (34–37). High success rates indicated that the lever position and task phase (temporal specificity) dimensions of this task were reliably and simultaneously encoded within single trials tested over hundreds of instances in different animals (Table 3). The unimodal normal distributions of withinclass scores (Figs. 2 and 3) suggested that each specific contingency was represented by a single pattern which re-emerged within the ensemble with appropriate codes on each individual trial. The ensemble encoded both features simultaneously with distributed, but mainly a high, fidelity. This reliable binding of a lever position within its appropriate temporal context (task phase) in real-time on single trials is consistent with suggestions that the hippocampus encodes episodic memory (38,39). The linear model applied here made few assumptions regarding the nature of the ensemble firing patterns and did not explicitly model time
472
Simeral et al.
series or noise. Its success in single trial pattern detection based on relatively small numbers of neurons (8–12) might be due to a high degree of redundancy within the ensemble (40). Alternatively, the multivariate nature of the model may have captured consistent firing relationships between select neurons tuned to simultaneously encode several different features of the task (2,15,33,41–43). It is likely that a more fine-grained algorithm (for example, a nonlinear systems analysis) would improve pattern recognition and reduce inter-trial variability (Berger et al. this volume, also Refs. 4,10,11, since variability in classification success may reflect nonstationarity, or natural firing fluctuations due to sampling error (10). However, the chromoraster plots (Fig. 5) revealed that at least some misclassifications may have reflected actual neural ‘‘miscodes’’ associated with behavioral errors during the task (21,27). 4.2.
Dynamics of Information Representation Within Trials
Analysis of the dynamics of ensemble firing throughout the session revealed that event related representations were present on single trials, but that they sometimes erupted only briefly from ongoing background ensemble activity, or were observed during nontask-related epochs as well. For example, the emergence of position representations during the delay period, even though not contained in the original discriminant analysis, was unexpected. As in many previous reports, delay activity for the first 10–15 sec was not selective for position or task phase (18,19,22,44). However, there was a definite bias toward encoding the appropriate nonmatch position on trials with delays longer than 15 sec. It is tempting therefore, to hypothesize that these patterns during the delay period are the neural signature of prospective coding of the nonmatch behavior required to complete the trial after the delay (45). These analyses demonstrate that it is possible to identify neural representations of task-relevant cognitive information in ensembles of hippocampal neurons on individual DNMS trials. The presence on single trials of phase and position information within the ensembles was consistent with extracted ‘‘average’’ patterns (i.e., discriminant functions) determined from large numbers of trials. Single trial codes are a necessary prerequisite for eventual implementation of neural prosthetics—because such devices will have to detect patterns of neural activity that are present in response to unitary events, and also be stable across different types of trials. The characterization of prospective encoding on single trials is therefore critical for neural prosthetic development, since it will be necessary to detect neural codes that predict and guide goal-directed behavior, not just represent it at the time of the behavioral response. This will require a finer temporal structure that is provided by the strict linear analyses applied here. However, the ability to detect such codes on single trials increases the likelihood that hippocampal neural prosthetics can achieve a level of functional viability appropriate to sustain and encode new events when needed.
Neural Codes in Hippocampal Ensembles
473
REFERENCES 1.
2.
3. 4.
5. 6.
7. 8. 9. 10.
11.
12.
13. 14.
15.
16. 17.
Georgopoulos AP, Kalaska JF, Caminiti R, Massey JT. On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex. J Neurosci 1982; 2:1527–1537. Nicolelis MA, Ghazanfar AA, Stambaugh CR, Oliveira LM, Laubach M, Chapin JK, Nelson RJ, Kaas JH. Simultaneous encoding of tactile information by three primate cortical areas. Nat Neurosci 1998; 1:621–630. Stanley GB, Li FF, Dan Y. Reconstruction of natural scenes from ensemble responses in the lateral geniculate nucleus. J Neurosci 1999; 19:8036–8042. Chapin JK, Moxon KA, Markowitz RS, Nicolelis MA. Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex [see comments]. Nat Neurosci 1999; 2:664–670. Wessberg J. Real-time prediction of hand trajectory by ensembles of cortical neurons in primates. Nature 2000; 408:361–365. Gao Y, Black MJ, Bienenstock E, Shoham E, Donoghue JP. Probabilistic inference of hand motion from neural activity in motor cortex. Advances in Neural Information Processing Systems 14. Boston: MIT Press, 2002. Serruya MD, Hatsopoulos NG, Paninski L, Fellows MR, Donoghue JP. Instant neural control of a movement signal. Nature 2002; 416:141–142. Taylor DM, Helms Tillery SI, Schwartz AB. Direct cortical control of 3D neuroprosthetic devices. Science 2002; 296:1829–1832. Wilson MA, McNaughton BL. Dynamics of the hippocampal ensemble code for space. Science 1993; 261:1055–1058. Brown EN, Frank LM, Tang D, Quirk MC, Wilson MA. A statistical paradigm for neural spike train decoding applied to position prediction from ensemble firing patterns of rat hippocampal place cells. J Neurosci 1998; 18: 7411–7425. Zhang K, Ginzburg I, McNaughton BL, Sejnowski TJ. Interpreting neuronal population activity by reconstruction: unified framework with application to hippocampal place cells. J Neurophysiol 1998; 79:1017–1044. Jensen O, Lisman JE. Position reconstruction from an ensemble of hippocampal place cells: contribution of theta phase coding. J Neurophysiol 2000; 83: 2602–2609. Warland DK, Reinagel P, Meister M. Decoding visual information from a population of retinal ganglion cells. J Neurophysiol 1997; 78:2336–2350. Nawrot M, Aertsen A, Rotter S. Single-trial estimation of neuronal firing rates: from single-neuron spike trains to population activity. J Neurosci Methods 1999; 94:81–92. Maynard EM, Hatsopoulos NG, Ojakangas CL, Acuna BD, Sanes JN, Normann RA, Donoghue JP. Neuronal interactions improve cortical population coding of movement direction. J Neurosci 1999; 19:8083–8093. Wible CG, Findling RL, Shapiro M, Lang EJ, Crane S, Olton DS. Mnemonic correlates of unit activity in the hippocampus. Brain Res 1986; 110:37–110. Eichenbaum H, Otto T, Cohen NJ. The hippocampus—what does it do? Behav Neural Biol 1992; 57:2–36.
474
Simeral et al.
18. Otto T, Eichenbaum H. Neuronal activity in the hippocampus during delayed non-match to sample performance in rats: evidence for hippocampal processing in recognition memory. Hippocampus 1992; 2:323–334. 19. Hampson RE, Heyser CJ, Deadwyler SA. Hippocampal cell firing correlates of delayed-match-to-sample performance in the rat. Behav Neurosci 1993; 107: 715–739. 20. Deadwyler SA, Hampson RE. Ensemble activity and behavior: what’s the code? Science 1995; 270:1316–1318. 21. Deadwyler SA, Bunn T, Hampson RE. Hippocampal ensemble activity during spatial delayed-nonmatch-to-sample performance in rats. J Neurosci 1996; 16: 354–372. 22. Wiebe SP, Staubli UV. Dynamic filtering of recognition memory codes in the hippocampus. J Neurosci 1999; 19:10562–10574. 23. Wood E, Dudchenko PA, Eichenbaum H. The global record of memory in hippocampal neuronal activity. Nature 1999; 397:613–616. 24. Suzuki WA, Eichenbaum H. The neurophysiology of memory. Ann NY Acad Sci 2000; 911:175–191. 25. Hampson RE, Deadwyler SA. Cannabinoids reveal the necessity of hippocampal neural encoding for short-term memory in rats. J Neurosci 2000; 20: 8932–8942. 26. Hampson RE, Simeral JD. ‘‘Keeping on track’’: firing of hippocampal neurons during delayed-nonmatch-to-sample performance. J Neurosci 2002; 22:RC198. 27. Hampson RE, Deadwyler SA. Ensemble codes involving hippocampal neurons are at risk during delayed performance tests. Proc Natl Acad Sci USA. 1996; 93:13487–13493. 28. Hampson RE, Deadwyler SA. Temporal firing characteristics and the strategic role of subicular neurons in short-term memory. Hippocampus 2003; 13: 529–541. 29. Marcoulides GA, Hershberger SL. Multivariate Statistical Methods. Mahwah, NJ: Lawrence Erlbaum Assoc, 1997. 30. Fisher RA. The use of multiple measurements in taxonomic problems. Ann Eugen 1936; 7:179–188. 31. Shannon CE, Weaver W. Mathematical Theory of Communication. Champaign, IL: University of Illinois Press, 1964. 32. Obermaier B, Neuper C, Guger C, Pfurtscheller G. Information transfer rate in a five-classes brain–computer interface. IEEE Trans Neural Syst Rehabil Eng 2001; 9:283–288. 33. Hampson RE, Deadwyler SA. LTP and LTD and the encoding of memory in small ensembles of hippocampal neurons. In: Baudry M, Davis, J, eds. Longterm Potentiation Vol. 3. Cambridge, MA: MIT Press, 1996:3:199–214. 34. Celebrini S, Newsome WT. Neuronal and psychophysical sensitivity to motion signals in extrastriate area MST of the macaque monkey. J Neurosci 1994; 14:4109–4124. 35. Britten KH, Newsome WT, Shadlen MN, Celebrini S, Movshon JA. A relationship between behavioral choice and the visual responses of neurons in macaque MT. Vis Neurosci 1996; 13:87–100.
Neural Codes in Hippocampal Ensembles 36.
37. 38. 39. 40. 41. 42. 43.
44.
45.
475
Dodd JV, Krug K, Cumming BG, Parker AJ. Perceptually bistable three-dimensional figures evoke high choice probabilities in cortical area MT. J Neurosci 2001; 21:4809–4821. Cook EP, Maunsell JH. Dynamics of neuronal responses in macaque MT and VIP during motion detection. Nat Neurosci 2002; 5:985–994. Tulving E, Markowitsch HJ. Episodic and declarative memory: role of the hippocampus. Hippocampus 1998; 8:198–204. Eichenbaum H. The hippocampus and declarative memory: cognitive mechanisms and neural codes. Behav Brain Res 2001; 127:199–207. Hampson RE, Simeral JD, Deadwyler SA. Distribution of spatial and nonspatial information in dorsal hippocampus. Nature 1999; 402:610–614. Tanaka K. Inferotemporal cortex and higher visual functions. Curr Opin Neurobiol 1992; 2:502–505. Young MP, Yamane S. Sparse population coding of faces in the inferotemporal cortex. Science 1992; 256:1327–1331. Hatsopoulos NG, Ojakangas CL, Paninski L, Donoghue JP. Information about movement direction obtained from synchronous activity of motor cortical neurons. Proc Natl Acad Sci USA 1998; 95:15706–15711. Colombo M, Gross CG. Responses of inferior temporal cortex and hippocampal neurons during delayed matching to sample in monkeys (Macaca fascicularis). Behav Neurosci 1994; 108:443–455. Frank LM, Brown EN, Wilson M. Trajectory encoding in the hippocampus and entorhinal cortex. Neuron 2000; 27:169–178.
21 Recent Advances Toward Development of Practical Brain–Machine Interfaces David J. Krupa Department of Neurobiology, Duke University Medical Center, Durham, North Carolina, U.S.A.
Recently, experimental results from several laboratories have demonstrated the feasibility of controlling electromechanical devices such as computer cursors or robotic arms in real time with signals derived directly from motor cortical areas in awake behaving animals (1–4). These results, along with numerous other recent technological advances, have generated renewed interest in the ongoing development of practical brain–machine interfaces (BMI) that could be used to restore lost functionality in patients suffering from paralysis or other severe neurological disorders (5–10). Although these recent experimental results are very promising, many significant problems remain to be solved before fully functional BMIs could be routinely used in humans as a therapeutic cure for disabilities such as paralysis. The present chapter will discuss various properties of a BMI, review recent advances in this developing technology, and examine problems remaining to be solved before such devices might progress from experimental prototypes to realistic solutions.
1.
BRAIN–MACHINE INTERFACES
The term brain–machine interface is not new (11), nor is the underlying concept behind such devices (12–14). A BMI broadly refers to a device that forms a communication link between the brain (and its ongoing neuronal 477
478
Krupa
activity) and some external electronic or electromechanical device, typically some form of prosthetic. Other terms, such as brain–computer interface or neural prosthetic generally refer to similar devices. Although scientists and engineers have been striving to develop and build practical, clinically useful BMIs for quite some time, progress has been slow. However, in recent years, simple forms of BMIs have emerged. Moreover, recent technological advances have inspired a renewed vigor and excitement in research and development of more sophisticated BMIs. A BMI typically can be classified into one of two general categories: input BMIs and output BMIs (5,9). As its name suggests, the primary function of an input BMI is to acquire information from the external environment and transduce this into a useable signal that can be communicated directly to the brain. As described below, some examples of input BMIs include auditory prostheses, that convert external acoustic signals into electrical stimulation of the auditory nerve, and visual prostheses that convert light signals (visual information) into stimulation of either the retina, optic nerve, or direct stimulation of the visual cortex. Needless to say, the purpose of these devices is typically to bypass a damaged or otherwise nonfunctional cochlea or retina so that information from these sensory realms can be communicated to the brain in a form that is useful to the implanted subject. In contrast with input BMIs, an output BMI acquires signals from the brain and transduces these signals into a form that can be used to control some external device or other mechanism such as a computer cursor, an artificial limb or robotic arm, or a stimulating device used to activate certain muscles. Output BMIs can typically be separated into two different categories, indirect output BMIs and direct output BMIs. An indirect BMI typically acquires brain activity through noninvasive techniques such as scalp EEG electrodes. A direct output BMI typically involves an invasive recording device that acquires brain activity by recording electrophysiological activity directly from individual neurons or small populations of neurons. A third category of BMI is one that combines both input and output BMIs to create a ‘‘closed loop’’ interface with the brain. In this particular type of BMI, command and control signals acquired from the brain by the output BMI are used to control some external device. At the same time, information from the external environment related to the ongoing activity generated by the output BMI is fed back directly to the brain via an input BMI so that output can be modulated and controlled more precisely. An example of a closed-loop input–output BMI might include a system in which an output BMI acquires motor control signals directly from motor cortical areas. Output from these motor areas are transduced and processed by the output BMI in order to control an artificial limb. In addition, tactile and=or proprioceptive information from the limb is fed back directly to the brain via an input BMI so that movement and positioning of the limb can be accurately and rapidly controlled (see Fig. 1). Although such a system
Development of Practical Brain–Machine Interfaces
479
Figure 1 Schematic diagram showing the major components of a hypothetical closed-loop input–output BMI. (A) The multichannel recording system would form the actual physical interface between brain and machine. This component would likely consist of multiple high-density intracortical recording arrays. (B) The signal-processing module would contain electronic circuits designed to extract single unit spiking activity from signals recorded by the recording electrodes. This module might contain high impedance input amplifiers, filters, and other components necessary to isolate and separate out individual action potentials. (C) This module would implement different mathematical algorithms optimized to extract command and control signals from the recorded neuronal data streams. (D) A telemetry interface would transmit the processed signals across the skin to a receiver located externally. (E) This device would consist of some form of controller interface that would convert the output command and control signals to drive the actual device. (F) A multidimensional artificial limb or robotic arm is one form of possible output device. Others might include computer cursors or possibly a stimulator device that could be used to stimulate and actuate the muscles of a paralyzed limb. In addition to these components, an output BMI would incorporate some form of feedback to form a closedloop control system. Visual feedback would be important in almost any device. In BMIs that use electromechanical devices such as a robotic limb, other forms of feedback would be required to convey information about loads or forces encountered by the device. This type of feedback may be in the form of vibromechanical stimulation of the skin or, possibly, a somatosensory input BMI that would convey somatosensory input from the device directly to the brain. (Adapted from Ref. 5.)
480
Krupa
would potentially provide invaluable assistance and improvement of life to countless patients who suffer from severe forms of paralysis, development of closed-loop input–output BMIs such as this are presently only in the very earliest experimental phases. As detailed below, several significant challenges remain to be solved before these types of BMIs might become practical solutions. 2.
INPUT BMIs
Although scientists and engineers have been working for many years on developing practical BMIs, at present, few BMIs have reached the level of development where they have become a commonly used solution in human patients. Below, two different types of input BMIs are briefly described. These two BMIs, auditory and visual prostheses, provide good examples of the progress and advancements that have been made in recent years in developing practical BMIs. 2.1.
Auditory Prostheses
As mentioned above, auditory prostheses are a particular class of input BMI that convert acoustic signals into electrical signals that are transmitted to the brain by stimulating neurons in the auditory pathway so that a patient using an auditory prosthesis can obtain usable auditory information about his or her surroundings. Auditory prostheses have been under development for several decades and were the first widely used form of BMI, of any type, in human patients (6). As of June 1999, over 30,000 multichannel auditory prostheses had been successfully implanted in deaf human patients ranging in ages from 12 months to over 80 years old. Indeed, at present, auditory prostheses are the only commercially manufactured BMIs in use. A multichannel auditory prosthesis consists of four main components. The first component is a microphone that captures acoustic signals and converts them to an electrical signal (15,16) that is then fed into the second main component, a signal processor=power supply module. The signal processor converts the raw signal from the microphone into multiple subcomponents using filters and other electronic processing devices (17). These multiple signals are then encoded and converted into electrical waveforms that are used to stimulate neurons in the auditory system (18,19). A variety of different encoding strategies may typically be used including: analog, which follows the actual temporal characteristics of the original signal; pulse trains that can be rate-modulated according to characteristics of the original signal; or amplitude-modulated pulse trains in which the amplitude of the pulses correspond to particular characteristics of the original signal. The power supply is typically a rechargeable battery which, along with the microphone and signal processor, are worn as external devices. The processed electrical
Development of Practical Brain–Machine Interfaces
481
signals are then passed to the stimulating electrodes via the third main component, typically a transcutaneous connector implanted behind the ear. Finally, the electrical signals from the transcutaneous connector are sent to an array of stimulating electrodes which are surgically implanted near neurons in the auditory pathway. The most common site for implanting the stimulating electrodes is in the cochlea, typically in the scala tympani. The stimulating electrodes can be either a bipolar configuration where current passes between two electrode contacts, or a monopolar configuration where current passes between an electrode contact and an extracochlear return electrode (20). In either case, stimulating currents from the electrodes activate auditory nerve fibers in the scala tympani thereby producing the signals that are used by the patient to deduce auditory information. In patients with auditory nerve damage, a cochlear implant would not be effective since input from cochlear stimulation could not be relayed to higher auditory centers. In these patients, arrays of stimulating electrodes may be placed on or directly in the cochlear nucleus itself (21,22). The design and location of the stimulation-electrodes remain an area of active research and development. For instance, thin-film high density multicontact electrodes implanted in the cochlear nucleus may provide advantages over cochlear implants (23). Although auditory prosthetic BMIs have been implanted in many thousands of patients with the vast majority of them providing at least some benefit, the degree of success remains highly variable from patient to patient (24,25). Improving the effectiveness of auditory prostheses is another area of active research. Among the variables that might affect the outcome of an implant include: the design of the stimulating electrode, the mode of stimulation, and postoperative training in how to use the prosthesis. Therefore, although auditory prostheses are presently the most well-developed BMIs available for human use, improving their design and use continues. 2.2.
Visual Prostheses
Another example of an input BMI is a visual prosthesis (7,8). Similar to auditory prostheses, the basic function of a visual BMI is to acquire signals in the visual spectrum and convert these light signals into electrical signals that can be used to stimulate a particular region of the visual system so that the user of the device can obtain useful visual input, such as reading text or navigating through an environment. Obviously, these particular types of input BMIs would be intended for patients with severe damage to or disease of a particular portion of their visual system. The visual prostheses would be used to bypass the damaged region of the visual system by communicating visual information, in the form of electrical stimulation, to the remaining intact portion of the visual system. As with auditory prostheses, research and development of visual prostheses has been underway for many years.
482
Krupa
Initial experiments, performed over a quarter century ago, used large numbers of subdural stimulating electrodes implanted over the primary visual cortex to evoke focal visual sensations termed phosphenes (26,27). Although initial results appeared promising, a practical visual prosthesis failed to emerge, primarily because of limitations of the surface stimulating electrodes (28). However, research and development in this field has continued and, using new technologies derived primarily from the semiconductor field, several new approaches towards developing a practical visual prosthesis are underway. These new approaches have focused on producing improved stimulating electrode arrays that are smaller and more efficient at producing focal stimulation than that achievable using subdural electrodes. Also, as described below, implantation of these new devices in different regions of the visual system such as visual cortex, the optic nerve, or the retina is being tested. One approach currently being tested is to implant arrays of high density stimulation electrodes directly into visual cortex and to use microstimulation of these electrodes to generate visual sensations (29–32). An example of the stimulation arrays that might be used in such a visual prosthesis is the Utah intracortical electrode array (33). This particular device typically consists of an array of 100 micro electrodes each about 1.5 mm in length that is rapidly inserted into cortical tissue (34). Once implanted, different patterns of stimuli are applied to the electrodes of the array in an effort to evoke distinct patterns of visual sensations that can be used by the patient, for instance, to identify different text objects. Some of the advantages of an intracortical implant over subdural electrodes are that much lower stimulus currents are necessary to evoke phosphenes, and their apparent density can be significantly enhanced (30). However, although this approach remains promising, development of this type of BMI remains only in the earliest phases and several issues remain to be addressed. For instance, the complex spatial organization of visual cortex is such that two adjacent cortical regions do not necessarily correspond to two adjacent areas in space. Therefore, patterned electrical stimulation of an intracortical array might not produce patterned visual perceptions (8). Also, small regions of visual cortex are highly specialized for parameters such as color, motion, eye preference, etc. Thus, it may be difficult to evoke simple perceptions even when stimulating relatively large numbers of neurons. Finally, another potential drawback of using intracortical stimulation is that the visual field is represented across a relatively large, convoluted cortical area, thereby making it difficult to stimulate large areas of the visual field using currently available stimulating electrode technologies. In contrast with direct intracortical stimulation, an alternative approach towards development of a visual prosthesis is to use optic nerve stimulation (35). With this approach, spiral nerve cuff electrodes (36–38) would be used to stimulate the optic nerve to produce visual sensations.
Development of Practical Brain–Machine Interfaces
483
One advantage of optic nerve stimulation is that the entire visual field is represented in a relatively small space, thus, it may be possible to evoke phosphenes in the entire visual space using multiple-contact cuff electrodes and complex patterns of stimulation (35,39,40). Alternatively, instead of using nerve cuff electrodes, it may be possible to use intraneural microelectrode arrays (41) to stimulate the optic nerve (7). This approach has the possible advantage of requiring smaller currents to induce phosphenes and it might provide greater control of the evoked visual sensations compared with what might be achievable with the surface electrode contacts used in cuff electrodes. However, as with direct intracortical stimulation, the possibility of using stimulation of the optic nerve as the basis of a visual BMI remains in the earliest phases of experimental development. For instance, long-term stability of the stimulating electrodes remains to be tested as well as the ability to evoke patterned visual sensations. A third class of visual prostheses involves retinal stimulation. Common causes of vision loss in humans are diseases such as retinitis pigmentosa, which result in degeneration of retinal pigment epithelial cells. However, with diseases such as this, a significant number of inner retinal nuclear cells remain intact. Thus, the possibility exists to stimulate these remaining retinal cells to evoke visual sensations. To achieve this, several groups are developing subretinal prostheses which are, effectively, artificial silicon retinas (ASR) (42–45). An ASR is a device that typically consists of a large number (many thousands) of individual subunits, each measuring 20 20 mm (46) that is implanted behind the retina between the sclera and the bipolar cells. Each subunit consists of a silicon microphotodiode and an associated stimulating electrode. The microphotodiode of each subunit effectively acts as a tiny solar cell that converts light energy in the 500–1100 nm spectrum into electrical voltages which can then be used to stimulate remaining functional retinal cells through the associated stimulating electrode. Unlike all of the BMIs discussed above, this particular device requires no external power supply or other associated electronics; it is essentially a self-contained man-made replacement for the lost retinal cells. However, although this particular visual prosthesis has many desirable characteristics, several limitations remain. First, such a device will only be useful in patients with outer retinal degeneration. Damage in other areas of the visual system, such as the optic nerve would preclude use of this device. Also, at present, the photodiodes are relatively inefficient. Thus, in order to generate stimulating voltages sufficient to evoke visual- sensations, unnaturally bright light or some form of image intensification is necessary. Finally, long-term biocompatibility remains to be tested. However, this particular form of input BMI remains a very promising solution for a significant portion of the population who suffer from retinal degenerative diseases (47,48). As with auditory prostheses, fully functional visual prostheses (with the possible exception of future ASRs) will require a number of components
484
Krupa
in addition to the stimulation electrodes described above that form the actual interface between brain and machine. The first component in the system will be a front end camera that captures visual (light) information and converts this to an electrical signal. As with the microphone in auditory prostheses, useful front end cameras should be developable using current technologies (31,49). Another component would include a signal-processing device that maps the visual space onto the retinotopy of the target structure to be stimulated so that visual perception can be optimized for each individual implant (50). A third component would include either a percutaneous connector or radio frequency telemetry device to transmit signals and power to the multichannel stimulation electrodes (32,51). Finally, some form of stimulus controller=driver circuits will be necessary so that input signals to the stimulating electrodes evoke consistent and repeatable visual sensations over extended periods. In summary, the auditory and visual prostheses described above provide good examples of the substantial progress that has been made in the past several years towards developing practical BMIs. The intricate and complex design of these BMIs exemplify just some of the very difficult technical challenges that must be overcome in order to directly communicate sensory information to the brain. Further, although much progress has been made in development of these input BMIs, many significant challenges remain unresolved.
3.
OUTPUT BMIs
As described above, the basic function of an output BMI is to acquire command and control signals from the brain and use these signals to control some external device. The intended purpose of an output BMI typically would be to restore lost motor function in patients suffering from severe paralysis or other disruption in neuromuscular control. Numerous diseases, stroke, or traumatic brain and spinal cord injuries resulting from accidents, violence or falls can result in severe or complete loss of voluntary muscle control (52–54). In many cases, patients require continuous assistance to accomplish even the most basic motor tasks. Presently, ongoing experimentation in many different fields is attempting to find cures for these disabilities. As described below, recent experimental results and technological advances raise the possibility that output BMIs might be used to restore at least some basic motor functions in patients suffering from severe motor impairments. Figure 1 shows the basic components of a typical output BMI (5). The first component represents the particular technique used to acquire the motor signals from the brain. These signals can be acquired indirectly, for instance, through scalp EEG electrodes, or directly, for instance, through
Development of Practical Brain–Machine Interfaces
485
chronically implanted intracortical recording electrodes. The second component is a device that performs signal processing such as amplification, thresholding and filtering of the raw, neural signals. The third component performs real-time analysis of the output signal to extract the critical control commands. The fourth component is a transcutaneous interface (not used with noninvasive, indirect BMIs). Finally, the last component is some form of actuator circuit that translates the extracted control signals into a signal appropriate for driving the output device, for instance a computer cursor or robotic arm. 3.1.
Indirect Output BMIs
As mentioned above, indirect BMIs use noninvasive techniques to record neural activity. Output motor control signals are then extracted from the recorded activity and these control signals are used to activate some external device, typically a computer cursor. Standard EEG electrodes are commonly used in many variations of indirect BMIs. The neural signal recorded by these types of electrodes represents the massed activity of many neurons. Alternatively, in some instances, more invasive techniques are used to implant similar electrodes on the dura or cortical surface. As described below, different brain signals such as visual evoked potentials (VEP), slow cortical potentials (SCP), or P300 evoked potentials can be used as the source command signal. Systems based on these indirect BMIs currently provide a mechanism for severely paralyzed humans to control an output device. Several output BMIs have been designed that use VEP to control a particular output device (55–57). In these systems, evoked potentials are typically recorded from the scalp over the visual cortex as the user gazes at different symbols on a computer screen. First, the users fixate their gaze on a particular symbol that they wish to select. Then, the different systems use various techniques to manipulate the images on the screen in order to modulate the VEP recorded by the scalp electrodes. By analyzing the effects of image manipulation on the VEP, the particular symbol that the user was gazing at and wished to select is accurately and efficiently extracted. Using such a system, a patient can communicate at a rate up to 10–12 words=min (57). Also, VEP based BMIs can provide a useful option for disabled patients. However, these systems depend on the user’s ability to control gaze direction, thus, VEP based systems might not be practical for all situations. These VEP systems fall into a category of BMIs called dependent BMIs since their output is dependent on active muscular control (53). An alternative to VEP is to use SCP as the source of control signals. SCPs are very low frequency EEG signals generated in cortex that shift over 0.5–10s (58). Using biofeedback training techniques, people can learn to control and adjust recorded SCP signals (59,60) thereby making these EEG signals potentially useful for controlling external devices. A spelling
486
Krupa
device has recently been described in which two severely paralyzed patients suffering from advanced amyotrophic lateral sclerosis learned to control a computer cursor by adjusting the direction their SCP deviated from baseline (61). Although writing sentences using this device is slow, approximately 2 characters=min, these completely paralyzed individuals now have the ability to communicate, a possibility that has not previously existed for such severely affected patients. In contrast with VEP based BMIs that depend on active muscle control, this type of SCP-based device is referred to as a ‘‘thought translation device’’ since it does not depend on muscular movements to generate cortical control signals (53). One drawback in the use of SCP signals to control an output BMI is that extensive training is required in order to learn to modulate the SCP. An alternative BMI that does not require initial training is based on P300 evoked potentials. If a particularly significant or novel auditory, visual, or somatosensory signal is presented amidst other frequent or routine stimuli, a positive peak occurs approximately 300 ms later (P300) in the EEG signal recorded over the parietal cortex (62,63). This P300 signal has been used in an output BMI in which a 6 6 matrix of letters, numbers, or other symbols is presented to a user (64,65). The user focuses attention on a particular letter or symbol in the matrix while rows and columns are flashed. Flashing the rows and columns containing the attended letter are ‘‘rare’’ events that evoke a P300 which can be detected and used as a control signal to select the particular letter or symbol. In addition to requiring no initial training in order to use this device, auditory or somatosensory stimuli can be used instead of visual stimuli to evoke a P300, a potentially important feature for patients who suffer visual impairment (66,67). However, although BMIs based on the P300 signal appear promising, thus far, this system has only been tested on healthy subjects and only for short periods of time. It is possible that the P300 could habituate over time with extended use so that performance of the BMI deteriorates (68). The indirect output BMIs described above are just a few examples of many such devices (53). As neural prosthetics intended to restore some level of communication or control to severely paralyzed patients, these devices have several advantageous features. For instance, they all typically use scalp EEG signals, thereby requiring no invasive surgical implants. Further, the electronic hardware and components used in the systems rely on currently available technologies, making them readily available and cost effective. Because no devices are surgically implanted, these BMIs can be easily upgraded as new technologies develop. However, indirect BMIs also have certain limitations. Routine use can be cumbersome; for instance, attaching multielectrode EEG recording systems can take an hour. Also, even under optimal conditions, the output of indirect BMIs is very slow, often only a few characters or choices per minute. Thus, writing entire sentences can take hours to complete (61). The slowness of these systems is a direct result of the
Development of Practical Brain–Machine Interfaces
487
nature of the brain signals used for their control. The EEG signals used in these indirect BMIs represent a highly filtered average of the summed activity of vast numbers of cortical neurons. Thus, using EEG, it is impossible to obtain a direct readout of the actual neuronal spiking patterns that encode intended movements. Also, these systems are all attention related; therefore, distractions can degrade their performance. Despite these limitations, however, indirect BMIs offer a currently available, useable means of communication for people who otherwise have extremely limited ability to communicate. 3.2.
Direct Output BMIs
In contrast with the indirect BMIs described above, a direct output BMI uses intracortically implanted electrodes to record simultaneously the activity of large populations of individual neurons that encode movement or its intent. From these ensembles of neural activity, command and control signals are derived in real time and used to actuate an external device such as a computer cursor or mechanical robotic arm. Because of the nature of the neural signals recorded with a direct BMI compared with indirect BMIs, control of external devices can be significantly more rapid, complex, versatile, and precise. However, as described below, the technological demands and difficulties of recording and processing signals from large populations of single neurons simultaneously and in real time are also significantly greater than EEG signals. Despite these technical demands, recent experimental results have demonstrated the feasibility of precise, real-time control of electronic or electromechanical devices with direct BMIs. It has long been known that activity patterns of cortical motor neurons encode motor control signals, including force and direction (69–71). Likewise, efforts to utilize these signals to control external devices have been underway for many years (14). Although results of this earlier work were encouraging, success in developing a practical direct output BMI was limited. This earlier work highlighted at least two areas that required further development before a feasible BMI could be implemented: electrode design=fabrication and signal-processing techniques. As described below, significant advances in both of these areas opened the way for the recent experimental demonstrations of functional direct output BMIs. The first demonstration that the simultaneously recorded activity of large ensembles of neurons could be converted into a signal that could control an external device (a one-dimensional robotic arm) in real time was a recent study by Chapin et al. (1) using rats. In this experiment, rats were trained to press a bar that moved a robotic arm that delivered a water reward to the rat. At the same time, the activity of several dozen neurons was recorded simultaneously in motor cortex and thalamic areas. Principal component analysis was used to determine that a burst of activity in many
488
Krupa
of the recorded neurons preceding the bar press could be used to predict the movement of the robotic arm. The weighted activity of 32 neurons was summed and used to derive a neural population function that could be used to drive the robotic arm to deliver the water reward. Then, control of the robotic arm was switched from the bar press to the neural population function. Upon switching control, two out of six rats were immediately able to control the robotic arm to obtain water reward with 100% efficiency. Two more rats reached 70–90% efficiency in controlling the arm using the neuronal population function. Importantly, because neural activity was controlling the arm movement and not the act of bar pressing, rats learned to press the bar less and in some cases stopped pressing entirely. This indicates that neuronal activity in the absence of any associated movement was sufficient to control the robotic device, an important result in terms of developing a prosthetic device for paralyzed patients. Although the results obtained in rats were an important demonstration that ensemble neuronal activity can be used to control devices in real time, a number of critical questions remained unanswered. In particular, could similar techniques be used to control a more sophisticated device in multiple dimensions and could this control be maintained in a stable manner for extended periods of time, for instance months or years? To examine these issues, Wessberg et al. trained two owl monkeys to perform two different arm related motor tasks (3,10). The first task involved one-dimensional hand movements to displace a manipulandum in one of two directions (left or right). In the second task, the monkeys were trained to make three dimensional hand movements to reach for small pieces of food placed randomly at one of four different positions on a tray. The monkeys were then implanted with multiple arrays of microwire recording electrodes into several cortical areas involved in generating motor movements including premotor, primary motor, and posterior parietal cortical areas. The activity of up to 100 neurons within these regions was recorded simultaneously as the monkeys performed the tasks. Both linear and nonlinear neural network based algorithms were used independently to extract motor control signals that were then used to predict hand position in real time. Using these methods, real-time predictions of hand movements were generated that were highly statistically correlated with the actual hand movements made by the monkey. Further, the output of the linear or non-linear algorithms was used to control, in real time, a multidimensional robotic arm (72) located near the behaving monkey or remotely, hundreds of miles away in a separate laboratory (10). Also, the contribution of each of the cortical areas where recordings were made was examined and quantified. It was found that neuronal activity in all areas contained at least some predictive information about the hand movements, although premotor areas provided the highest information content. Finally, the stability of these recordings and hand predictions were shown to remain for at least 24 months. Collectively, these
Development of Practical Brain–Machine Interfaces
489
results (and results of other recent studies, described below) provide an important initial demonstration of the feasibility of using a direct output BMI as a potential neural prosthetic device for severely paralyzed humans. 3.3.
Technical Challenges in Developing a Practical Output BMI for Humans
Although the results described above demonstrate the feasibility of a direct output BMI, many technical challenges remain to be solved before such devices might one day be considered a practical solution for patients suffering from severe forms of paralysis. As noted above, two areas of particular concern are improving the performance and long-term stability of the implanted multielectrode recording arrays and miniaturizing the associated electronic signal-processing devices necessary to extract the critical motor control signals from the recorded neural activity. Both of these areas are currently the focus of active research and development. 3.3.1. Recording Electrode Arrays Perhaps one of the most critical components regarding the long-term functionality of any direct output BMI is the implanted recording electrode arrays. Obviously, degradation in the quality and=or number of recorded neurons would have a significant negative impact on the performance of the BMI. Further, unlike virtually all other components of any direct output BMI, replacement or repair of a damaged or otherwise nonfunctional recording array would be very difficult if not impossible without causing significant trauma to the cortical tissue in which it was implanted. Thus, development of recording devices suitable for implantation in humans that can reliably record neuronal activity for many years will be essential for any neural prosthetic of this type. Recently, several new electrode designs have been developed that are intended to address these issues (for recent reviews, see Refs. 73,74). One particular design that has provided good results in several animal preparations including rats, owl monkeys, and, more recently, macaque monkeys are microwire electrode arrays (75–79). This electrode design has been shown to work effectively in monkeys for periods of over 2 years (3,5). These particular devices consist of arrays of ultrafine diameter microwire electrodes attached to miniaturized printed circuit boards (see Fig. 2). Among the advantages of this design are: (i) a very compact design allowing implantation of multiple arrays in closely neighboring cortical areas; (ii) the ability to individually insert small subsets of a larger array during surgery, thereby minimizing damage to the cortical tissue; (iii) the individual micro wires are much more flexible than electrodes based on rigid substrates, thereby allowing the electrodes to move with cortical tissue movement; (iv) the diameter of each electrode wire is small and fixed along its entire length
490
Krupa
Figure 2 Examples of some critical components in an output BMI. (A) An array of microwire recording electrodes in which electrodes can be moved deeper long after surgical implantation through the use of miniaturized microdrives. Electrodes are 12 mm diameter and spaced about 250 mm apart. Multiple arrays would be implanted side by side to increase electrode density. Scale bar ¼ 500 mm. (B) A high-density microwire recording array consisting of 128 fixed position electrodes in a 16 8 pattern. Arrays such as these have been used to record single unit activity in motor cortical areas in monkeys for periods of over 2 years (3,10). Scale bar ¼ 2 mm. (C) A 36 channel high-input impedance VLSI amplifier circuit designed by J. Morizio, I. Obeid, and P. Wolf, in the departments of Biomedical and Electrical and Computer Engineering at Duke University. This chip is used for the first stage of amplification of neuronal signals recorded from implanted microelectrodes. The actual VLSI circuit is the rectangular chip highlighted by the black arrow; other components are support connectors used to test the chip. In an output BMI, a device such as this would be incorporated directly with the microwire electrodes, thereby minimizing the overall package size. (D) A 16 channel preamplifier=signal processor also designed by J. Morizio, I. Obeid, and P. Wolf to amplify and preprocess recorded neuronal signals. Each channel consist of (1) a linear band-pass filter with a gain of 150, (2) a single pole high pass switch capacitor filter, (3) a two pole low pass switch capacitor filter, and (4) a low pass filter with a selectable gain of 10 or 40. Also included on the chip are a voltage controlled oscillator and clock dividers for clock generation and tuning, and an AC ground buffering amplifier. Overall device size is approx. 2 mm2. (Panel A adapted from Ref. 80; panels B–D adapted from Ref. 79.)
(unlike other designs that become significantly larger along the shaft), thereby minimizing the chance of displacement and damage to tissue located above the recording tip. Further, microwire electrodes are currently being used along with ultracompact microdrives that allow small subsets of
Development of Practical Brain–Machine Interfaces
491
microwires in an implanted array to be moved long after surgical implantation, thereby allowing electrodes with degraded signal quality to be moved slightly to improve the recorded signals (80). However, although these microwire electrode designs have many promising features, they have not yet been tested for long-term effectiveness in human subjects. Other types of intracortical electrode arrays are also currently in various phases of testing and development. For instance, the Utah intracortical electrode array (33), described above, has been used to record ensemble neuronal activity in macaque monkeys as they performed arm reaching tasks (2,81). Yet another class of intracortical electrodes are the various types of thin-film electrodes (74,82–86). Typically, these electrodes consist of exposed metal pads that are deposited on a rigid substrate using thin-film technologies developed for integrated circuit manufacture (74). This particular design has the advantage that multiple electrode contacts can typically be incorporated onto a single probe, thereby greatly increasing the density of contacts for a given area of cortex. A novel alternative to the electrodes described above is an electrode called the cone electrode (87). This particular type of electrode consists of an insulated gold wire fixed inside a hollow glass cone. A piece of sciatic nerve is placed in the glass cone before implantation. Cortical neurites grow into the sciatic nerve in the cone from surrounding neurons and their electrical activity is recorded via the wire (or wires) in the cone. This type of electrode has been used in monkeys and humans (88,89). At present, however, the electrodes described above are only in the earliest phases of testing in human subjects. Much more testing remains in order to demonstrate long-term performance, biocompatibility, and stability in human patients. Nevertheless, results obtained using these various types of electrodes in monkeys gives hope that development of an intracortical recording array suitable for use in a human BMI should be possible in the near term. 3.3.2. Miniaturized Signal Processing Electronics A second area of development crucial for designing a practical direct output BMI for human use is miniaturizing the electronic signal processing components necessary to obtain and process the intracortical neuronal signals. For instance, in order to control a computer cursor with a direct BMI, the activity of many tens to hundreds of neurons must be recorded simultaneously (90). In typical experimental situations, the raw neuronal activity from each recording electrode is passed directly to a multichannel signal processing unit located near the restrained animal via high-density multiconductor cables and connectors. This external signal processor typically includes high gain=low noise amplifiers, filters, and digitizers that process the recorded activity so that single unit spiking activity can be extracted and used to obtain motor control signals. However, a practical output BMI for human
492
Krupa
use will require that much of these signal-processing elements be incorporated with the implanted intracortical electrode arrays (5,9,74). Miniaturization of these electronic components will require the design and manufacture of custom built high-density very large scale integrated (VSLI) circuits. This work is currently underway (see Fig. 2). For example, a 16 channel device that incorporates the low-noise high-gain amplifiers and filters onto a single integrated circuit has recently been developed, manufactured, and tested (91). A second generation of this device incorporating 32 or 64 channels of amplification and filtering is currently being tested (79). In addition to these signal-processing components, a telemetry system for multiplexing and transmitting the processed neuronal signals through the skin to an external device that extracts the motor commands and converts these to control signals is also being developed (92). Although these VLSI devices are only in relatively early phases of development, the progress achieved thus far using state of the art VLSI design and manufacturing techniques indicates that, within the next few years, integration of the signalprocessing components with the intracortical electrode arrays into a package suitable for implantation in humans should be possible. In summary, the output BMIs described above exemplify the significant progress that has been achieved in recent years towards developing practical solutions for aiding severely paralyzed humans in controlling external devices. Although the indirect output BMIs have certain limitations as a result of the signal recording methods (EEG), these systems offer potential solutions today or in the very near future to patients who otherwise have very limited options. The recent experimental results using direct output BMIs demonstrate the feasibility of such systems. However, a significant amount of development and testing remains before a direct output BMI could become a realistic option.
4.
CLOSED-LOOP INPUT–OUTPUT BMIs
Although the results of Chapin et al. (1) and Wessberg et al. (3), using direct output BMIs (above), were an important demonstration of the feasibility of controlling external devices with ensemble neuronal activity recorded directly in motor cortical areas, the design of the BMIs used in those experiments would not be directly applicable to a human prosthetic device since no feedback was incorporated into the system. For instance, the monkeys performing the three-dimensional hand reaching task had no knowledge of the position of the robotic arm while it was being controlled by their cortical activity. In contrast, a prosthetic device intended for use in humans would incorporate some form of feedback since this would significantly improve the performance and enhance control of the BMI (5,9,10). An output BMI that incorporates some form of feedback would be classified
Development of Practical Brain–Machine Interfaces
493
as a closed-loop BMI. As described below, feedback in its simplest form might involve visually monitoring the performance of the output device. A more sophisticated system might incorporate some form of tactile or proprioceptive feedback that could be transmitted to the subject by an input BMI, creating a closed-loop input–output BMI. However, although such a system would likely provide substantial assistance to severely paralyzed patients, development of a closed-loop input–output BMI is presently only in early conceptual phases; significant obstacles remain to be overcome before the feasibility of such a system could be demonstrated.
4.1.
Feedback Improves BMI Performance
Perhaps the simplest method of incorporating some form of feedback into a direct output BMI is to provide visual feedback about the motion of the controlled device. For example, a subject might view the movement of a cursor that is being controlled by direct output from motor cortex on a video monitor or computer screen. The ability to view the cursor’s movement should allow the subject to improve control of the movement by adjusting for errors in the cursor’s trajectory in real time. Recent work that incorporated visual feedback with direct output BMIs demonstrates that closedloop BMIs using visual feedback produce improved control compared with open-loop BMIs that do not use feedback (2,4,10,93). For example, in one study, monkeys were trained to control the movement of a cursor on a computer monitor by moving a manipulandum with their hand, while, at the same time, single unit neuronal activity from about 18 cells was recorded from motor cortex (4). An algorithm that tracks changes in cortical tuning parameters was used to extract movement control signals from the recorded neuronal activity. Then, control of the cursor movement was switched from hand control to brain control, but, in this case, subjects visually monitored the cursor movements created from their cortical activity in real time. It was found that switching from the handcontrolled task to the brain-controlled task with visual feedback resulted in global changes in the tuning parameters of the recorded neuronal population. As a result, cursor movements produced with visual feedback were significantly improved compared with cursor movements generated off-line without feedback. Also, the use of visual feedback allowed the cursor to be controlled with smaller populations of cells than without feedback. Results of the experiments that have incorporated some form of visual feedback clearly demonstrate that using feedback to create a closed-loop BMI significantly improves performance of the device. Thus, visual feedback would likely play a key role in any neural prosthetic designed for use in humans. However, there are certain limitations to the use of visual feedback. For instance, visual feedback would require focused attention on
494
Krupa
the movement of the device under control; any distractions would likely result in degradation of performance. Also, if the BMI is intended to control an electromechanical device, for example, a robotic arm with some form of a gripping device, feedback about the forces applied by the device would greatly enhance its usability. This type of feedback would not be possible using visual feedback alone. In this case, some form of tactile and=or proprioceptive feedback would be necessary. 4.2.
Other Types of Feedback
Although the potential value of tactile=proprioceptive feedback for a neural prosthetic of this type is understood, combining this type of feedback with a direct output BMI is presently only in conceptual phases (9,10). One possible means of providing tactile feedback would be to attach vibromechanical stimulators to a patch of skin and use different patterns of vibration frequencies to provide feedback about the ongoing activity of the robotic limb such as the force necessary to grasp or move an object (10). An alternative method of producing tactile=proprioceptive feedback would be to develop a somatosensory input BMI that uses electrical stimulation of some region of the somatosensory system to simulate somatosensory sensations, similar in concept to the visual BMIs described above. Recent studies have shown that animals can be trained to respond to different patterns of electrical stimulation of somatosensory cortical areas (94–96). However, much more work would be required before a somatosensory input BMI might be considered a feasible option for a human neural prosthesis. For instance, it is not known to what degree different patterned stimuli could simulate different types of tactile sensations. Also, the somatosensory system, as with other sensory systems, undergoes substantial plastic reorganization of sensory representations following perturbations of the system (97–100). For instance, a peripheral deafferentation causes significant reorganization of receptive field properties throughout the somatosensory system (101–104). This process begins immediately after the deafferentation and continues for months to years later. Thus, somatosensory sensations that might be evoked by a somatosensory input BMI could be relatively unstable over extended periods. However, as with the auditory and visual input BMIs described above, development of a practical somatosensory input BMI appears feasible. In summary, combining feedback with a direct output BMI to create a closed-loop system improves the performance of the device. At the minimum, visual feedback can be used to improve movement control. More sophisticated forms of feedback such as vibrotactile stimulation of a particular region of skin might be useful to convey input about forces or loads that the controlled device may encounter. A somatosensory input BMI might provide a more useable form of feedback. However, incorporation
Development of Practical Brain–Machine Interfaces
495
of such feedback systems with direct output BMIs have yet to be demonstrated, even at the experimental levels. 5.
CONCLUSIONS
The goal of restoring lost functionality, through the use of neural prosthetics, to the large number of people suffering from severe neurological disorders has been the focus of a substantial amount of research and development over the past several decades. Tremendous progress towards development of practical BMIs has been achieved in recent years. Many thousands of auditory input BMIs have been successfully implanted in deaf patients allowing them to obtain useful auditory information. Similarly, different types of indirect output BMIs currently provide a means of communication to severely paralyzed patients who, otherwise, lack this ability. Progress in development of visual input BMIs suggests that practical devices for use in humans might become available in the near term. Finally, recent experimental results with direct output BMIs demonstrate the feasibility of such systems. However, as promising as these advances may seem, several aspects of these devices will require significantly more development before they may afford users with the ability to fully interact with their surroundings in a manner that even partially resembles the abilities of an intact system. REFERENCES 1.
2. 3.
4. 5. 6.
7. 8.
Chapin JK, Moxon KA, Markowitz RS, Nicolelis MA. Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex. Nat Neurosci 1999; 2:664–670. Serruya MD, Hatsopoulos NG, Paninski L, Fellows MR, Donoghue JP. Instant neural control of a movement signal. Nature 2002; 416:141–142. Wessberg J, Stambaugh CR, Kralik ID, Beck PD, Laubach M, Chapin JK, Kim J, Biggs SJ, Srinivasan MA, Nicolelis MA. Real-time prediction of hand trajectory by ensembles of cortical neurons in primates. Nature 2000; 408: 361–365. Taylor DM, Tillery SI, Schwartz AB. Direct cortical control of 3D neuroprosthetic devices. Science 2002; 296:1829–1832. Nicolelis MA. Actions from thoughts. Nature 2001; 409(suppl):403–407. Pfingst BE. Auditory prostheses. In: Chapin JK, Moxon KA, eds. Neural Prostheses for Restoration of Sensory and Motor Function. Boca Raton: CRC Press, 2001:3–43. Maynard EM. Visual prostheses. Annu Rev Biomed Eng 2001; 3:145–168. Margalit E, Maia M, Weiland JD, Greenberg RJ, Fujii GY, Torres G, Piyathaisere DV, O’Hearn TM, Liu W, Lazzi G, Dagnelie G, Scribner DA, de Juan E Jr, Humayun MS. Retinal prosthesis for the blind. Surv Ophthalmol 2002; 47:335–356.
496
Krupa
9.
Donoghue JP. Connecting cortex to machines: recent advances in brain interfaces. Nat Neurosci 2002; (suppl 5):1085–1088. Nicolelis MA, Chapin JK. Controlling robots with the mind. Sci Am 2002; 287:46–53. Joseph AB. Design considerations for the brain–machine interface. Med Hypotheses 1985; 17:191–195. House WF, Urban J. Long term results of electrode implantation and electronic stimulation of the cochlea in man. Ann Otol Rhinol Laryngol 1973; 82: 504–517. Dobelle WH, Mladejovsky MG, Girvin JP. Artifical vision for the blind: electrical stimulation of visual cortex offers hope for a functional prosthesis. Science 1974; 183:440–444. Schmidt EM. Single neuron recording from motor cortex as a possible source of signals for control of external devices. Ann Biomed Eng 1980; 8:339–349. Hamacher V, Doering WH, Mauer G, Fleischmann H, Hennecke J. Evaluation of noise reduction systems for cochlear implant users in different acoustic environment. Am J Otol 1997; 18:S46–S49. Kompis M, Dillier N. Noise reduction for hearing aids: combining directional microphones with an adaptive beamformer. J Acoust Soc Am 1994; 96: 1910–1913. Dillier N, Frolich T, Kompis M, Bogli H, Lai WK. Digital signal processing (DSP) applications for multiband loudness correction digital hearing aids and cochlear implants. J Rehabil Res Dev 1993; 30:95–109. Kompis M, Vischer MW, Hausler R. Performance of compressed analogue (CA) and continuous interleaved sampling (CIS) coding strategies for cochlear implants in quiet and noise. Acta Otolaryngol 1999; 119:659–664. Battmer RD, Zilberman Y, Haake P, Lenarz T. Simultaneous analog stimulation (SAS)–continuous interleaved sampler (CIS) pilot comparison study in Europe. Ann Otol Rhinol Laryngol Suppl 1999; 177:69–73. Zwolan TA, Kileny PR, Ashbaugh C, Telian SA. Patient performance with the Cochlear Corporation ‘‘20 þ 2’’ implant: bipolar vs. monopolar activation. Am J Otol 1996; 17:717–723. Lenarz T, Moshrefi M, Matthies C, Frohne C, Lesinski-Schiedat A, Illg A, Rost U, Battmer RD, Samii M. Auditory brainstem implant: part I. Auditory performance and its evolution over time. Otol Neurotol 2001; 22:823–833. Lesinski-Schiedat A, Frohne C, Illg A, Rost U, Matthies C, Battmer RD, Samii M, Lenarz T. Auditory brainstem implant in auditory rehabilitation of patients with neurofibromatosis type 2: Hannover programme. J Laryngol Otol Suppl 2000; 15–17. Anderson DJ, Najafi K, Tanghe SJ, Evans DA, Levy KL, Hetke JF, Xue XL, Zappia JJ, Wise KD. Batch-fabricated thin-film electrodes for stimulation of the central auditory system. IEEE Trans Biomed Eng 1989; 36:693–704. Tyler RS, Kelsay DM, Teagle HF, Rubinstein JT, Gantz BJ, Christ AM. 7-year speech perception results and the effects of age, residual hearing and preimplant speech perception in prelingually deaf children using the nucleus and clarion cochlear implants. Adv Otorhinolaryngol 2000; 57:305–310.
10. 11. 12.
13.
14. 15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Development of Practical Brain–Machine Interfaces
497
25. Gantz BJ, Tyler RS, Knutson JF, Woodworm G, Abbas P, McCabe BF, Hinrichs J, Tye-Murray N, Lansing C, Kuk F, et al. Evaluation of five different cochlear implant designs: audiologic assessment and predictors of performance. Laryngoscope 1988; 98:1100–1106. 26. Brindley GS, Donaldson PE, Falconer MA, Rushton DN. The extent of the region of occipital cortex that when stimulated gives phosphenes fixed in the visual field. J Physiol 1972; 225:57P–58P. 27. Dobelle WH, Mladejovsky MG. Phosphenes produced by electrical stimulation of human occipital cortex, and their application to the development of a prosthesis for the blind. J Physiol 1974; 243:553–576. 28. Brindley GS. Effects of electrical stimulation of the visual cortex. Hum Neurobiol 1982; 1:281–283. 29. Bak M, Girvin JP, Hambrecht FT, Kufta CV, Loeb GE, Schmidt EM. Visual sensations produced by intracortical microstimulation of the human occipital cortex. Med Biol Eng Comput 1990; 28:257–259. 30. Schmidt EM, Bak MJ, Hambrecht FT, Kufta CV, O’Rourke DK, Vallabhanath P. Feasibility of a visual prosthesis for the blind based on intracortical microstimulation of the visual cortex. Brain 1996; 119 (Pt 2):507–522. 31. Normann RA, Maynard EM, Guillory KS, Warren DJ. Cortical implants for the blind. IEEE Spectrum 1996; 33:54–59. 32. Normann RA, Maynard EM, Rousche PJ, Warren DJ. A neural interface for a cortical vision prosthesis. Vis Res 1999; 39:2577–2587. 33. Maynard EM, Nordhausen CT, Normann RA. The Utah intracortical electrode array: a recording structure for potential brain–computer interfaces. Electroencephalogr Clin Neurophysiol 1997; 102:228–239. 34. Rousche PJ, Normann RA. A method for pneumatically inserting an array of penetrating electrodes into cortical tissue. Ann Biomed Eng 1992; 20:413–422. 35. Veraart C, Raftopoulos C, Mortimer JT, Delbeke J, Pins D, Michaux G, Vanlierde A, Parrini S, Wanet-Defalque MC. Visual sensations produced by optic nerve stimulation using an implanted self-sizing spiral cuff electrode. Brain Res 1998; 813:181–186. 36. Sweeney JD, Ksienski DA, Mortimer JT. A nerve cuff technique for selective excitation of peripheral nerve trunk regions. IEEE Trans Biomed Eng 1990; 37:706–715. 37. Naples GG, Mortimer JT, Scheiner A, Sweeney JD. A spiral nerve cuff electrode for peripheral nerve stimulation. IEEE Trans Biomed Eng 1988; 35: 905–916. 38. Sweeney JD, Mortimer JT. An asymmetric two electrode cuff for generation of unidirectionally propagated action potentials. IEEE Trans Biomed Eng 1986; 33:541–549. 39. Parrini S, Delbeke J, Legat V, Veraart C. Modelling analysis of human optic nerve fibre excitation based on experimental data. Med Biol Eng Comput 2000; 38:454–464. 40. Parrini S, Delbeke J, Romero E, Legat V, Veraart C. Hybrid finite elements and spectral method for computation of the electric potential generated by a nerve cuff electrode. Med Biol Eng Comput 1999; 37:733–736.
498
Krupa
41. Branner A, Normann RA. A multielectrode array for intrafascicular recording and stimulation in sciatic nerve of cats. Brain Res Bull 2000; 51:293–306. 42. Zrenner E, Miliczek KD, Gabel VP, Graf HG, Guenther E, Haemmerle H, Hoefflinger B, Kohler K, Nisch W, Schubert M, Stett A, Weiss S. The development of subretinal microphotodiodes for replacement of degenerated photoreceptors. Ophthalmic Res 1997; 29:269–280. 43. Zrenner E, Gekeler F, Gabel VP, Graf HG, Graf M, Guenther E, Haemmerle H, Hoefflinger B, Kobuch K, Kohler K, Nisch W, Sachs H, Schlosshauer B, Schubert M, Schwahn H, Stelzle M, Stett A, Troeger B, Weiss S. Subretinal microphotodiode array as replacement for degenerated photoreceptors. Ophthalmologe 2001; 98:357–363. 44. Peachey NS, Chow AY. Subretinal implantation of semiconductor-based photodiodes: progress and challenges. J Rehabil Res Dev 1999; 36:371–376. 45. Peyman G, Chow AY, Liang C, Chow VY, Perlman JI, Peachey NS. Subretinal semiconductor microphotodiode array. Ophthalmic Surg Lasers 1998; 29:234–241. 46. Chow VY, Peachey NS. The subretinal microphotodiode array retinal prosthesis. Ophthalmic Res 1998; 30:195–198. 47. Curcio CA, Medeiros NE, Millican CL. Photoreceptor loss in age-related macular degeneration. Invest Ophthalmol Vis Sci 1996; 37:1236–1249. 48. Bunker CH, Berson EL, Bromley WC, Hayes RP, Roderick TH. Prevalence of retinitis pigmentosa in Maine. Am J Ophthalmol 1984; 97:357–365. 49. Wyatt JJR. Ocular implants for the blind. IEEE Spectrum 1996; 33:47–53. 50. Eckmiller R. Learning retina implants with epiretinal contacts. Ophthalmic Res 1997; 29:281–289. 51. Troyk PR, Schwan MA. Closed-loop class E transcutaneous power and data link for microimplants. IEEE Trans Biomed Eng 1992; 39:589–599. 52. Carter GT. Rehabilitation management in neuromuscular disease. J Neurol Rehabil 1997; 11:69–80. 53. Wolpaw JR, Birbaumer N, McFarland DJ, Pfurtscheller G, Vaughan TM. Brain–computer interfaces for communication and control. Clin Neurophysiol 2002; 113:767–791. 54. Nobunaga AI, Go BK, Karunas RB. Recent demographic and injury trends in people served by the model spinal cord injury care systems. Arch Phys Med Rehabil 1999; 80:1372–1382. 55. Middendorf M, McMillan G, Calhoun G, Jones KS. Brain–computer interfaces based on the steady-state visual-evoked response. IEEE Trans Rehabil Eng 2000; 8:211–214. 56. Vidal JJ. Toward direct brain–computer communication. Annu Rev Biophys Bioeng 1973; 2:157–180. 57. Sutter EE. The brain response interface: communication through visually induced electrical brain responses. J Microcomput Appl 1992; 15:31–45. 58. Birbaumer N, Elbert T, Canavan AG, Rockstroh B. Slow potentials of the cerebral cortex and behavior. Physiol Rev 1990; 70:1–41. 59. Elbert T, Rockstroh B, Lutzenberger W, Birbaumer N. Biofeedback of slow cortical potentials. I. Electroencephalogr Clin Neurophysiol 1980; 48:293–301.
Development of Practical Brain–Machine Interfaces
499
60. Kotchoubey B, Schleichert H, Lutzenberger W, Birbaumer N. A new method for self-regulation of slow cortical potentials in a timed paradigm. Appl Psychophysiol Biofeed 1997; 22:77–93. 61. Birbaumer N, Ghanayim N, Hinterberger T, Iversen I, Kotchoubey B, Kubler A, Perelmouter J, Taub E, Flor H. A spelling device for the paralysed. Nature 1999; 398:297–298. 62. Donchin E, Smith DB. The contingent negative variation and the late positive wave of the average evoked potential. Electroencephalogr Clin Neurophysiol 1970; 29:201–203. 63. Sutton S, Braren M, Zubin J, John ER. Evoked-potential correlates of stimulus uncertainty. Science 1965; 150:1187–1188. 64. Farwell LA, Donchin E. Talking off the top of your head: toward a mental prosthesis utilizing event-related brain potentials. Electroencephalogr Clin Neurophysiol 1988; 70:510–523. 65. Donchin E, Spencer KM, Wijesinghe R. The mental prosthesis: assessing the speed of a P300-based brain–computer interface. IEEE Trans Rehabil Eng 2000; 8:174–179. 66. Roder B, Rosier F, Hennighausen E, Nacker F. Event-related potentials during auditory and somatosensory discrimination in sighted and blind human subjects. Brain Res Cogn Brain Res 1996; 4:77–93. 67. Glover AA, Onofrj MC, Ghilardi MF, Bodis-Wollner I. P300-like potentials in the normal monkey using classical conditioning and an auditory ‘oddball’ paradigm. Electroencephalogr Clin Neurophysiol 1986; 65:231–235. 68. Ravden D, Polich J. On P300 measurement stability: habituation, intra-trial block variation, and ultradian rhythms. Biol Psychol 1999; 51:59–76. 69. Georgopoulos AP. Population activity in the control of movement. Int Rev Neurobiol 1994; 37:103–119; discussion 121–103. 70. Humphrey DR, Schmidt EM, Thompson WD. Predicting measures of motor performance from multiple cortical spike trains. Science 1970; 170:758–762. 71. Evarts EV. Pyramidal tract activity associated with a conditioned hand movement in the monkey. J Neurophysiol 1966; 29:1011–1027. 72. Salisbury JK, Srinivasan MA. A phantom-based haptic interaction with virtual objects. Comput Graph Appl 1997; 17:6–10. 73. Nicolelis MA. Methods for Neural Ensemble Recordings. Boca Raton: CRC Press, 1998. 74. Moxon KA, Morizio JC, Chapin JK, Nicolelis MA, Wolf PD. Designing a brain–machine interface for neuroprosthetic control. In: Chapin JK, Moxon KA, eds. Neural Prostheses for Restoration of Sensory and Motor Function. Boca Raton: CRC Press, 2001:179–219. 75. Nicolelis MA, Ghazanfar AA, Faggin BM, Votaw S, Oliveira LM. Reconstructing the engram: simultaneous, multisite, many single neuron recordings. Neuron 1997; 18:529–537. 76. Williams JC, Rennaker RL, Kipke DR. Long-term neural recording characteristics of wire microelectrode arrays implanted in cerebral cortex. Brain Res Brain Res Protoc 1999; 4:303–313.
500
Krupa
77. Kralik JD, Dimitrov DF, Krupa DJ, Katz DB, Cohen D, Nicolelis MA. Techniques for long-term multisite neuronal ensemble recordings in behaving animals. Methods 2001; 25:121–150. 78. Nicolelis MA, Ribeiro S. Multielectrode recordings: the next steps. Curr Opin Neurobiol 2002; 12:602–606. 79. Lehew GC, Krupa DJ, Oliveira LM, Morizio JC, Wolf PD, Obeid I, Nicolelis MA. A compact high density multi-electrode array for long-term chronic recording of large ensembles of single unit neuronal activity. Soc Neurosci Abst 2002; 28:450–455. 80. Krupa DJ, Laubach M, Nicolelis MA. Somatosensory processing in different layers of rat barrel cortex. Soc Neurosci Abst 2002; 28:257–254. 81. Maynard EM, Hatsopoulos NG, Ojakangas CL, Acuna BD, Sanes JN, Normann RA, Donoghue JP. Neuronal interactions improve cortical population coding of movement direction. J Neurosci 1999; 19:8083–8093. 82. Blum NA, Carkhuff BG, Charles HK Jr, Edwards RL, Meyer RA. Multisite microprobes for neural recordings. IEEE Trans Biomed Eng 1991; 38:68–74. 83. Pochay P, Wise KD, Allard LF, Rutledge LT. A multichannel depth probe fabricated using electron-beam lithography. IEEE Trans Biomed Eng 1979; 26:199–206. 84. Prohaska OJ, Olcaytug F, Pfundner P, Dragaun H. Thin-film multiple electrode probes: possibilities and limitations. IEEE Trans Biomed Eng 1986; 33:223–229. 85. Prohaska O, Olcaytug F, Womastek K, Petsche H. A multielectrode for intracortical recordings produced by thin-film technology. Electroencephalogr Clin Neurophysiol 1977; 42:421–422. 86. Rousche PJ, Pellinen DS, Pivin DP Jr, Williams JC, Vetter RJ, Kipke DR. Flexible polyimide-based intracortical electrode arrays with bioactive capability. IEEE Trans Biomed Eng 2001; 48:361–371. 87. Kennedy PR. The cone electrode: a long-term electrode that records from neurites grown onto its recording surface. J Neurosci Methods 1989; 29: 181–193. 88. Kennedy PR, Bakay RA, Moore MM, Adams K, Goldwaithe J. Direct control of a computer from the human central nervous system. IEEE Trans Rehabil Eng 2000; 8:198–202. 89. Kennedy PR, Bakay RA. Activity of single action potentials in monkey motor cortex during long-term task learning. Brain Res 1997; 760:251–254. 90. Nicolelis MA. Opinion: brain–machine interfaces to restore motor function and probe neural circuits. Nat Rev Neurosci 2003; 4:417–422. 91. Obeid I, Morizio JC, Moxon KA, Nicolelis MA, Wolf PD. Two multichannel integrated circuits for neural recording and signal processing. IEEE Trans Biomed Eng 2003; 50:255–258. 92. Obeid I, Nicolelis MA, Wolf PD. A multichannel neural telemetry system. Soc Neurosci Abst 2002; 28:405–417. 93. Crist RE, Carmenal JM, Wessberg J, Nicolelis MA. Sensory feedback fro realtime control of a robotic device based on cortical ensemble recordings from the cortex of Rhesus macaques. Soc Neurosci Abst 2002; 28:351–357.
Development of Practical Brain–Machine Interfaces
501
94. Talwar SK, Xu S, Hawley ES, Weiss SA, Moxon KA, Chapin JK. Rat navigation guided by remote control. Nature 2002; 417:37–38. 95. Romo R, Hernandez A, Zainos A, Brody CD, Lemus L. Sensing without touching: psychophysical performance based on cortical microstimulation. Neuron 2000; 26:273–278. 96. Wickersham I, Groh JM. Neurophysiology: electrically evoking sensory experience. Curr Biol 1998; 8:R412–R414. 97. Kaas JH. Plasticity of sensory and motor maps in adult mammals. Annu Rev Neurosci 1991; 14:137–167. 98. Buonomano DV, Merzenich MM. Cortical plasticity: from synapses to maps. Annu Rev Neurosci 1998; 21:149–186. 99. Krupa DJ, Nicolelis MA. Network level properties of short-term plasticity in the somatosensory system. Prog Brain Res 2000; 128:161–172. 100. Nicolelis MA, Katz D, Krupa DJ. Potential circuit mechanisms underlying concurrent thalamic and cortical plasticity. Rev Neurosci 1998; 9:213–224. 101. Merzenich MM, Nelson RJ, Stryker MP, Cynader MS, Schoppmann A, Zook JM. Somatosensory cortical map changes following digit amputation in adult monkeys. J Comp Neurol 1984; 224:591–605. 102. Pons TP, Garraghty PE, Ommaya AK, Kaas JH, Taub E, Mishkin M. Massive cortical reorganization after sensory deafferentation in adult macaques (see comments). Science 1991; 252:1857–1860. 103. Calford MB, Tweedale R. Immediate and chronic changes in responses of somatosensory cortex in adult flying-fox after digit amputation. Nature 1988; 332:446–448. 104. Krupa DJ, Ghazanfar AA, Nicolelis MA. Immediate thalamic sensory plasticity depends on corticothalamic feedback. Proc Natl Acad Sci USA 1999; 96:8200–8205.
Index
9-hole box apparatus, 407 32 kDa dopamine- and cAMP-regulated phosphoprotein, 366
Actual amount of use test, 289 Adaptive task practice, 304 Adhesion molecules, 312 Alzheimer’s disease, 15, 53, 131, 175 Ampakines, 14 Amphetamine, 243 Amyotrophic lateral sclerosis, 344 Analog-to-digital conversion, 439 Apoptosis, 190 Apoptotic cell death, 191 Application specific integrated circuts, 447 Artificial Neural Nets, 45 Astrocytes, 341 Auditory prostheses, 478 Axonal arborization, 91, 276 Axonal regeneration, 276 Axonal sprouting, 75
B-toxin, 216 Basal synaptic transmission, 4 Basic fibroblast growth factor (bFGF), 166
Beam-walking task, 403 Best case session, 463 Bilateral vestibular damage (BVD), 47 Bioengineering, 457 Biotinylated dextran amine, 77, 240 Blood cord-derived stem cells, 350 Blood–brain barrier, 83, 243 Body weight-supported treadmill training (BWSTT), 275 Body-weight-assisted treadmill training, 286 bone marrow, 347 Botulinum toxin type A, 222 Brain aging, 53 Brain injury, 231–232 Brain plasticity, 39, 209 Brain structure and function, 159 Brain-derived neurotrophic factor, 89, 368, 409 Brain–machine interfaces, 475 auditory prostheses, 476 visual prostheses, 476 Broca’s aphasia, 257 Broca’s area, 258 Brunnstrom criteria, 310
Cache County study, 136 Cajal’s dictum, 396 503
504 Calcium–calmodulin dependent kinases, 106 Calpain, 4,194,199 activation, 4 inhibition, 199 Caspase-8, 192 Cathepsin D, 55 Cell loss, 200 Cell-replacement strategy, 364 Cell-to-cell interaction, 398 Center of gravity, 244 Central nervous system (CNS), 179, 190, 329, 395 Cerebral blood flow, 237 Cerebrovascular accident, 288 Cervical spinal cord, 339 Chimeric animals, 350 Cholinergic and glutaminergic cells, 273 Cholinergic neurone rich transplants, 401 Chromorasters, 469 Chronic hemiparesis, 283 CI movement therapy, 211 Ciliary neurotrophic factor, 368 Cingulate motor area, 316 Clarke’s nucleus neurons, 346 Closed loop, 476 Closed-loop BMI, 491 Cognitive function, 139 Cognitive impairment, 16 Communication activity log, 295 Compensatory function, 283–284 Cone electrode, 489 Conformal multielectrode arrays, 448 Conjugated equine estrogens, 137 Constraint-induced (CI) therapy, 284 Constraint-induced movement therapy, 244, 282 Contralateral striatum, 76 Contralesional arborization, 240 Cortical neurites, 489 Corticospinal tract, 272 Corticostriatal neurons, 272 Crossed-corticostriatal pathway, 77
Index Cyclin-dependent kinase 5 (cdk5), 182
Delayed nonmatch-to-sample (DNMS) tests, 458 Dendritic branching, 239–240 Dentate gyrus, 337 Dependent BMIs, 483 Diaschesis, 286 Direct output BMI, 485 DNMS behavioral paradigm, 458 Dopaminergic activation, 405 Dorsolateral striatum, 77 Drug therapy and rehabilitation, 41
Electrical stimulation, 213, 223 Electrode design=fabrication, 485 Electromyography, 284 Embryonal carcinoma (EC) cells, 342 Embryonic stem (ES) cells, 337 Embryonic-striatal cell transplantation, 407 EMG-triggered electrical stimulation, 303 Endocytosis, 57 Energy cost, 46 Estradiol, 176 Estrogen analogs, 179 Estrogen receptor, 140, 179 Estrogen therapy (ET), 131, 176 Estrogen-replacement therapy, 136 Estrogens, 175 Excitatory postsynaptic current frequency, 100 Excitatory postsynaptic potential, 432 Extremity constraint induced therapy evaluation, 282 Extrinsic pathway, 191
Fetal transplantation, 163 Film autoradiography, 170 Filopodia, 28 Finger tapping protocol, 237 Finite–state machine, 439
Index Fischer-344 rats, 230 FMRI scans, 263 Focal ischemic brain injury, 283 Focal spasticity, 209 Foot-slip tests, 403 Fugl-Meyer assessment, 290 Function electrical stimulation, 424 Functional magnetic resonance imaging (fMRI), 172 Functional recovery, 115
GABA modulation, 293 Gait symmetry, 222 Gait velocity, 222 g-aminobutyric acid, 236 Gene mutation, 160 Glial precursor cells, 341 Glial proteins, 312 Glutamate, 10 Gonadotrophin releasing hormone, 410 Graft-derived recovery, 402 Graft–host circuitry, 407 Graft–host connections, 402 Growth-associated protein-43, 240
Head posturograms, 47 Hebb-Williams maze task, 402 Hematopoietic stem cells, 347 Hemiparetic patients, 309 Hemiplegic stroke, 271 Hippocampal circuit dynamics, 430 Hippocampal neurons, 457 Hippocampal slice, 451 Hippocampal-prosthetic devices, 458 Hippocampus, 90 Hit rate, 462 Hormone replacement therapy, 133 Hormone therapy (HT), 131, 176 Human machine interface (HMI), 47 Human neural prosthesis, 492 Huntingtin, 363 Mutant huntingtin, 363 Normal huntingtin, 363 Huntington’s disease, 361, 397 Hybrid neuronal networks, 44
505 Indium tin oxide, 449 Infinite impulse response, 441 Information content, 462 Input=output properties, 432 Intra- and extracellular microelectrode, 43 Intracerebral transplantation, 396 Intracortical connections, 79 Intracortical electrode arrays, 489 Intracortical microstimulation, 233 Intraneuronal trafficking, 63 Intrastriatal dopaminergic grafts, 403 Interimpulse interval, 428 Intrinsic circuits, 425 Ipsilateral forelimb binding, 242 Ischemia-inducing thermocoagulation, 76 Ischemic infarcts, 234 Ischemic lesions, 79
Kainic acid, 338 Kinematics and electromyographic activity, 273
Laguerre parameter, 443 Laguerre polynomials, 444 Lateral ganglionic eminence, 366 Learned nonuse, 287, 308 Learning and memory, 101 Learning to use a graft, 406 Learning, 3–4 Left hemisphere lesions, 255 Lesion size, 261 Lesion-specific sprouting, 79 Letter fluency, 261 Linear discriminant methodology, 460 Linear statistical pattern recognition technique, 460 Long–Evans rats, 458 Long-term depression, 399 Long-term potentiation, 2, 97, 399 Lysosomal dysfunction, 53 Lysosomes, 61
506 Magnetic resonance imaging, 272–273, 376 Magnetic resonance-morphometric measures, 309 Massive glutamatergic projection, 76 Medroxyprogesterone acetate, 137 Meganeurites, 56 Memory, 1 Mesenchymal (or stromal) stem cells, 348 Micro PET system, 164 Microelectrode arrays, 459 Microglia, 61, 195 Microphotodiode, 481 Microwire electrode arrays, 487 Middle cerebral artery occlusion, 229, 402 Mild cognitive impairment, 15 Miniaturization, 490 Monitor gene therapy, 169 Morphological plasticity, 25 Morris water maze, 101 Motor activity log, 288 Motor learning, 287, 315 MR spectroscopy, 172 Multisite electrode arrays, 429 Myelin basic protein, 240 Myelin-deficient rat, 340
Necrosis, 190 Neocortex, 90 Nerve growth factor, 57, 90, 368 Neural cell adhesion molecules, 398 Neural cell lines, 345 Neural functions, 457 Neural plasticity and repair, 159 Neural prosthetic, 429 Neural recruitment, 261, 262 Neural stem cell implantation, 166 Neural stem cells, 330 Neural Transplantation, 365 Neuroadapted Sindbis virus, 342 Neurobehavioral engineering, 457 Neurodegenerative diseases, 163 Neurofibrillary tangles (NFTs), 181 Neurogenesis, 378
Index Neurohumoral transmission, 43 Neuroimaging, 256 Neurological deficits, 230 Neurological disorders, 229, 493 Neurological impairments and disabilities, 271 Neuronal circuitry, 94 Neuronal degeneration, 61 Neuronal plasticity, 75, 107 Neuronal-restricted precursors, 338 Neuroplasticity and neuropathology, 160 Neuroprotection, 175 Neurotransmitter receptors, 312 Neurotrophin, 57, 89 Niemann-Pick type C disease, 60 Nigrostriatal dopamine system, 169 NMDA receptor activation, 9 NMDA-receptor antagonist, 236 Nonhuman primate models, 231 Nonsynaptic diffusion neurotransmission (NDN), 39 Nonsynaptic interneuronal communication, 41 Normalized mean square error, 435 Ntera-2 cells, 343
Oligodendrocyte-type-2-astrocyte (O-2A) precursor, 340 Oligodendrocytes, 85, 195 Operant conditioning, 287 Operant learning, 288 Optimal function, 284 Output BMIs, 482
P-zones, 366 p75 neurotrophin receptor, 192 Paraptosis, 191 Parkinson’s disease, 397 Peri-infarct cortex, 240 Peripheral myelin, 349 Perspex window, 406 Phosphenes, 480 Physiological and cellular processes, 161 Placebo, 138
Index Plasma-enhanced chemical vapor deposition, 449 Polymorphic delta activity, 83 Positron emission tomography (PET), 160, 373 Post-traumatic brain injury, 284 PremPro, 138 Primary embryonic neurons, 397 Primary progressive aphasia, 266 Procedural learning, 271 Progestin, 132 Proof-of-concept, 429 Protein kinase C, 107 Proteolytic cleavage, 363
Quantal analysis of synaptic transmission, 10
Randomized-clinical trials, 287 Reactive oxygen species, 192 Real-time population spike extraction, 439 Recording methodology, 458 Redundancy recovery, 259 Regional anesthesia, 245 Rehabilitation, 239 Repetitive transcortical magnetic stimulation, 259 Retinal ganglion cell, 95 Retinal stimulation, 481 Retinitis pigmentosa, 481 Reward–punishment elements, 293 Rhythmic auditory stimulation, 222
Scala tympani, 479 Schizophrenia, 15 Schwann cells, 348 Second order kernels, 435 Selective estrogen receptor modulators, 151 Semantic fluency, 261 Sensorimotor cortex, 77, 231 Short-term potentiation, 10 Signal-processing techniques, 485
507 Sit-to-stand and lie-to-sit transfers, 293 Skilful learning, 210 Slow cortical potentials, 483 Somatosensory cortex, 30, 403 Somatosensory deafferentation, 287 Somatosensory function, 311 Spasticity, 216 Spinal cord injury (SCI), 189, 271, 329 Sprague–Dawley rats, 343 Squirrel monkeys, 233 Static linear model, 463 Stem cells, 365 Stem-cell therapy, 363 Stimulus–response associative learning, 407 Striatal grafts, 367 Striatal-like cells, 409 Stroke and ischemia, 230 Stroke, 209 Stroke-hemiparesis, 286 Structural-functional correlations, 402 Subventricular zone, 412 Supplementary motor area (SMA), 272 Synaptic activity, 26, 34 Synaptogenesis, 28, 242 Synaptophysin reactivity, 241
T-value testing, 462 Tactile vestibular substitution system (TVSS), 47 Tactile=proprioceptive feedback, 492 Tau protein, 240 Thalamic and striatal metabolism, 166 Therapeutic treatment, 344 Therapeutic window, 178 Thermocoagulatory lesions, 76 Three-dimensional anisotrophy contrast, 310 Training and classification testing procedure, 462 Transcortical sensory aphasia, 258 Transcutaneous interface, 483 Transdifferentiation, 347 Traumatic brain injury, 115 Treadmill therapy, 220
508 Trk receptors, 90 Tumor necrosis factor, 191 Tyrosine hydroxylase, 369
Ubiquitin, 193 Uninjured brains, 238 Upper limb rehabilitation, 224
Varicosity turnover, 31 Ventral premotor cortex, 236–237 Very large scale integrated system, 428 Visual evoked potentials, 483
Index Visual feedback, 491 Visual prostheses, 479 Volterra–Poisson modeling approach, 432 Volterra–Poisson kernel, 434 Volume transmission, 39
Wallerian degeneration, 310 Wiener–Poisson kernels, 434 Wernicke’s area, 260 Women’s Health Initiative Memory Study, 131 Worst case session, 465
About the Editors MICHEL BAUDRY is Professor, Department of Biological Sciences, Neurology and Biomedical Engineering, University of Southern California, Los Angeles. He received the B.E. degree (1971) from Ecole Polytechnique, Paris, France, and the Ph.D. degree (1977) from the University of Paris VII, France. XIAONING BI is Associate Professor, Department of Anatomy and Neurobiology, University of California, Irvine. She received the Ph.D. degree from the University of Southern California, Los Angeles. STEVEN S. SCHREIBER is Professor of Neurology and Anatomy and Neurobiology, College of Medicine, University of California, Irvine, and Chief of Neurology, VA Long Beach Healthcare System, Long Beach, California. He received the M.D. degree (1979) from Albany Medical College, New York, and completed neurology residency and postdoctoral training in molecular biology and neuroscience at the University of Southern California, Los Angeles.