INTERNATIONAL REVIEW OF
Neurobiology VOLUME 18
Associate Editors W. R. ADEY
H. J. EYSENCK
D. BOVET
C. HEBB
Josh DELGADO
S. KETY
SIR JOHN ECCLES
A. LAJTHA
0. ZANGWILL
Consultant Editors R. BALDESSARINI
P.
F. BLOOM
K. KILLAM
P. BRADLEY
C. KORNETSKY
R. J. BRADLEY
B. A. LEBEDEV
J. ELKES
P. MANDEL
K. FUXE
H . OSMOND
R. HEATH
S. H. SNYDER
B. HOLMSTEDT
s. SZARA
JANSSEN
INTERNATIONAL REVIEW OF
Neurobiology Edited by CARL C. PFEIFFER Brain Bio Center 1225 State Road Princeton, New Jersey
J O H N R. SMYTHIES Department of Psychiatry and the Neurosciences Program University of Alabama Medical Center Birmingham, Alabama
V O L U M E 18
1975
ACADEMIC PRESS
New York
San Francisco
London
A Subsidiary of Harcourt Brace Jovanovich, Publishers
COPYRIGHT 0 1975, BY ACADEMIC PRESS, INC. ALL RIGHTS RESERVED. NO PART O F THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.
ACADEMIC PRESS, INC.
111 Fifth Avenue, New York, New York 10003
United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road. London N W l
LIBRARY OF CONGRESS CATALOG CARDNUMBER:59-13822
ISBN 0-12-366818-2 PRINTED IN THE UNITED STATES OF AMERICA
CONTENTS CONTRIBUTORS.
.
.
.
.
.
.
.
.
.
.
.
.
ix
Integrative Properties and Design Principles of Axons STEPHENG. WAXMAN
I . Introduction . . . . . . . . I1. The Axon as a Simple Transmission Line . . I11. The Axon as a Delay Line . . . . . I V. The Axon as a Filtering System . . . . V . External Effects on Axons . . . . . VI . Electrotonic Coupling by Axonal Pathways . VII . Structure-Function Relations for Central Axons VIII . Functions of Axons in the Normal Nervous System
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
1 2 4
6 18 19 20 32
I X. Demyelination : Pathophysiological Aspects of Delayed Conduction and Intermittence . . . . . . . . . . . . X . Conclusions and Summary . . . . . . . . . . References . . . . . . . . . . . . .
33 34 36
Biological Transmethylation Involving S-Adenosylmethionine: Development of Assay Methods and Implications for Neuropsychiatry R o s s J . BALDESSARINI
I . Introduction . . . . . . . . . . . . . I1. Biochemical Assays for the Study of Transmethylation: Assays of the
41
. . . . . . . . . . . . Methyl Donor 111. Other Assays Related to Transmethylation . . . . . . IV . Clinical Implications : Need for New Strategies for Clinical Metabolic . . . . . . . . . Research in Schizophrenia . . . . . . . . . . . . . . References
44
57 61 63
Synaptochemistry of Acetylcholine Metabolism in a Cholinergic Neuron
BERTALANCSILLIK I. I1. I11. IV.
Introduction . . . . . . . . . . . Histochemistry of Acetylcholinesterase in the Spinal Motoneuron Indirect Information on Cholinergic Mechanisms . . . Molecular Anatomy of Transmitter Release . . . . References . . . . . . . . . . .
. . . . .
. . . . .
112 119 133
. . .
141 142 151
69
77
Ion and Energy Metabolism of the Brain at the Cellular Level LEIF HERTZA N D ARNESCHOUSBOE
.
I Introduction . . I1. Complexity of Brain I11. Energy Metabolism
. . .
. . .
. . .
. . . V
. . .
. . .
. . .
. . .
. . .
. . .
vi
CONTENTS
I V. Ion and Water Metabolism V . Concluding Remarks . References . . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
176 191 193
.
.
. .
213 220 232 237 253 255
Aggression and Central Neurotransmitters
S. N . PRADHAN
I. Introduction I1. I11. I V. V.
.
.
.
.
.
.
.
.
.
.
Neuroanatomical and Neurochemical Correlation of Aggression . Chemostimulation of Discrete Brain Areas and Induced Aggression Neuropharmacological Manipulation of Aggression . . . . Summary and Conclusion . . . . . . . . . References . . . . . . . . . . . .
.
. . .
A Neural Model of Attention. Reinforcement and Discrimination Learning
STEPHENGROSSBERG
I . Introduction . . . . . . . . . . . . I1. Drives. Rewards. Motivation. and Habits . . . . . . I11. The Rebound from Fear to Relief . . . . . . . I V . Short-Term Memory and Total Activity Normalization . . . V . Sensory-Drive Heterarchy . . . . . . . . . V I . Conditionable Ct+ S Feedback and Psychological Set . . . V I I . T h e Persistence of Learned Meanings . . . . . . V I I I . Overshadowing and the Triggering of Arousal by Unexpected Events IX . Pavlovian Fear Extinction vs Persistent Learned Avoidance . . X . Frustration . . . . . . . . . . . . X I . Partial Reinforcement Acquisition Effect . . . . . . XI1. Generalization Gradients in Discrimination Learning . . . XI11. Habituation and the Hippocampus . . . . . . . X I V . Overshadowing vs Enhancement XV . Novelty and Reinforcement . . . . . . . . . XVI . Motivation and Generalization . . . . . . . . XVII . Predictability and Ulcers . . . . . . . . . X V I I I . Orienting Reaction . . . . . . . . . . X I X . A Learned Expectation Mechanism X X . Regulation of Orienting Arousal . . . . . . . . XXI . Hippocampal Feedback, Conditioning. and Dendritic Spines . . XXII . Nervous Eating and Attentional Deficits Modulated by Arousal . Appendix . . . . . . . . . . . . . References . . . . . . . . . . . .
.
.
.
.
. . . . .
264 274 276 282 288 290 29 1 294 297 297 300 301 305 306 308 309 310 311 313 316 319 321 323 325
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
329 330 345 351 353
.
.
. .
.
.
.
.
.
.
.
.
.
.
. .
. . . . .
.
. .
.
.
.
.
.
.
Marihuana. learning. and Memory
ERNESTL . ABEL I . Introduction
.
I1. Animal Studies I11. Human Studies
. . .
I V. Summary and Further References . .
. . .
. . .
. . .
. . .
. . . Considerations . . . . . . .
vii
CONTENTS
Neurochemical and Neuropharmacological Aspects of Depression
B. E . I. EONARD Introduction . . . . . . . . . . . . . Characteristics of the Affective Disorders . . . . . . . The Biogenic Amine Hypothesis of Affective Disorders . . . . Cyclic AMP and Possible Connection with Affective Disorders . . . Some Biochemical Effects of Drugs Used in the Treatment of Affective Disorders . . . . . . . . . . . . . V I . Conclusion . . . . . . . . . . . . . References . . . . . . . . . . . . .
I. I1. I11. I V. V.
SUBJECTINDEX .
.
.
.
.
.
.
.
.
.
.
.
.
CONTENTS O F PREVIOUS VOLUMES.
.
.
.
.
.
.
.
.
357 359 360 367 368 380 381 389
.
393
This Page Intentionally Left Blank
CONTR IBUTORS Numbers in parentheses indicate the pages on which the authors’ contributions begin.
ERNESTL. ABEL, Research Institute on Alcoholism, Buffalo, N e w York (329) Ross J. BALDESSARINI, Psychiatric Research Laboratories, General Hospital, and Department of Psychiatry, Harvard Medical School, Boston, Massachusetts (41)
BERTALAN CSILLIK,Department of Anatomy, University Medical School, Szeged, Hungary (69) STEPHENGROSSBERG, Department of Mathematics, Massachusetts Institute of Technology, Cambridge, Massachusetts (263) LEIF HERTZ,Department katoon, Canada (141)
of
Anatomy, University of Saskatchewan, Sas-
B. E. LEONARD,* Pharmacology Department, Organon International B. V., Oss, T h e Netherlands (357) S. N. PRADHAN, Department of Pharmacology, Howard University College of Medicine, Washington, D.C. (213) ARNE SCHOUSBOE, Department of Biochemistry A, University hagen, Copenhagen, Denmark (141)
of
Copen-
STEPHENG. WAXMAN,?Harvard Neurological Unit, Boston City Hospital, Boston, Massachusetts, and Department of Neurology, Harvard Medical School, Boston, Massachusetts (1 )
* Present address : Department of Pharmacology, University College, Galway, Republic of Ireland. t Present address: Harvard Medical School Neurological Unit, Beth Israel Hospital, Boston, Massachusetts 022 15 and Department of Biology and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02 139. ix
This Page Intentionally Left Blank
INTERNATIONAL REVIEW OF
Neurobiology VOLUME 18
INTEGRATIVE PROPERTIES AND DESIGN PRINCIPLES OF AXONS By Stephen G. W a x m a n '
Harvard Neurological Unit, Boston City Hospital, Boston, Massachusetts, a n d Department of Neurology, Harvard Medical School, Boston, Massachusetts
.
I. Introduction 11. The Axon as a Simple Transmission Line . 111. The Axon as a Delay Line . IV. The Axon as a Filtering System . A. Space-Time Transformations in the Central Nervous System B. Intermittent Conduction in Vertebrates . C. Intermittent Conduction in Invertebrates D. Differentiation of Nodal Morphology and Functional Implications . V. External Effects on Axons . VI. Electrotonic Coupling by Axonal Pathways . VII. Structure-Function Relations for Central Axons . A. Nodes and Internode Spacing B. Diameter Spectra . C. Critical Diameter for Myelination . VIII. Functions of Axons in the Normal Nervous System. IX. Demyelination: Pathophysiological Aspects of Delayed Conduction and Intermittence X. Conclusions and Summary References
.
.
.
. .
.
. . . . . .
. .
. . . . . . .
.
. .
1 2 4 6 6 7 10 13 18 19 20 20 26 29 32 33 34 36
I. Introduction
Neurophysiology has classically treated the axon as a simple transmission line which functions so as to conduct neural messages from one site to another with a minimum of delay and without alteration in content or form. This concept has held a central place in the development of ideas concerning Present address: Harvard Medical School Neurological Unit, Beth Israel Hospital, Boston, Massachusetts 02215 and Department of Biology and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02 139. 1
2
STEPHEN G. WAXMAN
the neuron. I t is clear that, at numerous sites, especially in the peripheral nervous system, maximization of conduction velocity and reliability of transmission have been primary criteria in axonal design. However, a number of lines of reasoning indicate that maximization of conduction velocity and safety factor may not be the primary criteria of design for all axons, and there are data indicating that neural information may, in fact, undergo significant transformations within the axonal component of the neuron. I t is the purpose of this paper to challenge the generality of the transmission line hypothesis, and to review the evidence supporting the alternative notion, that axons are not necessarily designed so as to conduct impulses as rapidly and reliably as possible, but that they may rather function so as to distribute and filter neural information in the spatial and temporal domains. This alternative model, which regards axons as integrative structures, implies that not only the dendrites, perikarya, and associated synapses, but also the axon and its branches, may play a role in determining the logical infrastructurc of the neuron.
II. The Axon as a Simple Transmission Line
The classical concept of axonal function, which represents the axon as a simple transmission line, derives in large part from dimensional arguments. Following studies on the invariance of form of the action potential (see, e.g., Hodgkin, 1964) and demonstrations of saltatory conduction in myelinated axons (Tasaki and Takeuchi, 1941; Huxley and Stampfli, 1949), Rushton (1951) demonstrated that if fibers exhibited the same specific menibrane properties and exhibited “dimensional similarity,” conduction velocity should be proportional to diameter for myelinated fibers, whereas conduction velocity should be proportional to diameterw for nonmyelinated fibers. Dimensional similarity required that
and that
d 2 / L 2 a l/log, ( D / d ) where d = axon diameter = internal diameter of myelin, D = fiber diameter = external diameter of myelin, a = area of nodal membrane, and L = internode length. Rushton presented evidence that the conditions of dimensional similarity did apply to peripheral axons, and argued that nerves tended to conform to the theoretical conditions because these were optimal
DESIGN PRINCIPLES OF AXONS
3
in terms of maximizing conduction velocity and safety factor. Rushton cited histological evidence (Gasser and Grundfest, 1939; Sanders, 1948) that the ratio d l D for peripheral axons was close to the value for which conduction velocity would be maximal. The dimensional arguments predict that internodal conduction time should be the same for all fibers. [While this may be true for some groups of myelinated fibers (see, e.g., Tasaki, 1959; Rasminsky and Sears, 1972), precise measurements are in general not available for central fibers, and there is evidence that the internodal conduction time for small peripheral myelinated fibers is greater than for large fibers (Coppin and Jack, 1972), as would be predicted from the increased duration of the rising and falling phases of the action potential in small diameter fibers (Paintal, 1966; see also Waxman and Bennett, 1972).] Pickard ( 1969), using the assumptions ( i ) that corresponding points along axons will pass through the same state at corresponding times, and (ii) that the rise of the action potential is initially exponential, has argued that the morphology of myelinated fibers is such as to maximize conduction velocity, provide high reliability, and minimize energy consumption during impulse propagation. One set of conditions sufficient to ensure a proportionality between conduction velocity and fiber diameter includes the structural constraints L a D, d oc D, and awlad = 0, where w = the width of the unmyeelinated gap a t the node. Pickard cited evidence (Cragg and Thomas, 1964; Friede and Samorajski, 1967; Dodge and Frankenhauser, 1959) that the structural constraints do apply in some cases. Goldman and Albus (1968) represented the myelinated axon as being composed of lengths of passive, leaky cable with periodic interruptions by short lengths of excitable membrane. Their dimensional analysis showed that the conditions of proportionality between internode length and fiber diameter, and of the constancy of the ratio between axon diameter and fiber diameter, could not be relaxed individually without compromising the linear relationship between conduction velocity and fiber diameter. Dun (1970), using a transmission line model based on the assumptions of proportionality of myelin thickness to fiber diameter, proportionality of internode length to diameter, constant internodal conduction time, and proportionality of conduction velocity to fiber diameter, has computed the length and diameter of the node of Ranvier as functions of fiber diameter, and has presented some evidence which suggests that the predicted relations may apply to fibers in peripheral nerve. The data are consistent with a transmission line model for some axons and suggest that maximization of conduction velocity may be, for some fibers, a primary criterion of design. It is interesting, in this regard, that there may be a correspondence between presynaptic and postsynaptic fiber sizes in afferent systems (Bishop, 1966), providing a form of “velocity matching” for the pre- and postjunctional axons.
4
STEPHEN G . WAXMAN
111. The Axon as a Delay line
That the axon could provide functionally significant temporal delays in conduction was shown as early as 1938 in the studies of Pumphrey and Young on the innervation of the mantle of the cephalopods Sepia and Loligo. Conduction velocities of the giant fibers innervating the circular mantle muscles increased with the 0.614 power of diameter. In Sepia, the conduction distances to the various muscles are nearly equal, and the diameters of the nerves are similar. I n Loligo, on the other hand, the muscles are located at various distances from the command nucleus. In this system, where synchrony of contraction of muscle at different distances from the stellate ganglion relay is of importance in the generation of maximal propulsive force, the axons exhibit a spectrum of sizes, and muscle closer to the stellate ganglion is innervated by thinner axons of slower conduction velocity. A similar organizational principle applies to the teleost electromotor systems, in which electroplaques at different distances from command or relay nuclei must discharge synchronously so as to generate the electric organ signal. Bennett ( 1968) has demonstrated three compensatory mechanisms in electromotor systems: equalization of path length, compensatory differences in conduction velocity, and localized compensatory delays, determined by variations in conduction properties of preterminal axon branches. In the eel Electrophorus, compensation for differences in conduction time along the spinal cord involves both increased delay at spinal relays and increased delay from activity in the ventral roots to impulse initiation in electrocytes ( Albe-Fessard and Martins-Ferreira, 1953; Bennett, 197l a ) . Morphological studies indicate a reduction in the ratio of internode distance to fiber diameter along some preterminal fibers (Waxman, 1971; see also Section VII, A ) . Light and electron microscopic studies on bulbospinal and electromotor axons indicate that differences in fiber diameter, and in the ratios of myelin thickness and internode length to fiber diameter, could account for compensatory delays in conduction to rostra1 and caudal electrocytes (Meszler and Bennett, 1972; Meszler et al., 1974). The data suggest that synaptic delays do not contribute significantly to compensatory delays, which are determined by axonal conduction. In the cerebellum also, there is evidence that axons function as delay lines. Braitenberg (1967) suggested that the spacing of Purkinje cells along beams of parallel fibers might mediate their activation in a definite sequence, allowing the cerebellar cortex to act as a clock. Freeman ( 1969) and Freeman and Nicholson (1970) computed first-order serial correlations for pairs of frog Purkinje cells separated by known distances along the same beam of parallel fibers. The physiological data indicate that, in response to synchronous afferent volleys, Purkinje cells lying along the same beam of parallel fibers fire in a precise sequence, with the delay be-
DESIGN PRINCIPLES OF AXONS
5
tween firing proportional to the distance between cells. This finding suggests that the parallel fiber system functions as a tapped delay line, exciting Purkinje cells in a precise temporal sequence (Freeman, 1969; Freeman and Nicholson, 1970). Anatomical data suggest that delay line operation is not limited to specialized motor systems and the cerebellum. Lorente de No ( 1953), in studies of the presynaptic arborizations in the oculomotor nucleus and ventral nucleus of cochlear nerve, concluded that, because of differences in length of the thin presynaptic branches, invasion of the various endings must take place a t different times. Scheibel and Scheibel, on the basis of light microscopic studies of axonal branching patterns in the brain stem reticular core (1958) and in thalamic systems (1970), have suggested the possibility of multiplexing and parallel processing in the axonal arborizations, which appear structurally to determine a divergence of information with a spectrum of latencies. Evidence for reduction in internode distances along central axons is summarized in Section VII, A. Theoretical considerations indicate that for any given fiber diameter, there is an optimal internode distance for maximal conduction velocity (Hardy, 1971 ; Huxley and Stampfli, 1949), and it has been suggested that this optimum is close to the internode distance exhibited in normal peripheral nerve. Ito and Takahashi (1960) demonstrated a delay of impulse conduction through spinal ganglia which they explained as arising on account of the structure of the afferent axons, along which the internode distance :diameter ratio is smaller than in peripheral nerve. T h e data of Hardy (1971) suggest an internode distance of 1.0-1.5 mm and a nodal surface area of 22 pm2 as the dimensions which maximize conduction velocity for a fiber with 14 pm outer diameter; both of these are close to the values observed in peripheral nerve. Significant reduction in internode length or increase in nodal surface area should decrease the conduction velocity. There is evidence for both mechanisms in the teleost central nervous system. Nodes of Ranvier are very closely spaced in teleost oculomotor and electromotor nuclei, where internode distances can be less than 10 pm. At some of the fibers in these nuclei (and in other regions in teleosts and other species; see below) nodal surface area is markedly increased (Waxman, 1971 ) . T h e presence of en passant synapses arising at nodes of Ranvier (Bodian and Taylor, 1963 ; Khattab, 1966; Waxman, 1972) also suggest delay-line operation with a high degree of temporal resolution. Since the synapses arise directly at the node, they represent collaterals of negligible path length. Conduction time per internode is of the order of 20-30 psec in peripheral nerve (Rasminsky and Sears, 1972), so that an interval of at least 20 psec must occur between firing of the en passant synapse and a synapse at the next node for fibers with internodal conduction times in this range. Rasminsky
6
STEPHEN G. WAXMAN
and Sears ( 1972) have demonstrated that internodal conduction time may be increased to more than 600 psec in demyelinated fibers. I n view of the morphological similarities between some central axons and demyelinated fibers (see Section VII, A ) , it may be expected that the internodal conduction time along these fibers may exceed those of normal peripheral fibers.
IV. The Axon as a Filtering System
A. SPACE-TIME TRANSFORMATIONS I N T H E CENTRAL NERVOUS SYSTEM The idea that neural information can be coded in both space and time has a long history and arises at least in part from cybernetic and behavioral issues. McCulloch and Pitts ( 1943) demonstrated the formal equivalence of spatial and temporal processes in their logical calculus, which was based on the properties of formal neurons. Lashley (1951), in his monograph on the problem of serial order in behavior, explored the biological bases for temporal patterning in motor activity. He paid particular attention to the interaction of temporal and spatial systems, stating that in the nervous system “temporal sequence is readily translated into a spatial concept” and that conversely “translation from the spatial distribution . . . to temporal sequence seems to be a fundamental aspect of the problem of serial order.” Uttley ( 1954), in his comments on classification of signals in the nervous system, directed further attention to neural space-time transformations. He suggested three stages in the transformation of variable signals into ones suitable for neural classification : ( i ) analog-digital conversion, ( i i ) differentiation, and (iii) multiple delays. With regard to the last process, Uttley specifically noted the property of multiple delay lines of distinguishing the degree of temporal separation of input signals and for transforming temporal into spatial sequences. I n discussing relative timing between impulses in the nervous system, MacKay (1954) noted in particular that changes in fiber diameter could significantly effect conduction delays and thus temporal relationships, and suggested that changes in delay in transmission could modify the behavior of neural networks. Efron (1963a-d), in a series of papers based on clinical observations, has presented data dealing with the perception of simultaneity and temporal order in man. The data strongly suggest that temporal discrimination of simultaneity and order occurs in the dominant hemisphere. Differences in conduction distance are not corrected for, but the errors, which are of the order of 10-20 msec, for simultaneous stimuli to finger and toe, and which correspond to the difference in latency for cortical potentials evoked by finger and toe stimulation, can be shown only statistically, since judgment
DESIGN P R I N C I P L E S OF AXONS
7
of simultaneity, when there are no differences in peripheral conduction dis-
tance, has an error of 10-20 msec (Efron, 1963a). O n the other hand, at the unit level, intervals of less than 1 msec may be discriminated (see, e.g., Yasargil and Diamond, 1968). If it is assumed that information is coded by temporal patterns of impulses determined at the initial segment, variations in conduction times may be interpreted as introducing noise and limiting the information capacity of neural channels (the alternative, that the variance may itself represent information, is discussed below). The effects of noise on nerve channel capacity were studied by Harris and Stark (1971), who analyzed dispersion curves of conduction times in a crayfish photoreceptor ncrve channel. Although the means were equal for short, medium, and long impulse intervals, the standard deviation was greater for short intervals. Channel capacity for a 1 cm length of nerve, calculated by maximizing the computed mutual information rate over all biologically possible input interval distributions, was determined to be 360 bits per second. Noise present in synaptic transmission limited information capacity to approximately the same degree as noise in axonal transmission (Harris and Stark, 1973).
B. INTERMITTENT CONDUCTION IN VERTEBRATES Barron and Matthews ( 1935) initially demonstrated intermittent conduction in the cat and frog spinal cord. They presented evidence that recurrent branches leave dorsal column fibers via dorsal roots. Antidromic activity was recorded in these branches after stimulation of the appropriate afferent dorsal root or the dorsal columns. However, the antidromic discharge differed from that in the fiber entering the cord in that at fairly regular intervals it stopped abruptly without any change of frequency, so that conduction of impulses along the recurrent branch was intermittent (Fig. 1) . A similar degree of intermittence was observed when stimulating and recording electrodes were reversed, indicating that the intermittence was not unidirectional. Barron and Matthews postulated that intermittent conduction
FIG. 1. Intermittent conduction, as first demonstrated by Barron and Matthews ( 1935). Continuous series of impulses are transformed into intermittent series. Blockage occurs at regions of low safety factor.
8
STEPHEN G. WAXMAN
block was occurring at branch points, perhaps due to electrotonic influences of the collaterals, and suggested that this could provide a “mechanism of nervous integration . . . which does not involve a synapse.” The electrophysiological observations suggested several factors which might modify intermittence. The duration of blocked periods was reduced by contralateral dorsal root section or cord transection, and electrical stimulation of nerves or cutaneous stimulation also modified the degree of block (Barron and Matthews, 1935) . Interestingly, the degree of intermittence was temperature-related, increases in temperature corresponding to an increased degree of blockage. Recent studies of partially demyelinated fibers (Davis and Jacobson, 1971; Rasminsky, 1973) have shown an increased susceptibility to conduction failure at increased temperature. This probably reflects the decrease in the time integral of inward nodal current which occurs at high temperatures, and which becomes significant in terms of conduction failure at sites of low safety factor (Rasminsky, 1973). Similar mechanisms may account for the temperature-sensitivity reported by Barron and Matthews. Wall et al. (1956) made a quantitative study on impulse transmission from sciatic nerve to dorsal root and to dorsal column in cats. The data indicated that safety factor for transmission at high frequency was higher in sciatic nerve or dorsal root than in the sciatic nerve-dorsal column channel or dorsal root-dorsal column channel. I n the opposite direction, from dorsal column to sciatic nerve, frequency limitation was the same as in a peripheral axon. Similarly, it was shown that bursts of activity of short duration effected subsequent passage of impulses, a partial block for the second volley following the first for as long as 30-40 msec; this effect was present only for orthodromically conducted impulses. These findings suggested impulse blockage along continuous axons at regions where side branches emerge, the blockage occurring only for orthodromic impulses. Increased frequency of branching and presence of unmyelinated segments of greater length and higher frequency close to the point of entry into the spinal cord was suggested as one explanation for the condition failure. Alternatively, it was suggested that activity in side branches of the axons or neighboring axons could effect impulse conduction. Presynaptic failure of impulse propagation has been described in the rat phrenic nerve-diaphragm preparation (Krnjevic and Miledi, 1959) . In this case, intermittent conduction failure in presynaptic fibers occurred at frequencies of less than 50 per second within 2-5 minutes, both in uitro and in situ. Repetitive discharges occurred in three types of sequences: clear alternation of impulses and failures, a cyclic pattern of alternation of groups of impulses and groups of failures, and irregular sequences of impulses and failures with no obvious pattern. Krnjevic and Miledi (1959) presented
DESIGN PRINCIPLES OF AXONS
9
evidence that muscle fibers belonging to the same motor unit could fail at different frequencies, and suggested that the conduction block occurred at branch points, where safety factor is reduced. Characteristics of impulse trains in dimming fibers from the frog retina and in “ectodromic” fibers carrying impulses outward in dorsal roots have been studied by Chung et al. (1970). The findings suggest that variations in interspike interval, as well as mean impulse rate, may represent information about stimulus parameters, so that single units may code information about several stimulus parameters. Interspike interval records for type I dimming fibers contain high- and low-frequency bands. With changes in background illumination, the temporal patterning of discharges, as reflected in the interspike interval records, is modified while average discharge frequency is essentially unchanged. Similar changes in temporal pattern occurs for type I1 fibers. Ratliff et al. (1968) presented evidence that variations in interspike interval reflect fluctuations in membrane potential in eccentric cells in Limulus, so that impulse trains could code information about both light intensity and state of adaptation, one parameter being coded by variation in interspike intervals. O n the basis of a second set of experiments on ectodromic impulses in cat dorsal roots, Chung et al. (1970) suggested that temporal patterns may be resolved into spatial patterns within the axonal tree. The evidence arises from the demonstration of regions of low safety factor along the intraspinal cord part of the axonal pathway from the afferent dorsal root fibers to ectodromic dorsal root axons in cats maintained at 38-40OC (Raymond and Lettvin, 1969). T h e data suggest a regularity in alternation between conduction and block, with a strong relationship between interspike interval and safety factor. For short interspike intervals (ca. 10 msec) , blockage occurred approximately 50% of the time, with regular alternation between conduction and block. Conduction safety factor was also shown to be sensitive to discharge in nearby dorsal rootlets, with an increased blockage during the long negative wave of the dorsal root potential. I t was suggested that bifurcations and other asymmetrical aspects of axonal geometry could provide a morphological basis for the intermittence of axonal conduction, the manner in which impulses are distributed within the fiber depending on previous distributions, so that axonal arborizations might transform temporal impulse patterns in the parent axon into spatial patterns in the terminals. Blum (1972) has recorded from rapidly and slowly conducting cortical neurons which fit a number of criteria of pyramidal neurons. The smallest conduct impulses at rates as low as 10 m per second. Collision experiments suggest that for some neurons in both the slowly and rapidly conducting groups, propagation into the pyramidal tract axon is intermittent. The distribution and pattern of axon branching of pyramidal tract cells has been
10
STEPHEN G . M'AXMAN
studied by Endo et al. (1973), who analyzed antidromic invasion following stimulation of subcortical areas, Antidromic responses without preceding depolarizations and with fixed latencies at stimulation frequencies of 100-200 per second were recorded after stimulation in ten subcortical regions, including ventralis anterior, ventralis lateralis, ventralis posterior lateralis nuclei of thalamus, red nucleus, mesencephalic reticular formation, and dorsal column nuclei. I n most cases, there were large differences in latencies between antidromic spikes evoked by stimulation in subcortical structures and those evoked by stimulation of adjacent pyramidal tract, suggesting that conduction velocities along the axon collaterals were considerably slower than those along the parent axons. The distribution of collaterals suggested that pyramidal cells may fall into several functional subclasses on the basis of collateral branching patterns (Endo et al., 1973). Intracellular recordings from cat motoneurons have demonstrated that for some terminals, the probability of release of a quantum of transmitter is less than one per impulse (Kuno, 1964). Merrill and Wall (1972) have recorded the activity of spinal cord cells with perikarya in Rexed lamina IV having low-threshold cutaneous receptive fields with abrupt edges that do not move with changes in excitability of the cell. Experiments in which peripheral nerves or dorsal roots were blocked reversibly, showed that the afferent fibers which excite these cells after natural stimulation, run in a restricted part of the peripheral nerve and dorsal root. After the fibers mediating response to natural stimuli in the receptive field were blocked, electrical stimulation of other large myelinated fibers in nearby roots produced monosynaptic firing. The demonstration of two classes of afferent synapses, one effective in firing the cell after natural stimulation, and the other having no effect after natural stimulation, but firing the cells in response to electrical stimulation, suggests either that one class of synapse is ineffective in depolarizing the cells unless synchronously activated by electrical stimulation, or that one class of terminals is normally blocked, but during synchronous activity carry impulses. That activity in nerve fibers can have effects on transmission in adjacent fibers has, in fact, been shown by Katz and Schmitt ( 1940), Marrazzi and Lorente de No ( 1944), and Arvanitaki ( 1942). Physiological data do not yet permit differentiation between pre- or postsynaptic mechanisms in the case of the lamina IV cells, but it is clear that the extended synaptic region, extending from the terminal arborizations to the postsynaptic dendrites, constitutes a low safety factor region at which failure of transmission may occur.
C. INTERMITTENT CONDUCTION I N INVERTEBRATES There have been a number of demonstrations of intermittent conduction in invertebrate axons, in which branching may occur close to the cell body.
DESIGN PRINCIPLES OF AXONS
11
Tauc and Hughes (1963) studied the modes of initiation and propagation of spikes in the branching axons of neurons in the pleural and abdominal ganglia of the gastropod mollusc Aplysia. Antidromic responses in some cells could be recorded intracellularly after stimulation of more than one nerve, indicating that axonal branches were distributed in different nerves. Differences in amplitude of antidromic response after stimulation of different nerves indicated that spikes coming from one axonal branch did not necessarily invade other branches. Collision experiments yielded similar results. In some cases the conduction failure was asymmetrical in the sense that spikes from axon A would invade axon B, but would not propagate in the opposite direction. Tauc and Hughes suggested, on the basis of these findings, that some cells might exhibit a pleurality of trigger zones for spike initiation, and that spikes initiated in one branch might not be transmitted to other branches. In the crayfish opener muscle, several classes of fibers contribute differently to muscle tension at high and low impulse frequencies in the single motor axon (Bittner, 1968). This differentiation is not due to differences in electrical properties of the muscle membrane fibers, but rather reflects rate-related differences in probability of transmitter release. The evidence suggests that this is not due to differences in transmitter mobilization or in the relation between terminal depolarization and transmitter release, but rather to differences in the degree of terminal invasion. Parnas (1972) studied high frequency differential block in the branches of the single axon which innervates the deep extensor abdominal medialis (DEAM) and deep extensor abdominal lateralis (DEAL) muscles in crayfish and lobster. Bursts of impulses at frequencies of up to 50 per second are recorded during spontaneous activity in the abdominal flexor muscles, which are in many respects similar to the extensor muscles (Kahn, 1971). The DEAM response at 1-20 per second showed facilitation. At 15-35 per second, gradual reduction in amplitude (fatigue) was observed. During stimulation at frequencies of 40-50 per second, responses in DEAM were abruptly blocked after 40-80 stimuli, while there was facilitation of the DEAL response. Responses reappeared abruptly in DEAM at 100-1000 msec after lowering of stimulation frequency. High-frequency block of DEAM was not effected by high concentrations of Mg” or by reduction in extracellular Ca2+concentration. Extracellular recordings of nerve terminal potentials indicated that conduction block reflects failure of invasion of the finer axonal branches. Grossman et al. (1974) have confirmed the presence of conduction block by intracellular recording from the axonal branch innervating DEAM, which is not invaded at frequencies at which the branch to DEAL is invaded. Refractory period at the bifurcation region (3.7-4 msec) was longer than that in the nonbranching parent axon (2-3.3 msec) . The giant axons in the nerve cord of the cockroach Periplanata ameri-
12
STEPHEN G . W A X M A N
cana run continuously from the sixth abdominal ganglion to the subesophageal ganglion. Reversible blockage at points of low safety factor has been demonstrated at frequencies as low as 50 per second (Parnas et al., 1969). The low safety factor regions are “unidirectional” in that descending impulses but not ascending impulses were blocked. Bullock and Turner (1950) also reported unidirectional conduction block, in Lumbricus axons. More recently, Spira e t al. ( 1974) have demonstrated reduction in amplitude, decrease in afterhyperpolarization, appearance of prepotentials and increases in delay in spike initiation, and failure of spike invasion for high frequency descending impulses, and have suggested that this “frequency filtering” may be due to potassium accumulation outside of axons. The initial anatomical evidence (Spira et al., 1969) suggested that the cockroach axons progressively taper in the thoracic cord, with periodic “isthmuses” at which diameter is reduced. Numerical computations applying the Hodgkin-Huxley equations to spike-train transmission along nonhomogeneous axons, taking into account the effect of K+ in the periaxonal space, indicate that changes in fiber diameter could account for the band-pass characteristics of the axon (Parnas et al., 1973). Application of nicotine or carbamylcholine causes depolarization and conduction increase associated with conduction block for descending spikes, while curare prevents these effects, suggesting that cholinergic synapses are present on the fiber (Yarom et al., 1973). More recent electron microscopic studies have demonstrated branching of the giant axons, and it has been suggested that the branching may be responsible for the formation of low safety factor areas and provide a site for synaptic inputs (Spira et al., 1974). Sensory adaptation and axonal conduction block have also been demonstrated in sensory neurons of the leech Hirudo medicinalis, where there is evidence for a relation to membrane hyperpolarization (Van Essen, 1973). Simultaneous recordings from the cell body and from peripheral axons during repetitive stimulation demonstrated failure of invasion of the cell body, in some cases after only a few seconds of activity at frequencies of 20-40 Hz. I n experiments where several axon branches were recorded from, intraganglionic conduction failure was demonstrated at frequencies of 40 per second. The patterns of impulse failure exhibited a degree of specificity in that some axonal branches consistently failed to conduct while others did not. Conduction block was also demonstrated at peripheral sites. During block of impulse conduction, electrotonic potentials were recorded from the cell body or from the nerve roots. The electrotonic potentials were usually constant in size, with changes in amplitude occurring in discrete steps, suggesting that conduction block occurred at specific sites. Examination of cells marked with Procion yellow indicated that conduction block occurred more readily for impulses traveling from a small branch to a larger axon than
DESIGN PRINCIPLES OF A X O N S
13
in the opposite direction. The data suggest that in this system, conduction block is due at least in part to membrane hyperpolarization. Most of the increase in threshold following repetitive stimulation was attributed to hyperpolarization, and hyperpolarizing and depolarizing currents were shown to directly effect conduction. In addition, conduction block was in part relieved by strophanthidin (Van Essen, 1973).
D. DIFFERENTIATION OF NODALMORPHOLOGY AND FUNCTIONAL IMPLICATIONS An opportunity for the study of regional differentiation of axons is provided by the neurogenic electrocytes of certain sternarchid fish. In the Sternarchidae, the electric organs are neurogenic, i.e., derived from peripheral axons, in contrast to the electric organs of most other gymnotids, which are derived from muscle (Bennett, 1970, 1971a). T h e electrocyte axons end blindly within the electric organ. Comparative arguments suggest that an electric organ derived from muscle was originally present but was lost in the course of evolution. As would be expected from its neurogenic rather than myogenic origin, the discharge, which is of high frequency (700-1500 per second), is not affected by curare (Bennett, 1966, 1970). The fibers run from the spinal cord to the electric organ, where they run a hairpin course, initially running anteriorly for several spinal segments, then turning sharply to run posteriorly for several segments, finally tapering and ending blindly in a connective tissue filament. Light microscopic examination reveals differences in morphology at different regions along the axon (Bennett, 1971a: Waxman et al., 1972). Where the fibers enter the electric organ and where they turn around, they are about 20 pm in diameter. Anteriorly and posteriorly running parts of the axon dilate to a diameter of approximately 100 pm. Where the fibers enter the electric organ and where they turn around, nodes appear normal and extend approximately 1 pm along the fiber. In proximal parts of the anteriorly and posteriorly running segments, the nodes also appear small. In the distal part of the anteriorly running segment, the nodes are much larger, extending for 50 pm or more along the axis of the fiber (Fig. 2 ) . The changes in nodal morphology have been confirmed by light microscopy of intact fibers isolated from the electric organ and by light and electron microscopy of sectioned fibers (Waxman et al., 1972). Electron microscopy reveals that the nodes of Ranvier fall into two classes. Where the fibers enter the electric organ and where they turn around, and in proximal parts of the anteriorly and posteriorly running segments, the nodal morphology is similar to that of typical peripheral nodes of Ranvier (cf. Robertson, 1959; Elfvin, 1961) ; the nodal gap extends less
14
STEPHEN G . WAXMAN
FIG. 2. This light micrograph shows the distal part of the posteriorly running segment of a Sternarchus electrocyte axon, which has been dissected from the organ. Nodes appear as darkly stained bands in this preparation, which was stained with 0.25% toluidine blue. T h e nodes are indicated by arrows. Note the variation in nodal size, the largest nodes (e,f) being located in the most distal part of the fiber. The bar indicates 100 pm. x 100.
than 1 pm along the fiber and there is a distinct electron-dense undercoating subjacent to the axon membrane. Fingerlike extensions of the paranodal Schwann cell cytoplasm extend into the nodal gap. The large nodes in distal parts of the anteriorly and posteriorly running segments exhibit a distinct structure (Fig. 3 ) . At these nodes, myelin is absent for as far as 50 pm or more along the axon. The axonal surface is elaborated to form a layer of irregular polypoid processes, further increasing surface area (Waxman et al., 1972). The cytoplasmic dense undercoating is not present. At the enlarged nodes, but not at the small nodes, the paranodal myelin begins to terminate at distances of up to 200 pm from the nodal gap; similar features have been described in diphtheritic demyelination (Waxman, 1973; cf. Harrison et al., 1972). The cellular basis for the electric organ discharge has been studied by Bennett (1970, 1971a). The discharge is generated by the synchronous activ-
FIG. 3. Electron micrograph of a longitudinal section through a Sternarchus electrocyte axon (AX) at the site of a large node. Compact myelin begins to terminate a t the small arrows. ?‘he unmyelinated gap (between large arrows) extends approximately 30 pm along the axis of the fiber. T h e axon surface is elaborated at the node to form a layer of irregular processes ( P ) which further increase the nodal surface area. The inset shows, for comparison, a node of Ranvier from a similar fiber near its site of entry into the electric organ; the nodal gap measures less than 1 pm along the axis of the fiber. Both nodes are shown at the same magnification. e = extracellular space. x 2560.
16
STEPHEN G . WAXMAN
ity of the electrocytes, and is diphasic (initially head-positive) . The physiological data indicate that impulses in the electrocyte axons propagate to involve both the anteriorly and posteriorly running segments, the first generating the head-positive phase and the second generating the head-negative phase of the discharge (Fig. 4 ) . Spikes are generated only by nodes with normal morphology; the enlarged nodes in distal portions of the anteriorly
A
Recording sites
Potentials hood
+
phase
,,
hood-
phase
Current directions
a x i a l resistance
B excitable nodes
I A e r i e s . capacity
%-3 external resistance
exter no I potential
FIG. 4. Axonal function during electric organ discharge in Sternarchus. ( A ) Intracellular recordings are shown in the center column as they would be obtained from sites along the axon as indicated on the left. A single cycle of the externally recorded organ discharge is shown in the uppermost trace of the center column. Narrow nodes in the proximal part of the anteriorly running segment become active and pass inward current during the head-positive phase of organ discharge. T h e large nodes in the distal part of the anteriorly running segment are inexcitable. During the head-positive phase external current runs in a caudal direction ((diagram on right). Narrow nodes in the proximal part of the posteriorly running segment are active, and the enlarged distal noden inactive, during the head-negative phase. Modified from Bennett ( 1971a). ( B ) Equivalent circuit of electrocyte segment, illustrating the effect of a series capacity. From Waxman et al. (1972).
DESIGN PRINCIPLES OF AXONS
17
and posteriorly running segments do not generate spikes. Initially, only the nodes in the proximal part of the anteriorly running segment are active. The enlarged nodes in the distal part of the anteriorly running segment are inexcitable and pass outward current. External currents run posteriorly, generating a head-positive phase. The normal-appearing nodes in the region where the fiber turns around are excited by the reduced spike. The narrow nodes in the proximal parts of the posteriorly running segments fire subsequently, and since the large nodes in the distal part are inactive, external current flows in an anterior direction, so as to generate the head-negative phase of the organ discharge (Bennett, 1970, 1971a). Several results suggest that the enlarged nodes act as a series capacity (Fig. 4 B ) . The evidence is that there is no dc component to the discharge, indicating that there is no net current flow averaged over a single discharge cycle. When propagation into the posteriorly running segment is blocked by anoxia, there is still no net current flow, demonstrating that the outputs of the segments exhibit no net current flow. I n addition, in other sternarchids where the anteriorly running segment is reduced or absent, the head-positive phase is reduced or absent, but the discharge still exhibits no dc component (Bennett, 1970, 1971a). The large surface area at the distal nodes, which is augmented by the polypoid elaboration of the axon surface (Fig. 5 ) , provides a morphological correlate for the increased capacity. I n other electrocytes and electroreceptors whrre membranes act as a series capacity (Bennett, 1970, 1971a,b), the membranes are elaborated as in the large nodes
FIG. 5. The gap in the myelin at a large node from Sternarchus electric organ. The axoplasm is indicated AX. Fiber axis runs horizontally. The unmyelinated gap extends between the arrows. Note the elaboration of the axon surface to form a layer of irregular polypoid processes (P.) e = extracellular space. ~ 7 5 0 0 .
18
STEPHEN G . WAXMAN
of the Sternarchus electrocytes (Schwartz, 196%; Schwartz and Pappas, 1968; Bennett, 1971b). The Sternarchus electrocyte axons provide an example of two principles of axonal design. First, they demonstrate that axons need not be uniform structures with the same morphology throughout their course, but may rather exhibit a high degree of regional differentiation, in terms of both morphology and physiology. Second, they illustrate that axons need not function as simple conduits which transmit information with high security from one site in the nervous system to another as rapidly as possible. I n this case the axons mediate a transformation of spikes into diphasic external signals.
V. External Effects on Axons
Presynaptic inhibition has been demonstrated at numerous sites, both in invertebrates (Dude1 and Kuffler, 1961; Tauc, 1960) and in inframammalian (Furakawa et al., 1963) and mammalian vertebrates (see e.g., Wall, 1958; Frank, 1959; Andersen et al., 1962; Horcholle-Bossavit and Tyc-Dumont, 1969; Pappas and Waxman, 1972). Clearly, the terminals of some axons are subject to modulation depending on the level of presynaptic inhibition. Available data do not yet indicate whether conductance or voltage changes, in an element to which an axon is electrotonically coupled, effect conduction properties in the axon. There is evidence for nonsynaptic “ephaptic” electrical interactions between nearby nerve fibers ( Katz and Schmitt, 1940; Arvanitaki, 1942; Marrazzi and Lorente de No, 1944; Renshaw, 1946). I t is not clear what role such effects may have in normal integrative processes, although the data certainly suggest the possibility of excitability changes in axons as a result of activity in adjacent structures. Baylor and Nicholls ( 1969a) have demonstrated significant increases in extracellular potassium concentrations following activity in the central nervous system of the leech. Their studies have also demonstrated long-lasting hyperpolarization following activity in leech sensory neurons and their processes (Baylor and Nicholls, 1969b) . Sensitivity of the neuronal membrane potential to external potassium concentration increased during hyperpolarization. I t was suggested that neighboring neurons or neural processes might interact by a nonsynaptic potassium-mediated mechanism, depending on the previous history of the cell. Such effects might be especially significant at sites of low safety factor. Chung et al. (1970), in a study on cat spinal cord, correlated the degree of conduction block in “ectodromic” dorsal root fibers with the degree of activity in neighboring dorsal rootlets. The degree of conduction block increased during a period beginning just after the peak of the long negative wave of the dorsal root potential. Physiological stimulation
DESIGN PRINCIPLES OF AXONS
19
(moving or touching the hind limb ipsilateral to the recorded rootlet) also led to variations in safety factor. Evidence for external electrical effects has also been adduced by Lurie (1973), whose studies on the dc-recorded electroretinogram of intact frog eye suggest that long-term voltage changes, including the c-wave, are the result of a bleached rhodopsin signal dependent on the integrity of the pigment epithelium. There was a close correlation between the time course of the c-wave and the pattern of activity in class I V optic nerve (off-) fibers recorded simultaneously. Small fluctuations in the slow voltages were accompanied by changes in fiber activity. Similar changes in fiber activity were produced by applied transretinal currents, and during the slow wave produced by intra-arterial administration of sodium azide. In addition, selective elimination of the slow voltages by retinal detachment was accompanied by absence of slow changes in the pattern of activity, suggesting that extracellular currents might modify activity in neural elements, and that information about sensitivity (state of adaptation) might be transmitted from receptor to ganglion cell by means of extracellular current flow.
VI. Electrotonic Coupling by Axonal Pathways
Electrotonic coupling of cells via axonal pathways was first demonstrated in teleost electromotor systems (Pappas and Bennett, 1966; Bennett et al., 1967a). In some neural systems, neurons are directly coupled by somatosomatic, dendrosomatic, or dendrodendritic electrotonic junctions. I n other systems, where there is physiological evidence for coupling between neurons, electrotonic junctions between the coupled neurons are not observed by electron microscopy. The presence of gap junctions between axons and neuronal somata or dendrites suggests that axonal pathways are responsible for electrotonic coupling, a group of neurons being coupled to each other by virtue of each being coupled to the same prejunctional fiber (Pappas and Bennett, 1966) . Intracellular recording techniques have demonstrated physiologically the coupling of several neurons to a single presynaptic fiber. Criteria for identification of prejunctional fibers include absence of spikes in response to antidromic stimulation, absence of postsynaptic potentials preceding spontaneous or evoked discharges, and lack of effect of applied hyperpolarizing current on spontaneous or evoked discharges (Kriebel et al., 1969). Coupling of a presynaptic fiber with a group of neurons is demonstrated by the presence in this fiber of short latency graded depolarizations in response to antidromic stimulation of the neurons. Gradedness of the short latency antidromic response reflects difference in threshold for antidromic stimulation and constitutes physiological evidence for coupling of a prejunctional
20
S T E P H E N G . WAXMAN
fiber to more than one postsynaptic neuron. Morphological confirmation for the concept of axonal coupling pathways has recently been provided by Meszler et al. (1972, 1974), who demonstrated by electron microscopy, in serial or appropriately oriented single sections, electrotonic junctions between single prejunctional fibers and several electromotoneurons in the spinal cord of the electric eel Electrophorus. Similarly, single axons have been shown to establish gap junctions with several electromotor neurons in the spinal cord of Sternarchus albifrons (Pappas et al., 1975). Although the initial demonstrations of electrotonic junctions were in inframammalian species, there have within the past several years been demonstrations of electrotonic synapses in several areas in the mammalian central nervous system. I n at least one of these regions, the data strongly suggest an axonal coupling pathway. Korn et al. ( 1973) have demonstrated electrical coupling between giant neurons in the rat lateral vestibular nucleus. The evidence is based on the presence of graded antidromic depolarizations in giant neurons in response to vestibulospinal tract stimulation and on collision experiments in which the graded antidromic response was not blocked by directly evoked spikes which did: however, block antidromic spikes. Electron microscopy revealed gap junctions between axon terminals and cells bodies, but not between neighboring perikarya or between dendrites. The evidence suggests electrotonic coupling of neurons via prejunctional pathways in the mammalian brain.
VII. Structure-Function Relations for Central Axons
A. NODESA N D INTERNODE SPACING The geometry of central myelinated fibers differs from that in peripheral nerve in that the nodes of Ranvier may be larger; in addition, internode distances in neuropil may be much shorter than in peripheral nerve or white matter. I n the peripheral nervous system, the nonmyelinated gap at nodes of Ranvier usually extends less than 1 pm along the axis of the fiber (Hess and Young, 1952; Robertson, 1959). A dense cytoplasmic coating is present subjacent to the axon membrane at the node (Elfvin, 1961) ; a similar undercoating is present at the axon initial segment (Palay et al., 1968). At central nodes of Ranvier, the nodal gap can extend for less than 1 pm or can be considerably larger. Hess and Young (1952) noted, on the basis of light microscopic studies, that at central nodes a longer stretch of axon could be left bare than in peripheral nerve. Chang (1952) described what he regarded as “segments of myelin sheath widely separated by unmyelinated stretches” in Golgi-Cox preparations of cerebral cortex. Electron microscopy has confirmed the existence of nodes extending 10 pm or more in the central
21
DESIGN PRINCIPLES O F AXONS
nervous system (Metuzals, 1965; Gray, 1970; Waxman, 1971, 1972; Witkovsky, 1971). An exmple is shown in Fig. 6. At other nodes bulbous protrusions of the axon increase the surface area (Waxman, 1971). There is also evidence at a number of sites that synapses may arise at nodes (Bodian and Taylor, 1963; Khattab, 1966; Bennett et al., 1967b; Sotelo and Palay, 1970; Waxman, 1970). These may be of either the chemical or electrotonic type. ‘4t most central nodes, the dense cytoplasmic undercoating is present subjacent to the unmyelinated axon membrane. At some enlarged nodes, however, it is observed only subjacent to a small part of the axon membrane (see below) . In peripheral nerve there is an approximately linear relationship between internode distance and fiber diameter. Tasaki et al. (1943) reported a ratio of 205 between internode distance and fiber diameter for amphibian sciatic nerve. Other workers have reported the relationship between diameter x and internode distance Y to be of the form Y = A Bx where A = the 3’ intercept and B = a slope coefficient. The values for A and B varied depending on species and size. For a 40-cm specimen of Raia, the data indicated that A = -0.14 mm and B = 0.15 (Thomas and Young, 1949). For human sural nerve the values of A range between -0.14 mm and $0.14 mm and B between $0.03 and +0.12 (Gutrecht and Dyck, 1970). Internode distances of less than 200 pm occur in peripheral nerve but are rare (Lubinska, 1958). Hess and Young (1952) reported a similar monotonic relationship between internode distance and fiber diameter in white matter (ventral and lateral funiculi) of rabbit spinal cord, where minimum internodes (which corresponded to the smallest fibers, with diameter of less than
+
FIG. 6. Electron micrograph of a central node of Ranvier, from the teleost oculomotor nucleus (Chilornycterus) . T h e unmyelinated gap (between arrows) extends more than 10 pm along the axis of the fiber (ax). A synapse is established with a spine, which is cut in cross section. ~ 9 4 0 0 .
22
STEPHEN G . WAXMAN
5 pm) were of the order of 200 pm. In gray matter, there is evidence that the internode distances may be shorter. Chang (1952) illustrated but did not give measurements of short myelinated internodes along axons in mouse cerebral cortex. Bodian’s (1951 ) study of internode distances in preoptic area, hypothalamus, hypoglossal root and pyramidal tract of adult opposum indicated a roughly linear relationship between internode distance and fiber diameter, but included observations of internodes only 50 pm long. Studies by Haug (1967) and Waxman and Melker (1971) indicate that nodes of Ranvier along fibers in mammalian neuropil may be somewhat more closely spaced than would be predicted on the basis of their diameters from the internode distance-diameter relationships for white matter and peripheral nerve. I n some parts of the teleost brain, internode distances may be strikingly reduced (see Fig. 7 ) ; fibers several micrometers in diameter with internode distances of less than 10 pm have been described (Waxman, 1970, 1972). There is no question that the properties of axons may change along their course. Sunderland and Roche (1958) have suggested that the chemical characteristics of myelin may change along the course of axons. The properties of myelin obviously change along fibers that traverse the central nervous system-peripheral nervous system boundary (Tarlov, 1937). Changes in diameter at different levels along the same fiber are well documented and Hildebrand’s ( 1972) preliminary demonstration of a correlation between myelin period and fiber diameter suggests possible qualitative differences in myelin at regions of different diameter. Nodal geometry may also change. One obvious example is the peripheral-central nervous system interface. There is also evidence for changes in the structure of nodes along fibers confined to the central nervous system. In regions of the teleost nervous system where nodes of Ranvier are closely spaced, it is possible, in serial or appropriately cut single thin sections, to follow a fiber for several successive internodes. Figure 7 illustrates one such fiber along which the nodes exhibit variation in terms of surface area. There is also evidence for possible differentiation of central axons in terms of nodal membrane properties (Waxman, 1974). At most peripheral and central nodes of Ranvier, there is a dense cytoplasmic undercoating approximately 200 8, thick subjacent to the axon surface (Elfvin, 1961; Andres, 1965; Peters, 1966). Figure 8 illustrates the undercoating, at a node from rhesus monkey oculomotor nucleus. A similar undercoating is present at the axon initial segment (Palay et al., 1968; Peters et al., 1968). The undercoating is absent or attenuated in regions where synaptic terminals contact the initial segment. Because of the distribution of the dense undercoating, it was suggested by Palay et al. (1968) that it may represent a structural modification of the axon membrane related to specific membrane
FIG. 7. Myelinated fiber from the pacemaker region of the electromotor nucleus in the gymnotid Sternopygus. Four nodes of Ranvier (N, - N,) are separated by internode distances of less than 10 pm. The surface area of node N4 is significantly greater than that of nodes N,, Nz, and N:,. Nodes N? and N4 are enlarged in the insets. At N,, there is a close apposition (arrow, upper inset) with a dendrite ( D ) . C = capillary. ~ 7 7 0 0 insets ; ~13,400.
24
STEPHEN G . W A X M A N
FIG. 8. This electron micrograph shows part of a myelinated axon ( a x ) at a node of Ranvier from the rhesus monkey oculomotor nucleus. A dense cytoplasmic undercoating is present subjacent to the unmyelinated axon membrane at the node (between arrows). Adjacent to the axon are a dendritic process ( p ) and a n enlarged extracellular space ( e ) . x82,OOO.
properties. As noted in Section IV, D of this paper, along the electrocyte axons in the gymnotid Sternarchus there are specialized nodes at which spikc generation does not occur (Bennett, 1970, 1971a). At these nodes, surface area is increased, and in contradistinction to most other nodes, the dense undercoating is absent (Waxman et al., 1972). At most central nodes, the dense undercoating is present and forms a continuous layer beneath the unmyelinated nodal membrane. However, the undercoating is absent or limited in extent at nodes at which synapses arise or unmyelinated collaterals emerge (Figs. 9 and 10). Serial or appropriately oriented single sections of collaterals arising at nodes demonstrate that the dense undercoating in most cases extends for only several micrometers or less along the axon membrane into the collateral ; more distant membrane is devoid of the dense cytoplasmic layer (Waxman, 1974). The morphological data are thus consistent with the hypothesis that differences in nodal membrane properties, in addition to differences in surface area and geometry, may contribute to the differentiation of central myelinated axons. T h e majority of neural models require nodal surface area to be proportional to fiber diameter (see, e.g., Rushton, 1951; Dun, 1970; Goldman and Albus, 1968). Livingston et al. (1973) have studied the morphology of the glia-axonal junctions, which exhibit a degree of differentiation in both peripheral and central nodes of Ranvier. In view of the dis-
DESIGN PRINCIPLES OF AXONS
25
FIG. 9. Absence of dense undercoating at a synaptic node of Ranvier from monkey oculomotor nucleus. The terminating myelin is indicated by m. Vesicles (ves) are clustered at a site of synaptic contact with a n adjacent dendrite ( D ) . There is no dense undercoating a t this node. T h e inset shows the axon surface at a nearby node (nonsynaptic) a t which a cytoplasmic undercoating is present (indicated by d ) . ~ 3 0 , 0 0 0 .
crepancies between calculated and observed areas of nodal membrane (Stampfli, 1954), the glia-axonal junction may contribute to nodal activity. Computer simulations of transmission properties of partially demyelinated axons (Koles and Rasminsky, 1972) predict changes in conduction velocity and in safety factor at sites of myelin loss; small changes in fiber geometry had significant effects on transinission properties. McDonald and Sears (1970) and Davis (1972) have demonstrated reduction in conduction velocity and failure of transmission of high-frequency impulses at sites of demyelination, and internodal conduction times of more than 600 p e c for partially demyelinated fibers have been reported by Rasminsky and Sears (1972). In the gymnotid neuroeffector axons described above there is a definite relationship between the structure of nodes and their electrophysiological properties. The data also indicate a relation between internode distance and conduction velocity. There is theoretical evidence that, for any given fiber diameter, there is an optimal internodal distance for maximal conduction
26
STEPHEN G. WAXMAN
FIG. 10. Node of Ranvier from the oculomotor nucleus of the chameleon Anolis carolinensis. Terminating myelin lamellae are labeled m. A collateral ( C ) arises a t the node and extends to the upper left of the figure, where it forms synaptic contacts with adjacent dendritic elements. T h e cytoplasmic dense undercoating extends for only a short distance along the nodal membrane from which the collateral arises (between arrows). ~ 5 8 , 0 0 0 .
velocity (Huxley and Stampfli, 1949; Hardy, 1971), and there is physiological evidence for delayed conduction along fibers with relatively short internode distances in spinal ganglia ( I t o and Takahashi, 1960). Some of the sites in the teleost nervous system at which the internode distances are very short, are known to involve delay mechanisms (see Section 111). Although the internode distance-velocity relationship may have a broad maximum, for fibers in these areas (at which in some cases nodal surface area approaches myelinated surface area) , conduction velocity is probably substantially below the maximal possible for fibers of that diameter.
B. DIAMETER SPECTRA The now classic studies of Erlanger and Gasser (1937) clarified the relationship between conduction velocity and fiber diameter in peripheral nerve
DESIGN PRINCIPLES OF AXONS
27
trunks. Bishop (1966) has commented on the fact that, while large myelinated fibers are more common in the cortical spectrum in mammals than in inframammalian species, the myelinated fiber population in mammalian cortical systems still contains relatively few large fibers compared to peripheral nerve. Bishop and Smith ( 1964) have demonstrated fibers considerably smaller than 1 pm in mammalian and reptilian cortical white matter. Myelinated fibers with diameters of 0.3 pm or less have been described in mammalian caudate nucleus (Adinolfi and Pappas, 1968), teleost oculomotor (Waxman and Pappas, 1971) and electromotor nuclei (Waxman, 1971 ) , and reptilian oculomotor nuclei (Waxman and Bennett, 1972). Fibers begin to acquire myelin sheaths at diameters of 0.2 pm in dorsal funiculus of rat spinal cord (Matthews and Duncan, 1971). This is in distinct contrast to peripheral nerve, where 1 pm is the critical diameter at which myelin is first seen (Vizoso and Young, 1948; Matthews, 1968). Bishop has noted that, in mammalian cortex, not over 20% of the fibers have diameters greater than 3 pm. From observations on cat optic nerve, Bishop et al. (1969) derived diameter spectra with peaks at approximately 1 pm, and with a majority of fiber diameters less than 3 pm. Hildebrand and Skogland (1971) have presented data on fiber caliber spectra from cat gracile and cuneate fasciculi, dorsal part of dorsal columns, anterior and posterior lateral funiculi, and pyramidal tract. I n adult cats, the largest dorsal column fibers were 12-15 pm, with only 30-45c/o having diameters of 4 pm or more in the dorsal part of dorsal column, and 17% measuring 4 pm or more in gracile fasciculi. In posterior lateral funiculi, 13-22% of fibers measured 4 pm or morc. I n pyramidal tract, a large proportion (50-60'/c) of fibers had diameters of approximately 1 pm, with only 6-9% having diameters of more than 4 pm. There have been several systematic studies of diameter spectra in nuclear regions of the central nervous system. Gobel and Purvis (1972) have presented data on myelinated axon diameters in the deep bundles of the spinal V nucleus in cats; 80-90% of the axons have diameters between 0.3 pm and 1.5 pm. Myelinated fibers in cat caudate nucleus range in size from 0.3 pm to 1.6 pm. The majority are approximately 0.6 pm in diameter (Adinolfi and Pappas, 1968). In the reptilian oculomotor nucleus, 84% of myelinated fibers have diameters of less than 2 pm, with 48% smaller than 1 pm (Waxman and Bennett, 1972). Suriderland and Roche (1958) have noted that the cross-sectional shape and the diameter of nerve fibers may vary significantly along single internodes. Williams and Wendell-Smith ( 1971 ) have demonstrated changes in fiber diameter and in the relations of myelin thickness to diameter and of internodal distance to diameter in populations of nerve fibers sampled at different points along their course. Fraher (1972), in discussing the varia-
28
STEPHEN G . WAXMAN
tions in axon circumference associated with a given sheath thickness, suggests that axon caliber may change longitudinally, and that the thickness of the myelin sheath may be different at different parts of the internode. The morphological data indicate a relative paucity of large myelinated fibers in the central nervous system. A large proportion of the myelinated fibers in the central nervous system have diameters of 1 pm or less (Fig. 11 ) , Dimensional arguments (Waxman and Bennett, 1972 ; see also Section VII, C ) suggest that myelinated fibers 1 pm in diameter have conduction velocities that are approximately 2.6 times larger than those of nonmyelinated fibers of similar size, and that the relative increase in conduction velocity is smaller than this for smaller fibers. This estimate does not take into account the increase in rise and fall time of the spike in small myelinated fibers (Paintal, 1966), which, as pointed out by Huxley (cf. Waxman and Bennett, 1972), would tend to decrease myelinated fiber conduction velocity (Coppin and Jack, 1972).
FIG. 11. Electron micrograph of neuropil from the oculomotor nucleus of the lizard A n o h carolinensis, including a synapse between an axon ( A ) and dendrite (D) . The profiles of myelinated fibers of varying diameters are present. The diameter of fiber m, is 0.6 pm and g = 0.78 for this fiber. The diameter of fiber m2 is 0.4 pm, and the value of g for this fiber is 0.75. Fiber m3 has a diameter of 1.2 pm and g = 0.85 for this fiber. ~ 2 4 , 8 0 0 .
DESIGN I’RINCIPLLS
OF A X O N S
29
C. CRITICAL DIAMETER FOR MYELINATION As shown above, a significant proportion of myelinated axons in the central nervous system have diameters of less than 1 pm, the smallest myelinated fibers having diameters of about 0.2 pm. Since conduction velocity for myelinated fibers varies directly uith diameter while conduction velocity for nonmyelinated fibers varies with the square root of the diameter, the relationships between conduction velocity and diameter must cross a t some point, suggesting that below this diameter the nonmyelinated fibers will conduct more rapidly than myelinated fibers of similar diameter. Rushton ( 1951) , noting that 1 pm is the diameter at which fibers are myelinated in peripheral nerve (Visozo and Young, 1948; see also Matthews, 1968), presented a series of arguments leading to the conclusion that 1 pin was the diameter at which the two diameter-conduction velocity relationships crossed ; i.e., that 1 pm corresponded to a critical diameter above which “myelin increases conduction velocity’’ and below which “conduction is faster without myelination.” This conclusion was based on the relationships shown in Fig. 12, which are redrawn from Rushton’s ( 1951 ) Fig. 5. The diameter-conduction velocity relationship for nonmyelinated fibers is a parabola perpendicular to the ordinate at the origin; it was drawn on the basis of the proportionality of conduction velocity to the square root of fiber diameter, using Gasser’s (1950) measurements of diameter ( 1.1 pm) and conduction velocity (2.3 m/sec) for the largest fibers. The relation between conduction velocity and diameter for myelinated fibers intersects the parabola a t a point corresponding to a 1 pm diameter, predicting that the smallest central myelinated fibers conduct at slower rates than nonmyelinated fibers of the same diameter. This prediction in itself is surprising, although not in theory impossible. However, extrapolation downward of the velocity-diameter relationship for myelinated fibers leads to intersection with the abscissa at a diameter of 0.6 pm, suggesting that fibers smaller than this should not conduct impulses at all. Together with Dr. M. V. L. Bennett, the present author has reexamined the arguments leading to the prediction of a critical diameter of 1 pm in myelinated fibers (Waxman and Bennett, 1972). The derivation of the diameter-conduction velocity relation for myelinated fibers was based, in Rushton’s formulation, on the relation
V
Dg d - l o g , g
(3) where g is defined as axon diameter divided by overall fiber diameter, and ‘v is conduction velocity. Rushton used Sanders’ (1948) measurements for g to compute the left side of the equation, and fit the resulting curve to Hursh’s (1939) data relating Conduction velocity and diameter, as shown
30
STEPHEN G. WAXMAN
Fiber diameter
(pm)
FIG. 12. Relations between conduction velocity and fiber diameter for small myelinated and nonmyelinated fibers. Modified from Rushton’s (1951) Fig. 5 as indicated in the text. The circled point represents Gasser’s (1950) measurements for the largest C fibers. The revised linear relation for myelinated fibers (-.-. ) intersects a t a point corresponding the parabolic relation for nonmyelinated fibers (-) to a diameter of about 0.2 pm. I t is suggested that this value rather than the 1 pm intersection provided by Rushton’s relation for myelinated fibers (- - -) is the critical diameter above which myelinated fibers can be expected to conduct more rapidly than nonmyelinated fibers of the same size. From Waxman and Bennett (1972).
in Fig. 13. The extrapolated region of the curve (dashed line) was replotted on an expanded scale in Rushton’s Fig. 5 (see Fig. 12). Sanders’ data for g were derived from light microscopic observations on rabbit peripheral nerves, and suggest that the value of g decreases rapidly for small fibers (Fig. 14). Extrapolation of g to zero at a diameter of 0.6 pm accounts for the prediction of conduction failure at and below this diameter, since axonal core resistance becomes infinite. More recent studies using electron microscopy (Waxman and Bennett, 1970, 1972; Waxman, 1975) indicate the value of g for central fibers is approximately constant and does not vary appreciably with diameter (Fig. 1 3 ) . Schnepp and Schnepp ( 1971 ) have shown that electron microscopy of cross sections of peripheral nerve yields a nearly constant value for g, while light microscopy on the same nerves yields values similar to those reported by Sanders. If it is assumed that the value of g is constant, it follows from Eq. ( 3 ) that conduction velocity should be proportional to diameter, and the velocity-diameter relationship should intersect the origin. The revised, linear relationship between conduction velocity and diameter, together with Rushton’s relationships for myelinated and nonmyelinated fibers, are shown in Fig. 12. The revised relation for myelinated fibers intersects the relation for non-
31
DESIGN PRINCIPLES OF AXONS
120
..
-
-I 6
4
8
10
12
14
18
16
Fiber diameter (pm)
FIG. 13. Relations between conduction velocity and fiber diameter for myelinated axons, modified from Rushton’s (1951) Fig. 3. Open circles and dots represent Hursh’s ( 1939) observations on fibers from kittens and cats, respectively. Rushton’s relation computed using Sanders’ measurements of g (the ratio of axon diameter to overall fiber diameter) is indicated by the solid curve with dashed extrapolation for small diameters. T h e linear relation assuming constant g is indicated by the broken line; its slope is 5.5 m sec-’ pmP. From Waxman and Bennett (1972).
. .
*
+ + +++
go.6: 04
+
+ ++
+*
++
+ + +++ 3: ++ +t; * ++ti+++ +++ + +A%+, f ++ ++ ++ + +++ * + + +
++; + + i ++ ** +++
++
+
+
+++ +++ )*+ +
0
f*++:
2
4
6 8 FIBER DIAMETER ( k m )
10
12
FIG. 14. Values of g as a function of myelinated fiber diameter. Fibers from the oculomotor nucleus of the lizard Anolir carolinensis are represented by dots. Data obtained by electron microscopy. Twenty-four of the fifty fibers have diameters under 1 pm; g is independent of diameter and ranges between 0.54 and 0.88 with a mean of 0.77. Sanders’ (1948) data for rabbit fibers are indicated by crosses (taken from his Fig. 3 ) . With his light microscopic techniques, g appears to decrease markedly for small fibers. Modified from Waxman and Bennett (1972).
32
STEPHEN G . WAXMAN
myelinated fibers at a point which corresponds to a diameter of 0.2 pm, suggesting that this is the critical diameter above which myelinated fibers should conduct more rapidly then nonmyelinated fibers of the same size. This is, in fact, the diameter of the smallest central myelinated fibers which have been reported. I t is unlikely, on morphological grounds, that much smaller myelinated fibers are present, since the minimal sheath consisting of a single layer of myelin is approximately 200 A thick, implying a diameter of 0.1 pm if the optimal value of g (0.6) obtains. The absence of myelinated fibers smaller than 1 pm in peripheral nerve may be related to several recent observations. The assumption that specific mcmbrane properties are constant for myelinated fibers of different diameter is contradicted by recent data indicating that rise and fall time of the spike are greater for small-diameter myelinated fibers (Paintal, 1966). As might be expected from this result, internodal conduction time is greater in fibers of small than of large diameter (Coppin and Jack, 1972). A second possible limiting factor for reliable operation of small myelinated fibers has been suggested by Hille ( 1970), who has commented on the unreliability in terms of the state of sodium channels at nodes of small-diameter fibers.
VIII. Functions of Axons in the Normal Nervous System
The foregoing indicates a multiplicity of functions for axons. It is clear that some axons function as simple transmission lines, in which speed of conduction and a high degree of fidelity for the transmission of each impulse represent primary criteria of design. Thus, the transmission line hypothesis does apply to some fibers. In other cases, however, nerve fibers are not constrained to function as simple conduits, but rather mediate transformations on neural information. Thus, in addition to transmitting information from one neural locus to another, axons may function, in some cases, as delay lines (in which the transformation is one of phase-shifting) or may mediate more complex spatiotemporal transformations by frequency-dependent impulse intermittency or filtering (see Fig. 15). Several “local” functions are also suggested by the physiological and morphological data. The evidence for external influences on axonal properties suggests an interactive function, which may in turn be reflected in conduction properties, such as safety factor. Finally, axons may function in a local context by providing pathways for electrotonic coupling. Interactive and local effects may be reflected more globally within the axonal tree in terms of the resetting of contextual parameters, so that the axonal tree must be represented as a complex network with properties that vary along both the spatial and temporal domains. I t is not surprising, in this context, that axons exhibit a high degree of local
DESIGN PRINCIPLES OF AXONS
33
FIG. 15. Integrative properties of axons. ( A ) Simple transmission line model. ( B ) Delay line model. ( C ) Intermittent conduction (transformational element) model. The available evidence suggests a multiplicity of functional properties for axons, which may mediate transformations of neural information in both the spatial and temporal domains. T h e reader is referred to the text for details.
differentiation, both in terms of morphology and in terms of physiological properties and principles of design. As will be discussed below (see Section X ) , the differentiation of axons implies that neural information is subject to transformation at a number of sites within the neuron. IX. Demyelination: Pathophysiological Aspects of Delayed Conduction and Intermittence
Most studies on the pathophysiology of axonal conduction in demyelinating diseases have stressed the deviations from normal axonal function that result from demyelination. However, as noted above, there is evidence that similar deviations from classical axonal physiology occur in the normal nervous system. These similarities are commented on here because they suggest that some neural properties which are usually considered pathological may in fact have significance for normal nervous function, and since they also suggest that experimental demyelinating lesions, which may be produced in the laboratory, may provide a model for the study of intermittent conduction in less accessible regions of the normal brain and spinal cord. In the paranodal type of demyelination, myelin loss occurs near nodes of Ranvier, so that the adjacent internodes are separated by an unmyelinated gap larger than the usual 1 pm (Harrison et al., 1972). At remyelinated areas, internode distances are often reduced. Early studies by Mayer and Denny-Brown ( 1964) demonstrated decrease in conduction velocity and conduction block along some peripheral nerve fibers at sites of demyelination. Conduction failures at frequencies as low as 60 per second have been
34
STEPHEN G . WAXMAN
observed in segmental demyelination of peripheral nerve (Lehmann et al., 1971), and more recently at frequencies as low as 25 per second in peripheral nerves from guinea pigs with experimental autoimmune neuritis (Davis, 1972). Studies of diphtheritic demyelination in the central nervous system (McDonald and Sears, 1970) have shown reduction in conduction velocity, prolongation of refractory period, and failure of transmission of high-frequency (290 per second) impulse trains at sites of focal demyelination. Rasminsky and Sears (1972) noted intermittent propagation after 90 seconds of stimulation at 80 per second. Computer simulations of the behavior of demyelinated fibers indicate that small changes in the geometry of the myelin sheath may significantly effect transmission properties (Koles and Rasminsky, 1972). The data from demyelinating lesions suggest there may be some similarities in the relationships between morphological and functional properties for pathological and nonpathological systems. Physiological mechanisms such as reduction in conduction velocity or frequency-related conduction block, which have classically been considered pathological, appear to play a role in normal integrative processes. Experimental demyelinating lesions may therefore provide a laboratory model for the study of “nonoptimal” conduction properties. In recent studies of internodal conduction in undissected demyelinated nerve fibers (Rasminsky and Sears, 1972), recordings of extracellular longitudinal currents from demyelinated spinal roots indicated that membrane currents were confined to regions less than 200 pm long, separated by distances in some cases of 1 mm or more. The length of demyelinated regions along the fibers was not determined. However, the persistence of saltatoiy conduction in demyelinated fibers suggests that, with appropriate morphological investigations of the extent of demyelination, it should be possible to determine whether the internodal axon membrane, which is usually covered by myelin, is electrically excitable (Rasminsky and Sears, 1972).
X. Conclusions and Summary
Foregoing sections of this review have focused on the spectum of integrative properties and multipicity of design principles exhibited by axons. A large body of information indicates that axons are not necessarily uniform structures, but may rather exhibit regional differentiation, in terms of both morphological and physiological properties. It seems clear that the axon need not be regarded as a simple conduit, but that it may rather exhibit more complex properties and function as a filtering system or transformational element. Recent studies on the morphology and electrophysiology of dendrites (Purpura, 1967, 1971; Rall et al., 1967) have elaborated the mechanisms
DESIGN PRINCIPLES OF AXONS
35
for spike electrogenesis, and dendritic inhibition and summation of postsynaptic potentials, and, together with morphophysiological studies on the differentiation of dendritic systems (Purpura, 1971) , indicate the importance of dendrites as local elements in integrative processes. These data, together with data derived from studies on axons, suggest a complex picture of the functional organization of the neuron. The model which begins to emerge is one of a hierarchical array of logical operators, which sequentially process information first at dendritic loci, next at the initial segment, and finally in the axon and its terminals (Waxman, 1972). Figure 16 illustrates this conceptual model of the neuron. Superimposition of dendritic integrative mechanisms (phase I ) on threshold operations a t the initial segment (phase 11), together with transformations in the axon (phase 111) and the axon terminals (phase I V ) , endow the neuron with a rich logical structure far exceeding that of a simple threshold element. The “multiplex” model of the neuron thus exhibits the c,haracteristics of a cascaded array of logical elements. Relaxation of the constraint of bistable behavior suggests the possibility of a neural representation for higher-order calculi. The multiplicity of integrative mechanisms and hierarchical structure imply that the func-
Ia
FIG. 16. ’I‘he multiplex ncurun. Impulse initiation sites in the dendrites and cell body are indicated by shading. Transformation of neural information occurs sequentially, first in the dendritic zone (phase I ) , then by initiation of series of impulses at the axon initial segment (phase I I ) , and by transformations within the axonal tree (phase 111), and finally by modulation of activity a t axonal terminals by presynaptic inhibition (phase I V ) . Information is transformed in both the spatial and temporal domains. T h e formal equivalent is a cascaded array of transformations.
36
STEPHEN G . WAXMAN
tional properties of nerve cells are determined not only by patterns of connectivity, but also by a complex logical infrastructure. The richness of structure exhibited by even the single neuron imparts a formidable complexity to morphophysiological analysis. This holds true particularly for axonal systems, in which the processes may be of fine caliber, with complex patterns or arborization. Nevertheless, it is not unlikely that future studies will further clarify the functional significance of the array of structural patterns exhibited by neurons and in particular by axons, and it seems not unreasonable to expect the development of models which reflect the dynamic, as well as static, properties of axons. I t may also be expected that future investigations will lead to a fuller understanding of developmental mechanisms, and of the pathophysiology of axons. Hopefully the newer data will contribute to a more complete picture of the functional architecture of the nervous system. ACKNOWLEDGMENTS The author’s research has been supported by grants from the National Institute of Neurological Diseases and Stroke (NB-07512, NS-12307, 1K04-NS-00010) and the National Institute of General Medical Sciences (5T5-GM-1674) and by a grant from the Epilepsy Foundation. I t is a pleasure to acknowledge the advice and support of Drs. G. D. Pappas, M. V. L. Bennett, and D. P. Purpura, without whose help my investigations could not have been initiated. I also wish to thank Dr. N. Geschwind for stimulating comments and encouragement, and Dr. P. D. Wall for many helpful discussions.
REFERENCES Adinolfi, A. M., and Pappas, G. D. (1968). J . Comp. Neurol. 133, 167. Albe-Fessard, D., and Martins-Ferreira, H. (1953). J . Physiol. (Paris) 45, 533. Andersen, P., Eccles, J. C., and Schmidt, R. F. (1962). Nature ( L o n d o n ) 194, 741. Andres, K. H. (1965). 2. Zellforsch. Mikrosk. Anat. 65, 701. Arvanitaki, A. ( 1942). J . A europhysiol. 5 , 89. Barron, D. H., and Matthews, B. C. (1935). J . Physiol. ( L o n d o n ) 85, 73. Baylor, D. A,, and Nicholls, J. G. (1969a). J . Physiol. ( L o n d o n ) 203, 555. Baylor, D. A,, and Nicholls, J. G. (1969b). 1. Physiol. ( L o n d o n ) 203, 571. Bennett, M. V. L. (1966). Fed. Proc., Fed. Amer. Soc. Ex$. Biol. 25, 569. Bennett, M. V. L. (1968). I n “The Central Nervous System and Fish Behavior” (D. J. Ingle, ed.), p. 14;. Univ. of Chicago Press, Chicago, Illinois. Bennett, M. V. L. (1970). A n n u . R e v . Physlol. 32, 471. Bennett, M. V. L. (1971a). Fish Physiol. 5 , 347. Bennett, M. V. L. (1971b). Fish Physiol. 5, 493. Bennett, M. V. L., Nakajima, Y . , and Pappas, G. D. (1967a). /. Neurophysiol. 30, 209. Bennett, M. V. L., Pappas, G. D., Gimenez, M., and Nakajima, Y. (196713). J . Neurophysiol. 30, 236. Bishop, G. H. (1966). I n “Pain” (R. S . Knighton and P. R. Dumke, eds.), p. 83. Churchill, London.
DESIGN PRINCIPLES OF AXONS
37
Bishop, G. H., and Smith, J. M. (1964). Exp. Neurol. 9, 483. Bishop, G. H., Clare, M. H., and I,andau, W. M. (1969). E x @ . Neurol. 24, 386. Bittner, G. D. (1968). J. Gen. Physiol. 51, 731. Blum, B. (1972). Kybernetik 11, 170. Bodian, D. (1951). J . Comp. Neurol. 94, 475. Bodian, D., and Taylor, N. (1963). Science 139, 330. Braitenberg, V. ( 1967). Progr. Brain Res. 25, 334. Bullock, T. H., and Turner, R. S. (1950). J . Cell. Comp. Physiol. 36, 59. Chang, H.-T. (1952). Cold Sfiring Harbor Symp. Quant. Biol. 17, 189. Chung, S., Raymond, S. A., and Lettvin, J. Y. (1970). Brain Behauior, B Evolution, 3. 72. Coppin, C. M. I>., and Jack, J. J. B. (1972). J. Physiol. ( L o n d o n ) 222, 91P. Cragg, B. G., and Thomas, P. K. (1964). J . Physiol. ( L o n d o n ) 171, 164. Davis, F. A. (1972). J. Neurol., Neurosurg. Psychiat. 35, 537. Davis, F. A,, and Jacobson, S. ( 1 97 1 ) . J. h’eurol., Neurosurg. Psychiat. 27, 106. Dodge, F. A,, and Frankenhauser, B. (1959). J . Physiol. ( L o n d o n ) 148, 188. Dudel, J., and Kuffler, S. W. (1961). J. Physiol. ( L o n d o n ) 155, 543. Dun, F. T. (1970). IEEE Trans. Bio-Med. Eng. 17, 21. Efron, R. (1963a). Brain 86, 261. Efron, R. (1963b). Brain 86, 285. Efron, R. ( 1 9 6 3 ~ )Brain . 86, 295. Efron, R. (1963d). Brain 86, 403. Elfvin, I,. G. (1961). J. Ultrastruct. Res. 5, 374. Endo, K., Araki, T., and Yagi, N. (1973). Brain Res. 57, 484. Erlanger, J., and Gasser, H. S. (1973). “Electrical Signs of Nervous Activity.” Univ. of Pennsylvania Press, Philadelphia. Fraher, J. P. (1972). J . Anat. 112, 99. Frank, K. (1959). I R E Trans. il4ed. Electron. 6, 85. Freeman, J. A. ( 1969). I n “Neurobiology of Cerebellar Evolution and Development” ( R . Llinas, ed.), p. 397. Amer. Med. Ass., Chicago, Illinois. Freeman, J. A., and Nicholson, C. N. (1970). Nature ( L o n d o n ) 226, 640. Friede, R. L., and Samorajski, T. (1967). J. Comp. Neurol. 130, 223. Furukawa, T., Fukami, Y., and Asada, Y. (1963). J. Neurophysiol. 26, 759. Gasser, H. S. (1950). J . Gen. Physiol. 127, 393. Gasser, H. S., and Grundfest, H. (1939). Amer. J . Physiol. 127, 393. Gobel, S., and Purvis, M. B. (1972). Brain Res. 48, 27. Goldman, L., and Albus, J. S. (1968). Biophys. 1.8, 596. Gray, E. G. (1970). Comp. Biochenz. Physiol. 36, 419. Grossman, Y., Spira, M. E., and Parnas, I. ( 1974). Isr. J . M e d . Sci. (in press). Gutrecht, J. A,, and Dyck, P. J. (1970). J . Comp. Neurol. 138, 117. Hardy, W. L. (1971). Biophys. Soc. Abstr. p. 238a. Harris, D. A,, and Stark, I,. (1971). I E E E Trans. Syst. M a n &? Cybernetics 1, 67. Harris, D. A,, and Stark, L. (1973). Brain Res. 51, 340. Harrison, B. M., McDonald, W. I., Ochoa, J., and Ohlrich, G. D. (1972). 1. Neurol. Sci. 16, 489. Haug, H. (1967). Z . Zellforsch. Mikrosk. Anat. 83, 265. Hess, A., and Young, J. Z. (1952). Proc. Roy. Sac., Ser. B 140, 301. Hildebrand, C. (1972). J . Neurocytol. 1, 223. Hildebrand, C., and Skogland, S. (1971). Acta Physiol. Scand., Suppl. 364, 5.
38
STEPHEN G . WAXMAN
Hille, B. (1970). Progr. Biophys. Mol. B i d . 21, 1. Hodgkin, A. L. (1964). “The Conduction of the Nervous Impulse.” Thomas, Springfield, Illinois. Horcholle-Bossavit, G., and Tyc-Dumont, S. (1969). Exp. Brain Res. 8, 201. Hursh, J. B. (1939). A n e r . J . Physiol. 127, 131. Huxley, A. F., and Stampfli, R. (1949). J . Physiol. ( L o n d o n ) 108, 315. Ito, M., and Takahashi, I. (1960). I n “Electrical Activity of Single Cells” ( Y . Katsuki, ed.), p. 159. Ikagu Shoin Ltd., Tokyo. Kahn, L. B. (1971). Comp. Biochen. Physiol. 40, 1. Katz, B., and Schmitt, 0. H. (1940). J. Physiol. ( L o n d o n ) 97, 471. Khattab, F. I. (1966). Anat. Rec. 156, 91. Koles, Z. J., and Rasminsky, M. (1972). J . Physiol. ( L o n d o n ) 227, 351. Korn, H., Sotelo, C., and Crepel, F. (1973). Exp. Brain Res. 16, 255. Kriebel, M. E., Bennett, M. V. L., Waxman, S. G., and Pappas, G. D. (1969). Science 166, 520. Krnjevic, K., and Miledi, R. (1959). J . Physiol. ( L o n d o n ) 149, 1. Kuno, M. (1964). J . Physiol. ( L o n d o n ) 175, 81. Lashley, K. S. (1951). I n “Cerebral Mechanisms in Behavior” (L. A. Jeffries, ed.), p. 112. Wiley, New York. Lehmann, H. J., Lehmann, G., and Tackmann, W. (1971). Z. Neurol. 199, 67. Livingston, R. B., Pfenninger, K., Moor, H., and Akert, K. (1973). Brain Res. 58, 1. Lorente de No, R. (1953). I n “The Spinal Cord” (J. L. Malcolm, J. A. B. Gray, and G. E. W. Wolstenholme, eds.), p. 132. Little, Brown, Boston, Massachusetts. Lubinska, L. (1958). Acta Biol. Exp. ( Warsawa) 18, 117. Lurie, M. ( 1973). Doctoral Dissertation, Massachusetts Institute of Technology, Cambridge. McCulloch, W. S., and Pitts, W. (1943). Bull. Math. Biophys. 5, 115. McDonald, W. I., and Sears, T. A. ( 1970). Brain 93, 583. MacKay, D. M. (1954). Synthese 9, 182. Marrazzi, A. S., and Lorente de No, R. (1944). J. Neurophysiol. 7, 83. Matthews, M. A. (1968). Anat. Rec. 161, 337. Matthews, M. A,, and Duncan, D. (1971).J . Comp. Neurol. 142, 1. Mayer, R. F., and Denny-Brown, D. (1964). Neurology 14, 714. Merrill, E. G., and Wall, P. D. (1972). J . Physiol. ( L o n d o n ) 226, 825. Meszler, R. M., and Bennett, M. V. L. (1972). Anat. Rec. 172, 367. Meszler, R. M., Pappas, G. D., and Bennett, M. V. L. (1972). Brain Res. 36, 412. Meszler, R. M., Pappas, G. D., and Bennett, M. V. L. (1974). J . Neurocytol. 3, 251. Metuzals, J. (1965). Z . Zellforsch. Mikrosk. Anat. 65, 719. Paintal, A. S. (1966). J . Physiol. ( L o n d o n ) 184, 791. Palay, S. L., Sotelo, C., Peters, A., and Orkand, P. M. (1960). J . Cell Biol. 38, 193. Pappas, G. D., and Bennett, M. V. L. (1966). Ann. N . Y . Acad. Sci. 137,495. Pappas, G. D., and Waxman, S. G. (1972). I n “The Structure and Function of Synapses” (G. D. Pappas and D. P. Purpura, eds.), p. 1. Raven Press, New York. Pappas, G. D., Waxman, S. G., and Bennett, M. V. L. (1975). J . Neurocytol. 4 (in press). Parnas, I. (1972). J . Neurophysiol. 35, 903.
DESIGN PRINCIPLES OF AXONS
39
Parnas, I., Spira, M. E., Werman, R., and Bergmann, F. (1969). J . Exp. Biol. 50, 635. Parnas, I., Hochstein, S., Parnas, H., and Spira, M. (1973). Insr. J . Med. Sci. 9, 681. Peters, A. (1966). Quart. J. Exp. Physiol. Cog. M e d . Sci. 51, 229. Peters, A,, Proskauer, C. C., and Kaiserman-Abramof, I. R. (1968). J . Cell Bid. 39, 604. Pickard, W. F. (1969). M a t h . Riosci. 2, 111. Pumphrey, R. J., and Young, J. Z. (1938). J . Exp. B i d . 15, 453. Purpura, D. P. (1967). I n “The Neurosciences: A Study Program” ( G . C. Quarton, T . Melnechuk, and F. 0. Schmitt, eds.), p. 372. Rockefeller Univ. Press, New York. Purpura, D. P. (1971). I n “Handbook of Electroencephalography and Clinical Neurophysiology” (A. Remond, e d . ) , Vol. I, Part B, p. IB2. Elsevier, Amsterdam. Rall, W., Burke, R. E., Smith, T. G., Nelson, P. G., and Frank, K. (1967). J. Neurophysiol. 30, 1169. Rasminsky, M. (1973). Arch. Neurol. (Chicago) 28, 287. Rasminsky, M., and Sears, T. A. (1972). J. Physiol. ( L o n d o n ) 227, 323. Ratliff, F., Hartline, H. K., and Lange, D. (1968). Proc. Nut. Acad. Sci. U S . 60, 464. Raymond, S. A,, and Lettvin, J. Y. (1969). Mass. Inst. Technol., Res. Lab. Electron. Quart. P r o g r . Rep. 92, 431. Renshaw, B. (1946). Amer. J . Physiol. 146, 443. Robertson, J. D. (1959). Z. Zellforsch. Mikrosk. Anat. 50, 553. Rushton, W. A. H. (1951). J . Physiol. ( L o n d o n ) 115, 101. Sanders, F. K. (1948). Proc. Roy. SOL.,Ser B 135,323. Scheibel, M. E., and Scheibel, A. R. (1958). In “Reticular Formation of the Brain” ( H . Jasper et al., eds.), p. 31. Little, Brown, Boston, Massachusetts. Scheibel, M. E. and Scheibel, A. B. (1970). I n “The Neurosciences: Second Study Program” (F. 0. Schmitt, ed.), p. 443. Rockefeller Univ. Press, New York. Schnepp, P., and Schnepp, G. (1971). Z. Zellforsch. Mikrosk. Anat. 119, 99. Schwartz, I. R. ( 1968). Doctoral Dissertation, Yale University, New Haven, Connecticut. Schwartz, I. R., and Pappas, G. D. (1968). Anat. Rec. 160, 424. Sotelo, C., and Palay, S. L. (1970). Brain Res. 18, 93. Spira, M. E., Parnas, I., and Bergmann, F. ( 1969). J . Exp. B i d . 50, 615. Spira, M. E., Castel, M., and Parnas, I. (1974). 1 5 7 . J . M e d . Sci. (in press). Stampfli, R. (1954). Physiol. Rev. 34, 101. Sunderland, S., and Roche, A. E. (1958). Acta Anat. 33, 1. Tarlov, I. M. (1937). Arch. Neurol. Psychiat. 37, 555. Tasaki, 1. (1959). I n “Handbook of Physiology” (Amer. Physiol. SOC., J. Field, ed.), Sect. 1, Vol. I, p. 75. Williams & Wilkins, Baltimore, Maryland. Tasaki, I., and Takeuchi, T. (1941). Pfluegers Arch. Gesamte Physiol. Menschen Tiere 244, 696. Tasaki, I., Ishii, K., and Ito, H. (1943). J u p . /. M e d . Sci. 3 9, 189. Tauc, L. (1960). J . Physiol. ( L o n d o n ) 152, 36P. Tauc, T., and Hughes, G . M. (1963). J . Gen. Physiol. 46, 533. Thomas, P. K., and Young, J. Z. (1949). J . Anat. 83, 336. Uttley, A. M. (1954). Electroencephalogr. Clin. Neurophysiol. 6, 479. Van Essen, D. C. (1973). J . Physiol. ( L o n d o n ) 230, 509.
40
STEPHEN G . WAXMAN
Vizoso, A. D., and Young, J. Z. ( 1948). J. Anat. 82, 110. von Schwarzacher, H. (1954). A c t a . Anat. 21, 26. Wall, P. D. (1958). J. Physiol. ( L o n d o n ) 142, 1. Wall, P. D., Lettvin, J. Y., McCulloch, W. S., and Pitts, W. H. (1956). Syrnp. Inform. T h e o r y Biol., p. 329. Waxman, S. G. (1970). Nature ( L o n d o n ) 227, 283. Waxman, S. G. (1971). Brain Res. 27, 189. Waxman, S. G. (1972). Brain Res. 47, 269. Waxman, S. G. (1973). J . Neurol. Sci. 19, 357. Waxman, S. G. (1974). Brain Res. 65, 338. Waxman, S. G. (1975). J . Neurol. Sci. (in press). Waxman, S. G., and Bennett, M. V. L. (1970). J. Cell B i d . 47, 222a. Waxman, S. G., and Bennett, M. V. I,. (1972). Xature ( L o n d o n ) , New Bid. 238, 217. Waxman, S. G., and Melker, R. J. (1971). Brain Res. 32, 445. Waxman, S. G., and Pappas, G. D. (1971). J . Conzp. h’eurol. 143, 41. Waxman, S. G., Pappas, G. D., and Bennett, M. V. L. (1972). J. Cell B i d . 53, 210. Williams, P. L., and Wendell-Smith, C. P. (1971). J. Anat. 109, 505. Witkovsky, P. (1971). J. Comp. h'eurol. 142, 205. Yarorn, Y., Spira, M. E., and Parnas, I. (1973). Zsr. J . M e d . Sci. 9, 680. Yasargil, G. M., and Diamond, J. (1968). Nature ( L o n d o n ) 220, 241.
BIOLOGICAL TRANSMETHYLATION INVOLVING S-ADENOSYLMETHIONINE: DEVELOPMENT OF ASSAY METHODS AND IMPLICATIONS FOR NEUROPSYCHIATRY' By Ross J. Baldessarini'
Psychiatric Research Laboratories, General Hospital, and Department of Psychiatry, Harvard Medical School, Boston, Massachusetts
I. Introduction
.
11. Biochemical Assays for the Study of Transmethylation: Assays of the
Methyl Donor . A. Assay of S-Adenosylmethionine . B. Turnover of S-Adenosylmethionine . C. Effect of Substrate Supply and Increased Utilization on Levels of S-Adenosylmethionine D. Metabolic Effects of Methionine Loading . E, Is S-Adenosylmethionine the Only Methyl Donor?: The Case of Methyl Tetrahydrofolate 111. Other Assays Related to Transmethylation A. Assays of Methyl Acceptors: N-Acetykerotonin and Histamine. B. Assay of ATP:CMethionine Adenosyltransferase . C. Assay of Methionine . IV. Clinical Implications: Need for New Strategies for Clinical Metabolic Research in Schizophrenia . References .
. .
.
.
41
. . .
44 44 46
.
47 51
.
55 57 57 59 59
. .
. . . .
.
61 63
I. Introduction
For many years, there has been considerable interest in the chemistry and pharmacology of biological transmethylation in the field of neuropsychiatry. This interest was largely stimulated by the fact that many natural Based in part on a chapter to be published in Italian in: Transmetilationi S A M e Dipendenti nel Sistema Nervoso Centrale: Ruolo nei Disturbi del Cornportament (Frazio, C . , E d . ) , Tamburini Editore, Milano, 1975. ' Supported in part by USPHS ( N I M H ) Research Grant MH-16674 and Career Development Award MH-47370. 41
42
ROSS J . BALDESSARINI
or synthetic substances which produce hallucinations, or other reactions that also occur in psychotic illness, are methylated amines (see Baldessarini, 1966a). As early as 1952, Osmond and Smythies (1952) reported a suggestion of the biochemist Harley-Mason that abnormal transmethylation of an endogenous amine, possibly dopamine, might produce a psychotomimetic compound like mescaline (3,4,5-trimethoxyphenethylamine) . More direct evidence consistent with this hypothesis was the observation that methionine, uniquely among several amino acids, and especially when combined with a n inhibitor of monoamine oxidase, led to striking but transient exacerbations of the psychotic symptoms of chronic schizophrenic patients (Pollin et al., 1961). This clinical phenomenon is perhaps the only biochemical finding in schizophrenia that has been confirmed by several groups and so far contradicted by none (Brune and Himwich, 1962; Alexander et al., 1963; Haydu et al., 1965; Park et al., 1965; Kakimoto et al., 1967; Spaide et al., 1968; Ban, 1969; Antun et al., 1971a; see also Coper et al., 1972; Cohen et al., 1974). Moreover, a similar result was obtained with betaine, another substance capable of contributing a methyl group to intermediary metabolism in mammalian tissues (Brune and Himwich, 1963). Also, an unconfirmed report, which is not easily interpreted, is that methionine sulfoximine, a metabolic antagonist of niethionine, may have had beneficial effects in a small number of schizophrenics (Heath et al., 1966). Another observation is that schizophrenia-like psychoses appear in unexpectedly high frequencies in patients with homocystinuria and in their relatives, and this inborn error is usually associated with a deficiency of cystathionine synthetase and high circulating levels of niethionine (Carey et al., 1968; Freeman et al., 1975), although it was recently reported that homocystinuria can be associated with a deficiency of methylenetetrahydrofolate reductase resulting in increased homocysteine levels in blood and urine, with normal levels of methionine, but with psychosis and mental retardation (Freeman et al., 1975). The latter observation suggests that psychosis in homocystinuria may be unrelated to increased tissue levels of methionine ; it may also be unrelated to increased levels of homocysteine since relatively few homocystinurics become psychotic. There have also been repeated suggestions that there may be unusual methylated phenylethylamines (Friedhoff and Van Winkle, 1962) or indoleamines in the urine of schizophrenic patients (see Fischer and Spatz, 1970; Tanamukai et al., 1970; Narasimhachari and Himwich, 1973a,b). O n the other hand, the significance of the latter findings has been questioned or not supported by several recent studies (Creveling and Daly, 1967; Heslinga et al., 1970; Sharma and Sinari, 1971; Narasimhachari e t al., 1972; Wyatt et al., 1972,1973a; Lipinski et al., 1974). Nevertheless, reports of abnormal excretion of possibly psychodysleptic N-methylated tryptamines have contin-
BIOLOGICAL TRANSMETHYLATION
43
ued to appear, even with the application of less ambiguous analytical methods (Tanamukai et al., 1970; Narasimhachari and Himwich, 1973a,b). These findings are particularly interesting in light of the observation that the hallucinogen N,N-dimethyltryptamine may not produce tolerance to its behavioral effects in the cat (Gillen et al., 1973), and it should not if it is an endogenous toxin that contributes to the appearance of chronic psychosis in man. There is also an unconfirmed report of uncertain significance that the ability of blood samples froin schizophrenic patients to support the methylation of nicotinamide may be higher than normal (Buscaino et al., 1969), as well as an observation of clinical worsening of schizophrenics upon injection of a preparation of a plant extract containing catechol O-methyltransferase ( C O M T ) activity (Hall et al., 1969), possibly on the basis of toxicity of the material given. Although there have also been suggestions that antipsychotic drugs may inhibit a variety of methyltransferase reactions (Salvador and Burton, 1965; Antun et al., 1971b; Hartley et al., 1972; Narasimhachari and Lin, 1974), these effects on arnine methylation are weak and of dubious functional significance. Another report has suggested that the decarboxylation of labeled dihydroxyphenylalanine (dopa) to dopamine by erythrocytes of schizophrenic patients may be more active than normal (Tran-Manh et al., 1972 ) , thereby possibly increasing the availability of an aromatic amine capable of accepting methyl groups. Interest in the possibility that transmethylation niight be abnormal in schizophrenia has also been stimulated recently by reports that the activity of monoamine oxidase ( M A O ) may be decreased in the blood platelets of schizophrenic patients (Murphy and Wyatt, 1972; Wyatt et al., 197313; Meltzer and Stahl, 1974), although apparently not in their brains (Domino et al., 1973; Schwartz et al., 1974), nor has it consistently been found decreased even in platelets (Friedman et al., 1974). l h e r e is also a recent observation that a methyltransferase dependent on S-adenosylmethionine ( SAMe) may be more active in blood platelets of schizophrenic patients than of comparison subjects, possibly owing to the drcreased availability of a dialyzable inhibitor of the enzyme in schizophrenics (Wyatt et al., 1 9 7 3 ~ )This . enzyme appears to be similar to the nonspecific N-methyltransferase that occurs in many tissues along with a dialyzable inhibitor (Saavedra et al., 1973b) ; it has even been reported to occur in low activity in human brain tissue (Mandell and Morgan, 1971 ; see also Rhikharidas r t al., 1975), although it is probably not increased in activity in the brains of schizophrenics (Domino et al., 1973). In the affective disorders, there is also considerable, though somewhat inconsistent evidence to suggest that there may be an abnormality of amine metabolism (see Baldessarini, 1975), including an unconfirmed report of decreased activity of erythrocyte C O M T in depressed women (Cohn et al., 1970). There is also a preliminary report of abnormal metabolism of methio-
44
ROSS J . BALDESSARINI
nine in schizophrenic and depressive states as estimated by the rate of appearance of radioactive C O , in the breath following intravrnous injection of "C-methyl-labeled rnethionine (Israelstam et al., 1970) . Recently there have been preliminary studies suggesting that injections of SAMe may be of therapeutic benefit to depressed patients by an uncertain mechanism (Fazio et al., 1973). The weight of these several obscrvations has supported the idea that studies of transinethylation of biogenic amines in the major mental illnesses might be of some iniportance in attempting to understand their pathophysiology, and possibly to gain insights into their causes and more effective treatment. II. Biochemical Assays for the Study of Transmethylation: Assays of the Methyl Donor
A. ASSAYOF S-ADENOSYLMETHIONINP: The observations relating to the unique exacerbation of psychosis when patients were treated with niethionine (or betaine), with or without an inhibitor of MAO, but not with other amino acids, strongly suggested that methionine might be acting by donating methyl groups after its conversion with ATE' by methionine adenosyltransferase to the important methyl donor, S-adenosylmethionine (SAMe) . Some aspects of this topic have been approached by studies of the physiological chemistry of SAMe. An initial problem was the requirement of a sensitive and specific assay for tissue levels of this methyl donor. One approach to this problem resulted in the development of a double-isotopic enzymic assay for SAMe (Baldessarini and Kopin, 1963, 1966; Kopin and Baldessarini, 1971 ) . The basic principle involved is the isotope dilution of radioactive SAMe with the endogenous compound present in acid extracts of the tissue, and estimation of the specific radioactivity of the diluted SAMe by the enzymic formation of melatonin from the methyl donor and N-acetylserotonin. The specificity of the assay depends on the selectivity of the enzyme hydroxyindole 0-methyltransferase (HIOMT) for SAMe as methyl donor and the absence of appreciable amounts of N-acetylserotonin in most tissues (with the notable exception of the pineal gland). The assay could be conducted with only methyl-labeled SAMe, but preliminary experiments revealed that the efficiency of production of melatonin was low and somewhat variable, and failed to yield a linear relationship between the amount of unlabeled SAMe present and the amount of melatonin produced. Thus, in order to monitor the efficiency of the production of melatonin, a second label was introduced in the acetyl group of the cosubstrate, N-acetylserotonin. Ordinarily, [3H-acetyl]N-acetylserotonin and [I'C-rnethyl1SAMe are used, largely so as to take advantage
BIOLOGICAL TKAN SMETHYLATION
45
of the relative chemical stability of the l'C-labeled SAMe. However, when assays of relatively low concentrations of SAMe are required, as in blood specimens, it is advantageous to increase the sensitivity by reversing the labels and to use tritiated SAMe of high specific radioactivity and I'C-labeled N-acetylserotonin (Matthysse and Baldessarini, 1972), I t can be predicted mathematically that the ratio of the two labels in the recovered melatonin should be linearly related to the amount of unlabeled SAMe present, and this prediction has been verified experimentally (Baldessarini and Kopin, 1966). More recently, the principle of this assay has been applied in a chromatographic assay lor SAMe, which is elegant in its simplicity (Salvatore et al., 1971). In the chromatographic assay, again radioactive SAMe is added to acid homogenates of tissue to establish the specific radioactivity, and SAMe is recovered by I)o\vex-Na+ ion-exchange chromatography ; the specific activity of SAMe in the sulfuric acid elutates as estimated by counting and by spectrophotometric assay of adenine compounds is proportional to the endogenous SAMe. Estimates of tissue levels of SAMe by this method agree quite well with those provided by the enzymic method, although they are generally somewhat lower (as much as 50%), possibly owing to greater purity of the authentic SAMe used to establish standard curves for the assays. The materials required for the enzymic assay of SAMe include partially purified methylstransferase enzyme ( H I O M T ) prepared from beef pineal gland, which is available from commercial sources. The methyl acceptor, N-acetylserotonin, is easily and quickly prepared by allowing serotonin to react with radioactive acetic anhydride in a mildly alkaline medium, and separating the products by preparative paper chromatography. SAMe, either unlabeled or radioactively labeled with ''C or is also readily available commercially. The tissue is extracted with trichloroacetic acid, and the labeled SAMe can be introduced directly into the homogenates to avoid problems of recovery or losses of the endogenous SAMe by establishing the specific radioactivity of the SAMe immediately. Even without this precaution, the recovery of authentic SAMe is virtually quantitative (>%c/o. The samples can then be frozen and assayed later at one's convenience. Large numbers of samples can be handled easily at one time. The materials can be prepared at one time and kept frozen, and they are stable for many months. When the SAMe preparations, methyl acceptor and H I O M T are allowed to react, the product, doubly labeled melatonin, is recovered by extraction into chloroform ; the organic pliase is washed with NaOH solution, and then counted for 3H and 'Y:. Quantitative recovery of the product is not required since the assay depends merely on the ratio of the two labels, and it is important only to recover sufficient melatonin for counting and to be certain that melatonin is the only labeled molecule recovered. The
46
ROSS J . BALDESSARINI
authenticity of the recovered product was verified by chromatography in several solvent systems with authentic melatonin. Furthermore, it was shown that negligible radioactivity was recovered by incubation of the methyl acceptor and methyl donor with tissue extracts in the absence of H I O M T , or incubation of labeled SAMe with tissue extracts and H I O M T . Thus, the tissue extracts do not have any significant amounts of H I O M T activity or of substances that accept methyl groups and are extracted into chloroform under the conditions of the assay; moreover, the contribution of methyl acceptors by the partially purified and dialyzed H I O M T preparation is also insignificant. Of several potential methyl donors, only SAMe was found to yield melatonin under conditions of the assay, although it appeared that S-adenosylethionine (not normally present in tissue, but found after treatment with high doses of ethionine) can transfer its ethyl group to N-acetylserotonin in the presence of H I O M T . The method is capable of detecting as little as 500 pmoles of SAMe, when [WISAMe is used, and the use of [3H]SAMe increases the sensitivity by about an order of magnitude. The precision of the assay is very high. Measurable quantities of endogenous SAMe were detected in all tissues examined (Table I ) , the highest levels being found in the adrenal and pineal glands. Most tissues contained 10-50 pg per gram of wet tissue, while blood or serum contained 0.5-1.0 pglml. Brain tissue contained about 10-15 pglgrn, with no impressive regional distribution. The concentrations of SAMe in brain and liver tended to fall as a function of age in rats. These values may all be slightly high since comrncrcially available authentic SAMe was used as a standard without further repurification and is now known to be <90% pure.
B. TURNOVER OF S-ADENOSYLMETHIONINE It is also possible to utilize this assay to estimate the turnover of SAMe in tissues in vivo by introducing labeled L-methionine by intravenous administration and allowing the tissues to generate labeled SAMe. The assay can then be used to estimate the specific activity of SAMe over time, and this activity will change as a function of the rate of synthesis and utilization of endogenous SAMe (Baldessarini and Kopin, 1966). In this way., it was found that the rate of production of labeled SAMe in rat liver was extremely rapid: labeled SAMe was detected in less than 5 minutes and its half-life was on the order of 10 minutes; brain was found to produce SAMe somewhat less rapidly and to consume it much more slowly than liver, and this difference correlates well with the much lower activity of methionine adenosyltransferase in brain (Baldessarini and Kopin, 1966; Matthysse et d.,
1972).
47
BIOLOGICAI. TRANSMETHYLATION
TABLE 1
TISSUE CONCENTRATIONS OF S-ADENOSYLMETHIONINE SAME)^
Adrenal Pineal (beef) Liver Heart Spleen Kidney Lung Brain Whole Blood (human) Leucocytes (human) Lymphocytes (human) Serum
48 38c 29 26 24 20 11 11 1 .o 1.5 2.2 0.5
200
158 121 108 100 83.3 45.8 45.8 3.3 6.3 9.1 1.2
“ D a t a are mean values for N = 3 to 20 assays with rat tissues, except as noted otherwise and are summarized from Baldessarini and Kopin (1963, 1966), Baldessarini and Carbone (1965) (Copyright 1965 by the American Association for the Advancement of Science) and Matthysse and Baldessarini (1972), 128, 1310-1312. Copyright 1972, the American Psychiatric Association. An approximate value based on assumption that SAMe distributes in tissue water (60% of wet weight). C A n approximate value not corrected for the effect of endogenous N-acetylserotonin.
C. EFFECTOF SUBSTRATE SUPPLYA N D INCREASED UTILIZATION O N LEVELS OF S-ADENOSYLMETHIONINE Up to this point, the most important conclusion to be drawn from the preliminary survey of tissue concentrations of SAMe was that this important metabolic intermediary accumulates to measurable levels and that it is neither so labile nor so rapidly utilized as to be unmeasurable. Moreover, the rate of synthesis and turnover of SAMe occurs with a time course that can easily be measured in vivo. The next aspects of the metabolism of SAMe to be considered were whether its availability might be enhanced by increased input of methionine, or, conversely, whether its utilization might be increased by increased availability of methyl acceptors. I t was found that the systemic administration of methionine in the rat produced striking increases of SAMe in liver and other peripheral tissues and smaller increases
48
ROSS J . BALDESSARINI
in the brain as well (Baldessarini, 1966b) (Table I I A ) , Ltnd this finding has been confirmed independently in the rabbit (Salvatore et al., 1971) and the rat (Rubin et al., 1974; Schatz and Sellinger, 1975). Increased availability of the precursor methionine probably can enhance the rate of production of SAMe, since the normal tissue levels of methionine (50-90 nmoles/gm, or about 80-150 p M ) (Lombardini et al., 1971; Rubin et al., 1974) are close to the K , for methionine (about 90 p M ) (Matthysse et al., 1972) for methionine adenosyltransferase, which is thus probably not normally saturated with substrate. These observations were thus consistent with the hypothesis that the administration of methionine to psychiatric patients might increase the availability of SAMe for amine methyltransferase reactions. It was also noted, when end-to-side portacaval venous anastomoses were made surgically in the rat, that there was a marked decrease in the concentration and total amount of SAMe in the liver (Baldessarini and Fischer, 1967). This effect was apparently not due to nonspecific toxic effects on the liver, since the activity of methionine adenosyltransferase was not
TABLE I1 TISSUE CONTENTOF S-ADENOSYLMETHIONINE (SAME) 60 MINUTESAFTER TREATMENT WITH METHIONINE (MEANCONTENT t SEM)a
A. Tissues of intact rats (SAMe, pg/gm)
Tissue Liver Heart Brain
Control 26.0 25.7 11.5
k 1.0 k 6.2 0.5
Methionine-treated (100 mg/kg, i.p.) 115.0 t 15.0 42.8 f 0 . 6 15.0 rt 0.6
B. Liver a t 6-8 weeks after end-to-side portcaval venous anastomosis (SAMe, pg/liver)b
Condition
Control
Sham-operated Porta-caval shunted
330 k 15 111
It
14
Methionine-treated (200 mg/kg, i.p.) 875 775
+ 75 k 75
Data from Baldessarini (196613) and Baldessarini and Fischer Surg. 62, 311-318 (1967).
* Concentrations are expressed per organ since liver weight was slightly decreased after shunting. For all increasesp < 0.01, N 2 6.
BIOLOGICAL TRANSMETHYLATION
49
deficient in the “shunted” livers, and furthermore, when more of the substrate methionine was made available by systemic administration, the shunted rats were able to produce nearly as much hepatic SAMe as their “sham-operated” controls (Table IIB) . Thus, it appeared likely that the production of SAMe in liver was highly dependent on the availability of the precursor L-methionine. Other experiments have also provided more indirect evidence that is consistent with the same conclusions. For example, it was found that human leukocytes from patients with chronic myelocytic leukemia have striking elevations of SAMe levels (Baldessarini and Carbone, 1965), while the activity of methionine adenosyltransferase was not different from that of normal white blood cells (Baldessarini and Bell, 1966). It was later reported that similar leukemic white blood cells had greater-than-normal uptake of labeled methionine in vitro (Sloane and Bridges, 1968). These findings taken together suggest that the increased SAMe levels in leukemic cells might be due to increased availability of the precursor, methionine. More recently, it was noted that SAMe levels in rat brain (Wurtman et al., 1970; Chalmers et al., 1971) (Table 111) and in human blood (Matthysse et al., 1971) are depleted by large doses of L-dopa, which is rapidly methylated by catechol 0-methyltransferasc ( C O M T ) with SAMe. One of the findings in these experiments was that, several hours after a dose of L-dopa, the reduced levels of SAMe in the rat brain not only had returned to normal, but were actually 30-40% higher than normal (Chalmers et al., 1971). A possible explanation for this change follows from the observation that 3-0-methyl-dopa (3-methoxytyrosine) accumulates in brain and disappears with a time course strikingly similar to the observed increase of SAMe (Bartholini and Pletscher, 1970). Moreover, when isolated nerve endings were preloaded with this methylated derivative of L-dopa, the uptake of labeled niethionine was markedly enhanced, suggesting that methionine can be taken up in exchange for other amino acids (Baldessarini and Karobath, 1972). This phenomenon might provide for increased availability of methionine to cells containing the adenosyltransferase and might at least partly explain the increased levels of SAMe several hours after an acute dose of 1.-dopa. In addition to the %Me-depleting effects of L-dopa, it had been noted earlier that a variety of polyphenolic substrates for C O M T can lead to the depletion of tissue levels of SAMe (Baldessarini, 1966b) (Table 111).These compounds evidently are able to decrease tissue levels of SAMe by increasing its utilization beyond the capacity of synthesis to keep pace with the demand for new molecules of the methyl donor. The effect of L-dopa and of polyphenols was more striking in the brain, a tissue with relatively less ability to produce SAMe, than in the liver, although even in liver, large doses of
50
ROSS J . BALDESSARINI
these agents or their chronic administration did lead to the partial depletion of hepatic SAMe (Table 111). These results were not associated with decreased activity of methionine adenosyltransferase in the case of one polyphenolic substance, pyrogallol (Baldessarini, 1966b) . Thus, these findings indicate that the availability of SAMe is dependent not only on the availability of methionine, but also on the demands of methyl acceptors for methyl groups. I t also follows that the rate of production of methylated metabolites might increase after large doses of methionine, as required by the hypothesis that methionine produces exacerbations of psychosis in schizophrenic patients by increasing the availability of certain methylated products, possibly including psychotogenic amines. An interesting negative observation among these experiments was that even a large dose (100 mg/kg, i.p.) of nicotinamide failed to alter hepatic concentrations of SAMe, and there was no effect on liver or brain SAMe even when nicotinamide was given intravenously (Table 111) (Baldessarini, 1966b). The failure of exogenous nicotinamide to decrease tissue levels of SAMe may reflect the observation that its normal tissue concentrations of about 10 pM (Chaudhuri and Kodicek, 1949) are already about 100 times above the apparent K , of nicotinamide N-methyltransferase (Salvador and
TABLE 111 EFFECTSOF METHYLACCEPTORS ON LEVELSOF S-ADENOSYLMETHIONINE SAME)^ SAMe (% of control f SEM) Substance* L-Dopa (acute)
L-Dopa (chronic) Pyrogallol Purpurogallin Nicotinamide
Dose (mg/kg, i.p.1
10 30 100 100 100 100 100
-
Liver
Brain
-
95.2 f 6 . 5 63.6 f 1.8" 32.7 k 2.4c 1 6 . 7 2.0" 25.0 k 5.0" 70.7 +_ 4 . O C 128.0 f 1 . 0 (i.v.)
108.0 f 9 . 5 78.8 f 6 . 1 c 30.4 f 1 . 8 c 35.3 f 8.OC 111.0 rt 1 . 1
a Data from Baldessarini (1966b), Wurtman et al. (1970) (Copyright 1970 by the American Association for the Advancement of Science) and Chalmers et al. (1971). Drugs were given 30-60 minutes before sacrifice after a n acute dose or the last dose of chronic treatment for 10 days. The data on brain after nicotinamide involved intravenous infusion of the agent (100 mg/kg) over 50 minutes and sacrifice a t 60 minutes: N = 3 (R. J. Baldessarini, unpublished data, 1962). ' p < 0.05 or less for N 2 6 rats.
BIOLOGICAL TRANSMETHYLATION
51
Burton, 1965). This lack of an SAMe-depleting action of nicotinamide, which is known to be methylated by SAMe (Cantoni, 1951), is of some interest in light of the highly controversial proposal that nicotinamide or nicotinic acid might be of value in the treatment of schizophrenia, and that such an effect might be mediated by the action of nicotinamide as an acceptor of methyl groups (Hoffer and Osmond, 1964; see also Lipton, 1973). There is even evidence that methionine can decrease the excretion of N-methyl nicotinamide (Horwitt et al., 1956; Sprince, 1970). I t is also interesting that large doses of nicotinic acid failed to prevent the exacerbation of psychosis following methionine and a n MAO-inhibitor (Ban, 1969; see also Lipton, 1973).
D. METABOLIC EFFECTSOF METHIONINE LOADING I n order to propose that methionine induces exacerbations in schizophrenia by increasing the production of psychotogenic methylated amines, it was necessary to demonstrate that more SAMe is formed, and helpful to find that its rate of utilization is dependent on the availabiliy of substrates. However, the next crucial question is whether the increased availability of the methyl donor can increase the rate of methylation of methyl acceptors, particularly of metabolites of catechol- or indoleamines. This question has been dealt with only partially and incompletely in preclinical as well as human experiments. I t has been demonstrated in the rat in viuo that intravenously injected 14C-methyl-labeled methionine results in the formation of methyl-labeled catecholamine metabolites such as vanillylmandelic acid (VMA) (Antun et al., 1971b). However, this finding does not clarify the question of whether methylation can be increased with additional methyl donor. We have recently administered labeled L-dopa to rats and mice following large doses of L-methionine or of SAMe itself (R. J. Baldessarini and S. Will, unpublished observations) . These treatments failed to produce substantial increases in the production of methylated products. However, methionine had an apparent inhibitory effect on the uptake of L-dopa into brain (Table IV) . The latter result may help to explain the clinical observation that methionine can antagonize the clinical efficacy of L-dopa in patients with parkinsonism (Pearce and Waterbury, 1974). A rather indirect experiment pertinent to the question of whether increased availability of SAMe can enhance methylation was to study the potency and duration of action of apomorphine ( a catechol-aporphine, and putative direct agonist of central dopamine-receptors) following pretreatment with pyrogallol (to inhibit catechol-0-methylation) or L-methionine or even SAMe. While the COMT-inhibitor potentiated and prolonged the stereotyped behavioral response to apomorphine in the rat, even large doses ( u p to 1000 mg/kg, i.p.) of 1.-methionine or SAMe failed to have
52
ROSS J. BALDESSARINI
TABLE IV EFFECTOF METHIONINE O N UPTAKE AND 0-METHYLATION OF [ 3 H ] ~ - D o ~ ~ " A. Total radioactivity taken up by tissue Total 3H (dpm X
f SEMI
Tissue
Control
L-Methionine
yo of control
Brain Heart
23.1 f 2 . 4 41.6 f 5.6
10.1 f 0 . 6 31.8 f 5 . 8
44 %b 77 %
B. Ratio of methylated metabolites: catechol metabolites of [ 3H]~-dopa Ratio of metabolites f SEM Tissue ~~
Control
L-Methionine
% of control
10.82 f 0.27 1 . 8 6 f 0.38
9.47 k 0.44 1.46 f 0.07
88% 78 %
~
Brain Heart
~
Rats were given 500 mg/kg, i.p., L-methionine at 60 and 30 minutes prior to injection of radiographically pure [G-3H]~-dopa (25 pCi, i.p.). Tissues were removed 30 minutes later, extracted in 0.4 N perchloric acid, counted and fractionated by alumina column chromatography into methylated metabolites (effluent washes) and catechol mztabolites (material eluted with 5 ml of 2 N HCI). The radioactivity in each fraction was calculated as SEM ( N = 9 rats). dpm/gm tissue. Data are means *By t-test = p < 0.01 or less. These mcthods are described i n detail by Baldessarini and Greiner (1973).
+
an antagonistic effect (Table V ) . Since 0-methylation is a major means of inactivating apomorphine (Missala et al., 1973), these observations argue against an appreciable increase in a C O M T pathway with increased availability of SAMe. Another biochemical consideration also casts some doubt on the hypothesis that methionine loads might actually increase the rate of methylation of amines. The data in Table I indicate that the concentration of SAMe, at least in rat tissues, ranges from about 50 to 200 p M . Since the half-maxima1 velocity of most of the known amine-methyltransferases occurs at SAMe concentrations (K,) of about 5-50 pM (Deguchi and Barchas, 1971 ; Baudry et al., 1973; Saavedra et al., 1973b; Laduron et al., 1974), it seems probable that additional SAMe will not force the reactions to an important extent.
53
BIOLOGICAL TKANSMETHYLATION
TABLE V
EFFECT OF METHYLATION O N THE BEHAVIORAL ACTIONSOF APOMORPHINE Condition ( N ) a Control (22) Pyrogallol (100 mg/kg) (6) 500 mg/kg)(12) L-Methionine (500 SAMe (250 250 nig/kg)(6)
+
+
Mean stereotypy score If: SEM % of control 13.1 k 0.5 22.8 k 0 . 6 13.4 k 0 . 9 12.2 f. 1 . 0
100% 174%b 102% 93 %
a The drugs were given intraperitoneally to N rats 30 minutes prior to the subcutaneous injection of the catechol-containing central dopamine receptor agonist apomorphine .HCI (Merck), 5.0 mg/kg. The doses of L-methionine and SAMe (Samyr, donated by Errekappa, Milan) were also repeated at 5 minutes prior to the apomorphine. Control animals were given saline or the vehicle for SAMe, their scores did not vary significantly and so were pooled from several experiments. Stereotyped behavior was evaluated every 15 minutes for 2 hours by a blind observer with a rating-scale technique described by Tarsy and Baldessarini (1974). The effect of pyrogallol, an inhibitor of catechol-0-methyltransferase, capable of depleting endogenous SAMe (Baldessarini and Greiner, 1973), was to increase the maximal response of apomorphine and to prolong its action by 15 minutes, presumably by preventing its inactivation by methylation. * p < 0.001 by t test.
Thus, it would be well to consider alternative metabolic effects of high doses of methionine with which to explain its effects on schizophrenia. One interesting possibility is suggested by the recent observations that S-adenosylhomocysteine, an end product of methyl-transferase reactions utilizing SAMe, exerts a strong competitive inhibitory action on several amine-methyltransferase reactions in micromolar conccntrations (Zappia et al., 1969; Deguchi and Barchas, 1971; Baudry et al., 1973; Lin et al., 1973). Thus it is conceivable that increased levels of SAMe produced by large doses of methionine might alter the ratio of SAMe: S-adenosylhomocysteine to favor the methylation of amines. The possible clinical metabolic significance of the administration of methionine or methyl acceptors is still not well known. I t is quite possible that methionine or one of its metabolites may exert toxic actions in the central nervous system. It is not known whether methionine loads increase levels of homocysteine, and, although psychosis has been associated with a form of homocystinuria without methioninemia, homocysteine is not likely to be the sole toxic metabolite since the occurrence of psychosis in homocystinuria is only occasional (Freeman et al., 1975). The currently available evidence concerning the effects of methionine in patients is generally not supportive of the hypothesis that methionine produces exacerbation of psychosis in schizophrenia by increasing the availability of psychotogenic methylated amines. Thus, in one report, while the urinary excretion of metanephrine and
54
ROSS J . BALDESSARINI
normetanephrine was markedly increased by treatment with an MAO-inhibitor, and apparently increased even somewhat more on the addition of methionine (20 gm) , the effect of methionine was not quite significant statistically (Kakimoto et al., 1967). Also, large doses of methionine plus tryptophan failed to increase the excretion of vanillylmandelic acid (VMA) (Berlet et al., 1965). In another study, when large doses (up to 20 gm) of methionine without an inhibitor of M A 0 were given to schizophrenic patients, there was a clear exacerbation of psychosis, but no increased excretion of the methylated catecholamine metabolites, VMA and methoxyhydroxyphenylethyleneglycol ( M H P G ) (Antun et al., 1971a). A similar failure to find increased excretion of a number of 0- or N-methylated metabolites after 10 gm of methionine has been reported more recently (Coper et al., 1972). There have been suggestions that certain methylated tryptamines might be excreted in abnormally high quantities by schizophrenic patients given large doses of methionine (Spaide et al., 1968) ; however, these findings were obtained with analytical methods of very low specificity, and furthermore, they are inconsistent with other results (Sprince et al., 1963). One of the problems with such studies is that methionine as well as other amino acids, including the precursors of methionine, cysteine and homocysteine, appear to increase the excretion of indoleamines generally, particularly tryptamine, evidently by altering the metabolism of tryptophan (Sprince, 1970). This effect of cysteine has been used to increase the excretion of tryptamines in schizophrenic patients, and the administration of cysteine plus an inhibitor of M A 0 has been reported to increase the excretion of N-methylated tryptamines in such patients (Tanamukai et al., 1970), but again, the significance of these findings is not clear and they were obtained with assay methods that are not unambiguous. T o date, the effects of methionine-loading in schizophrenic and other subjects upon the excretion or blood levels of N-methylated indoleamines, as measured by recently developed and very specific and sensitive gas chromatographic and mass spectrometric methods (Narasimhachari and Himwich, 1973a,b; Wyatt et al., 1973a), has not been reported. There are several other findings that are generally not supportive of the idea that methylation of amines may be abnormal in schizophrenia or other psychoses. For example, mean circulating levels or activities of SAMe, methionine adenosyltransferase, and COMT have all been reported to be statistically within the range of values obtained with normal or other comparison groups (Matthysse and Baldessarini, 1972; Dunner et al., 1973). Nevertheless, the existence of a subgroup of schizophrenics not obviously identified on clinical grounds alone, but havinq an unusual ability to methylate amines remains a possibility (Matthysse and Baldessarini, 1972). When a large number of subjects were challenged with the methyl acceptor protocatechuic
BIOLOGICAL TRANSMETHYLATION
55
acid, there was no difference in the excretion of methylated products by the schizophrenics (Price, 1972), although such tests should be repeated with precursors of the indoleamines and the newer assays for their methylated products, and they should also be repeated with and without methionine loads. In another loading-type experiment, nonschizophrenic subjects given huge doses of amphetamine, failed to excrete the hallucinogenic substance p-methoxyamphetamine, although the authentic exogenous substance could be detected by the methods employed (Schweitzer et al., 1971) ; on the other hand, this finding is not directly relevant to the idiopathic psychoses and even the psychosis lvhich follows large doses of amphetamine need not be the result of a psychotoniimetic metabolite. There has been some hope that substances which accept methyl groups would have clinically beneficial effects in schizophrenia. Unfortunately, most of the materials available are toxic and unsuitable for clinical use. L-Dopa has been given to schizophrenic patients, but its central excitatory effects are evidently so great as to exacerbate the psychosis in many cases (YaryuraTobias and Merlis, 1970; Murphy, 1973) ; an alternative, but less likely, hypothesis would be that this clinical worsening might reflect the activity of methylated metabolites of L-dopa. Polyphenolic compounds inhibit catechol 0-methyltransferase, and can accept methyl groups from S-adenosylmethionine (Baldessarini and Greiner, 1973). While most of these substances are toxic, butylgallate has been given in doses u p to 5 gm to chronic schizophrenics with disappointing results, including gradual clinical worsening, possibly due to coincidental deprivation of antipsychotic medications, and evidence of hepatic toxicity at doses over 2 gm a day (Simpson and Varga, 1972). There was no evidence that the drug was in fact an effective inhibitor of methylation in the patients; even if this did occur, it is not likely that doses below the toxic level of 2 gm a day have an important SAMe-depleting effect, judging by analogy to the relatively mild and short-lasting effects of large doses (several grams a day) of I.-dopa on blood concentrations of SAMe in Parkinson patients (Matthysse ~t al., 1971 ) .
E. Is S-ADENOSYLMETHIONINE T H I ONLYMETHYLDONOR? THE CASEOF METHYLTETRAHYDROFOLATE A provocative aspect of transrnethylation reactions which has recently been widely discussed is that certain amine methyltransferase reactions might be supported preferentially by N'-methyl tetrahydrofolate ( M T H F ) rather than SAMe, including the direct N-methylation of dopamine (followed by p-hydroxylation) as a possible alternative to the synthesis of epinephrine by way of norepinephrine (Laduron, 1972). This reaction is proposed to be an additional form of methyl transfer to that now well established in
56
ROSS J . BALDESSARINI
bacteria and mammalian liver, in which M T H F can donate its methyl group in the synthesis of niethionine from homocysteine (Elford et al., 1965). More recently, M T H F was reported to be a possible, even a preferred, cosubstrate for the transfer in vitro of methyl groups to nitrogen or oxygen atoms of certain indoleamines, including tryptamine (Banerjee and Snyder, 1973; Hsu and Mandell, 1973). The possible physiological significance of these proposed reactions is not clear. A biochemical consideration that tends to cast some doubt on their importance is that the K , for the M T H F in such reactions is similar to the values for SAMe in many methyltransferase reactions (about 10-25 p M ) (Banerjee and Snyder, 1973; Laduron et al., 1974), while the concentrations of the methylfolate in mammalian brain are much lower (never more than about 1 p M , and in most regions several orders of magnitude less) ; nevertheless, the highest levels have been reported, interestingly, to occur in regions rich in endogenous indoleamines (Korevaar et al., 1973). These findings suggest either that the tissue concentrations of M T H F are likely to be an important factor limiting the rate of production of methylated indoleamines, or that these reactions are not very important in vivo. I t is not yet clear whether large doses of substances including methionine and betaine, capable of donating methyl groups to the “1-carbon’’ metabolic pool from which M T H F acquires its methyl group, can increase the availability of M T H F in the mammalian brain, and this possibility should be investigated. M T H F takes on added potential importance in neuropsychiatry since the congenital deficiency of the enzyme, 5,l O-methylenetetrahydrofolate reductase, that produces M T H F from its immediate precursor has been associated with mental retardation and psychosis (Freeman et al., 1975). Recent information has tended to cast considerable doubt on the importance of M T H F in the transfer of methyl groups to oxygen or nitrogen atoms of indoleamines or catecholamines. Thus, it has been noted that some of the products of reaction of M T H F with indoleamines are not N-w-methylated products (Lin and Narasimhachari, 1974). Furthermore, it now appears likely that cyclic derivatives incorporating the amino group can form from indoleamines (variously called tetrahydropyridoindoles or tetrahydrop-carbolines or “tryptolines”) in the presence of M T H F (Rarchas et al., 1974; Hsu and Mandell, 1975; Laduron et al., 1974; Mandel et al., 1974; Meller et al., 1974, 1975; Wyatt et al., 1975). These cyclic derivatives may represent products of nonenzymic condensation with formaldehyde (Hsu and Mandell, 1975; Meller et al., 1975), which can be produced enzymically in some tissues from M T H F (Donaldson and Keresztesy, 1961) including brain (Leysen and Laduron, 1974), probably by the ubiquitous enzyme, 5,lO-methylenetetrahydrofolate reductase (Taylor and Hanna, 1975), and in some tissues also from SAMe by an uncertain, but presumably enzymic
BIOLOGICAL TKAN SMETHYLATION
57
mechanism (Meller et al., 1974). It is also strongly suspected (Lin and Narasimhachari, 1974) that the proposed N-methylation of dopamine (Laduron et al., 1974) and the O-methylation of 5-hydroxyindoles by M T H F (Banerjee and Snyder, 1973) may also have been confused with condensation reactions in the presence of formaldehyde formed from MTHF. Furthermore, the possibility that formaldehyde may form from SAMe in tissues other than erythrocytes (Meller et al., 1974) is not resolved, although pyridoindoles did not form when indoleamines were incubated with SAMe in the presence of heart or brain extracts (Mandel et al., 1974). Presumably, the formation of formaldehyde from SAMe is related to the ability of erythrocyte and pituitary enzymic activity to form the more reduced 1-carbon product, methanol from SAMe (Axelrod and Cohn, 1971 ) .
Ill. Other Assays Related to Transmethylation
A. ASSAYSOF METHYLACCEPTORS : N-ACETYLSEROTONIN AND
HISTAMINE
In addition to the assay of SAMe by the radioactive-enzymic method, analogous concepts and techniques also permit the measurement of other important metabolic components of methylating systems. For example, if the specific radioactivity of SAM? in the reaction described above for the assay of SAMe is held constant, and that of the labeled cosubstrate N-acetylserotonin is decreased by the addition of the unlabeled authentic indole, there again results a linear relationship between the radioactivity ratio in the product of methyl transfer, melatonin, and the amount of N-acetylserotonin present (R. J. Baldessarini, unpublished observations, 1965) . In principle, this strategy might be used for the assay of tissue levels of a variety of methyl acceptors in addition to the precursor of melatonin. The requirements of such assays would include sufficient activity of a methyltransferase enzyme so as to yield detectable quantities of a methylated product, and sufficient selectivity of the partially purified enzyme preparation so that it methylates only the substrate of interest, but not other methyl acceptors in the tissue. Alternatively, relatively crude enzyme preparations may suffice if simple and efficient separation techniques will permit the isolation of only a single niethylated product. Even the presence of large amounts of an alternative methyl acceptor (e.g., after treatment with large doses of L-dopa) should not interfere with such assays, unless the competing reaction effectively prevented the recovery of measurable radioactivity in the product desired; even gross alterations in the kinetic efficiency of the assay reaction
58
ROSS J . BALDESSARINI
produced by tissue components or drugs should not matter since only a ratio of ?H:"C in the final methylated product, not its quantitative recovery, is required. Such an assay has been developed for tissue histamine (Snyder et al., 1966). This method is highly sensitive and specific, and it avoids the problem that has plagued previous bioassay and spectrophotofluorimetric assays of histarnine: the occurrence of spuriously high values due to the presence of other polyamines, notably spermidine, and particularly in assays of blood and brain tissue. In this assay, SAMe is present at constant specific radioactivity (labeled in the methyl group with "C or 'H) , and is allowed to react with labeled histamine (with the opposite radioisotope), buffered extracts of tissue (which have been boiled to inactkaate enzymes and endogenous SAMe) containing endogenous histamine (which survives boiling), and partially purified histamine N-methyltransferase prepared from guinea pig cerebral cortex. The doubly labeled radioactive product, N-methylhistamine, is then extracted into chloroform from the NaOH-treated reaction mixture as the only labeled product recovered, and the ratio of the two labels is determined. This ratio was found to bear the theoretically predicted linear relationship with the amount of unlabeled histamine present. More recently, several technical modifications have been made to increase the sensitivity of the enzymic assay for histamine, and it is now possible routinely to measure about 50 pg of tissue histamine, and under some condition.;, perhaps even less (Taylor and Snyder, 1972; Lipinski et al., 1973). For many purposes, it is not even necessary to utilize two labels, and one can take advantage of the high specific radioactivity of [?H]SAMe for a very sensitive singlelabel microassay for histamine (Taylor and Snyder, 1972). These methods are sufficiently specific and sensitive to permit the assay of histamine in small regions of the brain, including human brain post mortem, in which high concentrations of histamine are found in the hypothalamus (about 1 pglgm) , the colliculi, certain midbrain nuclei, and the olfactory tubercles (0.1-0.3 pglgm) (Lipinski et al., 1973). Blood histamine can also be measured in small samples in this way, and these methods are being applied to clinical studies (J. F. Lipinski, pcrsonal communication). A number of radioisotopic-derivative, enzymic assays of othw amines are based on the preparation and selective recovery of their methylated derivatives from labeled SAMe. Most of these techniques have been devised by Axelrod and his associates. Examples of these methods include assays for octopamine and other p-hydroxylated phenylethylamines, using phenylethanolamine N-methyltransferase ( Molinoff et al., 1969) ; catecholamines, using catechol 0-methyltransferase (Coyle and Henry, 1973; see also Engelman and Portnoy, 1970) ; and serotonin, using serotonin N-acetyltransferase and hydroxyindole 0-methyltransferase (Saavedra et al., 1973a) . These techniques are much more powerful than older chemical or fluorometric
BIOLOGICAL T R A N S M E T H Y LATION
59
methods and permit the microassay of 10-100 pg quantities of amines in small samples of brain tissue (Palkovits et al., 1974a,b).
B. ASSAYOF ATP: L-METHIONINE ADENOSYLTRANSFERASE Variations on the same chemical theme can also be utilized to assay the enzyme which synthesizes SAMe, methionine adenosyltransferase ( ATP :Lmethionine S-adenosyltransferase, EC 2.5.1.6) . Alternative methods for the assay of this enzyme are based on the principle that labeled methionine and the product SAMe have dissimilar binding to ion-exchange resins or papers, thus permitting chroniatographic separation of substrate from product, and an attempt must be made to recover the product quantitatively (see Matthysse et al., 1972; McKenzie and Gholson, 1973). In the first step of the enzymic double-isotopic assay, I.-["H-methyllmethionine is allowed to react with an excess of ATP, Mg", reduced glutathione and various amounts of tissue preparations containing rnethionine adenosyltransferase activity; in the middle of this reaction, ["CISAMe is introduced to provide an automatic means of correcting for losses of newly synthesized SAMe which might take place even as it is being synthesized. Next, the reaction mixture is deproteinized by the addition of trichloroacetic acid, which is removed by washing with ether, and the second step of the assay is conducted with unlabeled N-acetylserotonin and partially purified beef pineal hydroxyindole O-methyltransferase ( H I O M T ) as in the assay [or SAMe described above (Matthysse et al., 1972). In this case, the ratio of ''H:l'C in the doubly labeled melatonin finally recovered in a chloroform extract was predicted and found to be linearly related to the amount of SAMe synthesized in step one, and hence to the activity of the adenosyltransferase. The advantage of this method is that it provides for correction for losses of the rather unstable product, SAMe, not only after the reaction is terminated, but also during the assay itself, when reactions leading to the utilization and destruction of SAMe can occur, although they are almost never taken into account in the traditional chromatographic techniques for separating and recovering SAMe.
C. ASSAYOF MI:THIONINE Finally in addition to the assay of SAMe and of the eni.yme which synthesizes it, as well as of molecules which accept methyl groups from SAMe, it has recently been demonstrated that similar principles can be applied to the assay of L-methionine by an cnzymic derivative, double-isotope method (Lombardini et al., 1971) . In this assay, ['H-methyl]~-methionineis reacted with [8-"C]ATP, Mg", glutatliionine, and an excess of bacterial methionine
60
R O S S J. DALDESSARINI
TABLE V I DOUBLE-LABEL ENZYMIC ASSAYS RELATED TO METHYL AT ION^ Assay L-Methionine (Me)
Me
Reaction
Referencesb
+ [3H]Me+ [14C]ATP+ Mg2+
(a)
M4T
+[W,14C]SAMe S-Adenosylmethionine (SAMe) Methionine adenosyltransferase (MAT)
SAMe
+ PPi + Pi
+ [14C]SAMef [3H]NAS
(b)
HIOMT
----+ [14C,3H]MEL
-]
(1) [3H]Me
+ ATP + Mg2+ [3H]SAMe + PPi + Pi
Tissue MAT
(2) [3H]SAMe +NAS [14C]SAMe
HIOMT
N- Acetylserotonin WAS)
-
Histamine (HIST)
HIST
['*C]SAMe HIOMT
t3H,l4CC]MEL
+ [3H]NAS + NAS
[14C,3H]MEL
+ [3H]HIST + [14C]SAMe
HNMT
(d)
(el
[3H,14C]Me-HIST
a The abbreviations used are; Ac, acetyl; HIST, histamine; HIOMT, hydroxyindole0-methyltransferase; HNMT, histamine-N-methyltransferase;MAT, ATP-L-methionine adenosyltransferase; Me, L-methionine; MEL, melatonin; NAS, N-acetylserotonin; SAMe, S-adenosyl-L-methionine. Key to references; (a) Lombardini el al., 1971: (b) Baldessarini and Kopin, 1966: ( c ) Matthysse et al., 1972: (d) R. J. Baldessarini, unpublished observations: ( e ) Snyder et al., 1966.
adenosyltransferase purified from Escherichia coli; tissue extracts containing endogenous methionine are also included after their treatment on small columns of Dowex anion-exchange resin to remove endogenous ATP. After the enzymic reaction is terminated, newly synthesized, doubly labeled SAhte is separated from the radioactive cosubstrates on columns of Dowex (NH,') cation-exchange resin which are washed with large volumes of water; finally, the doubly labeled SAMe is eluted with NH,OH and the radioactivity ratio (3H:14C) is counted and was found to bear the predicted linear relationship to the amount of authentic 1.-methionine present in the reaction mixture. T h e method is capable of detecting 150 ng of methionine easily, and can be made sensitive enough to detect 1.5 ng. The results of this assay were found to accord well with the data provided by a Beckman amino acid analyzer, demonstrating in rat tissue for example, levels of methionoine of
BIOLOGICAL TRANSMETHYLATION
61
80-86 nmoles/gm in liver and 65-68 nmoles/gm in brain (Lombardini et al., 1971). I n summary, the principle of using two radioactive substrates with dissimilar labels has permitted the development of highly sensitive and specific enzymic assays for estimating tissue levels or activity of several important components in methyltransferasc systems in mammalian tissues : L-methionine, S-adenosylmethionine, L-methionine adenosyltransferase, and molecules which accept methyl groups from SAMe, including histamine (Table VI) . These methods provide linear standard plots of radioactivity ratios (3H:1*C)vs the amount of authentic material being assayed. While they involve the utilization of relatively exotic materials, these are either commercially available or easily prepared in a short time and are stable for many months of use. The assays themselves are relatively simple and rapid and can be done in large numbers. One advantage of several of them is that corrections for recovery and reaction efficiency are provided automatically. The methods have been successfully applied to various fractions of human blood, and so they are appropriate for use in clinical metabolic experiments.
IV. Clinical Implications: Need for New Strategies for Clinical Metabolic Research in Schizophrenia
The observation that 10-20 gni loading doses of methionine, of several amino acids, uniquely led to exacerbation of psychosis in schizophrenic patients is highly unusual among biological findings in severe psychiatric illness in that it has been independently confirmed repeatedly in the past decade. The initial suggestion that methionine might act by increasing the production of potentially psychotomimetic methylated amines by way of increased availability of S-adenosylmethionine, is only partially supported by the available evidence. Thus, there is excellent metabolic evidence in animals, and very little in man, that the availability of SAMe is dependent on the availability of its precursor, methioninc, and on the rate of its utilization, which can be increased by giving large doses of certain methyl acceptors. O n the other hand, the evidence that methionine loading increases blood or tissue levels of SAMe in patients remains untested even though the means of doing so have been available for 10 years. Furthermore, the attempts to document increases of the methylation of amines after methionine loading in animals or man have variously been unsuccessful (catecholamines), or when successful (indoleamines), have not been studied with unambiguous analytical methods, or have not been confirmed by independent replication. The more recent introduction of SAMe for parenteral administration in clinical trials suggests yet another more direct approach to the question of whether in-
62
ROSS J . BALDESSARINI
creased availability of this methyl donor can increase the methylation of amines. The more recent clinical experiments with blood platelets which suggest that there may be either increased synthesis of methylated amines by an indoleamine N-methyltransferase or decreased catabolism of amines by M A 0 are other important findings urgently in need of independent corroboration. An important question at the present time is whether the methylation hypothesis can or should be pursued clinically. There has recently been a tendency to find many reasons not to pursue this and other metabolic hypotheses. Part of the apparent pessimism may be based on the large numbers of unsuccessful attempts to demonstrate statistically significant mean differences between groups of comparison subjects in the past 20 years. These failures may indicate that no difference in fact exists, but it is also possible that we may have been missing some good opportunities. One of the common experimental strategies which surely must help to obscure metabolic idiosyncrasies of certain psychotic patients is to study groups of schizophrenic or depressed or manic patients as if “schizophrenia” or “manic-depressive illness” were unitary phenomena. Even the subclassification of patients on the basis of rigorous clinical criteria, as has become more popular for the affective disorders recently, may not be sufficient, particularly among the schizophrenic disorders. If the clinical groups studied are in fact highly heterogeneous biologically, it is then almost inevitable that specific biological measures of mean differences between groups will fail to detect even significant biological differences unless these are enormous and obvious. One alternative strategy is to take advantage of a given biological “fact” as a starting point from which to acquire further relevant biological information. A striking but neglected example of such an opportunity is the observation that methionine loads induce transient psychotic exacerbations in only a limited proportion of schizophrenic patients: as few as 15-50c/c of chronic patients so tested. I t is reasonable to study such patients much more intensively, and to compare them not only with normal subjects or patients in other diagnostic categories, but also with other schizophrenics who are clinically similar except that they did not become more psychotic with large doses of methionine. Progress in this direction has been slow in recent years, perhaps to a considerable extent owing to the recent tendency toward excessive conservatism in regard to human experimentation in many countries, and particularly in psychiatry. Clearly, there are very compelling ethical dilemmas presented by the type of experimentation being suggested : one does not consider an experiment which deliberately seeks to provoke even a transient exacerbation of illness casually, particularly when it requires the gaining of “informed consent” from institutionalized patients whose mental capacity is reduced by reason of chronic psychosis. Nevertheless, to conclude
RIOLOGICAL T R A N S M E T H Y L A T I O N
63
that such experimentation should not be done under any circumstances seems shortsighted and irrational. Once such patients as the “methionine responders” are identified among a population of schizophrcnics, they should not only be studied metabolically, but they should also be given experimental treatments specifically aimed at decreasing methylation or the formation of certain aromatic amines, as they become available. At the present time there are several drugs that have already been developed at least to the stage of dinical trials, including inhibitors that selectively block the synthesis of catecholamines (cu-methyltyrosine) or 5hydroxyindoleamines (p-chlorophenylalanine ) , inhibitors of aromatic amino acid decarboxylases (MI(-486, IiO-44602 ) , inhibitors of phenylethylamine p-hydroxylase (disulfiram, fusaric acid), an acceptor of niethyl groups from SAMe which inhibits at least COMT (butylgallate). It might also be possible to devise other relatively nontoxic strategies to interfere with methylation reactions ; for example, guanidineacetic acid is a natural metabolite which utilizes a large number of niethyl groups daily in the synthesis of the biologically almost inert product creatine, and it might be given safely in large doses, which can decrease tissue levels of SAMe in the rat (S. Matthysse, personal communication). In attempting to grapple with the ethical issues raised by these experiments, the risk of short-term exacerbation of an already virtually incurable illness such as chronic poor-prognosis schizophrenia (dementia praecox) should be balanced against the chance of offering more rational, though experimental, treatments as well as the considerable nonspecific benefits that may derive from the intensive study and treatment of patients who might otherwise receive little more than custodial care. In short, the effect of metliionine in schPLophrenia and the ceveral metabolic findings related to the synthesis and degradation of methylated indoleamines in that syndrome are among the very few promising clues to a pathophysiological substratum of a major psychiatric illness based on clinical metabolic experimentation, and, as such, they can and should be pursued vigorously. REFERENCES Alexander, F., Curtis, G. C., Sprince, H., and Crosley, A. P. (1963). J . N e r u . M e n t . Dis. 137, 135-142. Antun, F. T., Burnett, G. B., Cooper, A. J., Daly, R. J., Smythies, J. R., and Zealley, A. K. (1971a). J . Psychint. Res. 8, 63-71. Antun, F. T., Eccleston, D., and Smythies, J. R . (1971b). In “Brain Chemistry and Mental Disease’’ (B. T. Ho and W. M. McIsaac, eds.), pp. 61-71. Plenum, New York. Axelrod, J., and Cohn, C. K. ( 1 9 7 1 ) . J . Phnrmacol. E x $ . T h e r . 176, 650-654. Baldessarini, R. J. (1966a). I n “Amines and Schizophrenia” ( H . Himwich, S. Kety, and J. Smythies, eds.) , pp. 199-206. Pergamon, Oxford. Baldessarini, R. J. (1966b). Biochem. Phartizacol. 15, 741-748.
64
ROSS J. BALDESSARINI
Baldessarini, R. J. (1975). I n “The Nature and Treatment of Depression” (S. C. Draghi and F. F. Flach, eds.), Chapter 9, pp. 347-385. Wiley, New York. Baldessarini, R. J., and Bell, W. R. (1966). Nature ( L o n d o n ) 209, 78-79. Baldessarini, R. J., and Carbone, P. P. (1965). Science 149, 644-645. Baldessarini, R. J., and Fischer, J. E. (1967). Surgery 62, 311-318. Baldessarini, R. J., and Greiner, E. ( 1973). Biochem. Pharmacol. 22, 247-256. Baldessarini, R. J., and Karobath, M. (1972). Neuropharmacology 11, 715-720. Baldessarini, R. J., and Kopin, I. J. (1963). Anal. Biochem. 6, 289-292. Baldessarini, R. J., and Kopin, I. J. (1966). J. Neurochem. 13, 769-777. Ban, T. A. (1969). Psychopharmacol. Bull. 5, 5-20. Banerjee, S. P., and Snyder, S. H. (1973). Science 182, 74-75. Barchas, J. D., Elliott, G. R., DoAmaral, J., Erdelyi, E., O’Connor, S., Bowden, M., Brodie, H . K. H., Berger, P. A., Renson, J., and Wyatt, R. J. (1974). Arch. Gen. Psychiat. 31, 862-867. Bartholini, G., and Pletscher, A. (1970). Brit. 1. Pharmacol. Chemother. 40, 461-467. Baudry, M., Chast, F., and Schwartz, J.-C. (1973). J. Neurochem. 20, 13-21. Berlet, H. H., Matsurnoto, K., Pscheidt, G. R., Spaide, J., Bull, C., and Himwich, H. E. (1965). Arch. Gen. Psychiat. 13, 521-531. Bhikharidas, B., Mann, L. R., and McLeod, W. R. (1975). J. Neurochem. 24, 203-205. Brune, G. G., and Himwich, H. E. (1962). J. Neru. M e n t . Dis. 134, 447-450. Brune, G. G., and Hirnwich, H. E. ( 1963). Recent Advan. Biol. Psychiat. 5, 144-160. Buscaino, G. A., Spadetta, V., and Carella, A. (1969). Acta Neurol. (Naples) 24, 113-1 18. Cantoni, G. L. (1951). J . B i d . Chem. 189, 203-216. Carey, M. C., Donovan, D. E., Fitzgerald, O., and McCauley, F. D. (1968). Amer. J. M e d . 45, 7-25. Chalrners, J . P., Baldessarini, R. J., and Wurtrnan, R. J. (1971). PTOC.Nut. Acad. Sci. U S . 68, 662-666. Chaudhuri, D. K., and Kodicek, E. (1949). Biochem. J. 44,434-438. Cohen, S. M., Nichols, A,, Wyatt, R., and Pollin, W. (1974). B i d . Psychiat. 8, 209-2 25. Cohn, C. K., Dunner, D. I>., and Axelrod, J. (1970). Science 170, 1323-1324. Coper, H., Deyhle, G., Fahndrich, G., Fahndrich, E., Rosenberg, L., Strauss, S., Blurn, A., and Dufour, H. ( 1972). Pharmakopsychiat. Neuro-Psychopliarmakol. (Stuttgart) 5, 177-187. Coyle, J. T., and Henry, D. (1973). J. Neurochem. 21, 61-67. Creveling, C. R., and Daly, J. W. (1967). Nature ( L o n d o n ) 216, 190-191. Deguchi, T., and Barchas, J. ( 197 1 ) . J . B i d . Chem. 246, 3 175-3 181. Domino, E. F., Krause, R. R., and Bowers, J. (1973). Arch. Gen. Psychiat. 29, 195-20 1. Donaldson, K. O., and Keresztesy, J. C. (1951). Riochem. Biophys. Res. Commun. 5, 289-292. Dunner, D. L., Cohn, C. K., Weinshilboum, R. M., and Wyatt, R. J. (1973). B i d . Psychiat. 6,215-220. Elford, H. I,., Katzen, H. M., Rosenthal, S., Smith, I,. C., and Buchanan, J. M. (1965). I n “Transmethylation and Methionine Biosynthesis” (S. K. Shapiro and F. Schlenk, eds.), pp. 157-171. Univ. of Chicago Press, Chicago, Illinois. Engelman, K., and Portnoy, B. (1970). Circ. Res. 26, 53-57. Fazio, C., Andreoli, V., Agnoli, A,, Casacchia, M., and Cerbo, R. (1973). Minerua M e d . ( R o m e ) 64, 1-14.
BIOLOGICAL TRANSMETHYLATION
65
Fischer, E., and Spatz, H. ( 1970). Hiol. Psychiat. 2, 235-240. Freeman, J. M., Finkelstein, J. D., and Mudd, S. H. (1975). A’. Engl. /. M e d . 292, 491-496. Friedhoff, A. J., and Van Winkle, E. (1962). J. N e w . M e n t . Dis. 135, 550-555. Friedman, E., Shopsin, B., Sathananthan, G., and Gershon, S. (1974). A m e r . J. Psychiat. 13 1, 1392-1 394. Gillen, J. C., Cannon, E., Magyar, R., Schwartz, M., and Wyatt, R. J. (1973). Biol. Psychiat. 7, 213-220. Hall, P., Hartridge, G., and van I,eeuwen, G. H. (1969). Arch. Gen. Psychiat. 20, 573-575. Hartley, R., Padwick, D., and Smith, J. A. (1972). /. Pharm. Pharmacol. 24, Suppl., 100P-103P. Haydu, G. G., Dhrymiotis, A , , Korenyi, C., and Goldschmidt, I,. (1965). Amer. /. Psychiat. 122, 560-564. Heath, R. G., Nesselhof, W., Jr., and Timrnons, E. (1966). Arch. Gen. Psychiat. 14, 213-217. Heslinga, F. J. M., van Tilburg, W., and Starn, F. C. (1970). Psychiat. Neurol. Neurochir. ( A m s t . ) 73, 157-164. Hoffer, A,, and Osmond, H. (1964). Acta Psychiat. Scand. 40, 171-189. Horwitt, M. K., Harvey, C. C., Rothwell, W. S., Cutter, J. I,., and Hoffron, D. (19.56 ) . /. h utr. 60, Suppl. 1, 1-43, HsL~, I,. L.,and Mandell, A. J. (1973). Life Sci. 13, 847-858. Hsu, L. L., and Mandell, A. J. (1975). J. Neurochem. 24,631-636. Israelstarn, D. M., Sargent, T., Finley, N. N., Winchell, H. S., Fish, M. B., Motto, J., Pollycove, M., and Johnson, A. (1970). J . Psychiat. Res. 7, 185-190. Kakimoto, Y . , Sano, I., Kanazawa, A,, Tsujio, T., and Kaneko, 2. (1967). Nature ( L o n d o n ) 216, 1110-1111. Kopin, I. J., and Baldessarini, R. J. (1971 ) . I n “Methods in Enzymology” ( H . Tabor and C. W. Tabor, eds.), Vol. 17B, pp. 397-400. Academic Press, New York. Korevaar, W. C., Geyer, M. A,, Knapp, S., Hsu, L. L., and Mandell, A. J. (1973). Nature ( L o n d o n ) ,N e w Biol. 245, 244-245. Laduron, P. (1972). Nature ( L o n d o n ) ,N e w B i d . 238, 212-213. I.aduron, P. M., Gommersen, W. R., and Leysen, J. E. (1974). Biochem. Pharmacol. 23, 1599-1608. Leysen, J., and Laduron, P. (1974). FEBS ( F e d . Eur. Biochem. Soc.) Lett. 47, 299-303. I h , R.-I,,, and Narasimhachari, N. ( 1974). Res. Commun. C h e m . Pathol. Pharmacol. 8, 535-542. Lin, R.-L., Narasimhachari, N., and Himwich, H . E. (1973). Biochem. Biophys. Res. C o m m u n . 54, 751-759. Lipinski, J. F., Schaumburg, H . H., and Baldessarini, R. J. (1973). Brain Res. 52, 403-408. Lipinski, J. F., Mandel, L. R., Ahn, H . S., Van den Heuvel, W. J . A., and Walker, R. W. (1974) . B i d . Psychiat. 9, 89-91. Lipton, M. A,, ed. (1973). “Megavitamin and Orthomolecular Therapy in Psychiatry,” Task Force Rep. No. 7, pp. 1-54. Amer. Psychiat. Ass., Washington, D.C. Lombardini, 5. B., Burch, M. K., and Talalay, P. (1971). J. Biol. C h e m . 246, 4465-4470. McKenzie, R. M., and Gholson, R. K. (1973). Anal. Uiochem. 53, 384-391.
66
R O S S J . BALDESSARINI
Mandel, L. R., Walker, R. W., Rosegay, A., Van den Heuvel, W. J. A., and Rokach, J. (1974). Science 186, 741-743. Mandell, A. J., and Morgan, M. (1971). Nature ( L o n d o n ) ,N e w Biol. 230, 85-87. Matthysse, S., and Baldessarini, R. J. (1972). Amer. J. Psychiat. 128, 1310-1312. Matthysse, S., Lipinski, J., and Shih, V. (1971). Clin. Chim. Acta 35, 253-254. Matthysse, S., Baldessarini, R. J., and Vogt, M. (1972). Anal. Biochem. 48, 410-421. Meller, E., Rosengarten, H., and Friedhoff, A. J. (1974). Life Sci. 14, 2167-2178. Meller, E., Rosengarten, H., Friedhoff, A. J., Stebbins, R . D., and Silber, R. (1975). Science 187, 171-173. Meltzer, H. Y., and Stahl, S. M. (1974). Res. Commun. Chem. Pathol. Pharmacol. 7, 419-431. Missala, K., Lal, S., and Sourkes, T . L. (1973). Eur. J . Pharmacol. 22, 54-58. Molinoff, P. B., Landsberg, L., and Axelrod, J. (1969). J . Pharmacol. Exp. Ther. 170, 253-261. Murphy, D. L. (1973). Annu. Reu. Med. 24, 209-216. Murphy, D. L., and Wyatt, R. J. (1972). Nature ( L o n d o n ) 238, 225-226. Narasimhachari, N., and Himwich, H. E. (1973a). Life Sci., Part I 2 12, 475-478. Narsimhachari, N., and Himwich, H. E. (1973b). Biochem. Biophys. Res. Commun. 55, 1064-1071. Narsimhachari, N., and Lin, R.-L. (1974). Rep. Commun. Chem. Pathol. Pharmacol. 8, 341-351. Narasimhachari, N., Plaut, J., and Himwich, H. E. (1972). J . Psychiat. Res. 9, 325-328. Osmond, H., and Smythies, J. (1952). J. Ment. Sci. 98, 309-315. Palkovits, M., Brownstein, M., and Saavedra, J. M. (1974a). Brain Res. 80, 237-249. Palkovits, M., Brownstein, M., Saavedra, J. M., and Axelrod, J. (197410). Brain Res. 77, 137-149. Park, L. C., Baldessarini, R. J., and Kety, S. S. (1965). Arch. Gen. Psychiat. 12, 346-351. Pearce, L. A,, and Waterbury, L. D. (1974). Neurology (Minneapolis) 24, 640-641. Pollin, W., Cardon, P. V., Jr., and Kety, S. S. (1961). Science 133, 104-105. Price, J. (1972). J . Psychiat. Res. 9, 345-351. Rubin, R. A,, Ordonez, L. A,, and Wurtman, R. J. (1974). J . Neurochem. 23, 227-2 3 1. Saavedra, J. M., Brownstein, M., and Axelrod, J. (1973a). J. Pharmacol. Enp. Ther. 186, 508-515. Saavedra, J. M., Coyle, J. T., and Axelrod, J. (1973b). J . Neurochem. 20, 743-752. Salvador, R. A., and Burton, R. M. (1965). Biochem. Pharmacol. 24, 1185-1 196. Salvatore, F., Utili, R., and Zappia, V. (1971). Anal. Biochem. 41, 16-28. Schatz, R. A,, and Sellinger, 0. Z. (1975). J . Neurochem. 24, 63-66. Schwartz, M. A,, Aikens, A. M., and Wyatt, R. J. (1974). Psychopharrnacologia 38, 319-328. Schweitzer, J. W., Friedhoff, A. J., Angrist, B. M., and Gershon, S. (1971 ) . Nature ( L o n d o n ) 229, 133-134. Sharma, S., and Sinari, V. P. (1971). Dis. N e r v . Syst. 32, 831-834. Simpson, G. M., and Varga, V. (1972). J . Clin. Pharmacol. New Drugs 12, 417-421. Sloane, K. M., and Bridges, J. M. (1968). Acta Haematol. 40, 18-27. Snyder, S. H., Baldessarini, R. J., and Axelrod, J. (1966). /. Pharmacol. Exp. Ther. 153, 544-549. Spaide, J., Tanimukai, H., Bueno, J. R., and Himwich, H. E. (1968). Arch. Gen. Psychiat. 18, 658-666.
BIOLOGICAL T RAN SMETHYLATION
67
Sprince, H. (1970). Biol. Psychiat. 2, 109-117. Sprince, H., Parker, C. M., Jameson, D., and Alexander, F. (1963). /. N e r . Ment. Dis. 137, 246-251. 'I'anamukai, H., Ginther, R., Spaide, J., Bueno, J. R., and Himwich, H. E. (1970). Brit. J. Psychiat. 117, 421-430. 'I'arsy, D., and Baldessarini, R. J. (1974). Neuropharmacology 13, 927-940. Taylor, K. M., and Snyder, S. H. (1972). 1.Neurochem. 10, 1343-1358. Taylor, R. T., and Hanna, M. L. ( 1975). Life Sci. (in press). Tran-Manh, N., Laplante, M., Saint-Laurent, J., and LeBel, E. (1972). Reu. Can. Biol. 31, Suppl., 255-262. Wurtrnan, R. J., Rose, C. M., Matthysse, S., Stephenson, J., and Baldessarini, R. J. (1970). Science 169, 395-397. Wyatt, R. J., Termini, B. A,, and Davis, J. M. (1972). Schizophrenia Bull. 4, 10-77. Wyatt, R. J., Mandel, L. R., Ahn, H. S., Walker, R. W., and Van den Heuvel, W. J. A. (1973a). Psychopharmacologia 31, 265-270. Wyatt, R . J., Murphy, D. L., Belmaker, R., Cohen, S., Donnelly, C. H., and Pollin, W. (197310). Science 179, 916-918. Wyatt, R. J., Saavedra, J. M., and Axelrod, J. ( 1 9 7 3 ~ ) A. m e r . 1. Psychiat. 130, 754-760. Wyatt, R. J., Erdelyi, E., DoAmaral, J. R., Elliott, G. R., Renson, J., and Barchas, J. D. (1975). Science 187, 853-855. Yaryura-Tobias, J. A., and Merlis, S. (1970). 1.A m e r . M e d . Ass. 211, 1857-1858. Zappia, V., Zydek-Cwick, C. R., and Schlenck, F. (1969). 1. Biol. Chem. 244, 4499-4509.
This Page Intentionally Left Blank
SYNAPTOCHEMISTRY OF ACETYLCHOLINE METABOLISM IN A CHOLINERGIC NEURON' By Bertalan Csillik
Department of Anatomy, University Medical School, Szeged, Hungary
I. Introduction . A. Evergreen Cholinesterase , B. Acetylcholinesterase : Basic Characteristics and Cytochemical Techniques . C. Acetylcholinesterase: Recent Developments D. Acetylcholinesterase : Role in Synaptic Transmission . 11. Histochemistry of Acetylcholinesterase in the Spinal Motoneuron . A. The Perikaryon . B. TheAxon C. Axon Terminals: The Motor End Plate . D. Recurrent Initial Axon Terminals: Renshaw Elements . 111. Indirect Information on Cholinergic Mechanisms . A. Role of Choline in Transmitter Synthesis: Autoradiography of [ I4C] Hemicholinium B. Dynamics of Thiamine Pyrophosphatase: Localization in Synapses IV. Molecular Anatomy of Transmitter Release . A. Origin of Synaptic Vesicles . B. Charging of Synaptic Vesicles . C. Discharging of Synaptic Vesicles D. Transmitter Binding to the Receptor: Postsynaptic Amplification E. Interrelations between Acetylcholinesterase and the Acetylcholine Receptor: Similarity or Identity? References .
.
.
.
. . . . . .
.
. . .
69 69 71 72 75 77 77 82 84 108 112
. . . .
.
. . .
112 117 119 119 121 123 128
.
.
130 133
.
I. Introduction
A. EVERGREEN CHOLINESTERASE Has cholinesterase anything to do with synaptic transmission? Iconoclastic and ungrateful, as it is, this question reflects serious doubts, tacitly or explicitly underlying many recent neurochemical discussions. Iconoclastic, This review was completed during the author's stay a t the Neurosciences Research Program, Boston, Massachusetts. 69
70
BERTALAN CSILLIK
since up-to-date synaptochemistry is based mainly upon the pioneering Koelle ( 1951, 1954; Koelle and Friedenwald, 1949) and Couteaux (1951 ; Couteaux and Taxi, 1952) histochemical studies that, for the first time, enabled visualization of the postsynaptic membrane on the basis of its acetylcholinesterase ( AChE?) activity. Ungrateful, too, since discovery of this unique feature of the “subneural apparatus” was not only the first irrefutable proof for the cellular independence of pre- and postsynaptic membranes but, a t the same time, it was also the first direct demonstration of a neurochemical transmission mechanism at the microscopic level. I n spite of these studies, doubts as to the usefulness of AChE techniques in demonstrating cholinergic transmission mechanisms, especially in the central nervous system, are often justified. The presence of this enzyme in noncholinergic synapses (Silver, 1967; Ishii and Friede, 1967; Manocha and Shantha, 1969; Haj6s et al., 1970; Gwyn and Flumerfelt, 1971 ; etc.) might indicate nonspecificity of the reaction or ubiquity (hence nonspecificity ) of AChE. Also the localization of the AChE reaction product at the electron microscopic level, even in well proved cholinergic synapses, raised serious questions ; the functional importance of a presynaptic enzyme localization, consistently accompanying the postsynaptic bulk, is not easily reconciled with current physiological and pharmacological theories of impulse transmission. Unfortunately, the transmitter of the cholinergic synapse cannot be directly visualized; until now, efforts to locate ACh by histochemical techniques, were unsuccessful. Catecholamine histochemistry, due to the fabulous fluorescence of aldehyde condensation products, is in a far more advantageous position. Neither is the direct approach to locate the ACh-synthesizing machinery by means of a chromogenic reaction, more successful. I n the present state of art, ChAc histochemistry is neither specific nor sensitive enough (Burt and Silver, 1973) ; the immune histochemical method, offering a specific and sensitive demonstration of ChAc, is still in a preliminary state ( MaltheSorensen et al., 1973). The indirect approach, outlined in Section 111, can be regarded only as a useful approximation, suggesting sites of ACh synthesis. Thus, the main source of structural information on the sites of cholinergic transmission mechanism that we can rely upon is still the demonstration of AChE activity at the lizht and the electron microscopic levels. Just as with other evergreens, however, the where, the when, and the how of its application is an extremely delicate task.
’ Abbreviations used : AChE, acetylcholinesterase (acetylcholine hydrolase, EC 3.1.1.7) ; ACh, acetylcholine; AChR, acetylcholine receptor; ChAc, choline acetylase (acetyl-CoA: choline-O-acetyltransferase, EC 2.3.1.6).
SYNAPT0C:HEMISTRY O F A CHOLINERCIC N E U R O N
71
B. ACETYLCHOLINESTERASE : BASICCHARACTERISTICS AND CYTOCHEMICAL TIXHNIQUES After Dale’s ( 1914) and Loewi’s (Loewi and Navrratil, 1926) hypotheses on the existence of an enzyme responsible for the inactivation of ACh, and Stedman, Stedman, and Easson’s first observations on an enzyme actually hydrolyzing ACh (1932), “true” or “specific” AChE was first detected in red blood cells by Mendel, Mundell, and Rudney (1943) and analyzed in detail by Augustinsson and Nachrnansohn (1949). O n the basis of Nachiriansohn’s review ( 1955), the main characteristics of this enzyme were summarized several years ago by Leuzinger (1969) as follows: 1. AChE is present in all types of conducting tissues throughout the animal kingdom. 2. AChE is relatively specific for ACh and distinctly different from other csterases. 3. AChE hydrolyzes ACh in a few microseconds. 4. AChE is located in excitable rnerribranes of axons, muscle fibers, nerve terminals, and postsynaptic membranes. An extraordinarily high concentration is found in thc electric organs of electric fish. 5. Characteristic of AChE is a substrate inhibition at increasing substrate concentrations. 6. AChE is less sensitive to organophosphates (DFP, Ambenonium, Mipafox, etc.) than pseudoChE. Owing to the abundance of excellent reviews and summaries concerning various aspects of AChE biochemistry, physiology, and pharmacology, it would be entirely superfluous to go into details with respect to what can bc described as the “classical knowledge,” summarized in Koelle’s Handbook : “Cholinesterases and Anticholinesterase Agents” ( 1963) . For more recent details, the essays of Hebb and Morris (1969), Marchbanks ( 1970), Whittaker (1969a,b), Potter (1970a), Molinoff and Potter (1972), Jones (1972), and Fonnurn (1973) should be consulted. Techniques for the histochemical demonstration of AChE activity have also been reviewed in detail (Koelle, 1963; Zacks, 1964; Csillik, 1965, 1967; Friedenberg and Seligman, 1972) . Essentially, these techniques fall into two categories. Most of them are variants of the original acetylthiocholine procedure, developed 25 years ago by Koelle and Friedenwald (1949). I n spite of its shortcomings, owing to the problematic penetration of the substrate and the histological substantivity of the end product, this technique is still the only one the substrate of which differs from the physiological substrate of the enzyme only in a single atom [S in thiocholine instead of 0 in choline (Fig. 1 ) 1. All other techniques, employing substrates essentially differing
72
BERTALAN CSILLIK
AChE
+ cuso,
.+I
Cu- Thiocholine
CuS (brown precipitate, medium electron dense)
FIG. 1. Chemistry of the Koelle-Friedenwald histochemical reaction.
from ACh (naphthylacetate, indoxyl acetate, thiolacetic acid, benzene diazonium salts with thiol acetate functions, etc.) , though useful in providing additional information, suffer from the basic inadequacy due to substrate aspecificity. These problems especially ardent as we approach the sizes of molecular dimensions in electron microscopic histochemistry, and accentuated also by artificial translocation of enzyme reaction products, will be discussed in the context of the interpretation of fine structural studies (Section
11, c, 3 ) . C. ACETYLCHOLINESTERASE : RECENTDEVELOPMENTS One of the most important features of AChE is its membrane-bound state within the neuron (Potter, 1970a). Such membranes include the surface membrane of the cell as well as the internal membranes of the endoplasmic reticulum; early fractionation studies revealed this latter to be a “microsomal” enzyme (Whittaker, 1965). More recent studies emphasize that, among surface membranes, the postsynaptic membrane is a major source of enzyme activity; in fact, McRride and Cohen (1972) succeeded in histochemically demonstrating AChE activity in the postsynaptic membranes of brain synaptosomes, obtained by differential centrifugation. This does not exclude the possibility that at least part of the AChE revealed in synaptosomal preparations might be due also to presynaptic elements. Another important facet of AChE biochemistry is related to the molecular structure of the enzyme. AChE isolated from the electroplax is a multisubunit enzyme (Leuzinger, 1969) ; more recent studies prove that also AChE of the mammalian brain is a polymer (Chan et al., 1972). According to Hollunger and Niklasson (1973) the monomer of the mammalian brain AChE has a molecular weight of 80,000; these monomers undergo aggregation, yielding oligomers with molecular weights of 250,000 and 510,000, respectively. I t has been suggested that aggregation of the monomers is
S Y N A P T O C H E M I S T R Y OF A CHOLINERCIC N E U R O N
73
brought about by some factor, probably by a phospholipid (Grafius et al., 1971), or by the aggregating protein of Kremzner and Fei (1971). Recent studies performed on AChE isolated from the electroplax of Electrophorus electricus suggest, furthermore, not only that AChE is a multisubunit enzyme, but also that the subunits themselves differ from each other both in size and shape (Bauman et al., 1972). Whereas the membranebound subunit appears to be elongated, globular subunits ( M W = 170,000) seem not to be associated with membranes. Similar results were reported also by Ron et al. (1973) and by Rieger et al. (1973a,b). I n this context it may be recalled that, on the basis of histochemical studies, Koelle et al. (1970) have actually predicted such differences between isoenzymes of AChE, in line with Eranko’s concept on “desmo” and “lyo” esterases (Eranko et al., 1964). T h e existence of a soluble form of AChE raises the intriguing question whether or not, AChE transported by axoplasmic flow mechanisms, being en route from the site of synthesis (perikaryon) to the site of functional activity (terminal), might be in a globular, soluble form. These recent results on a soluble portion of AChE should not be mistaken for those early observations (Ord and Thompson, 1951 ; etc.) when various detergents and organic solvents were used for solubilization of the enzyme. Recent techniques avoid all these drastic procedures; up-to-date separation of enzyme-active fractions is performed preferably in ion-free media. T h e fact that AChE is a multisubunit enzyme raises the important question whether or not, AChE is an allosteric enzyme. Recently Kato et al. ( 1972) observed sigmoidicity in the reaction velocity of acetylthiocholine hydrolysis by AChE, that suggests cooperativity between the individual subunits; also earlier studies (Changeux, 1966) point at the allosteric nature of AChE. Important discoveries are related also to the molecular biology of the active center of the AChE molecule. I t has been known for a long time that the active center of AChE consists of an anionic and an esteratic site (Wilson, 1967). The esteratic site, equipped with a nucleophilic oxygen atom of a serine residue, attacks the carbonyl atom of ACh; at the same time, the esteratic site exerts also an electrophilic attack upon one of the oxygens of the substrate. (Fig. 2 ) . Thus, the net result of these attacks is an acylated enzyme intermediary product, an important step in enzyme mechanism that gave rise to the development of the thiolacetic acid histochemical technique for AChE ( CrCvier and BClanger, 1955). During all these events, the role of the anionic site is to bind the quaternary nitrogen of ACh by electrostatic forces, quasi immobilizing the substrate during hydrolysis. Recent studies, employing the photochromic ligand p-phenylazophenyl trimethyl ammonium chloride as an AChE inhibitor, revealed that the ac-
74
BERTALAN CSILLIK
Esteratic site
Anionic site
T I
CH3- N-CH2CH3
CH,-
,
O-C=O
c *3
FIG. 2. The active center of acetylcholinesterase (Wilson, 1967)
tive site of the enzyme contains two hydrophobic regions, consisting of phenylalanine, tyrosine, tryptophan, and/or histidine residues (Galley et al., 1973) (Fig. 3 ) . O n the other hand, Mooser and Sigman (1972) have shown that, in addition to the active site, AChE possesses also a ligand binding site with a special affinity for d-tubocurarine. Such a peripheral binding site had been postulated earlier on the basis of kinetic studies (Changeux, 1966) and by nuclear magnetic resonance studies (Kato et al., 1972). The existence of such a ligand-binding site, differing from the active center, appears now to be unequivocally proved. Since such a site is a prerequisite for a regulatory enzyme, these studies are crucial in proving the allosteric character of AChE. O n the other hand, the fact that d-tubocurarine is bound to AChE, but not at the active site, is of considerable importance from the point of view of interpreting the reliability of electron histochemical AChE localization (page 9 6 ) . Amonic site
FIG. 3. The active center of acetylcholinesterase (Galley et al., 1973).
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
75
D. ACETYLCHOLINESTERASE : ROLEI N SYNAPTIC TRANSMISSION According to the classical theory of neurochemical impulse transmission, AChE is located at the postsynaptic membrane. This is a strategic localization, suited for the inactivation of transmitter molecules that already had performed their task: having combined with the AChR and changed its molecular conformation, they have induced an increased permeability of the postsynaptic membrane (Nachmansohn, 1955) . I n fact, polarization optical studies (Csillik, 1963) indicate that ACh induces a molecular rearrangement of nonlipid constituents of the postsynaptic membrane ; according to Changeux e t al. ( 1969), configurational changes convert receptor proteins into ionophores. Recent studies (Meunier et al., 1973) suggest that these molecular transformations represent a cooperative process in the postsynaptic membrane. Thus, AChR is assumed to be an integral part of the postsynaptic membrane (Miledi and Potter, 1971 ) ; AChR molecules are supposed to alternate with AChE molecules in a mosaic pattern (Waser, 1967; Barnard e t al., 1971) , or they might simply represent functional groups of the same macromolecules that exert AChE activity (kupanGE, 1967). Hou ever, electron microscopic histochemistry demonstrates AChE activity not only at the postsynaptic, but also at the presynaptic, membrane, seriously jeopardizing this Olympic order. AChE at the presynaptic site apparently would inactivate the transmitter, right before it could have reached its site of destination, i.e., the postsynaptically located receptor. Thus, there is either a more sophisticated process involved in synaptic transmission or, rather, the presynaptic localimton of AChE is not meaningful (vulgo, it is a histochemical artifact). In order to overcome this difficulty (which he foresaw long before presynaptic AChE was actually l o c a l i d by electron microscopy) , Koelle forwarded an ingenious hypothesis ( 1961 ) , suggesting that ACh exerted both a presynaptic and a postsynaptic action. Accordingly, the first quanta of ACh, released by the axon terminal, would not cross the synaptic gap; but rather, acting upon the presynaptic membrane as a percussion cap, would release additional quanta of ACh. Only this secondary amount of the transmitter, released in such a high amount that it could survive the attack of presynaptic AChE, would pass the synaptic gap and combine with the receptor. The role of the presynaptic AChE would be to terminate the effect of the primary ACh, wherea? postsynaptic AChE would terminate the effect of the secondary amount of Ach. T h e concept of presynaptic amplification was later extended by Koelle (1962) ; in a more general form, it was applied also to other types of transmission mechanisms, leading to the idea of a “gencral neurotransmitter role”
76
UERTALAN CSILLIK
of acetylcholine. Such an arrangement would account for the presence of AChE in apparently noncholinergic synapses, like, for example, those in the cerebellar cortex. However, even though the basic mechanism underlying Koelle’s concept was proved to take place in synapses of the superior cervical ganglion, other areas (especially in the central nervous system) failed to obey this law. Also the pretransmitter role of ACh in liberating norepinephrine (Burn and Rand, 1959) could be proved only in a few species-specific instants. Thus, even though the interactions between different transmission mechanisms still represent a fruitful field for daring pharmacologists (Vizi, 1974), neither the theory of presynaptic amplification nor the general transmitter role of ACh has been generally accepted (Koelle, 1971). Accordingly, the role of presynaptic AChE and, especially, its deleterious effect upon the freshly released transmitter, remained an enigma. Ironically, even the role of postsynaptic AChE in the neuromuscular junction (where, until recent times, it held the strongest foothold) appears to be shaky, or at least much less simple than was supposed earlier. Neuromuscular impulse transmission proceeds normally in rats injected with sublethal doses of organophosphates; the only sign of AChE inhibition is the absence of posttetanic potentiation after supramaximal (tetanizing) volleys (Gerebtzoff et al., 1964). Recent studies by Ferry and Marshall (1971) prove that, if junctional AChE is inhibited by antiChE drugs (edrophonium, BW 284 C 5 1, dyflos, ecothiophate, neostigmine) , neuromuscular transmission is not effected under normal conditions of stimulation, i.e., when relatively small amounts of transmitter are released. Apparently, diffusion alone is sufficient to dispose of the surplus ACh. Only at higher frequencies of stimulation, when large amounts of transmitter are released, leads AChE inhibition to any marked prolongation of ACh action. O n the other hand, if endplate AChE is detached from the postsynaptic membrane by collagenase treatment (Hall and Kelly, 1971 ; Betz and Sakmann, 1971 ; McMahan et al., 1972), resting potentials, action potentials and the input resistances were unchanged as compared to control muscles; it was only the duration of the end-plate potential that was prolonged in such preparations. Furthermore, not only the role, but also the origin, of postsynaptic AChE in the neuromuscular junction is puzzling enough to deserve a more thorough investigation. Does it arrive from presynaptic stores, crossing the synaptic gap like the transmitter? But, if so, how does it survive degeneration of the motor axon? O r is it synthesized in the postsynaptic cytoplasm of the sole plate; but, if so, why does the sole plate not exhibit any AChE activity? Finally, what is the relation between AChE and AChR; how about their fine structural localizations in the postsynaptic membrane? How do these macromolecules fit into the thin protein layer provided by the unit mem-
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
77
brane, still widely regarded as the basic structural arrangement of cell membranes? And, most importantly, what kind of molecular anatomical events take place during the release of ACh; how, or why, does the presynaptic contingent of AChE not interfere with these events? Thus, it is by no means like forcing an open door if we attempt to review the structural aspects of cholinergic transmission, with special regard to the synaptic localization of AChE, in order to answer the above questions. The model system for this review is the spinal motoneuron, one of the few wellproved cholinergic neurons that have been an Archimedean point for neurophysiologists for nearly a hundred years-a solace for the frustrated histochemist.
II. Histochemistry of Acetylcholinesterase in the Spinal Motoneuron
A. THEPERIKARYON Located in lamina I X of the Rexed system, motoneurons occupy a pool measuring about one-third of the ventral gray. This area, in the lateral aspect of the ventral horn, contains only a few smaller-sized cells besides the large motor nerve cells. The dendritic system of the motoneuron extends mainly in the rostrocaudal direction; a fact noted as early as by Ram6n y Cajal ( 1911 ) and emphasized recently by the Scheibels ( 1966). Motoneuronal axons proceed via ventral roots to spinal nerves. I n addition to these large A a motoneurons, smaller motor nerve cells, called Ay cells, appear in a restricted number in the same location. While Aa motoneurons innervate extrafusal striated muscle fibers, Ay motoneuTons furnish motor innervation of intrafusal muscle fibers of muscle spindles. Though both Aa and Ay motoneurons are cholinergic, and thus they share common histochemical characteristics, for the sake of simplicity, only ACY motoneurons will be dealt with in detail in the following. Light microscopically, motoneurons exhibit the shape of a characteristic multipolar cell ; accordingly, in histological sections, their perikaryon appears to be multangular, with slightly concave surfaces. The perikaryon contains clusters of basophilic material, known as the Nissl substance ; these clusters are, in fact, highly organized samples of granulated (rough-surfaced) endoplasmic reticulum. Mitochondria are present both within these clusters and in the hyaloplasm among them. In contrast, the Golgi apparatus, consisting of regularly arranged cisterns and vesicles, is located always within the spaces left free from Nissl granules. Both the Nissl granules and the Golgi cisterns extend into larger dendrites. These latter are characterized by the regular, parallel arranged system
78
BERTALAN C.SILI.IK
of neurotubulcs and neurofilaments, present also, in a less regular arrangement, in the perikaryon. The motoneuronal surface, both of the perikaryon and the dendrites, is covered by numerous synaptic boutons. Nonsynaptic areas, as a rule, are covered by glial and feet (Fig. 4 ) .
FIG. 4. Electron micrograph of a spinal motoneuron (Macacus rhesus). Note nucleus ( N ) , rough-surfaced endoplasmic reticulum (r-ER) , constituting Nissl bodies, and synaptic boutons ( B ) on the surface of the cell membrane.
SYNAPTOCHEMISTRY OF A C:HOLINERGIC N E U R O N
79
Synaptic boutons, impinging upon the perikaryon, contain either spherical or flattened vesicles. Routons synapsing with dendrites are equipped with spherical vesicles only. Dense-core vesicles, usually of the larger (800-900 A ) variety, can be found in various numbers in both kinds of boutons (Bodian, 1966). An outstanding structural characteristic of axosomatic synapses is the occurrence of sandwichtype boutons. I n such cases, a second bouton makes synaptic contact with the first one, impinging upon the perikaryal surface. Both the first- and the second-order boutons of such sandwich synapses contain spherical vesicles. Indirect evidence, obtained mainly from neurophysiological experiments performed with intracellular recording techniques in the cerebellar cortex, suggests that synaptic boutons containing spherical vesicles are excitatory whereas those equipped with flattened vesicles exert postsynaptic inhibition. I t should be pointed out, however, that this idea, originally suggested by Uchizono (1965), was, in fact, never directly proved with regard to rnotoneurons. Sandwich boutons are generally assumed to represent the structural basis of presynaptic inhibition; recent studies (Magherini et al., 1971 ) prove that, in contrast to previous doubts, presynaptic inhibition is, in fact, involved in the functional organization of motoneuronal performance. In all the aforementioned types of synapses, the postsynaptic membrane exhibits a well defined thickening ; from this, a postsynaptic “web” extends into the motoneuronal soma. Beneath this web, as a rule, a Nissl granule can be seen; arrangement of the endoplasmic cisterns is strictly parallel to the synaptic surface, giving rise to a “subsurface cistern” in most cases. T h e presynaptic membrane is equipped with conspicuous projections, producing a structure similar to an egg-crate (Akert et al., 1974). Within this “crate”, synaptic vesicles are arranged in a quasi-geometrical form. The synaptic gap between boutons and perikaryal cell membrane measures -200 8, or slightly more. Within the gap (or cleft), there can often be seen a system of transversely oriented filaments; the role of this “intersynaptic organelle” is, in spite of repeatedly discussed proposals, still obscure. T h e axon of the motoneuron originates from an area of the perikaryon called axon hillock. Within this area, Nissl granules are absent. The surface membrane of the axon hillock and that of the initial segment of the axon, emerging from this area, is structurally different from other surface membranes of both the perikaryon and the rest of the axon (Palay et al., 1968; Peters et al., 1968), suggesting a specific role of this area in impulse initiation. AChE has been located light microscopically in motoneuronal perikarya ever since the very beginnings of neurohistochemical studies, both in tissues
80
BERTALAN CSILLIK
(for reviews, see Koelle, 1963; Silver and Wolstencroft, 1971) and in cells (Hosli and Hosli, 1971; Kim, 1971) . Electron histochemical investigations, performed by means of various modifications of the acetylthiocholine technique (Lewis and Shute, 1964, 1965, 1966; Eranko et al., 1967; Navaratnam and Lewis, 1970) revealed that AChE is located within the cisterns of the granulated (rough-surfaced) endoplasmic reticulum (Fig. 5). Clusters of rough-surfaced endoplasmic reticulum (r-ER) constitute the Nissl gran-
FIG. 5. Electron microscopic localization of AChE in the cytoplasm of a spinal motoneuron (rat). N, nucleus; Nucl, nucleolus; both are devoid of any enzyme reaction. AChE activity is confined to the cisterns of the rough-surfaced endoplasmic reticlum (r-ER) .
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
81
FIG. 6. High power electron microscopic localization of acetylcholinesterase (AChE) in the cytoplasm of a cholinergic neuron (rat). Note that the enzyme reaction within the cisterns of the rough-surfaced endoplasmic reticulum (r-Er) spares globular nonreacting “endoplasmic units” (Csillik and KnyihAr, 1968b), probably related to the final modeling of the tertiary structure of AChE molecules.
ules, displaying AChE activity at the light microscopic level. Variations in the intensity of the AChE reaction in individual cisterns of individual motoneurons may reflect genuine differences between these cells ; or, alternatively, may be due to differences between various areas of the r-ER; may suggest different functional states (Hebb and Morris, 1969) or may depend, simply, on various extents of osmic acid decolorization, brought about by postfixation, i.e., the secondary application of buffered osmium (Reale and Luciano, 1964). Invariably, however, the enzyme reaction product within the cisterns of the r-ER outlines nonreactive globular units, measuring 250-300 A in diameter, possibly related to the final modeling of the tertiary protein structure of the AChE molecules (Csillik and Knyihir, 1968b; Fig. 6 ) . Ultramicrosassay of AChE by means of the Cartesian diver technique (Giacobini, 1959) proves that the cytoplasmic localization of AChE is a substantial one and not due to any histochemical artifact. I n tissue homogenates (Toschi, 1959; Aldridge and Johnson, 1959) cytoplasmic AChE was found in the microsomal fraction; it can be separated from AChE bound to synaptosomal membranes (Whittaker, 1969a). Also synaptic boutons impinging upon motoneuronal somata were reported to stain for AChE, both light microscopically (Gerebtzoff, 1959; Koelle, 1963) and electron microscopically (Eranko et al., 1967). Since the transmission mechanism activating motoneurons was never convincingly proved to be cholinergic, this activity might well be due to noncholinergic synapses.
BERTALAN CSILLIK
B. THEAXON The main axon of the motoneuron proceeds via the ventral root to the spinal nerve, in order to innervate cross-striated muscles. An initial axon collateral abandons the main axon within the ventral gray or after a short distance, while traveling within the white matter. This initial axon collateral plays the key role in the recurrent (Renshaw) inhibition of motoneurons (Eccles, 1964) . The structure of the rnotoneuronal axon is identical with that of other thick myelinated nerve fibers. Surrounded by the axolemmal membrane, its axoplasm contains neurotubuli and/or neurofilaments and elongated mitochondria, aligned with the tubular and filamentous components. The main axon acquires a myelin sheath shortly after emerging from the motoneuronal pool ; the sheath consists of concentric gliocellular membranes while within the boundaries of the central nervous system; at the level of the pial membranes, it gives place to a morphologically similar laminated myelin sheath built up of Schwann cell membranes. AChE activity of motor nerve fibers has been detected light microscopically during the early years of AChE histochemistry (for details, see Koelle, 1963). Recent studies by Gruber and Zenker (1973) confirm earlier observations and prove that, even though sensory axons, too, exhibit AChE activity to a lesser extent, they can easily be distinguished from the strongly active motor nerve fibers when appropriate histochemical parameters are used. Electron microscopically, AChE activity of motoneuronal axons is confined to the axolemmal surface membrane. This localization, demonstrated for the first time by Torack and Barrnett (1962), using the thiolacetic acid procedure, has been repeatedly confirmed by numerous authors, using more specific histochemical substrates, including acetylthiocholine. Occasionally intra-axonal structures may exhibit AChE activity that has been regarded a sign for AChE being en route. On the other hand, staining of myelin lamellae (or more properly, thin intralamellar lines exhibiting AChE reaction) is probably only a diffusion (or precipitation) artifact (Fig. 7 ) . I n ligated axons, where axoplasmatic transport is subjected to an artificial blockade, various axoplasmatic constituents and enzymes, including AChE (Lubinska and Niemierko, 1971), are accumulated above and below the ligature. Biochemical investigations concerning AChE accumulation proximal and distal to ligatures were recently summarized and discussed by Fonnum et al. (1973). On the basis of electron histochemical investigations, it has been suggested that, in ligated axons, AChE is associated with neurotubules ( K h , 1968), probably related to an axonal transport mechanism of the enzyme. However, the structures exhibiting AChE activity in such electron micrograms are not identical with characteristic micro- (or
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
83
FIG. 7. Electron histochemical localization of AChE in motoneuronal axons within the rat’s spinal cord. Note the delicate localization of the enzyme reaction a t the axolemmal surface membrane (arrow). M, mitochondria; Nt, neurotubuli; Nf, neurofilaments; My, myelin sheath. ( A ) Cross section; ( B ) longitudinal section.
84
BERTALAN CSILLIK
neuro-) tubules; they are rather “irregular structures resembling smooth endoplasmic reticulum,” as observed recently by Davison (1970). Thus, the morphological substrate responsible for the transport and renewal of axolemma1 AChE is obscure for the time being.
C. AXONTERMINALS: THE MOTOREND PLATE Terminals of spinal motoneuronal axons fall, in essence, in two categories. The main axon, undergoing dichotomous divisions during its course to the musculature, gives rise to varying numbers of motor end plates. Dichotomization of the axon occurs, as a rule, at the nodes of Ranvier. The number of terminals a motoneuron gives off is dependent on the functional characteristics of the muscle ; motoneurons innervating external ocular muscles or the tongue, give off only a few, or even single, end plates, obviously because of the fine functional precision demanded by these muscles; on the other hand, large voluminous muscles of the pelvic girdle, like the gluteus, that do not require such a fine functional coordination, are innervated at much rougher scales, one motoneuron giving rise to 200-300 end plates. Axons of Ay motoneurons (page 77) terminate on intrafusal muscle fibers in muscle spindles, both in mammals and in amphibia; in addition, tonic muscles of amphibia are innervated also by axons of Ay motoneurons. These endings are known as terminaisons en grappe as contrasted to the terminaisons en plaque of Aa motoneurons. The main axon of the motoneuron gives off a recurrent initial axon collateral right in the ventral gray or during its passage through the white matter. Terminals of these initial axon collaterals constitute the Renshaw elements, impinging upon dendritic dilatations of Renshaw cells (Section 11, D) (Fig. 8 ) . 1. T h e Motor End Plate: History and Histology
Discovery of nerves terminating in striated muscle (DoyCre, 1840) was an important milestone in the history of neuroanatomy. I n contrast to the belief held by most physiologists of that time, namely, that motor nerve fibers form closed loops in the muscles and return to the spinal cord as sensory fibers, Doy2re was the first to prove that nerve fibers actually ended in tiny eminences on the surfaces of muscle fibers. Metal impregnation techniques, introduced in the mid- 1850s, enabled contemporary histologists to distinguish two main components of the motor nerve ending. One of these, the Geweih of German scientists (arborisation terminale of the French) is the nerve terminal proper; the other, called Sohlenplatte (semelle) was found below and around the nerve terminal. We will refer to this latter as sole plate. For detailed literature, see Csillik ( 1965, 1967).
SYNAPTOCHEMISTRY O F A CHOLINERGIC NEURON
85
FIG. 8. Geometry of the spinal motoneuron. T h e main axon ( A ) , undergoing several dichotomizations, innervates striated muscle fibers by means of neuromuscular junctions. As initial collateral ( I C ) returns to the ventral gray, in order to innervate Renshaw cell(s) ( R ) . Inset: the actual innervation of the Renshaw cell by the motoneuronal axon collateral. Dendritic dilatations of the Renshaw cell are surrounded by initial axon collaterals, establishing Renshaw elements. T h e Renshaw cell, in turn, innervates a neighboring motoneuron (MN?) by means of a n inhibitory nerve ending.
While the nerve terminal proper had been extensively studied, no major studies were performed on the sole plate until the 1940s. Using Janus green as a supravital stain, Couteaux (1917) was first to discover that around and below the axon terminal the muscle surface membrane constitutes a peculiar structure, consisting of hundreds of semicircular (doughnutlike)
86
B E RT AL AN CSILLIK
organites. This subneural apparatus was shown subsequently to be the site of AChE activity at the light microscopic histochemical level (Couteaux, 1951 ; Couteaux and Taxi, 1952). The great epoch of electron microscopy that followed Sjostrand’s important technical innovations (plastic embedding and glass knife ultramicrotomy ) , soon furnished information on the fine structural arrangement of the motor end plate. Organization of the neuromuscular “ ~ y n a p s e ”was ~ first summarized in Robertson’s famous diagram ( 1956)-an outstanding example of lasting, correct interpretation of relatively poor electron microscopic pictures. The electron microscopic structure of a mammalian motor end plate (Fig. 9 ) is characterized by the terminal axon, containing synaptic vesicles and mitochondria; the axon proceeds in a groove or gutter furnished by the postsynaptic membrane, which, in the case of a-motor endings, is thrown into multiple junctional folds. The cytoplasmic area between the postsynaptic membrane and the contractile elements of the muscle is the sole-plate; in addition to nuclei, mitochondria, a few cisterns of r-ER, and a great number of free ribosomes are characteristic of this area.
2. Light and Electron Microscopic Histochemistry Using appropriate substrates, pH, temperature, and incubation time, AChE appears to be concentrated in the subneural apparatus, i.e., in the postsynaptic membrane, of the motor end plate. For detailed accounts of the applicable techniques, see the reviews by Gerebtzoff (1959), Koelle (1963), Zacks (1964), and Csillik (1965, 1967). Sizes and shapes of the AChE-active subneural structures differ according to species and muscle (Namba, 1971). In spite of this, the fundamental structural arrangement characterizing mammalian motor end plates can be summarized as follows. The structure is usually pretzel-like, measuring 30-40 pm in diameter. AChE activity is exerted by thousands of tiny, semicircular organites, located normal to the plane of the muscle surface and normal also to the course of the axon terminal. These organites constitute AChEactive gutters (troughs) accommodating for the terminal nerve fiber. In between the “holes” of the pretzel, and around the entire structure, nuclei belonging to the muscle fiber (sole-plate nuclei), or to the Schwann envelope (teloglial nuclei) are present; nuclei do not display AChE activity at optimal incubation parameters but can be seen by phase contrast or interference contrast (Nomarski) optics, or after staining with basic dyes. Neither does the sole-plate cytoplasm stain for AChE (Fig. 10A). a T o use the term synapse for the junction between a neuron and a muscle cell may be a barbarism; to avoid it on the grounds of pedantry would be a purism.
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
87
FIG. 9. Electron microscopic structure of a neuromuscular junction (m. flexor digitorum brevis, r a t ) . T h e terminal of the motor axon ( A T ) appears in cross section; within this profile, mitochondria ( M ) and synaptic vesicles (sv) are present. Pre, presynaptic membrane (axolemma) ; post, postsynaptic membrane (cell surface membrane of the cross-striated muscle fiber). Note the homogeneous, slightly electron dense material (basement membrane, B M ) within the synaptic cleft. JF, junctional folds of the postsynaptic membrane containing BM material ; R, free ribosomes in the sole-plate area. Arrows point to globular structures resembling “endoplasmic units” in the secondary synaptic gap.
88
BERTALAN CSILLIK
FIG. 10. Light microscopic histochemical appearance of AChE in neuromuscular junctions. ( a ) Rat, gastrocnemius (tetanic muscle, a-junction) ; ( h ) frog, thoracohumeral (tetanic muscle, a-junction) ; ( c ) frog, Tonusbiindel of the iliofihular (tonic muscle, y-junction) . I n a-junctions, the suhneural apparatus displaying AChE activity consists of semicircular (doughnutlike) organites, arranged normal to the surface of the muscle fiber and normal also to the course of the axon terminal. I n mammalian junctions, only the tips (arrow) of the relatively small organites can he distinguished, whereas in amphibians, owing to their size, individual organites can clearly he identified. As follows from polarization optical and electron microscopic reconstruction, organites consist of fingerlike pouches (mammals) or of parallel sheets (amphibians) that give rise to the junctional folds characterizing electron micrographs. I n contrast, in the grapelike y-junctions, innervating intrafusal muscle fibers of muscle spindles, and also tonic skeletal muscles of amphibians, the postsynaptic membrane is smooth; even if a few shallow junctional folds are present, these do not constitute organites.
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
89
I n addition to AChE, also nonspecific esterases and pseudoChE were demonstrated in neuromuscular junctions (Barron et al., 1967 ; Eranko and Teravainen, 1967; Chokroverty et al., 1971). The function of these enzymes is unknown ; electron microscopically they are located in the synaptic gap of the junction (see below). The above structural and histochemical arrangement can be found most characteristically in large tetanic muscle fibers of mammals (e.g., gastrocnemius and diaphragm of the r a t ) . In tonic muscles, the subneural structures are more delicate, more fragile, and exert a slightly weaker AChE activity. In small muscles of the extremities (e.g., flexor digitorum brevis of the rat) the light microscopic appearance of the subneural apparatus is much simpler; however, in all cases, it can readily be seen that it consists of “organites” that, under the electron microscope, correspond to multiplex systems of junctional folds, both in tetanic and in tonic mammalian muscles (Duchen, 1971). I n tetanic muscles of amphibians, organites are extremely large (4-6 pin) and, according to the straight courses of the terminal nerve fibers, they are arranged in a geometrically regular, palisade-like order. I n essence, however, there is no difference between their arrangement as compared to that of mammalian tetanic muscles (Fig. 1OB) . An entirely different kind of subneural structure characterizes the endings of the small nerve system, originally described by Kuffler (1953). Accordingly, in tonic muscles of amphibians and in intrafusal muscle fibers of both mammals and amphibians, AChE activity is confined to grapelike accumulations of spherical units, measuring 2-4 pm in diameter. (Fig. 1OC). These latter are not composed of organites, however; the spiny appearance characterizing a-motor endings can never be observed in these y-subneural apparatuses. Electron microscopy reveals that this difference is due to the absence of junctional folds in y-endings. I t should be noted that, under identical parameters of histochemical techniques, y-subneural apparatuses display a distinctly weaker AChE activity than a-subneural structures (Csillik et al., 1961 ; Heaton et al., 1972). Electron microscopic histochemistry of end-plate-bound AChE has been recently reviewed by Friedenberg and Seligman ( 1972). I n essence, the chromogenic techniques developed for the electron microscopic localization of AChE are either ( a ) variants of the Koelle acetylthiocholine procedure, or belong to one of the following groups: ( b ) thiolacetic acid techniques; ( c ) hexazotized pararosaniline diazo coupling methods; ( d ) osmiophilic thiol ester methods. I t will be noted that, except for the thiocholine technique, the chromogenic substrates of all the other reactions are distinctly different from the physiological substrate of the enzyme; a compromise dedicated to overcome the poor tissue penetration of thiocholine esters, due to the positively charged N’ atom (Fig. 11 ) . Thus, these substrates represent
90
BERTALAN CSILLIK CH CH~-P-%H,-CH CH3
-S-CO-CH~
-
Acetylthiocholine (AThCh) 0-CO-CH-
CH CH -Y&H 3CH3
,-cH~-
-
s- C O - C ~ H ~
aButyrylthiocholine (BuThCh)
0-CO
a-Naphthyl acetate
- CH3
0-Naphthyl acetate
Indoxyl butyrate
Indoxyl acetate CH3-CO-SH
0-
Thiolacetic acid
( CI-I3-C0) 2s Diacetyl sulfide
CH2-5-GO-CH3
N+or~N-
d
Acetylthiol- m -toluene diazonium (ATTD)
0 -CO
-CH3
3-Acetoxy-5-indole diazonium (AID)
c q C,H3 G ,- - C H
C H 2- S
- CO - C
H
.CH3 Quaternary carbon analog of acetyl thiocholine
FIG. 11, C h e m i c a l s t r u c t u r e s of c h r o m o g e n i c s u b s t r a t e s u s e d for acetylcholinesterase (and e s t e r a s e ) histochemistry.
the two basic trends of AChE histochemistry in general; the only difference is that, in electron histochemistry, the end product has to be electron dense, in contrast to the optical density required from light microscopic techniques. Each of the above substrates has its merits and its drawbacks; sound conclusions as to the actual localization of AChE can be drawn if one is using any of these substances in a sufficiently sophisticated system, whereas all suffer from the basic handicap of chromogenic techniques employing substrates similar to, but not identical with, the physiological substrate of the enzyme. O n the other hand, it should be kept in mind that, although drastic artifacts in enzyme localization due to tissue disorganization can be avoided
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
91
by using appropriate techniques for tissue preparation, there always remain the prospective pitfalls caused by moderate tissue penetration, by impregnation and by tissue substantivity (see Bogusch, 1973). These are, unfortunately, inherent in the very chemistry of the substrates used; and, therefore, it is our belief that none of the electron histochemical techniques do ever allow for a “direct,” a t first glance acceptable, pictorial demonstration of enzyme localization. I n this respect, the fundamental feature of electron microscopy, viz. its indirect character (see page 97) is even more accentuated by the complicating factors of substrate aspecificity. In order to illustrate this point, it should be recalled that, at the light microscopic level, all these techniques reveal AChE activity of the subneural apparatus. At the electron microscope level, however, there are wide variations of the fine structural localization within the light microscopic entity of the synaptolemma. Activity was reported in the pre- and postsynaptic membranes or both, or in the junctional cleft between them. The preferential localization obtained by any of these techniques is dependent, in addition to the specificity of the substrate, on the following factors: ( a ) chemistry and duration of the fixation; ( b ) pH, ionic strength, and duration of incubation; ( c ) chemistry and duration of postfixation. Thus, the initial uncertainties outlined above are increased in a logarithmic order. Since a priori proofs for a real localization as differing from an artifactual one, are not available, it is only logical reasoning that can answer this vital question. Keeping in mind all the above shortcomings of AChE electron histochemistry, it may appear a bold suggestion that, even if one uses appropriate parameters and exerts sharp criticism, it will be possible to arrive at a definite conclusion as to the molecular anatomical localization of AChE in the neuromuscular junction. I n spite of this, as exemplified on the basis of our own studies (Csillik and Knyihzir, 1968a), we shall attempt to achieve this goal (Figs. 12-16). When the lead-copper-uranylthiocholine variant of the Koelle procedure is used on formalin-fixed muscles, electron microscopy reveals that the end product of the reaction is deposited on the external surfaces of the preand the postsynaptic membranes of the neuromuscular junction. If a very short fixation is followed by a similarly short incubation, a clear-cut localization of AChE can be achieved; in this case, electron dense granules are located on both membranes, facing the (virtually inactive) synaptic cleft material (Fig. 1 2 ) . Density of granules is identical per unit length of the membranes, both in the primary and the secondary synaptic areas (i.e., the intensity of the reaction is identical in presynaptic and postsynaptic menibranes. the latter establishing the junctional folds) . Accordingly, AChE is not preferentially located in the postsynaptic membrane as suggested by light microscopy; it is only the intricate system of the junctional folds, exceeding
92
RERTALAN CSILLIK
12
13
14
15
16
SYNAPTOCHEMISTRY O F A CHOLINERGIC
NEURON
93
the membrane material of the presynaptic membrane by a factor of at least 10 in a-motor end plates, that results in an overwhelming accumulation of the enzyme at the postsynaptic side.4 Prolonged incubation of the sections (which is inevitable to achieve any reaction if a substantial fixation is preferred in order to preserve ultrastructural integrity of other tissue components) will result in a more or less gross diffusion of the enzyme reaction product throughout the synaptic cleft. (Fig.
13). When using indoxyl acetate as a substrate, either after short or long prefixation, the reaction end product will be uniformly distributed within the synaptic cleft, without any prefrrential localization in any component of the synaptolemma (Fig. 14). ‘ I n y-subneural apparatuses, equipped with only a few, if any, shallow folds, the amount of presynaptic AChE is virtually identical with that of the postsynaptic enzyme contingent. Apparently, this is the reason why y-subneural apparatuses display a much weaker AChE activity in light microscopical histochemical sections than do a-subneural apparatuses. FIGS. 12 and 13. Electron microscopic histochemical localization of AChE in neuromuscular junctions of the rat diaphragm. Acetylthiocholine technique; coupling with Cu-Pb-U02. FIG. 12. brief prefixation in formaldehyde, 5 minutes of incubation. Ultrastructural preservation is obviously poor; yet the localization of the enzyme reaction product at the external surfaces of pre- and postsynaptic membranes is evident. Within the synaptic cleft, only a minor precipitate occurs. (Figures read from top to bottom.) FIG. 13. Substantial fixation ( 3 hours) in formaldehyde, 15 minutes of incubation. Ultrastructural preservation is improved ; note mitochondria ( M ) and synaptic vesicles (sv) within the axon terminal. Owing to the longer incubation, however, the enzyme reaction product is no longer membrane-bound; rather it fills the synaptic cleft. A membrane-bound localization can be observed only a t the site marked by a n arrow (Csillik and KnyihBr, 1968a). FIG. 14. Electron microscopic localization of indoxyl acetate esterase in a neuromuscular junction of the rat diaphragm; 3 hours’ fixation in formaldehyde; 10 minutes’ incubation. Enzyme reaction product fills the primary and the secondary synaptic cleft. Note synaptic vesicles (sv) and mitochondria ( M ) in the axon terminal, the junctional folds (JF) of the postsynaptic membrane, nuclei of teloglial cells ( N ) , and myofibrils ( M f ) . Z, Z lines. FIG. 15. Electron microscopic localization of indoxyl butyrate esterase in a neuromuscular junction of the rat diaphragm; 3 hours’ fixation in formaldehyde, 10 minutes’ incubation. Enzyme reaction clearly outlines the middle dense layer, or basement membrane, of the synaptic cleft. M, mitochondria; sv, synaptic vesicles; JF, junctional folds; Mf, myofibrils. FIG. 16. Electron microscopic localization of thiolacetic acid hydrolase in a neuromuscular junction of the rat diaphragm ; 3 hours’ fixation in formaldehyde, 10 minutes’ incubation. Enzyme reaction is mostly confined to the postsynaptic membrane; also the presynaptic membrane and several synaptic vesicles (sv) exert some reaction. JF, junctional folds.
94
BERTALAN CSILLIK
Another nonspecific substrate, indoxyl butyrate, results in a clear-cut localization of the enzyme reaction in the middle dense layer of the synaptolemma1 complex, i.e., in the basement membrane consisting of the joint glycoprotein external coats of pre- and postsynaptic membranes (Fig. 15). Finally, a membrane-bound localization of the enzyme reaction can be observed when thiolacetic acid is used as substrate,j or aromatic substrates with a thiolacetic function (Davis et al., 1972). I n this case, however, the reaction is heavier on the postsynaptic membrane. At the same time, also several synaptic vesicles are stained; Barrnett (1962) suggested that this might be due to their choline acetylase activity (Fig. 16) . Essentially, all the electron histochemical investigations published hitherto revealed one or the other of the fine structural localization of AChE in neuromuscular junctions described above. Therefore, without trying to discuss the plausibility of any of the fine structural localizations achieved by others, we shall attempt to analyze our own studies in a critical way. O n the premises of relevant biochemical studies, we will try to show that the localization seen after a short fixation and a short incubation with acetylthiocholine reflects the very biological organization of the postsynaptic membrane.
3. Double-Membrane Localization of A C h E : Fact or Artifact?
+
One of the main arguments against a double-membrane (presynaptic postsynaptic) localization of neuromuscular AChE is that it can be visualized only by means of the acetylthiocholine technique. Is it not a simple metallic impregnation of adjacent membranes that show up in the shape of a double-membrane localization? And is it not only tissue substantivity that induces a preferential staining of the synaptic gap material when using nonmetallic, azo-dye techniques? Is there any real reason to assume that there is more than one hydrolytic enzyme associated to the junctional complex; is it not the very same enzyme, the fine structural localization of which remains unknown (ignoramus and ignorabimus) , since it displays different localizations when using different histochemical techniques? a. Some of the nonspecific substrates, such as a-naphthylacetate and indoxyl acetate (if used on frozen sections, and the hydrolysis product is coupled with hexazotized pararosaniline to yield an azo dye), result in a homogeneous localization within the primary and the secondary synaptic clefts, without any preferential localization at any components of the synaptolemma. O n the other hand, indoxyl butyrate (used under identical parameters, and the resulting hydrolysis product, identical with the indoxyl that According to Koelle et al. (1968), the actual substrate giving rise to this reaction is diacetyl sulfide, a common contaminant of commercial thiolacetic acid preparations.
95
SYNAPTOCHEMISTRY O F A CHOLINERGIC N1:UKON
results from the hydrolysis of indoxyl acetate coupled similarly with hexazotized pararosaniline) , gives rise to a clear-cut concentration of the azo-dye end product in the middle dense layer of the synaptolenima, accompanied by only a weak, diffuse staining, if any, uithin the rest of the synaptic gap. Since the end products of both the indoxyl acetate and the indoxyl butyrate reactions are chemically identical, as are also other parameters of the incubation, the differential localizations of thc two reactions cannot be traced back to tissue substantivity. The only logical explanation of the results is that acetate-splitting arylesterases are located throughout the synaptic cleft whereas the middle dense layer of the synaptolemma is the specific site for arylesterases hydrolyzing substrates esterified with a longer carboxylic acid. Although the biological importance of either arylesterase in the synaptic gap is unknown (except for the possibility of hydrolyzing "waste products" leaking from the synaptically coupled elements), the preferential localization of two different enzymes in two different layers of the gap indicates topochemical specificity within a very restricted anatomical area (Fig. 1 7 ) . b. Similarly, the ultrastructural localizations achieved by using- acetylthiocholine and thiolacetic acid as substrates and Pb" as a capturing agent differ from each other characteristically. Acetylthiocholine, if used after a
(cH,),-N+
-cH,-CH,--S
-CO-CH,
0 -CO - CH,
(hexazo)
(Cu, Pb, U)
OT0
-CO-C,H, (hexazo)
CH,CO-SH
or (CH,CO),S (Pb)
FIG. 17. Summary of the electron microscopic localizations of AChE and other esterases in the neuromuscular junction (Csillik and Knyihir, 1968a).
96
BERTALAN CSILLIK
very short fixation that allows a short incubation period, results in a doublemembrane localization, whereas thiolacetic acid, under similar parameters, gives rise to a strong postsynaptic activity. T h e end products of both reactions are identical (PbS) ; thus the differences between the actual fine structural localizations of the reaction end product prove again the validity of the double-membrane localization of AChE. ( T h e apposed membranes are liable for a nonspecific metallic impregnation in both cases; the difference between the two patterns proves that the localization is not due to an impregnation artifact.) c. More difficult is to rule out the possibility that it is neither the metallic ions themselves nor the metallosulfide end product that bind with the membranes, but the molecules of the substrate, acetylthiocholine, or the immediate breakdown product, thiocholine. N+ in these substances has a positive charge (cationic head) ; it can be assumed that receptor moieties in any membranes have a preferential affinity for the cationic head of the substrate. I n this case, be it hydrolyzed before or after this combination, thiocholine would inevitably give rise to a double-membrane localization of the histochemically demonstrable end product ; any further treatment (capturing reaction with copper, lead, uranyl, or gold ions) would actually visualize only the thiocholine molecule, already bound to the membrane ( s ) by ionic forces. This argument, relying upon the molecular structure of the substrate, so similar to the physiological transmitter molecule, is the most serious conjecture that was ever raised against the double-membrane localization of junctional AChE. In fact, there is no way to rule out such a possibility on the basis of logical reasoning, except for the admittedly very shaky counterargument that ionic forces that determine transmitter-receptor binding in uiuo, are probably less active after aldehyde fixation. In this context, however, it should be noted that pretreatment by d-tubocurarine, does not change either the light microscopic appearance or the electron histochemically characteristic double-membrane-bound AChE reaction of the motor end plate. Since d-tubocurarine is known to interact with acetylcholine receptors, there is a good reason to assume that, in this case, cation-binding receptor moieties are blocked ; thus the binding of the substrate to the synaptic membranes by ionic forces is prevented. [Apparently, the d-tubocurarine-binding site of AChE (Belleau and DiTullio, 1971 ; Mooser and Sigman, 1972) does not interfere with the hydrolytic activity and with the double-membrane localization of AChE.] Thus, the electron histochemical patterns obtained after d-tubocurarine treatment suggest that AChE molecules are, in fact, located at (or on) both pre- and postsynaptic membranes. Whether, by the same token, this experiment would indicate also that enzyme and receptor are two distinct entities should be corroborated by more specific inhibitors of the AChR. Studies with a-bungarotoxin,
S Y N A P T O C H E M I S T R Y OF A CHOLINERGIC N E U R O N
97
the best available blocker of AChR in the neuromuscular junction, are in progress to confirm the results obtained with d-tubocurarine. O n the other hand, the fine structural localization of the reaction obtained by the quaternary carbon analog of acetylthiocholine (Nyberg-Hansen et al., 1969) might also prove helpful. While analyzing the fine structural localization of AChE in the neuromuscular junction, the basic question ol electron microscopy emerges : Is it the true image of the living structure what we see on the screen or not? This fundamental doubt is justified not only in electron histochemical studies (where, owing to various manipulations, to substrate nonspecificity and to end-product substantivity, the chances of an artifact are logarithmically increased), but also in any kind of the so-called direct visualization of submicroscopical organization, close to the resolving power of the technique, i.e., in the range of molecular anatomy. Several years ago (Csillik, 1967) admittedly in a quite naive way, I tried to compare these difficulties to Heisenberg’s uncertainty principle. Anyway, even dispensing with mathematical formulas, it is not difficult to realize that, since a priori proofs do not exist (whether or not the pattern seen in the electron microscope is identical with the organization existing in the living state or even a t the moment of the cellular death), our conclusions are subject to structural deformations caused all the steps of microtechnical manipulations by many factors-including and also the energy transmitted by the instrument of investigation. 4. Ontogenesis
AChE activity of motor end plates appears for the first time on the fifteenth embryonic day in the rat. At this age, motor axons outgrowing from the spinal cord exert intense AChE activity; also enzyme active are numerous cells streaming out of the neural crest that follow the course of spinal nerves. As soon as these (Schwann) cells arrive at the musculature, they settle at the sarcolemmal surface, in a close topographic relation to the immature motor ending formed by the AChE-active axon (Fig. 18). Light microscopic histochemistry suggests that, though a slight AChE activity is exerted also by the muscle fibers proper, the structure and enzyme activity of the subneural apparatus is mainly elaborated by the AChE-active motor axonal endings and the similarly AChE-active teloblastic Schwann cells (Zelenj and Szentigothai, 1957 ; Csillik, 1960) . An essentially similar sequence of events was observed in the course of ontogenetic development by other authors, too; both in viuo (for references, see Filogamo and Machisio, 1971; Lentz, 1969, 1970, 1972; Juntunen and Teravainen, 1972) and in vitro (Pappas Pt al., 1971; Koenig, 1973). Electron microscopic histochemical studies revealed that junctional AChE is derived from at least three different sources (Csillik, 1974). ( 1 )
98
BERTALAN CSILLIK
FIG. 18. Light microscopic appearance of AChE in motor axons and in cytoplasms of teloblastic Schwann cells in the striated musculature of a newborn rat. Arrows point to teloblasts ( T ) (nuclei counterstained with hematoxylin) ; arrows with asterisks point to fine bundles of motor axons.
The axolemmal membrane of the immature nerve terminal is a priori loaded with AChE. ( 2 ) An intense enzyme synthesis takes place in the perikaryon of the teloblastic Schwann cell. Here, the enzyme is concentrated mainly in the perinuclear cistern of the endoplasmic reticulum. It is not clear, however, how this teloblastic enzyme contributes to the general histochemical architecture of the neuromuscular junction.G ( 3 ) A third source of AChE is the sarcotubular system of the underlying muscle fiber. Enzyme synthesis appears to take place mainly in the longitudinal elements of the sarcotubular reticulum; on the other hand, transverse tubules seem to be involved in the transport of the enzyme to their definitive sites a t the postsynaptic membrane (Figs. 19-2 1) . Embryonic and early postnatal neuromuscular AChE is highly dependent on the innervation. Transection of the motor nerve, either before or shortly after birth, results in a rapid dispersion of enzyme activity at the surface of the muscle fiber and, within a few days, in complete disappearance of AChE activity. Whether this reflects the dependence of AChE on presynaptically liberated ACh or on other “trophic” factors is still open for discussion; some of the relevant arguments are summarized (Section 11, C, 6 ) . I n any I t should be noted that teloblastic AChE is slightly inhibited by DFP treatment ( 3 x lo-’ M ) ; thus, its sensitivity is between that of adult AChE and pseudoChE. I t may be assumed that, during ontogenesis, the chemical pattern of this enzyme undergoes alterations; probably the end result is the arylesterase that, in adults is located within the synaptic cleft (page 93).
FIG. 19 ( t o p ) . Electron microscopic localization of AChE in the perinuclear cistern of a lemmoblastic Schwann ccll, overlying a developing neuromuscular junction. Note nuclear pores (arrows). Three-day-old rat puppy. FIG. 20 ( b o t t o m ) . Electron microscopic localization of AChE in a neuromuscular junction of a 3-day-old rat puppy. F , junctional folds. Arrows point at enzyme reaction u ithin the sarcotubular system apparently supplying the postsynaptic AChE continSent. Within the primitive axon terminal ( A T ) , filamentous and/or tubular structures exert AChE activity, probably contributing to the presynaptic enzyme contingent. FIG. 2 1 (inset). Electron microscopic localization of AChE in the sarcotubular system underlying a developing motor end plate. lhree-day-old rat puppy.
100
BERTALAN CSILLIK
case, the behavior of embryonic (and early postembryonic) neuromuscular AChE is, in this respect, entirely different from that seen in adult animals, where neuromuscular AChE survives degeneration for an amazingly long time.
5 . Degeneration and Regeneration Long before the advent of AChE histochemistry, biochemical investigations revealed that neuromuscular ChE survives degeneration of the motor nerves for a considerable time (for a review of the pertinent literature, see Csillik, 1965, 1967) . Recent studies, using more sophisticated biochemical methods, confirmed these early studies in many details (Crone and Freeman, 1972). Also scores of light microscopic histochemical studies were performed during the last decades to disclose the histochemical and structural alterations of the subneural apparatus after denervation. Although the speed at which these alterations occur differs to some extent, the net result may be summarized simply in that the subneural apparatus and its AChE activity, though slowly disorganized in structure and with a decreased enzyme activity, survives several months after the nerves have completely degenerated. I n our own studies dedicated to this question (Sivay and Csillik, 1956), we observed a survival of the subneural apparatus in the rat gastrocnemius, after repeated resection of the nerves, u p to 6-8 months. Results obtained by other authors were summarized by Koelle (1963), by Csillik (1965, 1967), and by Tuffery (1971). Electron histochemistry proves that the first week after denervation is characterized by two outstanding features. First, AChE activity of the deteriorating presynaptic membrane disappears rapidly ; at the same time, degenerating fragments of the axoplasm exhibit thiolacetic acid hydrolase (probably proteolytic) activity. Structure and AChE activity of the postsynaptic membrane, including junctional folds, remains virtually intact (Fig. 2 2 ) . Structural disorganization of the junctional folds, accompanied by a decrease in their AChE activity, takes place in rat skeletal muscles only several months after denervation. This outstanding viability of the subneural apparatus explains the excellent regenerative capacity of the neuromuscular junction. Therefore, if regeneration takes place in an early period, i.e., when the subneural structures (junctional folds and their AChE activity) are still in a relatively healthy state, regenerating motor axons will simply reinnervate the abandoned postsynaptic structures (Csillik and Sivay, 1958). The possibility of such a simple reinnervation is of utmost importance with respect to peripheral neurosurgery (Liillmann-Rauch, 1971) . Only if regeneration of the motor axons is impeded (by experimental retardation, or, in clinical traumatology, by amputation neuroma) , and thus
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
101
FIG. 22. Electron microscopic localization of thiolacetic acid hydrolase activity in a neuromuscular junction, 11 days after transection of the motor nerve. Adult rat, diaphragm. Note the unchanged appearance of postsynaptic enzyme activity, especially in the junctional folds ; presynaptic activity disappeared completely. Signs of activity characterizing degenerating fragments of the late axon terminals are due, probably, to proteolytic enzymes.
the original postsynaptic structures become unsuited for a simple reinnervation, will new subneural apparatuses be established. Apparently, this elaborate and time-consuming process, described by us earlier (Csillik and SAvay, 1958) ensues also after treatment of the muscles with tetanus toxin (Duchen and Tonge, 1973), the significance of which will be discussed in more detail in Section 11, C, 6 ) . Electron microscopy of end-plate degeneration and regeneration has recently been described in detail by Gonzenbach and Waser (1973).Unfortunately, until now, electron histochemical studies were not performed to reveal the structural events resulting in the formation of new subneural appa-
102
BERTALAN CSILLIK
ratuses after a retarded regeneration. Such studies could reveal the source of freshly synthesized AChE as well as the morphogenetic factors responsible for the formation of junctional folds. Denervation induces also a conspicuous spread of AChR-s along the surface of the muscle fiber, resulting in the phenomenon of ACh supersensitivity (for references, see Csillik, 1965, 1967; Thesleff, 1973). Spread of AChR-s is not associated with any similar alterations in the localization of AChE. The only fine structural alteration that follows denervation is a widespread accumulation of free ribosomes and r-ER in the subsarcolemmal sarcoplasm throughout the lengths of denervated muscle fibers during the period of supersensitivity (Gauthier and Schaeffer, 1974). This is, probably, related to the synthesis of new AChR-molecules: in fact, nascent protein synthesis is increased after denervation (Kimura and Kimura, 1973). According to Grampp et al. (1972), de nouo synthesis of nonjunctional AChR molecules and, hence, denervation supersensitivity to ACh can be prevented by drugs that interfere with protein synthesis (e.g., actinomycin D ) .
6. Origin of Postsynaptic AChE: Induction and Gene Expression Ontogenetically, the postsynaptic contingent of AChE in the neuromuscular junction, is derived from sarcotubular and schwannocellular sources, whereas presynaptic AChE is established from the axolemmal (and axoplasmatic) elements. Once, however, this final arrangement of synaptolemmal cytochemistry is achieved, it has to be sustained for a lifetime. Resupply of presynaptic AChE is maintained, obviously, by transport mechanisms (Lubinska and Niemerko, 1971; Ranish and Ochs, 1972; Fonnum et al., 1973), carrying enzyme synthesized in the motoneuronal perikaryon. More difficult is to account for the resupply of the postsynaptic AChE, since the sole-plate area does not display any activity; neither do teloglial Schwann cells or the sarcotubular system of the underlying muscle fiber exert any AChE activity in adult animals. Theoretically, it could be assumed that the same transport mechanism that supplies presynaptic AChE, is responsible also for the maintenance of the postsynaptic contingent, by means of enzyme translocation (through the synaptic cleft) or by means of some mysterious enzyme reduplication. Since, however, under optimal electron histochemical parameters, there can be seen only a negligible amount of AChE activity within the synaptic cleft, translocation is improbable; on the other hand, the extremely long survival of postsynaptic AChE after motor nerve transection, would require an unbelievably long turnover (half-life) of the postsynaptic AChE, were it derived from presynaptic sources. Obviously, there is some other explanation for the origin of postsynaptic AChE. During the last decades, there accumulated an increasing body of evi-
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
103
dence that suggests that synthesis of AChE might be induced by its substrate. Jones et al. (1956) have shown that explants of chick embryo intestine, cultured in embryo extracts for several days, lost most of their ChE activity, unless ACh was included in the culturing medium. More specifically, Rurkhalter et al. (1957) found that addition of ACh to chick lung explants actually resulted in an increased AChE activity. These early observations have been, however, subjected to severe criticism. Burdick and Strittmatter (1965) did not observe any substrate induction by ACh of AChE in cultures of whole chick brains; Goodwin and Sizer (1965), studying tissue explants of chick embryonic skeletal muscle, concluded that cholinergic drugs did not evoke a true “inductive” response of AChE synthesis since their effect was restricted to the prevention of loss of AChE activity, normally encountered if the tissues are incubated alone. More recently, Turbow and Burkhalter (1968) found that, in cultures of chick spinal cord explants, ACh caused a n essentially similar prevention of the otherwise inevitable decrease of AChE activity. Although this might be related to substrate induction, it can equally well be explained also by other factors (increased cell membrane permeability, etc.) . Thus, these early studies, summarized and discussed by Guth (1968, 1969), suggest rather a trophir influence of nerve on muscle. More recent studies, however, appear to indicate a straightforward substrate induction. According to Filogamo and Marchisio (1971), ACh, released from nerve terminals is the determinant factor of AChE synthesis; AChE of the subneural apparatus is incapable of self-differentiating and, thus, depends on ACh released by presynaptic axons. For example, after irreversible inhibition of the avian end-plate AChE by DFP, the enzyme reappears only in normally innervated neuromuscular junctions, whereas in previously denervated ones, i.e., in end plates containing only degenerating fragments of the motor axon terminal, AChE activity does not reappear after DFP treatment (Filogamo and Gabella, 1961, 1967). Essentially similar results were obtained also in rodent muscle (Rose and Glow, 1967; Filogamo, 1969). O n the other hand, it appears that ACh-sensitivity is not a prerequisite for the morphogenesis of the neuromuscular junction; Cohen (1972) observed the development of functionally normal neuromuscular junctions in combined explants of skeletal muscles and nerves, even in the presence of d-tubocurarine. Recently, Giacobini (1972) pointed out that there is a striking coincidence in the appearance of ChAc activity in myotubes and limb buds of chick embryos and the time when exploratory nerve fibers make their first appearance in the musculature. I n his opinion, this fact supports the hypothesis that nerve fibers control AChE synthesis in myoblasts by means of ACh release. Also O h and Johnson (1972), who repeated Goodwin’s and Sizer’s studies (see above), observed that, under more favorable conditions, cholin-
104
BERTALAN CSILLIK
ergic drugs actually increased the AChE activity of chick embryonic skeletal muscle cultures, both in the synaptic area and in the entire sarcotubular system throughout the muscle fiber. Thus, it appears that, during ontogenesis, ACh actually exerts an inductive e8ect on the AChE system. I t is worth noting that studies performed on mouse neuroblastoma cells also suggest a similar substrate induction of AChE by ACh in nerve cell bodies. Schubert et al. (1971) observed an increased AChE activity in neuroblastoma tissue cultures during neurite formation; a second increase of AChE occurs when the cells in culture start to deteriorate. Harkins et al. (1972) observed a 37-fold increase in the specific activity of AChE in neuroblastoma tissue cultures following addition of ACh. This might be looked upon as a brilliant example of substrate induction; however, there are also alternative explanations. I t may be assumed that interaction between ACh and the AChR might lead to gene induction or gene derepression; on the other hand, ACh might interact with a number of positive or negative cellular control elements in a much less specific way, also resulting in an increased AChE activity. Recent studies indicate that gene expression might be mediated by cyclic AMP, or rather by dibutyryl cyclic AMP; as shown by Furmanski et al. (1971), dibutyryl cyclic AMP is a more potent activator of gene expression phenomena, like neurite extension and AChE synthesis, than cyclic AMP itself. Since synaptic vesicles contain cyclic AMP (Johnson et al., 1973), it might be hypothesized that it is actually not ACh itself but cyclic AMP released concomitantly that is involved in gene expression phenomena also at the postsynaptic membrane. An entirely new line of evidence for ACh being involved in the maintenance of postsynaptic AChE follows from the study of the effect of tetanus toxin on the neuromuscular transmission. Duchen and Tonge (1973) found that, in addition to induce a presynaptic blockade of transmission by inhibiting transmitter release, tetanus toxin induces also a spectacular sprouting of axon terminals. Although the axon terminals within the original end plates do not degenerate, the underlying postsynaptic structures undergo a degenerative process similar to that observed after transection of the motor nerves. At the same time, newly formed axonal sprouts tend to establish new AChE-active subneural structures in the neighborhood of the original, decaying postsynaptic membranes. It appears the whole sequence of events is essentially identical to that observed by us (Csillik and Sivay, 1958) in the course of retarded end-plate regeneration. I n other words, Duchen and Tonge’s studies indicate that the survival of the postsynaptic AChE (and, possibly, also of the junctional folds) does not depend merely on the structural integrity of the presynaptic structures, but rather on their functional activity. Apparently, the release of ACh from its presynaptic store is a prerequisite for the structural and functional viability (AChE activity) of the
SYNAPTOCHEMISTRY O F A CHOLINERGIC N E U K O N
105
postsynaptic membrane; under favorable conditions, the combined effect of nerve fibers and ACh actually induces de nouo appearance (synthesis) of postsynaptic AChE. Summing up the results of the pertinent recent literature it appears that, though in itself unable to induce a specific AChE synthesis, ACh is an indispensable factor to prevent deterioration of already existent AChE and, in conjunction with the “trophic” action of the nerve (perhaps by involving cyclic AMP or dibutyryl cyclic A M P ) , it is able to induce AChE-active postsynaptic structures. Therefore, degeneration of the motor axon terminal (after transection of the nerve) or its inability to release ACh (after tetanus toxin treatment) results in a slow but inevitable destruction of the postsynaptic AChE. Apparently, the continuous bombardment by ACh is an “exercise” for the postsynaptic AChE that kecps it “in shape”; otherwise the enzymeactive complex of the postsynaptic membrane looses its ability to accoinmodate for the hydrolysis of ACh molecules. This, in turn, results in the disintegration of the postsynaptic structures. I n this respect, the effect of ACh is comparable to that of a “trophic substance,” as postulated by Guth (1968, 1969) and by Drachman (1968, 1972). If so, however, the question still arises: How does the postsynaptic AChE survive degeneration of the motor axon? The time of survival differs considerably in different muscles; in one of the longest experiments (Sivay and Csillik, 1956), a survival time of more than 6 months has been reported. Such a long turnover (half-life) of AChE is more than improbable. Corresponding values for other enzymes are many times shorter. Fonnum et al. (1973) estimated the half-life of ChAc to be in the range of 16-21 days. Other enzymes, like cytochrome b, (60 hours), cytochrome c reductase (40 hours), and coenzyme-A reductase (2-3 hours) have even shorter halflives and only nicotinamide adenine dinucleotide glycohydrolase was reported to have a turnover of 16 days (Dehlinger and Schimke, 1971). In a recent study, Wenthold et al. (1974) reported that the half-life of rat brain AChE is 2.84 5 0.13 day. Thus it is reasonable to assume that, after degeneration of the motor nerve terminal, there still persists an AChE production, even though to a constantly decreasing extent. In order to account for this permanent synthesis of AChE, and in view of the above literature data, we have to account both for a persistent release of ACh at the end plate area after motor nerve degeneration and also for a locally present AChE-synthesizing apparatus. I t has to be admitted that, in a first approximation, both suggestions appear to be absurd. What tissue element, if not nerve, can account for the release, of even small amounts, of ACh from sites of denervated neuromuscular junctions? And what tissue element might be responsible for the synthesis of postsynaptic AChE, except for the ribosomal apparatus of the sole plate, which obviously does not exert any AChE-activity?
106
DERTALAN CSILLIK
With regard to the site of postsynaptic AChE synthesis, it might be useful, by analogy, to scrutinize once more the histochemical pattern of AChE activity seen in the perikarya of cholinergic neurons. I n spite of the light microscopic concentration of AChE in Nissl bodies, the fine-structural analysis reveals that ribosomes themselves never stain for AChE (except for extremely overincubated specimens). As a rule, the end product of the enzyme reaction is confined to the lumina of r-ER cisterns, suggesting that perikaryal AChE, synthesized from amino acids in a due sequential order at the surfaces of ribosomes, gains its specific hydrolytic capacity (probably its tertiary structure) only within the cisterns of the endoplasmic reticulum. Furthermore, the electron histochemical AChE reaction, a negative stain in this sense of the word, reveals the existence of -250-300 A globular, nonreacting, intracisternal elements, called endoplasmic units (Csillik and KnyihAr, 1968b). Endoplasmic units were visualized also by other electron microscopic techniques, including the zinc iodide-osmic acid reaction that consistently reveals the existence of endoplasmic units in a variety of nerve cells (Madar6sz and H h o r i , 1971; HalAsz et al., 1971; Job et al., 1973). Faint outlines of endoplasmic units can be observed within the cisterns of the r-ER in many other cells, too, even in conventional electron micrographs. Circumstantial evidence indicates that the role of the endoplasmic units is the elaboration of the tertiary structure of enzyme protein molecules, responsible for their hydrolytic activity (Csillik and Knyihjr, 196810). In contrast to nerve cell perikarya, sole plates underlying the junctional folds contain mainly free ribosomes and only a very few r-ER cisterns. NO wonder there is not any AChE activity associated to the sole plate. O n the other hand, the fingerlike expansions of the junctional folds intrude into the ribosome-containing sole-plate quite similar to endoplasmic cisterns ; in fact, faint outlines of globular structures, resembling endoplasmic units can often be seen within the intracleft material of the secondary synaptic gap (arrows in Fig. 5 ) . Thus, by analogy, it can be assumed that specific amino acid sequences characterizing AChE are synthesized at the surfaces of soleplate ribosomes ; however, these polypeptides acquire their tertiary structure (and their hydrolytic activity) only within the cisternlike junctional folds of the subneural apparatus, probably by virtue of interacting with ACh. The final question to be considered is the origin of ACh responsible for enzyme induction. Under normal conditions, the main source is undoubtedly the axon terminal ; the reasons outlined above, however, suggest that some additional source persists, although in a diminishing amount, also after degeneration of the terminal. In order to account for the apparent amplification of ACh action in a-motor end plates, we have suggested (Csillik, 1965, 1967) that ACh is released also from the junctional folds themselves. Participation of nonneural
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U K O N
107
structures of the neuromuscular junction in the release ACh is supported by the following experimental facts : 1. I n addition to the early observations of Birks e t al. (1960), recent studies by Dennis (1972) show that, after the axon terminal has undergone degeneration, ACh can be released from the end-plate area by electrically stimulating teloglial Schwann cells. 2. Acetylcholine can be detected in the perfusate of skeletal muscles even after complete degeneration of the nerve endings (McIntyre, 1959; McIntyre et al., 1950). 3. According to biochemical investigations, most of the ChAc activity of the muscle is concentrated in the end-plate region (Hebb et al., 1964). Although, after degeneration of the motor nerves, ChAc activity decreases markedly, there still remains a slightly elevated ChAc activity at the formerly innervated area. 4. I n electron microscopic autoradiographic studies, aiming to demonstrate the fine structural localization of “C-labeled hemicholinium, in addition to the overwhelming localization of reduced silver grains in teloglial Schwann cells, there could consequently be seen grains in the area of junctional folds. Autoradiographic localization of [‘-’C]heniicholinium is a marker for intensive choline metabolism, probably related to ACh synthesis (Section 111, A ) : thus the presence of reduced silver grains in the junctional fold area suggests an active cholinc metabolism comparable to that of the teloglial Schwann cells, obviously providing choline, as a raw material, for synthesis of ACh by ChAc. 5. Histochemical studies revealed that ACh induces a Ca?+ release in the sole-plate area (Csillik and SLvay, 1963, 1965). At 2-4 days after transection of the motor nerve, Cn-reaction of the sole-plate disappears; 8 or more days after the transection, a linear reaction appears on the surfaces of muscle fibers, extending 100-200 pm at both sides of the original end plate area. Both under normal conditions and after nerve degeneration, Ca” release can be evoked not only by ACh and by carbamoylcholine, but also by eserine (or by other anti-ChE agents). Thus, in fact, there has to persist an ACh release even after a complete degeneration of the motor nerve fibers; otherwise the inhibition of junctional AChE would not result in protection of ACh that, in turn, evoked the “linear” Ca‘+ release. ( I n this context, it will be noted that the cxtended linear Ca’+ reaction on the surface of denervated muscle fibers is related to the spread of acetylcholine receptors. ) Thus, there appears to be sufficient evidence to substantiate a release of ACh from the end-plate area, even after the complete degeneration of the motor nerve fibers. It is more difficult to decide whether the source of this “nonneural” ACh is the tcloglial Schwann cell, the junctional folds,
108
BERTALAN CSILLIK
or other structural elements. Though virtually irrelevant from the point of view of the inductive effect exerted by ACh after denervation (i.e., with regards to the survival of AChE) this question bears great importance in the theory of synaptic amplification. Therefore, let us briefly summarize the arguments in favor of a release of ACh from the junctional folds: 1. Depolarization of Schwann cells results in acetylcholine release (Dennis, 1972) ; this, however, does not prove that the source of acetylcholine was the Schwann cell: it may well be that depolarization of the Schwann cell releases ACh from another store. 2. Schwann cells are in an intimate contact with the denuded postsynaptic membrane only for a short time after axoterminal degeneration. Within a few days, connective tissue and collageneous fibers invade the abandoned synaptic gutter, isolating overlying teloglial Schwann cells from the AChEactive postsynaptic membrane (Csillik and KnyihAr, 1968a; Waser and Nickel, 1969). 3. Lemmoblastic Schwann cells are unable to synthesize ACh, as follows from the abolished ChAc activity in degenerated nerves (Hebb and Waites, 1956). 4. There can never be seen any synaptic vesicles in Schwann cells, either under normal conditions or after axoterminal degeneration. Resemblance to synaptic vesicles of the “honeycomb-like structures” observed by Birks et al. (1960) in teloglial Schwann cells is entirely superficial. Thus, in conclusion, it appears that AChE of the postsynaptic membrane is synthesized by the ribosomal apparatus of the sole plate, underlying the motor nerve terminal. The enzyme acquires its functional maturity (hydrolytic capacity) within the junctional folds. ACh, released from the nerve terminal under normal conditions and liberated, to a smaller extent, also by the junctional folds themselves, contributes to the final elaboration of AChE enzyme molecules, playing the role of an inductive substance. After degeneration of the nerve terminal, the gradually decreasing ACh production of the postsynaptic structures sustains this effect for a time, but fades out completely later, when the junctional fold apparatus itself undergoes degeneration.
INITIAL AXONTERMINALS : RENSHAWELEMENTS D. RECURRENT Ever since the discovery of the powerful recurrent inhibition of spinal motoneurons by antidromic ventral root volleys (Renshaw, 1941, 1946), the morphological substrate of this fundamental inhibitory mechanism was within the center of animated discussions. Functional characterization of this mechanism was far more advanced than its structural basis; repeated studies performed with the multibarreled microelectrode technique unequiv-
SYNAPTOCHEMISTRY O F A CHOLINERGIC N E U R O N
109
ocally proved the nicotinic cholinergic character of the motoneuronal axon collaterals impinging upon (that time) hypothetical Renshaw cells (Eccles, 1964). I n addition, on the basis of more sophisticated neurophysiological studies, Curtis and Ryall (1966) postulated also muscarinic and “third type” cholinergic receptors on Renshaw cells. It was also suggested that nionoaminergic inputs may influence the activity of Renshaw cells. Although the very identity of the inhibitory transmitter substance released by Renshaw axons (producing a postsynaptic inhibition, i.e., hyperpolarization of the motoneuronal surface membrane) remained unknown, circumstantial evidence suggested that it might be glycine. In order to pinpoint cells responsible for Renshaw inhibition, various histological and neurophysiological techniques were employed. Light microscopic neurohistological studies lrad Szenthgothai ( 1951, 1958) and Ralthazar (1952) to the opinion that Renshaw inhibition might be elicited by short-axoned interneurons. However, the Scheibels ( 1961, 1966, 1971) categorically denied the existence of such short-axoned (Golgi 11) cells in the ventral gray. More recent studies (Szentigothai, 1967) proved that terminals of motoneuronal axon collaterals are scattered in the lamina IX of Rexed, i.e., in the motoneuronal pool, and in between ventral root fibers emerging from the anterior horn. This localization is essentially identical with the sites from where Renshaw inhibition can be elicited, as follobvs from the dye-ejection studies of Willis and Willis (1964), Willis (1971), and Thomas and Wilson (1965). Another important line of evidence follows from the studies reported by Erulkar et al. (1968). Sites from where Renshaw inhibition could br elicited were labeled by Fast Green FCF, a nondiffusible dye ejected from the stimulating micropipette at the end of the recording experiment. A subsequent AChE reaction of cryostat sections obtained from such spinal cord material revealed small, nondescript structures in the very neighborhood of motoneurons. Unfortunately, because of the very poor tissue preservation, it could not have been decided whether the structures stained by Fast Green FCF and exhibiting AChE reaction were small nerve cells, glial cells, axons, or dendrites; therefore, Erulkar et al. used the term Renshaw element as referring to these ill-defined structures. The first direct evidence for structural characteristics of Renshaw cells is derived from electrophysiological studies coupled by the ejection of the fluorescent dye Procion yellow. It had been shown earlier (Stretton and Kravitz, 1968) that this dye, if injected into living nerve cells, would stain the complete axonal and dendritic arborization. In the hands of Jankowska and Lindstrom (1971, 1972) Procion yellow, ejected from a multibarreled micropipette during a recorded Renshaw inhibition evoked by stimulation through another barrel, stained large, spindle-shaped multipolar cells in the
110
BERTALAN CSILLIK
ventral horn, equipped with dendrites exhibiting periodic bulbous enlargements. Prior to this, we followed the path of Erulkar et al. and concluded that the “Renshaw elements” are, in fact, bulbous dilatations of dendrites of the (that time unknown) Renshaw cells, surrounded by scores of AChE-positive terminals of motoneuronal axon collaterals (Csillik and T6th, 1972 ) . While searching for cells in the ventral gray that are equipped with bulbous dendritic enlargements, we found that the cell designated No. 3 by the Scheibels (1966) exactly fits the criteria obtained on the basis of histochemical staining. I n the course of more detailed studies (Csillik et al., 1973b) we systematically collected samples of AChE-active Renshaw elements from serial cross- and longitudinal sections of the rat and the cat spinal cord. The central part or “core” of these elements measures 3-5 pm in the rat and 4-8 pm in the cat; this core does not exert any AChE activity, neither does it stain for basic dyes (Figs. 23 and 24). Possible interpretations, other than a bulbous dendritic dilatation (like neuronal and glial nuclei or parikarya, axonal profiles, and cross sections of capillaries) were ruled out on the basis of structural and/or histochemical characteristics. The number of strongly AChE-active initial axon terminals impinging upon the core is in the range between 10 and 20. Ligation of the ventral roots resulted in an increase of the AChE activity of the initial axon terminals; decrease of motoneuronal AChE due to the same experimental intervention enhanced the optical contrast thus resulting in outstanding pictures of the Renshaw elements. I n a few cases, Renshaw elements were located among the emerging ventral root fibers, in a localization where the background AChE activity did not interfere with their structural correlations. I n such cases it could often be seen that AChE-active axon terminals impinging upon the bulbous central core derived, in fact, from initial collaterals of motor axons. Through-focus series photomicrographs prove that possibly not only scores but hundreds of AChEactive axon terminals impinge upon every single dendritic bulb ; accordingly, the shape and size of this AChE-active “collier” changes from a ringlike structure to a crescentlike one in different optical depths of the section. (Fig. 25). Electron microscopic localization of AChE revealed that, like in the neuromuscular junction, the enzyme is located in both the pre- and the postsynaptic membranes of Renshaw elements. Here again, the origin of the presynaptic AChE can readily be explained, since the entire initial axon collateral exerts AChE activity; thus the enzyme at the presynaptic membrane might simply be supplied by the translational (lateral) movement of the axolemmal membrane components. With respect to the postsynaptic contingent, it may be assumed that the enzyme is synthesized by the Renshaw cell itself; in fact, a low AChE activity was reported to be present in the
S Y N A P T O C H E M I S T R Y OF A CHOLINERGIC N E U R O N
111
FIG. 23 ( t o p ) . AChE activity of Renshaw elements (arrows) between motoneurons ( M N ) ; lumbar ventral gray of the rat. Activity of Renshaw elements is increased, whereas that of the motoneurons is decreased, owing to ligature of the ventral roots. FIG. 24 ( b o t t o m ) . AChE activity of Renshaw elements; high power light microscopic pattern. Note the homogeneous center of elements, corresponding to bulbous dendritic dilatations of Renshaw cells; AChE activity is exerted by scores of intensely reacting recurrent initial axon terminals impinging upon these dendritic bulbs.
perikarya of neurons within the ventral gray, tentatively identified with Rerishaw cells (Navaratnam and Lewis, 1970). The exceptionally intense AChE activity of initial axon terminals impinging upon Renshaw dendritic bulbs raises an intriguing possibility. There is a good reason to assume that these are identical with the nicotinic choliner-
112
BERTALAN CSILLIK
FIG. 25. Three-dimensional reconstruction of a Renshaw dendrite. Note bulbous dilatations surrounded by scores of AChE-active terminals of initial axon collaterals, establishing Renshaw elements. These latter are supposed to represent nicotinic cholinergic receptors, while interbulbous (thin) dendritic areas may house muscar i n k and other receptors.
gic receptors, whereas the localization of muscarinic and other cholinergic receptors remain unknown for the time being, but supposedly located on the perikaryal surface membrane or on the interbulbous portions of the dendrites. In this context, it will be noted that, in the mammalian nervous system, the intensity of AChE activity of Renshaw elements is comparable only to that of motor end plates and of autonomic ganglionic synapses. O n the other hand, muscarinic cholinergic junctions (in the parasympathetic autonomic innervation apparatus, or in the striatum, or in the piriforni cortex, etc.) exert a considerably weaker AChE activity. Thus, it appears that the intensity of the histochemical AChE reaction reflects, to some extent, the physiological characteristics of the underlying receptor.
111. Indirect Information on Cholinergic Mechanisms
A. ROLE OF CHOLINEI N TRANSMITTER SYNTHESIS: AUTORADIOGRAPHY O F [“C]HEMICHOLI NIUM ACh synthesis in cholinergic terminals involves the activity of ChAc and the transport of choline to the sites of the ACh-synthesizing machinery
7% HO -CH,
-CH,-
N+ -CH,
I
CH3
y
-
oc
3
CO-CH,-N+-CH,-CH,-OH
I
CH3
Ftc. 26. Structural formula of a hemicholinium (HC-3).
FIG. 27. Light microscopic autoradiographs of the flexor digitorum brevis muscle of the rat. Cryostat section prepared 35 minutes after a n i.p. injection of “C-labeled
hemicholinium (10 mg/kg body weight). ( a ) Low power. Note concentration of reduced silver grains in motor end plates (arrows) and in nerve bundles ( N ) . Muscle tissue proper exhibits a slisht background reaction. ( b ) High power. T h e grains outline the end-plate area.
SYNAPTOCHEMISTRY OF A C:I-IOLINERGIC N E U R O N
115
(Hanin et al., 1972). Since ChAc activity cannot be localized with sufficient reliability for the time being, indirect informations at the structural level appear to be of considerable importance. One of these is to trace the transport mechanism of choline. Unfortunately, the direct approach (i.e., to locate isotope-labeled choline injected to the animal or added to a bathing solution) is subjected to serious pitfalls due to the intermediary metabolism of choline. More promising is to use and to trace substances, like hemicholiniums, that, according to the classical studies of MacIntosh ( 1959), interfere with the uptake, transport and/or activation of choline in the ACh synthesizing machinery. HC-3, the most potent member of the hemicholinium series (Fig. 26) synthesized by Schueler ( 1955), was labeled by “C. Rats injected intraperitoneally with 2-10 mg/kg body weight of [11C]-HC-3were killed at the heights of the pharmacological symptoms of poisoning, i.e., 30-40 minutes after the injection. Striated muscles and the spinal cord were dissected, cut in a cryostat, dried, and processed for light microscopic autoradiography. Fixation and embedding was omitted in order to prevent presumable loss of the water-soluble drug (Csillik et al., 1970) . In the light microscopic autoradiograms, reduced silver grains were concentrated in intramuscular nerve trunks and, especially, in motor end plates (Fig. 2 7 ) . Since, however, the patterns obtained in neuromuscular junctions were not characteristic either of the axoterminal arborization or the subneural apparatus, the exact site where HC-3 exerted its inhibitory action could not be decided. I n order to overcome this difficulty, electron microscopic autoradiographic studies were performed (Csillik et al., 1973a). Here, however, we have to encounter the possibility that a t least a part of HC-3, is not firmly bound to proteins; loss of an unknown amount of the drug is a major handicap in such studies. I n spite of this, it is of considerable importance that we never succeeded in tracing any single grain within the terminals of the motor axons; on the other hand, most of the grains was concentrated in
FIG. 28 ( t o p ) . Electron microscopic autoradiograph of a neuromuscular junction in the rat’s flexor digitorum brevis muscle. T h e rat was killed 40 minutes after a n i.p. injection of “C-labeled hemicholinium (10 mg/kg body weight). Ax, axon terminal, containing mitochondria ( M ) and synaptic vesicles (sv) ; JF, junctional folds; Schw, teloglial Schwann cell. Note concentration of reduced silver grains in the cytoplasm of the Schwann cell. FIG. 29 ( b o t t o m ) . Electron microscopic autoradiograph of a neuromuscular junction in the rat’s flexor digitorum brevis muscle. Technique identical as in Fig. 28. Reduced silver grains are located postsynaptically in the region of the junction folds. Mf, myofilaments; N, nucleus of the overlying (teloglial) Schwann cell.
116
BERTALAN CSILLIK
FIG. 30. Light microscopic autoradiograph of motoneuron ( M N ) and a neighboring Renshaw element ( R ) . Cryostat section of the lumbar spinal cord, prepared 35 minutes after i.p. injection of I4C-labeled hemicholinium ( 10 mg/kg body weight). Note accumulation of reduced silver grains within the cytoplasm of the motoneuron, sparing the nucleus; and a similar, or even stronger accumulation of grains within (or around) initial recurrent axon terminals impinging upon the central core ( = bulbous dendritic dilatation) of the Renshaw element. Compare this latter with the histochemical patterns of AChE in Renshaw elements (Figs. 23 and 2 4 ) .
the cytoplasms of the teloglial Schwann cells, overlying the terminal (Fig. 2 8 ) . I n a few cases, grains were found in the junctional fold area, i.e., in postsynaptic structures (Fig. 29). Therefore, it appears that choline, the “raw material” for ACh synthesis, is provided by satellite elements of the cholinergic synapse. In other words, notwithstanding the fact that ChAc is a synaptosomal protein (Aquilonius et al., 1973), teloglial cells play a vital role in supplying choline for the synthesis of ACh. I n this respect, it is an important fact that both the perikarya of spinal motoneurons, and, especially, the Renshaw elements described earlier, accumulate great amounts of [“Cl-HC-3 (Fig. 30). Electron microscopic autoradiography of the spinal cord will probably reveal whether also in Renshaw elements, HC-3 is bound to glial structures or to postsynaptic (dendritic)
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
117
sites rather than to nerve terminals proper. There is no doubt about, however, that HC-3 is accumulated by the motoneuronal cytoplasm itself; the dimensions of these cells are sufficient to exclude any translocational phenomena.
B. DYNAMICS OF THIAMINE PYROPHOSPHATASE: LOCALIZATION IN
SYNAPSES
Thiamine pyrophosphatase (TPPase) , an enzyme responsible for the hydrolysis of cocarboxylase ( T P P ) , was shown by the Novikoff school to be a specific marker for the Golgi apparatus of vertebrate cells, including neurons (Novikoff et al., 1964, 1971). I n fact, TTPase reaction outlines dictyosomes of the Golgi system, both in the perikaryon and in proximal dendrites of spinal motoneurons (Fig. 3 1 ) . After transection of motor axons, both the light- and the electron microscopic patterns of TPPase undergo marked alterations that offer the possibility of tracing cells of origin of different pathways (Lriszlb and Knyihrir, 1974). Recent studies (Knyihrir et al., 1973; Griffith and Bondareff, 1973) revealed that, in addition to the Golgi apparatus, axon terminals, too, exhibit a marked TPPase activity. The reaction is mainly confined to synaptic vesicles. A more detailed examination of axon terminals in the spinal cord revealed that TPPase reaction characterizes the spheroid, -400 A synaptic vesicles, whereas flattened (ovoid) vesicles and, especially, dense-core vesicles, are devoid of TPPase activity. Even more intriguing is the fact that the reaction product is attached to the external surfaces of some of the vesicles, and, that, in the close vicinity of synapses, the end product of the reaction is often seen in a free form, dispersed in the terminal axoplasm (Figs. 32 and 3 3 ) . Dynamics of TPPase in the neuron appears to follow that of other neuroproteins. Under normal conditions, axons contain only small amounts of TPPase reaction product; 72 hours after ligating the sciatic nerve, however, a massive accumulation of the reaction above the ligature was observed. Thus, it appears, TPPase is manufactured in the perikaryon, concentrated (perhaps also coupled with macromolecular carbohydrates) in the Golgi system and, finally, it is transported via axonal flow to the terminals (Csillik et al., 1974). This, of course, suggests that TPPase might be involved in some important activities of the axon terminal. In fact, some of the biochemical processes closely related to transmitter synthesis, require the participation of TPP.
118
RERTALAN C S I L L I K
FIG. 3 1. Histochemical localization of thiamine pyrophosphatase (TPPase) in the Golgi apparatus of a spinal motoneuron. ( a ) Light microscopy, ( h ) electron histochemistry ; note the electron dense reaction product in the internal Golgi cisterns and the absence of reaction from other subcellular organelles.
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
119
Biochemical studies (TuEek and Cheng, 1970; Cheney et al., 1969; Koeppe et al., 1964; Browning and Schulman, 1968; Heinrich et al., 1973) prove that acetyl groups in acetyl coenzyme A, providing active acetate for ACh biosynthesis, are derived mainly from pyruvate (whereas acetate itself plays a negligible role in acylation of coenLyme A ) . I n the multienzyme system responsible for pyruvate decarboxylation, T P P is an essential factor. Thus T P P plays a decisive role in ACh synthesis, as follows from the critically decreased levels of ACh after thiamine deficiency (Bhagat and Lockett, 1962; Heinrich et al., 1973). O n the other hand, thiamine (vitamin B,, aneurine) appears to play also a direct role in axonal conductivity and synaptic efficiency, as proposed originally by von Muralt (1946, 1958) and recently by Itokawa et al. (1972). As a matter of fact, thiamine is transported via axons to the terminals ( Tanaka et al., 1973). Thus, the enzyme involved in T P P hydrolysis may actually play a vital role in neurotransmissional processes. O n the other hand, the fine-structural localization of the end product of the TPPase reaction in axon terminals can be regarded a useful marker of how the content of a synaptic vesicle may contribute to the complex functional activity required for the transmission process. I t stands to reason to assume that TPPase, transported by the axonal neurotuhules, is incorporated by the synaptic vesicles, possibly by a rotational movement of the -45 A subunits constituting the membrane of the synaptic vesicle (see Section IV, A ) . As soon, however, as the vesicle arrives a t the active region of the synapse, i.e., to the area of presynaptic “dense” projections (see Section IV, C, l ) , friction of the vesicle with the projections is bound to induce an inverse rotation of the subunits, leading to the release of at least some of the content of the vesicle. Such a step, if extrapolated in terms of other contents of synaptic vesicles, e.g., ACh, might be regarded as a preparation for the release of the content of the vesicle by the exocytotic cycle (see Section IV, C, 1 ) . I t will be noted that the role of TPPase in synaptic transmission can be accepted only by using several steps of analogy and extrapolation. I n spite of this, the intriguing correlations outlined above justify further efforts to prove the role of T P P and TPPase in synaptic processes, using electron histochemical techniques in conjunction with antivitamin treatment. IV. Molecular Anatomy of Transmitter Release
A. ORIGIN OF SYNAPTIC VESICLES I t is a well-established fact that ACh, like other neurotransmitters, is associated with synaptic vesicles (De Robertis, 1958; De Robertis and Rod-
120
BERTALAN CSILLIK
FIG. 32 ( t o p ) . Electron histochemical localization of TPPase in synaptic vesicles in the spinal cord. Note reaction product concentrated within the cavities of most of the vesicles (e.g., sv-1) ; in the vesicle sv-2, the reaction product is associated to the external and the internal surfaces of the vesicular membrane whereas vesicle
S Y N A P T O C H E M I S T R Y OF A CHOLINERCIC N E U R O N
121
riguez de Lores Arnaiz, 1969; Whittaker, 1959, 1969a,b; Potter, 1972) ; recent studies suggest that part of the ACh might be “loosely bound” to the vesicles, perhaps to their surfaces, as suggested by us several years ago (Csillik and Job, 1967), whereas the bulk of the transmitter appears to be concentrated within the cavity of the vesicle (Fonnum, 1973). Genesis of synaptic vesicles, once a major item of ardent discussion, appears to be settled by the recent studies employing powerful labeling techniques, like horse radish peroxidase and dextran (Clark et al., 1972; Ceccarelli et al., 1972, 1973; Heuser and Reese, 1973). Accordingly, synaptic vesicles are derived from recycled membrane material, pinched off from nonsynaptic areas of the axoterminal surface membrane. The membrane material is provided by coated vesicles, that, by means of endocytosis, travel into the center of the terminal ; here, they coalesce with endoplasmic cisterns and/or tubules, in order to give rise to synaptic vesicles. While merging with these intraterminal endoplasmic elements, coated vesicles loose their coat, which occasionally, can be found within the terminal axoplasm. The coat material, being in essence a remainder of the internal lining of the axolemma, contributes to the electron density of several types of axon terminals. The fine molecular mechanism the membrane material undergoes after being engulfed by the intraterminal cisterns is unknown. I t can only be assumed that, being dismantled to some extent, there will be elaborated a newly fashioned membrane material that fits the functional demands of a “transmitter-carrying’’ synaptic vesicle. I n fact, the fine electron microscopic structure of the synaptic vesicle is distinctly different from that of other membranous elements of the axon terminal: instead of exhibiting a unit membrane appearance, membranes of synaptic vesicles are composed of -45 A subunits (Fig. 34). [Essentially similar, though slightly larger (-70 A ) particles constitute the membranes of chromaffine granules (Pollard et al., 1973) .] Also the chemical composition of synaptic vesicle membranes differs considerably from that of the membranes of the synaptosome (Whittaker, 1969a; Morgan et al., 1973), indicating a fundamental transformation of the membrane material during the genesis from axoterminal cisterns.
B. CHARGING OF SYNAPTIC VESICLES Incorporation of the transmitter substance by the newly formed synaptic vesicles is the next step in the procedures leading to synaptic transmission. sv-3 is empty. Here, however, the subunit structure of the vesicle membrane can clearly be seen. FIG. 33 ( b o t t o m ) . Electron histochemical localization of TPPase in the neuromuscular junction. Flexor digitorum brevis muscle of the rat. Note concentration of enzyme reaction product in synaptic vesicles. Pre, presynaptic membrane; post, postsynaptic membrane; JF, junctional folds.
-
122
BERTALAN CSILLIK
45
A
\
I
/
I
400
%,
-
\
\
Fro. 34. Molecular anatomy of a spherical synaptic vesicle. T h e membrane of the vesicle consists of -45-A globular subunits, themselves representing phospholipid micelles. The interspaces between the spherical subunits are occupied by integral protein molecules.
This process, conveniently called charging, might occur either during formation of synaptic vesicles within and from the axoterminal endoplasmic system or, after that, while en route to the presynaptic membrane. T h e electron microscopic structure of the vesicle membrane and, especially, its relation to an easily detectable particulate material (end product of the TPPase reaction) suggest that charging takes place after the vesicles were released from the axoterniinal endoplasmic system. Since the membrane of the synaptic vesicles consists of -45 A globular subunits (Section IV, A ) , it is reasonable to assume that, if such subunits are able to rotate freely, they would convey any material attached to their external surfaces into the cavity of the synaptic vesicle. Such a process can logically bc reconstructed on the basis of the electron histochemical patterns of TPPase in synaptic terminals (Section 111, B) . Extrapolating this mechanism for smaller molecules (e.g., ACh) it would appear that both the assumption that ChAc is located freely within the terminal axoplasm (Whittaker et al., 1964) as well as the notion that ChAc is associated with synaptic vesicles under different experimental parameters (McCaman et al., 1965;
S Y N AI'TOC H E & I SI T R Y OF A C: I1 0L.I N liROI( 1 N 1: UKO Pi
123
De Kobertis arid Rodriguez de Lores Arnaiz, 1969) are entirely consistent with such a proposed uptake mechanism.
C. DISCHARGING OF SYNAPTIC \'ESICI~ES 1. Preparatory S t e p s I n addition to differences in the electron densities of pre- and postsynaptic membranes (Gray, 1959), syninietrical (type 11) and asymmetrical (type I ) synapses appear to differ from each other also with respect to the finestructural organization of the terminal axoplasin (Jones, 1972). I n spite of this, basically the very sanie microstructural arrangement characterizes the internal surfaces of the presynaptic membrane in virtually all the vertebrate synapses studied hitherto. Presynaptic electron dense projections intruding into the terminal axoplaslii from the presynaptic membrane were first describrd by Gray (1963, 1966) ; they were analyzed in more details by Bloom and Aghajanian (1966, 1968), using phosphotungstic acid as a selective marker stain of these structures. Following this path, Akert (1971) and his school (Akert and Sandri, 1970: Akert et al., 1971, 1972, 1974; Pfenninger et al. 1969, 1971, 1972) introduced various electron dense stains and the technique of freeze-fracturing to disclose the fine structure, geornetrical organization and three-dimensional arrangement of these projections. Presently, there emerges a clear-cut picture of a regularly arranged presynaptic vesicular grid that accornrnodates for synaptic vesicles before being discharged into the synaptic gap. In Akert's terminology, the presynaptic vesicular grid is similar to a n egg crate: the individual projections that establish the grid are electron-dense pyramids, measuring 400-600 A ; these are arranged in a hexagonal pattern covering the active region of the presynaptic membrane. Synaptic vesicles destined to discharge their content fit into the interspaces between the individual pyramidlike projections. Although the presynaptic vesicular grid was primarily described in central synapses, recent studies by Couteaux and P6cot-1)echavassine ( 1970) and by Dreyer et al. (1973) revealed that similar structures are present also in the neurornuscular junction. An important feature of this system is that it is located exactly opposite to the junctional folds; this organization emphasizes the importance of the junctional fold system in synaptic transniission in the neuromuscular junction, probably related to postsynaptic amplification of ACh effects (Section I V , D ) . O u r studies on the fine-structural appearance of thiamine pyrophosphatase (TPPase) activity in both central synapses (in the spinal cord) and in the neuromuscular junction suggest that these presynaptic electron dense projections may, in fact, play a n important role in those preparatory steps
124
BERTALAN CSILLIK
that, finally, result in the release of the transmitter from the synaptic vesicle. Extrapolating the localization of the TPPase reaction product to other contents of the synaptic vesicle, e.g., ACh, it would appear that a t least a part of the transmitter, most probably the loosely bound compartment (Fonnum. 1973) might be released by friction of the subunits constituting the vesiclc membrane, with the presynaptic projections (Fig. 3 5 ) . Since ACh is known to decrease the surface tension of membranes (Jain, 1972) it stands for reason to assume that this frictionally released ACh contingent might sensitize the presynaptic membrane in order to introduce the process of exocytosis. I n this respect, it may be suggested that such a preparatory step is not confined to cholinergic synapses only; as a matter of fact, synaptic. vesicles are known to contain some (admittedly negligible) amount of ACh even in noncholinergic synapses, providing a convenient explanation for AChE activity in presynaptic membranes of noncholinergic synapses.
3 ,
Presynaptic membrane
*
*
0
I
I
\
\
FIG. 35. Charging and discharging of the synaptic vesicle as hypothesized on the basis of incorporation and extrusion of TPPase. A-B-C: the process of charging (incorporation) ; D-E-F: the process of discharging (extrusion). For simplicity, in this diagram both processes are illustrated on the same synaptic vesicle, even though charging is supposed to take place on the center of the axon terminal and discharging to occur only as the vesicle arrived at the area of presynaptic projections. T h e -45-A spherical subunits constituting the vesicle membrane are supposed to rotate freely: 1, Particles (enzyme and/or transmitter molecules) dispersed in the terminal axoplasm; 2, particles concentrated within the cavity of the vesicle; 3, particles extruded from the vesicle by rotation of the subunits. Thin arrows indicate direction of rotation of subunits; thick arrows show the movement of the vesicle. From Csillik et al. (1974).
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
125
Another possibility might be the release of cyclic AMP from synaptic vesicles by the friction with presynaptic dense projections. It has been suggested (Goldberg and Singer, 1969; Singer and Goldberg, 1970) that cyclic AMP plays a n important role in neuromuscular transmission ; according to recent studies (Johnson et d.,1973), synaptic vesicles contain 5000-6000 cy-AMP molecules that may contribute to the induction of exocytosis.
2. Exocytosis Obviously, the simplest mechanism for discharging the content of a synaptic vesicle is exocytosis; in spite of this, ever since De Robertis’ (1958) first suggestion, this basic mechanism of transmitter release was subject to suspicion, probably because of the abundance of oversimplified textbook diagrams and the scarcity of convincing omega profiles in actual research reports. [The term omega profile emphasizes the similarity between the Greek capital letter omega ( 0 ) and the electron microscopic pattern of a synaptic vesicle in the process of exocytosis.] I n fact, omega profiles were observed in an amazingly small number, yet consistently, in the course of studies aiming to disclose microstructural equivalents of synaptic transmission. Thermodynamic unstability, owing to the high free energy content of such a membrane configuration, can be expected to be even more accentuated by the well known tendency of histological fixatives to uniformize (“smear”) membrane structures. This follows inter alia from the molecular reorganization of lipid bilayers by osmic acid, as shown recently in spin-labeling studies by Jost and Griffith (1973). On the other hand, the actual amount of exocytotic figures is also surprisingly small; Ceccarelli et al. (1972) calculated that, if vesicles fuse with the presynaptic membrane at a peak frequency of 1000/second, along a 1-pm portion of the presynaptic mernbrane, 0.03 fusion can be expected to occur every second. T h e situation is analogous to the occurrence of mitotic figures in conventional light microscopic preparations. I t is common knowledge that dividing cells tend to complete the mitotic cycle with all their efforts, even within the lethal environment of the fixative. However, if stopping the mitosis by administration of colchicine, the number of mitotic figures will be considerably higher-not as though colchicine increased the mitotic rate, but because cell divisions are interrupted in metaphase. I n analogy, the snake venom p-bungarotoxin “freezes” the process of exocytosis of cholinergic vesicles in the stage of the omega profile (Chen and Lee, 1970). Accordingly, in rats treated by p-bungarotoxin, the presynaptic membranes of neuroinuscular junctions are virtually crowded by omega profiles. Once the available surface area is occupied by vesicles frozen
126
I3ERTALAN CSILLIK
in the stage of exocytosis, there remains no place for any further synaptic vesicles to discharge their content ; thus, after a short latency period, pbungarotoxin results in a presynaptic transmission blockade. [Another toxin, isolated from the venom of Bungarus multicinctus, a-bungarotoxin, combines with ACh receptors of the neuromuscular junction (Changeux et al., 1970b).] We have seen (Section IV, C, 1 ) that friction of the synaptic vesicles with the electron dense presynaptic projections can be assumed to provide local circumstances suitable to induce the process of exocytosis. I t stands to reason to suppose, furthermore, that invasion of the axon terminal by the nerve impulse increases the affinity of the vesicles to the presynaptic membrane, as follows from the accumulation of synaptic vesicles in the active zone (the “front line”) after a supramaximal stimulation (Jones and Kwanbunbumpen, 1970; Csillik and Bense, 1971; Korneliussen, 1972). Fusion of the vesicle membrane with the presynaptic membrane, however, is extremely difficult to reconcile with the classical unit membrane hypothesis. While fusion of phospholipid bilayers with each other is a common physicochemical phenomenon, a protein “coating” of the membranes, as postulated by the unit membrane theory, would impede such a merging. Recent advances in membrane biophysics might furnish explanation for this problem. An increasing body of evidence suggests that the organization of cell surface membranes, including those of axons, differs considerably from the trilaminar structure seen in electron micrograms. According to the fluid mosaic model (Singer, 1971 ; Singer and Nicolson, 1972 ; Singer and Rothfield, 1973), the backbone of the membrane is a smectic phospholipid bilayer ; proteins are randomly distributed within this bilayer, exhibiting a remarkable translational (lateral) mobility. Some of the proteins span the entire thickness of the membrane (integral proteins), whereas others, called peripheral proteins, are only partly embedded into the phospholipid scaffolding. The surfaces of both integral and peripheral proteins are, at least partly, exposed to the aqueous environments of intra- (and/or extra-) cellular fluids, while hydrophobic portions of protein molecules are embedded into the essentially nonionic (hydrophobic) hydrocarbon chain region of the phospholipid bilayer. In contrast to the unit membrane arrangement, such a fluid mosaic membrane organization offers a logical explanation for the structural fusion of synaptic vesicles with the presynaptic membrane, especially if the peculiar membrane organization of the synaptic vesicle, as iollows from their electron microscopic appearance, is taken into consideration. It has been mentioned already (Section IV, A) that, even under the very same parameters of histological fixation and processing that yield a trilaminar appearance of axon surface membranes, the surface of synaptic vesicles appears to consist of
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
127
-45 A globular (spherical) subunits. Sizes and shapes of biologically important phospholipid molecules (Stein, 1967), their preferential organization in spherical micelles (Johnsion and Roots, 1972) and the high-resolution electron micrographs obtained from phospholipid micelles (Stoeckenius, 1963) suggest that, in fact, the -45 A subunits of the synaptic vesicles represent phospholipid micelles. I11 such micelles, hydrophilic groups of the individual phospholipid molecules are directed outwardly, with their hydrocarbon chains pointing a t the renter of the micelle, in a radial arrangement. I t can be assumed that the i.hin interspaces between the subunits are OCCUpied by integral proteins. Given synaptic vesicles of this organization, and a presynaptic membrane essentially complying to a fluid mosaic pattern, fusion of the two structures is expected to take place at areas free of intervening membrane proteins. I t can be assumed that, owing to the random distribution of integral proteins, in the resting state, such suitable areas are restricted both in size and in amount. This might be the reason why, in the resting state, only a few synaptic vesicles undergo exocytosis, producing only minature end-plate potentials. Invasion of the axon terminal by the nerve action potential, however, might produce temporary aggregations of integral and peripheral proteins, that, as mentioned already, exhibit a considerable translational mobility within the fluid mosaic. The resulting “free” phospholipid areas of the presynaptic membrane, obviously, are confined to the pits between neighboring presynaptic dense protrusions ; thus, exocytotic fusion of vesicles with the presynaptic membrane would take place a t these predilectional points, sensitized in advance by ACh released from the surface of as yet intact synaptic vesicles. Once the fusion gets started, the process of exocytosis tends to decrease the thermodynamic improbability of the omega profile; this appears to be supported by contractile proteins lining the internal surface of the presynaptic membrane (Berl et al. 1973). Formation of S-S bonds in the vicinity of the omega profile (possibly in the region of the presynaptic protrusion) would pull the neighboring membrane areas aside (Kosower and Werman, 1971) , thus completing the mdocytotic process. Why do synaptic vesicles not merge with each other? and why do vesicles not merge with the presynaptic membrane except for the arrival of the nerve impulse? This question has been dealt with by Bass and Moore (1966) and, recently, by Van der Kloot .and Kita (1974). In their opinion, “rigid” water layers on the external surfaces of the vesicles and on the internal surface of the presynaptic membrane, and also the electrostatic repulsion due to the fixed charges on their surfaces, prevent merging under resting conditions. Both the water layers and the electrostatic repulsion might be obliterated by Ca?+ ions, entering the lerminal following an action potential. It stands
128
BERTALAN CSILLIK
to reason that Ca’+ ions, known to be essential for ACh release (Eccles, 1964; Cooke et al., 1973; Cooke and Quastel, 1973a,b), play their fundamental role at this step. Probably this mechanism is synergistic with the actions of neurostenin (Berl et al., 1973) and of S-S groups (Kosower and Werman, 1971). As a net result, the content of the vesicle will be discharged into the synaptic gap (Fig. 36). The molecular anatomical sequence of events outlined above implies two important consequences. (1) Since fusion of the synaptic vesicle with the presynaptic membrane takes place at “free” lipid surface areas, neither integral nor peripheral proteins of the membrane may interfere with, or even have access to, the content extruded from the synaptic vesicle. Accordingly, AChE molecules, being peripheral proteins of the presynaptic membrane, cannot attack ACh liberated from the synaptic vesicle. Thus the argument that presynaptic AChE is a nuisance for presynaptically liberated ACh is virtually ruled out on the basis of molecular anatomy. ( 2 ) Exocytosis results in a net gain in presynaptic membrane material. Since the lateral compressibility of the membrane is restricted, the extension of the membrane surface that follows from the translational movement can be expected to trigger membrane retrieval (endocytosis) at nonsynaptic axolemmal areas, in full accord with the relevant electron microscopic studies (Clark et al., 1972; Ceccarelli et al., 1972, 1973; Heuser and Reese, 1973).
D. TRANSMITTER BINDING TO
THE
RECEPTOR: POSTSYNAPTIC
AMPLIFICATION Once the ACh molecules are released from the presynaptic terminal, they are subject to diffusion within the polysaccharide material of the junctional cleft. Since the physicochemistry of the gap substance is only partly known, the rate of diffusion of the transmitter can only be guessed; calculations range from 1 mm to 1 cm per second. Characterization and purification of the receptor protein (AChR) , using different sources of biological material and different approaches for isolation, has been carried out during the last decade in several laboratories (for details, see Hammes et al., 1973). The receptor appears to be a macromolecular proteolipid (in the range of 1 million daltons) ; recent studies suggest that it is a membrane-bound glycoprotein that has to be reconstituted with phospholipids in order to exert full receptor properties (Raftery et al., 1973). Receptor molecules are located in the postsynaptic membrane, 600-800 A apart from the presynaptic site of ACh release. Interaction between the receptor and the transmitter molecule is, apparently, a cooperative process, owing to the allosteric character of the receptor (Edelstein, 1972). Binding results in alterations of the molecular structure of the receptor, as shown
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
129
FIG. 36. Microstructural events that supposedly occur during exocytosis of synaptic vesicles. ( A ) Synaptic vesicle cores to the active region (between two presynaptic protrusions); note that the external surface of the vesicle as well as the internal surface of the presynaptic membrane is covered by structured water (***). Electrostatic forces (not shown) also exert a repulsive force. ( B ) Release of loosely bound ACh from the surface of the vesicle (dots), by means of friction with presynaptic . protrusions; sensitization of the active region of the presynaptic membrane ( C ) Arrival of axon potential: entrance of Ca2+ions that obliterate structured water and electrostatic repulsion. ( D ) ,4ttachment of synaptic vesicle to the active region; neighboring membrane areas are being pulled apart by the contraction of neurostenin bonds. ( E ) Omega profile (exocytosis); the content and by the formation of S-S of the vesicle ( A C h ) is discharged into the synaptic cleft. ( F ) Translational movement (extension) of the presynaptic membrane after exocytosis, triggering off membrane retrieval from nonsynaptic areas of the axolemma.
(m)
by us earlier by means of polarization microscopy (Csillik, 1963). Changes in the molecular conformation of the receptor results, probably by ionophoric or ion carrier mechanisms, in an increased permeability of the postsynaptic membrane. The immense ion:ic fluxes induced by this process produce the
130
BERTALAN CSILLIK
end-plate potential (or, in more general terms, the excitatory postsynaptic potential). Apparently, the system of junctional folds in a-neuromuscular junctions is a device for amplifying this postsynaptic effect. I t is only a t junctions having numerous and deep junctional folds that the end-plate potential reaches the critical threshold level that triggers off propagatory muscle action potentials. I n 7-end plates, where the postsynaptic membrane is either completely smooth or thrown only into a few shallow folds, the resulting small nerve end-plate potential, restricted to the end-plate area, is insufficient to induce a propagatory action potential of the muscle. The molecular mechanism by which this postsynaptic amplification system operates is unknown. We have proposed (Csillik, 1965, 1967) that interaction of the receptor with ACh results, in addition to an increased ion flux, also in the release of additional amounts of ACh from the postsynaptic membrane itself. These additional amounts of ACh could trigger off receptor transformation in more distad areas of the postsynaptic membrane, in a cascade reaction. Accordingly, the molecular transformations induced by presynaptic ACh would successively invade the entire postsynaptic membrane, including the depths of the junctional folds, probably not available for presynaptic ACh released in a distance as far as 0.5-1 pm apart. Experimental evidence for a postsynaptic release of ACh was summarized on page 107. BETWEEN ACETYLCHOIJNESTERASE AND THE E. INTERRELATIONS ACETYLCHOLINE RECEPTOR: SIMILARITY OR IDENTITY?
The idea that the enzyme responsible for the hydrolysis of ACh might be closely related to or identical with the protein responsible for reacting with ACh has been raised repeatedly during the last four decades. Even before AChE and AChR were either functionally or biochemically characterized, Roepke (1937) mentioned that “there might be some justification in using choline esterase as a convenient model for cell receptors.” In a more explicit form, Karassik (1946) suggested that AChR might be identical with a negative group on the active surface of ChE, different from its esteratic center. Identity of AChE with AChR was hypothesized and discussed in more concrete terms by Wurzel (1956, 1967) and by ZupanEiE (1967). I n a more recent formulation, this theory suggests that “ACh receptor sites and ACh hydrolyzing sites are located on the same macromolecule, thzugh on different parts of it, probably on different polypeptide chains” (Stalc and ZupanEiE, 1972). I n fact, there are many features that are common to AChE and AChR. Both obey the law of a lock and key arrangement; both are located in the postsynaptic membrane ; and according to several calculations the numbers
S Y N A P T O C H E M E T R Y OF A CHOLINERGIC N E U R O N
131
of AChE-active sites and AChR sites are amazingly close to each other, both in electric organs of eels and electric fish, and in mammalian neuromuscular junctions (Changeux et aZ., 1970a; Miledi and Potter, 1971; Barnard et al., 1971). Also the light microscopic localization of AChR is identical with that of AChE in motor end plates, as follows from our early polarization optical studies (Csillik, 1963:i as well as from recent studies, aiming to locate a-bungarotoxin-binding sites in motor end plates. Such autoradiographic essays, performed with a-bungarotoxin labeled with 3H (Barnard et al., 1971) or with lZ5I(Fambrough and Hartzell, 1972), as well as the fluorescence microscopic studies of Anderson and Cohen ( 1973), performed with fluorescein-labeled a-bungarotoxin prove that the toxin-binding sites, identical in all probability with AChR, are located, both in mammalian and in amphibian muscles, in the very same subneural apparatus that exerts AChE activity. Allosteric mechanisms characterize the function of the neuromuscular junction (Rang, 1973) ; allosteric properties of AChE (Section I, C ) were regarded as fundamental for the cholinoreceptor function of this enzyme (AiranEiE, 1971). I n fact, :similarities between the subunit structures of AChE (Section I, C ) and of AChR are striking; results obtained in various laboratories, though differing from each other with respect to the molecular weight, agree in the polymeric nature of AChR. The monomers of the receptor protein were estimated to have a molecular weight of 45,000 (Raftery, 1973), 52,000 (Changeux, 1973), or 90,000 (Edelstein, 1974) ; these are supposed to form tetramers or hexamers that, in turn, aggregate as a macromolecule with a molecular weight around 1,000,000. Accordingly, the receptor is a regulatory protein, characterized by “active” and “resting” conformations (Meunier et al., 1973) ; binding of ACh is a cooperative process. Such cooperative effects might be associated either with the oligomeric structure of the receptor, i.e., with the interactions between the subunits, or rather with the latticelike organization of AChR, demonstrated recently by freeze etching (Cartaud et al., 1973). On the other hand, cooperative cross-linking of receptor subunits by ACh may lead to the formation of ion-conducting channels, as proposed recently by Allison (1972). I t is remarkable that virrually all AChR preparations contain traces of AChE that can be dissociated only by relatively drastic procedures (Eldefrawi and Eldefrawi, 1972). O n the other hand, bimolecular (“black”) lipid membranes, being genuinely sensitive for ACh (Leuzinger and Schneider, 1972), appear to acquire specific ACh-receptor properties if combined with AChE (Del Castillo et al., 11967; Parisi et al., 1971; Grafius et aZ., 1971; Jain et al., 1973). Another intriguing fact is that chronic treatment of rats with neostigmine, while pari:ially inhibiting AChE activity, also results in a marked decrease of AChR-s; the number of receptors is 2.1 X lo’ in nor-
132
BERTALAN CSILLIK
ma1 end plates and 1.2 x lo7 in neostigmine-treated animals (Chang et al., 1973). Whether or not, the close structural and functional correlations between AChE and AChR, both in the electric tissue and in the postsynaptic membrane, are due to a common macromolecular basis or rather to a peculiar membrane mosaic consisting of AChE and AChR, as proposed by Barnard et al. (1971), cannot be decided for the time being. We certainly feel, however, that there is no reason to reject the idea of a membrane mosaic on the grounds of “crowdedness,” as argued by Miledi and Potter (1971). Ion channels and/or ionophores might be associated with receptor molecules, as suggested recently by Raftery (1974) ; and, if the calculations of Salpeter and Eldefrawi (1973) are correct, AChR and AChE molecules would occupy only 25% of the total surface area of the postsynaptic membrane, that cannot be taken as too crowded. Therefore, in contrast to Miledi and Potter’s assumption, we do not think the enzymatic detachment of AChE from the postsynaptic membrane (Hall and Kelly, 1971; Betz and Sakmann, 1971; McMahan et al., 1972) is a proof against AChE being a component of the membrane. More reasonable appears the conclusion that AChR, an integral protein in the sense of Singer (1971 ), spans across the entire thickness of the postsynaptic membrane (Meunier et al., 1973; Karlin and Cowburn, 1973), whereas AChE, a peripheral protein, is only partially embedded in, or rather only superficially associated to, the bimolecular phospholipid sheet (Fig. 37). It should be kept in mind, however, that such a close structural and polysaccharide “basement membrane” of the synaptic gap
muscle cell cytoplasm
FIG. 37. Molecular anatomy of the postsynaptic membrane. AChE molecules (cross-hatched) are peripheral proteins of the membrane; AChR molecules (black) are integral proteins. Close apposition of AChE to AChR accounts for stoichiometry and for structural/functional similarities between these two main constituents of the postsynaptic membrane. Ionic channels (white slits) are formed within AChR-s, probably between subunits, or between AChE and AChR molecules.
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
133
functional correlation between AChE and AChR is a unique speciality of the normally innervated postsynaptic membrane. AChE is present in the myotendinous junction that does not exert any AChR properties ; on the other hand, after denervation, AChR-s spread along the surface of the muscle fiber without any similar spread of AChE activity. I t appears that ACh released from the nerve terminal under normal circumstances keeps the delicate molecular anatomy of the postsynaptic membrane in due balance. REFERENCES Akert, K. (1971). Klin. Wochenschr. 49, 509-519. Akert, K., and Sandri, C. (1970). I n “Excitatory Synaptic Mechanisms” (P. Andersen and J. K. S . Jansen, eds.), pp. 27-41. Universitetsforlaget, Oslo. Akert, K., Moor, H., and Pfenninger, K. (1971 ). Advan. Cytopharmacol. 1, 273-290. Akert, K., Pfenninger, K., Sandri, C., and Moor, H. (1972). I n “The Structure and Function of Synapses” (G. D. Pappas and D. P. Purpura, eds.), pp. 67-86. Raven Press, New York. Akert, K., Livingston, R. B., Moor, H., and Streit, P. (1974). I n “Neurovegetative Transmission Mechanisms” (B. Csillik and J. Ariens Kappers, e d . ) , pp. 1-12. Springer-Verlag, Berlin and New York. Aldridge, W. N., and Johnson, M. K. (1959). Biochem. J . 73, 270-276. Allison, A. C. (1972). I n “Cell Interactions” (L. G. Silvestri, ed.), pp. 156-161. North-Holland Publ., Amsterdam. Anderson, M. J., and Cohen, M. W. (1973). Abstr. SOL. Neurosci., 3rd Annu. Meet. p. 228. Aquilonius, S.-M., Flentge, F., Schuberth, J., Sparf, B., and Sundwall, A. (1973). J. Neurochem. 20, 1509-1521. Augustinsson, K. B., and Nachmansohn, D. (1949). Science 110, 198. Balthazar, K. (1952). Arch. Psychiat. Nervenkr. 188, 345-378. Barnard, E. A., Wieckowski, J., and Chiu, T. H. (1971). Nature ( L o n d o n ) 234, 207-209. Barrnett, R. J. (1962). J . Cell B i d . 12, 247-262. Barron, K. D., Bernsohn, J., and Ordinario, A. T. (1967). J . Histochem. Cytochem. 15, 782. Bass, L., and Moore, W. J. (1966). Proc. N u t . Acad. Sci. U S . 55, 1214-1217. Bauman, A,, Benda, P., and Rieger, F. (1972). Brain Res. 45, 183-192. Belleau, B., and DiTullio, V. (1971). Can. J . Biochem. 49, 1131-1133. Bed, S., Puszkin, S., and Nicklas, W. J. (1973). Science 179, 441-446. Betz, W., and Sakmann, B. (1971). N a t u f e ( L o n d o n ) , N e w Biol. 232, 94-95. Bhagat, B., and Lockett, M. F. (1962). J . Pharm. Pharnacol. 14, 37. Birks, R., Katz, B., and Miledi, R (1960). J . Physiol. ( L o n d o n ) 150, 145-168. Bloom, F. E., and Aghajanian, G. K. (1966). Science 154, 1575-1577. Bloom, F. E., and Aghajanian, G. K. (1968). J . Ultrastruct. Res. 22, 361-375. Bodian, D. (1966). Bull. Johns Hopkins Hosp. 119, 16-45. Bogusch, G. (1973). Histochemie 33, 39-46. Bon, S., Rieger, F., and Massoulie, J. (1973). Eur. J . Biochem. 35, 373-379. Browning, E. T., and Schulman, M. P. (1968). J . Neurochem. 15, 1391-1405. Burdick, C. J., and Strittmatter, C . F. (1965). Arch. Biochem. Biophys. 109, 293-301. Burkhalter, A,, Jones, M., and Featherstone, R. M. (1957). Proc. S O L . Exp. Biol. Med. 96, 747-750. I
134
BERTALAN CSILLIK
Burn, J. H., and Rand, M. J. (1959). Nature ( L o n d o n ) 184, 163-165. Burt, A. M., and Silver, A. (1973). Brain Res. 62, 509-516. Cartaud, J., Benedetti, L., Cohen, J. B., Meunier, J. C., and Changeux, J. P. (1973). FEBS (Fed. Eur. Biochem. S o c . ) Lett. 33, 109-113. Ceccarelli, B., Hurlbut, W. P., and Mauro, A. (1972). J. Cell B i d . 54, 30-38. Ceccarelli, B., Hurlbut, W. P., and Mauro, A. (1973). J . Cell B i d . 57, 499-524. Chan, S. L., Shirachi, D. Y . , and Trevor, A. J. (1972). J. Neurochem. 19,437-447. Chang, C. C., Chen, T. F., and Chuang, S. T. (1973). J . Physiol. ( L o n d o n ) 230, 613-618. Changeux, J. P. (1966). M o l . Pharmacol. 2, 369-392. Changeux, J. P. (1973). Neurosci. Res. Program, Bull. 11, 246-252. Changeux, J. P., Podleski, T. R., and Meunier, J. (1969). J. Gen. Physiol. 54, 225-244. Changeux, J. P., Kasai, M., Huchet, M., and Meunier, J. C. (1970a). C. R. Acad. Sci., Ser. D. 270, 2864-2867. Changeux, J. P., Kasai, M., and Lee, C. Y. (1970b). Proc. Nut. Acad. Sci. U.S. 67, 1241-1247. Chen, I.-L., and Lee, C. Y. (1970). Virchows Arch., B 6, 318-325. Cheney, D. L., Gubler, C. J., and Jaussi, A. W. (1969). 1.Neurochem. 16, 1283-1291. Chokroverty, S., Paramesvar, K. S., and Co, C. (1971). J Histochent Cytochem. 19, 798-800 Clark, A W., Hurlbut, W. P., and Mauro, A. (1972). J . Cell B i d . 52, 1-14. Cohen, M. W. (1972). Brain Res. 41,457-463. Cooke, J. D., and Quastel, D. M. (1973a). J . Physiol. ( L o n d o n ) 228, 377-405. Cooke, J. D., and Quastel, D. M. ( 1973b). J. Physiol. ( L o n d o n ) 228, 407-434. Cooke, J. D., Okamoto, K., and Quastel, D. M. (1973). J . Physiol. ( L o n d o n ) 228, 459-497. Couteaux, R. (1947). Reo. Can. B i d . 6, 563-711. Couteaux, R. (195 1 ) . Arch. Int. Physiol. 59, 526-537. Couteaux, R., and Ptcot-Dechavassine, M. (1970). C. R. Acad. Sci., Ser. D 271, 2346-2349. Couteaux, R., and Taxi, J. (1952). Arch. Anat. Microsc. Morphol. Exp. 41, 352-392. Crkvier, M., and BClanger, L. F. ( 1955). Science 122, 556-557. Crone, H. D., and Freeman, S. E. (1972). J. Neurochem. 19, 1207-1208. Csillik, B. (1960). Z . Zellforsch. Mikrosk. Anat. 52, 150-162. Csillik, B. (1963). J . Cell Biol. 17, 57 1-586. Csillik, B. (1965). “Functional Structure of the Post-Synaptic Membrane in the Myoneural Junction.” Akadtmiai Kiadb, Budapest. Csillik, B. (1967). Ph.D. Thesis, Szeged (in Hungarian). Csillik, B. (1974). I n “Neurovegetative Transmission Mechanisms” (B. Csillik and J. Ariens Kappers, eds.), pp. 13-42. Springer-Verlag, Berlin and New York. Csillik, B., and Bense, S. (1971). A c t a Biol. (Budapest) 22, 131-139. Csillik, B., and Job, F. (1967). Nature ( L o n d o n ) 213, 508-509. Csillik, B., and Knyihlr, E. (1968a). J. Cell Sci. 3, 529-538. Csillik, B., and Knyihlr, E. (1968b). Acta Biochim. Biophys. 3, 165-170. Csillik, B., and Slvay, G. (1958). Acta Neurooeg. 19, 41-52. Csillik, B., and Sgvay, G. (1963). Nature ( L o n d o n ) 198,399. Csillik, B., and Slvay, G. ( 1965). A c t a Physiol. 26, 337-342. Csillik, B., and Tbth, L. (1972). J. Histochem. Cytochem. 20, 385-387. Csillik, B., Schneider, I., and Kblmln, G. (1961). Acta Neuroveg. 22, 212-224.
SYNAPTOCHEMISTRY OF A CHOLINERGIC NEURON
135
Csillik, B., Job, F., KBsa, P., and SBvay, G. (1966). Acta Histochem. 25, 58-70. Csillik, B., Haarstad, V. B., and KnyihBr, E. (1970). J. Histochem. Cytochem. 18, 58-59. Csillik, B., Haarstad, V. B., Kiss, J., and Knyihhr, E. (1973a). Neurobiology 3, 131-139. Csillik, B., T6th, L., and Karcsu, S . (1973b). J . Neurocytol. 2, 441-455. Csillik, B., Knyihir, E., LBsz16, I., and Boncz, I. (1974). Brain Res. 70, 179-183. Curtis, D. R., and Ryall, R. W. (1966). Exp. Brain. Res. 2, 49-106. Dale, H. H. (1914). J . Pharmacol. Exp. Ther. 6, 147-190. Davis, D. A,, Wasserkrug, H. L., and Heyman, I. A. (1972). J . Histochem. Cytochem. 20, 161-172. Davison, P. F. (1970). Zn “The Neurosciences: Second Study Program” (F. 0. Schmitt, ed.), pp. 851-857. Rockefeller Univ. Press, New York. Dehlinger, P. J., and Schimke, R. T. (197 1 ) . J . B i d . Chem. 246, 2574-2583. Del Castillo, J., Rodriguez, A., and Romero, C. A. (1967). Ann. N . Y . A d . Sci. 144, 803-818. Dennis, M. (1972). J. Physiol. (London) 226, 79P-81P. De Robertis, E. (1958). Exp. Cell Res. Suppl. 5, 347-369. De Robertis, E., and Rodriguez de Lores Arnaiz, G. (1969). In “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 2, pp. 365-392. D o y h , M. P. L. N. (1840). Ann. Sci. Natur. Zool. B i d . Anim. [2] 14, 269-361. Drachman, D. B. (1968). Growth Il‘erv. Syst., Ciba Found. Symp. pp. 251-272. Drachman, D. B. (1972). J . Physiol. (London) 226, 619-627. Dreyer, F., Peper, K., Akert, K., Sandri, C., and Moor, H. (1973). Brain Res. 62, 373-380. Duchen, L. W. (1971). J . Neurol. Sci. 14, 37-45. Duchen, L. W., and Tonge, D. A. (1973). J . Physiol. (London) 228, 157-172. Eccles, J. C. ( 1964). “The Physiology of Synapses.” Spring-Verlag, Berlin and New York. Edelstein, S. J. (1972). Biochern. Biophys. Res. Commun. 48, 1160-1 165. Edelstein, S . J. (1974). I n “Functional Linkage in Biomolecular Systems” (F. 0. Schmitt, ed.). In press. Eldefrawi, M. E., and Eldefrawi, A. T . (1972). Proc. Nat. Acad. Sci. U.S. 69, 1776-1780. Eranko, O., and Teravainen, H. (1967). J . Histochem. Cytochem. 15, 399-403. Eranko, O., Harkonen, M., Kokko, A., and Raisinen, L. (1964). J . Histochem. Cytochem. 12, 570-581. Eranko, O., Rechardt, L., and Hanninen, L. (1967). Histochemie 8, 369-376. Erulkar, S. D., Nichols, C. W., Popp, M. B., and Koelle, G. B. (1968). J . Histochem. Cytochem. 16, 128-135. Fambrough, D. M., and Hartzell, H. C . (1972). Science 176, 189-191. Ferry, C. B., and Marshall, A. R. (1971). J. Physiol. (London) 219, 33P-34P. Filogamo, G. ( 1969). Excerpta M e d . Found. Znt. Congr. Ser. 199, 709. Filogamo, G., and Gabella, G. (1961 ) . Acta Anat. 63, 199-214. Filogamo, G., and Gabella, G. (1967). Arch. B i d . 78, 9-60. Filogamo, G., and Marchisio, P. C. ( 197 1) . I n “Neurosciences Research” (S. Ehrenpreis and 0. C. Solnitzky, eds.), Vol. 4, pp. 29-64. Academic Press, New York. Fonnum, F. (1973). Brain Res. 62, 497-507. Fonnum, F., Frizell, M., and Sjpstrand, J. (1973). J . Neurochem. 21, 1109-1120.
136
UERTALAN CSILLIK
Friendenberg, R. M., and Seligman, A. M. (1972). J . Histochem. Cytochem. 20, 771-792. Furmanski, P., Silverman, D. J., and Lubin, M. (1971). Nature ( L o n d o n ) 233, 4 13-4 15. Galley, K. T., De Sorgo, M., and Prim, W. (1973). Biochem. Biophys. Res. Commun. 50, 300-307. Gauthier, F. G., and Schaeffer, S. F. (1974). J. Cell Sci. 14, 113-137. Gerebtzoff, M. A. (1953). Acta Anat. 19, 366-379. Gerebtzoff, M. A. ( 1959). “Cholinesterases.” Pergamon, Oxford. Gerebtzoff, M. A., Goffard, P., and Csillik, B. (1964). Cited by Csillik (1967). Giacobini, E. (1959). Acta Physiol. Scand. 45, Suppl. 156, 1-45. Giacobini, G. (1972). J . Neurochem. 19, 1401-1403. Goldberg, A. L., and Singer, J. J. (1969). Proc. Nut. Acad. Sci. U S . 64, 134-141. Gonzenbach, H. R., and Waser, P. G. (1973). Brain Res. 63, 167-174. Goodwin, B. C., and Sizer, I. W. (1965). Decelop. B i d . 11, 136-153. Grafius, M. A., Bond, H. E., and Millar, D. B. (1971). Eur. 1. Biochem. 22, 382-390. Grampp, W., Harris, J. B., and Thesleff, S. (1972). 1. Physiol. ( L o n d o n ) 221, 743-754, Gray, E. G. (1959). J. Anat. 93, 420. Gray, E. G. (1963). J . Anat. 97, 101-106. Gray, E. G. (1966). Int. Rev. Gen. Exp. 2001.2, 139. Griffith, D. L., and Bondareff, W. (1973). Amer. J . Anat. 136, 549-556. Gruber, H., and Zenker, W. (1973). Brain Res. 51, 207-214. Guth, I,. ( 1968). Annu. R e v . Physiol. 48, 645-687. Guth, L. ( 1969). Neurosci. Res. Porgram, Bull. 7 , 1-73. Gwyn, D. G., and Flumerfelt, B. A. (1971). Brain Res. 34, 193-198. Hajbs, F., Prymak, E. K., and Kerpel-Fronius, S. (1970). Acta Histochem. 35, 114-122. HalLsz, N., PLrducz, A,, and Job, F. (1971). Acta B i d . ( S t e g e d ) 17, 143-152. Hall, Z. W., and Kelly, R. B. (1971). Nature ( L o n d o n ) ,New B i d . 232, 62-63. Hammes, G. G., Molinoff, P. B., and Bloom, F. E. (1973). Neurosci. Res. Porgram, Bull. 11, 115-294. Hanin, I., Massarelli, R., and Costa, E. (1972). In “Studies of Neurotransmitters at the Synaptic Level” (E. Costa, L. L. Iversen, and R. Paoletti, eds.), pp. 181-202. Raven Press, New York. Harkins, J., Arsenault, M., Schlesinger, K., and Kates, J. (1972). Proc. Nut. Acad. Sci. U S . 69, 3161-3164. Heaton, J., Buckley, G. A., and Evans, R. H. (1972). Experientia 28, 503-504. Hebb, C. O., and Morris, D. ( 1969). I n “The Structure and Function of Nervous Tissue” (G. H. Bourne, ed.), Vol. 3, pp. 25-60. Academic Press, New York. Hebb, C. O., and Waites, G. M. H. (1956). J. Physiol. ( L o n d o n ) 132, 667-671. Hebb, C. O., KrnjeviE, K., and Silver, A. (1964). J . Physiol. ( L o n d o n ) 171, 504-513. Heinrich, C. P., Stadler, H., and Weiser, H. (1973). J. Neurochem. 21, 1273-1281. Heuser, J. E., and Reese, T. S. (1973). J. Cell Biol. 57, 315-344. Hollunger, E. G., and Niklasson, B. H. (1973). J. Neurochem. 20, 821-836. Hosli, E., and Hosli, L. (1971). Brain Res. 30, 193-197. Ishii, T., and Friede, R. L. (1967). Znt. Reu. Neurobiol. 10, 231-275. Itokawa, Y., Schulz, R. A., and Cooper, J. R. (1972). Biochim. Biophys. Acta 266, 293-299.
SYNAPTOCHEMIS’TRY OF A CHOLINERGIC NEURON
137
Jain, M. F. (1972). “The Bimolecular Lipid Membrane,” p. 92. Van Nostrand-Reinhold, Princeton, New Jersey. Jain, M. K., Mehl, L. E., and Cordes, E. H. (1973). Biochem. Biophys. Res. Commun. 51, 192-197. Jankowska, E., and Lindstrom, S. (1971). Acta Physiol. Scand. 81, 428-430. Jankowska, E., and Lindstrom, I;. (1972). J. Physiol. ( L o n d o n ) 226, 805-823. Johnson, G. A., Boukma, S. J., Lahti, R. A., and Mathews, J. (1973). 1.Neurochem. 20, 1387-1392. Johnston, P. V., and Roots, B. I. (1972). “Nerve Membranes,” p. 94. Pergamon, Oxford. Jones, D. G. (1972). I n “The ,Structure and Function of Nervous Tissue” (G. H. Bourne, ed.), Vol. 6, pp. 81-129. Academic Press, New York. Jones, M., Featherstone, R. M., and Bonting, S . L. (1956). J . Pharmacol. ExP. Ther. 116, 114-118. Jones, S. F., and Kwanbunbumpen, S. (1970). J. Physiol. ( L o n d o n ) 207,31-50. Job, F., Halisz, N., and Pirducz, .A. ( 1973). J . Neurocytol. 2, 393-405. Jost, P. C., and Griffith, 0. EL (1973). Arch. Biochem. Biophys. 159, 70-81. Juntunen, J., and Teravainen, H. (1972). Histochemie 32, 107-1 12. Karassik, V. M. (1946). Usp. Sorvem. R i d . 21, 1. Karlin, A., and Cowburn, D. (1973). P r o c . N a t . Acad. Sci. U.S. 70, 3636-3640. K h a , P. (1968). A ature ( L o n d o n ) 218, 1265-1267. Kato, G., Tan, E., and Yung, J. (1972). Nature ( L o n d o n ) , New Biol. 236, 185. Kim, S. V. (1972). Experienta 28, 537-538. Kimura, M., and Kimura, I. (1973). Nature ( L o n d o n ) , New B i d . 241, 114-115. Knyihir, E., Liszl6, I., and Csillik, B. (1973). Neurobiology 3, 327-334. Koelle, G. B. (1951). 1. Pharmacol. Exp. T h e r . 103, 153-171. Koelle, G. B. (1954). J. Comp. h’eurol. 100, 211-228. Koelle, G. B. (1961). Nature ( L o n d o n ) 190, 208-211. Koelle, G. B. (1962). J . Pharna. Pharmacol. 14, 65-90. Koelle, G. B. (1963). I n “Handbuch der experimentellen Pharmakologie” (G. B. Koelle, subed.), Vol. 15, pp. 187-298. Springer-Verlag, Berlin and New York. Koelle, G. B. (1971). A n n . N . Y . .4cad. Sci. 183, 5-20. Koelle, G. B., and Friedenwalcl, J. S. (1949). Proc. SOC.Exp. B i d . M e d . 70, 6 17-622. Koelle, G. B., Davis, R., and Devlin, M. (1968). J . Histochem. Cytochem. 16, 754-764. Koelle, W. A,, Sharifi Hossaini, K., Akbarzadeh, P., and Koelle, G. B. (1970). J . Histochem. Cytochem. 18, 812-819. Koenig, J. (1973). Brain Res. 62, 361-365. Koeppe, R. E., O’Neal, R. M., and Hahn, C. H. (1964). J . Neurochem. 11, 695-699. Korneliussen, H. (1972). Z . Zellforsch. Mikrosk. Anat. 130, 28-57. Kosower, E. M., and Werman, R. (1971). Nature ( L o n d o n ) ,New B i d . 233, 121-123. Kremzner, L. T., and Fei, S. C. (1971). Fed. Proc., Fed. Amer. SOC. Exp. Biol 30, 1193. Kuffler, S . W. ( 1953). Naunyn-,Schnziedebergs Arch. Exp. Pathol. Pharmakol. 220, 116-135. LPszl6, I., and Knyihir, E. (1975). J. Neural Transmission ( W i e n ) 36, 1232141. Lentz, T . L. (1969). J. Cell B i d . 42, 431-443. I,entz, T. L. (1970). J. Cell B i d . 47, 423. Lentz, T. L. (1972). J. Cell Biol. 55, 93-103.
138
BERTALAN CSILLIK
Leuzinger, W. (1969). Progr. Brain Res. 31, 241-245. Leuzinger, W., and Schneider, M. (1972). Experientia 28, 256-257. Lewis, P. R., and Shute, C. C. D. (1964). J . Physiol. (London) 175, 5P-6P. Lewis, P. R., and Shute, C. C. D. (1965). J . Anat. 99, 941P. Lewis, P. R., and Shute, C. C. D. (1966). J . Cell Sci. 1, 381-390. Loewi, O., and Navratil, E. (1926). Pf7uegers Arch. Gesamte Physiol. Menschen Tiere 214. 678. Lubinska, L., and Niemerko, S. (1971 ) . Brain Res. 27, 329-342. Liillmann-Rauch, R. (1971). Z . Zellforsch. Mikrosk. Anat. 121, 593-603. McBride, W. J., and Cohen, H. (1972). Brain Res. 41, 489-493. McCaman, R. E., Rodriguez de Lores Arnaiz, G., and De Robertis, E. (1965). J . Neurochem. 12, 927-935. MacIntosh, F. C. (1959). Can. J . Biochem. 37, 343-356. McIntyre, A. R. (1959). I n “Curare and Curare-like Agents” (D. Bovet, F. BovetNitti, and G. B. Marini-Bettolo, eds.), pp. 21 1-218, Elsevier, Amsterdam. McIntyre, A. R., Downing, F. M., Bennett, A. L., and Dunn, A. L. (1950). Proc. SOC.Exp. B i d . M e d . 74, 180-185. McMahan, V. J., Spitzer, N. C., and Peper, K. (1972). Proc. Roy. Ser. SOC. B 181, 421-430. Madarkz, B., and HLmori, J. (1971). Acta B i d . (Budapest) 22, 309-319. Magherini, P. C., Pompeiano, O., and Thoden, U. (1971). Brain Res. 33, 514-518. Malthe-S~rensen,D., Eskeland, T., and Fonnum, F. (1973). Brain Res. 62, 517-522. Manocha, S. L., and Shantha, T . R. (1969). Zn “The Structure and Function of Nervous Tissue” (G. H. Bourne, ed.), Vol. 2, pp. 137-200. Academic Press, New York. Marchbanks, R. M. (1970). Symp. Znt. SOC.Cell Biol. 8, 115-135. Mendel, B., Mundell, D. B., and Rudney, H. (1943). Biochem. J . 37, 473-4-76. Meunier, J. C., Sugiyama, H., Cartaud, J., Sealock, R., and Changeux, J. P. (1973). Brain Res. 62, 307-315. Miledi, R., and Potter, L. T. (1971). Nature (London) 233, 599-603. Molinoff, P. B., and Potter, L. T. (1972). Biochem. Psychopharmacol. 111-134. Mooser, G., and Sigman, D. S. (1972). Biochem. Biophys. Rex. Commun. 48, 559-564. Morgan, I. G., Zanetta, J. P., Breckenbridge, W. C., Vincendon, G., and Gombos, G. (1973). Brain Res. 62, 405-411. Nachmansohn, D. (1955). Ergeb. Physiol. Biol. Chem. Exp. Pharmakol. 48, 575-683. Namba, T . (1971). Exp. Neurol. 33, 322-328. Navaratnam, V., and Lewis, P. R. (1970). Brain Res. 18, 41 1-425. Novikoff, A. B., Essner, E., and Quintana, N. (1964). Fed. Proc., Fed. Amer. Soc. Exp. B i d . 23, 1010. Novikoff, P. M., Novikoff, A. B., Quintana, N., and Hauw, J.-J. (1971). J . Cell B i d . 50, 859-886. Nyberg-Hansen, R., Rinvik, E., Aarseth, P., and Barstad, J. A. B. (1969). Histochemie 20, 40-45. Oh, T. H., and Johnson, D. D. (1972). Exp. Neurol. 37, 360-370. Ord, M. G., and Thompson, R. H. S. (1951). Biochem. J . 49, 191 Palay, S. L., Sotelo, C., Peters, A., and Orkand, P. M. (1968). J . Cell B i d . 38, 193-201. Pappas, G. D., Peterson, E. R., and Masurovsky, E. B. (1971). Ann. N . Y . Acad. Sci. 183, 33-45.
SYNAPTOCHEMISTRY OF A CHOLINERGIC N E U R O N
139
Parisi, M., Rivas, E., and De Robertis, E. (1971). Science 172, 56-57. Peters, A,, Proskauer, C. C., and Kaiserman-Abramof, I. R. (1968). J. Cell Biol. 39, 604-619. Pfenninger, K., Sandri, C., Akert, K., and Eugster, C. H. (1969). Brain Res. 12, 10-18. Pfenninger, K., Akert, K., and !jandri, C. (1971). Phil. Trans. Roy. Soc. London, S e r . B 216, 387. Pfenninger, K., Akert, K., Mom, H., and Sandri, C. (1972). J . Neurocytol. 1, 129-149. Pollard, H. B., Miller, A,, and Cox, G . C. (1973). J. Supramol. Struct. 1, 295-306. Potter, L. T . (1970a). Zn “Handbook of Neurochemistry” ( A . Lajtha, ed.), Vol. 4, pp. 263-284. Plenum, New York. Potter, I,. T . (1970b). Advan. Biochem. Psychopharmacol. 2, 163-168. Potter, L. T. (1972). I n “The !jtructure and Function of Nervous Tissue” (G. H. Bourne, ed.), Vol. 4, pp. 1055128. Academic Press, New York. Raftery, M. A. (1973). Neurosci. .Res. Program, Bull. 11, 264-268. Raftery, M. A. (1975). I n “Functional Linkage in Biomolecular Systems” (F. 0. Schmitt, ed.). In press. Raftery, M. A,, Schmidt, J., Martinez-Carrion, M., Moody, T., Vandlen, R., and Duguid, J. (1973). J . Supranzol. Struct. 1, 360-367. Ram& y Cajal, S. (191 1 ) . “Histologie du systeme nerveux de I’homme et des vertibrts.” Maloine, Paris. Rang, H . P. (1973). Neurosci. Rer. Program, Bull. 11, 220-224. Ranish, N., and Ochs, S. (1972). J. Neurochem. 19, 2641-2649. Renshaw, B. (1941). J . Neurophysiol. 4, 167-183. Renshaw, B. (1946). J. Neurophy.riol. 9, 191-204. Reale, E., and Luciano, L. (1964). J. Histochem. Cytochem. 12, 713-715. Rieger, F., Benda, P., and Bauman, A. (1973a). FEBS (Fed. Eur. Biochem. Soc.) Lett. 32, 62-68. Rieger, F., Bon, S., and MassouliC, J. (197313). Eur. 1. Biochem. 34, 539-547. Robertson, J. D. (1956). J. Bioph;vs. Biochem. Cytol. 2, 381-394. Roepke, M. H. (1937). J . Pharmacol. 59, 264. Rose, S., and Glow, P. H. (1967). Exp. Neurol. 18, 267-275. Salpeter, M. M., and Eldefrawi, Pd. E. (1973). J . Histochem. Cytochem. 21, 769-778. Sivay, G., and Csillik, B. (1956). Acta Morphol. 6, 289-297. Scheibel, M. E., and Scheibel, A. B. (1964). Anat. Rec. 148, 322P. Scheibel, M. E., and Scheibel, A. 13. (1966). Arch. Ztal. Biol. 104, 328-353. Scheibel, M. E., and Scheibel, A. B., (1971). Brain, Behaoiour, B Evolution 4, 53-93. Schubert, D., Tarikas, H., Hami!;, A. J., and Heinemann, S. (1971). Nature ( L o n d o n ) , New Biol. 233, 79-80. Schueler, F. W. (1955). J . Pharmacol. Exp. T h e r . 115, 127-143. Silver, A. (1967). Znt. Reu. Neurobiol. 10, 57-109. Silver, A,, and Wolstencroft, J. H. (1971). Brain Res. 34, 205-227. Singer, S. J. (1971). In “Structure and Function of Biological Membranes” (L. I. Rothfield, ed.), pp. 145-222. Academic Press, New York. Singer, S. J., and Goldberg, A.. L. (1970). A d x z n . Biochem. Psychopharmacol. 3, 335-348. Singer, S . J., and Nicolson, G. L. (1972). Science 175, 720-731. Singer, S. J., and Rothfield, L. I. (1973). Neurosci. Res. Program, Bull. 11, 1-86.
140
BERTALAN CSILLIK
Stalc, A., and ZupanEiE, A. 0. (1972). Nature ( L o n d o n ) , New B i d . 239, 91-92. Stedman, E., Stedman, E., and Easson, L. H. (1932). Biochem. J. 26, 2056. Stein, W. D. (1967). “The Movement of Molecules across Cell Membranes” pp. 10, 23, and 29. Academic Press, New York. Stoeckenius, W. (1963). J. Cell B i d . 12, 221-229. Stretton, A. 0. W., and Kravitz, E. A. (1968). Science 162, 132-134. Szentigothai, J. (1951). Acta Morphol. 1, 81-94. Szentigothai, J. (1958). Acta Morphol. 8, 287-309. Szentigothai, J. ( 1967). Electroencephalogr. Clin. Neurophysiol., Suppl. 25, 4-19. Tanaka, C., Itokawa, Y., and Tanaka, S. (1973). J. Histochem. Cytochem. 21, 81-86. Thesleff, S. (1973). Neurosci. Res. Program, Bull. 11, 227-231. Thomas, R. C., and Wilson, V. J. (1965). Nature (London) 206, 21 1-213. Torack, R., and Barnett, R. J. (1962). Exp. Neurol. 6, 224-244. Toschi, G. (1959). Exp. Cell Res. 16, 232-255. TuEek, S., and Cheng, S.-C. ( 1970). Biochim. Biophys. Acta 208, 538-540. Tuffery, A. R. (1971). J. Anat. 110, 221-247. Turbow, M. M., and Burkhalter, A. (1968). Develop. Biol. 17, 233-244. Uchizono, K. (1965). Nature (London) 207, 642-643. Van der Kloot, W., and Kita, H. (1974). BioScience 24, 13-17. von Muralt, A. ( 1946). “Die Signalubermittlung im Nerven.” Basel. von Muralt, A. (1958). Exp. Cell Res., Suppl. 5, 72-79. Vizi, E. S. (1974). In “Neurovegetative Transmission Mechanisms” (B. Csillik and J. Ariens Kappers, eds.) , pp. 61-78. Springer-Verlag, Berlin and New York. Waser, P. G. (1967). Ann. N . Y . Acad. Sci. 144, Art. 2, 737-755. Waser, P. G., and Nickel, E. (1969). Progr. Brain Res. 31, 157-169. Wenthold, R. J., Mahler, H. R., and Moore, J. (1974). J. Neurochem. 22, 941-943. Whittaker, V. P. (1959). Biochem. J. 72, 694-706. Whittaker, V. P. (1965). Progr. Biophys. Mol. Biol. 15, 39-91. Whittaker, V. P. (1969a). I n “Handbook of Neurochemistry” (A. Lajtha, e d . ) , Vol. 2, pp. 327-364. Plenum, New York. Whittaker, V. P. (1969b). I n “The Structure and Function of Nervous Tissue” (G. H. Bourne, ed.), Vol. 3, pp. 1-24. Academic Press, New York. Whittaker, V. P., Michaelson, I. A., and Kirkland, R. J. A. (1964). Biochem. J. 90, 293-303. Willis, W. D. (1971). Brain, Behaviour, d Evolution 4, 5-52. Willis, W. D., and Willis, J. C. (1964). Nature (London) 204, 1214-1215. Wilson, I. B. (1967). Ann. N . Y . Acad. Sci. 144, Art. 2, 664-674. Wurzel, M. (1956). Bull. Res. Counc. Isr., Sect. A 5, 303. Wurzel, M. (1967). Ann. N . Y . Acad. Sci. 144, Art. 2, 694-704. Zacks, S. I. ( 1964). “The Motor Endplate.” Saunders, Philadelphia, Pennsylvania. Zelenh, J., and Szentigothai, J. (1957). Acta Histochem. 3, 284-296. ZiranEiE, A. 0. (1971). Glas, Srp. Akad. Nauka Vmet., O d . M e d . Nauka 24, 37-41. ZupanEiE, A. 0. (1953). Acta Physiol. Scand. 29, 63. ZupanEiE, A. 0. (1967). Ann. N . Y . Acad. Sci. 144, Art. 2, 689-693.
ION AND ENERGY METABOLISM OF THE BRAIN AT THlE CELLULAR LEVEL By Leif Hertz and Arne Schousboe
Department of Anatomy, University of Saskatchewan, Saskatoon, Canada, and Department of Biochemistry A., University of Copenhagen, Copenhagen, Denmork
I. Introduction
.
11. Complexity of Brain
A. CellTypes
.
.
B. Interstitial Space . C Techniques Used for Stud,y of Metabolism in Individual Cell Types D. Measurements of Ion Concentrations in the Interstitial Space . E. Ontogenetic and Phylogenetic Development F. Compartmentation . III. Energy Metabolism . A. Oxygen Consumption. B. High-Energy Phosphates . C. Intermediary Metabolism . IV. Ion and Water Metabolism A. Comparison between Ion and Water Metabolism in Vi’ivoand in Vitro B. Events at the Cellular Level . V. Concluding Remarks . References
.
.
.
.
.
. . . . . . . . . . . . .
. . . .
141 142 142 144 145 147 148 150 151 151 160 166 176 176 186 191 193
I. Introduction
I n spite of an extensive knowledge of energy metabolism and ion composition of the brain and its constituent structures (e.g., cortex) as a whole, only little information was, until recently, available about ion and energy metabolism a t the cellular 1 evel. Owing to the pronounced morphological heterogeneity of the tissue such information is, however, essential for a deeper understanding of brain function. This heterogeneity involves that procedures leading to oppositely directed changes in different cell types (cf., e.g., HydCn, 1967a) or in cells and interstitial fluid (e.g., release of potassium from cells to their surroundings) may become totally obscured when conventional macroanalysis IS performed. Recently developed micromethods have given at least some insight into such phenomena, and in the following, the emergent pattern of ion and energy metabolism a t the cellular level 141
142
LEIF HERTZ A N D ARNE S C H O U S B O E
will be discussed. Since the metabolic characteristics of some of the preparations used conceivably might be heavily impaired, emphasis will be placed upon a comparison between findings at the cellular level, in other, less disintegrated in vitro preparations and in vivo. The reason for treating ion and energy metabolism together is that these two parameters probably are closely linked (cf., Section V, Concluding Remarks). Accordingly, effects of ions (especially K') on energy metabolism and of adverse metabolic conditions on ion distribution will be treated in some detail.
II. Complexity of Brain
A. CELLTYPES Two main elements, i.e., nerve cells and glia cells, constitute the bulk of nervous tissue, but each of these may on morphological grounds be divided into several subgroups. The nerve cells may be classified morphologically according to the size and shape of their cell bodies (perikarya) and the number, length, and mode of branching of their processes, or pharmacologically according to the transmitter involved in synaptic transmission. Only little is known about differences in energy metabolism between different types of neurons (Kato and Lowry, 1973). The nerve cell perikarya account for at most 5-10% of the volume in mammalian brain cortex (von Economo, 1926; Peters and Flexner, 1950; Haug, 1956, 1960; Rebhan, 1956; SchadC and Baxter, 1960). The volume occupied by their processes is several times larger, and it has been estimated that the proximal and distal dendrites together account for about one-fourth of the total volume of the gray matter (SchadC and Baxter, 1960; SchadC et al., 1964). The nerve endings account for another 15-25% of the brain volume (Salganicoff and Koeppe, 1968; Bloom and Iversen, 1971) . The glia cells are traditionally divided into fibrous and protoplasmatic astrocytes, oligodendrocytes, and microglia. I n gray matter, the characteristic glia cells are the protoplasmatic astrocyte (Rambn y Cajal, 1909; Wolff, 1970) and the oligodendrocyte, whereas the fibrous astrocyte is found mainly in white matter. The volume occupied by easily recognizable glia cell bodies is relatively small, although the glia cells may outnumber the nerve cells by a factor of about 5 to 10 (Nurnberger and Gordon, 1957; Pope, 1958; Hild, 1961; Blinkov and Glezer, 1968; Johnston and Roots, 1972). Thus, a considerable fraction of the tissue volume remains to be accounted for, and in gray matter the main part of this fraction is made u p by the so-called neuropil. This term refers to a rather badly defined, intricately interwoven network of minute neuronal and glial processes, which (Fig. 1) taper into
BRAIN METABOLISM AT THE CELLULAR LEVEL
143
FIG. 1. Line drawing of a n electron micrograph of stratum radiatum of regio superior of the rat hippocampal cortex. T h e micrograph was one of series facilitating identification of the structures. Scale is I fim. D, Dendritic shaft or larger branch; d, finer dendritic branch; s, spine; t, axon swelling with synaptic vesicles; g, astroglial process; A,axon. From Westrum and Blackstad (1962).
lamellar sheets 200-1000 A thick (Pappas and Purpura, 1961 ; Wolff, 1965). The ratio between glia cells and nerve cells in the neuropil is uncertain and probably varies in different parts of the nervous system. I n the hippocampal cortex-from which Fig. 1 was drawn-astroglia cells thus seem to constitute only 556% (Blackstad, 1967), whereas Wolff (1970) found 25-30% astrocytes in the occipital cortex of the rat, and Hydtn (196713) estimated that about nine-tenths of the Deiters’ cell surroundings are made up by glia cells. The previously mentioned observation by SchadC and coworkers (Schadt and Baxter, 1960; Schadt et al., 1964) that the dendrites altogether occupy about one-fourth of the brain cortex, supports the concept that about half of its volumcx may be occupied by glia cells (cf. also Schultz et al., 1957; Friede, 1970). It is an impression shared by several authors that the glia cells contain fewer mitochondria than the nerve cells (e.g., Windle, 1958; Abood, 1969),
144
LEIF HERTZ A N D ARNE S C H O U S B O E
but others have described a relatively abundant presence of mitochondria in at least some types of glia cells (Wyckoff and Young, 1956; Farquhar and Hartmann, 1957; De Robertis and Gerschenfeld, 1961; Mugnaini and Walberg, 1964).
B. INTERSTITIAL SPACE Besides the cellular elements the brain contains an extracellular compartment, and for years a controversial question has been the magnitude of the extracellular space. Basing their conclusions on electron micrographs, many investigators (e.g., Horstmann and Meves, 1959; Kuffler and Potter, 1964) have estimated that only about 5 % of the total volume in the brain cortex is occupied by extracellular fluid. This fluid is mainly localized in intercellular clefts of about 150 A in width. On the other hand, Van Harreveld and co-workers have reported that the extracellular space is considerably larger in vivo, but shrinks during the period between the arrest of the circulation and conventional fixation (Van Harreveld, 1962, 1966) ; this concept is supported by their finding that a special fixation technique yields an extracellular space of about 20% (Van Harreveld et al., 1965; cf., however, also Kuffler and Nicholls, 1966; Ibata et al., 1971). This space is mainly found between nonmyelinated axons, whereas other cellular elements (and thus also neurons and glia cells) are separated by the narrow slits described above (Van Harreveld and Malhotra, 1967). Accordingly, compounds (e.g., ions) which are released from cells may in any case become confined to such a small volume that their concentrations become high enough to exert biochemical effects on adjacent cells, and an elevation of the extracellular potassium concentration during neuronal activity has in recent years been convincingly demonstrated (Grossman et al., 1969; Fertziger and Ranck, 1970; Grossman and Rosman, 1971; Dichter et al., 1972; Krnjevic and Morris, 1972; Vyklicky et al., 1972; Vyskocil et al., 1972; Prince et al., 1973; Ransom and Goldring, 1973b; Futamachi et al., 1974). As a further complication for evaluation of ion content and transport in the extracellular space the presence of an extracellular anionic substance has been reported (Bondareff, 1966, 1967; Pease, 1966). Conceivably some binding of cations to this substance may occur (cf. Adey, 1967), or the structuration of the space may cause a reduction or an increase (cf. Metzner, 1965) of diffusion rates. The demonstration of ATPase activity in the extracellular space of brain and retina (Torack and Barnett, 1963; Marchesi et al., 1964; Torack, 1965; ODaly, 1967; cf., however, also Torney, 1966) may even indicate capacity for active transport. Determinations of extracellular space in vitro are generally performed by aid of extracellular marker substances such as sucrose, thiocyanate, or
BRAIN METABOLISM AT T H E CELLULAR LEVEL
145
inulin (e.g., Davson and Spaziani, 1959; Pappius and Elliott, 1956a; Harvey and McIlwain, 1968). During recent years heterogeneity of the marker space (Cohen et al., 1968; Cohen and Lajtha, 1969, 1970, 1971; Cohen, 1972), and intracellular penetration of the markers (Allen, 1955; Nicholls and Wolfe, 1967; Schousboe and Hertz, 1969, 1971a), have, however, been demonstrated. By kinetic analysis of inulin washout, it is possible to obtain a more exact determination of the “true” extracellular space (Lund-Andersen and Hertz, 1973; Lund-Andersen, 1974), which is essential for determination of intracellular concentrations.
C.
T E C H N I Q U E S USED FOR STUDY OF METABOLISM I N INDIVIDUAL CELLTYPES
The histological heterogeneity of brain tissue has necessitated the development of methods to separate its constituent cells from each other. By aid of microdissection (Lowry, 1953, 1957; HydCn, 1959, 1960, 1967a), it is possible to obtain single nerve cells or clumps of glia cells, which are exceedingly well separated from each other, as can be seen in Fig. 2. A drawback of the method is that the nerve cells lose part of their dendrites, which is then a source of contamination in the glia cell samples (HydCn and Lange, 1961; Hamberger, 1963; Kuffler and Nicholls, 1966). I t must also be emphasized that experiments on isolated glia cell clumps do not allow any distinction between metabolic processes occurring inside the glia cells and those possibly occurring in the interstitial space (cf. Adey et al., 1963; Adey, 1967). Damage of the membranes has been observed (Roots and Johnston, 1964; Johnston and Roots, 1965), but this criticism was rejected by HydCn ( 1967a), and metabolically the samples seem surprisingly intact (cf. Tables I1 and 111). Cells from cultures grown under suitable conditions (Fig. 3) may similarly show a well maintained energy metabolism (Hartman et al., 1970; Lehrer et al., 1970; Schousboe et al., 1970; Booher et al., 1971b; Dittmann et al., 1973a,b,c), and cultivated glia cells and neurons have been useful in several biochemical and physiological studies (e.g., Hild et al., 1958; Hild, 1964; Sensenbrenner et al., 1968; Scott and Fisher, 1970). Large quantities of certain cell types may be obtained from established cell lines (e.g., Benda et al., 1968:; Seeds e t al., 1970; Shein et al., 1970; De Vellis and Kukes, 1973; Mandel et al., 1973; Pontkn 1973; Sato, 1973). Bulk-prepared nerve and glia cells (Fig. 4) may be obtained using differential centrifugation (e.g., Roots and Johnston, 1964; Rose, 1965, 1967; Bradford and Rose, 1967; Azcurra et al., 1969; Blomstrand and Hamberger, 1969, 1970; Rose and Sinha, 1969; Norton and Poduslo, 1970; Blomstrand, 1971) . Also these preparations may show considerable metabolic and physio-
146
LEIF HERTZ AND ARNE SCHOUSBOE
FIG. 2. U p p e r row: Three Deiters’ nerve cells dissected freehand, cleaned from surrounding glia, and photographed in a phase contrast microscope. Lower row: Three collections of glia cells, each of which originally surrounded the nerve cell situated above. From HydCn (1967a).
FIG. 3. Phase contrast micrographs of ( A ) migrated astrocytes from spinal cord of a 10-day-old chick embryo after 33 days of cultivation in a Rose chamber and ( B ) nerve cell clump from dissociated cerebral hemispheres of a 7-day-old chick embryo after 7 days of cultivation in a Falcon flask. Note homogeneity of cells. ( A ) From Booher et al. (1971a) ; ( B ) from Dittmann et al. ( 1 9 7 3 ~ ) .
BRAIN METABOLISM A T T H E CELLULAR LEVEL
147
FIG. 4. A group of neurons ( A ) and a glia cell clump ( B ) prepared by gradient centrifugation of rat brain cortex homogenates. The cell preparations were stained with acidic thionine. The bar indicates 5 pm in ( A ) and 10 pm in ( B ) . From Nagata et al. (1974).
logic activity, though they in all probability remain less intact than the microdissected cells (Dittmann et al., 1 9 7 3 ~ ) .By the use of the separation techniques, by investigations on tumor tissue (e.g., Victor and Wolf, 1937) or by aid of histological (e.g., Diamond et al., 1966) and autoradiographic (Altman, 1963; Droz and Leblond, 1963) techniques a considerable amount of information has been obtaincd about metabolism and function of individual cell types in the brain. I t must, however, also be kept in mind that each of these procedures has its own-often serious-limitations [see, e.g., the criticism raised against the Rose differential centrifugation technique by Cremer et al. (1968) and against the use of tumor tissue by Utley (1963)l. This may explain the often conflicting results and emphasize the need of attacking the same problem with several different techniques (cf. also Hertz, 1969). D. MEASUREMENTS OF IONCONCENTRATIONS IN THE INTERSTITIAL SPACE Direct measurements of ion concentrations (or rather activities) in the interstitial space have recently been made possible by the development of microelectrodes specifically sensitive to potassium or other ions (Hinke, 1961; Walker, 1971; Krnjevic and Morris, 1972; VyskoEil and KEG, 1972; Prince et al., 1973). Although the diameter of the electrode tip is small ( 1-3 pm) , it is still considerably larger than the narrow interstitial clefts of about 150 A (Horstmann and Meves, 1959; Van Harreveld and Malhotra, 1967) . The recorded increases in interstitial potassium concentrations after activity of adjacent neurons may thus reflect alterations in a somewhat expanded extracellular area arid therefore represent minimum values.
148
LEIF HERTZ AND ARNE SCHOUSBOE
The direct measurements of K+ activities have confirmed previous conclusions by the Kuffler group (Kuffler and Nicholls, 1966; Orkand e t al., 1966) using the selective potassium permeability of the leech packet glia cell to determine changes in extracellular K+ concentration on the assumption of an unaltered intracellular K+ concentration (cf. also Ransom and Goldring 1973a,b). AND PHYLOCENETIC DEVELOPMENT E. ONTOGENETIC
Another approach to the study of differences between nerve cells and glia cells is the use of immature animals or of phylogenetically less developed species. Even between relatively closely related mammals, great temporal differences occur in brain development as reflected by the different time periods for maximum DNA synthesis. I n the rat brain the rate of DNA synthesis is high during the first 2-3 weeks after birth (Mandel et al., 1964; Dobbing, 1968; Fish and Winick, 1968; Mori et al., 1970; Schousboe, 1972), whereas cell proliferation ceases earlier in the guinea pig (Mandel et al., 1964). The late synthesis of DNA seems predominantly to represent a proliferation of glia cells, which in the rat brain cortex mainly or exclusively are formed postnatally (Altman, 1969). Histochemical studies have shown that the enzymic activities of the astrocytes remain low longer than those of the oligodendrocytes (Franck e t al., 1970). The number of neurons per volume brain cortex decreases during postnatal ontogenesis (Brizzee and Jacobs, 1959a,b; Brizzee et al., 1964; Vernadakis and Woodbury, 1965; Tower and Bourke, 1966; Franck, 1970), but the development of the dendrites occurs at an ontogenetically relatively late stage (e.g., SchadC and Baxter, 1960; cf. also Altman, 1967). I t is thus difficult with certainty to correlate the onset of metabolic events with maturation of either dendrites or glia cells although there seems to be a rather abrupt decline in DNA synthesis, i.e., glia cell proliferation about postnatal day 15 in the rat (Fig. 5 ) whereas the development of neuronal processes (Fig. 6 ) , may occur more gradually during the whole first month (Eayrs and Goodhead, 1959; Aghajanian and Bloom, 1967). As a further complication, differences may, however, occur between the rate of development in the same species kept in different laboratories (cf. Franck, 1970; Schousboe, 1972). Also during phylogenesis great alterations occur in the distribution between nerve cells and glia cells. The relative amount of the neuropil seems thus to increase with an increasing brain size leading to a decreased neuronal density (e.g., Nissl, 1898; von Economo, 1926; Tower and Elliott, 1952; Friede, 1954; Tower, 1954; Hawkins and Olszewski, 1957; Haug, 1960; Friede and Van Hauten, 1962; Altman, 1967; cf. however, also Nurnberger
BRAIN METABOLLSM AT T H E CELLULAR LEVEL
-
5000
-
4
Q
P
z
0
ul
2E 4000-
a u
149
0
p"
4
z
--
3000-
0
h
..-> .-0
2000-o
.0 u a ul
1000-
I ) 0-
II
16 24 DAYS AFTER BIRTH
FIG. 5. Rate of DNA synthesis measured as in uivo incorporation of [2--"C]thymidine into DNA (cpm/mg DNA) of the rat brain cortex 12 hours after intraperitoneal injection of ['4C]thymidine to rats, between 1 day of age and maturity. The brain region used was the superficial 0.5 mm of the outer convexity of each hemisphere, i.e., corresponding to the first brain-cortex slice cut from the lateral convexity. From Schousboe (1972).
and Gordon, 1957). The interesting hypothesis has been forwarded by Roitbak (1970) that it should be possible to form conditioned reflexes only in species with a relatively high degree of differentiation of glia cells. Phylogenetically less developed species often contain special forms of glia cells (Johnston and Roots, 19;'2), e.g., the leech packet glia cell. This cell is well suited for neurophysiological investigation on account of its large size, and it has been extensively used in several important studies on glial function by the Kuffler group (see, e.g., Kuffler and Nicholls, 1966; Kuffler, 1967). T h e packet glia cell differs, however, morphologically very much from the astrocytes and oligodendrocytes found in mammalian brain. In contrast to mammalian glia cells, it shows no potassium-induced increase in oxygen uptake (L. Hertz and C. Nissen, unpublished experiments). Since the leech central nervous system in addition contains a centrally located
150
LEI7 HERTZ AND ARNE SCHOUSBOE
11 days
I
30 days
100 Krn
FIG. 6. Characteristic changes in appearance of pyramidal cells in the rat brain cortex (lamina ganglionaris) during ontogenesis. From Eayrs and Goodhead ( 1959).
neuropil (Coggeshall and Fawcett, 1964) in which all the synaptic contacts occur (Johnston and Roots, 1972), great caution should probably be taken in attributing a too general validity to results obtained with this and other invertebrate preparations.
F. COMPART ME NTATION The presence of metabolic compartmentation in brain tissue in viuo or in vitro (Berl et al., 1961, 1968; Balazs and Cremer, 1973) is a further indication of cellular or subcellular heterogeneity. A great advantage of compartmentation studies is their use of histologically intact preparations but, on the other hand, compartmentation studies alone give no information about the cellular or subcellular localization. This is also the case with experiments involving graphical resolution of curves describing ion fluxes (e.g., Franck,
BRAIN METABOLISM AT T H E CELLULAR LEVEL
151
1970, 1973) which have been used to distinguish between various cellular compartments characterized by different membrane permeability for ions.
111. Energy Metabolism
A. OXYGEN CONSUMPTION
1. Comparison between Respiration in V i v o and in Vi t r o The mean oxygen consumption in normal human brain is 3.6 ml/minute per 100 gm (Kety, 1957), which equals 95 Fmoles/hour per gram wet weight (Table I ) . A decrease of the metabolism to about 60% is found in coma or under anesthesia (Kety, 1957), and an increase occurs during convulsions (Schmidt et al., 1945). Approximately identical results have been obtained in the isolated and perfused cat brain (Table I ) by Geiger and Magnes ( 1947). Species variations occur and in the rat the oxygen consumption is about 5-6 ml/minute per 100 gm (Bergen et al., 1953). Regional differences in metabolic rate have been demonstrated (Kennedy et al., 1974; Sokoloff et al., 1974). The least damaged in vitro preparation probably is the brain slice. For comparison between oxygen consumption in this preparation and under in uivo conditions, attention should be paid to both regional and species differences. Several authors agree that the rate of oxygen consumption also in brain slices decreases with an incrrasing size of the animal (Quastel and Wheatley, 1932; Elliott and Henderson, 1948; Krebs, 1950; Elliott, 1952, 1957; Davies, 1961; Tower and Young, 1973; cf. however, also Locker and Kaps; 1960). I n the same species, slices from cerebral and cerebellar cortex respire at approximately equal intensity. I n basal ganglia the respiratory rate is the same or somewhat lower (Dixon and Meyer, 1936; Bollard and McIlwain. 1957; Hertz and Clausen, 1963; Ridge, 1967; Weiss et al., 1972). The rate of oxygen uptake is still lower in the brain stem, spinal cord, and central white matter (Bollard and McIlwain, 1957; Hertz and Clausen, 1963; Ridge, 1967). Peripheral nerve respires at an intensity of only about 10% of that in the cortex (Holmes, 1932; Chang et al., 1935; Chang and Tai, 1936; Hertz and Clausen, 1963; Rang and Ritchie, 1968), whereas peripheral ganglia show a relatively high rate of oxygen uptake (Larrabee, 1958; Nagata, 1969; cf., however, also Holmes, 1932 ; Dixon and Meyer, 1936). The great topographical variation introduces difficulties in comparing respiration by a cortical brain dice with that of the total brain, but, roughly, the rate of oxygen consumption in the total rat brain (per unit weight)
152
L E I F H E R T Z AND ARNE S C H O U S R O E
CORRECTED RATESOF
OXYGEN
TABLE I UPTAKE BY DIFFERENT BRAINPREPARATIONS"**
Preparation
In uiuo, normal conditions In uiuo, coma In vivo, convulsions Perfused brain, normal conditions Perfused brain, coma Perfused brain, convulsions Brain slices, resting conditions Brain slices, electrical stimulation Brain slices, excess potassium Homogenates, resting conditions Homogenates, excess potassium Isolated neurons, resting conditions Isolated neurons, excess potassium Isolated glia cells, resting conditions Isolated glia cells, excess potassium Isolated neurons plus glia cells in a ratio similar to that in brain cortex As above, excess potassium
Animal species Man (1)" Man (1) Monkey (2) Cat (3) Cat (3) Cat (3) Man (4) Man (4) R a t (5) R a t (6) R a t (6) Cat (7) Cat (7) R a t (7) R a t (7) Rat Rat
+ cat (7)
+ cat (7)
Rate of oxygen uptake Gmoleslhour per gram wet weight 95 50 140 90 50 180 45 90 65 40 35 150 140 25 50
30 55
From Hertz (1973a). T h e rates of oxygen uptake by the different structures were measured in different species (indicated in the table), and have been recalculated to the supposed values in human tissue. This was done by multiplying the observed rates of oxygen uptake in rat, cat, and monkey tissue by, respectively, 0.55, 0.70, and 0.80 (cf. Elliott and Henderson, 1948). The average respiratory intensity in the rat brain corresponds to about 80% of that in a rat brain-cortex slice. O n the assumption that this correlation also is valid in other species and after homogenization or microdissection, all results obtained with slices, homogenates, or isolated cells were furthermore multiplied by 0.80 to give the corresponding respiratory rates by the whole brain. As a result of the corrections only approximated values are presented. No attempt has been made to indicate the scatter of the figures given in the literature, but the values shown are supposed to be representative. I n general the deviations from these values are small. Results originate from: (1) Kety, 1957; (2) Schmidt ~t al., 1945; (3) Geiger and Magnes, 1947; (4)McIlwain, 1953c; (5) Hertz and Schou, 1962; (6) Elliott and Libet, 1942 (average values from their table 2); (7) Hertz, 1966. a
has been estimated as corresponding to about 80% of that in cortical brain slices (Elliott, 1952). The rate of oxygen uptake of 40-50 pmoles/hour per gram wet weight in human brain-cortex slices during incubation in a physiological medium (Elliott and Henderson, 1948; McIlwain, 1953a), may thus correspond to a whole-brain respiration of 32-40 pmoles/hour per gram wet
BRAIN METAROI.ISM
AT T H I CELI.ULAR
LEYEI,
153
weight, i.e., less than half of the respiratory intensity recorded in uivo (Table I ) . Exposure to, for example, high concentrations of potassium (Ashford and Dixon, 1935; Dickens and Greville, 1935) or electrical pulses (McIlwain, 1951; McIlwain and Joanny, 1963) may, however, lead to an increase of almost 100% in the respiratory rate by brain-cortex slices or the locally perfused cerebral cortex in v i m (Grenell, 1959). The magnitude of the stimulated respiration is accordingly much more similar to that of the functioning brain (cf. Quastel and Quastel, 1961; McIlwain, 1966). The potassium-induced stimulation of brain cortex respiration requires the presence of a cwtain concentration of sodium (Dickens and Greville, 1935; Canzanelli et al., 1942; Gore and McIlwain, 1952; Tsukada and Takagaki, 1955; Bachelard et al., 1962; Hertz and Schou, 1962; Kozawa and Naito, 1966; Nakazawa and Quastel, 1968), but is competitively inhibited by excess sodium (Hertz and Schou, 1962). At least 20 mM K+ is needed and a maximum effect is obtained with about 50 mM K+ (Hertz and Schou, 1962; Huttunen, 1969). The increase in oxygen uptake occurs very rapidly after the increase in the potassium concentration (Fig. 7 ) , i.e., 01
E E
760
C
.-YI
u 0 "
380
0
-
I
I
I
I
I
1
I
2
3
4
5
6
I
7 8 Time, min
FIG. 7. Typical record obtained with an oxygen electrode in a 3.0 ml chamber and showing the rate of oxygen uptake (slope of the curve) for rat brain-cortex slices in a bicarbonate-buffered, initially physiological medium. Addition of excess potassium (arrow) is seen to evoke an immediate increase in the respiratory rate. Unpublished experiment by G. B. Weiss, F. Goodman, and L. Hertz.
154
LEIF HERTZ AND ARNE SCHOUSBOE
before any pronounced changes have occurred in the tissue concentrations of monovalent cations (cf. Lund-Andersen and Hertz, 1970). The “stimulated” respiration shows, however, a relatively rapid decline (Hertz and Schou, 1962; Mase et al., 1962; Machiyama et al., 1970). The stimulating group of ions also comprises ammonium (Weil-Malherbe, 1938; Gore and McIlwain, 1952), cesium, rubidium (Dickens and Greville, 1935), lithium (Canzanelli et al., 1942; Rybova, 1959), and choline (Hertz and Schou, 1962). Also exposure to a calcium-deficient medium may lead to an increase in oxygen consumption (Dickens and Greville, 1935; Kratzing, 1953; Quastel and Quastel, 1961; Bourke and Tower, 196613). Lack of sodium (replacement with sucrose) leads, in contrast, to a respiratory decrease (Gore and McIlwain, 1952; Hertz and Schou, 1962; Hertz and Clausen, 1963; Hertz, 1966). Only tissue from structures that show a relatively high rate of oxygen uptake during incubation in physiological media (cerebrum, cerebellum, basal ganglia, and brain stem) seems to react metabolically to excess potassium (Hertz and Clausen, 1963; Ridge, 1967). No stimulation is observed with mammalian peripheral nerve (Chang et al., 1935; Pietra, 1961; Tamarit and Gallego, 1962; Hertz and Clausen, 1963; cf., however, also Shanes and Hopkins, 1948; Rang and Richie, 1968) or with white matter from the central parts of the brain (Hertz and Clausen, 1963; Ridge, 1967). A small potassium-induced increase in rate of oxygen uptake is observed in the spinal cord of frog or chicken (Chang and Tai, 1936; H. Fosmark and L. Hertz, unpublished experiments) whereas no such stimulation is found in calf spinal cord (Hertz and Clausen, 1963). The potassium-induced stimulation of oxygen uptake is absent in brain cortex slices from rats younger than 2 weeks (Himwich et al., 1942) and also the responsiveness to electrical stimulation is decreased at this age (Greengard and McIlwain, 1955; McIlwain, 1956). High concentrations of potassium or electrical pulses have no stimulatory effect on oxygen uptake by brain homogenates (Elliott and Libet, 1942; Ghosh and Quastel, 1954; Narayanaswami and McIlwain, 1954; Jones and Banks, 1970a’l. I t may therefore seem peculiar that excess potassium increases the respiratory rate of isolated brain mitochondria (Sugawara and Utida, 1961; Utida and Sugawara, 1963; Krall et al., 1964; Ozawa et al., 1967). The possibility of a contamination of brain mitochondrial preparations was pointed out by Opit and Charnock (1965), but Clark and co-workers (Clark, 1970; Clark and Nicklas, 1970; Nicklas et al., 1971) confirmed that 100 mM potassium causes a distinct increase in the rate of oxygen uptake by a “brain mitochondrial preparation known to be relatively free from synaptosomal and cytoplasmic contamination” ; in contrast to the potassium-induced respiratory stimulation in brain slices no presence of sodium was required, the
BRAIN METABOLISM A T T H E CELLULAR LEVEL
155
response was not abolished by uncoupling, and the content of ATP was unaffected (cf. also Katzman and Pappius, 1973).
2. Respiration by Neurons The oxygen uptake by large neurons (Table 11) obtained by microdissection from normal rat or rabbit brain amounts to 50 to 100 x pl/hour per cell (Epstein and O'Connor, 1965; HydCn and Lange, 1965; Hertz, 1966). This value is in relatively good agreement with a mean nerve cell respiration of about 2 to 14 >( 10-j pl/hour per cell calculated on the basis of respiratory rates in fractions from white and gray matter (Heller and Elliott, 1955; Elliott and Heller, 1957; Korey and Orchen, 1959') and with a respiratory rate of 0.7 x $/hour per cell in small cultivated cells (Dittmann et al., 1 9 7 3 ~ )I.t also fits with an average rate of oxygen uptake per rat spinal ganglion neuron of 4.5 x lo-' pl/hour (Tobias et al., 1942). Using neuroblastoma cells it lhas been shown that the respiration is dependent upon differentiation, being 0.15 x and 0.49 X pl/hour per cell in, respectively, proliferating and differentiating cells (Nissen et al., 1972; Mandel et al., 1973). A similar effect of maturation has been observed in neurons from young chick embryos (Dittmann et al., 1973b). Expressed per unit volume (or weight) some discrepancies are found. I n bulk-prepared neurons obtained by differential centrifugation, the rate of oxygen uptake has been found to be 80-200 pmoles/hour per gram wet weight (Table 11).The respiration of microdissected and of cultivated neurons is higher, i.e., between 260 (Hertz, 1966) and 1080 (Epstein and OConnor, 1965) pmoles/hour per gram wet weight. No systematic differences were found between neurons from the brain and from dorsal root ganglia or between microdissected and cultivated cells (Table 11). Based upon the work of Lowry et al. (1954), it is often quoted in the literature that brain metabolism mainly occurs in the dendrites. These authors investigated microsamples from different layers, but no attempt was made to distinguish specifically between neuronal and glial components (0.H. Lowry, personal communication) . Nerve-ending particles prepared by differential and density gradient centrifugation respire at a rate of 50-60 pmoles/hour per 100 mg of protein (Bradford, 1967, 1969; Whittaker, 1969; De Belleroche and Bradford, 1972), which probably corresponds to about 50-60 prnoles/hour per gram wet weight.
3. Respiration by Glia Cells The rate of oxygen uptake per glia cell (Table 11) has been both meapl/hour per cell (Heller and sured and calculated to 0.5 to 1.0 X Elliott, 1955; Korey and Orlzhen, 1959; HydCn and Lange, 1965; Hertz,
' Cf. also RuSEBk
and RuSEikovd, 1971.
TABLE I1 RATESOF OXYGEN UPTAKEFOR NEURONS AND GLIA CELLS",~ uptake per cell (p1 X hour-') 101 x 10-5 100 x 10-6 50 x 10-5 3.0 X 10-5 0 2
0 2 uptake per gram wet weight (pmoles X hour-') Authors 1080 Epstein and O'Connor (1965) H y d h and Lange (1965) 263 Hertz (1966) 79 Rose ( i 9 m Haljamae and Hamberger (1971) 204 Hemminki and Holmila (1971) 95 88 Hemminki and Holmila (1971) 1000 Hartman et al. (1970) Booher et al. (1971b) 567 Dittmann et 01. (1973~) Heller and Elliott (1955) Elliott and Heller (1957) Korey and Orchen (1959) Epstein and O'Connor (1965) H y d h and Lange (1965) 40-80 Hertz (1966) 76 Rose (1967) 187 Haljamae and Hamberger (1971) 71 Hemminki and Holmila (1971) 40 Hemminki and Holmila (1971) Heller and Elliott (1955) Korey and Orchen (1959) 51 Tobias et al. (1942) 130 Dittmann et al. (1973~) 135 Abood et al. 11951)
Cell Origin Species Method Neuron Cortex Cat Microdissection Neuron Brain stem Rabbit Microdissection Cortex Neuron Cat Microdissection Cortex Neuron Rat Gradient centrifugation Rabbit Cortex Neuron Gradient centrifugation Rat Cortex Neuron Gradient centrifugation 0.15 x 10-5 Cortex Neuron Gradient centrifugation Rat 0.11 x 10-5 Spinal ganglion Chick Neuron Microdissection 4 x 10-5 Spinal ganglion Chick Neuron Cultivation 4.3 x 10-5 Chick Cortex Neuron Cultivation 0.66 x 10-5 Cortex Neuron Cat, dog Calculation 6-9 x 10-5 Cortex Neuron Cat, dog Calculation 2-3 x 10-5 Neuron Cortex Calculation Lamb 1 4 x 10-5 Glia cell Cortex Microdissection Rat 0 Glia cell Rabbit Brain stem Microdissection 0-10 x 10-5 Glia cell Cortex Microdissection Rat 0.5-1 x 10-5 Glia cell Cortex Gradient centrifugation Rat Glia cell Cortex Gradient centrifugation Rabbit Glia cell Cortex Gradient centrifugation Rat 0.11 x 10-5 Glia cell Cortex Gradient centrifugation Rat 0.04 x 10-5 Glia cell C. callosum Cat, dog Slices 0.5-0.6X Glia cell C. callosum Homogenate Lamb 1 . 2 x 10-5 Glia cell Filum terminale Cat Astrocyte Cortex Cultivation Rat 0.75 x 10-5 Microalia SDinal cord Chick Cultivation Modified from Dittmann et a/. (1973~). Rates of oxygen uptake (per cell and per volume) given in the literature for presumably normal, differentiated neurons and glia cells. The values were obtained by microdissection, cultivation, gradient centrifugation, or calculation (based upon cell counts and respiratory rates in slices or homogenates of corpus callosum). The oxygen uptake rates per cell obtained by gradient centrifugation have been recalculated from the dry weight or the content of protein per cell given in the original papers, and the corresponding respiratory rates per volume have been calculated on the assumption that solids constitute 20% and protein 10% of the cell volume. Based upon the volume data presented in some of the papers the neurons may be classified as large (microdissected samples), medium sized (spinal ganglion), or small (Dittmann et al., 1973~).The calculated results and probably also the values for cells obtained by centrifugation (cf. Rose, 1967) refer to cells of average size. ~
~
~~~
c
cn m
r
8 ! I ib
1 N
t
3
9 7J
3 m 0
2 W
BRAIN METABOLISM A T THE CELLULAR LEVEL
157
1966; RulEAk and RukEAkovA, 1971; Dittmann et al., 1 9 7 3 ~ ) The . microdissected glial samples obtained by Epstein and OConnor (1965) showed practically no oxygen uptake at all, but a long period of preincubation ( 2 hours) had been employed, and the respiration by isolated glia clumps declines rapidly (Hertz, 1966). Cells obtained by the different gradient centrifugation techniques show a variable and often low rate of oxygen uptake (Table 11; cf. also Rose and Sinha, 1969). Relatively good agreement is, however, found that the respiration in normal glia cells expressed per unit volume amounts to about 50-200 pmoles/hour per gram wet weight (Table 11). Cells extracted from the NN cell line (Shein et al., 1970) have an oxygen uptake rate of 0.3 x $/hour per cell (C. Nissen, J. Ciesielski-Treska, D. Beya, L. Hertz, and P. Mandel, unpublished experiments), and 138 MG cells (PontCn and MacIntyre, 1968) respire at about 0.6 X pl/hour per cell (Walum et al., 1974). The latter value was obtained with cells extracted from older, presumably more differentiated cultures, whereas cells from younger cultures had a much lower respiratory intensity, probably reflecting the low rate of oxygen uptake observed in astrocytoma (Victor and Wolf, 1937; Heller and Elliott, 1955; Brierley and McIlwain, 1956; Allen, 1972). Also reactive cortical glia cells (RulEAk et al., 1967) have a low oxygen consumption. Oligodendroglioma cells may, however, respire at a considerably higher rate (Victor and Wolf, 1937; Heller and Elliott, 1955; Allen, 1957; cf., however, also Rrirrley and McIlwain, 1956; Weber, 1959; Mahaley, 1966) .
4. Response to Stimulation in Individual Cell T y p e s The results obtained with slices allow no conclusions whether it is the glia cells or the neurons (or both) which react metabolically to an increased potassium concentration. All available information indicates that glia cells react to a rise of the K+ concentration with a considerable increase in rate of oxygen uptake (Fig. 8 ) . This has been shown micromanometrically (Hertz, 1966; Aleksidze and Blomstrand, 1969) with preparations microdissected as described by HydCn (1959) or with cultivated astrocytes (Hertz et al., 1973). The response is short-lasting and the stimulated respiration declines rapidly. Also glia cells prepared in bulk by the gradient centrifugation method of Blomstrand and Hamberger (1969, 1970) showed a large (85%) K+ induced stimulation (Haljamae and Hamberger, 1971) whereas the preparation used by Bradford and Rose (1967) gave a more moderate response (25% ) . A glial localization is in accordance with the ontogenetically late appearance of the potassium-induced stimulation. The question of a potassium-induced increase of neuronal respiration is more ambiguous. Little or no stimulation was observed in the experiments
158
LEIF HERTZ AND ARNE SCHOUSBOE
70
30
10
50
70
60
Durotion of incubation ( m i n
1
FIG. 8. Cumulative rates of oxygen uptake by neurons (0) and glia cells ( 0 )from Deiters’ nucleus in the rabbit compared to the apparent respiration in control divers ( A ) . Excess potassium was added at arrows without opening the divers (Hamberger, 1968). The respiration is shown in arbitrary units (change in gas volume of flotation chamber (mm burette) to maintain buoyancy of diver). From Aleksidze and Blomstrand (1969).
by Hertz ( 1966), Aleksidze and Blomstrand (1969), Haljamae and Hamberger (1971), and Hertz et al. (1973) using, respectively, microdissected, bulk prepared and cultivated neurons (Fig. 8 ) . The oxygen consumption by cultivated neuroblastoma cells is also not increased by excess potassium (Nissen et al., 1973). Bradford and Rose (1967) observed, however, approximately the same reaction to addition of potassium in their bulk-prepared neurons and glia cells. Using a microspectrophotometric technique, Hultborn and HydCn (1974) observed a K+-induced stimulation of oxygen uptake in microdissected neurons. The rate of oxygen uptake by synaptosomes is increased by electrical stimulation (De Belleroche and Bradford, 1972) or excess potassium (Whittaker, 1969).
5 . Relatiue Contribution t o Total Respiration by Individual Cell T y p e s The observation that the rates of oxygen uptake per volume of isolated cells or synaptosomes are of the same magnitude (glia cells, synaptosomes) or even higher (neuronal perikarya) than that in brain slices indicates metabolic integrity of these preparations. The values give, however, no direct answer to the pertinent question how much each cell type contributes to the total respiration in brain cortex. Figure 9 represents a n attempt to perform such an evaluation. The upper part (1) indicates the relative fractions of the total volume occupied by, respectively interstitial fluid, neuronal
BRAIN METABOLISM AT T H E CELLULAR LEVEL
1
3
15 151 15
25
I
191
32.5 (?)
I
33
I
32.5
159
(?I
33
perikarya, synaptosomes, and the remaining part (65% ) which tentatively has been equally divided between glia cells and neuronal processes (cf. Section 11, A and B) . The middle part ( 2 ) shows the relative respiratory intensities in each of these components (cf. above) with the respiration of a rat brain cortex slice (approximately 100 pmoles/hour per gram wet weight) used as the reference (i.e., equal to 1 ) . From these two sets of values the results shown in the lower part ( 3 ) were obtained showing that about onefourth of brain cortex respiration occurs in neuronal perikarya, one-tenth in synaptosomes, and one-third in glia cells. The remaining one-third thus probably occurs in neuronal processes. Since the latter elements account for a comparable part of the volume, this estimate is against the concept that dendritic respiration should be specially high. The approximations involved in these estimates should be emphasized. The relative volumes occupied by, respectively, neuronal processes and glia cells are not known with certainty, and if the influence of potassium on metabolism predominantly is exerted on glia cells, the relative contribution of these cells to the total oxygen consumption would increase. Previous calculations showing that about onefourth of the total oxygen uptake occurs in glia cells and three-fourths in the neurons (Heller and Elliott, 1955; Elliott and Heller, 1957; Korey and Orchen, 1959; Tower, 1960) are in reasonably good accordance with the present conclusion. Based upon sensitivity to sodium deficiency, a larger glial contribution has been suggested by Hertz (1966), who in analogy with the present estimate found that 20% of the oxygen uptake occurs in neuronal perikarya. A large contribution of glia cells to the total oxidative metabolism is also indicated by the finding of Howe and Mellors (1945) that two-thirds of the cytochrome oxidase activity in thalamus remains intact after degeneration of the neurons by decortication. On the other hand, it has also been suggested (Hess, 1961) that only 5% of the total oxygen uptake in the brain
160
LEIF H E R T Z A N D A R N E SCHOUSROE
cortex should occur in glia cells, but this conclusion was based upon rates of oxygen uptake in undifferentiated glial tumors.
B. HIGH-ENERGY PHOSPHATES
1. Comparison between High-Energy Phosphate Leuels in V i u o and i n Vitro Measurement of high-energy phosphates in viuo is complicated by their extremely fast turnover, but evidence is accumulating that the true i n uiuo concentrations of ATP and creatine phosphate are 2.5-3.0 and 4.0-5.0 pmoles/gram wet weight, respectively (FolbergrovA et al., 1969; Granholm et al., 1969; Ponttn et al., 1973; Veech et al., 1973). Hypoxia leads to a reduction of the creatine phosphate level whereas ATP remains constant for a considerable period (Ferrendelli et al., 1973; Bachelard et al., 1974). The concomitant loss of electrical brain activity and of excitability occurs before the level of the high-energy phosphates has fallen drastically (Gurdjian et al., 1947; Albaum et al., 1953; Richter, 1955; Ferrendelli, 1975). The concentrations of both creatine phosphate and ATP are low in freshly cut slices. However, some one-half to three-fourths of the in viuo level may rapidly (i.e., within 10 minutes) be restored during incubation (e.g., McIlwain et al., 1951; McIlwain, 1952; McIlwain and Gore, 1953; Swanson, 1968; Jones and Banks, 1970b; Okamoto and Quastel, 1970; Banay-Schwartz et al., 1971; Piccoli et al., 1971). Exposure to excess potassium leads almost immediately to a considerable decrease in the levels of creatine phosphate and ATP in brain slices (Gore and McIlwain, 1952; McIlwain, 1952 ; Brossard and Quastel, 1963; Biesold, 1967; Rolleston and Newsholme, 1967), and the threshold concentration of potassium is about 25 mM (Huttunen, 1969; Takagaki, 1972). Simultaneously the rate of incorporation of 3?P into acid-soluble phosphate is increased (Brossard and Quastel, 1963; cf., however, also Fonyo et al., 1958). A minimum concentration of sodium is required for the response, and in potassium-rich media the level of energy-rich phosphates may be higher in the absence of sodium than in the presence of this ion (Gonda and Quastel, 1966) . Application of electrical pulses causes a lowering of the level of acid-soluble phosphate, but the time course of the decrease differs slightly from that observed after exposure to high concentrations of potassium, and the decrease in the ATP level may be only transient (Heald, 1954; cf., however, also Jones and Banks, 1970a). I n the absence of sodium no decrease is evoked (Kato, 1970).
BRAIN METABOLISM AT T H E CELLULAR LEVEL
161
Initially the rate of breakdown of creatine phosphate becomes as high as 1200-1400 pmoleslhour per gram during electrical stimulation (Heald, 1960; McIlwain, 1966). As pointed out by Heald, this extremely high rate suggests that the breakdown is not the result only of an inhibition of oxidative phosphorylation. After cessation of the pulses the previous level of creatine phosphate is rapidly (i.e., within 90 seconds) reobtained (Heald, 1954). This is probably one of the processes mirrored by the stimulation of oxygen consumption which also continues beyond the actual period of stimulation (McIlwain, 1954) .
2. Na+-K+-ATPase I t has repeatedly been suggested that the potassium-induced stimulation of energy metabolism in brain might be a metabolic manifestation of an increased breakdown of ATP by Na+-K+-ATPases (e.g., Hess, 1962; Minakami et al., 1963; Henn et al., 1972) or related enzymes (Hertz and Schou, 1962). The activity of this enzyme is low in the absence of sodium and potassium. Addition of sodium alone, but not of potassium alone, causes a slight stimulation, whereas the simultaneous presence of both ions increases the activity considerably (Skou, 1957, 1960, 1962; Post et al., 1960). In erythrocyte ghosts it has been shown that the enzyme is activated by potassium externally and by sodium internally (Glynn, 1962; Whittam, 196213; Garrahan and Glynn, 1967). To exert maximum stimulation of the brain enzyme, 10-20 m M potassium (Fig. 10) are required (Samson and Quinn, 1967; Henn et al., 1972). Potassium may be replaced by several other ions (e.g., NH,', Cs+, Rb+, Li') , but the sodium effect on the enzyme is rather specific, since only lithium is capable of causing some stimulation in the total absence of sodium (Skou, 1957; Bader and Sen, 1966; Whittam, 1962a). Excessively high concentrations of either sodium (Post et al., 1960; Whittam and Ager, 1962; Priestland and Whittam, 1968) or potassium (Skou, 1957, 1962; Ahrned et al., 1966) cause an inhibition of the enzyme. These characteristics remind strikingly of the potassium-induced stimulation of oxygen consumption which also requires the presence of sodium but is competitively inhibited by high concentrations of sodium (Hertz and Schou, 1962; Fig. 3 in Elliott and Bilodeau, 1962), and in which only lithium is able to replace sodium (Hertz and Schou, 1962). The Na+-K+-ATPase activity in brain is much lower in newborn rats than in adult animals and the increase in the activity seems to occur especially fast (Fig. 11) during the first 2-3 weeks after birth (Abdel-Latif et al., 1967; Samson and Quinn, 1967; Medzihradsky et al., 1972; Nagata et al., 1974), i.e., possibly also somewhat before the onset of the potassiuminduced increase in oxygen consumption and swelling. I n the chicken a cor-
162
LEIF H E R T Z A N D ARNE SCHOUSBOE
10
0
0
10
20 30 mM K'
40
50
FIG. 10. Specific activity of neuronal and glial Na+-K+-ATPase as a function of the external potassium concentration a t unaltered sodium concentration. T h e
activities (in micromoles of phosphate liberated per hour per milligram) are expressed as the increments of activity due to addition of sodium and potassium. Both the glial ( A ) and neuronal ( m ) enzymes were incubated at 37°C for 30 minutes in media containing 75 mM Na+, 5 m M Mg", and 50 mM Tris.HC1, p H 7.4. SD values are given as vertical bars. From Henn et al. (1972).
responding increase may be observed before hatching (Bignami et al., 1966; Zaheer et al., 1968). The activity of the Na+-K+-activated ATPase is low in DBA mice at the time period when they are susceptible to seizures (Hertz et al., 1974).
7
F
6
x
5
2
2
E = 1
0
20
30
LO
50
Age of r a t (days1
FIG. 11. Specific activity of the Na+-K+-ATPase in rat brain homogenates exposed to 100 m M Na' and 20 m M K' in the absence ( A ) and in the presence (A) of 5 x lo-* M ouabain as a function of the age of the animals. Each point is the average of 20 experiments, and values are expressed in terms of wet weight of original brain. From Samson and Quinn ( 1967).
BRAIN METABOLISM A T T H E CELLULAR LEVEL
163
The Nat-K+-ATPase is inhibited by ouabain (e.g., Post et al., 1960; Skou, 1960, 1962; Schwartz et al., 1962; Gibbs et al., 1965; Swanson and McIlwain, 1965; Israel and Titus, 1967; Henn et al., 1972). Intracranial injection of this drug leads to motor hyperactivity, convulsions, and finally death (Bignami and Palladini, 1966; Cornog e t al., 1967).
3. High-Energy Phosphates in Individual Cell T y p e s The concentration of ATP in neurons obtained by microdissection or cultivation (Table 111) equals the average ATP level in vivo and are thus, if anything higher than the concentrations in brain cortex slices. The pronounced difference in content per cell reflects the variation in cell volume (cf. Table 11). Approximately identical concentrations of both ATP and creatine phosphate are found in cell bodies and neuropil (Passonneau and Lowry, 1971). Neurons obtained by gradient centrifugation seem to have a somewhat lower content of ATP, and also isolated synaptosomes have a low concentration of ATP and creatine phosphate (Bradford, 1969). Cultivated glia cells have the same ATP concentration as neurons, whereas the level seems to be lower in microdissected samples (Table 111). Again the concentration is relatively low in preparations obtained by gradient centrifugation. The glia cells are not essential for the synthesis of ATP in neurons, since isolated nerve cells without surrounding glia cells are capable of incorporating inorganic phosphate into high-energy phosphates (Hillman and HydCn, 1965b). The potassium-induced reduction of ATP content observed in brain slices seems to occur predominantly in glia cells (astrocytes), since addition of 50 mM potassium to cultivated astrocytes leads to a considerable decrease in the content of ATP but has no effect on cultivated neurons (Schousboe et al., 1970).
4. Na+-K’-ATPase Activity in Individual Cell T y p e s Studies on neurons or glia cells obtained by microdissection (Cummins and HydCn, 1962) or gradient centrifugation (Medzihradsky et al., 1971, 1972, 1974; Henn et al., 1972; Nagata et al., 1974) have shown that the Na+-K'-ATPase activity (expressed in relation to volume or protein content) is higher in glia cells than in neuronal perikarya. The enzyme activity reaches its maximum already at about 10 days of age (Medzihradsky et al., 1972), and maximum stimulation is obtained with 10 m M K+ (Fig. 1 0 ) . Corresponding investigations of microsamples from different layers of the cortex were, however, interpreted to show a localization in the plexi of dendrites and axons (Hess et al., 1963; Lewin and Hess, 1964). Also nerve ending particles possess a potent Na+-K'ATPase (Hosie, 1965 ; Kurokawa et al., 1965; Bradford et al., 1966; Whittaker, 1969) with maximum activity
TABLE I11 ATP CONTENT IN NEURONS AND GLIA CELLS' ATP
Cell
Origin
Species
Method
pmoleslgm pmoles/cell wet weight
Authors
r
E! -¶
Neuron Neuron Neuron Neuron Undiff. brain cells Glia cell Astrocyte Astrocyte Glia cell
Brain stem Dorsal root ganglion Brain hemispheres Brain cortex Brain hemispheres
Rabbit Chick
Microdissection Cultivation followed by microdissection
0.24 0.008
3 2
Cummins and Hydtn (1962) Schousboe et al. (1970)
Chick Rat Chick
Cultivation followed by microdissection Gradient centrifugation Dissection
0.006
-
Dittmann et al. (1973b) Rose and Sinha (1969) Dittmann et al. (197313)
Brain stem Spinal cord Brain hemispheres
Rabbit Chick Rat
Microdissection Cultivation followed by microdissection Cultivation
Brain cortex
Rat
Gradient centrifugation
-
0.008
-
1.2 2.8
-
<<3.0 3 2
-
0.7
0.012
Cummins and Hydtn (1962) Schousboe et al. (1970) A. Schousboe and L. Hertz, unpublished results Rose and Sinha (1969)
a The content of ATP per gram for microdissected neurons was calculated on the basis of the content per cell and the cell volume given by Cummins and Hydtn (1962). The values from Rose and Sinha (1969) were recalculated as micromoles per gram wet weight on the assumption of a protein content of the samples of 10% of the wet weight, T h e exact value for the ATP content (per cell or gram) of the microdissected glia cells (Cummins and Hydtn, 1962) could not be calculated, but it has been given as much lower than the value for the corresponding neurons (3.0 pmoles/gram).
t
3
BRAIN METABOLISM AT THE CELLULAR LEVEL
165
a t about 10 m M K + (Appel et al., 1969) and ontogenetic maturation 10-20 days after birth (Abdel-Latif et al., 1967; Gaballah and Popoff, 1971). The Na+-K+-ATPase activity in astrocytoma is low (Hess et al., 1963, 1972). Among the established cell lines used for study of glial function both the N-nitrosomethylurea-induced rat astrocytoma cells, C6 (Benda et al., 1968) and the spontaneously transformed astrocytoma line NN, originally developed from the brains of newborn hamsters (Shein et al., 1970; Eichberg et al., 1971; Silberstein et al., 1972), have an extremely low Na+-K+ATPase activity (Embree et al., 1971a,b; Hess et al., 1972; cf., however, also Kimelberg, 1974). This may conceivably indicate a less complete metabolic differentiation. All the above-mentioned experiments to determine the localization of Na+-K+-ATPase activity have been performed employing more or less altered or disintegrated preparations. T o conclude that normal glia cells are charactcrized by a high Na+-K+-ATPase activity it is therefore of major importance that intracranial injection of ouabain (Fig. 12) leads to a pronounced swelling and vacuolization of astroglia (and of certain presynaptic terminals)
FIG. 12. Electron micrograph of brain cortex from a rat 2 hours after intracerebral injection of ouabain. Perivascular astrocyte ( A ) shows peripheral condensation of nuclear chromatin and swollen, clear cytoplasm containing abnormal membranous profiles (arrowheads). Neuron ( N ) is partly surrounded by swollen but unidentifiable cell processes, many of which are also present in neuropil (arrows). x4300. From Cornog et al. (1967).
166
L E I F H E R T Z AND A R N E SCHOUSBOE
whereas neuronal perikarya, axons, dendrites, and oligodendroglia remain largely intact (Bignami and Palladini, 1965, 1966; Cornog et al., 1967).
METABOLISM C. INTERMEDIARY 1. Pathways and Substrates There is good evidence that the normal intact brain predominantly utilizes glucose as its substrate and source of energy (e.g., Kety, 1957; Balazs, 1970) though fatty acids probably may be metabolized neonatally (cf. Klee and Sokoloff, 1967) and after prolonged fasting (Owen et al., 1967). This does not imply that glucose exclusively or predominantly is metabolized along the glycolytic pathway and subsequent tricarboxylic acid cycle, since both the pentose phosphate shunt and the GABA shunt (Roberts, 1956) might contribute considerably to cerebral glucose metabolism. The GABA shunt has been found metabolically active in vivo (Lajtha et al., 1959) and the pentose phosphate shunt has by one author (Moss, 1964) been reported to be a major pathway for cerebral glucose oxidation, but by other authors (e.g., Sacks, 1957) to be of practically no importance in v i m . Also during incubation of brain slices most of the energy production is derived from breakdown of glucose as reflected by a respiratory quotient ( R Q ) relatively close to 1.0 (e.g., Dickens and Simer, 1931 ; Elliott et al., 1937; Elliott and Penfield, 1948; Nishimura and Kimura, 1965). The function of the pentose phosphate shunt in mammalian and frog brain tissue and retina during in vitro experiments has been reported by several authors (e.g., Cohen and Noell, 1960; Hotta, 1962; Reading, 1964; De Piras and Zadunaisky, 1965; Nishimura and Kimura, 1965). Also the GABA shunt has been found active in brain slices, but different estimates of its quantitative importance have been reported. The reasons for these discrepancies were discussed by Hammond et al. (1970), who found that the GABA flux in vitro accounts for about 8% of the total flux through the tricarboxylic acid cycle (Balazs et al., 1970). The turnover via the pentose phosphate shunt is not increased during stimulation with electrical pulses (O’Neill et al., 1965) or high concentrations of potassium (De Piras and Zadunaisky, 1965; Kozawa, 1961) , whereas the tricarboxylic acid cycle (and aerobic glycolysis) operates at an increased rate under the influence of excess potassium (Machiyama et al., 1970; cf. also Kimura and Niwa, 1953). If the passage through the GABA shunt is increased in potassium-rich media (Machiyama et al., 1967), the absolute increase is only small and the relative importance of this pathway decreases (Machiyama et al., 1970). Glucose, fructose, pyruvate, lactate, and oxaloacetate all support identical, well-maintained rates of oxygen uptake during incubation in a physio-
BRAIN METABOLISM AT T H E CELLULAR LEVEL
167
logical medium (e.g., McIlwain, 1953a; Ghosh and Quastel, 1954; Ito, 1960), and a respiratory stimulation occurs after addition of excess potassium (e.g., Dickens and Greville, 1935; Dixon and Holmes, 1935; Lipsett and Crescitelli, 1950; Ghosh and Quastel, 1954; Ito, 1960). Relatively high concentrations of high-energy phosphates are found with either glucose, fructose, or pyruvate (McIlwain, 1952; Heald, 1960; Woodman and McIlwain, 1961; Abadom and Scholefield, 1962b; Okamoto and Quastel, 1970). Almost the same concentrations of potassium are obtained in slices with either of the five substances, but the sodium concentrations are lowest with glucose or lactate (Joanny and Hillman, 1963, 1964; Joanny et al., 1963). With each of these compounds as the substrate, application of electrical pulses leads to a potassium loss (and sodium gain) followed by a reaccumulation (and reextrusion) after the termination of the stimulation (Joanny and Hillman, 1963) . Both glucose, pjrruvate, lactate, and oxaloacetate maintain high membrane potentials (Hillman, 1961 ) . Other tricarboxylic acid cycle intermediates than oxaloacetate are less able to maintain the content of high-energy phosphates and potassium; application of excess potassium does not lead to an increased rate of oxygen uptake, and no potassium accumulation occurs after electrical stimulation ; the rate of oxygen uptake is often initially higher than with glucose as the substrate but shows a relatively rapid decline (for references, see Hertz, 1973a). According to Kerpel-Fronius and Hajos (1971 ) , presynaptic mitochondria are unable to oxidize a-ketoglutarate and succinate, whereas perikaryal dendrites and glial mitochondria have an intact tricarboxylic acid cycle ; only the “nonaxonal terminal” compartment should thus be able to perform a complete oxidation. I t is in accordance with the presence of an intact tricarboxylic acid cycle in both neuronal perikarya and glia cells that each of the two cell types is able to oxidize pyruvate, succinate, a-ketoglutarate and glutamate (Hydkn and Pigon, 1960; Hamberger, 1961, 1963; Rose, 1967) . In neurons these four substrates maintain approximately similar rates of oxygen uptake which are much higher than endogenous respiration (Table I V ) , but probably lower than the oxygen consumption with glucose as the substrate (Table 11).Glia cells seem on the other hand to show different rates of oxygen uptake with the various substrates, and the respiratory activity is remarkably high (i.e., above the level in a medium containing glucose) with succinate and possibly also a-ketoglutarate as the substrate (Table I V ) . Increased oxidative metabolism is accompanied by an acceleration of the turnover of respiratory coenzymes such as NADH (Chance and Williams, 1955; Chance et al., 1962). Interpretation of changes in NAD’ :NADH ratios and, especially, quantitative estimates may, hobvever,
TABLE IV RATESOF OXYGEN UPTAKEOF NEURONS AND GLIA CELLSUTILIZING DIFFERENT SUBSTRATES~
0 2
Substrate
Cell
Origin
Species
Method
Endogenous Pyruvate Succinate Succinate
Neuron Neuron Neuron Neuron
Brain stem Cortex Brain stem Brain stem
Rabbit Rat Rabbit Rabbit
Microdissection Gradient centrifugation Microdissection Microdissection
a-Ketoglutarate a-Ketoglutarate Glutamate Glu tamate Endogenous Pyruvate Succinate Succinate a-Ketoglutarate a-Ketoglutarate Glutamate Glutamate Glutamate
Neuron Neuron Neuron Neuron Glia Glia Glia Glia Glia Glia Glia Glia Glia
Brain stem Brain stem Brain stem Cortex Brain stem Cortex Brain stem Brain stem Brain stem Brain stem Brain stem Brain stem Cortex
Rabbit Microdissection Rabbit Microdissection Rabbit Microdissection Rat Gradient centrifugation Rabbit Microdissection Rat Gradient centrifugation Rabbit Microdissection Rabbit Microdissection Rabbit Microdissection Rabbit Microdissection Rabbit Microdissection Rabbit Microdissection Rat Gradient centrifugation
0 2 uptake per uptake per cell gram wet weight (p1 X hour-') (pmoles X hour-')
0 . 4 x 10-5 3 x 10-5 22 x 10-5 38 x 10-5 11 22 22
x
x x
0.4
10-5 10-5 10-5
x
10-5
6 X 7 x 10-5 1 x 10-5 4 x 10-5 1 x 10-5 2 x 10-5
2 74 123 212 61 123 123 43 17 69 251 304 50 172 33 76 43
Authors Hamberger (1961) Rose (1967) Hydtn and Pigon (1960) Aleksidze and Blomstrand (1968) Hamberger (1961) Hamberger (1963) Hamberger (1961) Rose (1967) Hamberger (1961) Rose (1967) HydCn and Pigon (1960) Hamberger (1963) Hamberger (1961) Hamberger (1963) Hamberger (1961) Hamberger (1963) Rose (1967)
a The rate of oxygen uptake per neuron obtained by gradient centrifugation has been calculated from the content of protein per cell given in :he original paper and the rates per weight unit on the assumption that the solids constitute 20%. Oxygen consumption per microdissected glia cell was deduced on the assumption of eight glia cells per sample (Hamberger, 1963) and values for capillary and neuronal glia (Hamberger, 1963) were averaged.
r m v
I
BRAIN METABOLISM AT T H E CELLULAR LEVEL
169
be complicated, as indicated by the fact that the potassium-induced stimulation of oxygen consumption by brain slices is accompanied by an increased oxidation of NAD ( P )H, whereas the electrically induced increase in metabolic rate is accompanied by an increase in the reduced component (Cummins and Bull, 1971). Conceivably this difference may be related to the fact that the intracellular potassium concentration is decreased during electrical stimulation but increased during exposure to potassium-rich media. Measurements of NAD :NADH ratios have, however, the advantage that they may be performed under i n vivo conditions, and by aid of such determinations an increase in oxidative metabolism has been observed during seizures (O’Connor et al., 1972; Lewis et al., 1974). At the same time the extracellular K+ concentration is increased (Fertziger and Ranck, 1970; Dichter et al., 1972; Prince et al., 1973). Based upon histological estimates [i.e., paucity of glia cells in hippocampus (cf., however, also Bignami and Dahl, 1974)], it was concluded that mainly neurons are affected (O’Connor et al., 1972). This conclusion seems discordant with a mainly glial localization of the potassium-induced stimulation of oxygen uptake and ATP utilization. Reconciliation of these i n uiuo and i n uitro findings seems essential for further understanding of cellular metabolic events during brain activity. 2. Metabolic Compartnaentation Based upon product-precursor relationships after exposure to different labeled substrates, it has been demonstrated that at least two metabolically distinctive tricarboxylic acid cycles are operating in the brain in uivo (Lajtha et al., 1959; Berl et al., 1961; Gaitonde, 1965; Garfinkel, 1966; Van den Berg et al., 1969; Van den Berg and Garfinkel, 1971) and i n vitro (Berl et al., 1968; Berl and Clarke, 1969; Ralazs et al., 1970; Cheng, 1973; Clarke and Berl, 1973). For the mouse brain, Van den Berg and Garfinkel (1971) deduced fluxes of 1.25 and 0.3 pmoles/minute per gram wet weight (Fig. 13) in, respectively, a large and a small compartment, combined with a flux between the two compartments of 0.14 pmole/minute per gram wet weight. In the rat, the flux between the two compartments seems to be about half of this value (Dzubow and Garfinkel, 1970), possibly reflecting the differences in rate of oxygen consumption. Conceivably the compartmentation could be due to metabolic differences between, for example, mitochondria and cytoplasm, cell bodies, axons and synaptosomes, or neurons and glia cells (e.g., Berl and Clarke, 1969; Pate1 and Balazs, 1970; Rose, 1970; Balazs, 1971). Several attempts including cytochemical and ontogenetic studies and investigations of preparations obtained by gradient centrifugation (Rose, 1970) have been made to distinguish between these possibilities (Balazs et al., 1973; Garfinkel, 1973; Van den Berg, 1973), but it seems at present not possible with certainty to corre-
170
LEIF HERTZ AND ARNE SCHOUSBOE
GLUCOSE A0.8 (1.6)
f
ACETATE
1
1
ACETYL-CoA
ACETYL-CoA
I
10.3
.-
uF 1mm.111c
1
GLUTAMATE
GLUTAMINE C-
.
0.u
I
cell membrane
FIG. 13. Overall scheme of compartmentation model for the mouse brain in uiuo. Values are fluxes in micromoles/minute per gram wet weight; L- and S-glutamate designate the glutamate of, respectively, the large and the small compartment. From Van den Berg and Garfinkel (1971 ) .
late any compartment with specific anatomical structures, and “the major direction for future work no doubt will be the correlation of metabolic compartments with cellular and subcellular structures” (Berl and Clarke, 1969).
3. Amino Acids and Related Compounds Several amino acids are found in the adult (but not in the neonatal) brain in high concentrations. N-Acetylaspartic acid and y-aminobutyric acid, which do not occur in significant concentrations in other tissues in the mammalian body, thus reach concentrations between 1 and 10 pmoles/gm wet weight (e.g., Roberts and Frankel, 1950; Tallan et al., 1956; Baxter et al., 1960; Waelsch, 1962; McIlwain, 1966; Baxter, 1970). The glutamine concentration is 4-5 pmoles/gm and the concentration of glutamic acid is higher (about 10 pmoles/gm) than in any other organ (e.g., Tower, 1960; Berl and Purpura, 1963; Agrawal et al., 1966; Mangan and Whittaker, 1966; Van den Berg, 1970b). The latter amino acid is rapidly formed from glucose in rats older than 2-3 weeks (Gaitonde and Richter, 1966; cf. also Van den Berg, 1970a) and is formed in at least two, only partly intercommunicating, compartments (Fig. 13)-i.e., a larger pool comprising most of the glutamate (and associated tricarboxylic acid cycle, cf. above) and a smaller pool (with associated tricarboxylic acid cycle) from which glutamine is formed (e.g., Lajtha et al., 1959; Berl et al., 1961; Garfinkel, 1962; 1966; Gaitonde, 1965; O’Neal and Koeppe, 1966; Berl and Clarke, 1969; Van
BRAIN METABOLISM AT T H E CELLULAR LEVEL
171
den Berg et al., 1969; Rose, 1970; Van den Berg and Garfinkel, 1971). Evidence is found (Van den Berg and Garfinkel, 1971) that the two pools may be partly connected by a flow of glutamine from the small to the large compartment (Fig. 13) and a movement of GABA in the opposite direction (cf. also Berl et al., 1970; Benjamin and Quastel, 1972). I t has been suggested that a considerable fraction of the glutamine content should be present in nonneuronal elements (Tower, 1960; Margolis et al., 1968; Benjamin and Quastel, 1972; cf. also Bradford and Thomas, 1969). Ontogenetic and topographic findings indicate that GABA mainly is formed in neurons (Baxter et al., 1960; McKhann et al., 1960; &had@ and Baxter, 1960; Tower, 1960; Rerl and Purpura, 1963; Balazs et al., 1971; Roberts, 1971). This neuronal localization has been verified by measurements on isolated neurons (Otsuka et al., 1971), and Fonnum and Walberg (1973a,b) have calculated that the concentration of GABA at certain inhibitory terminals reaches 50-100 mA4. A very high concentration of glutamate (10 pmoles/gm wet weight) has been observed in cultivated astrocytes (Schousboe et al., 1975). Both glutamate and GABA in all probability have transmitter functions (Krnjevic, 1965; Krnjevic and Whittaker, 1965; Galindo et al., 1967; Krnjevic and Schwartz, 196713). Active uptake of several amino acids with high affinity and large tissue: medium ratio has been demonstrated in brain slices (for references, see Lajtha, 1974). I t is unrelated to the level of ATP (Abadom and Scholefield, 1962a; Banay-Schwartz et al., 1971; Piccoli et al., 1971), but requires the presence of sodium (e.g., Margolis and Lajtha, 1968). For several amino acids pronounced alterations occur during brain maturation (Lajtha and Piccoli, 1971 ) . The bulk prepared glia cells show a smaller pool size but a greater uptake of most amino acids than corresponding neurons (Bradford and Rose, 1967; Rose, 1968). Among the amino acids which are accumulated to a considerable extent into glia cells are GABA and glutamate (Hamberger, 1971; Henn and Hamberger, 1971 ; Hokfelt and Ljungdahl, 1971 ; Okamoto and Quastel, 1972; Henn et al., 1973; Hutchison et al., 1973, 1974; Kelly et al., 1973; Young et al., 1973; Schon and Kelly, 1974a,b). This indicates an inactivation of transmitters by uptake into glia cells (Krnjevic and Schwartz, 1967a) and even acetylcholine may possibly be taken up into glia cells (Koelle and Koelle, 1959; Koelle, 1962). Also isolated synaptosomes are able to accumulate amino acids (Weinstein et al., 1965; Logan and Snyder, 1971). Incorporation of radioactivity from 14C-labeled glucose into especially glutamate (cf. Vrba et al., 1962), but also into aspartate, GABA, glutamine, serine, and alanine (Belloff-Chain et al., 1955; Kini and Quastel, 1959; Balazs et al., 1970) occurs rapidly. The incorporation is faster in bulk-pre-
172
LEIF HERTZ A N D ARNE S C H O U S B O E
pared glia cells than in bulk-prepared neurons (Rose, 1968) ; only little glutamine is formed in synaptosomes (Bradford and Thomas, 1969). Excess potassium leads to an increased release of 14C from preparations of brain cortex or retina loaded with 14C-labeled glutamate (Hertz and Arnfred, 1967; Hertz, 1968, 1969; Katz et al., 1969; Bradford, 1970; Van Harreveld and Fifkova, 1970; Arnfred and Hertz, 1971). A similar potassium-induced release of endogeneously formed glutamate has been observed in synaptosomes (De Belleroche and Bradford, 1972) but it may be smaller in brain slices (Machiyama et al., 1967, 1970; Snodgrass and Iversen, 1973). Chromatography has shown that the increased amount of 14C, which is released from labeled glutamate, largely is due to an efflux of unmetabolized glutamate (Van Harreveld and Fifkova, 1970; Arnfred and Hertz, 1971). Also the efflux of GABA is increased by excess potassium (Machiyama et al., 1967, 1970; Srinivasan et al., 1969; Arnfred and Hertz, 1971; De Belleroche and Bradford, 1972). This effect of high concentrations of K+ differs both quantitatively and ontogenetically ( Weiss and Hertz, 1974; A. Schousboe, V. Lisy, and L. Hertz, unpublished experiments), from that on ion metabolism and swelling (cf. below). Application of electrical pulses resembles excess potassium in raising the efflux of glutamate, aspartate, and GABA from brain slices, synaptosomes, or retina (Katz et al., 1969; Srinivasan et al., 1969; Bradford, 1970; Van Harreveld and Fifkova, 1970; De Belleroche and Bradford, 1972). Calcium is required for a maximum stimulation of the release of GABA (Katz et al., 1969; Srinivasan et al., 1969; Arnfred and Hertz, 1971; Roberts, 1974), aspartate and glutamate (De Belleroche and Bradford, 1972; Roberts, 1974), which is in keeping with the concept that these compounds may function as transmitters. The role of proteins and nucleic acids in brain function has received increasing attention in recent years (e.g., HydCn, 1967b,c; Lodin and Rose, 1968; Altman, 1969; Lajtha, 1970; Richter, 1970). Certain proteins seem to be localized mainly or exclusively in either neurons or glia cells. S-100 has thus a predominantly glial localization (HydCn and McEwen, 1966) and is not found in rats until the age of about 2 weeks (Moore, 1969; Herschman et al., 1971). A glia fibrillary acidic protein, GFA, is specific for astrocytes (Eng et al., 1971), and occurs in large quantities (E. Bock and L. Dittmann, personal communication) in the astrocytic culture developed by Booher and Sensenbrenner (1972). The 14-3-2 protein (Moore and Perez, 1968) has, in contrast a predominantly neuronal localization (Herschman et al., 1973), and the major part of overall protein synthesis seems to occur in neurons (e.g., Blomstrand and Hamberger, 1969; Tiplady and Rose, 1970, 1971; Blomstrand, 1971; Lisy and Lodin, 1973; cf. however, also Johnson and Sellinger, 1971; Sellinger et al., 1971).
BRAIN METABOLISM A T T H E CELLULAR LEVEL
173
4. Glycolysis and C 0 2 Production The R Q of brain in vivo is 1.00 (Gibbs et al., 1942) and the amount of glucose consumed corresponds almost quantitatively to the production of CO, measured as oxygen consumption (Kety, 1965; cf., however, also Sacks, 1969). This indicates that the production of lactate is small under ordinary experimental conditions. I n spite of sufficient oxygen supply, it increases greatly when convulsions are evoked ( McIlwain, 1966; Bolwig and QuistorfT, 1973; cf. also Klein and Olsen, 1947). I n brain slices, the rate of lactate production seems to be higher than in the brain in vivo, and most authors reach estimates of about 25 pmoles/hour per gram wet weight (e.g., Elliott and Henderson, 1948; Dixon, 1949; McIlwain, 1953a,c; Thomas and McIlwain, 1956; Takagaki, 1968). Also the total absence of lactate production has, however, been reported (e.g., Dickens and Greville, 1935). The reason for this discrepancy probably is that the aerobic glycolysis declines rapidly with time (Elliott and Henderson, 1948; McIlwain, 1953b; Cremer, 1967; cf., however, also Kimura, 1940). Together, the production of lactate and the respiration account relatively well for the consumption of glucose (McIlwain et al., 1952). I n most tissues the rate of glycolysis increases under anaerobic conditions (the Pasteur effect). The values given in the literature for anaerobic lactate production in brain slices vary from about 40 pmoles/hour per gram wet weight in the rat (Elliott and Henderson, 1948) to 100-200 pmoles/hour per gram wet weight in man and in the rabbit (Elliott and Penfield, 1948; Dixon, 1949). This means that the rate of anaerobic glycolysis is higher in species of bigger body size; i.e., the correlation is the reverse of that found with respect to oxygen uptake (cf. Elliott, 1952, 1957). Also in brain slices the production of carbon dioxide is almost equal to the consumption of oxygen, giving an R Q close to 1.00 (e.g., Dickens and Simer, 1931; Elliott et al., 1937; Elliott and Penfield, 1948; Nishimura and Kimura, 1965). After addition of labeled glucose the respiratory 14C02 production increases only slowly, indicating that the equilibration of the labeled metabolites with the pools requires appreciable time (Beloff-Chain et al., 1955; Cremer, 1967; Balazs et al., 1970). Absence of sodium leads to an increase in the rate of aerobic lactate production by brain slices (Takagaki and Tsukada, 1957; Takagaki et d., 1959; RulEdk and Whittam, 1967; cf., however, also Gore and McIlwain, 1952) and synaptosomes (Diamond and Fishman, 1973). Also the anaerobic glycolysis in brain slices is inhibited by the presence of sodium (Racker and Krimsky, 1945; Utter, 1950; Racker, 1952). A considerable increase in the aerobic glycolysis by brain-cortex slices is evoked by electrical stimulation (McIlwain et al., 1952; McIlwain and Tresize, 1956) and by high
174
LEIF HERTZ A N D ARNE SCHOUSBOE
concentrations of potassium, rubidium, cesium, or ammonium. A maximum response is obtained with, for example, 50-100 m M potassium (Ashford and Dixon, 1935; Kimura, 1937; Dixon, 1949; Gore and McIlwain, 1952), and the threshold concentration is about 20 mM (Takagaki, 1968, 1972). The potassium-induced stimulation requires the presence of sodium (Takagaki and Tsukada, 1957; RugEBk and Whittam, 1967). Anaerobic glycolysis is in contrast inhibited by high concentrations of potassium leading to a “reversal of the Pasteur effect” (e.g., Ashford and Dixon, 1935; Dickens and Greville, 1935; Dixon, 1949; Elliott and Bilodeau, 1962). The formation of I4CO2from labeled pyruvate or glucose is increased by excess potassium, electrical stimulation or addition of dinitrophenol (Quastel, 1959; Gonda and Quastel, 1962; De Piras and Zadunaisky, 1965; O’Neill et al., 1965), and the output of labeled CO, from glucose seems to be more enhanced than the oxygen consumption (Hoskin, 1960; Cremer, 1967 ; Machiyama et al., 1970; cf., however, also Gainer et al., 1962). An intense aerobic glycolysis has been reported in neurons (Sotelo, 1967) whereas the aerobic glycolysis in synaptosomes is 15-20 pmoles/hour per gram wet weight (Bradford and Thomas, 1969), i.e., the same value as in brain slices. Microdissected neurons and glia cells from the rabbit show an anaerobic glycolysis corresponding to about 8 x p1 of CO, per hour per sample (20 x gm dry weight) or approximately 200 pmoles/hour per gram wet weight (Hamberger and HydCn, 1963), i.e., again the same value as that observed with slices. Undifferentiated cells from the chick brain exhibit a similar glycolytic rate which, however, shows some decrease during ontogenetic development (Dittman et al., 197313). Cell cultures from normal tissue have at their early stages a rate of aerobic (Lehrer and Bornstein, 1968) and anaerobic (Dittmann et al., 197313) glycolysis which is identical to that in the tissue from which they were cultured; in contrast to the findings in the living animal there is a moderate increase with time of cultivation (cf. also Bornstein, 1973). The C6 astrocytoma has a high glycolytic rate (Lust et al., 1974). I n a series of elegant experiments, Lowry and co-workers (e.g., Passonneau and Lowry, 1971) have measured metabolic rates expressed as changes in energy reserves during the first minute after cutting off blood supply. Under these conditions the contents of glucose and of glycogen decrease considerably faster in the neuropil than in the nerve cells bodies (Fig. 14).
5. Enzymes Cerebral enzymes have been extensively reviewed by Seiler (1969). Since then, studies of cultivated cells (e.g., Blume et al., 1970; Lehrer et al., 1970; Ciesielski-Treska et al., 1972; Maker et al., 1972; Tholey et al., 1972; Wilson et d.,1972; Mandel et d.,1973) and of cells obtained from gradient
175
BRAIN METABOLISM AT T H E CELLULAR LEVEL
f 60 0
X
m c -
L
Y
I
m o o
:LO
.
s 2
E - 20 el m
E
0 0
0 0
la 0
LO
0
Time a f t e r decapitation Isec
40
1
FIG. 14. Values for glycogen ( a ) and glucose ( b ) in anterior horn cell bodies ( - - - ) and neuropil ( . -. -) from four anesthetized mice, two of which were frozen immediately after decapitation and the other two frozen after a delay of 40 seconds. Values for average anterior horn are given for comparison ( - ) . From Passonneau and Lowry (1971).
centrifugation (e.g., Lisy et al., 1971; Radin et al., 1972; Sinha and Rose, 1972; Nagata et al., 1974) have added further information about regulation and cellular localization of enzymes. Some enzymes show such cellular specificity that they may be used as “markers” for either neurons [e.g., pgalactosidase (Sinha and Rose, 1972) and ceramide glycosyltransferase (Radin et al., 1972)] or glia cells [carbonic anhydrase (e.g., Giacobini, 1961, 1962, 1964; Friede, 1966; Rose, 1967)l. The possible physiological importance of the carbonic anhydrase has been discussed by Tschirgi (1958), by Giacobini (1964), and by Bourke and co-workers (Bourke and Nelson, 1972; Kimelberg and Bourke, 1973). The Na+-K+-ATPase has already been discussed (cf. Section 111, B, 2 ) . Na+ is also of importance for other enzymic activities in astrocytes (Friede, 1964). Several enzymes, e.g., pyruvate kinase (Boyer et ul.,’ 1942, 1943; Kachmar and Boyer, 1953; Melchior, 1965) are activated by K’, and the potassium-induced stimulation of aerobic glycolysis is probably due to a direct or indirect (via ATP) effect on several enzymes including the pyru-
176
LEIF HERTZ A N D ARNE S C H O U S B O E
vate kinase (Takagaki, 1968), the phosphofructokinase and the glyceraldehyde-3-phosphate-dehydrogenase (Takagaki, 1972). The formation of acetylcoenzyme A from acetate is stimulated by excess potassium and inhibited by sodium (von KorlT, 1953; Evans et al., 1964; Webster, 1966), and Quastel and co-workers have suggested that also the rate of conversion of pyruvate to acetyl coenzyme A may be stimulated in media containing high concentrations of potassium (e.g., Kini and Quastel, 1959, 1960; Quastel, 1960, 1962; cf. also Bilodeau and Elliott, 1963). As a purely hypothetic possibility, K + might, besides the ordinary reactions in glycolysis and tricarboxylic acid cycle, affect the transport of metabolites between different metabolic compartments.
IV. Ion and Water Metabolism
A. COMPARISON BETWEEN IONAND WATERMETABOLISM in vim A N D in vitro 1. W a t e r Content The solid content accounts for 20% in cortex from adult animals (e.g., Manery and Hastings, 1939; Pappius and Elliott, 1956a; Franck, 1970) and 30% in white matter (Reulen et al., 1970). The water content decreases during ontogenesis and shows some regional but little interspecies variation (for references, see Katzman and Pappius, 1973). The water phase comprises intracellular and extracellular (interstitial) water. As discussed in Section II,B, the magnitude of the extracellular space of the brain in uiuo is not known with certainty. Furthermore, a larger extracellular space has been observed in species with larger brain size than in those with smaller brain size (Bourke et al., 1965), and the extracellular space seems to decrease during maturation (Flexner and Flexner, 1949; Vernadakis and Woodbury, 1965; Ferguson and Woodbury, 1969; Caley and Maxwell, 1970; cf., however, also Pappius, 1969). In smaller adult mammals the extracellular space probably does not exceed 25% (e.g., Davson, 1967; Oldendorf and Davson, 1967; Woodward et al., 1967; Pappius, 1969; Marchbanks, 1970), and a value of 15-20% has been obtained by the use of the diffusion profile technique (Rall et al., 1962; Fenstermacher et al., 1970; Katzman and Pappius, 1973). During incubation of brain slices from adult animals, the water content increases gradually from its in uivo value. Already 10 seconds of contact with a physiological medium may lead to a swelling (calculated as percent of the initial wet weight) of between 3% (Franck, 1970) and 5-10%
RRAIN METABOLISM AT THE CELLULAR LEVEL
177
(Varon and McIlwain, 1961; Pappius et al., 1962; Keesey et al., 1965; Bourke and Tower, 1966a). In the course of the subsequent incubation at 37OC under optimum conditions in an oxygenated, physiological medium, the tissue may keep its weight constant for hours (Bourke and Tower, 1966a; Franck et al., 1968; Franck, 1970; Lund-Andersen and Hertz, 1970) leading to a total swelling of only 10-15%. A swelling of 40-50% is, however, often encountered. Anaerobiosis or glucose deficiency increases the swelling considerably ( Pappius and Elliott, 1956a; Bourke and Tower, 1966a; Franck et al., 1968; Okamoto and Quastel, 1970). Small regional differences seem to occur (Hertz et al., 1970) , and in the rat the tendency of the brain slices to swell develops during the first postnatal weeks (Franck and Schoeffeniels, 1972; Schousboe, 1972). Originally the swelling was supposed to occur extracellularly (cf. Terner et al., 1950; Pappius and Elliott, 1956a,b), but more recent studies have shown that it is not exclusively or even not predominantly due to an increase in the extracellular water phase. The fluid which is taken up almost immediately after exposure to a physiological medium is accessible to extracellular markers and may represent an increase in the extracellular volume (Torack et al., 1965; Pate1 et al., 1971; Mraller et al., 1974), adhering medium, or a damaged part of the slice (Pappius and Elliott, 1956a; Varon and McIlwain, 1961; Pappius et al., 1962; Keesey et al., 1965; Pappius, 1965). The remaining part of the swelling occurs intracellularly (Pappius and Elliott, 1956a; Varon and McIlwain, 1961; Pappius et al., 1962; Bourke and Tower, 1966a; Franck et al., 1968; Franck, 1970; Ibata et al., 1971; Schousboe and Hertz, 1971a; Lund-Andersen, 1974; Mraller et al., 1974). The swelling has caused considerable difficulties since a prerequisite for discussion of, for example, ion distribution and transport is that the content of the substance in the incubated tissue is expressed in relation to a n unchanging weight unit. Accordingly, concentrations of ions (e.g., K') or other compounds have often been expressed in relation to the initial wet weight with a correction made for a presumed extracellular swelling. The demonstration that the swelling is not exclusively an extracellular phenomenon implies that such a correction can no longer be regarded as permissible.
2. Ion Content The ion concentrations in brain tissue removed immediately after the death of the animal are probably representative of those occurring in uiuo (Hillman, 1964). The potassium concentration in adult gray matter is about 100 pmoles/gm wet weight (Yannet, 1940; Ames and Nesbett, 1958; Bachelard et al., 1962) and the sodium concentration is about half of this value (cf. also Grossman et al., 1968). I n white matter the potassium concentration may be slightly lower, and the sodium concentration slightly
178
LEIF HERTZ AND ARNE SCHOUSBOE
higher, but the differences between gray and white matter are probably small (Stewart-Wallace, 1939; Levy et al., 1965; Tower, 1969; Katzman and Pappius, 1973). I n cortex from newborn animals the potassium and sodium concentrations are of approximately equal magnitude, i.e., 70-90 pmoleslgm wet weight (Wender and Hierowski, 1960; Vernadakis and Woodbury, 1962; Valcana and Timiras, 1969; Franck, 1970). The cation concentration shows no major interspecies variations in mammals (Pappius and Elliott, 1954). The chloride concentration has, in contrast, been found to increase with an increasing brain volume (Bourke et al., 1965; cf., however, also Katzman and Pappius, 1973). I n the rat cerebral cortex the chloride concentration is about 30, and in the rabbit about 40, pmoles/gm wet weight (e.g., Bourke et al., 1965). I t may be slightly lower in white matter (Tower, 1969). The concentration of potassium found in vitro is lower than that occurring in vivo, whereas the sodium and chloride concentrations are higher. The reason for the low K+ concentration during incubation is a pronounced loss of this ion during and immediately after the preparation of the tissue; the loss is followed by a more or less complete reaccumulation (e.g., Krebs et al., 1951; Pappius and Elliott, 1956b; Bachelard et al., 1962; Bourke and Tower, 1966b; Israel et al., 1966; Franck, 1970). The highest K+ concentrations in brain-cortex slices from adult rats after incubation in physiological media seem to be those of 70-75 pmoles/gm final wet weight obtained by Franck et al. (1968)) Franck (1970), and Lund-Andersen and Hertz (1970) using a specially designed microtome (Franck et al., 1968) and massive oxygenation (cf. also Arnfred et al., 1970). Under anoxic conditions the potassium concentration becomes much lower (e.g., Pappius and Elliott, 195613; Bourke and Tower, 196613; Franck et al., 1968; Lund-Andersen and Hertz, 1970). I n brain-cortex slices from newborn rats and mice the potassium concentration is significantly higher than in mature tissue (Franck, 1970; Schousboe and Hertz, 1971b; Schousboe, 1972; Hertz et al., 1975). An augmentation in sodium content occurs concomitantly with the potassium loss during and immediately after the preparation of brain slices; a subsequent extrusion of sodium seems more difficult to obtain than the corresponding uptake of potassium ions, but may be observed if the tissue is treated with special care (e.g., Bachelard et al., 1962; Franck et al., 1968). Also the concentration of chloride is normally higher in brain slices than in vivo (Cummins and McIlwain, 1961; Bachelard et aZ., 1962; Keesey et al., 1965). The high concentration of chloride indicates that a considerable amount of this ion is found intracellularly (Allen, 1955; Varon and McIlwain, 1961; Keesey et al., 1965; Bourke, 1969a,b; Schousboe and Hertz, 1971a).
BRAIN METAROLISM AT T H E CELLULAR LEVEL
179
The calcium concentration in incubated brain slices depends upon that in the medium but is generally somewhat higher (Charnock, 1963; Lolley, 1963; Ramsay and McIlwain, 1970). 3. Transport of Ions and Water Transport in uiuo between the brain and its surroundings represents special problems due to the presence of barrier systems and will not be treated in this review (see, e.g., Davson, 1967; Katzman and Pappius, 1973). T h e uptake of water during the development of cerebral edema is of considerable practical and theoretical interest, but also the discussion of this subject seems to be beyond the scope of the present paper (see, e.g., Bakay, 1965; Pappius, 1965, 1969; Tower, 1967; Bourke et al., 1970; Katzman and Pappius, 1973). Measurements of influx or efflux (e.g., Levi and Ussing, 1949; Harris, 1960; Huxley, 1960; Solomon, 1960; Winegrad and Shanes, 1962; Hagemeijer and van Remoortere, 1969) have been performed both in brain slices (Krebs et al., 1951; Cummins and McIlwain, 1961; Keesey and Wallgren, 1965; Franck and Cornette, 1966; Bourke and Tower, 1966a; Hertz, 1968; Franck, 1970, cf. also Brinley, 1963), in isolated sympathetic ganglia (Harris and McLennan, 1953; Brinley, 1967), and in the total frog brain (Zadunaisky and Curran, 1963; Bradbury et al., 1968). The efflux of extracellularly located inulin is rapid (Schousboe and Hertz, 1971a; Lund-Andersen and Hertz, 1973; Lund-Andersen, 1974) indicating free extracellular diffusion (McLennan, 1957). The diffusion coefficients for K+ and Na' are about 10 times larger (cf. Harris and Burn, 1949; Harris and McLennan, 1953), than that reported for inulin (Phelps, 1965), and a half time (tip) of 0.5-1.0 minutes may be expected for extracellular diffusion from 0.5 mm thick slices of these ions provided their charges do not cause any retardation (cf. Harris and McLennan, 1953; McLennan, 1957). All potassium in brain slices from adult rats is exchangeable within about 1 hour, whereas the exchange occurs much more slowly in brain slices from newborn animals (Fig. 15). The desaturation curves describing the washout of 4'K from slices previously loaded with the radioisotope indicates the presence of at least two, kinetically different, compartments (Hertz, 1968). By refinement of the graphical analysis, Franck ( 1970) distinguished between one extracellular and two kinetically different, rapidly exchanging, cellular compartments. The slowest of these rapidly exchanging compartments ( t w 12 minutes) comprises the largest amount of radioactive potassium in the tissue and was envisaged to represent nerve cell bodies and dendrites (for further details, see Franck, 1970). From the potassium concentrations and the rate constants the approximate magnitudes of the fluxes between this compartment (or the tissue as a Ivhole) and the medium can be estimated
180
LEIF HERTZ AND ARNE SCHOUSBOE
1.0
0.9 0.8
0.7 0.6 0.5
0.4 0.3 0.2 0.1
1 1 1 1 1 1 1 1 15 30 45 60 75 90 105 120 Time from s t a r t of i n c u b a t i o n I m i n 1
FIG. 15. Uptake of "K into brain cortex slices from adult
(0) and
newborn
( 0 )rats as a function of time of incubation in radioactive physiological medium. The uptake is shown as the ratio between the specific activity (s.a.) of the radioisotope in the medium and in the slices; i.e., 1.0 indicates a complete exchange. Results for adult rats are from Franck ( 1970), and those for newborn animals are unpublished data of A. Schousboe and L. Hertz.
to 1-3 pmoles/minute per gram wet weight during incubation in oxygenated, physiological media (Franck and Cornette, 1966; Hertz, 1968; Franck, 1970) though a somewhat higher value (6 pmoles/minute per gram) was observed by Cummins and McIlwain ( 1961) . The desaturation curve observed after loading with radioactive sodium resembles in principle that obtained with potassium, since again two or more compartments are observed. These may or may not correspond anatomically to the same potassium compartments, and in this case it is the fastest of the two rapidly exchanging cellular fractions described by Franck (1970) and suggested to represent glia cells, which accounts for most of the radioactivity. The fluxes between the rapidly exchanging cellular fraction(s) and the medium are faster than the corresponding potassium fluxes (Hertz, 1968; Franck, 1970) and from the rate constants of about 0.07 min-l obtained in the frog brain by Zadunaisky and Curran (1963) and in rat braincortex slices by Hertz (1968) the efflux can be calculated to about 5 pmoles/minute per gram final wet weight (cf. also Keesey and Wallgren, 1965). Also the chloride fluxes in brain-cortex slices amount to a few micromoles per minute per gram wet weight (Bourke, 1969a). Both influx and efflux depend upon the chloride concentration in the medium, and the influx into the cellular compartment conforms to Michaelis-Menten kinetics indicating mediated but not necessarily active transport (Bourke, 1969a).
BRAIN METABOLISM AT T H E CELLULAR LEVEL
181
The exchange of water in brain slices occurs extremely rapidly (Cohen et al., 1968; Okamoto and Quastel, 1970). 4. Stimulatory Effects on Contents and Transport of Ions and Water An augmentation of the potassium concentration above 20 mM leads to a distinct increase in the degree of swelling in brain-cortex slices or the perfused cortex (Elliott, 1955; Pappius and Elliott, 1956a; Bourke and Tower, 1966a; Bourke, 1969b; Bourke et al., 1970; Bourke and Nelson, 1972; Lipton, 1973) and to a decrease in the interstitial space (Schousboe and Hertz, 1971a; Lund-Andersen, 1974; Meller et al., 1974). This increase in the intracellular swelling may be competitively counteracted by an increase in the sodium concentration, but requires, on the other hand, the presence of a certain amount of sodium (Lund-Andersen and Hertz, 1970). The presence of chloride [or nitrate (L. Hertz, unpublished experiments) ] is obligatory (Bourke and Tower, 1966a; Bourke, 196913; Bourke et al., 1970; Lund-Andersen and Hertz, 1970), and no potassium-induced swelling occurs if all chloride is replaced with the indiffusible isethionate ion. This reflects the fact, that the swelling fluid appears to be an isotonic-fluid expansion composed predominantly of K+ and C1- ions (Bourke, 196913). The potassium-induced increase in swelling of brain-cortex slices occurs both in hypertonic media containing an unaltered concentration of sodium (e.g., Lund-Andersen and Hertz, 1970) and in isotonic media where the augmentation of the potassium concentration is compensated by a reduction of the sodium concentration (e.g., Bourke, 1969b). As discussed in detail by Boyle and Conway ( 1941) and by Shanes ( 1916), the swelling after incubation in media with reduced sodium content can both qualitatively and quantitatively be explained as a result of osmotic, electrical, and Donnan equilibria, whereas no swelling can be expected when potassium chloride is added to a medium with an unaltered sodium concentration. The uptake of water under hypertonic conditions can accordingly not be accounted for by Donnan forces, but must be explained in another way. The observations that the potassium-induced swelling is diminished by a lowering of the temperature (Bourke, 1969b) and that the concomitant uptake of 3GClfollows Michaelis-Menten kinetics [with respect to both chloride and potassium (Bourke, 1969a)l may indicate participation of an active transport mechanism. I t has therefore been suggested that the water uptake is secondary to an accumulation of potassium and chloride ions (Pappius, 1965; Bourke, 196913; Lund-Andersen and Hertz, 1970; Bourke et al., 1970; Bourke and Nelson, 1972). The finding that the potassium-induced swelling, like the potassium-induced stimulation of oxygen uptake, requires the presence of sodium, but, on the other hand, is competitively inhibited by excess sodium
182
LEIF HERTZ A N D ARNE SCHOUSBOE
--
-f
-
I
'
I
I
I
I
130 120'
.-
.c
E,
110
%
- loo -E, 90 Y)
a
0
v
80 c
.'
70
3 a
External K* conc. (mM)
FIG. 16. Concentrations (pmolesigram final wet weight) of sodium ( 0 ) and in rat brain-cortex slices as a function of the external potassium potassium (0) concentration after 1 hour of incubation in a physiological medium to which different amounts of KCI had been added. Results are means of 7-25 experiments, with SEM indicated by vertical bars if they extend beyond the symbols. Potassium concentrations are from Lund-Andersen and Hertz (1970) and the sodium concentrations are unpublished results of H. Lund-Andersen and L. Hertz.
is compatible with the concept of an active uptake involving an enzyme system akin to the Na+-K+-ATPase. Prolonged application of electrical pulses may induce an increase in swelling (Thomson and McIlwain, 1961 ; Bachelard et al., 1962; Okamoto and Quastel, 1970; cf., however, also Varon and McIlwain, 1961). An increase to between 16 mM (Bourke and Tower, 1966b) and 20 mM potassium (Lund-Andersen and Hertz, 1970; cf. also Franck, 1970) in the medium leads to a significant decrease in the sodium content (and concentration) in the tissue (Fig. 16). A further increase of the potassium concentration to between 20 and 50 m M causes a steep rise; this rise is attenuated, but not abolished in chloride-free media (where no increased swelling occurs) and is accordingly not a simple consequence of the swelling (Lund-Andersen and Hertz, 1970) ; it is partially inhibited by procaine (C. S. Kjeldsen and L. Hertz, unpublished experiments). Also application of electrical pulses leads to a considerable gain of sodium ions; the sodium uptake is approximately equal to a concomitant loss
BRAIN METABOLISM AT T H E CELLULAR LEVEL
183
of potassium (Varon and McIlwain, 1961; Bachelard et al., 1962; Joanny and Hillman, 1963, 1964; cf. however, also Cummins and McIlwain, 1961), and it has been suggested (McIlwain, 1963; Keesey et al., 1965) that the increase in intracellular sodium concentration during the electrical stimulation might trigger the metabolic response. The previous tissue concentration is more or less completely regained during the first minutes after termination of the stimulation (Hillman et al., 1963; Keesey et al., 1965). A rise of the potassium concentration from the about 5 m M found in physiological media leads to an increase in the potassium concentration in the tissue which is steep in the interval 5-20 m M (Fig. 16), but negligible in the range 20(25)-50 mM (Franck, 1970; Lund-Andersen and Hertz, 1970). The high potassium concentration at an external potassium concentration around 20 m M is mirrored by the low sodium concentration indicating a, probably active, exchange between the two ions. T h e hypothesis has been forwarded (Fig. 17) that the depolarization evoked by a further increase of the external K+ concentration leads to a potassium loss and a sodium gain, and that the potassium loss is balanced by a compensatory accumulation of potassium and chloride (cf. Bourke, 1969a,b), which for osmotical reasons leads to an increased swelling. No corresponding extrusion of sodium seems to occur (Hertz, 1968; Franck, 1970). The compensatory uptake of KCl is absent neonatally (Schousboe, 1972). Addition of excess potassium as the chloride salt causes an increase of the chloride concentration in the tissue, and also the ratio between the chloride concentration in the tissue (expressed per gram final wet weight) and that in the medium, i.e., the chloride space, may be considerably larger in potassium-rich than in physiological media (Bourke and Tower, 1966a; Bourke, 1969a; Schousboe and Hertz, 1971a). Such an increase is to be expected if a depolarization is evoked and the chloride ion is in electrochemical equilibrium, but solid evidence for stimulation of an active transport mechanism has also been presented (Bourke, 1969a; cf. also Gill et d., 1974). Application of electrical pulses causes a decrease in the potassium concentration of brain slices ; the decline begins immediately, is independent upon the substrate used and reaches a maximum of about 20 pmoles/gram wet weight after 5-10 minutes (Cummins and McIlwain, 1961; Joanny and Hillman, 1964). During the first 2-3 minutes after termination of the stimulation an almost quantitative reaccumulation of potassium occurs (Keesey et al., 1965). A considerable increase in potassium efflux and influx in the rapidly exchanging cellular fraction (s) is evoked by excess potassium (Hertz, 1968 ; Franck, 1970). Also the potassium concentration in the tissue increases and the magnitudes of the fluxes increase from 1-2 pmoles/minute per gram
184
LEIF HERTZ AND ARNE S C H O U S B O E
0P
K+ N d
a
b
K+
Na+
C
FIG. 17. Diagrammatic representation of movements of potassium and sodium ions in brain slices exposed to 5 , 20, or 50 m M external potassium. T h e solid-line arrows indicate the net uptakes or releases that would occur if only diffusion took place, whereas the dashed arrows show movements that probably are evoked by active transport. Each square represents the same amount (dry weight) of tissue. I n the physiological medium ( a ) the active accumulation of potassium equals the potassium loss by diffusion and the active extrusion of sodium equals the sodium uptake by diffusion. Accordingly, the potassium content is kept constant above, and the sodium content below, the medium concentrations. No swelling occurs. An increase of the external potassium concentration to 20 m M ( b ) alters the potassium gradient across the membrane, but does not seem to affect the membrane permeability to either sodium or potassium. This is indicated by solid-line arrows of the same magnitude as in ( a ) . T h e active uptake of potassium and the active extrusion of sodium is slightly increased, leading to a decrease in sodium concentration and to a greater rise in potassium concentration than could be expected from the increase in diffusion alone. T h e cellular osmolarity remains unchanged. A further increase of the external K' concentration to 50 mM ( c ) causes a large rise of the permeability to both sodium and potassium. Simultaneously the active uptake of potassium is further enhanced without any corresponding increase of sodium extrusion. Accordingly, the net loss of potassium becomes smaller than the net gain of sodium. A net uptake of negatively charged ions, probably chloride, must occur together with the uptake of potassium ions. This leads to a cellular hypertonicity and thus to an increased swelling. From Lund-Andersen and Hertz (1970).
final wet weight in a physiological medium to 10 pmoles/minute per gram in a potassium-rich medium. The potassium-induced increase of K+ fluxes may, in contrast to the effect on oxygen uptake and swelling be observed in slices from newborn rats (Schousboe and Hertz, 1971b). Also during application of electrical pulses, both in- and efflux of potassium increase to about 10 pmoles/minute per gram wet weight (Cummins and McIlwain, 1961; cf. also Franck and Cornette, 1966; Franck, 1970). The increased influx continues during the first minutes after termination of the electrical stimulation and leads to a considerable (4-10 pmoles/minute per gram wet weight) net gain of potassium (Cummins and McIlwain, 1961; Bachelard et al., 1962; Keesey et al., 1965).
BRAIN METABOLISM AT T H E CELLULAR LEVEL
185
The influx of sodium into brain-cortex slices seems to be independent of an increase in the external potassium concentration (Hertz, 1968; Franck, 1970), but it should be remembered that the tissue is not in a steady state since the sodium content increases. The rate constant of the efflux from the rapidly exchanging compartment (Hertz, 1968), or at least from its quantitatively dominant, second compartment (Franck, 1970), is also not affected by excess potassium. Franck (1970) found analogously electrical stimulation to have no effect on the efflux of 24Na, but it has been claimed by Keesey and Wallgren (1965) that both influx and efflux of sodium are increased by application of electrical pulses. The identical slopes of washout curves with and without stimulation (Fig. 4 in Keesey and Wallgren, 1965) is, however, not in agreement with this conclusion. Addition of excess potassium during the washout of 36Clfrom brain-cortex slices has no effect on the rate constant (Hertz, 1968; Schousboe and Hertz, 1971a), but the chloride content is increased. The initial velocity of the chloride influx into brain-cortex slices is significantly increased by excess potassium and follows Michaelis-Menten kinetics with a K , (for K’) of about 30 mM (Bourke, 1969a). Both influx and efflux of calcium seem to be transitorily increased by electrical stimulation of brain slices (Lolley, 1963), whereas the rate of water exchange is slightly decreased (Okamoto and Quastel, 1970). An in vivo phenomenon which seems to bear a close correlation to effects on ion content and transport exerted by excess potassium or electrical stimulation is Le2o’s spreading depression (LeBo, 1944) recently reviewed by Burel et al. ( 1974). This is a peculiar neurophysiological phenomenon where a wave of suppression of the normal electrical activity spreads from a stimulated point of the exposed brain over almost its entire surface. The propagation is accompanied by a release of radioactive potassium ions from brain tissue which has been preloaded with 42K (Brinley et al., 1960; KEivinek and Burel, 1960; Brinley, 1963; VyskoEil et al., 1972), but no concomitant decline occurs in total brain potassium content after a prolonged period of spreading depression (Burel et al., 1974). A key role in the process has been attributed to the potassium ion (Grafstein, 1956) and to glutamate (Van Harreveld and Fifkova, 1970, 1973). The spreading depression shows certain metabolic analogies with the potassium-induced stimulation of oxygen uptake (e.g., Burel et al., 1960; Ochs, 1962; cf. also Hertz, 1965; Van Harreveld, 1966), and the two phenomena appear simultaneously (i.e., at about 14 days in the rat) during ontogenesis (Burel, 1957; Burel et al., 1964). I n younger rats a similar potassium release is evoked by procedures normally leading to spreading depression, but the neurophysiological phenomenon itself is absent (Ki.iv6nek and Burel, 1960).
186 B. EVENTSAT
LEIF HERTZ AND ARNE SCHOUSBOE
THE
CELLULAR LEVEL
1. Membrane Potentials Membrane potentials in individual cells are determined by the distribution of certain ions (Goldman, 1943; Hodgkin and Katz, 1949), of which K generally is the most important. If the intracellular concentration of potassium is regarded to be in electrochemical equilibrium with the external concentration, i.e., the cell behaves as a perfect potassium electrode, one may write (Nernst, 1898)
Em = 61 log (Ko+/Ki+) where Em is the membrane potential in millivolts , 61 is the Nernst coefficient at 37"C, and KO+and Ki+ are the concentrations of potassium outside and inside the cell. Knowledge of Em and either KO+or Ki+ may therefore under these conditions allow calculations of K;+ and of KO+(Kuffler and Nicholls, 1966; Orkand et d.,1966). The high potassium concentration in the brain is consistent with the presence of resting in uiuo potentials up to -60 to -90 mV in neurons (Phillips, 1956; Takahashi, 1965; Lux and Pollen, 1966) and -70 to -95 mV in glia cells (Grossman and Hampton, 1968; Dennis and Gerschenfeld, 1969; Grossman et al., 1969; Grossman and Rosman, 1971; Dichter et al., 1972; Glotzner, 1973; Ransom and Goldring, 1973a) ; the average potential (Krnjevic and Schwartz, 1967a) seems to be considerably higher in the latter cell type (-62 mV) than in the former (-36 mV) , but it is probably at present not possible to obtain experimental determination of potentials in the complicated interlacement of neuronal and glial processes constituting the neuropil. In vitro measurements have shown the presence of membrane potentials of up to -80 mV in cells of brain slices (Li and McIlwain, 1957; Hillman, 1961; Hillman and McIlwain, 1961; Gibson and McIlwain, 1965; Bradford and McIlwain, 1966), -50 mV in microdissected (Hillman and HydCn, 1965a) or cultivated neurons (Scott et aZ., 1969; Athias et al., 1973; Lawson and Biscoe, 1973), and -70 mV in glia cell cultures (Hild et al., 1958; Hild and Tasaki, 1962; Hild, 1964). Attempts to record potentials from bulk-prepared neurons have failed (Bradford and Rose, 1967). An especially useful preparation for the study of glial membrane potentials is the packet glia cell of the leech which, owing to its large size, is well suited for measurements of membrane potentials. Potentials of -60 to -80 mV have been obtained in this preparation (Kuffler and Potter, 1964) and in glia cells from the Necturus optic nerve (Kuffler and Nicholls, 1966; Kuffler et al., 1966).
BRAIN METABOLISM AT THE CELLULAR LEVEL
187
High external concentrations of potassium have been found to abolish membrane potentials in brain slices (Hillman and McIlwain, 1961; Gibson and McIlwain, 1965), microdissccted neurons (Hillman and Hydkn, 1965a), mammalian glia cell cultures (Hild et al., 1958; Hild, 1964; Wardell, 1966), leech packet glia cells (Nicholls and Kuffler, 1964), and Necturus optic nerve (Kuffler et al., 1966). I n the Zatter two preparations the contained glia cells behave as perfect potassium electrodes (Nicholls and Kuffler, 1964; Kuffler and Nicholls, 1966; Kuffler et al., 1966; Kuffler, 1967; cf. also Dennis and Gerschenfeld, 1969), whereas this is not the case for mammalian glia cells (Pape and Katzman, 1970, 1972; Ransom and Goldring, 1973a). Stimulation of neurons may lead to a depolarization of adjacent glia cells due to release of potassium ions and resulting increase in the potassium concentration of the interstitial fluid (Kuffler and Nicholls, 1966; Orkand et al., 1966; Grossman et al., 1969; Fertziger and Ranck, 1970; Grossman and Rosman, 1971; Dichter et al., 1972; Krnjevic and Morris, 1972; Vyklicky’ et al., 1972; VyskoEil e t al., 1972; Prince et al., 1973; Ransom and Goldring, 1973b). During spreading depression the interstitial K+ concentration may reach as high a level as 80 mM (Lux et al., 1972; VyskoEil et al., 1972) leading to a glial depolarization (Karahashi and Goldring, 1966; Higashida et al., 1974). The glia cell membrane reacts to potential changes without any “all-or none” action potential (Hild et al., 1958; Hild and Tasaki, 1962; Kuffler and Potter, 1964; Kuffler and Nicholls, 1966; Kuffler et al., 1966; Wardell, 1966; cf. also Dennis and Gerschenfeld, 1969), but potential changes may spread between individual glia cells over a distance of 0.2 to more than 1.0 mm (Kuffler and Potter, 1964; Kuffler and Nicholls, 1966; Kuffler et al., 1966; Walker and Hild, 1969). The electrical contacts between individual cells probably occur at specialized “junctions” (Gray, 1961; Peters, 1962; Coggeshall and Fawcett, 1964; Brightman and Reese, 1969). I n species where the glia cells behave as potassium electrodes, the potassium ions released during nervous activity may enter the joined cells passively in one region, traverse the cells and leave in another region, i.e., the glia cells may function as a passive “spatial buffer” which removes potassium ions from the immediate extraneuronal environment (Orkand et al., 1966; Kuffler and Nicholls, 1966; Kuffler, 1967; Orkand, 1971; Trachtenberg and Pollen, 1970; Trachtenberg et al., 1972). Evoked potentials have been observed in brain slice preparations (Yamamot0 and McIlwain, 1966a,b; Richards and McIlwain, 1967; Yamamoto and Kawai, 1967; Richards and Sercombe, 1968; Kawai and Yamamoto, 1969; Yamamoto and Kurokawa, 1970), and different types of cell cultures may offer excellent opportunities for study of neurophysiological events (e.g., Crain, 1972). Using tissue cultures, it has thus been demonstrated that cer-
188
LEIF HERTZ AND ARNE SCHOUSBOE
tain characteristics of spreading depression do not require the presence of neurons (Walker and Hild, 1972). The Schwann cells of the squid nerve fiber have a high sodium concentration (Villegas et al., 1965; Villegas, 1968), and it has repeatedly been suggested that the sodium concentration should be high also in glia cells from the central nervous system (Gerschenfeld et al., 1959; De Robertis and Gerschenfeld, 1961; Katzman, 1961; Koch et al., 1962; Hartmann, 1966; cf. also Franck, 1970). No conclusive evidence has been reported, however, and findings in the leech and in cultivated glia cells (cf. below) do not support this concept. A high concentration of potassium (i.e., 75 m M or more) has been observed both in microdissected glia cells (Hamberger and Rockert, 1964), in cultivated astrocytes (Lees and Shein, 1970; Gill et al., 1974) and in dissociated cells cultivated from chick brain hemispheres (Latzkovits et al., 1974a), but not in cultured human embryonic glia cells (Trachtenberg et al., 1972). Also glia cells obtained by gradient centrifugation show some, though probably not as efficient, accumulation of potassium (Bradford and Rose, 1967; Haljamae and Hamberger, 1971; Hemminki and Holmila, 1971; Nagata et al., 1974). Cultured astrocytoma cells (C6) have a higher capacity for active potassium uptake than C1300 neuroblastoma cells (Kimelberg, 1974) and microdissected neurons show a continuous potassium loss (Hamberger and Rockert, 1964). Bulk-prepared neurons show little (Rose and Sinha, 1969; Haljamae and Hamberger, 1971; Hemminki and Holmila, 1971) or no accumulation of potassium (Bradford and Rose, 1967), but some uptake occurs in synaptosomes (Bradford, 1969; Escueta and Appel, 1969). All evidence thus points toward a greater capacity for potassium uptake in glia cells than in neurons. This difference seems also to be present in media containing high (50-100 mM) concentrations of potassium (Henn et al., 1972). I t is in agreement with this concept that a transfer of potassium from a neuronal to a glial compartment has been observed after exposure to excess potassium (Franck, 1973), and that the potassium-activated uptake of C1- (Bourke, 1969a) has been demonstrated in cultured NN astrocytoma cells (Gill et al., 1973, 1974). As pointed out by Grossman and Rosman ( 1971) , such an inward transport of chloride into glia cells could explain the hyperpolarization of mammalian glia cells, which follows the initial depolarization after neuronal excitation (Glotzner and Grusser, 1968; Sypert and Ward, 1971; Ransom and Goldring, 1 9 7 3 ~ ) . Active transport of K+ in exchange with Na+ has been observed in an astrocytoma cell line (Kukes et al., 1973). Using cultures of dissociated chick embryo brain containing neurons and glia cells in different proportions Latzkovits et al. (1974a,b) have obtained evidence (Fig. 18) for a transport
189
LO
80
120
LO
Time from exposure to radioisotope
80
120
(rnin)
FIG. 18. Accumulation of potassium analog "Rb into pure glial ( A ) and mixed neuronal/glial ( B ) chick embryo cultures. T h e simple exponential course in the glial culture is contrasted by a complex uptake curve in the mixed culture. This curve was studied by tracer kinetic model analysis which suggested that, besides direct uptake from the medium into both neurons and glia cells, a glial/neuronal interrelationship in transport occurs. For further details, see Latzkovits et al. ( 1974a). From Latzkovits et al. (1974a).
of K+ between neurons and glia cells and have suggested that neuronal uptake is maintained and regulated by intact glial function (for details, see legend to Fig. 1 8 ) .
2. Water In order to determine the cellular localization of the swelling occurring in brain slices, histological examination must be resorted to. Conceivably the histological picture may be distorted due to artifacts during fixation, etc., but recently good concordance has been demonstrated between histological and biochemical (use of inulin as extracellular marker) determination of extra- and intracellular swelling (Mdler et al., 1974). After incubation in a physiological medium (Fig 19A) electron micrographs have shown that mainly glia cells and their processes are swollen (Gerschenfeld et al., 1959; De Robertis and Gerschenfeld, 1961; Cohen and Hartmann, 1964; Torack et al., 1965), but some swelling has also been observed in neurons (Allen, 1955; Ibata et al., 1971; Selwood, 1971; Mdler et al., 1974). The increased swelling after addition of excess potassium (Fig 19B) is almost exclusively glial (Zadunaisky et al., 1963; Cohen and Hartmann, 1964; Bourke et al., 1970; Lodin et al., 1971 ; Bourke and Nelson, 1972; Msller et
190
LEIF HERTZ A N D ARNE S C H O U S B O E
FIG. 19. Electron micrographs of rat brain cortex slices incubated for 1 hour in a physiological ( A ) or a potassium-rich (40 m M K') medium ( B ) a t 37°C or in a physiological medium a t O"(C). All x25,OOO. ( A ) T h e extracellular space ( e ) is extended. G indicates swollen glial process. ( B ) Glial process ( G ) shows extensive swelling, and the extracellular space ( e ) is less increased than in panel ( a ) . ( C ) An extremely swollen process can be identified as dendritic (Den) by the synapse (arrowhead). From Mdler et al. (1974).
BRAIN METABOLISM A T T H E CELLULAR LEVEL
191
al., 1974), indicating that the compensatory accumulation of KCl (and water) into brain tissue (cf. Fig. 17) mainly or exclusively occurs into glia cells. Under other conditions, e.g., exposure to low temperature (Fig. 19C), sodium lack, or metabolic inhibitors, it is, in contrast, mainly the neurons that show increased swelling (Ibata et al., 1971; Mdler et al., 1974). The glial localization of the swelling in physiological and potassium-rich media is supported by the finding that brain slices from newborn kittens or rats (which have few glia cells) do not swell (Tower and Bourke, 1966; Franck, 1970; Okamoto and Quastel, 1970; Schousboe and Hertz, 1971a; Schousboe, 1972), and a parallelism between ontogenetic development of swelling and astrocytic maturation has been pointed out by Franck et al. (1970).
V.
Concluding Remarks
The comparison between metabolism in the brain in vivo and in brain slices and isolated cells has shown that ion and energy metabolism of the dissociated cell types to a remarkable extent may correspond to that in more complex in uitro preparations and under in vivo conditions. Observations in such preparations of individual cells are therefore in all probability relevant for interpretation of physiological and pathological events at the cellular level of the brain cortex. I t should, however, also be kept in mind that some of the preparations and less differentiated cell lines have metabolic characteristics different from those found in viuo and that observations on ion and energy metabolism accordingly should be evaluated with care. From a physiological point of view it is probably of major importance that the potassium concentration in the interstitial fluid becomes considerably increased during neuronal activity. Owing to the pronounced influence of K+ on membrane polarization, efficient and fast removal of excess extracellular potassium is essential. The restoration of a low extracellular potassium concentration may conceivably be brought about by diffusion, active uptake into any cell type or the current-carried redistribution proposed for the leech by Kuffler and co-\rorkers' (Fig. 20). Evidence is found that the latter mechanism is of no major importance in the mammalian brain cortex (Glotzner, 1973; Lux and Neher, 1973). Participation of an active mechanism is suggested by the relatively faster potassium removal after more intense stimulation (i.e., higher extracellular K+ concentration) observed by Ransom and Goldring ( 1 9 7 3 ~ )and the overshoot, i.e., lowering below the original level observed by Lux and co-workers (Heinemann and Lux, 1973; Lux and Neher, 1973). I t seems likely that such an active removal of potassium plays a major role among the energy-requiring processes triggering the increase in energy metabolism evoked by high concentrations of potassium. *Kuffler and Nicholls, 1966; Orkand et al., 1966; Kuffler, 1967.
192
LEIF HERTZ AND ARNE S C H O U S B O E
; GLlA
NEURON
K'
I -
<--
--__ > <----
K*
{
K* vs. Na* Active transport K* + Cl' Passive transport
diffusion current- carried
FIG. 20. Schematic representation of possible potassium homeostasis mechanisms at the cellular level. The potassium release from neurons during excitation leads to a high potassium concentration in the narrow extracellular clefts. This may be counteracted by diffusion along the clefts and by passive current-carried K' transport through glia cells as described by Kuffler (e.g., 1967). I n addition, the mammalian brain cortex seems to possess an active transport mechanism leading t o a potassium uptake into glial cells (together with C1- or in exchange with Na') and reflected by potassium effects on ATPases, energy metabolism, water content, and ion content and fluxes. Ultimately the displaced potassium ions must be reaccumulated into neurons. From Hertz (197313).
I t may seem peculiar that very good evidence is found that this cellular uptake of potassium to a large extent occurs in glia cells. This represents, however, an analogy to the current-carried redistribution of potassium ions through glia cells suggested for the leech nervous system and may indicate a further phylogenetic development of the potassium-removing mechanism. Ultimately the displaced potassium ions must of necessity be reaccumulated into neurons. The potassium effects on neuronal energy metabolism observed by some authors may be a metabolic manifestation of this process, and also
BRAIN METABOLISM AT T H E CELLULAR LEVEL
193
neurophysiological evidence for an active uptake of potassium into neurons has been reported (Eccles, 1957; Ransom and Goldring, 1973a). Derangements of mechanisms for cellular reuptake of potassium may conceivably lie behind at least certain forms of epilepsies (Hillman, 1970; Pollen and Trachtenberg, 1970), since exposure of the brain to excess potassium is known to provoke seizure activity (e.g., Feldberg and Sherwood, 1957; Izquierdo et al., 1970; Zuckermann and Glaser, 1970; Glaser, 1972; OConnor and Lewis, 1974). Also the decreased Na+-K+-ATPase activity in mice susceptible to seizures (Hertz et al., 1974) points in this direction. Further work on the energy and ion metabolism of the brain cortex at the cellular level may accordingly provide important information about interactions between different cell types during normal brain activity as well as under pathological conditions. REFERENCES Abadom, P. N., and Scholefield, P. G. (1962a). Can. J. Biochem. Physiol. 40, 1575-1590. Abadom, P. N., and Scholefield, P. G. (1962b). Can. J . Biochem. Physiol. 40, 1603-1 6 18. Abdel-Latif, A. A,, Brody, J., and Ramahi, H. (1967). J. Neurochem. 14, 1133-1141. Abood, L., Cavanaugh, M., Tschirgi, R. D., and Gerard, R. W. (1951). Fed. Proc. Fed. Amer.
[email protected]. 10, 3. Abood, L. G. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 11, pp. 303-326. Plenum, New York. Adey, W. R. (1967). I n “Neurosciences: A Study Program” (G. C. Quarton, T. Melnechuk, and F. 0. Schmitt, eds.), pp. 615-633. Rockefeller Univ. Press, New York. Adey, W. R., Kado, R. T., Didio, J., and Schindler, W. J. (1963). ExP. Neurol. 7, 259-281. Aghajanian, G. K., and Bloom, F. E. (1967). Brain Res. 6, 716-727. Agrawal, H. C., Davis, J. M., and Himwich, W. A. (1966). J. Neurochem. 13, 607-6 15. Ahmed, K., Judah, J. D., and Scholefield, P. G. (1966). Biochim. Biophys. Actu 120, 351-360. Albaum, H. G., Noell, W. K., and Chinn, H. I. (1953). Amer. J . Physiol. 174, 408-41 2. Aleksidze, N., and Blomstrand, C. ( 1968). Brain Res. 11, 7 17-7 19. Aleksidze, N. G., and Blomstrand, C. (1969). Dokl. Akad. Nauk SSSR 186, 1429-1430; Proc. Acud. Sci. USSR, Sect. Biochem. 186, 140-141 (1969). Allen, J. N. (1955). A M A Arch. Neurol. Psychiat. 73, 241-248. Allen, N. (1957). J . Neurochem. 2, 37-44. Allen, N. (1972). In “The Experimental Biology of Brain Tumors” (W. M. Kirsch, E. G. Paoletti, and P. Paoletti, eds.), pp. 243-274, Thomas, Springfield, Illinois. Altman, J. (1963). J . Histochem. Cytochem. 11, 741-750. Altman, J. (1967). I n “Neurosciences: A Study Program” (G. C. Quarton, T. Melnechuk, and F. 0. Schmitt, eds.), pp. 723-743. Rockefeller Univ. Press, New York.
194
LEIF HERTZ AND ARNE SCHOUSBOE
Altman, J. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 11, pp. 137-182. Plenum, New York. Ames, A., and Nesbett, F. B. (1958). J. Neurochem. 3, 116-126. Appel, S. H., Autilio, L., Festoff, B. W., and Escueta, A. V. (1969). J . Biol. Chem. 244, 3166-3172. Arnfred, T., and Hertz, L. (1971). J . Neurochem. 18, 259-265. Arnfred, T., Hertz, L., Lolle, L., and Lund-Andersen, H. (1970). Exp. Brain Res. 11, 3 73-3 75. Ashford, C. A,, and Dixon, K. C. (1935). Biochem. J. 29, 157-168. Athias, P., Sensenbrenner, M., and Mandel, P. (1973). J. Physiol. (Paris) 67, Suppl. 3, 330A. Azcurra, J . M., Lodin, Z., and Sellinger, 0. Z. (1969). Abstr., Znt. M e e t . Znt. SOL. Neurochem., 2nd, 1969 pp. 76-77. Bachelard, H. S., Campbell, W. J., and McIlwain, H. (1962). Biochem. J . 84, 225-232. Bachelard, H. S., Lewis, L. D. Ponten, U., and Siesjo, B. K. (1974). J . Neurochem. 22, 395-401. Bader, H., and Sen, A. K. (1966). Biochim. Biophys. Actu 118, 116-123. Bakay, L. (1965). Progr. Bruin Res. 15, 155-183. Balazs, R. (1970). I n ‘Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 111, pp. 1-36. Plenum, New York. Balazs, R. (1971). Abstr., Znt. Meet. Znt. SOC. Neurochem, 3rd, 1971 p. 408. Balazs, R., and Cremer, J. E., eds. (1973). “Metabolic Compartmentation in the Brain.” Macmillan, New York. Balazs, R., Machiyama, Y., Hammond, B. J., Julian, T., and Richter, D. (1970). Biochem. J. 116, 445-467. Balazs, R., Patel, A. J., Johnson, A. L., and Richter, D. (1971). Abstr., Znt. Meet. Znt. SOC.Neurochem. 3rd., 1971 p. 228. Balazs, R., Patel, A. J., and Richter, D. (1973). In “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 167-184. Macmillan, New York. Banay-Schwartz, M., Piro, I,., and Lajtha, A. (1971). Arch. Biochem. Biophys. 145, 199-210. Baxter, C. F. (1970). In “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 111, pp. 289-353. Plenum, New York. Baxter, C. F. Schadt, J. P., and Roberts, E. (1960). I n “Inhibition in the Nervous System and Gamma-aminobutyric Acid” ( E . Roberts et al., eds.), pp. 214-218. Pergamon, Oxford. Beloff-Chain, A., Catanzaro, R., Chain, E. B., Masi, I., and Pocchiari, F. (1955). Proc. R o y . SOC.,Ser. B 144, 22-28. Benda, P., Lightbody, J., Sato, G., Levine, L., and Sweet, W. (1968). Science 161, 370-371. Benjamin, A. M., and Quastel, J. H. (1972). Biochem. J. 128, 631-646. Bergen, J. R., Hunt, C. A,, and Hoagland, H . (1953). Amer. J . Physiol. 175, 327-332. Berl, S., and Clarke, D. D. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 11, pp. 4 4 7 4 7 2 . Plenum, New York. Bed, S., and Purpura, D. P. (1963). J. Neurochem. 10, 237-240. Berl, S., Lajtha, A., and Waelsch, H. (1961). J. Neurochem. 7, 186-197. Bed, S., Nicklas, W. J., and Clarke, D. D. (1968). J . Neurochem. 15, 131-140. Bed, S., Nicklas, W. J., and Clarke, D. D. (1970). J . Neurochem. 17, 1009-1015.
BRAIN METABOLISM AT T H E CELLULAR LEVEL
195
Biesold, D. (1967). Biochem. J. 102, 20P-21P. Bignami, A., and Dahl, D. (1974). J. Comp. Neurol. 153, 27-38. Bignami, A., and Palladini, G. (1965). A t t i Accad. Naz. Lincei, C1. Sci. Fis., Mat. Natur., Rend. [8] 38, 253-258. Bignami, A,, and Palladini, G. (1966). Nature ( L o n d o n ) 209, 413-414. Bignami, A,, Palladini, G., and Venturini, G. (1966). Brain Res. 3, 207-209. Bilodeau, F., and Elliott, K. A. C. (1963). Can. J. Biochem. Physiol. 41, 779-792. Blackstad, T. W. (1967). I n “The Neuron” (H. Hyden, ed.), pp. 49-118. Elsevier, Amsterdam. Blinkov, S. M., and Glezer, I. I. (1968). “The Human Brain in Figures and Tables,” p. 416. Plenum, New York. Blomstrand, C. ( 197 1 ) . Thesis, University of Goteborg, Sweden. Blomstrand, C., and Hamberger, 4 . (1969). J. Neurochem. 16, 1401-1407. Blomstrand, C., and Hamberger, A. ( 1970). J . Neurochem. 17, 1187-1 195. Bloom, F. R., and Iversen, L. L. (1971). Nature ( L o n d o n ) 229, 628-630. Blume, A,, Gilbert, F., Wilson, S., Farber, J., Rosenberg, R., and Nirenberg, M. (1970). Proc. Nut. Acad. Sci. U S . 67, 786-792. Bollard, B. M., and McIlwain, H. (1957). Biochem. J. 66, 651-655. Bolwig, T . G., and Quistorff, B. (1973). J. Neurochem. 21, 1345-1348. Bondareff, W. (1966). Z . Zellforsch. Mikrosk. Anat. 72, 487-495. Bondareff, W. (1967). Z . Zellforsch. Mikrosk. Anat. 81, 366-373. Booher, J., and Sensenbrenner, M. (1972). Neurobiology 2, 97-105. Booher, J., Hertz, L., and Lodin, Z. (1971a). Neurobiology 1, 27-31. Booher, J., Fosmark, H., and Hertz, L. (1971b). Neurobiology 1, 32-36. Bornstein, M. B. (1973). I n “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 267-283. Macmillan, New York. Bourke, R. S. (1969a). Exp. Brain Res. 8, 219-231. Bourke, R. S. (1969b). Ex$. Brain Res. 8, 232-248. Bourke, R . S., and Nelson, K. M. (1972). J . Neurochem. 19, 663-685. Bourke, R. S., and Tower, D. B. (1966a). J . Neurochem. 13, 1071-1097. Bourke, R. S., and Tower, D. B. (1966b). J. Neurochem. 13, 1099-1117. Bourke, R. S., Greenberg, E. S., and Tower, D. B. (1965). Amer. J . Physiol. 208, 682-692. Bourke, R. S., Nelson, K. M., Naumann, R. A,, and Young, 0. M. (1970). Exp. Brain Res. 10, 427-466. Boyer, P. D., Lardy, H . A., and Phillips, P. H. (1942). J . B i d . Chem. 146, 673-682. Boyer, P. D., Lardy, H. A., and Phillips, P. H . (1943). J. B i d . Chem. 149, 529-541. Boyle, P. J., and Conway, E. J. (1941). J . Physiol. ( L o n d o n ) 100, 1-63. Bradbury, M. W. B., Villamil, M., and Kleeman, C. R. (1968). Amer. J. Physiol. 214, 643-65 1. Bradford, H. F. (1967). Abstr. Commun., I n t . Meet. I n t . SOC.Neurochem., l s t , 1967 p. 30. Bradford, H. F. ( 1969). J . Neurochem. 16, 675-684. Bradford, H. F. (1970). Biochem. 1. 117, 36P. Bradford, H. F., and McIlwain, H. (1966). J. Neurochem. 13, 1163-1 177. Bradford, H. F., and Rose, S. P. R. (1967). J. Neurochem. 14, 373-375. Bradford, H. F., and Thomas, A . J. (1969). J . Neurochem. 16, 1495-1504. Bradford, H. F., Brownlow, E. K., and Gammack, D. B. (1966). J . Neurochem. 13, 1283-1297. Brierley, J. B., and McIlwain, H. (1956). J . Neurochem. 1, 109-118. Brightman, M. W., and Reese, T . S. (1969). J. Cell B i d . 40, 648-677.
196
LEIF HERTZ A N D ARNE S C H O U S B O E
Brinley, F. J. (1967). J . Neurophysiol. 30, 1531-1560. Brinley, F. J., Kandel, E. R., and Marshall, W. H. (1960). J . Neurophysiol. 23, 246-256. Brinley, F. J., Jr. (1963). Znt. Reu. Neurobiol. 5, 183-242. Brizzee, K. R., and Jacobs, L. A. (1959a). Anat. Rec. 134, 97-105. Brizzee, K. R., and Jacobs, L. A. (1959b). Growth 23, 337-347. Brizzee, K. R., Vogt, J., and Kharetchko, X. (1964). Progr. Brain Res. 4, 136-149. Brossard, M., and Quastel, J. H. (1963). Can. J . Biochem. Physiol. 41, 1243-1256. BureS, J. (1957). Electroencephalogr. Clin. Neurophysiol. 9, 121-130. BureS, J., BureSova, O., and Kfivlnek, J. (1960). I n “Structure and Function of the Cerebral Cortex” (D. B. Tower and J. P. Schadk, eds.), pp. 257-265. Elsevier, Amsterdam. Burel, J., Buregova, O., and Kfivlnek, J. (1974). “The Mechanism and Applications of Leao’s Spreading Depression of Electroencephalographic Activity.” Academic Press, New York. BureS, J., Fifkovl, E., and MareS, P. (1964). I n “Neurological and Electroencephalographic Correlative Studies in Infancy” (P. Kellaway and I. Petersen, eds.), pp. 27-36. Caley, D. W., and Maxwell, D. S. (1970). J . Comp. Neurol. 138, 31-48. Canzanelli, A., Rogers, G., and Rapport, D. (1942). Amer. J . Physiol. 135, 309-315. Chance, B., and Williams, G. R. (1955). J . B i d . Chem. 217,409-427. Chance, B., Cohen, P., Jobsis, F., and Schoener, B. (1962). Science 137, 499-508. Chang, T. H., and Tai, F. I. (1936). Contrib. Biol. Lab. Sci. SOC. China, Zool. Ser. 11, 239-266. Chang, T. H., Shaffer, M., and Gerard, R. W. (1935). Amer. J . Physiol. 111, 681-696. Charnock, J. S. (1963). 1. Neurochem. 10, 219-223. Cheng, S . 4 . (1973). In “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 107-1 18. Macmillan, New York. Ciesielski-Treska, J., Mandel, P., Tholey, G., and Wurtz, B. (1972). Nature (Lond o n ) , New B i d . 239, 180-181. Clark, J. B. (1970). Fed. Proc., Fed. Amer. SOC.Exp. Biol. 129, 471. Clark, J. B., and Nicklas, W. J. (1970). J. B i d . Chem. 245, 4724-4731. Clarke, D. D., and Bed, S. (1973). In “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 97-106. Macmillan, New York. Coggeshall, R. E., and Fawcett, D. W. (1964). J. Neurophysiol. 27, 229-289. Cohen, L. H., and Noell, W. K. (1960). J . Neurochem. 5,253-276. Cohen, M. M., and Hartmann, J. F. (1964). I n “Morphological and Biochemical Correlates of Neural Activity” (M. M. Cohen and R. S. Snider, eds.), pp. 57-74. Harper, New York. Cohen, S. R. (1972). Res. Methods Neurochem. 1, 179-219. Cohen, S. R., and Lajtha, A. (1969). Fed. Proc., Fed. Amer. SOC. Exp. B i d . 28, 843. Cohen, S. R., and Lajtha, A. (1970). Brain Res. 23, 77-93. Cohen, S . R., and Lajtha, A. (1971). Znt. J . Neurosci. 1, 251-258. Cohen, S. R., Blasberg, R., Levi, G., and Lajtha, A. (1968). J. Neurochem. 15, 707-720. Cornog, J. L., Jr., Gonatas, N. K., and Feierman, J. R. (1967). Amer. J. Pathol. 51, 573-590. Crain, S . M. (1972). In “Experimental Models of Epilepsy” (D. P. Purpura et al., edr.), pp. 291-316. Raven Press, New York.
BRAIN METABOLISM AT THE CELLULAR LEVEL
197
Cremer, J. E. (1967). Biochem. J . 104, 212-222. Cremer, J. E., Johnston, P. V., Roots, B. I., and Trevor, A. J. (1968). 1.Neurochem. 15, 1361-1370. Cummins, J. T., and Bull, R. (1971). Biochem. Biophys. Acta 253, 29-38. Cummins, J. T., and HydCn, H. (1962). Biochim. Biophys. Acta 60, 271-283. Cummins, J. T., and McIlwain, H. (1961). Biochem. J . 79, 330-341. Davies, M. (1961). J . Cell. Comp. Physiol. 57, 135-147. Davson, H. ( 1967). “Physiology of the Cerebrospinal Fluid.” Churchill, London. Davson, H., and Spaziani, E. (1959). J . Physiol. ( L o n d o n ) 149, 135-143. De Belleroche, J. S., and Bradford, H. F. (1972). I. Neurochem. 19, 585-602. Dennis, M. J., and Gerschenfeld, H. M. (1969). J . Physiol. ( L o n d o n ) 203, 211-222. De Piras, M., and Zadunaisky, J. A. (1965). J . Neurochem. 12, 657-661. De Robertis, E., and Gerschenfeld, H. M. (1961 ) . Znt. Rev. NeurobioZ. 3 , 1-65. De Vellis, J., and Kukes, G. (1973). T e x . R e p . B i d . M e d . 31, 271-293. Diamond, I., and Fishman, R. A. (1973). Nature ( L o n d o n ) 243, 519-520. Diamond, M. C., Law, F., Rhodes, H., Lindner, B., Rosenzweig, M. R., Krech, D., and Bennett, E. L. (1966). J . Comp. Neurol. 128, 117-125. Dichter, M. A., Herman, C. J., and Selzer, M. (1972). Brain Res. 48, 173-183. Dickens, F., and Greville, G. D. (1935). Biochem. J . 29, 1468-1483. Dickens, F., and Simer, F. (1931). Biochem. J . 25, 985-993. Dittmann, L., Hertz, L., Fosmark, H., Sensenbrenner, M., and Mandel, P. (1973a). Biochem. SOC.Trans. 1, 136-138. Dittmann, L., Hertz, L., Schousboe, A,, Fosmark, H., Sensenbrenner, M., and Mandel, P. (197313). Exp. Cell Res. 80, 425-431. . Neurochem. Dittmann, L., Sensenbrenner, M., Hertz, L., and Mandel, P. ( 1 9 7 3 ~ ) 1. 21, 191-198. Dixon, K., and Holmes, E. (1935). Nature ( L o n d o n ) 135,995-996. Dixon, K. C. ( 1949). J . Physiol. ( L o n d o n ) 110,87-97. Dixon, T. F., and Meyer, A. (1936). Biochem. J . 30, 1577-1582. Dobbing, J. (1968). Progr. Brain Res. 29, 417-425. Droz, B., and Leblond, C. P. (1963). J . C o m p . Neurol. 121, 325-337. Dzubow, L. M., and Garfinkel, D. (1970). Brain Res. 2 3 , 4 0 7 4 1 7 . Eayrs, J. T., and Goodhead, B. (1959). J . Anat. 93,385-402. Eccles, J. C. (1957). “The Physiology of Nerve Cells.” Johns Hopkins Press, Baltimore, Maryland. Eichberg, J., Hauser, G., and Shein, H. M. (1971). Biochem. Biophys. Res. Commun. 45, 43-50. Elliott, K. A. C. (1952). I n “The Biology of Mental Health and Disease”, pp. 54-70. Hoeber, New York. Elliott, K. A. C. (1955). Can. J . Physiol. 33, 4 6 6 4 8 0 . Elliott, K. A. C. ( 1957). In “Physiological Problems of the Central Nervous System.” (A. F. Solowiew et al., eds.), pp. 610-616. Publ. Sci. Acad. USSR, Moscow. Elliott, K. A. C., and Bilodeau, F. (1962). Biochem. J. 84, 421-428. Elliott, K. A. C., and Heller, I. H. (1957). I n “Metabolism of the Nervous System” (D. Richter, ed.), pp. 286-290. Pergamon, Oxford. Elliott, K. A. C., and Libet, B. (1942). J . B i d . Chem. 143, 227-246. Elliott, K. A. C . , and Henderson, N. (1948). J. Neurophysiol. 11, 471-484. Elliott, K. A. C . , and Penfield, W. (1948). J . Neurophysiol. 11, 485-490. Elliott, K. A. C., Greig, M. E., and Benoy, M. P. (1937). Biochem. J . 31, 1003-1020. Embree, L. J., Hess, H. H., and Shein, H. M. (1971a). Brain Res. 27, 422-425. Embree, L. J., Hess, H. H., and Shein, H. M. (1971b). Exp. Neurol. 31, 383-390.
198
LEIF HERTZ A N D A R N E S C H O U S R O E
Eng, L. F., Vanderhaeghen, J. J., Bignami, A., and Gerstl, B. (1971). Brain Rex. 28, 351-354. Epstein, M. H., and O’Connor, J. S. (1965). J. Neurochem. 12, 389-395. Esmeta, A. V., and Appel, S. H. (1969). Biochemistry 8,725-733. Evans, H. J., Clark, R. B., and Russell, S. A. (1964). Biochim. Biophys. Acta 92, 582-594. Farquhar, M. G., and Hartmann, S . F. (1957). J. Neuropathol. Exp. Neurol. 16, 18-39. Feldberg, W., and Sherwood, S. L. (1957). J . Physiol. ( L o n d o n ) 139, 408-416. Fenstermacher, J. D., Li, C.-L., and Levin, V. A. (1970). Exp. Neurol. 27, 101-1 14. Ferguson, R. K., and Woodbury, D. M. (1969). Exp. Brain Res. 7, 181-194. Ferrendelli, J. A. (1975). In “The Working Brain” (D. H. Ingvar, N. A. Lassen, and J. Hess Thaysen, eds.), The Alfred Benzon Symp. VII. Munksgaard, Copenhagen (in press). Ferrendelli, J. A,, Gay, M. H., Sedgwick, W. G., and Chang, M. M. (1972). J . Neurochem. 19, 979-987. Fertziger, A. P., and Ranck, J. B., Jr. (1970). Exp. Neurol. 26, 571-585. Fish, J., and Winick, M. (1968). Neurology 18, 292. Flexner, L. B., and Flexner, J. B. (1949). J. Cell. Comp. Physiol. 34, 115-127. FolbergrovP, J., Passonneau, J. V., Lowry, 0. H., and Schultz, D. W. (1969). J. Neurochem. 16, 191-203. Fonnum, F., and Walberg, F. (1973a). Brain Res. 54, 115-127. Fonnum, F., and Walberg, F. (197Bb). Brain Res. 62,577-579. Fonyo, A,, Kovach, A. G. B., Maklary, E., Leszkovzky, G., and Meszaros, J. (1958). Acta Physiol. 14, 305-307. Franck, G. (1970). Thesis, UniversitC de LiPge. Franck, G. (1973). Abstr., Znt. Meet. I n t . SOL.Neurochem., 4th, 1973 pp. 117-118. Franck, G., and Cornette, M. (1966). Rev. Neurol. 115, 312-314. Franck, G., and Schoffeniels, E. (1972). J. Neurochem. 19,395-402. Franck, G., Cornette, M., and Schoffeniels, E. (1968). J . Neurochem. 15,843-857. Franck, G., Schoffeniels, E., and Gerebzoff, M. A. (1970). Acta Neurol. Belg. 70, 424-438. Friede, R. (1954). Acta Anat. 20, 290-296. Friede, R. L. (1964). J . Cell B i d . 20, 5-15. Friede, R. L. (1966). “Topographic Brain Chemistry,” pp. 170-177. Academic Press, New York. Friede, R. L. (1970). Triangle 9, 165-173. Friede, R. L., and Van Hauten, W. H. (1962). Proc. Nut. Acad. Sci. US. 48, 817-82 1. Futamachi, K., Mutani, R., and Prince, D. A. (1974). Brain Res. 75, 5-25. Gaballah, S., and Popoff, C. (1971). Brain Res. 25, 220-222. Gainer, H., Allweis, C. L., and Chaikoff, I. L. (1962). J . Neurochem. 9, 433-442. Gaitonde, M. K. (1965). Biochem. J . 95, 803-810. Gaitonde, M. K., and Richter, D. (1966). J . Neurochem. 13, 1309-1318. Galindo, A., Krnjevic, K., and Schwartz, S. (1967). J. Physiol. ( L o n d o n ) 192, 359-3 7 7. Garfinkel, D. (1962). J. Theor. Biol. 3 , 4 1 2 4 2 2 . Garfinkel, D. (1966). J . B i d . Chem. 241, 3918-3929. Garfinkel, D. (1973). In “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 129-136. Macmillan, New York.
BRAIN METABOLISM A T T H E CELLULAR LEVEL
199
Garrahan, P. J., and Glynn, I. M. (1967). J. Physiol. ( L o n d o n ) 192, 217-235. Geiger, A,, and Magnes, J. (1947). Amer. J. Physiol. 149, 517-537. Gerschenfeld, H. M., Wald, F., Zadunaisky, J. A., and De Robertis, E. D. P. (1959). Neurology 9, 412-425. Ghosh, J., and Quastel, J. H. (1954). h ature ( L o n d o n ) 174, 28-31. Giacobini, E. (1961). Science 134, 1524-1525. Giacobini, E. (1962). J. Neurochem. 9, 169-177. Giacobini, E. ( 1964). I n “Morphological and Biochemical Correlates of Neural Activity” (M. M. Cohen and R. S. Snider, eds.), pp. 15-38. Harper, New York. Gibbs, E. L., Lennox, W. G., Nims, L. F., and Gibbs, F. A. (1942). J. B i d . Chem. 144, 325-332. Gibbs, R., Roddy, P. M., and Titus, E. (1965). J. Biol. Chem. 240, 2181-2187. Gibson, J. M., and McIlwain, H. (1965). J. Physiol. ( L o n d o n ) 176, 261-283. Gill, T. H., Young, 0. M., and Tower, D. B. (1973). Trans. Amer. SOC.Neurochem. 4, 139. Gill, T. H., Young, 0. M., and Tower, D. B. (1974). J . A eurochern. 23, 1011-1018. Glaser, G. H. (1972). In “Experimental Models of Epilepsy” (D. P. Purpura et al., eds.), pp. 315-345. Raven Press, New York. Glotzner, F. L. (1973). Brain Res. 55, 159-171. Glotzner, F., and Griisser, 0. J. (1968). Arch. Psychiat. Nervenkr. 210, 313-339. Glynn, I. M. (1962). J. Physiol. ( L o n d o n ) 160, 18P-19P. Goldman, D. E. (1943). J. Gen. Physiol. 27, 37-60. Gonda, O., and Quastel, J. H. (1962). Nature ( L o n d o n ) 193, 138-140. Gonda, O., and Quastel, J. H. (1966). Biochem. J. 100, 83-94. Gore, M. B. R., and McIlwain, H. (1952). J. Physiol. ( L o n d o n ) 117,471-483. Grafstein, B. (1956). J. Neurophysiol. 19, 154-171. Granholm, L., Lukjanova, L., and Siesjo, B. K. (1969). Acta Physiol. Scand. 77, 179-190. Gray, E. G. (1961). I n “Electron Microscopy in Anatomy” (J. D. Boyd, F. Johnson, and J. D. Lever, eds.), pp. 54-73. Arnold, London. Greengard, P., and McIlwain, H. ( 1955). In “Biochemistry of the Developing Nervous System” (H. Waelsch, ed.), pp. 251-256. Academic Press, New York. Grenell, R. G. (1959). Proc. Int. Congr. Biochem., 4th, 1958 Vol. 3, pp. 115-123. Grossman, R. G., and Hampton, T . (1968). Brain Res. 11, 316-324. Grossman, R. G., and Rosman, L. J. (1971). Brain Res. 28, 181-201. Grossman, R. G., Lynch, L., and Shires, G. T. (1968). Neurology 18, 292. Grossman, R. G., Whiteside, L., and Hampton, T. L. (1969). Brain Res. 14, 401-415. Gurdjian, E. S . , Webster, J. E., and Stone, W. E. (1947). Res. Publ., Ass. Res. Neru. Ment. Dis. 26, 184-204. Hagemeijer, F., and Van Remoortere, P. (1969). J. Theor. B i d . 25, 236-254. Haljamae, H., and Hamberger, A. (1971). J. Neurochem. 18, 1903-1912. Hamberger, A. (1961). J. Neurochem. 8, 31-35. Hamberger, A. (1963). Acta Physiol. Scand. 58, Suppl. 203, 1-52. Hamberger, A. (1971). Brain Res. 31, 169-178. Hamberger, A,, and HydCn, H. (1963). J. Cell B i d . 16, 521-525. Hamberger, A,, and Rockert, H. (1964). J. Neurochem. 11, 757-760. Hamberger, L. (1968). Acta Physiol. Scand. 74, 91-95. Hammond, B. Y., Julian, T., Machiyama, Y., and Balazs, R. (1970). Biochem. J. 116, 461-467.
200
L E I F H E R T Z AND A R N E SCHOUSBOE
Harris, E. J. ( 1960). “Transport and Accumulation in Biological Systems,” 2nd ed. Butterworth, London. Harris, E. J., and Burn, G. P. (1949). Trans. Faraday Soc. 45, 508-528. Harris, E. J., and McLennan, H. (1953). J . Physiol. ( L o n d o n ) 121, 629-637. Hartman, J., Hertz, L., Fosmark, H., and Lodin, 2. (1970). Physiol. Bohemoslou. 19, 381-383. Hartmann, J. F. (1966). Arch. Neurol. (Chicago) 15, 633-643. Harvey, J. A., and McIlwain, H. (1968). Biochem. J . 108, 269-274. Haug, H. (1956). J . Comp. Neurol. 104,473-492. Haug, H. (1960). Zn “Structure and Function of the Cerebral Cortex” (D. B. Tower and J. P. Schade, eds.), pp. 28-35. Elsevier, Amsterdam. Hawkins, A,, and Olszewski, J. (1957). Science 126, 76-77. Heald, P. J. (1954). Biochem. J . 57, 673-679. Heald, P. J. ( 1960). “Phosphorus Metabolism of Brain.” Pergamon, Oxford. Heinemann, U., and Lux, H. D. (1973). Electroencephulogr. Clin. Neurophysiol. 34, 735. Heller, I. H., and Elliott, K. A. C. (1955). Can. J . Biochem. Physiol. 33, 395-403. Hemminki, K., and Holmila, E. (1971). Acta Physiol. Scand. 82, 135-142. Henn, F. A,, and Hamberger, A. (1971). Proc. Nut. Acad. Sci. U S . 68, 2686-2690. Henn, F. A,, Haljamae, H., and Hamberger, A. (1972). Bruin Res. 43, 437-443. Henn, F. A., Goldstein, M. W., and Hamberger, A. (1973). Trans. Amer. SOC. Neurochem. 4, 135. Herschman, H. R., I.evine, I,., and De Vellis, J. (1971). J . Neurochem. 18, 629-633. Herschman, H. R., Grading, B. P., and Lerner, M. P. (1973). Zn “Tissue Culture of the Nervous System” (G. Sato, ed.), pp. 187-202. Plenum, New York. Hertz, L. (1965). Nature ( L o n d o n ) 206, 1091-1094. Hertz, L. (1966). J . Neurochem. 13, 1373-1387. Hertz, I,. (1968). 1.Neurochem. 15, 1-16. Hertz, L. (1969). Zn “The Biological Basis of Medicine” (E. E. Bittar, ed.), Vol. 5, pp. 3-37. Academic Press, New York. Hertz, L. (1973a). “Ion Effects on Metabolism in the Adult Mammalian Brain in Vitro.” FADL’s Forlag, Copenhagen. Hertz, L. (1973b). Abstr., Znt. Meet. Znt. SOC.Neurochem., 4th, 1973 pp. 113-114. Hertz, L., and Arnfred, T. (1967). Abstr., Znt. M e e t . Znt. Soc. Neurochem, lst, 1967 p. 95. Hertz, L., and Clausen, T. (1963). Biochem. J . 89, 526-533. Hertz, L., and Schou, M. (1962). Biochem. J . 85, 93-104. Hertz, L., Schousboe, A., and Weiss, G . (1970). Acta Physiol. Scand. 79, 506-515. Hertz, L., Dittmann, L., Sensenbrenner, M., and Mandel, P. (1973). Trans. Amer. SOC.Neurochem. 4, 142. Hertz, L., Schousboe, A,, Formby, B., and Lennox-Buchtal, M. (1974). Epilepsia 15, 619-631. Hess, H. H. (1961). In “Regional Neurochemistry” (S. S. Kety and J. Elkes, eds.), pp. 200-212. Pergamon, Oxford. Hess, H. H . (1962). J . Neurochem. 9, 613-621. Hess, H. H., Schneider, G . , Warnock, M., and Pope, A. (1963). Fed. Proc., Fed. Amer. SOC.Exp. Biol. 22, 333. Hess, H. H., Embree, L. J., and Shein, H. M. (1972). Progr. Exp. Tumor Res. 17, 308-317. Higashida, H., Mitarai, G . , and Watanabe, S. (1974). Bruin Res. 65, 41 1-425.
BRAIN METABOLISM AT THE CELLULAR LEVEL
20 1
Hild, W. (1964). I n “Brain Function” Vol. 11: RNA and Brain Function. Memory and Learning (M. A. B. Brazier, ed.), pp. 109-134. Univ. of California Press, Berkeley. Hild, W., and Tasaki, I. (1962). J . Neurophysiol. 25, 277-304. Hild, W., Chang, J. J., and Tasaki, I. (1958). Experientia 14, 220-221. Hillman, H. H. (1961). J . Neurochem. 8,257-261. Hillman, H. H. (1964). I n “Comparative Neurochemistry” (D. Richter, ed.), pp. 249-260. Pergamon, Oxford. Hillman, H. H. (1970). Lancet 2, 23-24. Hillman, H. H., and Hyden, H. (1965a). J . Physiol. ( L o n d o n ) 177, 398-410. Hillman, H. H., and Hydtn, H. ( 1965b). Histochemie 4, ,446-450. Hillman, H. H., and McIlwain, H. (1961). J . Physiol. ( L o n d o n ) 157, 263-278. Hillman, H. H., Campbell, W. J., and McIlwain, H. (1963). J . Neurochem. 10, 325339. Himwich, H. E., Bernstein, A. O., Fazekas, J. F., Herrlich, H. C., and Rich, E. (1942). Amer. J . Physiol. 137, 327-330. Hinke, J. A. M. (1961). J . Physiol. ( L o n d o n ) 156, 314-335. Hodgkin, A. L., and Katz, B. (1949). J . Physiol. ( L o n d o n ) 108, 37-77. Hokfelt, T., and Ljungdahl, A. (1971). Abstr. Meet. Nord. Soc. Cell Res., 7 t h , 1971 p. 11. Holmes, E. G. (1932). Biochem. J. 26, 2005-2009. Horstmann, E., and Meves, H. (1959). Z . Zellforsch. Mikrosk. Anat. 49, 569-604. Hosie, R. J. A. (1965). Biochem. J. 96,404-412. Hoskin, F. C. G . (1960). Arch. Biochern. Biophys. 91, 43-46. Hotta, S. S. (1962). J . Neurochem. 9, 43-51. Howe, H. A., and Mellors, R. C. ( 1945). 1. Exp. Med. 81,489-500. Hultborn, R., and Hydtn, H. (1974). Exp. Cell Res. 87, 346-350. Hutchison, H. T., Werrbach, K., Vance, C., and Haber, B. (1973). Trans. Amer. Soc. Neurochem. 4, 136. Hutchison, H. T., Werrbach, K., Vance, C., and Haber, B. (1974). Brain Res. 66, 265-274. Huttunen, M. 0. (1969). Thesis, University of Helsinki, Finland. Huxley, A. F. (1960). I n “Mineral Metabolism” (C. L. Comar and F. Bronner, eds.), Vol. 1, Part 1 A, Appendix 2, pp. 163-167. Academic Press, New York. HydCn, H. (1959). Nature ( L o n d o n ) 184,433-435. HydCn, H. (1960). I n “The Cell” (J. Brachet and A. E. Mirsky, eds.), Vol. 4, pp. 215-323. Academic Press, New York. Hydkn, H. (1967a). I n “The Neuron” ( H . Hydbn, ed.), pp. 179-219. Elsevier, Amsterdam. HydCn, H. (1967b). I n “The Neurosciences: A Study Program” (G. C. Quarton, T . Melnechuk, and F. 0. Schmitt, eds.), Vol. 1, pp. 248-266. Rockefeller Univ. Press, New York. HydCn, H. ( 1 9 6 7 ~ ) I. n “The Neurosciences: A Study Program” (G. C. Quarton, T. Melnechuk, and F. 0. Schmitt, eds.), Vol. 1, pp. 765-775. Rockefeller Univ. Press, New York. Hyden, H., and Lange, P. W. (1961). I n “Regional Neurochemistry” ( S . S. Kety and J. Elkes, eds.), pp. 190-199. Pergamon, Oxford. HydCn, H., and Lange, P. W. (1965). Acta Physiol. Scand. 64, 6-14. Hydkn, H., and McEwen, B. (1966). Proc. Nut. Acad. Sci. U S . 55,354-358. HydCn, H., and Pigon, A. (1960). J . Neurochem. 6,57-72.
202
LEIF HERTZ AND ARNE SCHOUSBOE
Ibata, Y., Piccoli, F., Pappas, G . D., and Lajtha, A. (1971). Brain Res. 30, 137-158. Israel, Y., and Titus, E. (1967). Biochim. Biophys. Acta. 139, 450459. Israel, Y., Kalant, H., and Le Blanc, A. E. (1966). Biochem. J . 100, 27-33. Ito, K. (1960). Jap. J. Exp. M e d . 30, 261-277. Izquierdo, I., Nasello, A. G., and Marichich, E. S. (1970). Arch. Znt. Pharmacodyn. Ther. 187, 318-328. Joanny, P., and Hillman, H. H. (1963). J . Neurochem. 10, 655-664. Joanny, P., and Hillman, H. (1964). J . Neurochem. 11, 413422. Joanny, P., Corriol, J., and Hillman, H. H. (1963). J . Physiol. (Paris) 55, 154-155. Johnson, D. E., and Sellinger, 0. Z. (1971). J . Neurochem. 18, 1445-1460. Johnston, P. V., and Roots, B. I. (1965). Nature ( L o n d o n ) 205, 778-780. Johnston, P. V., and Roots, B. I. (1972). “Nerve Membranes.” Pergamon, Oxford. Jones, C. T., and Banks, P. (1970a). Biochem. J. 118, 791-800. Jones, C. T., and Banks, P. (1970b). Biochem. J. 118, 801-812. Kachmar, J. F., and Boyer, P. D. (1953). J . Biol. Chem. 200,669-682. Karahashi, Y., and Goldring, S. D. ( 1966). Electroencephalogr. Clin. Neurophysiol. 20, 600-607. Kato, J. (1970). Kyoto Furitsu Zka Daigaku Zasshi 79, 923-940; Chem. Abstr. 74, 275 (1971). Kato, T., and Lowry, 0. H. (1973). J. Neurochem. 20, 151-163. Katz, R. I., Chase, T. N., and Kopin, I. J. (1969). J . Neurochem. 16, 961-967. Katzman, R. (1961). Neurology 11, 27-36. Katzman, R., and Pappius, H. M. (1973). “Brain Electrolytes and Fluid Metabolism.” Williams & Wilkins, Baltimore, Maryland. Kawai, N., and Yamamoto, C. (1969). Int. Meet. I n t . SOC.Neurochem., 2nd, 1969 pp. 238-239. Keesey, J. C., and Wallgren, H. (1965). Biochem. J . 95, 301-310. Keesey, J. C., Wallgren, H., and McIlwain, H. (1965). Biochem. J. 95, 289-300. Kelly, J. S., Iversen, L. L., Minchin, M., and Schon, F. (1973). Abstr., Znt. M e e t . Znt. Soc. Neurochem. 4th, 1973 p. 27. Kennedy, C., Des Rosiers, M., Patlak, C. S., Pettigreno, K. D., Reivich, M., and Sokoloff, L. (1974). Trans. Amer. Soc. Neurochem. 5, 86. Kerpel-Fronius, S., and Hajos, F. (1971). Neurobiology 1, 17-26. Kety, S. S. (1957). Zn “Metabolism of the Nervous System” (D. Richter, ed.), pp. 22 1-237. Pergamon, Oxford. Kety, S. S : (1965). Nature ( L o n d o n ) 208, 1252-1257. Kimelberg, H. K. (1974). 1.Neurochem. 22, 971-976. Kimelberg, H. K., and Bourke, R. S. (1973). J . Neurochem. 20, 347-359. Kimura, Y. (1937). Sci. Pap. Znst. Phys. Chem. Res., Tokyo 33, 231-245. Kimura, Y. (1940). Sci. Pap. Inst. Phys. Chem. Res., Tokyo 37, 488-518. Kimura, Y., and Niwa, T. (1953). Nature ( L o n d o n ) 171, 881-882. Kini, M. M., and Quastel, J. H. (1959). Nature ( L o n d o n ) 184, 252-256. Kini, M. M., and Quastel, J. H. (1960). Science 131, 412-414. Klee, C. B., and Sokoloff, L. (1967). J . Biol. Chem. 242, 3880-3883. Klein, J. R., and Olsen, S. N. ( 1947). J . Riol. Chem. 167, 747-756. Koch, A., Ranck, J. B., and Newman, B. 1,. (1962). E x p . Neurol. 6, 186-200. Koelle, G. B. (1962). J. Pharm. Pharmacol. 14, 65-90. Koelle, W. A., and Koelle, G. B. (1959). J . Pharmacol. Exp. Ther. 126, 1-8. Korey, S. R., and Orchen, M. (1959). J. Neurochem. 3, 277-285. Kozawa, S. (1961). Biochem. Pharmacol. 8, 41.
BRAIN METABOLISM AT T H E CELLULAR LEVEL
203
Kozawa, S., and Naito, K. (1966). J Q ~J.. Pharmacol. 16, 451-477. Krall, A. R., Wagner, M. C., and Gozansky, D. M. (1964). Biochem. Biophys. Res. Commun. 16, 77-81. Kratzing, C. C. (1953). Biochem. J. 54, 312-317. Krebs, H. A. (1950). Biochim. Biophys. Acta 4, 249-269. Krebs, H. A., Eggleston, L. V., and Terner, C. (1951). Biochem. J. 48, 530-537. Kiivlnek, J., and BureS, J. (1960). Physiol. Bohemoslou. 9, 494-503. Krnjevic, K. (1965). PTOC.Znt. Congr. Physiol. Sci., 23rd, 1965 Vol. 4, pp. 435-443. Krnjevic, K., and Morris, M. E. (1972). Can. J. Physiol. Pharmacol. 50, 1214-1217. Krnjevic, K., and Schwartz, S. (1967a). Exp. Brain Res. 3, 306-319. Krnjevic, K., and Schwartz, S. (1967b). Exp. Brain Res. 3, 320-336. Krnjevic, K., and Whittaker, V. P. (1965). J. Physiol. ( L o n d o n ) 179, 298-322. Kuffler, S. W. (1967). Proc. Roy. Soc., Ser. B 168, 1-21. Kuffler, S. W., and Nicholls, J. G. (1966). Ergeb. Physiol., Biol. Chem. Exp. Pharmakol. 57, 1-90. Kuffler, S. W., and Potter, D. D. (1964). J . Neurophysiol. 27, 290-320. Kuffler, S. W., Nicholls, J. G., and Orkand, R. K. (1966). J. Neurophysiol. 29, 768-787. Kukes, G., De Vellis, J., and Elul, R. (1973). Trans. Amer. SOC. Neurochem. 4, 141. Kurokawa, M., Sakamoto, T., and Kato, M. (1965). Biochem. J . 97, 833-844. Lajtha, A., ed. (1970). “Protein Metabolism of the Nervous System.” Plenum, New York. Lajtha, A. (1974). In “Aromatic Amino Acids in the Brain” (G. E. W. Wolstenholme and D. W. Fitzsimons, eds.), pp. 25-41. Elsevier, Amsterdam. Lajtha, A,, and Piccoli, F. (1971). I n “Cellular Aspects of Neural Growth and Differentiation” (D. C. Pease, ed.), pp. 419-446. Univ. of California Press, Berkeley. Lajtha, A,, Berl, S., and Waelsch, H. (1959). J. Neurochem. 3, 322-332. Larrabee, M. G. (1958). J. Neurochem. 2, 81-101. Latzkovits, L., Sensenbrenner, M., and Mandel, P. ( 1974a). J . Neurochem. 23, 193-200. Latzkovits, L., Sensenbrenner, M., and Mandel, P. (197413). J. Neurochem. (in press). Lawson, S. N., and Biscoe, T . J. (1973). J . Cell. PhyJiol. 82, 285-298. LeHo, A. A. P. (1944). J . Neurophysiol. 7, 359-390. Lees, M. B., and Shein, H. M. (1970). Brain Res. 23, 280-283. Lehrer, G. M., and Bornstein, M. B. ( 1968). Proc. Amer. h’eurol. Ass. 93, 174-175. Lehrer, G. M., Bornstein, M. B., Weiss, C., Furman, M., and Lichtman, C. (1970). Exp. Neurol. 27, 410-425. Levi, H., and Ussing, H. H. (1949). Acta Physiol. Scand. 16, 232-249. Levy, W. A,, Herzog, I., Suzuki, K., Katzman, R., and Scheinberg, L. (1965). J . Cell Biol. 27, 119-132. Lewin, E., and Hess, H. H. (1964). J. Neurochem. 11, 473481. Lewis, D. V., O’Connor, M. J., and Schuette, W. H. (1974). Electroencephalogr. Clin. Neurophysiol. 36, 347-356. Li, C., and McIlwain, H. (1957). J. Physiol. ( L o n d o n ) 139, 178-190. Lipsett, M. N., and Crescitelli, F. (1950). Arch. Biochem. 28, 329-337. Lipton, P. (1973). J. Physiol. ( L o n d o n ) 231, 365-383. Lisy, V., and Lodin, 2. (1973). Neurobiology 3, 320-326.
204
LEIF HERTZ A N D ARNE S C H O U S B O E
Lisy, V., Kovaru, H., Faltin, J., and Lodin, Z. (1971). Physiol. Bohemoslov. 20, 229-234. Locker, A., and Kaps, R. M. (1960). Z. Gesammte E x p . M e d . 134, 29-39. Lodin, Z., and Rose, S. P. R., eds. (1968). “Macromolecules and the Function of the Neuron.” Excerpta Med. Found., Amsterdam. Lodin, Z., Hartman, J., Kage, M. P., Korinkova, P., and Booher, J. (1971). Neurobiology 1, 69-85. Logan, W. J., and Snyder, S. H. (1971). Nature ( L o n d o n ) 234, 297-299. Lolley, R. N. (1963). J . Neurochem. 10, 665-676. Lowry, 0. H. (1953.) J . Histochem. Cytochem. 1,420-428. Lowry, 0. H. (1957). I n “Metabolism of the Nervous System” (D. Richter, ed.), pp. 323-328. Pergamon, Oxford. Lowry, 0. H., Roberts, N. R., Leiner, K. Y., Wu, M.-L., Farr, A. L., and Albers, R. W. (1954). J. B i d . Chem. 207, 39-49. Lund-Andersen, H. (1974). Brain Res. 65, 239-254. Lund-Andersen, H., and Hertz, L. (1970). Exp. Brain Res. 11, 199-212. Lund-Andersen, H., and Hertz, L. (1973). Biochem. SOL. Trans. 1, 123-126. Lust, W. D., Schwartz, J. P., Shirasawa, R., and Passonneau, J. V. (1974). Trans. Amer. SOL.Neurochem. 5, 86. Lux, H. D., and Neher, E. (1973). E x p . Brain Res. 17, 190-205. Lux, H. D., and Pollen, D. A. (1966). J. Neurophysiol. 29, 207-220. Lux, H. D., Neher, E., and Prince, D. (1972). Pfluegers Arch. 332, R89. Machiyama, Y., Balazs, R., and Richter, D. (1967). J . Neurochem. 14, 591-594. Machiyama, Y., Balazs, R., Hammond, B. J., Julian, T., and Richter, D. (1970). Biochem. J. 116, 469-48 1. McIlwain, H. (1951). Biochem. J. 49, 382-393. McIlwain, H. (1952). Biochem. J. 52, 289-295. McIlwain, H. ( 1953a). J. Neurol., Neurosurg. Psychiat. 16, 257-266. McIlwain, H. (1953b). Biochem. J. 53, 403-412. McIlwain, H. ( 1 9 5 3 ~ )Biochem. . J. 55, 618-624. McIlwain, H. (1954). J. Physiol. ( L o n d o n ) 124, 117-129. McIlwain, H. (1956). Physiol. Rev. 36, 355-375. McIlwain, H. ( 1963). “Chemical Exploration of the Brain.” Elsevier, Amsterdam. McIlwain, H. ( 1966). “Biochemistry and the Central Nervous System.” Churchill, London. McIlwain, H., and Gore, N. B. R. (1953). Biochem. J. 54, 305-312. McIlwain, H., and Joanny, P. (1963). J . Neurochem. 10, 313-323. McIlwain, H., and Tresize, M. A. (1956). Biochem. J. 63, 250-257. McIlwain, H., Buchel, L., and Cheshire, J. D. ( 1951) . Biochem. J. 48, 12-20. McIlwain, H., Anguiano, G., and Cheshire, J. D. (1952). Biochem. 1. 50, 12-18. McKhann, G. M., Albers, R. W., Sokoloff, L., Mickelsen, O., and Tower, D. B. (1960). I n “Inhibition in the Nervous System and Gamma-aminobutyric Acid” (E: Roberts et al., eds.), pp. 169-181. Pergamon, Oxford. McLennan, H. (1957). Biochem. Biophys. Acta 24, 1-8. Mahaley, M. S., Jr. (1966). Cancer Res. 26, 195-197. Maker, H. S., Lehrer, G. M., Weissbarth, S., and Bornstein, M. B. (1972). Brain Res. 44, 189-196. Mandel, P., Rein, H., Harth-Edel, S . , and Mardell, R. (1964). I n “Comparative Neurochemistry” (D. Richter, ed.) , pp. 149-163. Pergamon, Oxford. Mandel, P., Ciesielski-Treska, J., Hermetet, J. C., Hertz, L., Nissen, C., Tholey,
BRAIN METABOLISM AT T H E CELLULAR LEVEL
205
G., and Warter, F. (1973). In “Central Nervous System: Studies on Metabolic Regulation and Function” (E. Genazzani and H. Herken, eds.), pp. 223-230. Springer-Verlag, Berlin and New York. Manery, J. V., and Hastings, A. B. (1939). J. B i d . Chem. 127, 657-676. Mangan, J. L., and Whittaker, V. P. (1966). Biochem. J . 98, 128-137. Marchbanks, R. M. (1970). In “Membranes and Ion Transport” (E. E. Bittar, e d ) . Vol. 2, pp. 145-184. Wiley (Interscience), New York. Marched, V. T., Sears, M. L., and Barnett, R. J. (1964). Invest. Ophthalmol. 3, 1-21. Margolis, R. K., and Lajtha, A. (1968). Biochim. Biophys. Acta 163, 374-385. Margolis, R. K., Heller, A., and Moore, R. Y. (1968). Brain Res. 11, 19-31. Mase, K., Takahashi, Y., and Ogata, K. (1962). J. Neurochem. 9, 281-288. Medzihradsky, F., Nandhasri, P. S., Idoyaga-Vargas, V., and Sellinger, 0. Z. (1971). J. Neurochem. 18, 1599-1603. Medzihradsky, F., Sellinger, 0. Z., Nandhasri, P. S., and Santiago, J. C. (1972). J. Neurochem. 19, 543-545. Medzihradsky, F., Sellinger, 0. Z., Nandhasri, P. S., and Santiago, J. C. (1974). Brain Res. 67, 133-139. Melchior, J. B. (1965). Biochemistry 4, 1518-1525. Metzner, A. B. (1965). Nature ( L o n d o n ) 208, 267-268. Minakami, S., Kakinuma, K., and Yoshikawa, M. (1963). Biochim. Biophys. Acta 78, 808-81 1. Mfiller, M., Lund-Andersen, H., M@llg&d, K., and Hertz, L. (1974). E x p . Brain Res. 22, 299-314. Moore, B. W. (1969). In “Handbook of Neurochemistry” (A. Lajtha, ed.), pp. 93-99. Plenum, New York. Moore, B. W., and Perez, V. J. (1968). I n “Physiological and Biochemical Aspects of Nervous Integration” (F. D. Carlson, ed.), pp. 343-359. Prentice-Hall, Englewood Cliffs, New Jersey. Mori, K., Yamagami, S., and Kawakita, Y . (1970). J. Neurochem. 17, 835-843. Moss, G. (1964). Diabetes 13, 585-591. Mugnaini, E., and Walberg, F. (1964). Ergeb. Anat. Entwicklungsgesch. 37, 194-236. Nagata, Y. (1969). Bull. l a p . Neurochem. Soc. 8, 76-79. Nagata, Y., Mikoshiba, K., and Tsukada, Y. ( 1974). J . Neurochem. 22, 493-503. Nakazawa, S., and Quastel, J. H. (1968). Can. J. Biochem. 46, 355-362. Narayanaswami, A,, and McIlwain, H. (1954). Biochem. 1. 57, 663-666. Nernst, W. ( 1898). “Theoretische Chemie.” Enke, Stuttgart. Nicholls, J. G . , and Kuffler, S. W. (1964). J. Neurophysiol. 27, 645-671. Nicholls, J. G., and Wolfe, D. E. (1967). J . Neurophysiol. 30, 1574-1592. Nicklas, W. J., Clark, J. B., and Williamson, J. R. (1971). Biochem. J. 123, 83-95. Nishimura, K., and Kimura, Y. (1965). J u p . J. Exp. M e d . 35, 359-370. Nissen, C., Ciesielski-Treska, J., Hertz, L., and Mandel, P. (1972). Brain Res. 39, 264-267. Nissen, C., Ciesielski-Treska, J., Hertz, L., and Mandel, P. (1973). J . Neurochern. 20, 1029-1035. Nissl, F. (1898). Muenchen. M e d . Wochenschr. 45, 1023-1029. Norton, W. T., and Poduslo, S. E. (1970). Science 167, 1144-1 145. Nurnberger, J. I., and Gordon, M. W. (1957). Progr. Neurobiol. 2, 100-138. Ochs, S. (1962). Znt. R e v . Neurobiol. 4, 1-69.
206
LEIF HERTZ A N D ARNE SCHOUSROE
O’Connor, M. J., and Lewis, D. V. ( 1974). Electroencephalogr. Clin. Neurophysiol. 36, 337-345. O’Connor, M. J., Herman, C. J., Rosenthal, M., and Jobsis, F. F. (1972). J. Neurophysiol. 35, 471-483. O’Daly, J. A. (1967). Nature ( L o n d o n ) 216, 1329-1331. Okamoto, K., and Quastel, J. H. (1970). Biochem. J. 120, 25-36. Okamoto, K., and Quastel, J. H. (1972). Biochem. J. 128, 1117-1124. Oldendorf, W. H., and Davson, H. (1967). Arch. Neurol. (Chicago) 17, 196-205. O’Neal, R. M., and Koeppe, R. E. (1966). J . Neurochem. 13, 835-847. O’Neill, J., Simon, S. H., and Shreeve, W. W. (1965). J. Neurochem. 12, 797-802. Opit, L. J., and Charnock, J. S. (1965). Biochim. Biophys. Actu 110, 9-16. Orkand, R. K. (1971). I n “Ion Homeostasis of the Brain” (B. K. Siesjo and S. C. Sgrensen, eds.), The Alfred Benzon Symp. 111. Munksgaard, Copenhagen. pp. 124-1 3 7. Orkand, R. K., Nicholls, J. G., and Kuffler, S. W. (1966). J . Neurophysiol. 29, 788-806. Otsuka, M., Obata, K., Miyata, Y., and Tanaka, Y. (1971). J. Neurochem. 18, 287-295. Owen, 0. E., Morgan, A. P., Kemp, H. G., Sullivan, J. M., Herrera, M. G., and Cahill, G . F., Jr. (1967). J . Clin. Invest. 46, 1589-1595. Ozawa, K., Seta, K., Araki, H., and Honda, H. (1967). J . Biochem. ( T o k y o ) 62, 584-590. Pape, L. G., and Katzman, R. (1970). Physiologist 13, 278. Pape, L., and Katzman, R. (1972). Brain Res. 38, 71-92. Pappas, G. D., and Purpura, D. P. (1961). Ex$. Neurol. 4, 507-530. Pappius, H. M. (1965). Progr. Brain Res. 15, 135-154. Pappius, H. hii. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 11, pp. 1-10. Plenum, New York. Pappius, H., and Elliott, K. A. C. (1954). Can. J . Biochem. Physiol. 32, 484-490. Pappius, H. M., and Elliott, K. A. C . (1956a). Can. J . Biochem. Physiol. 34, 1007-1022. Pappius, H. M., and Elliott, K. A. C. (195613). Can. J . Biochem. Physiol. 34, 1053-1067. Pappius, H. M., Klatzo, I., and Elliott, K. A. C. (1962). Can. J . Biochem. Physiol. 40, 885-898. Passonneau, J. V., and Lowry, 0. H. (1971). I n “Recent Advances in Quantitative Histo- and Cytochemistry” (U. C. Duboch and U. Schmidt, eds.), pp. 198-209. Huber, Bern. Patel, A. J., and Balazs, R. (1970). J. Neurochem. 17, 955-971. Patel, K. K., Hartmann, J. F., and Cohen, M. M. (1971). J. Neurol. Sci. 12, 275-288. Pease, D. C. (1966). J. Ultrastruct. Res. 15, 555-588. Peters, A. (1962). J. Anat. 96, 237-248. Peters, V. B., and Flexner, L. B. (1950). Amer. J. Anat. 86, 133-161. Phelps, C. F. (1965). Biochem. J. 95,41-47. Phillips, C. G. (1956). Quart. J. Exp. Physiol. Cog. Med. Sci. 41, 69. Piccoli, F., Grynbaum, A., and Lajtha, A. (1971). J. Neurochem. 18, 1135-1148. Pietra, P. ( 1961). Boll. SOC.Ital. Biol. Sper. 37, 202-204. Pollen, D. A., and Trachtenberg, M. C. (1970). Science 167, 1252-1253. PontCn, J. (1973). I n ‘‘Tissue Culture: Methods and Applications” (P. Kruse, Jr. and M. K. Patterson, Jr., eds.), pp. 50-53. Academic Press, New York.
BRAIN METABOLISM AT T H E CELLULAR LEVEL
207
Ponten, J., and MacIntyre, E. H. ( 1968). Acta Pathol. Microbiol. Scand. 74, 465-486. Pontin, U., Ratcheson, R. A,, Salford, L. G., and Siesjo, B. K. (1973). J. Neurochem. 21, 1127-1138. Pope, A. (1958). I n “Biology of Neuroglia” (W. F. Windle, ed.), pp. 211-233. Thomas, Springfield, Illinois. Post, R. L., Merritt, C. R., Kinsolving, C. R., and Albright, C. D. (1960). J. Biol. Chem. 235, 1796-1802. Priestland, R. N., and Whittam, R. (1968). Biochem. J. 109, 369-374. Prince, D. A., Lux, H. D., and Neher, E. (1973). Brain Rex. 50, 489-495. Quastel, J. H. (1959). Proc. I n t . Congr. Biochem., 4th, 19.58 Vol. 3, pp. 90-114. Quastel, J. H. (1960). I n “Structure and Function of the Cerebral Cortex” (D. B. Tower and J. P. Schadi, eds.), pp. 374-385. Elsevier, Amsterdam. Quastel, J. H. (1962). I n “Neurochemistry” (K. A. C. Elliott, I. H. Page, and J. H. Quastel, eds.), pp. 226-237. Thomas, Springfield, Illinois. Quastel, J. H., and Quastel, D. M. J. (1961). “The Chemistry of Brain Metabolism in Health and Disease,” p. 11. Thomas, Springfield, Illinois. Quastel, J. H., and Wheatley, H. M. (1932). Biochem. J. 26, 725-744. Racker, E. (1952). I n “The Biology of Mental Health and Disease,” pp. 70-73. Hoeber, New York. Racker, E., and Krimsky, I. (1945). J . B i d . Chem. 161, 453-461. Radin, N. S., Breukert, A., Arora, R. C., Sellinger, 0. Z., and Flangas, A. L. (1972). Brain Res. 39, 163-169. Rall, D. P., Oppelt, W. W., and Patlak, C. S. (1962). Life Sci. 2, 43-48. Rambn y Cajal, S. (1909). “Histologie du systeme nerveux de l’homme et des vertkbrts,” p. 235. Maloine, Paris. Ramsey, R. L., and McIlwain, H. (1970). J. Neurochem. 17, 781-787. Rang, H. P., and Ritchie, J. M. (1968). J . Physiol. ( L o n d o n ) 196, 163-181. Ransom, B. R., and Goldring, S. (1973a). J . Neurophysiol. 36, 855-868. Ransom, B. R., and Goldring, S. (1973b). J . Neurophysiol. 36, 869-878. Ransom, B. R., and Goldring, S. ( 1 9 7 3 ~ ) J. . Neurophysiol. 36, 879-892. Reading, H. W. (1964). Nature ( L o n d o n ) 203, 491-492. Rebhan, I. (1956). Acta Anat. 27, 361-386. Reulen, H. J., Hase, U., Fenske, A,, Samii, M., and Schurmann, K. (1970). Acta Neurochir. 22, 305-325. Richards, C. D., and McIlwain, H . (1967). Nature ( L o n d o n ) 215, 704-707. Richards, C. D., and Sercombe, R. (1968). J. Physiol. ( L o n d o n ) 197, 667-683. Richter, D. (1955). I n “Biochemistry of the Developing Nervous System” (H. Waelsch, ed.), p. 225. Academic Press, New York. Richter, D. (1970). In “Protein Metabolism of the Nervous System” (A. Lajtha, ed.), pp. 241-257. Plenum, New York. Ridge, J. W. (1967). Biochem. J. 105, 831-835. Roberts, E. (1956). Progr. Neurobiol. 1, 11-25. Roberts, E. (1971). I n “Chemistry and Brain Development” (R. Paoletti and A. N. Davison, eds.), pp. 207-2 14. Plenum, New York. Roberts, E., and Frankel, S. (1950). J. B i d . Chem. 187,55-63. Roberts, P. J. (1974). Brain Res. 67, 419-428. Roitbak, A. I. (1970). Acta Neurobiol. Exp. 30, 81-94. Rolleston, F. S., and Newsholme, E. A. (1967). Biochern. J. 104, 524-533. Roots, B. I., and Johnston, P. V. (1964). J. Ultrastruct. Res. 10, 350-361. Rose, S. P. R. (1965). Nature ( L o n d o n ) 206, 621-622.
208
LEIF HERTZ AND ARNE SCHOUSBOE
Rose, S. P. R. (1967). Biochem. J. 102, 33-43. Rose, S. P. R. (1968). J. Neurochem. 15, 1415-1429. Rose, S. P. R. (1970). J. Neurochem. 17,809-816. Rose, S . P. R., and Sinha, A. K. (1969). J. Neurochem. 16, 1319-1328. RuSECk, M., and Ru3Ebkov6, D. (1971). “Metabolism of the Nerve Tissue in Relation to Ion Movements in Vitro and in Situ.” University Park Press, Baltimore, Maryland. RuSEik, M., and Whittam, R. (1967). J. Physiol. (London) 190, 595-610. RuSECk, M., RuiEikovA, D., and KonikovC, E. ( 1967). Biologia (Bratislava) 22, 337-348. Rybova, R. (1959). J. Neurochem. 4, 304-306. Sacks, W. (1957). J. Appl. Physiol. 10, 37-44. Sacks, W. (1969). In “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. I, pp. 301-324. Plenum, New York. Salganicoff, L., and Koeppe, R. E. (1968). J. Biol. Chem. 243, 3416-3420. Samson, F. E., and Quinn, D. J. (1967). J. Neurochem. 14,421-427. Sato, G., ed. (1973). “Tissue Culture of the Nervous System.” Plenum, New York. Schadt, J. P., and Baxter, C. F. (1960). Exp. Neurol. 2, 158-178. SchadC, J. P., van Backer, H., and Colon, E. (1964). Progr. Bruin Res. 4, 150-175. Schmidt, C . F., Kety, S. S., and Pennes, H. H. (1945). Amer. J. Physiol. 143, 33-52. Schon, F., and Kelly, J. S . (1974a). Brain Res. 66, 275-288. Schon, F., and Kelly, J. S . (197413). Brain Res. 66, 289-300. Schousboe, A. (1972). Exp. Brain Res. 15,521-531. Schousboe, A., and Hertz, L. (1969). Abstr. Int. Meet. Znt. SOG. Neurochem. 2nd, 2969 p. 354. Schousboe, A., and Hertz, L. (1971a). J. Neurochem. 18, 67-77. Schousboe, A., and Hertz, L. (1971b). Znt. J . Neurosci. 1, 235-242. Schousboe, A., Booher, J., and Hertz, L. (1970). 1.Neurochem. 17, 1501-1504. Schousboe, A., Fosmark, H., and Hertz, L. (1975). J. Neurochem. (in press). Schultz, R. L., Maynard, E. A,, and Pease, D. C. (1957). Amer. J. Anat. 100, 369-407. Schwartz, A., Bachelard, H. S., and McIlwain, H. (1962). Biochem. J. 84, 626-637. Scott, B. S., and Fisher, K. C. (1970). Exp. Neurol. 27, 16-22. Scott, B. S.,Engelbert, V. E., and Fisher, K. C. (1969). Exp. Neurol. 23, 230-248. Seeds, N. W., Gilman, A., Amano, T., and Nirenberg, M. W. (1970). Proc. Nut. Acad. Sci. U.S.66, 160-167. Seiler, N. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. I, pp. 325-468. Plenum, New York. Sellinger, 0. Z . , Azcurra, J. M., Johnson, D. E., Ohlsson, W. G., and Lodin, 2. (1971). Nature ( London) , New B i d . 230, 253-256. Selwood, L. (1971). Brain Res. 19, 15-25. Sensenbrenner, M., Lodin, Z., Treska, J., Jacob, M., and Mandel, P. (1968). C.R. Acad. Sci. 267, 1660-1662. Shanes, A. M. (1946). J. Cell. Comp. Physiol. 27, 115-118. Shanes, A. M., and Hopkins, H. S . (1948). J. Neurophysiol. 11, 331-342. Shein, H. M., Britva, A., Hess, H. H., and Selhoe, D. J. (1970). Brain Res. 19, 497-501. Silberstein, S. D., Shein, H. M., and Berv, K. R. (1972). Brain Res. 41, 245-248. Sinha, A. K., and Rose, S. P. R. (1972). Bruin Res. 39, 181-196. Skou, J. C. (1957). Biochim. Biophys. Acta 23, 394-401.
BRAIN METABOLISM AT THE CELLULAR LEVEL
209
Skou, J. C. (1960). Biochim. Biophys. Acta 42, 6-23. Skou, J. C. (1962). Biochim. Biophys. Acta 58, 314-325. Snodgrass, S. R., and Iversen, L. Id. (1973). Nature ( L o n d o n ) , New Biol. 241, 154- 156. Sokoloff, L., Reivich, M., Patlak, C. S., Pettigrew, K. D., Des Rosiers, M., and Kennedy, C. (1974). Trans. Anzer. Soc. h eurochetn. 5,85. Solomon, A. K. (1960). I n “Mineral Metabolism” (C. L. Comar and F. Bronner, eds.), Vol. I, Part l A , pp. 119-163. Academic Press, New York. Sotelo, C. (1967). Progr. Brain Res. 25, 226-250. Srinivasan, V., Neal, M. J., and Mitchell, J. F. (1969). J . Neurochem. 16, 1235-1244. Stewart-Wallace, A. M. ( 1939). Brain 62, 426-438. Sugawara, H., and Utida, S. (1961). Sci. Pap. Coll. Gen. Educ., Unic. Tokjlo 11, 139-151. Swanson, P. D. (1968). J . Neurochem. 15, 57-67. Swanson, P. D., and McIlwain, H. (1965). J. Neurochem. 12, 877-891. Sypert, G. W., and Ward, A. A,, Jr. (1971). Exp. Neurol. 33, 239-255. Takagaki, G. (1968). J . Neurochem. 15, 903-916. Takagaki, G. (1972). J. Neurochem. 19, 1737-1751. Takagaki, G., and Tsukada, Y. (1957). J . Neurochen. 2, 21-24. Takagaki, G., Hirano, S., and Nagata, Y. (1959). J. Neurochem. 4, 124-134. Takahashi, K. ( 1965). J. Neurophysiol. 28, 908-924. Tallan, H. H., Moore, S., and Stein, W. H. (1956). J . Biol. Chem. 219, 257-264. Tamarit, J., and Gallego, A. (1962). Abstr., Znt. Congr. Physiol. Sci. [Proc.], 22nd, 1962 Vol. 2, p. 1117. Terner, C., Eggleston, L. V., and Krebs, H. A. (1950). Biochem. J. 47, 139-149. Tholey, G., Ciesielski-Treska, J., Wurtz, B., and Mandel, P. (1972). C . R. Acad. Sci. 275, 1715-1718. Thomas, J., and McIlwain, H. (1956). J. Neurochern. 1, 1-7. ‘I’homson, C. G., and McIlwain, H. (1961). Biochem. J. 79, 342-347. Tiplady, B., and Rose, S. P. R. (1970). Biochern. J. 117, 65P. Tiplady, B., and Rose, S. P. R. (1971). J . Neurochem. 18, 549-558. Tobias, J. M., Clark, D. B., and Gerard, R. W. (1942). Fed. Proc., Fed. Amer. Soc. Exp. Biol. 1, 85. Torack, R. M., (1965). J . Histochern. Cytochem. 13, 191-205. Torack, R. M., and Barnett, R. J. (1963). J . Histochem. Cytochem. 11, 763-772. Torack, R. M., Dufty, M. L., and Haynes, J. M. (1965). Z. Zellforsch. Mikrosk. Anat. 66, 690-700. Torney, J. Mc D. (1966). Nature ( L o n d o n ) 210, 820-822. Tower, D. B. (1954). J. Comp. Neurol. 101, 19-46. Tower, D. B. (1960). I n “The Neurochemistry of Nucleotides and Amino Acids” ( R . 0. Brady and D. B. Tower, eds.), pp. 173-204. Wiley, New York. Tower, D. B. (1967). I n “Brain Edema” ( I . Klatzo and F. Seitelberger, eds.), pp. 303-332. Springer-Verlag, Berlin and New York. Tower, D. B. (1969). In “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. I, pp. 1-24. Plenum, New York. Tower, D. B., and Bourke, R. S. (1966). J . Neurochem. 13, 1119-1137. Tower, D. B., and Elliott, K. A. C. (1952). Amer. J . Physiol. 168, 747-759. Tower, D. B., and Young, 0. M. (1973). I . Neurochem. 20, 253-267. Trachtenberg, M. C., and Pollen, D. A. (1970). Science 167, 1248-1251.
2 10
LEIF HERTZ A N D ARNE S C H O U S B O E
Trachtenberg, M. C., Kornblith, P. L., and Hauptli, J. (1972). Brain Res. 38, 279-298. Tschirgi, R. D. (1958). I n “Biology of Neuroglia” (W. F. Windle, ed.), pp. 130-138. Thomas, Springfield, Illinois. Tsukada, Y., and Takagaki, G. (1955). Nature ( L o n d o n ) 175, 725-726. Utida, S., and Sugawara, H. (1963). J. Biochem. ( T o k y o ) 54, 553-554. Utley, J. D. (1963). Biochem. Pharmacol. 12, 1228-1230. Utter, M. F. (1950). J. Biol. Chem. 185, 499-517. Valcana, T., and Timiras, P. S. (1969). J . Neurochem. 16, 935-943. Van den Berg, C. J. (1970a). J. Neurochem. 17, 973-983. Van den Berg, C. J. (1970b). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 111, pp. 355-379. Plenum, New York. Van den Berg, C. J. (1973). In “Metabolic Compartmentation in the Brain” (R. Balazs and J. E. Cremer, eds.), pp. 137-166. Macmillan, New York. Van den Berg, C. J., and Garfinkel, D. (1971). Biochem. J. 123, 211-218. Van den Berg, C. J., Krzalic, Lj., Mela, P., and Waelsch, H. (1969). Biochem. J. 113, 281-290. Van Harreveld, A. (1962). Fed. Proc., Fed. Amer. SOC.Exp. Biol. 21, 659-664. Van Harreveld, A. ( 1966). “Brain Tissue Electrolytes.” Butterworth, London. Van Harreveld, A,, and Fifkova, E. (1970). J. Neurobiol. 2, 13-29. Van Harreveld, A,, and Fifkova, E. (1973). J . Neurobiol. 4, 375-387. Van Harreveld, A., and Malhotra, S. K. (1967). J. Anat. 101, 197-207. Van Harreveld, A,, Crowell, J., and Malhotra, S. K. (1965). J . Cell B i d . 25, 117-137. Varon, S . , and McIlwain, H. (1961). J . Neurochem. 8, 262-275. Veech, R. L., Harris, R. L., Veloso, D., and Veech, E. H. (1973). J. Neurochem. 20, 183-188. Vernadakis, A,, and Woodbury, D. M. ( 1962). Amer. J. Physiol. 203, 748-752. Vernadakis, A,, and Woodbury, D. M. (1965). Arch. Neurol. (Chicago) 12, 284-293. Victor, J., and Wolf, A. (1937). Res. Publ., Ass. Res. Nerv. Ment. Dis. 16, 44-58. Villegas, J. (1968). J . Gen. Physiol. 51, 61-71. Villegas, J., Villegas, L., and Villegas, R. (1965). J. Gen. Physiol. 49, 1-7. von Economo, C. (1926). Klin. Wochenschr. 5, 593-595. von KorfT, R. W. (1953). J. B i d . Chem. 203,265-271. Vrba, R., Gaitonde, M. K., and Richter, D. (1962). J. Neurochem. 9, 465-475. Vyklick?, L., Sykova, E., KEii, N., and Ujec, E. (1972). Brain Res. 45, 608-611. VyskoEil, F., and KEiZ, N. (1972). Pfiuegers Arch. 337, 265-276. VyskoEil, F., KFii, N., and BureE, J. (1972). Brain Res. 39, 255-259. Waelsch, H. (1962). I n “Neurochemistry” (K. A. C. Elliott, I. H. Page, and J. H. Quastel, eds.) , pp. 288-320. Thomas, Springfield, Illinois. Walker, F. D., and Hild, W. J. (1969). Science 165, 602-603. Walker, F. D., and Hild, W. J. (1972). J . Neurobiol. 3, 223-235. Walker, J. L. (1971). Anal. Chem. 43, 89A-93A. Walum, E., Nissen, C., Hertz, L., and Edstrom, A. (1974). J. Neurochem. 23, 881-883. Wardell, W. M. (1966). Proc. R o y . SOC.,Ser. B. 165, 326-361. Weber, G. (1959). Acta Neurochir., Suppl. 6, 21 1-218. Webster, L. T., Jr. (1966). J. B i d . Chem. 241, 5504-5510. Weil-Malherbe, H. ( 1938). Biochem. J. 32, 2257-2282.
BRAIN METABOLISM AT T H E CELLULAR LEVEL
21 1
Weinstein, H., Varon, S., Muhleman, D. R., and Roberts, E. (1965). Biochem. Pharmacol. 14, 273-208. Weiss, G. B., and Hertz, L. (1974). Biochem. Soc. Trans. 2,274-277. Weiss, G. B., Hertz, L., and Goodman, F. R. (1972). Biochem. Pharmacol. 21, 625-634. Wender, M., and Hierowski, M. (1960). J . Neurochem. 5, 105-108. Westrum, L. E., and Blackstad, T. W. (1962). J . Comp. Neurol. 119, 281-292. Whittaker, V. P. (1969). I n “Handbook of Neurochemistry” (A. Lajtha, ed.), Vol. 11, pp. 327-364. Plenum, New York. Whittam, R. (1962a). Nature ( L o n d o n ) 196, 134-136. Whittam, R. (1962b). Biochem. J . 84, 110-118. Whittam, R., and Ager, M. E. (1962). Biochim. Biophys. Acta 65, 383-385. Wilson, S. H., Schrier, B. K., Farber, J. L., Thompson, E. J., Rosenberg, R. N., Blume, A. J., and Nirenberg, M. W. ( 1972). J . Biol. Chem. 247, 3159-3169. Windle, W. F., ed. ( 1958). “Biology of Neuroglia.” Thomas, Springfield, Illinois. Winegrad, S., and Shanes, A. M. (1962). J . Gen. Physiol. 45, 371-394. Wolff, J. (1965). Z . Zellforsch. Mikrosk. Anat. 66, 811-828. Wolff, J. (1970). Abstr., Znt. Congr. Neuropathol., 6th, 1970 pp. 327-333. Woodman, R. J., and McIlwain, H. (1961). Biochem. J . 81,83-93. Woodward, D. L., Reed, D. J., and Woodbury, D. M. (1967). Amer. J . Physiol. 212, 367-370. Wyckoff, R. W. G., and Young, J. Z. (1956). Proc. Roy. Soc., S e r . B 144, 440-450. Yamamoto, C., and Kawai, N. (1967). Science 155, 341-342. Yamamoto, C., and Kurokawa, M. (1970). E x p . Brain Res. 10, 159-170. Yamamoto, C., and McIlwain, H. (1966a). Nature ( L o n d o n ) 210, 1055-1056. Yamamoto, C., and McIlwain, H. (196613). J . Neurochem. 13, 1333-1343. Yannet, H. ( 1940). Amer. J . Physiol. 128, 683-689. Young, J. A. C., Brown, D. A., Kelly, J. S., and Schon, F. (1973). Brain Res. 63, 479-486. Zadunaisky, J. A., and Curran, P. F. (1963). Amer. J . Physiol. 205, 949-956. Zadunaisky, J. A,, Wald, F., and De Robertis, E. D. P. (1963). Exp. Neurol. 8, 290-309. Zaheer, N., Iqbal, Z., and Talwar, G. P. (1968). J . Neurochem. 15, 1217-1224. Zuckermann, E. C., and Glaser, G . H. (1970). Arch. Neurol. (Chicago) 23, 358-364.
This Page Intentionally Left Blank
AGGRESSION AND CENTRAL NEUROTRANSMITTERS By
S. N. Pradhan
Department of Pharmacology, Howard University College of Medicine, Washington, D.C.
.
I. Introduction A. Definition . B. Types of Aggression . C. Factors Influencing Aggression: Induced Aggression D. Objective of This Review . 11. Neuroanatomical and Neurochemical Correlation of Aggression A. Neuroanatomical Correlates of Aggression. B. Neurochemical Correlates of Aggressive Behavior Induced by Manipulation of Brain Areas . C Neurochemical Correlates of Aggression . 111. Chemostimulation of Discrete Brain Areas and Induced Aggression. IV. Neuropharmacological Manipulation of Aggression A. Drugs Related to Noradrenergic Mechanism . B. Drugs Related t o Dopaminergic Mechanism C Drugs Related to Cholinergic Mechanism . D. Drugs Related t o Serotonergic Mechanism. V. Summary and Conclusion References .
.
.
. .
.
.
. .
. . . . . . . .
. . . . .
213 213 214 216 219 220 220 223 228 232 237 237 242 249 251 253 255
1. Introduction'
A. DEFINITION Aggression is an emotional behavior widely common to animals, including human beings. It is vital for the existence of the individual and plays ' T h e following abbreviations have been used i n the text, in addition to others mentioned in appropriate locations : ACh, acetylcholine ; AChE, acetylcholinesterase ; CA, catecholamine (s) ; ChA, choline acetyltransferase; DA, dopamine; DBH, dopamine-@-hydroxylase; DOPA, dihydrophenylalanine; FLA-63, bis (4)-methyl-1-homopiperazinylthiocarbonyl) disulfide, a DBH inhibitor; 5-HIAA, 5-hydroxyindoleacetic acid; 5-HT, 5-hydroxytryptamine (serotonin) ; i.c., intracisternal; i.m., intramuscular; i.p., intraperitoneal; i.v., intravenous; i.vent., intraventricular ; LH, lateral hypothalamus; MAO, monoamine oxidase ; MAOI, monoamine oxidase inhibitor; MFB, medial 213
214
S. N . PRADHAN
an important role in the process of evolution (Leakey, 1967). The term aggression has been used to indicate a particular oriented behavior which leads to, or appears to an observer to lead to, the damage or destruction of some goal. Some ethologists have used the term “agonistic behavior” to describe various sorts of behavior that occur during conflict and have described several patterns of this behavior. These patterns have been thought to be the result of a process of evolution within each species, perhaps involving different physiological mechanisms (Scott, 1966). Three patterns of this behavior have been distinguished in the cat: (1) defense, and ( 2 ) attack, as opposed to ( 3 ) flight. The term aggression has been associated with fighting and is used by some authors synonymously with attack and by others to include both defense and attack (Scott, 1958). However, it may be mentioned that aggression may or may not result in actual fighting, damage, or destruction, depending on the emotional level shown by subject or subjects, and may only involve certain manifestations indicative of rage, anger, hyperirritability or hyperreactivity.
B. TYPESOF AGGRESSION Aggression does not appear to be a unitary phenomenon. This behavior is evoked in response to a given stimulus and is in part genetically determined and in part influenced by learning. The stimulus situations that elicit this destructive behavior and the patterns of behavior that result in damage and destruction are so diverse that it appears to involve a variety of physiological substrates. There are a number of factors that determine this behavior and important among them are competition for food, social status, territoriality, or reproduction. Moyer ( 1968) tentatively classified aggression into seven categories on the basis of stimulus characteristics. Table I summarizes the characteristics of six of these categories (i.e., predatory, intermale, fearinduced, irritable, territorial, and maternal) along with the stimulus for their elicitation, and their neuroanatomical and endocrine substrates. Added to this list is the instrumental aggression which is a learned response in which any of the above-mentioned classes of aggression may result in a change in the environment that may reinforce the subject for elicitation of the aggressive behavior under a similar situation. I t can be observed from Table I that there is a considerable overlap between different types of aggression, and that the underlying neuroendoforebrain bundle; a-MT, a-methyl-p-tyrosine; n., nucleus; NE, norepinephrine ; 6-OHDA, 6-hydroxydopamine; 6-OHDOPA, 6-hydroxy-DOPA; PCPA, p-chlorophenylalanine; s.c., subcutaneous; TH, tyrosine hydroxylase; VMH, ventromedial hypothalamus.
215
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
TABLE I TYPES OF AGGRESSION”
Types Predatoryb
Stimulus for elicitation
Neuroanatomical substrate
Endocrine influence
Remarks
Lateral hypothal- Slightly reduced amus particuby castration a t an early age larly involved; facilitated by amygdaia and inhibited by olfactory bulb, prepyriform and frontal cortex, etc. InterStrange conspeSepta1 lesions are Male sex hormalebzc cific male mones may be facilitatory critical Fear-inSome threatening Reduced by duced object amygdalectomy and anterior cingulectomy, and facilitated by lesions of septum and ventromedial hypothalamus IrritaExtremely broad Many brain Affected by casblec,d areas, particurange of stimtration and uli; an attacklarly ventrocertain endoable animate medial hypocrine dysfuncor inanimate thalamus and tions medial nuclei object of amygdala are involved TerriPresence of an inGonadal hortorial truder (usually mones may be conspecific) in involved a n established territory Maternal Some agent Amygdala, sepIncrease during threatening to tum, or p p i lactation, form cortex which can be the young may be insuppressed by volved estrone A natural object of prey
A widely investigated subtype is “muricide” (mouse-killing behavior) by rats
Increased by isolation Preceded by escape attempts
Usually manifested as anger or rage; no specific stimulus; not preceded by escape attempts Relevance to environment (own territory)
~~~
a
Materials taken from Moyer (1968). Possesses specificity of stimulus. Has no relevance to environment. Increased by frustration, drive, slerp deprivation, aversive stimulation.
216
S. N. P R A D H A N
crinological mechanisms of many of these types are not clearly understood. The different types of aggression presented in the table are usually spontaneous and can be evoked in response to a stimulus that may be a subject of the same species (intraspecific or conspecific) or of a different species (interspecific) or even may be an inanimate object.
C. FACTORSINFLUENCING AGGRESSION : INDUCED AGGRESSION The aggressive manifestations, discussed earlier, can be facilitated or inhibited to varying degrees by several procedures, such as (1) brain lesion and stimulation, ( 2 ) aversive stimulation, ( 3 ) isolation, and (4) pharmacological treatment (Valzelli, 1967). 1. Brain Lesion and Stimulation Lesion or electric stimulation of several brain areas (e.g., olfactory bulb, septum, amygdala, hypothalamus) is known to modify aggressive manifestations (Tables I and 11),which will be discussed later.
2. Aversiue Stimulation Aversive stimulation can enhance irritable aggression that may be manifested against animate as well as inanimate objects. Electric foot-shock is a widely used experimental stimulus that elicits various manifestations of anger or rage. Fighting can be induced by subjecting a pair of rats to foot -shocks.
3. Isolation Prolonged social isolation may induce a marked tendency in a mouse to exhibit aggression against another strange mouse, the most aggressive display being violent attacks on, and biting of, the opponent. This procedure has the advantage of being a pure psychological reaction to an unusual situation and does not require any particular manipulation of the animals, and the behavioral change is reproducible and long-lasting. Following earlier observations of Allee ( 1942a,b), Scott ( 1946, 1958), and Seward ( 1945a,b, 1946) concerning behavior of mice in isolation, Yen et al. (1959) established requisite experimental conditions for isolation-induced aggression in mice. The most suitable species for laboratory investigation of isolation-induced aggression is the mouse, although other species, such as the rabbit (Wolf and von Haxthausen, 1960) and the rat (Bevan et al., 1951), have also been used. 4. Pharmacological Treatment Different types of aggressive behavior have been shown to be modified by a variety of drugs, such as tranquilizers and sedatives, antidepressants,
217
AGGRESSION A N D CENTRAL NEUROTRANSMITTERS
TABLE I1 NEUROANATOMICAL CORRELATES OF AGGRESSION Brain area, experimental manipulationa
Subject: Type of aggression/effect
Olfactory deafferentiation (removal of nasal mucosa)
Rat: No change in muricide
Olfactory bulbectomy
Olfactory bulbotomy Olfactory bulbotomy, L
Olfactory bulbotomy, L
Prepyriform cortex, L Genual portion of anterior cingulate gyrus, L Bilateral lesions of corpus striatum with or without overlying dorsal cortex Septum, L
Septum, medial, ES Amygdala, Centromedial area n., L
Amygdala, basal and medial, and dorsomedial nuclei,
ES
Reference
Cain (1974), Spector and Hull (1972) Mild increase in irritability Cain (1974) Cain (1974), Spector and Rat: Muricide elicited Hull (1972), Thorne et ~ l . and/or facilitated (1973), Gain and Paxinos (1974), Vergnes and Karli (1963), Bernstein and Moyer (1970), Malick (1970) Di Chiara et al. (1971) Rat: Muricide elicited Bandler and Chi (1972) Rat: Muricidal4 nonmuricidal and nonmuricidal muricidal Mice, 0 : Prevented arousal Rowe and Edwards (1971) of aggression by androgen: no effect on aggression due to competition for food Vergnes and Karli (1965) Rat: Elicited muricidal response Brutkowski et al. (1961), Dog: Angry behavior Brutkowski and Mempel (1961) Rat: No effect on rage reac- Fog et nl., 1970 tion induced by Eutonyl (MAOl) and DOPA Stark and Henderson (1966, Rat: Hyperreactivity 1972), Brady and Nauta (1953) Karli (1960), Mailey and Muricide Baenninger (1972) Malick (1970) Aggression (defensive) Thomas and Van Atta No change in irritability (1972) Rubinstein and Delgado Monkey: aggression in(1963) hibited Rat: Muricide
Cat: Increased irritability
Karli (1956), Karli and Vergnes (1965), Horovitz et al. (1966) Hilton and Zbrozyna (1963), Fernandez De Molina and Hunsperger (1959, 1962) (Continued)
218
S . N. PRADHAN
TABLE I1 (Continued) Brain area, experimental manipulationa Ventral amygdalohypothalamic fibers, ES L Stria terminalis and its nucleus, ES
L Hypothalamus, ES
Subject: Type of aggression/effect Cat: Increased irritability Rat: Muricide inhibited Cat: Increased irritability
Rat: Increased irritability and rage Cat: oppossum: Attack and killing
Rat: Affective and quite mouse-killing and biting Hypothalamus Anterior and preoptic, ES
Reference Hilton and Zbrozyna (1963) Vergnes and Karli (1964) Hilton and Zbrozyna (1963), Fernandez De Molina and Hunsperger (1959, 1962) Thomas and Van Atta (1972), Turner (1970) Wasman and Flynn (1962), Roberts and Bergquist (1968), Roberts et al. (1967) Panksepp (1971b)
Fernandez De Molina and Hunsperger (1959) King and Hoebel (1968), Lateral, ES Rat: Muricide Woodworth (1971) Lateral, L Rat: Muricide and shock-in- Panksepp (1971a) duced fighting decreased; muricide and shock-inMedial, L duced fighting intensified Glusman et al. (1961) VMH, L Cat: Savage behavior Malick (1970) Rat: Muricide Reis et al. (1970) Posterior, ES Cat: Attack behavior Tr. through hypothalamus Rat Grossman and Grossman Anterior (ant. to VMH) Intraspecies aggression (1970) inhibited No effect on aggression Posterior (post. to VMH) Between medial and Rat Sclafani (1971), Paxinos and lateral hypothalamus Increased irritability Bindra (1972) Paxinos and Bindra (1972) No change in muricide Varying changes in inter- Grossman (1970) species fighting Tr. of lateral connections Rat: No change in aggresGrossman (1 970) of lateral hypothalamus sion Central gray of midbrain, ES Cat: Increased irritability Fernandez De Molina and Hunsperger (1959) Cat: Increased irritability
(Continued)
AGGRESSION AND C E N T R A L NI'. UROTRAN S M I T T E R S
219
TABLE I1 (Continued) Brain area, experimental manipulation" Tegmentum, L Medial Lateral Midbrain reticular formation, L Inferior collicular Superior collicular
Subject: Type of aggression/effect
Reference
Cat Blocks elicited attack and Berntson (1972) reflex biting Exhibits spontaneous attack and reflexive biting Rat Increased nonspecific aggression No change
Kesner and Keiser (1973)
a Abbreviations: DOPA, dihydroxyphenylalanine; MAOI, monoamine oxidase inhibitor; n., nucleus, V M H , ventromedial hypothalamus; ES, electric stimulation; L, lesion; T r , transection.
psychostimulants, hallucinogens, hormones, and drugs affecting neurohormones (Avis, 1974; DaVanzo, 1969; Krsiak, 1974; Tedeschi et al., 1969; Valzelli, 1967). Using these procedures, it is not only possible to modify already existing aggressive manifestations in a subject, but also to induce aggression in an otherwise nonaggressive animal. Isolation-induced and shock-induced types of aggression are examples.
D. OBJECTIVE OF THISREVIEW The previous discussions indicate that aggression can be spontaneous or induced, and each of these varieties can be further subdivided into a number of types. Many of these types have been investigated extensively from ethological, neuroanatomical, neurochemical, psychophysiological, psychopharmacological, and other viewpoints. In addition to numerous reports, many reviews (Avis, 1974; Clemente and Chase, 1973; Krsiak, 1974; Krsiak and Steinberg, 1969; Moyer, 1968; Scott, 1966; Valzelli, 1967; Vernon, 1969) and books (Clemente and Lindsley, 1967; Garattini and Sigg, 1969; Johnson, 1972) have been published on various aspects of aggression. In recent years one interesting as well as challenging aspect of research on this behavior has been its neurochemical correlation, particularly correlation with putative central neurotransmitter mechanisms. There are very few reviews (Avis, 1974; Eichelman and Thoa, 1973) dealing with such neurochemical correlation of aggressive behavior. However, such correlation is
220
S . N . PRADHAN
extremely complex and difficult because of the heterogeneity of aggressive behavior and the existence of a number of putative central neurotransmitters. Furthermore, there is a great paucity of information in many areas. In the present review attempts have been made to correlate various types of aggression with several putative central neurotransmitter mechanisms, particularly in relation to norepinephrine, dopamine, serotonin, and acetylcholine. For the purpose of discussion here, only four types of aggression, which have been more extensively studied, are selected; two of these are spontaneous in nature-muricide and irritable-and two others are induced varieties-isolation-induced and shock-induced. Three approaches have been followed for the purpose of such correlation: 1. Use of stimulation or ablation procedure in certain areas of the brain to induce or modify aggressive behavior and subsequent correlation of the induced behavioral changes with associated neurochemical changes in various areas of the brain. 2. Measurement of the neurochemical changes in the whole or parts of the brain of animals showing aggressive manifestations that may be spontaneous or induced, but not resulting from stimulation or ablation of the brain areas. 3. Neuropharmacological manipulation in which drugs known to be related to various neurotransmitter mechanisms are used as tools to modify aggressive behavior and subsequent indirect correlation of the behavioral changes with the neurochemical mechanism.
II. Neuroanatornical and Neurochernical Correlation of Aggression
A. NEUROANATOMICAL CORRELATES OF AGGRESSION In a large number of investigations, effects of lesion or electrical stimulation of various brain areas of different types of aggressiveness have been studied. Results of some such experiments with the olfactory bulb, septum, amygdala, stria terminalis and its nucleus, hypothalamus, and central gray substance and tegmentum of the midbrain are briefly discussed below and summarized in Table 11. Lesions of the olfactory bulb in the rat elicit or facilitate muricide (mouse killing) behavior. The caudal region of the olfactory bulb (especially anterior olfactory nucleus) is involved in the muricidal behavior, while the rostra1 region may cause some change in the irritability of the animal. Bulbectomy also induced irritability depending on the extent of damage to the areas within the bulbs to which the stria terminalis projects (Cain, 1974). Removal of the sensory epithelium in the nose does not elicit
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
22 1
or facilitate this behavior. Thus, the olfactory bulb seems to save an inhibitory influence on muricide behavior in rats. The amygdala, in rats and cats, seems to play a positive role in the display of aggressive behavior. Lesions of the amygdala or its centromedial part (which includes the central nucleus, dorsal part of the medial nucleus and the medial part of the basal nucleus) inhibit muricide behavior in rats. Electric stimulation of the amygdala, especially its basal, medial or dorsomedial nucleus in cats results in increased irritability. Similar stimulation of either the ventral amygdalohypothalamic fibers or the stria terminalis and its nucleus in cats as well as rats causes increased irritability and rage, thus indicating a functionally positive role of these structures in the display of aggressive behavior. However, in the rat, lesions of the stria terminalis and its nucleus have also been shown to result in increased irritability. The role of the septal nuclei seems to be inhibitory on aggression, since ( 1 ) lesion of this structure causes hyperreactivity and elicitation or facilitation of muricide behavior in the rat, and ( 2 ) electric stimulation of septum (medial) in the monkey causes inhibition of aggression. However, small lesions confined to medial septal nuclei have been shown to result in greater passivity and hyperreactivity (Clody and Carlton, 1969). The hypothalamus plays a dual role, inhibitory as well as facilitatory, in the aggressive behavior of animals. The preoptic area, and the anterior, the lateral, and the posterior hypothalamus have a facilitatory role, while the medial hypothalamus (especially the VMH nucleus) is inhibitory on aggression. In the cat, electric stimulation of the preoptic area, anterior hypothalamus, or posterior lateral hypothalamus results in increased irritability and attack behavior, while lesions of the VMH result in savage behavior. Electric stimulation of the L H or lesion of the medial hypothalamus in the rat result in increased muricidal activity, while lesions of the L H decrease muricidal activity and shock-induced fighting. The medial and lateral parts of the hypothalamus are functionally interrelated, since a transection between these parts causes increased irritability in the rat ; however, transections of the possible lateral connections of the hypothalamus do not produce any change in aggressive behavior. T h e ventromedial portion of the central gray and the medial tegmentum of the midbrain appear to have a positive role, since electric stimulation of these areas in both rats and cats elicits aggressive behavior. O n the other hand, the lateral tegmentum may play a negative role, since lesions of this region result in spontaneous attack and reflexive biting. Among the cortical structures, the prepyriform cortex in the rat and genual portion of the cingulate cortex in the dog play an inhibitory role, since their lesions elicit aggressive behavior. The structures and their interconnections associated with the aggressive
222
S . N . PRADHAN
behavior are represented in Fig. 1. Rostra1 structures, such as the olfactory bulb (especially the olfactory nuclei), that are inhibitory in nature are connected to the prepyriform cortex and amygdala, among other structures. The amygdala, especially its centromedial part, plays an excitatory or positive role in the aggressive behavior. The prepyriform cortex that is inhibitory in nature is connected to the amygdala. The amygdala is in turn connected to the hypothalamus via two main pathways: (1) stria terminalis and (2) ventral amygdalohypothalamic fibers. The stria terminalis projects to the hypothalamus, septum, and olfactory nuclei. Although the stria, especially its bed nucleus, is known to play a role in aggressive behavior, recent investigations indicate that the ventral amygdalohypothalamic pathway is more critical in conveying a positive influence of the amygdala to the hypothalamus.
FIG. 1. Schematic diagram of the connections and interconnections of different central nervous structures predominantly involved in the facilitation (+++) or suppression (---) of muricide behavior in rats. A, amygdala; AC, anterior commissure; BNST, bed nucleus of stria terminalis; DLF, dorsal longitudinal fasciculus; DM, dorsomedial nucleus of the thalamus; ITP, inferior thalamic peduncle; LH, lateral hypothalamus ; LOT, lateral olfactory tract; MCG, mesencephalic central gray; MFB, medial forebrain bundle ; OB, olfactory bulb; O T , olfactory tubercle; PC, prepyriform cortex; S , septum; ST, stria terminalis ; VMH, ventromedial hypothalamus ; VMT, ventromedial midbrain tegmentum; VP, ventral amygdalofugal pathway). Modified from Karli et al. (1972).
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
223
In the hypothalamus, the medial part (especially VMH) has a inhibitory influence while the anterior, posterior, and lateral (especially the latter) parts of the hypothalamus have an excitatory role in aggressive behavior. The medial hypothalamus is connected with the lateral part, and transection of the connection between these parts releases the LH from the inhibitory influence of the medial hypothalamus. The hypothalamus (especially the medial and the lateral parts) receives inputs directly or indirectly from all the structures discussed above. These inputs are then integrated and conveyed caudally to the brainstem and the spinal cord via MFB and dorsal longitudinal fasciculus for manifestation of the aggressive behavior. Thus, the LH seems to act as an organizer in this behavior. The output of the hypothalamus may be further influenced on its way caudally by the midbrain central gray substance and the medial tegmentum which have excitatory influence, or by the lateral midbrain tegmentum that plays an inhibitory role.
B. NEUROCHEMICAL CORRELATES OF AGGRESSIVE BEHAVIOR INDUCED BY MANIPULATION OF BRAINAREAS In the preceding section aggressive behavior has been shown to be positively or negatively influenced by various brain areas by the use of lesioning and electric stimulation procedures. The same procedures have been used to study the changes in concentration and turnover rates of certain central neurotransmitters or changes in the other neurochemical parameters, and to correlate them with aggression or other behavioral changes in the subjects. Results from a number of investigations involving some brain areas are summarized in Table I11 and discussed below.
1, Telencephalon a. Amygdala. Sham rage in cats was produced by electric stimulation of certain regions of the amygdala (Reis and Gunne, 1965) or the hypothalamus (Gunne and Lewander, 1966) as well as by transection above the superior colliculi (“high decerebration” ) leaving the hypothalamus intact (Reis et al., 1967). In all these situations there was a selective decrease in NE level of the brainstem (also in the forebrain in stimulation experiments) and catecholamine (NE and E ) levels of the adrenal without any change in the levels of DA or 5-HT. Reduction of NE occurred in axon terminals throughout the lower brainstem without selective involvement of any one area. This reduction of brain NE during sham rage was probably the result of a marked increase in the activity, and consequently the release, of NE from axon terminals of noradrenergic neurons (Reis, 1972). b. Olfactory Bulb. Unilateral olfactory bulb lesions in the rat caused a selective decrease of NE in the telencephalon and its increase in the brain-
TABLE 111: NEUROCHEMXCAL CORRELATES OF AGGRESSION INDUCED
BY
MANIPULATION OF BRAINAREAS=
Changa in Brain area, manioulation. subiect
1\3 1\3
Brain area examined
NE
DA
5-HT
FB, BS (also adrenal) Pons and medulla
-
0 0
0 0
Sham rage Shamrage
TLC
-
0
0
BS
+
NE in BS increased by isolation
0
0
Remarks
References
.P
~
1. Telencephalon Amygdala, ES, cat Decerebrate “high”, cat
Olfactory bulb, L, rat
Bulbectomy, rat
Amygdala
Septum, L, rat
WB Cortex Hippocampus HB FB
2. Diencephalon Hypothalamus, anterolateral, ES, rat ES, cat
0
0 0
0
+ 0
0
Muricide
Poncey ct al. (1972)
P
Heller and Moore
?
(1973) Defense reaction, 5-HIAA
+ Sham rage ++ +,
+
VMH, L, rat
WB
LH, L, rat
WB
Flinch jump threshold
MFB, L
TLC (ipd) Frontal lobe
WB and areas in TLC TLC
Ebel ct al. (1973)
Salama and Goldberg
adrenal CA Flinch jump threshold
MFB (LH), L, rat
Reis and Gunne (1965) Reis ct al. (1967), Reis and Fuxe (1969), Reis (1972) Pohorecky and Chalmers (1971)
(1 9681, Moore (1 970)
+, T +++
-
+,
TH ChA AChE 0 Hyperreactivity, hyperphagia
FB and BS
FB and BS
+,
+ +
+ +++
Aphagia and adipsia
Kostowski and Giacalone (1969) Gunne and Lewander (1966) Poncey ct al. (1972)
Zigmond ct al. (1971) Jung and Hassler (1960) Moore (1970) Oltmans and Harvey (1972)
.d
Fz
MFB and MIC, Tr, rat NSB, L., rat
VL thalamus, L, monkey 3. Mesencephalon MR, ES, rat
TLC Neostriatum TLC (remaining)
-
-
-
Aphagia, adipsia
_-
---
-
Severe aphagia, adipsia, and disturbance of water regulation Relieves resting tremor induced by VM Tg lesion
-0,
Striatum
uo
-, T+ 5-HIAA
Forebrain
+, behavioral
depression
MR, L, rat
-
Brain and spinal cord
DMTg, L, rat VLTg, L, rat
WB WB and areas in TLC
VMTg, L, monkey
Striatum (ipsl)
Central gray, L, rat
L, C
-
-
-_
-
uo
0
U-
-
Also 5-HIAA No increase in sexual or aggressive behavior TH
Heller and Moore (1968) Oltmans and Harvey (1972) Battista et al. (1969a,b) 9
87J Aghajanian et af. (1967); Kostowski and Giacalone (1969); Gumulka etal. (1963); Kostowski et al. (1969) Giacalone and Kostowski (1968) Sheard (1973) Moore (1970)
-
Poirier and Sourkes (1965) Goldstein et al. (1966) Zigmond et al. (1971)
Abbreviations and symbols: AChE, acetylcholinesterase; BS, brainstem; C, content; CA, catecholamines; ChA, choline acetyl transferase; DA, dopamine; ES, electric stimulation; FB, forebrain; HB, hindbrain; 5-HIAA, 5-hydroxyindoleacetic acid; 5-HT, 5-hydroxytryptamine; ipsl, ipsilateral; L, lesion; LH, lateral hypothalamus; MFB, medial forebrain bundle; MIC, medial part of internal capsule; MR, midbrain raphe (medial nucleus) ; NE, norepinephrine; NSB, nigrostriatal bundle; T, turnover; Tg, tegmentum; TH, tyrosine hydroxylase; TLC, telencephalon; Tr, transection; U, uptake (in uitro); VL, ventrolateral; VM, ventromedial; VMH, ventromedial hypothalamus; (or - (or 0 indicate, respectively, increase, decrease, or no change usually of the contents. If prefixed by T o r U, WB, whole brain. they refer to turnover or uptake, respectively. Two or three or - signs indicate comparably increasing changes, and, wherever quantitatlve data are available, they represent changes by 26-50% or 51-75%, respectively. One 3- or - does not always provide quantitative representation.
+ T),
l),
+
m m
! z > 2
0
n P! z
i 7J 9
r
zm C 7J 0 +I 7J 9
5
s7 2 $
.I
226
S . N. PRADHAN
stem without any change in the concentration of DA or 5-HT; M A 0 activity also decreased in the telencephalon and increased in the brainstem of the lesioned animals. Isolation of the subjects further increased the brainstem NE concentration as well as degree of aggressiveness in the lesioned animals compared to the aggregated animals (Pohorecky et al., 1969; Pohorecky and Chalmers, 1971). Bulbectomy also increased T H and ChA activity; AChE activity was decreased, but not significantly (Ebel et al., 1973). c. Septum. Krnjevid and Silver (1965) demonstrated in the septum of the cat a number of cells strongly stained for AChE and AChE-containing fibers projecting to the cortex. Stimulation of the septum in the cat markedly increased ACh output from the cortex with only moderate EEG activation (Szerb, 1967). Bilateral lesions in the septal area produced both a significant decrease in the ACh content of the brain and also a significant increase in daily water consumption (Sorensen and Harvey, 1971) . Unilateral electrocoagulative lesion of the septum in rats decreased the ACh content in the brain, particularly in the cortex, and to some extent in the diencephalon and rostra1 midbrain (Pepeu et al., 1971). I t is possible that septal lesions cause degeneration of cholinergic fibers with the loss of ChA activity as shown by Lewis et al. (1967) after lesions of the fimbria in the rat. It was suggested (Pepeu et al., 1971) that the cholinergic fibers originating from the septum are widely distributed in the brain, particularly in the cortex including the hippocampus, but they do not reach the lower brainstem. Effects of septal lesions on monoamine metabolism have also been studied. I n rats with septal lesions showing hyperphagia and hyperreactivity, no significant change in the NE level was observed when the whole brain was used (Poncey et al., 1972). However, estimation of amine contents in discrete brain regions showed some changes. Thus, following septal lesions in rats, a significant decrease of NE (by 26%) and of 5-HT (by 15%) in the neocortex and of 5-HT (by 40%) in the hippocamp was observed (Heller and Moore, 1968; Moore, 1970). O n the other hand, a small but significant increase of NE level (by 12%) together with a 76% increase in its turnover rate in the hindbrain with no apparent change in the contents of 5-HT and NE in the forebrain was observed in septal-lesioned rats that showed muricidal responses (Salama and Goldberg, 1973). This is contrary to the observations in natural muricidal rats that showed normal brain 5-€IT level and turnover rate, but higher level and faster turnover rate of forebrain NE compared with nonkiller rats (Goldberg and Salama, 1969; Salama and Goldberg, 1970).
2. Diencephalon From Table I11 it is evident that lesions of the VMH or the LH, especially in the MFB (with or without transection of the medical part of
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
227
the internal capsule), in rats cause decrease in the content of NE as well as those of DA and 5-HT in the brain, particularly in the forebrain structures. Lesions of the V M H caused greater decrease of NE compared to that of the L H (Poncey et al., 1972). After MFB lesions, NE and 5-HT levels were shown to decrease (by 50-70%) in various forebrain structures (e.g., cortex, septum, striatum, amygdala, hippocampus), but not in the diencephalon or the brainstem (see Moore, 1970). I n rats with unilateral lesions of the LH, thcir ipsilateral telencephalon (but not the contralateral part or the brainstem) specifically showed a decrease (approximately by 50%) of the in vitro uptake of labeled NE in its synaptosome-rich homogenate along with a decline in its endogenous NE level. The uptake of NE by telencephalic homogenates was not affected by lesions of the central gray area of the mesencephalon, although it caused a 27% decrease in endogenous telencephalic 5-HT concentration. These results demonstrate a specific functional relationship between the lateral hypothalamus and the NE-containing fibers throughout the telencephalon (Zigmond et al., 1971). Stimulation of the anterolateral hypothalamus induced behavioral manifestations such as vocalization, defense position and signs of fear and escape along with an increase in both 5-HT and 5-HIAA in the forebrain and the brainstem. This increase in 5-HT could not be due to presence of the 5-HT neurons in the hypothalamus, since stimulation of other areas (e.g., frontal cortex, hippocampus, and thalamus) containing 5-HT terminals caused marked increase of 5-HIAA level in the forebrain and brainstem without changing their 5-HT contents (Kostowski and Giacalone, 1969). With lesions of the nigrostriatal bundle (NSB) whose fibers pass through the medial portions of the internal capsule close to the lateral hypothalamus, as compared to MFB lesions, a more severe aphagia, adipsia, and disturbance of water regulation along with greater decrease in the contents of DA (particularly in the neostriatum) and NE and less decrease in 5-HT content in the telencephalon were seen (Oltmans and Harvey, 1972).
3. Mesencephalon Serotonergic neurons originate in the nuclei of the medial ( M R ) or dorsal (DR) raphe of the mesencephalon and, with large projections, terminate in various forebrain areas, such as neocortex, limbic forebrain, striatum, thalamus, and hypothalamus (AndCn et al., 1965a,b; Fuxe et al., 1968). Electric stimulation of the M R in rats induced electroencephalographic patterns of sleep and behavioral signs of calmness (Kostowski et al., 1969). Its stimulation also caused a decrease of 5-HT along with an increase of 5-HIAA (Aghajanian et al., 1967; Kostowski and Giacalone, 1969) and an increase of 5-HT turnover in the forebrain of rats without any change
228
S.
N. PRADHAN
in the posterior part of the brain or spinal cord (Gumulka et al., 1969). Stimulation of the nucleus of the D R had little effect on the forebrain 5-HT, while there was an increase of forebrain 5-HIAA (Gumulka et al., 1971). Lesions of the M R also reduced the level of 5-HT and 5-HIAA in the brain and the spinal cord (Giacalone and Kostowski, 1968), and caused a persistent locomotor hyperactivity associated with behavioral and EEG arousal in rats (Kostowski et al., 1968). MR lesions also decreased slowwave sleep periods in cats (Jouvet, 1968). Compared to PCPA-treated rats, the MR-lesioned rats, in spite of comparable lowering of brain 5-HT and 5-HIAA, failed to display exaggerated sexual or aggressive behavior (Sheard, 1973). Unilateral lesions in the ventromedial tegmental areas that caused a severe loss of cells in the pars compacta of the ipsilateral substantia nigra was associated with a low concentration (18-42% of that of the intact side) of DA and NE of the corresponding striatum in monkeys. The CA concentrations were not altered in the absence of any cellular change in the pars compacta of the substantia nigra. This suggests that the latter structure normally exerts, through its efferent nervous pathways, a direct influence on the CA concentrations of the corresponding striatum (Poirier and Sourkes, 1965). Ventromedial as well as dorsolateral lesions in the midbrain tegrnentum also caused a decrease of NE content in the whole brain and also in discrete forebrain areas (e.g., cortex, septum and caudate, amygdala, and hippocampus) . I n addition, ventral tegmental lesions also caused a decrease of whole brain 5-HT content (see Moore, 1970). I n summary, attempts have been made for a neurochemical correlation of aggressive manifestations induced by stimulation or ablation of a number of brain areas. Pooling of some data from Tables I1 and I11 has permitted some correlation in a few cases, as summarized in Table IIIa. I t appears that during sham rage in cats, and muricide or other aggressive manifestations in rats, as induced by brain manipulations, NE content of the forebrain is usually decreased and, in lesion experiments in rats, that of brainstem or hindbrain is increased. I t is difficult to ascertain whether these effects are specific to aggression or are nonspecific to resulting acute stress. Lack of adequate information does not permit any further generalized conclusion for some correlation in other situations a t this stage.
C. NEUROCHEMICAL CORRELATES OF AGGRESSION In this section discussion will be mainly restricted to changes in the endogenous levels or turnover rates of the neurotransmitters in the whole or parts of the brain during development or manifestation of different types of aggressive behavior. This discussion does not include changes associated
229
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
TABLE IIIa NEUROCHEMICAL CORRELATES OF AGGRESSION INDUCED BY MANIPULATIONS OF BRAINAREAS(A SUMMARY)
Subject Cat Cat Rat Rat Rat
a
Brain manipulation
Aggressive manifestations
Amygdala, ESa Hypothalamus, ES Olfactory bulb, L Septum, L VMH, L
Sham rage Sham rage Muricide Muricide Muricide Shock-induced aggression
Brain NE
1 FB 1 BS
1 FB, 1 BS 1 TLC, BS O l L FB, T HB(T+)
1WB
1WB
Abbreviations and symbols as in Table 111.
with aggression modified by neuroanatomical, neurophysiological, or neuropharmacological manipulations, that are the subject matters of other sections. Available relevant data are summarized in Table IV.
1. Isolation-Induced Aggression I n most of the investigations dealing with this type of aggression, mice have been used, and their whole brain has been taken for neurochemical estimation. Such aggression does not usually appear to have any significant effect on the brain levels of NE, DA, or 5-HT (DaVanzo et al., 1966; Garattini et al., 1969) except that an increase in 5-HT level was observed in one study (Welch and Welch, 1968). However, the turnover rates of NE, DA, and 5-HT are slower in the isolated mice than in those housed in groups (see Welch and Welch, 1969; Modigh, 1973). This effect in mice appears to be strain-dependent, certain strains (e.g., MF-1 albino) being more sensitive (Goldberg et al., 1973). During isolation, the characteristic biochemical changes occur only in the animals which become aggressive, and the changes have different time courses. Thus, while the brain 5-HT level usually remains unaffected, 5-HT turnover rate and 5-HIAA level decreases in the brain beginning as early as the first day of isolation, while the brain level of N-acetyl-L-aspartic acid decreases progressively with period of isolation. O n the other hand, the brain levels of 5-HT, NE, DA and glutamic acid, or M A 0 and choline acetyltransferase remain unchanged during prolonged isolation (Garattini et al., 1969; Consolo and Valzelli, 1970; Giacalone and Kostowski, 1968). The maintenance of the brain level of 5-HT with decrease in 5-HT turnover rate and 5-HIAA level is probably accomplished by a decrease in metabolism of 5-HT.
TABLE I V NEUROCHEMICAL CORRELATES OF AGGRESSION^ Effects of brain amines NE Types of aggression IIA
Subjectb Mouse MF-l,albino,aggr. CF-l,albino,aggr. DBA/2J,gray,aggr. Mouse
+
+
I IA IIA
I IA Isolation fighting SIA Muricide
Mouse Mouse Mouse Rat Rat
C
T
0 0
0
0 0
- T-
DA
5-HT
Remarks
References
C-, T+
No clear relationship between behavior and T of NE
Goldberg et al. (1973)
0
O
co
0
- TT+
+ O/+ + +
TCO CO, T-
T+ CO, TO
A. S. Welch and Welch, (1968), B. L. Welch and Welch, (1969) DaVanzo et al. (1966) Garattini et 01. (1969) Consolo and Valzelli (1970) Modigh (1973)
Corticosterone O 5-HIAA -, ChA 0 AChE -, M A 0 0 5-HIAA -
+,
+
Serum DBH TH Changes in forebrain only ChA+, AChE 0 (in amygdala)
Lamprecht et al. (1972) Goldberg and Salama (1969), Salama and Goldberg (1970) Eble et al. (1973)
Abbreviations and symbols: AChE, acetylcholinesterase; aggr., aggression; C, level/concentration; ChA, choline acetyltransferase; DBH, dopamine-8-hydroxylase; IIA, isolation-induced aggression; SIA, shock-induced aggression; T, turnover rate; TH, tyrosine hydroxincrease; -, ylase (in hypothalamus); 5-HIAA, 5-hydroxyindoleacetic acid; MAO, monoamine oxidase; NE, norepinephrine; decrease; 0, no change. Usually the mouse whole brain has been used for neurochemical estimations in IIA.
+,
AGGRESSION A N D CENTRAL NEUROTRANSMITTERS
231
Moderate fighting among pairs of isolation-induced aggressive mice elicited within 10 minutes elevation of brain NE, DA, and 5-HT (Welch and Welch, 1971) , whereas more intense fighting appeared to lower the concentrations of NE and DA in the brainstem (Welch and Welch, 1969) and also to accelerate the turnover rates of CA and 5-HT in the brain (Modigh, 1973). During isolation-induced fighting in mice there was a decrease in the M A 0 activity in several brain areas preceded by its increase in the hypothalamus during the first 2 days of fighting (Eleftheriou and Boehlke, 1967). There was a n increase in adrenal TH activity (Maegnwyn-Davis et al., 1973). Although it is difficult to integrate these neurochemical data, it appears that during isolation the impulse flow in the central NE and DA neurons is probably retarded, and decrease in CA synthesis in isolated animals is likely to be causally related to the lowered nervous activity. During fighting which acts as an acute stress there is an increased demand for CA transmitters, which is mainly compensated by increase in their synthesis and also probably by partial inhibition of mitochondria1 MAO. Metabolism of 5-HT appears to be decreased, and ACh level remains unaffected. Further clarification is needed with respect to correlation of sequential development of various behavioral and neurochemical manifestations during prolonged isolation and induced aggression.
2. Muricide NE level was found to increase (by 25%) in the forebrain of killer rats. There was no significant difference between killer and nonkiller rats on the levels of NE in the hindbrain, or on 5-HT levels in the forebrain or hindbrain. The killer rats also showed a higher (52%) synthesis rate of NE in the forebrain (Goldberg and Salama, 1969). The spontaneous killer rats also showed higher choline acetyltransferase activity, but not significant alterations in AChE activity in their amygdala (Ebel et al., 1973).
3. Shock-Induced Aggression Very little neurochemical correlation has been studied in connection with this type of aggression. In one experiment (Lamprecht et al., 1972) rats subjected to 4 weeks of daily periods of immobilization stress showed a significant increase in shock-induced fighting as well as increase in the activity of serum DBH and of hypothalamic T H . The concentration of NE in the hypothalamus was not decreased. Four weeks after termination of immobilization stress DBH activity returned to normal, but the increases in shockinduced fighting as well as hypothalamic T H activity persisted.
232
S.
N. PKADHAN
111. Chemostimulation of Discrete Brain Areas and Induced Aggression
1. Cholinergic Stimulation Emotional excitement and aggressive behavior have been shown to be elicited by electric stimulation of restricted areas of the forebrain and the brainstem (Allikmets, 1974; Delgado, 1964; Flynn, 1967; Girgis, 1971; Karli, 1968). The neuronal pathways involved in aggression extend from the amygdala through the hypothalamus into the periaqueductal gray matter of the midbrain. The septum exerts an inhibitory influence on this behavior as shown by electric stimulation and lesion experiments. I n a recent review Allikmets (1974) has presented data from a large number of investigations to demonstrate that, as in the case of electric stimulation, chemical stimulation of certain areas of the brain can also trigger or modify aggressive reactions. Table V summarizes some of these data and shows that cholinergic stimulants are among the active chemicals that elicit muricide response and other aggressive manifestations. Aggression is elicited by cholinergic stimulation of certain areas of the brain extending from the limbic structures (e.g., amygdala, septum, hippocampus) through the thalamus and the hypothalamus into the midbrain. Microinjection of ACh, with or without an anticholinesterase (e.g., physostigmine), or of carbachol into certain parts of the amygdala in cats and rats induced aggression, attack, hyperreactivity, and vocalization (Baxter, 1967; Grossman, 1963; Hull et al., 1967; Allikmets, 1974). Local application of amitone [diethyl S-(2-diethylaminoethyl) thiophosphate], an anticholinesterase agent in the basolateral amygdala also increased aggressiveness and hyperreactivity, and, in some cases, muricide in rats. The increased hyperreactivity was depressed by atropine (IgiC et al., 1970). Demonstration of an intense cholinesterase activity mainly in the basolateral amygdala and also in the pathways between the LH and amygdala (Girgis, 1972) further supported involvement of a cholinergic mechanism in aggression mediated through the amygdala. Similar responses were also produced by application of these cholinergic agents into the hippocampus and the septum. Muricide responses were elicited by cholinergic stimulation of the thalamus and the hypothalamus (lateral part) in rats and were inhibited by microinjection of atropine into the thalamus (Bandler, 1969, 1970, 1971; Smith et al., 1970). Application of these agents into the medial periventricular part of the hypothalamus and the periaqueductal gray matter of the midbrain also caused generalized aggressive response in cats (see Allikmets, 1974).
AGGRESSION A N D CENTRAL NEUROTRANSMITTERS
233
TABLE V EFFECTOF CHOLINERGIC STIMULATION OF DIFFERENT BRAINAREASO N AGGRESSION Aggressive responses
Brain area Amygdala
Subject Cat
Amygdala Rat (dorsomedial) Amygdala Rat (basolateral) Amygdala Rat (centromedial) Cat Septum
ChemicalY
Muri- Sponcide taneous” AB
Carbachol
Baxter (1967), Grossman (1963), Hull et ul. (1967) ABC See Allikmets (1974) AB See Allikmets (1974)
ACh ACh, physostigmine
+
Amitone, 3 pg NE, methamphetamine ACh, carbachol, physostigmine
Hippocampus
Rat Cat
Amitone, 3 pg Carbachol
Hypothalamus
Cat
Carbachol
-
+
ACh
Hypothalamus (lateral)
Rat
Hypothalamus
Cat
Thalamus
Rat
Midbrain (periaqueductal gray)
Cat
References
Carbachol, neostigmine, ACh physostigmine Carbachol d-Tubocurarine Carbachol Atropine ACh Carbachol
+
+
+
ADc
Igid et al. (1970) Leaf et al. (1969)
ABC See Allikmets (1974), Hernindez-Pe6n c t al. (1963) AD Igid et al. (1970) Baxter (1967), MacLean AC (1957) Baxter (1967), Grossman AB (1965), Myers (1964), Varszegi and Decsi (1967) ABC See Allikmets (1974), Myers (1964), Varszegi and Decsi (1967) Smith et al. (1970), Bandler (1969, 1970) Rage Romaniuk et al. (1973a,b) Fear Bandler (1971) ABC Allikmets (1974) ABC Baxter (1968, 1969), Hernbndez-Pedn ~t al. (1963), Marczinski (1967)
ACh, acetylcholine; NE, norepinephrine. Spontaneous manifestations other than muricide: A, aggression; B, attack; C, vocalization; D, hyperreactivity. Hyperreactivity was depressed by atropine.
2 34
S. N . PRADHAN
Microinjection of carbachol through implanted cannula into various areas in the dorsal or the ventral parts of the anterior and the posterior hypothalamus of cats caused rage reactions. Injection of d-tubocurarine, an “antinicotinic agent” into the same structures produced autonomic, somatic, and behavioral responses characteristic of fear reaction. These evoked responses occurred with similar frequency and intensity in all the stimulated parts of the hypothalamus (Brudzynski et al., 1973; Romaniuk et al., 1973a,b).
2. Cholinergic Stimulation
us Electric Stimulation
Baxter ( 1967) using “chemitrodes” that permit application of either crystalline chemical compound or electric current at the same site, observed in cats that both cholinergic stimulation with carbachol and electric stimulation applied at the same locus in the hypothalamus produced some common responses (salivation, mydriasis, hissing, piloerection) ; however, the author indicated that electric and chemical stimulation applied at the same locus may excite somewhat different neural systems. When carbachol was injected into various amygdaloid and hippocampal sites, from which no emotional behavior could be elicited by electric stimulation, it produced emotional behavior similar to that produced in hypothalamus, further suggesting the involvement of two different neural systems. Allikmets and his associates (see Allikmets, 1974) also observed a similarity in aggressive manifestations elicited by electric stimulation and cholinergic stimulation. However, aggressive reactions elicited by electric stimulation of the hypothalamus were usually more pronounced. I n the hypothalamus and the mesencephalic gray matter, the areas eliciting aggressiveness in response to electric stimulation were wider than those responding to cholinergic stimulation. Thus, while aggressive-defensive responses were elicited by electric stimulation of 80 points in the hypothalamus and 55 points in the mesencephalon of cats, similar emotional responses were produced only in 60% and 55% of the cases after chemostimulation. The cholinoceptive points producing aggressive responses were mainly in the medial periventricular part of the hypothalamus extending from supraoptic area to the mammillary bodies, and in the central gray matter of mesencephalon. The points that failed to produce aggression on cholinergic stimulation in the hypothalamus were in the lateral parts, microinjection of ACh into which elicited sleep and inhibited behavioral responses. I n the midbrain such points were located laterally or ventrally to the central gray matter, and their cholinergic stimulation caused alerting and orienting responses. I t thus appears that although behavioral manifestations of cholinergic stimulation of certain areas resemble those of electric stimulation of these areas in some aspects, they differ in others. This may be due to the fact
AGGRESSION A N D CENTRAL N E U R O T R A N S M I T T E R S
235
that while the electric current stimulates the nerve cell and fibers, the cholinergic drugs act on the receptors of the postsynaptic membrane in the region of injection or, as mentioned earlier, different neural systems may be involved in these types of stimulation.
3. Noncholinergic Stimulation While ACh produced aggressive manifestations following its local application in many areas of the brain, other putative neurotransmitters had limited effects. Thus 5-HT, when injected into points in the medial part of the hypothalamus and the midbrain that showed cholinergically induced aggressive responses, evoked only vocalization that was weaker and shorter and could be suppressed by benactyzine. It appears that this effect of 5-HT is probably nonspecific. NE injected into the amygdala, septum, hypothalamus, or midbrain failed to produce any aggressive or defensive manifestations (see Allikmets, 1974). However, mouse-killing behavior has been shown to be inhibited by injection of NE, &methamphetamine (or also methylscopolamine) into medial and central amygdala (Leaf et al., 1969) and also by intraamygdalar injection of imipramine and chlorpromazine (Horovitz and Leaf,
1967). 4. Pharmacological Modification
of Cholinergic Stimulation of
Hypothalamus Studies were aimed at pharmacological modification of the aggressive manifestations elicited by microinjection of ACh into the periventricular part of the hypothalamus in cats. For this purpose, drugs affecting various neurotransmitter (e.g., ACh, NE, and 5-HT) mechanisms were used, as shown in Table VI. I t was observed that physostigmine, an anticholinesterase agent, shortened the latent period and prolonged the duration of AChinduced aggressive behavior. Conversely, drugs with marked antimuscarinic effects (e.g., atropine, scopolamine, or benactyzine) inhibited or blocked aggression, showing an involvement of m-cholinergic mechanism in this behavior. As mentioned earlier (Brudzynski et al., 1973; Romaniuk, 1973a,b), intrahypothalamic injection of d-tubocurarine and carbachol in cats produced fear and rage, respectively. Concurrent injections of a muscarinic blocker, atropine or a nicotinic blocker, betamon (tetraethylammonium bromide) , showed that carbachol-induced rage reaction was muscarinic in nature. O n the other hand, a d-tubocurarine-induced fear reaction could not be affected by either blockers, showing that the mechanism of this fear reaction was more complex. The authors believe that the whole of the hypothalamus maintains “a defensive system” that is subdivided into two circuits,
236
S. N. PRADHAN
TABLE VI TO VARIOUS NEUROTRANSMITTER MECHANISMS) ON EFFECTSOF DRUGS(RELATED AGGRESSIVE MANIFESTATIONS ELICITEDBY CHOLINERCIC STIMULATION OF THE HYPOTHALAMUS IN CATP Neurotransmitter mechanism involved in drug action Cholinergic
Catecholaminergic
Serotonergic
Drugsb showing effects on ACh-induced aggression Shortened/ suppressed Atropine, 1 Scopolamine, 1.5 Benactyzine, 1 Phentolamine, 10 dl-DOPA, 50 I-DOPA, 100 dl-Amphetamine, 1,2 Imipramine, 5-10 /-Tryptophan, 50
No change Arpanal, 3 Trasentin, 3 Propranonol, 5 Haloperidol, 3
BOL-148, 3
Lengthened/ intensified Physostigmine, 0.15
d-DOPA, 25 Iproniazide, 25 f DOPA, 50 Iproniazide, 25 Methysergide, 3
Modified from the table of Allikmets (1974). Drugs (mg/kg) were given intramuscularly or subcutaneously. DOPA, dihydroxyphenylalanine.
one for fear and the other for rage, both of which are mediated through the cholinergic mechanism, but of different natures. Effects of drugs related to catecholaminergic mechanisms, as shown in Table VI, are difficult to correlate and not clear. On the other hand, from the data in the table as well as those from other studies (Palermo and Carlini, 1972; MacDonnell et al., 1971), it appears that aggressive behavior is enhanced by a decrease in effective brain serotonin level and is inhibited by its increase. Thus, a functional antagonism appears to exist between cholinergic and serotonergic systems in the hypothalamus with respect to aggressive behavior. This fact is further corroborated by antagonism of the effects of carbachol on emotional behavior by intrahypothalamic microinjection of 5-HT (MacDonnell et al., 197 1) . Such functional antagonism between these two systems involving aggression at the level of the amygdala has also been observed (Allikmets, 1974). In summary, muricide and other aggressive manifestations have been shown to be elicited by cholinergic stimulation of certain discrete areas of the brain (eg., amygdala, septum, hippocampus, hypothalamus, thalamus, periaqueductal gray of the midbrain) . These manifestations can be inhibited by anticholinergic agents and further facilitated by anticholinesterases. Such manifestations resemble those elicited by electric stimulation of many of the
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
237
same areas, although there exist some points of difference between the two types. Adrenergic and serotonergic stimulations of these areas that have neither been promising nor intensively investigated, appear to produce inhibitory effects in some cases.
IV. Neurophamacological Manipulation of Aggression
I n the previous sections, attempts have been made to correlate different types of aggression (manifested spontaneously or induced by neuroanatomical manipulations) with the associated neurochemical changes. Further investigations on the neurochemical correlates of aggression have been extensively made by using pharmacological agents as experimental tools. These agents modify the metabolism and/or effects of specific neurotransmitters by acting at different stages (e.g., synthesis, storage, release, action, and disposal) of their life history and concomitantly alter the behavioral responses. The following discussion will be directed toward the effects on various types of aggression of such agents related to central noradrenergic, dopaminergic, cholinergic, or serotonergic mechanism.
A. DRUGSRELATED TO NORADRENERGIC MECHANISM 1. Isolation-Induced Aggression The effects of drugs related to noradrenergic mechanism on isolationinduced aggression have been somewhat confusing. From Table VII it can be observed that this type of aggression is inhibited by drugs that increase the concentration of NE at specific receptor sites by increasing its release (e.g., amphetamines), by decreasing its reentery into the storage sites (e.g., cocaine, imipramine, and related antidepressants) , or by preventing its oxidative destruction (e.g., M A 0 inhibitors). O n a few occasions, there were either no effects or even enhancement of aggression. These variations were due to such factors as drug dosage, method of evaluation of the effect, etc. Thus Welch and Welch (1969) reported dose-dependent biphasic effects of amphetamine and pargyline ( a M A 0 inhibitor) causing an increase of isolation-induced fighting at their low doses and a decrease at high doses. Furthermore, for evaluation of effective and nontoxic antiaggressive dose of an agent, the neurotoxic dose ( N T D 50) has been taken into consideration; since N T D 50 has been assayed by different investigators by different methods (DaVanzo et al., 1966; Sofia, 1969a), there have been some variations in evaluation of drug effect on aggression.
I0
w
TABLE VII EFFECTS OF DRUGS RELATED TO NORADRENERGIC MECHANISM ON VARIOUS TYPES OF AGGRESSION^
co
Effects*on aggression Dose Drugs
Adrenergic stimulants Amphetamine
Methamphetamine Cocaine M A 0 inhibitors Iproniazid Tranylcypromine Phenelzine
Pheniprazine Pargyline
Mg/kg
Route
1.5 i.p. 2 i.p. 3 i.p. >3 i.p. 5 i.p. 5 i.p. (2 X weekly) 6 10 i.p. 15 i.p. 2 >4 5 155 10 5 20 33.4 5-1 0
i.p. i.p. i.p. i.p. i.p. i.p.
50
i.p.
100
i.p.
Isolationinduced
+++ (1)
- - (3,s)
Shockinduced
Muricide
W"
v
I I
m
h
W N
r(
N
-
h
+
v
h
v 3
0
v
2
m
h
0
v
I
v
+
s l
v
l
h
0
I
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
h
-
v
0
h
z
o\ v
I
I
m
h
239
f f
c 'Z
0
TABLE VII (Continued) ~
~~~
Effectsa on aggression Dose Mg/k
Drugs Adrenergic depressants 6-OHDA FLA-63
Route
Isolationinduced
Shockinduced
Muricide
200 Pg i.c. 25 i.p. (daily X 3) 20 i.p. 10 i.p.
Propranolol Phentolamine
v
F
a FLA-63, bis(4-methyl-1-homopiperazinylthiocarbonyl)disulfide; aMT, a-methyl-fi-tyrosine; 6-OHDA, 6-hydroxydopamine; 6OHDOPA, 6-hydroxy-3,4-dihydroxyphenylalanine; i.c., intracisternal; i.p., intraperitoneal; i.vent., intraventricular; S.C. subcutaneous. 0, and - indicate increase, no change, and decrease, respectively. Two or three or - signs indicate changes by 26-50% or 51-75 %, respectively. One or - does not always provide quantitative representation. Number within parentheses indicates reference: (1) Welch and Welch, 1969; (2) DaVanzo et al., 1966; (3) S. Garattini and L. Valzelli, unpublished observations, 1966 (quoted in Valzelli, 1967); (4) Melander, 1960; (5) Valzelli et al., 1967; (6) Sofia, 1969a,b; (7) Cook and Weidley, 1960; (8) Yen et al., 1959; (9) Weischer, 1969; (10) Thoa et al., 1972a; (11) La1 et al., 1968; (12) Brunaud and Sou, 1959; (13) Lapin, 1967; (14) Navarro, 1960; (15) Tedeschi et nl., 1959; (16) Chen et nl., 1963; (17) Hingtgen and Hamm, 1969; (18) Thoa et al., 1972b; (19) Eichelman et al., 1972; (20) Stern et al., 1972; (21) Leaf et al., 1969; (22) Horovitz et al., 1965; (23) Horovitz rt al., 1966; (24) Karli, 1959a; (25) Karli, 1960; (26) Karli, 1959b; (27) Karli, 1958. Neurotoxic dose. Dose produced motor inactivation.
+,
+
+
2
AGGRESSION A N D CENTRAL NEUROTRANSMITTERS
24 1
Paradoxically, several agents that reduce the effective concentrations of NE at the adrenergic receptor sites [e.g., reserpine, which depletes catecholamines storage in tissues; lithium, which decreases brain NE content, probably by increasing its turnover and also by its deamination (Schildkraut et al., 1969) ; (u-MT, which inhibits CA synthesis; and phentolamine and propranonol, which block (u- and ,8-adrenergic receptors, respectively] also inhibit isolation-induced aggression. I n summary, the drugs having either stimulant or depressant effects on noradrenergic mechanism appear to depress isolation-induced aggression.
2 . Shock-Induced Aggression The effects of adrenergic agents on shock-induced aggression have also been confusing to some extent (Table V I I ) . This type of aggression was enhanced by adrenergic stimulants, such as amphetamine and cocaine, but not by an antidepressant, such as imipramine. Amphetamine has also been shown to reduce such aggression at lower doses, probably owing to its analgesic effects (La1 et al., 1968). On the other hand, 6-OHDA-treated rats showed more marked foot-shock-induced aggression when tested between 57 and 106 days after its i.c. injection (Stern et al., 1972). Marked and persistent facilitation of such aggression was demonstrated 3 4 days after a single i.c. dose of 6-OHDA in rats. This effect appeared to correspond with induced depletion of brain DA and NE and degeneration of DA and NE nerve terminals. A subsequent additional dose further increased the attack rates. However, 6-OHDA failed to show any effect on muricidal activity and jump thresholds (Eichelman et al., 1972). Facilitation of this type of aggression was also observed 4 days after administration of 6-OHDOPA, when brain NE but not DA, was reduced (Thoa et al., 197213). A decrease in shock-induced aggression was caused by CA depletors such as reserpine and tetrabenazine, but CA-synthesis blockers (e.g., a-MT) did not show any effect.
3. Muricide Response From Table VII, it appears that muricide responses can be inhibited by adrenergic stimulants, e.g., amphetamine, M A 0 inhibitors, and antidepressants. Sites of such action of the adrenergic agents have been demonstrated to a certain extent. Thus, lesions of the amygdala eliminated mousekilling in killer rats, and injections of NE, amphetamine, or imipramine directly into the central and medial amygdala temporarily suppressed the killing response (Horovitz and Leaf, 1967; Leaf et al., 1969; Leaf, 1970). Injection of NE into the lateral hypothalamus also suppressed the muricide response (Bandler, 1969). O n the other hand, (u-MT was shown to initiate killing in some nonkillers
242
S . N. PRADHAN
and also to block partially or completely amphetamine-induced inhibition of muricide response ; however, it failed to alter muricide inhibition induced by thiazesim (an antidepressant) or tripelennamine (an antihistaminic) both having no effect on DA metabolism. Thus these results suggest a noncatecholaminergic inhibition of muricide responses (Leaf et al., 1969). Furthermore, reserpine which causes CA depletion failed to affect the muricide response (Table VII) showing that the noradrenergic mechanism probably has limited control on this response. In summary, muricide behavior is inhibited by adrenergic stimulants and at least partially evoked or facilitated by adrenergic depressants, thus suggesting an inhibitory role of the adrenergic mechanism.
B. DRUGSRELATEDTO DOPAMINERGIC MECHANISM A role of DA in aggressive behavior has been implicated directly from genetic studies, as well as indirectly and extensively through pharmacological investigations. Everett (1968) studied two strains of mice, C57BL65 and BALB, both of which were found to have elevated brain CA levels. BALB mice which are spontaneously aggressive, had comparatively higher levels of DA, whereas the tamer C57BL65 mice showed higher levels of NE. The role of DA in aggressive behavior has been more extensively substantiated by pharmacological studies, results of which are discussed below. These studies have been done to modify the central dopaminergic activity in at least two ways by using drugs and chemicals. First, by the concentration or activity of DA at the specific central receptor sites can be increased. This has been possible by administration of its precursor DOPA that can pass through the blood-brain barrier. The concentration of DOPA within the CNS can be enhanced by using a peripheral decarboxylase inhibitor like R04-4602. DA activity can also be modified by combination with various drugs. Such data are summarized in Table VIII. Second, the central dopaminergic mechanism can be affected by some drugs (e.g., amphetamine, apomorphine) that would act on or through a central dopaminergic mechanism or would act directly on the central DA receptors. Such data are provided in Table IX.
1. DOPA Alone or in Combination with Other Drugs a. DOPA Alone or in Combination with a Decarboxylase Inhibitor. Administered in low doses, DOPA produced no effect or a sleeplike state (Rizzoli et al., 1969; Bryson and Bischoff, 1971) ; however, in high doses it evoked a number of neurological and behavioral changes which sometimes resemble aggression-like manifestations. I n mice, such manifestations include
243
AGGRESSION A N D C E N T R A L N E U R O T R A N S M I T T E R S
TABLE V I I I OF DOPA ALONEOR IN COMBINATION WITH EFFECTON AGGRESSION
DOPAa L, L,
Adjuvantb
50, i.p. 85-470, i.p.
Subject Mouse Mouse
No effect Sleeplike state
Mouse
Gnawing, biting, fighting, scatterjump syndrome Viciousness, biting
500-970, i.p.
DL,
400-500, i.v.
L,
1000, i.p.
Mouse
L,
250, i.v.
Moused
DL,
L,
6-62, i.p.
200, i.p.
DL,
400, i.p.
DL,
500, i.v.
R04-4602, 52, i.p., 10 min prior
Rat
R04-4602, 50, i.p., 40 min prior
Rat
d-amphetamine, 2, i.p.
Mouse
200, i.p.
Iproniazid, 62, Pargyline, 200
50
Pargyline, 100
L,
200-400
Pargyline, 100
L,
25,
Pargyline, 150 or Niamid, 500, S.C. 45-50 min prior
L,
S.C.
Aggressive manifestations
OTHER
DRUGS
References Rizzoli et al. (1969) Bryson and Bischoff (1971)
Vander Wende and Spoerlein (1962) Blaschko and Hyperactivity, Chruiciel (1960) jumping, fighting Kletzkin (1969) Biting and reactivity increased, but fighting decreased Lammers and Van Bizarre social beRossum (1968) havior, reduced by haloperidol (0.4, Lp.) Benkert et al. Fighting, potenti(1973a) ated by reserpine (2 and 16 hours prior) La1 et al. (1970b) Viciousness
MouseC Biting, potentiated by iproniazid or tranylcypromine Mouse Excitement, attack directed to a foreign object Mouse Aggression reduced Mouse Isolation-induced aggression reduced Rat Rage reactions (hissing, spitting, fighting) followed by stereotypy (sniffing, licking, biting of cage)
Yen et al. (1970)
Everett (1961), Everett and Weigund (1963) Karczmar and Scudder (1969) Welch and Welch (1969) Scheel-Kriiger and Randrup (1967), Randrup and Munkvad (1969)
(Continued)
244
S. N. PRADHAN
TABLE VIII (Continued) DOPA“ L,
20
L,
100, i.p.
L,
200-500
Adjuvantb Pheniprazine, 20, 90 min prior Reserpine Disulfiram, 200500 6-OHDA,d 200 pg, i.c., 4 days prior 4-M K 486, 10, i.p. 30 min prior PCPA-ME,d 400, i.p. daily, X4
Subject Cat
Rat
Mouse
Aggressive manifestations
References
Behavioral Reis (1972) excitement Latency of response shortened Response attenuated . Spontaneous Thoa et ~ l (1972~1 fighting
Excitement, jumpiness, fighting
Lycke et al. (1969)
a Optical isomer (L or DL, wherever known); dose (in mg/kg, unless otherwise mentioned), route of administration. * Drugs, dose (mg/kg), route, and time of administration; i.p., intraperitoneal; i.v., intravenous; s.c., subcutaneous. Three groups: normal, septal, isolated. 6-OHDA, 6-hydroxydopamine. ME, methyl ester.
gnawing, biting, jumping, viciousness, hyperreactivity, and fighting (Bryson and Bischoff, 1971; Vander Wende and Spoerlein, 1962; Yen et nl., 1970; Blaschko and Chruiciel, 1960; Kletzkin, 1969). In rats pretreated with a decarboxylase inhibitor (RO 4-4602), DOPA in small doses induced a “bizarre social behavior” that was reduced by haloperidol (Lammers and Van Rossum, 1968). Furthermore, DOPA-induced fighting has been shown to be potentiated by pretreatment with reserpine (Benkert et al., 1973a). b. Combination with M A 0 Inhibitors. Administration of M A 0 inhibitors (e.g., iproniazid, pargyline) alone and in combination with DL-DOPA produced graded increased alertness, responsiveness, irritability, and aggressiveness in mice. These behavioral responses have been correlated with the increasing degrees of M A 0 inhibition and the concomitant increase of both DA and NE in the brain (Everett, 1961; Everett and Weigund, 1963). DOPA-induced biting responses can be potentiated by M A 0 inhibitors (e.g., iproniazid and tranylcypromine) and suppressed by haloperidol, chlorpromazine, bretylium, and several other drugs (Yen et al., 1970). In rats, injection of a M A 0 inhibitor and DOPA also produced rage reactions (hissing, spitting, fighting) followed by stereotyped behavior (continuous sniffing, licking, biting) . Pretreatment with reserpine and DDC (diethyldithiocarbamate, a DBH-inhibitor that strongly inhibits formation
AGGRESSION A N D CENTRAL NEUROTRANSMITTERS
245
TABLE IX AGGRESSION AND DRUGACTIONS POSSIBLY INVOLVING DOPAMINE RECEPTORS"
Drugs
DDCb
+
Subject
Rat
Aggression increased
Mice
Aggressive activities and stereotyped sniffing, licking and biting of the cage Fighting
pargyline &Amphetamine, 15, i.p.
Apomorphine Rat 2.5, S.C. Apomorphine, 1, Rat i.v. Apomorphine, Rat 5-30, S.C.
Amantadine, 100 Depressed + 300 mg/ patients day, 10 days Morphine withdrawal
Aggressive manifestations
Rat
Intraspecific aggression Spontaneous fighting; increase shockinduced fighting Motor restlessness, aggression, hostile attack Aggression (attack/bite, rearing, vocalization)
Modification
References
Scheel-Kriiger and Randrup (1968) Aggression inhib- Hasselager el al. ited by small (1972) doses of spiramid or trifluperazine Schneider (1968) Antagonized by atropine Potentiated by reserpine
Senault (1970) McKenzie (1971)
Rizzo and Morselli (1972)
Increased by L-DOPA, (50), DL-DOPA (200), d-amphetamine (2), apomorphine (1.25) ; Decreased by haloperidol (0.63-2.5), a-MT (200), lesion of nigrostriatal bundle
Abbreviations: DOPA, dihydroxyphenylalanine; i.p., intraperitoneal; i.v., intravenous; a-MT, a-methyl-p-tyrosine; s.c., subcutaneous. Diethyldithiocarbamate, 500, 500, 50, 500 mg/kg a t 2 hour intervals with pargyline, 150 mg/kg, s.c., 2 hours prior.
246
S . N. P R A D H A N
of NE, but not that of DA) reduced the rage reaction, while stereotypy remained unaffected (Scheel-Kruger and Randrup, 1967 ; Randrup and Munkvad, 1969). DOPA-induced excitement in cats pretreated with a MAO-inhibitor ( pheniprazine) could also be attenuated by disulfiram, another DBH-inhibitor (Reis, 1972). Although aggressive behavior has been shown to be associated with increased activity of brain NE (Gunne and Lewander, 1966; Scheel-Kruger and Randrup, 1967; Reis, 1972), it has also been produced when DDC is given in combination with a M A 0 inhibitor (Scheel-Kriiger and Randrup, 1968). This indicates that production of rage or aggression is dependent on brain NE and DA. Presence of both of these neurotransmitters in balanced condition may be essential for this behavior. The effect of DOPA, however, appears to be different when administered to subjects already manifesting aggression that may be spontaneous or induced. Thus Karczmar and Scudder (1969) using “Mouse City” aggrega.ted male mice that manifested different types of spontaneous aggression, found marked reduction in the behavior following administration of DOPA alone or in combination with a M A 0 inhibitor, pargyline. The low dose of DOPA and differences in experimental procedures or conditions may be additional factors accounting for this difference in effects. Welch and Welch (1969) examined the effects of pargyline followed by high doses of DL-DOPA on male mice made aggressive by long-term isolation. These animals appeared aggressive and sounded as though they were fighting, but exhibited absolutely no coordinated aggressive activity. I n their most excited stage, they were often incapable of biting, even if the experimenter tried to put his finger into their mouths. c . 6-Hydroxydopamine ( 6 - O H D A ). 6-OHDA administered i.c. produces degeneration of CA nerve terminals and depletion of brain GAS (Bloom et al., 1969; Uretsky and Iversen, 1970; Breese and Traylor, 1970). I t causes an increase in foot-shock-induced fighting in rats along with a decrease in brain CAs. Both shock-induced fighting and CA depletion were reduced by desmethylimipramine. Inhibition of CA synthesis by either a-MT or FLA-63 failed to increase shock-induced fighting. L-DOPA in combination with a decarboxylase inhibitor (MI( 486) suppressed fighting facilitated by 6-OHDA; however, spontaneous fighting often occurred and persisted up to 2 hours after injection (Thoa et al., 1972a,c). Thus, drug-induced fighting appeared to be different from shock-induced fighting in certain aspects, although both types were influenced by dopaminergic mechanism. In some rats treated with 6-OHDA (100 pg, Lvent.), mock fighting behavior was observed after administration of 1.5 mg/kg of (+)amphetamine
AGGRI<SSION A N D CICNTRAL NEUROTRANSMITTERS
247
(Evetts et al., 1970). This behavior resembled the “bizarre social behavior” observed in rats treated with L-DOPA and R O 4-4602 (Lammers and Van ROSSUIII,1968). d . Serotonin Synthesis Inhibitors. Aggressive effects of L-DOPA were further demonstrated when given in combination with 5-HT synthesis inhibitors. Pronounced excitement, but not aggression, has been reported in mice treated with L-DOPA (200 mg/kg) along with PCPA (ChruSciel and Herman, 1969). Higher doses (200-500 mg/kg) of L-DOPA given 7 or more hours after injection of H 69/71 (methyl ester hydrochloride of DL-PCPA) produced an immensely aggressive behavior (sympathetic stimulation, rise of tail and neck, fighting posture, actual fight) in mice. Similar aggression was also observed following administration of L-DOPA combined with H 22/54 (another effective inhibitor of 5-HT synthesis, with a slight catechol-o-methyltransferase inhibitory effect) . Thus, aggressive behavior is induced not only by an increase in DA activity, but also by a decrease in 5-HT activity and thus by a disturbance in the balance of the activities of these neurotransmitters (Lycke et al., 1969).
2. Drugs Possibly Affecting Central Dopaminergic Mechanism The role of DA in elicitation of aggressiveness has been further explored indirectly through the use of some drugs (e.g., amphetamine, apomorphine) that have been considered to affect the central dopaminergic mechanism (Table I X ) . a. Amphetamine. d-Amphetamine, which affects both NE and DA in the brain, elicited both aggressive activities and stereotypy (e.g., sniffing, licking, and biting of the cage) in mice at a dose of 15 mg/kg. Small doses of the neuroleptics, spiramide (0.075 mg/kg) and trifluperazine (0.15 mg/kg) caused selective inhibition of aggressive activities without general sedation. Furthermore, aggressive behavior was depressed (as in the case of neuroleptics) by a-MT, which blocks synthesis of both NE and DA, but was not much affected by noradrenergic blocking agent (e.g., phenoxybenzamine or aceperone) or FLA-63, thus suggesting that aggressive behavior might be mediated by an increase in DA activity in the brain (Hasselager et al., 1972) . In mice pretreated with DOPA, amphetamine-induced aggression was enhanced (La1 et al., 1970b). b. Apomorphine. Apomorphine is considered to be a specific agonist at central dopaminergic receptors (Ernst, 1969; Roos, 1969; Ungerstedt et al., 1969). I t reduces the rate of disappearance of DA, but not that of NE, in the brain after TH inhibition, suggesting that apomorphine acts by stimulating DA receptors (Butcher and AndCn, 1969). It also induces aggressive behavior in rats (Schneider, 1968; Senault, 1970). This induced behavior
248
S . N . PRADHAN
can be enhanced by isolation especially in an opacified box (Senault, 1971) and by lesions of the septa1 region and can be inhibited by lesions of the amygdala or the substantia nigra (Senault, 1973), and by drugs, such as neuroleptics (except reserpine) , antianxiety tranquilizers, morphine, and atropine (Senault, 1970). Spontaneous aggression in male rats has been observed following S.C.injection of 10-30 mg/kg of apomorphine. Pain-induced (pinching of tail) aggressive response can be elicited 30 minutes after smaller doses (5-10 mg/kg) of apomorphine. After reserpine pretreatment, still smaller doses (0.5-2.5 mg/kg S.C. ) of apomorphine can induce spontaneous aggression, the thresholds for which thus appeared to be lowered (McKenzie, 1971). c. Amantadine. Amantadine, an antiviral agent, has been shown to possess antiparkinsonian effects (Schwab et al., 1969). I t has been shown to cause an increase of DA release from the caudate nucleus of cats (von Voightlander and Moore, 1971)) and its use as an antidepressant has been suggested (Vale et al., 1971). Given to depressed patients, it produced a progressive increase of motor restlessness, anxiety, and sudden bursts of violent aggressive behavior with hostile attacks upon ward attendants (Rizzo and Morselli, 1972). d. Morphine Withdrawal. Morphine withdrawal elicits social aggression (rearing, vocalization, attack-bites) in addicted rats. Pretreatment with DOPA, d-amphetamine, or apomorphine enhanced aggression severalfold. a-MT or haloperidol reduced or abolished morphine-withdrawal aggression (MWA), which might or might not be supersensitized by amphetamine treatment. MWA was also blocked by lesion of the dopaminergic nigrostriatal bundle (not by that of the medial forebrain bundle), but was reinstated with a small dose of apomorphine. These results suggest a dopaminergic basis of morphine-withdrawal aggression (Puri and Lal, 1973; Gianutsos et al., 1974). e. Pimozide. At low doses ( 5 0.5 mg/kg) , pimozide, a neuroleptic agent, appears to block DA receptors while not appreciably affecting NE receptors (Janssen et al., 1968; AndCn et al., 1970). In an investigation (Desmedt et al., 1973) concerning dominant-subordinate (D-S) relationship in pairs of rats competing for food, treatment of the D rat (but not S rat) in a pair with 0.16 mg/kg of pimozide weakened the D-S relationship, while that with 0.63 mg/kg dose resulted in a near-complete D-S reversal. The normalizing effect of pimozide on social interaction appears to be due to the inhibition of aggressive behavior and to be mediated through a dopaminergic mechanism. In summary, a dopaminergic mechanism appears to be involved in elicitation and facilitation of various types of aggressive manifestations. Since
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
249
a noradrenergic mechanism has also been shown to have facilitatory effects, and DOPA-induced aggression has been enhanced by a decrease of 5 - H T activity, it appears that a balance between these neurotransmitters is important for the modulation of this behavior.
C. DRUGSRELATEDTO CHOLINERGIC MECHANISM 1. Isolation-Induced Aggression A number of anticholinergic agents have been tested for their effects on isolation-induced aggression. Data in Table X show that anticholinergic (antimuscarinic) agents are markedly effective in reducing isolation-induced aggression. This effect may not be due to motor disabilities or mydriasis (DaVanzo et al., 1966; Janssen et al., 1960; Valzelli et al., 1967). Information on effects of cholinergic agents on such aggression is lacking.
2 . Shock-Induced Aggression Several investigators (Brunaud and Siou, 1959 ; Lapin, 1967 ; Powell et al., 1973) demonstrated inhibitory effects of anticholinergic agents on shock-induced aggression. As in the case of isolation-induced aggression, a similar depressant effect was also reported for cholinolytics of the benactyzine group (“central cholinolytics” ) on shock-induced fighting in mice (Denisenko, 1965) and rats (Allikmets, 1964). Information on the effect of cholinergic agents on such aggression or on the effect of such aggression of ACh metabolism is lacking, Imipramine, iproniazid, and amphetamine lowered the thresholds of shock-induced fighting in rats; cholinolytics had opposite effects (Lapin, 1967 ; Allikmets, 1964). The mechanism of these central cholinolytics on shock-induced aggression is not clear. As in isolation-induced aggression, the effects of these drugs are not due to motor or visual impairment. Inhibition of reticular formationmediated arousal phenomena appears not to be involved, since central sedation was absent. Cholinergic pathways in the medial and lateral hypothalamus that participate in elaboration of a number of behaviors, including aggression, may be involved in the action of cholinolytic drugs (see Powell et al., 1973). 3. Muricide As in the case of isolation-induced and shock-induced aggression, the muricide response is also inhibited by various anticholinergic agents. In addi-
250
S . N . PRADHAN
tion, the effects of several cholinergic agents enhanced this response (Table
X ) . Such effects of cholinergic agents are also corroborated by their local microinjections in amygdala, septum, thalamus, and hypothalamus (Table
V) . From the pharmacological nature of the agonists and antagonists, as also discussed in connection with the data in Table X, m- cholinergic mechanism appears to be involved in the muricide response.
TABLE X MODIFICATIONS OF AGGRESSION BY DRUGSAFFECTING CHOLINERGIC MECHANISM Effectsa on aggression
Drug, dose (mg/kg), routeb
Cholinergic Pilocarpine Tremorine Anticholinergics Atropine,
Isolationinduced
Shockinduced
Muricide
n.k. 60 n.k.
2 i.p. 7 . 9 i.p. 10 i.p. 20 i.p. 29 i.p. Bayer 1433, 0 . 2 5 15, i.p. Benactyzine 30, i.p. 32.7, i.p. 75, oral (14 days) 75, i.p. Benztropine 2 . 8 , i.p. 18, i.p. Scopolamine 1.05 0.5-3 30, i.p. 55, S.C.
Number within parentheses indicates reference: (1) McCarthy, 1966; (2) Avis, 1974; (3) McCarthy, quoted in Avis, 1974; (4) S. Garattini and L. Valzelli, unpublished observations, 1966, quoted in Valzelli, 1967; (5) Valzelli et al., 1967; (6) DaVanzo et a/., 1966; (7) Powell et al., 1973; (8) Horovitz et al., 1966; (9) Lapin, 1967; (10) Hoffrneister et al., 1964; (11) Kreisskott, 1963; (12) Mantegazzini et PI., 1960; (13) Brunuad and S o u , 1959; (15) Karli, 195913; (16) Janssen et al., 1960. 0, and - indicate increase, no change, and decrease, respectively. Two or three minus signs approximately indicate changes by 26-50% or 51-75%, respectively. i.p., intraperitoneal; S.C. subcutaneous; n.k., not known.
+,
25 1
AGGRESSION AND CENTRAL N E U R O T R A N S M I T T E R S
D. DRUGSRELATED TO SEROTONERGIC MECHANISM Available data on effects of drugs related to serotonergic mechanism are summarized in Table XI. 1, Isolation-Induced Aggression
Both 5-HTP and PCPA were shown to decrease isolation-induced fighting in mice (Welch and Welch, 1968; Yen et al., 1959; Benkert e t al., 1973b). Thus, alteration of 5-HT content of the brain in either direction has the same effect, e.g., decrease in fighting, as is also observed in studies with respect to the catecholaminergic mechanism.
2. Shock-Induced Aggression Studies on role of 5-HT in shock-induced aggression have been limited. Brunaud and Siou (1959) reported that a high dose of 5-HT could decrease shock-induced aggression. O n the other hand, PCPA treatment has been TABLE XI MODIFICATION OF AGGRESSION n Y DRUGSAFFECTING SEROTONERGIC MECHANISM Effectsa on aggression Drug, dose (mg/kg), route, interval after injectionb
Isolation induced
Shockinduced
Muricide
5-HT, 20 5-HTP, 200, i.p., 1 hour
PCPA, 400, ME,fii.p., 36 hours 360, i.p., 10 minutes 320, S.C. 100 daily X 6 450, i.p. 320, i.p., 17-24 hour
+
+
100, daily X 3 R04-4602, 50 400 daily X 3 p-Chloroamphetamine, 3, i.p.
+L-DOPA,~~~
a Number within parentheses indicates reference: (1) Brunaud and Siou, 1959; (2) Welch and Welch, 1968; (3) Ersparmer et al., 1960; (4) Kulkarni, 1968; (5) Di Chiara el al., 1971; (6) Conner et al., 1970; (7) Yen et al., 1971; (8) Sheard, 1969; (9) Karli et al., 1969; (10) Benkert et al., 1973b; (11) La1 et al., 1970a. f, 0, - indicate increase, no change, and decrease, respectively. Two or three minus signs approximately indicate changes by 26-50% or 51-75%, respectively. DOPA, dihydroxyphenylalanine; 5-HT, 5-hydroxytryptamine; i.p., intraperitoneal; ME, methyl ester; PCPA, p-chlorophenylalanine; s.c., subcutaneous.
252
S. N . P R A D H A N
reported to cause increased irritability and aggression on handling following drug-induced reduction of brain 5-HT (Koe and Weissman, 1966). Rats depleted of 5-HT either by PCPA treatment (Tenen, 1967) or by brain lesions (Lints and Harvey, 1969) exhibited lowered shock thresholds. However, in contradiction to the finding of a previous study that PCPA would increase shock-induced and other types of aggressive behavior (Conner and Levine, 1969), a later study from the same laboratory failed to show any effect of PCPA on such aggression, even when the brain 5-HT level was reduced to 10% of the control (Conner e t al., 1970). Furthermore, p-chloroamphetamine, that depletes central 5-HT without affecting central NE (Pletscher et al., 1964; Fuller et al., 1965), was shown to decrease markedly the frequency of attack and to increase the mean shock intensity to produce vocalization, stereotyped posture, or fighting (La1 et al., 1970a). Although these findings appear to be confusing, it may only be assumed at this stage that 5-HT plays a role in pain perception and as such in shockinduced aggression.
3. Muricide 5-HT has been shown to decrease the muricide response (Ersparmer e t al., 1960; Kulkarni, 1968; Di Chiara et al., 1971). A decarboxylase inhibitor (R04-4602; 25 mg/kg or more) reduced this inhibitory effect to some extent (Kulkarni, 1970), thus indicating that the site of such action is partially peripheral. On the other hand, in most of the studies (Di Chiara et al., 1971; Sheard, 1969), PCPA, the blocker of 5-HT synthesis, has been shown to increase the muricide response. But Benkert et al. (197313) failed to observe such a response to these drugs in rats. Furthermore, aggressive behavior following administration of PCPA could be suppressed by injection of 5-HTP (Sheard, 1969).
4. Irritability a n d O t h e r Aggressive Manifestations DOPA in combination with RO 4-4602 caused a decrease in isolationinduced fighting in mice, and did not show any muricide response in rats; however, pretreatment with PCPA potentiated the effects of this combination by producing more marked jumping, vocalization, hyperreactivity, and fighting in rats compared to the control or reserpine-treated subjects (Benkert et al., 1973b). As mentioned earlier, Lycke et al. (1969) also observed similar aggressive manifestations after administration of high doses of L-DOPA in mice pretreated with 5-HT synthesis inhibitors. Chronic treatment of rats with p-chloroamphetamine induced a decrease in the cerebral 5-HT levels concomitant with manifestation of “a bizarre social behavior” and an increase of “irritability” (Korf and Kuiper, 1971). These results
AGGRESSION AND CENTRAL NEUROTRANSMITTERS
253
were comparable to the manifestations of destruction of midbrain raphe, an area rich in 5-HT (Kostowski et al., 1968) or to those of depletion of cerebral 5-HT by PCPA (Tenen, 1967). I n cats, PCPA stimulated aggressive manifestations including vocalization and savage attacks against anesthetized rats (MacDonnell et al., 1971). Similar aggression has also been observed by others following PCPA in cats (Ferguson et al., 1970), but not in monkeys (Redmond et al., 1971).
V. Summary and Conclusion
The present discussions on neuroanatomical, neurochemical, and neuropharmacological aspects of several types of aggression have very clearly revealed their heterogeneity and complexity. These analyses have also exposed many gaps in available information pertaining to various aspects of aggression discussed here, especially in relation to their correlation with putative central neurotransmitter mechanisms. Attempts have been made to summarize in Table XI1 the extensive data reviewed here with a view to formulate a profile of central neurotransmitter mechanisms involved in the four types of aggression selected for present discussions. From the table it appears that the noradrenergic mechanism is usually depressed especially in muricide and isolation-induced aggression, and probably stimulated in shock-induced aggression. I t is difficult to decide whether the observed depression of this mechanism is specific to aggression, or occurs nonspecifically as a result of acute stress. Dopaminergic mechanism plays an excitatory role in irritable type of aggression, but in other types its role is uncertain. An appropriate ratio between the concentrations of NE and DA may be essential for proper behavioral expressions. I n addition, cholinergic mechanism appears to play an excitatory role and serotonergic mechanism an inhibitory role on all four types of aggression. Extremely oversimplified as this conclusion appears to be, it calls for further investigation for its verification, modification, and meaningful completion. Probably a few words are appropriate at this stage with respect to the problems in such investigations that might have contributed to the difficulties and deficiencies in this correlative review. This review indicates that the available information is far from being adequate for the intended neurochemical correlation of aggression. Moreover, many discrepencies exist between data collected in studies not only with different approaches (e.g., neurochemical, neuropharmacological) , but also with similar methods of procedure. To avoid many such variabilities, serious attention must be given to scrutiny and standardization of the materials and methods used in the investigation. Species and strain of animals are important sources of varia-
254
S. N. PRADHAN
TABLE XI1 NEUROTRANSMITTER PROFILEOF DIFFERENTTYPES OF AGGRESSION ~~~~
~
Pharmacological manipulations Types of aggression
Neurocheinical NTa
Isolation- N E induced
measurementsC
Facilitated by
Isolation, T fighting, T
+
DA ACh 5-HT
CO
T-, 5-HIAAAdrenergic stimulants: 6-OHDA, 6-OHDOPA
ShockNE induced
Muricide
DA ACh 5-HT NE
C-, T +
DA ACh
CO
5-HT TIrritable"
NE DA ACh 5-HT
Cholinergics (systemic or intracranial) 5-HT synthesis inhibitor
Neurotransmitter
Inhibited by
profileC
Adrenergic stimulants and depressants Antidepressants M A 0 inhibitors (and/or DOPA) Anticholinergics 5-HT precursors and synthesis inhibitors Adrenergic depressants
lt?
Anticholinergics 5-HT precursors NE and antidepressants (intraamygdalar) Adrenergic stimulants (systemic) Antidepressants M A 0 inhibitors An ticholinergics
t
5-HT precursor
L
5-HT precursor
t t 1
1 T? 1T
t?
I? 1
t
CDOPA Cholinergics 5-HT antagonists and synthesis blockers
a Neurotransmitter mechanism: ACh, acetylcholine; DA, dopamine; 5-HT, 5-hydroxytryptamine; NE, norepinephrine. * Includes spontaneous aggressive manifestations other than muricide (e.g., attack, biting, vocalization). (or t), - (or I), or 0 indicates respectively, increase, decrease, or no change. T, turnover; C, concentration; 5-HIAA, 5-hydroxyindoleacetic acid. DOPA, dihydroxyphenylalanine; MAO, monoamine oxidane.
+
AGGRESSION AND CENTRAL NEUKOTRANSMITTERS
255
tion. At the behavioral end, stimulus and environment play vital roles, and proper analysis of behavioral responses and their measurement are essential. A number of points are to be raised in connection with neurochemical studies. Simultaneous measurement should be done for several neurotransmitters in the same subjects. ACh must be included in addition to NE, 5-HT, and DA; histamine and amino acids may also be considered. I n many cases, measurement of endogenous concentrations of these substances in the brain may not be of any significance. Their turnover rates may provide more meaningful information. Furthermore, such estimations should be done in different appropriate brain areas, rather than in the whole brain. For pharmacological studies, the fact that a drug may have effects on more than one neurotransmitter system should be taken into consideration during evaluation of its effect. While investigators in the field are aware of these problems and of extreme difficulties in solving them, these are being reemphasized here for their heuristic values. From available information, it appears that each type of aggression involves more than one neurotransmitter and that these transmitters probably act in a balanced manner within each behavioral system, thus providing a characteristic multitransmitter profile for each type of aggression. Neuroanatomical or neurophysiological manipulations, and environmental, pharmacological, or other interactions alter the neurochemical profile and modify the behavior. As mentioned earlier, genesis of different types of aggression may have some relation with the evolutionary processes. If so, a molecular evolution with respect to neurotransmitters in both qualitative and quantitative directions might have concomiiantly occurred at different stages of development in different species of animals, thus setting up a molecular basis and pattern for each defined type of aggression. Conjectural as these statements are at the present momen:, it remains for the future investigators to refute, substantiate, or modify them. A(:K N O W LEDGM E N T The author is thankful to Drs. K. C. Gupta and B. H. Turner for their active participation in preparing the neuroanatornical part of this review, to Drs. S. N. Dutta and Leslie H. Hicks for helpful suggestions and criticisms and to Miss Lisa Banerjee for her help in preparation of this manuscript. REFERENCES Aghajanian, C. K., Rosecrans, J. A., and Sheard, M. H. (1967). Science 156, 402. Allee, W. C. (1942a). Science 95, 289. Allee, W. C. (194213). B i d . Syrnp. 8, 139. Allikrnets, I,. (1964). Quoted in Lapin (1967). Allikmets, L. H. (1974). M e d . B i d . 52, 19.
256
S.
N . PRADHAN
AndCn, N. E., Dahlstrom, A., Fuxe, K., and Larsson, K. (1965a). Amer. J . Anat. 116, 329. A n d h , N. E., Dahlstrom, A,, Fuxe, K., and Larsson, K. (1965b). Life Sci. 4, 1275. AndCn, N. E., Butcher, S. G., Corrodi, H., Fuxe, K., and Ungerstedt, U. (1970). Eur. J . Pharmacol. 11, 303. Avis, H. H. (1974). Psychol. Bull. 81, 47. Bandler, R. J., Jr. (1969). Nature ( L o n d o n ) 224, 1035. Bandler, R. J., Jr. (1970). Brain Res. 20, 409. Bandler, R. J., Jr. (1971). Nature ( L o n d o n ) 229, 222. Bandler, R. J., Jr., and Chi, C. C. (1972). Physiol. &’ Behau. 8, 207. Battista, A. F., Goldstein, M., Nakatani, S., and Anagnoste, B. (1969a). J . Neurosurg. 31, 164. Battista, A. F., Goldstein, M., Nakatani, S., and Anagnoste, B. (1969b). Arch. Neurol. (Chicago) 21, 611. Baxter, B. L. (1967). Exp. Neurol. 19, 412. Baxter, B. L. (1968). Exp. Neurol. 21, 1. Baxter, B. Id. (1969). Exp. Neurol. 23, 200. Benkert, O., Gluba, H., and Matussek, N. (1973a). Neuropharmacology 12, 177. Benkert, O., Renz, A,, and Matussek, N. (197313). Neuropharmacology 12, 187. Bernstein, S., and Moyer, K. E. (1970). Brain Res. 20 75. Berntson, G. G. (1972). J . Comp. Physiol. Psychol. 81, 541. Bevan, W., Jr., Bloom, W. L., and Lewis, G. T. ( 1951 ) . Physiol. Zool. 24, 231. Blaschko, H., and ChruSciel, T . L. (1960). J . Physiol. ( L o n d o n ) 15, 272. Bloom, F. E., Algeri, S., Groppetti, S., Revuelta, A,, and Costa, E. (1969). Science 166, 1284. Brady, J. V., and Nauta, W. J. H. (1953). J. Comp. Physiol. Psychol. 46, 339. Breese, G. R., and Traylor, T. D. (1970). J. Pharmacol. Exp. Ther. 174, 413. Brudzynski, S., Gronska, J., and Romaniuk, A. (1973). Acta Physiol. Pol. 5, 631. Brunaud, M., and S o u , G. (1959). Quoted in Valzelli (1967). Brutkowski, S., and Mempel, E. (1961). Science 134, 2040. Brutkowski, S., Fonberg, E., and Mempel, E. (1961). Acta B i d . Exp. ( W a r s a w ) 21, 199. Bryson, G., and Bischoff, F. (1971). Res. Commun. Chem. Pathol. Pharmacol. 2, 469. Butcher, L. L., and And+ N. (1969). Eur. J. Pharmacol. 6, 255. Cain, D. P. (1974). J. Cornp. Physiol. Psychol. 86, 213. Cain, D. P., and Paxinos, G. (1974). J . Comp. Physiol. Psychol. 86, 202. Chen, G., Bohner, B., and Bratton, A. C., Jr. (1963). Arch. Znt. Pharmacodyn. Ther. 142, 30. Chruiciel, T. L., and Herman, Z. S . (1969). Psychopharmacologia 14, 124. Clemente, C. D., and Chase, M. H. (1973). A n n u . Rev. Physiol. 35, 329. Clemente, C. D., and Lindsley, D. B., eds. (1967). “Aggression and Defense.” Univ. of California Press, Berkeley. Clody, D. E., and Carlton, P. L. (1969). J . Comp. Physiol. Psychol. 67, 344. Conner, R. L., and Levine, S. (1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 150. Wiley, New York. Conner, R. L., Stock, J. M., Barchas, J. D., Dement, W. C., and Levine, S. (1970). Physiol. B Behau. 5, 1221. Consolo, S.,and Valzelli, L. (1970). Eur. J . Pharmacol. 13, 129.
AGGRESSION AND CENTRAI. K E U R O T R A N S M I T T E R S
25 7
Cook, L., and Weidley, E. (1960). Fed. Proc., Fed. Amer. S O L . E x p . Biol. 19, 22. DaVanzo, J. P. (1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 263. Wiley, New York. DaVanzo, J. P., Daugherty, M., Ruckert, R., and Kang, L. (1966). Psychopharmacologia 9, 210. Delgado, J. M. R. (1964). I n t . R e v . Neurobiol. 6, 349. Denisenko, P. P. (1965). Mentioned in Lapin (1967). Desmedt, L. K. C., Van Bruggen, J. A. A., and Niemegeers, C. J. E. (1973). Psychopharmacologia 31, 49. Di Chiara, G., Camba, R., and Spano, P. R. (1971). Nature ( L o n d o n ) 233, 272. Ebel, A,, Mack, G., Stefanovic, V., and Mandel, P. (1973). Brain Res. 57, 248. Eichelman, B. S., Jr., and Thoa, N. B. (1973). Biol. Psychiat. 6, 143. Eichelman, B. S., Jr., Thoa, N. B., and Ng, K. Y. (1972). Physiol. B Behau. 8, 1. Eleftheriou, B. E., and Boehlke, K. W. (1967). Science 155, 1693. Ernst, A. M. (1969). Acta Physiol. Pharmacol. Neer. 15, 141. Ersparmer, V., Glasser, A,, and Mantegazzini, P. ( 1960). Experientia 16, 505. Everett, G. M. ( 1961 ) . Neuropsychopharmacology 2, 110. Everett, G. M. (1968). Pharmacologist 10, 181. Everett, G. M., and Wiegand, R. G. (1963). Proc. I n t . Pharmacol. Meet., Z s t , 1961 Vol. 8, p. 85. Evetts, K. D., Uretsky, N. J., Iversen, L. L., and Iversen, S. D. (1970). Nature ( L o n d o n ) 225, 961. Ferguson, J., Henriksen, S., Cohen, H., Mitchell, G., Barchas, J., and Dement, W. (1970). Science 168, 499. Fernandez De Molina, A., and Hunsperger, R. W. (1959). 1. Physiol. ( L o n d o n ) 145, 251. Fernandez De Molina, A , , and Hunsperger, R. W. (1962). 1. Physiol. ( L o n d o n ) 160, 200. Flynn, J . P. (1967). I n “Neurophysiology and Emotion” (D. C. Glass, ed.), p. 40. Rockefeller Univ. Press, New York. Fog, R., Randrup, A., and Pakkenberg, H. (1970). Psychopharmacologia 18, 346. Fuller, R. W., Hines, C. W., and Mills, J. (1965). Biochem. Pharmacol. 14, 483. Fuxe, K., Hokfelt, T., and Ungerstedt, U. (1968). Aduan. Pharmacol. 6A, 235. Garattini, S., and Sigg, E. B., eds. (1969). “Aggressive Behaviour.” Wiley, New York. Garattini, S., Giacalone, E., and Valzelli, I,. ( 1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 179, Wiley, New York. Giacalone, E., and Kostowski, W. (1968). Quoted in Kostowski et al. (1969). Gianutsos, G., Hynes, M. D., Puri, S. K., Drawbaugh, R. B., and Lal, H. (1974). Psychopharmacologia 34, 37. Girgis, M. (1971 ) . I n t . J. Neurol. 8, 327. Girgis, M. (1972). I n “The Neurobiology of the Amygdala” (E. Eleftheriou, ed.), p. 283. Plenum, New York. Glusman, M., Won, W., Burdock, E. I. and Ransohoff, J. ( 1961 ) . Trans. Amer. Neurol. Ass. 86, 216. Goldberg, M. E., and Salama, A. I. (1969). Biochem. Pharmacol. 18, 532. Goldberg, M. E., Insalaco, J. R., Hefner, M. A., and Salama, A. I. (1973). Neuropharmacology 12, 1049.
258
S. N. PRADHAN
Goldstein, M., Anagnoste, B., Owen, W. S., and Battista, A. F. (1966). Experientia 23, 98. Grossman, S. P. (1963). Science 142, 409. Grossman, S. P. (1965). Physiol. @ Behail. 1, 1. Grossman, S. P. (1970). Physiol. @ Behail. 5 , 1103. Grossman, S. P., and Grossman, L. (1970). Physiol. @ Behao. 5, 1313. Gumulka, W., Samanin, R., Garattini, S., and Valzelli, L. (1969). Eur. J . Pharmacol. 8, 380. Gumulka, W., Samanin, R., Valzelli, I,., and Consolo, S. (1971). J . Ncurochem. 18, 533. Gunne, L., and Lewander, T. (1966). Acta Physiol. Scand. 67, 405. Hasselager, E., Rolinski, Z., and Randrup, A. (1972). Psychopharmacologia 24, 485. Heller, A,, and Moore, R. Y. (1968). Adzsun. Pharmacol. GA, 191. HernLndez-Peh, R., Chivez-Ibarra, G., Morgane, P. J., and Timo-Iaria, C . (1963). Exp. Neurol. 8 , 93. Hilton, S. M., and Zbrozyna, A. W. (1963). J . Physiol. ( L o n d o n ) 165, 160. Hingtgen, J. N., and Hamm, H. D. (1969). Life Sci. 8. 1. Hoffmeister, F., Kreisskott, H., and Wirth, W. (1964). Arzneim.-Forsch. 14, 482. Horovitz, Z. P., and Leaf. R. C. (1967). Neuropsychopharmacology 5, 1042. Horovitz, Z. P., Ragozzino, P. W., and Leaf, R. C. ( 1965). Life Sci. 4, 1909. Horovitz, Z. P., Piala, J. J., High, J. P., Burke, J. C., and Leaf, R. C. (1966). Int. J . Neuropharniacol. 5, 405. Hull, C. D., Buchwald, N. A., and Ling, G. (1967). Brain Res. 6, 22. IgiL‘, R., Stern, P., and Basigid, E. ( 1970). Neuropharmacology 9, 73. Janssen, P. A. J., Jageneau, A. H., and Niemegeers, C. J. E. ( 1960). J . Pharmacol. E x p . Therap. 129, 471. Janssen, P. A. J., Niemegeers, C. J. E., Schellehens, K. H. L., Dresse, A., Lenaerts, F. M., Pinchard, A., and Schaper, W. K. A. (1968). Arzneirn.-Forsch. 18, 261. Johnson, R. N. (1972). “Aggression in Man and Animals.” Saunders, Philadelphia, Pennsylvania. Jouvet, M. (1968). Aduan. Pharmacol. 6B, 265. Jung, R., and Hassler, R. (1960). I n “Handbook of Physiology” (Amer. Physiol. SOC., J. Field, ed.), Sect. 1, Vol. 11, p. 863. Williams & Wilkins, Baltimore, Maryland. Karczmar, A. G., and Scudder, C. L. ( 1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 209. Wiley, New York. Karli, P. (1956). Behaviour 10, 81. Karli, P. (1958). C. R. Soc. Biol. 152, 1796. Karli, P. (1959a). J . Physiol. (Paris) 51, 497. Karli, P. (195913). C . R . Soc. Biol. 153, 467. Karli, P. (1960). Arch. Int. Pharmacodyn. T h e r . 122, 344. Karli, P. (1968). J . Physiol. (Paris) 60, Suppl., 3. Karli, P., and Vergnes, M. (1965). J . Physiol. (Paris) 56, 384. Karli, P., Vergnes, M., and Didiergeorges, F. (1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 47. Wiley, New York. Karli, P., Vergnes, M., Eclancher, F., Schmitt, P., and Chaurand, J. P. (1972). I n “The Neurobiology o f the Amygdala” (B. E. Elftheriou, ed.), p. 553. Plenum, New York. Kesner, R. P., and Keiser, G. (1973). J . Comp. Physiol. Psychol. 84, 194.
AGGRESSION A N D (:I-NTRAL
NI~,IJROTRANSMITTl~~KS
259
King, M. B., and Hoehel, B. G. (1968). Commun. Rehazi. Biol., Part A 2, 173. Kletzkin, M. (1969). In “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 253. Wiley, New York. Koe, B. K., and Weissman, A. (1966). J. Pharniacol. Exp. T h e r . 154, 499. Korf, J., and Kuiper, H. E. (1971). Z sychophaririacologia 21, 328. Kostowski, W., and Giacalone, E. (1969). Eur. J . Pharnzacol. 7, 176. Kostowski, W., Giacalone, E., Garattini, S., and Valzelli, 1,. (1968). Eur. J. Pharmacol. 6, 371. Kostowski, W., Giacalone, E., Garattini, S., and Valzelli, L. (1969). EUT.J . Pharmacol. 7, 170. Kreisskott, W. ( 1963). Naunyn-Schrniedebergs Arch. Ex#. Pathol. Pharmakol. 245, 255. Krnjevit, K., and Silver, A. (1965). J. Anat. 99, 711. Krsiak, M. (1974). Res. Commun. Cherrz. Pathol. Pharmacol. 7, 237. Krsiak, M., and Steinherg, H . (1969). J . Psychosom. Res. 13, 243. Kulkarni, A. S. (1968). Life Sci. 7, 125. Kulkarni, A. S. (1970). Pharmacologist 12, 207. I A , H., DeFeo, J., and Thut, P. (1968). Comniun. Rehair. Biol. 1, 333. I d , H., DeFeo, J., and Thut, P. (1970a). Riol. Psychiat. 2, 205. I A , H., Nesson, B., and Smith, N. (1970h). B i d . Psychiat. 2, 299. Lammers, A. J. J. C., and Van Rossrim, J. M. (1968). Eur. J. Pharmacol. 5, 103. Lamprecht, F., Eichelman, B., Thoa, N. B., Williams R. B., and Kopin, I. J. (1972). Science 177, 1214. I.apin, I. P. (1967). Psychopharnzacologia 11, 79. Leaf, R. C. (1970). I n “Drugs and Cerebral Fiinction” ( W . I,. Smith, e d . ) , p. 201. Thomas, Springfield, Illinois. Leaf, R. C., Lerner, L., and Horowitz, Z. P. (1969). I n “Aggressive Behaviour” (S. Garattini and E. B. Sigg, eds.), p. 120. Wiley, New York. Leakey, I.. S. B. (1967). I n “Aggression and Defense” (C. D. Clemente and D. B. Lindsley, eds.), p. 1. Univ. of California Press. Berkeley. LeLvis, P. R., Shiite, C. C. D., and Silver, A. (1967). J . Physiol. ( L o n d o n ) 191, 215. Lints, C. E., and Harvey, J. A. (1969). J . C o r ~ i pPhysiol. . Psychol. 67, 23. I.ycke, E., Modigh, K., and Roos, B. E. (1969). Experientia 25, 951. McCarthy, D. (1966). Fed. Proc., Fed. A m e r . SOC. E x p . Riol. 25, 385. MacDonnell, M. F., Fessock, I.., and Brown, S. H. (1971). Quart. J. S t u d . A l c . 32, 748. McKenzie, G. M. (1971). Brain Res. 34, 323. MacI,ean, P. D. (1957). - 4 M A A r c h . A.’eurol. Psychiat. 78, 113. Maengwyn-Davis, G. D., Johnson, D. G., Thoa, N. B., Weise, V. K., and Kopin, I. J. (1973). Psychopharmacologia 28, 339. Mailey, W. M., and Baenninger, R. (1972). Physiol. B Behau. 9, 379. Malick, J. B. (1970). Physiol. B Rehat . 5, 679. Mantegazzini, A,, Fabbri, S., and Magni, C. (1960). Arch. Ztal. Sci. Farmacol. lo, 347. Marczinski, T. J. (1967). Ergeb. Physiol., B i d . Chem. E x p . Pharmakol. 59, 86. Melander, B. (1960). Acta Pharmacol. Toxicol. 17, 182. hlodigh, K. (1973). Psychopharmacologia 33, 1. Moore, R. Y. (1970). I n t . Rev. l\ieurobiol. 13, 67.
260
S . N. PRADHAN
Moyer, K. E. (1968). C o m m u n . Behav. Biol., Part A 2, 65. Myers, R. D. (1964). Can. J. Psychol. 18, 6. Navarro, M. G. (1960). Acta Physiol. Lat. Amer. 10, 122. Oltmans, G. A., and Harvey, J. A. (1972). Physiol. B Behau. 8, 69. Palermo, J., and Carlini, E. A. (1972). Eur. J . Pharmacol. 17, 215. Panksepp, J. (1971a). Physiol. d Behav. 6, 31 1. Panksepp, J. (1971b). Physiol. d Behav. 6, 321. Paxinos, G., and Bindra, D. (1972). J. Comp. Physiol. Psychol. 79, 219. Pepeu, G., Mules, A., Ruffi, A,, and Sotgin, P. ( 1971). Life Sci. 10, 181. Pletscher, A,, Bartholini, G., Bruderer, H., Burkard, W. P., and Gey, K. F. (1964). J. Pharmacol. E x p . T h e r . 145, 344. Pohorecky, I,. A,, and Chalmers, J. P. (1971). Life Sci. 10, 385. Pohorecky, L. A., Zigmond, M. J., Heimer, L., and Wurtman, R. J. (1969). Proc. Nut. Acad. Sci. U.S. 62, 1052. Poirier, L. J., and Sourkes, T. L. ( 1965). Brain 88, 181. Poncey, M. P., Bernard, P. S., and Chernov, H. I. (1972). Neuropharmacology 11, 39. Powell, D. A., Milligan, W. L., and Walters, K. (1973). Pharmacol. Biochem. Behau. 1, 389. Puri, S., and Lal, H. (1973). Psychopharmacologia 32, 113. Randrup, A., and Munkvad, I. (1969). I n “Aggressive Behaviour” ( S . Garattini and E. B. Sigg, eds.)? p. 228. Wiley, New York. Redmond, D. E., Jr., Maas, J. W., Kling, A., Graham, C. W., and Dekirmenjian, H. (1971). Science 174, 428. Reis, D. J. (1972). Res. Publ., Ass. Res. Nerv. M e n t . Dis. 50, 266. Reis, D. J., and Fuxe, K. (1969). Proc. Nut. Acad. Sci. U.S. 64, 108. Reis, D. J., and Gunne, I,. M. (1965). Science 149, 450. Reis, D. J., Miura, M., Weinbren, M., and Gunne, L. M. (1967). Science 156, 1768. Reis, D. J., Moorhead, D. T., Rifkin, M., Joh, T. H., and Golstein, M. (1970). Trans. Amer. Neurol. Ass. 95, 104. Rizzo, M., and Morselli, P. L. (1972). Brit. M e d . J . 3, 50. Rizzoli, A. A., Agosti, S., and Galzigna, I,. (1969). J. Pharm. Pharmacol. 21, 465. Roberts, W. W., and Bergquist, E. H. ( 1968). I . Comp. Physiol. Psychol. 66, 590. Roberts, W. W., Steinberg, M. L., and Means, L. W. (1967). J. C o m p . Physiol. Psychol. 64, 1. Romaniuk, A,, Brudzynski, S., and Gronska, J. ( 1973a). Acta Physiol. Pol. 6 , 305. Romaniuk, A., Brudzynski, S., and Gronska, J. (197313). Acta Physiol. Pol. 6 , 809. Roos, B. E. (1969). J. Pharmacol. 21, 263. Rowe, F. A., and Edwards, D. A. (1971). Physiol. B Behau. 7, 889. Rubinstein, E. H., and Delgado, J. M. R. (1963). Amer. J . Physiol. 205, 941. Salama, A. I., and Goldberg, M. E. (1970). Biochem. Pharmacol. 19, 2023. Salama, A. I., and Goldberg M. E. (1973). Life Sci., Part I I 12, 521. Scheel-Kriiger, J., and Randrup, A. (1967). Life Sci. 6, 1389. Scheel-Kriiger, J., and Randrup, A. (1968). J. Pharm. Pharmacol. 20, 948. Schildkraut, J. J., Logue, T., and Dodge, B. (1969). Psychopharmacologia 14, 135. Schneider, C. (1968). Nature ( L o n d o n ) 220, 586.
AGGRESSION AND CENTRAL N E U R O T R A N S M I T T E R S
26 1
Schwab, R. S., England, A. C., Jr,, Poskanzer, D. C., and Young, R. R. (1969). J . Amer. &fed. Ass. 208, 1168. Sclafani, A. (1971). J . C o m p . Physiol. Psychol. 77, 70. Scott, J. P. (1946). J. C o m p . Psychol. 39, 379. Scott, J. P. (1958). “Aggression.” Univ. of Chicago Press, Chicago, Illinois. Scott, J. P. (1966). Amer. Zool. 6, 683. Senault, B. (1970). Psychopharmacologia 18, 271. Senault, B. (1971). Psychopharmacologia 20, 389. Senault, B. (1973). Psychopharmacologia 28, 13. Seward, J. P. (1945a). 1. Comp. Psychol. 38, 175. Seward, J. P. (194513). J . C o m p . Psychol. 38, 213. Seward, J. P. (1946). J . Comp. Psychol. 39, 51. Sheard, M. H. (1969). Bruin Res. 15, 524. Sheard, M. H. (1973). Physiol. B Behav. 10, 809. Smith, D. E., King, M. B., and Hoehel, B. G. (1970). Science 167, 900. Sofia, R. D. (1969a). Life Sci., Part 1 8 , 705. Sofia, R. D. (196913). Life Sci. Part Z 8, 1201. Sorensen, J. P., and Harvey, J. A. (1971 ) . Physiol. B Behuu. 6, 723. Spector, S. A,, and Hull, E. M. (1972). J . Comp. Physiol. Psychol. 80, 354. Stark, P., and Henderson, J. K. (1966). I n t . J . Neuropharmacol. 5, 379. Stark, P., and Henderson, J. K. ( 1972). Neuropharmacology 11, 839. Stern, W. C., Hartmann, E. L., Draskoczy, P. R., and Schildkraut, J. J. (1972). Psychol. R e p . 30, 815. Szerb, J. C. (1967). J. Physiol. ( L o n d o n ) 192, 329. Tedeschi, D. H., Fowler, P. J., Miller, R. B. and Macko, E. ( 1969). I n “Aggressive Behaviour” ( S . Garattini and E. B. Sigg, eds.), p. 245. Wiley, New York. Tedeschi, R. E., Tedeschi, D. H., Mucha, A,, Cook, L., Mattis, P. A., and Fellows, E. J. (1959). J . Pharmacol. E x p . Ther. 125, 28. Tenen, S. S. (1967). Psychopharmacologia 10, 204. Thoa, N. B., Eichelman, B., and Ng, L,. K. Y. (1972a). Bruin Res. 43, 467. ‘I’hoa, N. B., Eichelman, B. and Richardson, J. S., and Jacobowitz, D. (197213). Science 178, 75. Thoa, N. B., Eichelman, B., and Ng, K. Y. ( 1 9 7 2 ~ ) J. . Pharm. Pharmacol. 24, 337. Thomas, J. B., and Van Atta, L. (1972). Physiol. G? Behau. 8, 225. Thorne, B. M., Aaron, M., and Latham, E. E. (1973). J . Comp. Physiol. Psychol. 84, 339. Turner, B. H . (1970). J. Comp. Physiol. Psychol. 71, 103. Ungerstedt, U., Butcher, L. L., Butcher, S. G., Andkn, N. E., and Fuxe, K. (1969). Brain Res. 14, 461. Uretsky, N. J., and Iversen, L. L. (1970). 1. Neurochem. 17, 269. Vale, S., Espeiel, M. A., and Dominiguez, J. C. (1971). Lancet 2, 437. Valzelli, L. (1967). Aduan. Pharmacol. 5, 79. Valzelli, L., Giacalone, E., and Garattini, S. (1967). Eur. J. Pharmucol. 2, 144. Vander Wende, C., and Spoerlein, M. T . (1962). Arch. I n t . Pharmacodyn. T h e r . 137, 145. Varszegi, M. K., and Decsi, L. (1967). Acta Physiol. 32, 61. Vergnes, M., and Karli, P. (1963). C. R. Soc. Riol. 157, 1061. Vergnes, M., and Karli, P. (1964). C. R. Soc. B i d . 158, 856.
262
S. N . PRADHAN
Vergnes, M., and Karli, P. (1965). C. R . SOC.Biol. 159, 972. Vernon, W. M. (1969). Gen. Psychol. Monogr. 80, 2 . von Voightlander, P. F., and Moore, K. E. (1971). Science 174, 408. Wasman, M., and Flynn, J. P. (1962). Arch. Neurol. (Chicago) 6, 220. Weischer, M. (1969). Psychopharmacologia 15, 245. Welch, A. S., and Welch, B. L. (1968). Biochem. Pharmacol. 17, 699. Welch, A . S., and Welch, B. L. (1971). I n “Physiology of Fighting and Defeat” (B. E. Eleftheriou and J. P. Scott, eds.), p. 91. Univ. of Chicago Press, Chicago, Illinois. Welch, B. L., and Welch, A. S. (1969). I n “Aggressive Behaviour” (S. Garattini and %. B. Sigg, eds.), p. 188. Wiley, New York. Wolf, A,, and von Haxthausen, E. F. ( 1960). Arrneim.-Forsch. 10, 50. Woodworlh, C. H. (1971 ) . Physiol. &? Behau. 6 , 345. Yen, H. C. Y., Stanger, R. L., and Millman, S. (1959). Arch. I n t . Pharmacodyn. Ther. 123, 179. Yen, H. C. Y., Katz, M. H., and Krop, S. (1970). Toxicol. Appl. Pharmacol. 17, 597. Yen, H. C. Y., Katz, M. H., and Krop, S. (1971). Arch. Int. Pharmacodyn. T h e ? 190, 103. Zigmond, M. J., Chalmers, J. P., Simpson, J. R., and Wurtman, R. J. (1971). J . Pharmacol. E x p . T h e r . 179, 20.
A NEURAL MODEL OF ATTENTION, REINFORCEMENT AND DISCRIMINATION LEARNING By Stephen Grossberg'
Department of Mathematics Massachusetts Institute of Technology, Cambridge, Massachusetts
.
I. Introduction A. Blocking and Overshadowing B. Frustrative Nonreward C. Partial Reinforcement Acquisition Effect D. Steepening of Generalization Gradients Due to Discrimination Training E. Peak Shift and Behavioral Contrast . F. Orienting Reaction vs Discriminative Cues G. Novel Events as Context-Dependent Reinforcers . H. Motivation and Generalization . I. Predictability and Ulcers . J. Anatomy and Physiology. 11. Drives, Rewards, Motivation, and Habits 111. The Rebound from Fear t o Relief IV. Short-Term Memory and Total Activity Normalization . V. Sensory-Drive Heterarchy . VI. Conditionable Ct + S Feedback and Psychological Set . VII. The Persistence of Learned Meanings . VIII. Overshadowing and the Triggering of Arousal by Unexpected Events . IX. Pavlovian Fear Extinction vs Persistent Learned Avoidance , X. Frustration . XI. Partial Reinforcement Acquisition Effect . XII. Generalization Gradients in Discrimination Learning . XIII. Habituation and the Hippocampus . XIV. Overshadowing vs Enhancement . XV. Novelty and Reinforcement . XVI. Motivation and Generalization XVII. Predictability and Ulcers . XVIII. Orienting Reaction . XIX. A Learned Expectation Mechanism .
.
.
. .
.
.
.
.
. . . . .
. . . . . . . . .
. . . .
. . . . . . . . . .
264 265 266 267 267 268 269 269 270 271 272 274 276 282 288 290 291 294 297 297 300 301 305 306 308 309 310 311 313
Supported in part by the Alfred P. Sloan Foundation, the Office of Naval Research (N00014-67-A-O204-0051 ) , and the Advanced Research Projects Agency D A H C 15-73-C-0320) administered by Computer Corporation of America. 263
264
STEPHEN GROSSBERG
XX. Regulation of Orienting Arousal . XXI. Hippocampal Feedback, Conditioning, and Dendritic Spines . XXII. Nervous Eating and Attentional Deficits Modulated by Arousal. Appendix References
. .
. . . . .
316 319 321 323 325
I. Introduction
This paper describes a psychophysiological model aimed at discussing how animals pay attention to and discriminate among certain cues while ignoring others, based on criteria of relevance derived from past experience or innately preprogrammed in their neural apparatus. The model builds upon previous results (Grossberg, 1969a,b, 1970, 1971a,b, 1972a-d, 1973, 1974; Grossberg and Pepe, 1971) that introduce some psychophysiological mechanisms of classical and instrumental learning, and of pattern discrimination. These results include network mechanisms of drive, reward, punishment, escape and avoidance, motivation, short-term and long-term memory, serial learning, arousal, expectation, and various perceptual constancies (e.g., hue and brightness). They will be reviewed herein as needed to motivate the present work. A previous paper (Grossberg, 1974) reviews some of them more systematically. This collection of mechanisms comprises the theory of Embedding Fields. This theory derives neural networks from simple psychological facts that are taken as fundamental postulates. The theory tries to isolate postulates that act as guiding principles of neural design during individual development and the evolution of species. The networks that are hereby derived are capable of behavior that is far more complex and subtle than the postulates themselves, and also generate various new predictions. The theory is derived by a method of successive approximations; as more postulates are imposed, the networks become ever more sophisticated and reaslistic. At each stage of the derivation, basic mechanisms of network organization emerge, and are preserved as new postulates are imposed. Thus, each stage of the derivation ties a definite class of psychophysiological phenomena to a fixed list of elementary postulates, and successive stages of the derivation show how various phenomena of differing sophistication are interrelated. A central theme in the present model will be that two systems are continually readjusting each other. One system (an attentional system) strives toward an ever more stable response to patterns of fluctuating cues by focusing attention on important subclasses of cues. This system is incapable of adapting to unexpected environmental changes. The second system (an arousal system) overcomes the rigidity of the attentional system when unexpected events occur, and allows the network to adapt to new reinforcement
ATTENTION,
REINFORCEMENT, A N D DISCRIMINATION
265
contingencies. The following psychophysiological themes, which clarify this situation, will be discussed in the model, among others.
A. BLOCKING A N D OVERSHADOWING This theme is elegantly discussed by Honig ( 1970), Kamin ( 1968, 1969), Trabasso and Bower ( 1968), and Wagner ( 1969a), who should be consulted for details. Below are tersely summarized some main experimental facts taken from these sources. We will consider a sequence of three classical conditioning experiments. I n each experiment, two cues CS, and CS,, such as a sound and a flashing light, are the conditioned stimuli that will precede a prescribed unconditioned stimulus UCS, such as food or shock. Let the UCS be a shock of prescribed duration and intensity, for definiteness. I n experiment 1, let CS1 and CS2 be equally salient to the learning subject (3, and suppose that both cues are always presented together before the shock. On recall trials, will (3 be afraid of CS1 or CS2 presented separately? The answer is “yes”; thus, cues presented together can be conditioned separately. In experiment 2, first let CS1 be paired alone with shock, until 0 is afraid of CS1. Then present both CS1 and CS2before shock during the second phase of the experiment. O n recall trials, 0 is not afraid of CS2. Somehow, prior conditioning of CS1 to the UCS has “blocked,” or “overshadowed,” the possibility of conditioning CSZ to the UCS. This happens even though (3 “notices” CS2, and the amount of blocking depends on the amount of prior conditioning between CS1 and the UCS. A blocking effect can also be elicited in experiment 1 if CS1 is a more intense, or salient, cue than CS2. In a similar direction, Bitterman (1965) discussed evidence that a CS which is paired simultaneously with a UCS does not get conditioned to the UCR. In experiment 3, again pair CS1 with the UCS before pairing both CS1 and CS2 with the UCS; however, choose the UCS intensity at two different levels in the two phases of the experiment. Then the blocking effect is at least partially eliminated: 0 is afraid of CS2. (In general, one must also discuss whether a decrease in shock makes CS2a conditioned source of relief, rather than of fear.) These experiments can be interpreted as follows. I n the second phase of experiment 2, CS1 is a perfect predictor of the event UCS that is about to follow. Since CS2 is an irrelevant cue, (3 does not connect CS2 with the U C R even though (3 notices CS2. I n the second phase of experiment 3, however, CS1 is not a perfect predictor of UCS intensity. Hence some conditioning of CS2 to the new U C R (or UCR-like response) occurs. I n experiment 1, neither CSl nor CS2 is initially a predictor of the UCS. Hence 0 will learn connections from each CS1 to UCR. If CS1 is more salient or intense than
266
STEPHEN GROS SBERG
CS2, then faster conditioning of CSI to the UCS can eventually block conditioning of CS2 to the UCR. Such experiments suggest that various learning subjects act as minimal adaptive predictors; they enlarge the set of cues that control their behavior only when the cues that presently control their behavior do not perfectly predict subsequent events. In particular, somehow the results of (3’s acts can feed back in time to influence which cues will control these acts in the future. This phenomenon has broad implications, since it bears on such questions as: How do we decide which cues cause events and which are adventitious? How do we characterize the cues that define the objects with which we deal? Does the persistent unpredictability of a given source of cues increase the likelihood that this source will be treated more as a “subject” than as an “object”? B. FRUSTRATED NONREWARD A special case of an unpredictable event is one in which an expected reward does not occur. Suppose that 0 has learned to expect food as the end result of a particular sequence of motor acts, but that food is no longer available in the expected place. Were 0 to continue seeking food a t this place, (3 would starve to death. How does 0 countercondition this erroneous expectation, and thereby release exploratory behavior aimed at finding new sources of food, before starvation occurs? An aversive state that is activated by the nonoccurrence of expected events is “frustration” (Amsel, 1958, 1962 ; McAllister and McAllister, 1971 ; Wagner, 196913). Frustration can motivate avoidance behavior and has properties analogous to those of fear. Frustration can follow the nonoccurrence of expected rewards other than food. Thus if a sequence of events motivated by a given positive drive is suddenly interrupted, say by nonoccurrence of the expected reward at the end of a sequence of acts aimed a t getting the reward, then a negative (frustrative) reaction can occur. We will argue that this rebound effect, from positive to negative, can be given a mechanistic interpretation that is shared by rebound effects from negative to positive, such as the relief that is felt when a prolonged shock is unexpectedly terminated (Denny, 1970), or various other punishment contrast and reinforcement contrast effects ( Azrin and Holz, 1966). For example, let a pigeon be trained on a VI 1 schedule to peck for food. If a maintained level of punishment is suddenly removed, the pigeon will temporarily peck faster than it did in the absence of punishment. If the frequency of reward is suddenly increased, a temporary overshoot in pecking rate will again occur. The mechanism to be discussed herein also allows comparison with the facts that classically conditioned fear can
ATTENTION,
R E I N F O R C E M E N T , AND DISCRIMINATION
267
rapidly extinguish, even though learned asymptotic avoidance behavior can be very stable (Seligman and Johnston, 1973).
C. PARTIAL REINFORCI:MENT ACQLJISITION EFFECT Why can fearful or frustrating tasks that work out well in the end become so rewarding? What causes the extra “thrill” that some people seem to feel
after successfully carrying out dangerous tasks? An analogous boost in rebtard value is illustrated by thr following example. Consider the speed with which rats run down a straight alley to a positive goal. Compared to continuously rewarded animals, animals on a random partial reinforcement schedule run slower early in training, gradually catch up, and finally, late in training, run faster (Goodrich, 1959; Haggard, 1959). This effect has been attributed by several authors to frustration (Gray and Smith, 1969). We will suggest a property of the frustration mechanism that can fornially generate this effect, and can predict a relationship between an animal’s ability to carry out learned escape in the presence of fearful cues, the reinforcing effect of reducing J units of shock to J / 2 units of shock, the size of the partial reinforcement acquisition effect, and the animal’s arousal level, suitably defined.
D. STEEPENING OF GENERALIZATION GRADIENTS DUE TO DISCRIMINATION TRAINING Jenkins and Harrison (1960) showed that if pigeons are trained to peck a key in response to a 1000 cps tone (the S+) but not to peck in the absence of the tone (the S-) , then a sharper tonal generalization gradient is found than after training to peck at the S+ without discrimination training with S-. Newman and Baron (1965) used a vertical white line on a green key as S+ and the green key as S-. They tested generalization by tilting the line at various orientations. A generalization gradient was found, but no gradient occurred if the S- was a red key or if the S- was a vertical ivhite line on a red key. By contrast, Newman and Renefeld (Honig, 1970) used as S+ a vertical white line on a green key and as S- a green background, but tested and found generalization of the line orientation on a black key. They also tested generalization on a black key following training without a green S- and again found a significant generalization gradient, by contrast with the case where testing used a green key. This effect was interpreted to be one of “cue utilization during testing rather than cue selection during learning,”
268
STEPHEN GROSSBERG
since somehow removing green during testing unmasked prior learning on the orientation dimension. Honig (1969) used a blue key as S+ and a green key as S-. This was followed by dimensional acquisition with three dark vertical lines on a white key. Generalization testing was on the orientation dimension. This paradigm was called a true discrimination ( T D ) experiment. By contrast, another group of pigeons was rewarded half the time on the blue key and half the time on the green key before dimensional acquisition with the three vertical lines and generalization testing on the orientation dimension. This paradigm was called a pseudodiscrimination ( PD) experiment. The generalization gradient was marked in the T D case, but flat in the PD case. F. Freeman (unpublished master’s thesis, Kent State University, Kent, Ohio, 1967) modified this experiment by training pigeons to peck at a vertical line on a dark key (S+) but not to peck at a line tilted at 120° on the same dark background (S-) . Then dimensional acquisition with the vertical line on a green background was followed by generalization testing on the dimension of color. A steeper color gradient was found than in the absence of prior discrimination training on S-. This is an example of enhancement due to prior discrimination training, rather than blocking. Blocking can also be achieved, as Mackintosh and Honig showed (Honig, 1970). They trained pigeons with S+ and S- as above. Then they retrained them with two spectral values (501 and 675 nm) redundantly added after the animals had reached criterion. Control groups received only the second stage of training. A generalization test on four spectral values demonstrated steeper gradients for the control group.
E. PEAKSHIFTA N D BEHAVIORAL CONTRAST Let a pigeon be trained to peck at a key illuminated by a 550 nm light
(S+) but not to peck at a key illuminated by a light of x nm (S-), where x is chosen greater than 550 for definiteness. If the pigeon makes some errors in learning this discrimination, then it will, on test trials, peck most vigorously at a key lit by a light of y ( x ) nm, where typically y ( x ) # 550, y ( x ) < 550 if / x - 5501 is sufficiently small, and y(x) tends to increase as x increases (Hanson, 1959). This shift does not occur if the pigeon learns the discrimination without making errors (Terrace, 1966) . I n the same experimental setting, the influence of error-filled training at x nm can increase the rate of pecking at 550 nm if lx - 5501 is sufficiently large (“behavioral contrast”) (Hanson, 1959; Bloomfield, 1966’). These effects do not occur if the training is errorless (Terrace, 1966), and behavioral contrast disappears after long training sessions (Terrace, 1966).
ATTENTION, REINFORCEMENT, A N D DISCRIMINATION
269
Honig (1962) has noted that the peak shift occurs only if the S+ and S- are presented successively, but not if they are presented simultaneously. Grusec ( 1968) has shown that after errorless discrimination training, pairing a shock with the S- will create a peak shift. Bower (1966) has suggested that such contrast effects are due to frustration. Bloomfield (1969) has attempted to unify these results by stating that an “unexpected change for the worse” yields contrast and peak shift effects. Such changes include a sudden reduction in the frequency of reinforcement, or the introduction of shock.
F. ORIENTING REACTION vs DISCRIMINATIVE CUES T h e frustrative reaction is but one case of a general theme; namely, why can 0’s responses to a fixed unexpected, or novel, event be different in different contexts? For example, suppose that a human subject sits before a lever with no prior training and that a loud noise occurs abruptly to the left of the subject. There will ensue a strong tendency for the subject to orient toward the noise by turning his head to the left (Luria and Homskaya, 1970). By contrast, suppose that the subject is taught that the noise is a discriminative cue for rapidly pressing the lever to receive a valuable reward. Then the orienting reaction can be replaced by a rapid lever press. How does conditioning redirect the internal flow of activity that would otherwise activate the orienting reaction (Lynn, 1966) ? The orienting reaction is a form of attentional mechanism, but not the only one. For example, novel stimuli can attract more attention than nonnovel stimuli even if the stimuli are presented tachistoscopically (Berlyne, 1970; P. McDonnell, unpublished doctoral thesis, University of Toronto, 1968; Trabasso and Bower, 1968). We will distinguish between the two types of reaction in the mechanisms to be described below.
G. NOVELEVENTSAS CONTEXT-DEPENDENT REINFORCERS As we noted above, frustration can follow the nonoccurrence of an expected reward; thus, if a sequence of events motivated by a given positive drive is unexpectedly interrupted, say by nonoccurrence of the reward, then a negative (frustrative) reaction can ensue. By contrast, if the expected reward is replaced by an even more valued reward, then the frustrative reaction can be mitigated; for example, a check for $1,000,000 might well eliminate the frustration one might feel after opening a refrigerator and noting the absence of a n expected apple. In both cases, “surprise” might occur
270
S T E P H E N GROSSBERG
owing to the unexpectedness of the outcome, but this surprise is channeled differently in the two cases. Indeed, if an event is rewarding to an animal, then the effectiveness of the reward can be increased if it is also novel. Berlyne (1969) notes that novel events per se can be positively rewarding. He shows that a response-contingent change in the intensity of light in a rat’s cage can be used to reward bar pressing. We will suggest that the light change enhances the positive incentive-motivation that is motivating the rat during approach and pressing of the bar. This incentive motivation is not necessarily associated with a specific drive, such as hunger, and can merely be the motor arousal mechanism that is used for general approach behavior. Berlyne also notes that an increase in light level can be less rewarding if the animal’s arousal level is too high. He suggests that the rewarding value of an indifferent stimulus is an inverted U function of its novelty. T h e inverted U is also a function of the animal’s arousal level, so that a given novel stimulus can have different reward value if the animal’s arousal level is varied. Berlyne distinguishes the existence of an optimal arousal level from an optimal arousal increment and discusses the relationship between a given arousal level and its optimal arousal increment in terms of the inverted U . O u r model discusses related mechanisms of arousal with the property that various types of abnormal behavior can be elicited by overarousal ; cf. a schizophrenic’s difficulty in paying attention, or seizure activity. In summary, we will suggest that the nonspecific neural activity generated by a novel event filters through all internal drive representations. The effect of this activity on behavior will depend on the pattern, or context, of activity in all these representations when the novel event occurs. Sometimes the novel event can enhance the effect of an ongoing drive, sometimes it can cause a reversal in sign (as in the frustrative reaction), and sometimes it can introduce and enhance the effect of a different drive. We will be led to assume that every novel event has the capacity to activate orienting reactions, but whether it does or not depends on competition frorn the drive loci which the event also activates. The nonspecific activity generated by the novel event will also be assumed to reach internal sensory representations, where it helps determine which cues will enter short-term memory to influence the pattern of internal discriminatory and learning processes.
H. MOTIVATION AND GENERALIZATION Increasing an animal’s motivation during learning and performance can flatten its gradient during performance (Bersh et al., 1956; Jenkins et al., 1958; Kimble, 1961). By contrast, let a pigeon be trained to peck a key for food, and then trained using a 1000 cps tone as a warning for electric
ATTENTION,
REINFORCEMIiNT, A N D DISCRIMINATION
271
shock. O n testing trials, its generalization gradient for response suppression as a function of tonal frequency is steeper if the pigeon is hungrier (Hoffman, 1969). Note that in this experiment two drives (hunger and fear) compete, whereas in the experiments describing flattening of generalization gradients, only one drive is operative.
I. PREDICTABILITY A N D ULCERS Weiss ( 1971a,b,c) has carefully studied the influence of several parameters on the development of stomach ulcers in rats. I n his experiments, some rats can escape tail shock by turning a wheel. Each turn of the wheel delays the next onset of shock by a fixed amount of time. In some studies, each shock is preceded by a warning signal. I n other studies, each wheel turn is followed either by a tone or by a brief shock, but not both. In each study, there is a control group that is not shocked, and a yoked group that is shocked whenever the animals capable of avoiding or escaping the shock are shocked. Thc yoked group also hears the tone whenever the avoidanceescape group does. Weiss shows that ( a ) avoidance-escape subjects develop less ulceration than do the yoked animals; ( b ) a warning signal reduces the ulceration of both groups of rats; ( c ) the yoked animals develop less scvere ulcers than the avoidance-escape animals if both groups receive a brief shock after each avoidance-escape response; and ( d ) little ulceration develops in the avoidance-escape group, even if no warning signal precedes shock, if each avoidance-escape rcsponse is followed by a feedback stimulus, such as a tone. Weiss concludes from these results that two main factors contribute to thc development of ulcers: the number of roping responses that an animal makes, and the amount of relevant feedback that these coping responses produce. As the number of coping responses increases, the tendency to ulcerate also increases; but as the relevant fredback increases, the tendency to ulcerate decreases. For example, in ( d ), the avoidance-escape animals can make many coping responses, but they also receive a high level of relevant feedback, since each successful response is followed by a feedback stimulus that predicts an interval free from shock. In ( c ) , the avoidance-escape animals receive low relevant feedback, since they are shocked for coping. We will find that the magnitude of negative incentive-motivation in our model is a monotone increasing function of the amounts of ulceration that are described in ( a ) - ( d ) . A rebound from a source of net positive incentive motivation to a source of net ncgative incentive motivation produces the frustrative reaction in our modrl. This positive source is capable of motivating consunmiatory motor activity. 'I'he nrgative source linked with it is not the same as the source of fear. Thus our rcsults do not imply that amounts
272
S T E P H E N GROSSRERG
of fear equal to the amounts of negative incentive produced by the rebound will have the same effects on ulceration. They suggest, rather, that properties of the negative rebound source are triggered in parallel with, or themselves trigger, ulccrogenic agents.
J. ANATOMY A N D PHYSIOLOGY The networks will contain sevcral functionally distinct regions. The interactions between these regions call to mind familiar anatomical facts. I t will be apparent that the network regions are not presumed to be exact replicas of real anatomical fragments. Nonetheless, the anatomical relationships between the network regions, as well as their functional roles in total network processing, suggest natural analogs with real anatomies. These analogs will be pointed out both to suggest possible new insights about the functioning of real anatomies, and to serve as an interpretive marker for the networks that will arise in the future from additional postulates. The psychological validity of formal network interactions is, however, independent of how well we guess neuroanatomical labels for network components at this stage of theorizing, since the formal anatomy is still, at best, a lumped version of a real anatomy. A network region of particular interest is reminiscent of the hippocampus. This region supplies motivational feedback to several other network areas (Olds, 1969). This feedback is determined by a competition between channels corresponding to different drives. Each channel is influenced by sensory and drive inputs. The sensory pathways can be strengthened or weakened by reinforcing events (“conditioned reinforcers”) . If a given channel has a prepotent combination of input from conditioned reinforcers and drive, it will suppress other channels using its on-center off-surround anatomy (Anderson et al., 1969; Grossberg, 1973). This feedback has at least three functions. I t supplies signals to the region where the sensory pathways are being conditioned by reinforcing events. These signals help to determine the pattern of motivational activity that the sensory pathways will learn. Thus the mock-hippocampus receives input from a region that is implicated in reinforcement, and delivers feedback to this region. We therefore (undogmatically) interpret this second region as a mock-septum (Raisman et al., 1966). The mock-hippocampus also supplies conditionable nonspecific feedback, in the form of a late, slow potential shift, to sensory processing areas (e.g., mock-neocortex) of the network. This feedback, which is related to the network‘s arousal, drive, reinforcement, and motivational mechanisms, helps to determine which cues will be attended to by the network. An analogous wave, the contingent negative variation (CNV), has been reported
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
273
in uiuo (Walter, 1964). Finally, the mock-hippocampus controls a feedback pathway that helps to regulate the degree of motor arousal or suppression. If the mock-hippocampus is removed, then transfer of short-term memory into long-term memory is prevented, and difficulties in paying attention will ensue (Milner, 1958). The mock-septum is influenced by a source of drive input (mock-hypothalamus) and of nonspecific arousal (mock-reticular formation) . The level of nonspecific arousal is modulated by the degree of unexpectedness of external events. A mechanism whose motor command cells can be preset to fire only in response to expected events has been synthesized and has an anatomy reminiscent of cerebellar interactions (Grossberg, 1972a). This mechanism projects to the mock-reticular formation. Thus, although the arousal itself is nonspecific, its regulation can be dependent upon specific sensory cues. The nonspecific arousal filters through the drive-representing channels, and can either contrast enhance their activity, or cause a positive (negative) motivational bias to flip into a negative (positive) motivational bias. Thus nonspecific arousal can have specific effects on the pattern of motivational feedback. The nonspecific arousal also feeds into sensory processing areas (e.g., mock-neocortex) , where it influences which cues will generate enough neural activity to reverberate in short-term memory, and thereupon be able to influence processes of learning and discrimination. The nonspecific arousal that is triggered by unexpected events differs from the nonspecific conditionable feedback that is related to network drive, reinforcement, and motivational levels. Indeed, these two input sources can compete with each other in overshadowing experiments. In summary, at least two major feedback loops exist in the network. One feeds between external sensory and internal sensory (e.g., drive) processing areas (cortex + hippocampus + cortex). The other feeds within the internal sensory processing areas (septum + hippocampus + septum) . The drive representations are organized into dipoles, such that each dipole controls a positive and a negative incentive motivational channel; e.g., relief and fear, hunger and frustration. The regulation of motivational output from the dipoles, and of learning based on this output, has been interpreted as using two distinct transmitter systems, which are presumed to be analogous to adrenergic and cholinergic transmitters (Grossberg, 1 9 7 2 ~ ) The . need to synchronize the activity of the two parallel channels in a given dipole, and to sample the resultant activity in both dipole channels, suggests that the two transmitter systems are also organized in parallel across the two channels. The organization of drives into dipoles can induce a formal “poker-chip’’ organization in the input source that feeds them nonspecific arousal. A poker-chip anatomy for the reticular formation has been described (Scheibel and Scheibel, 1967).
274
STICPHEN GROSSHEKG
II. Drives, Rewards, Motivation, and Habits
The model is an extension of a previous model that has been derived from psychological postulates (Grossberg, 1969a; 1971a, 1972b,c, 1974). This extension is the result of imposing more postulates. The old Iiostulates describe basic properties of classical conditioning, yet the mechanisms that arise can also be used to discuss aspects of instrumental conditioning. The main postulates are described in Grossberg (1974). Two of these postulates are, for example, that ( 1) the time lags between CS and UCS on successil e learning trials can differ; and ( 2 ) after learning has occurred, the CS can elicit the U C R (or UCR-like event) in the absence of the UCS. Such obvious facts seem innocent enough; yet when several of them are taken together, and are translated into a rigorous mathematical description, the ensuing neural networks are capable of surprisingly subtle behavior. A heuristic discussion of various mathematical properties of these networks can be found in Grossberg (1974). Some mathematical theorems are proved in Grossberg (1972d, 1973). A review of relevant network properties is given below in several stages. Consider Fig. 1. I n Fig. la, the zth conditioned stimulus (CS,) among n possible stimuli excites the cell population U,1 of its sensory representation. I n particular, CS, has already been filtered on its way from the sensory periphwy of the network to UI1, so that i t reliably excites U,l but not irrelevant cells. Some mechanisms of sensory filtering (i.e., pattern discrimination) are derived in Grossberg (1970) and extended in Grossberg (19 7 2 4 . Sensory representations will be denoted generically by S. I n response to the CS, input, U,1 sends signals to stage U,Z of the zth sensory representation, as well as toward all thc populations @ = a,+, @-, . . .) of arousal cells. (In this presentation, we ignore effects due to spatial gradients in interaction strength.) Thus the
, I
NET INCENTIVE MOTIVATION
/ DRIVE
(b)
HUNGER SHOCK
FIG. 1. Interaction of reinforcement, drive, motivation, and habit strength in minimal network.
2 75
ATTENTION, R E I N F O R C E M E N T , AND DISCRIMINATION
s s
s
4 Q pathways are “nonspecific,” whereas the + pathways are “specific.” T h e arousal cells a h subserve the hunger drive, cells Qf+ subserve fear, and cells QJ- subserve relief from fear. T h e cells receive an internally generated drive input that is a monotone increasing function of hunger level. T h e cells Q.,+ receive an input that is a monotone increasing function of shock level. Offset of shock elicits transient excitatory activity in the relief center a,+ (Denny, 1970). Signals from the sources Q. are generated by activity at these sources and are, other things equal, monotone-increasing functions of this activity. The signals from Qh and Q f - to all U,z populations are excitatory, whereas the signals from a/+to all U,n populations are inhibitory. Since a signal from a is nonpopulation in Q. is sent to all populations Ut2,the pathway @ + specific. lJ,z can send signals to 312 only if it simultaneously receives a large signal from U,1 and a large net excitatory signal from a. In particular, a large excitatory Uz2 signal can be canceled by a large inhibitory Q,+-+ U,z signal, which thus prevents U t zfrom firing even if CS, is present. I n this way, consummatory activity compatible with hunger can be suppressed by shock. Suppose that shock is terminated by an avoidance response, or AR. (Learned escape responses can become avoidance responses; hence we use only the term “avoidance” below, for simplicity.) Then Qf- is excited and signal to all sensory repcreates a large, but transient, excitatory a,-+ resentations. Sensory feedback cues of the AR also excite particular sensory representations, which we denote by S(AR). T h e U,z stages of S(AR) cells thus receive U,1 and Q f - inputs. They can therefore fire and send signals to L3KCells in that receive only the a,- input cannot fire. Changes in long-term memory (LTM) can occur a t two locations in this picture: at the S + Q synaptic knobs, and at the + 3X synaptic knobs. The unit of L T M is a spatial pattern: the relative activities of all the longterm memory traces in the synaptic knobs of a given population. T h e U,I Q. synaptic knobs and the U,n+ 3K synaptic knobs can learn (“sample”) patterns of activity playing on the populations Q and 3X, respectively, only when these knobs are activated by U,1 or U,z signals. T h e U,I* Q synaptic knobs encode a weighted average of the “motivational” patterns that arc sequentially presented to Q populations when these knobs are Sampling. T h e U,z+ 3K synaptic knobs encode a weighted average of the << motor” patterns playing on the motor control cells ‘3n when these knobs are sampling. (The cells 3K need not be motor controllers; they can also be an arbitrary collection of sensory cells, for example.) Thus when the cells in the S(AR) representations are active after a n avoidance response, their U,1 stages can sample “relief” at @,-,and their Ua2stages can sample the motor pattern corresponding to the AR. O n a recall trial, the presence of
s
s
s
s
--f
2 76
STEPHEN GROSSRERG
avoidance or escape cues, such as seeing the wheel that delays or terminates shock when turned, activates S(AR). Signals S(AR) -+a,--+ generate positive feedback which combines with S(AR) -+ S(AR) signals to activate s ( A R ) + 311 synaptic knobs and reproduce (a frame of) the AR. I n this sense, the rebound from fear to relief when shock terminates provides the positive motivation for learning escape or avoidance motor acts. Figure 1b describes network variables in more conventional terms. Rewarding or punishing events change the pattern of activity across populations in a. These reinforcing events are encoded in L T M by the 8- @ synapses that are sampling at the crucial times. Drive inputs also perturb @. Ca. cells such as can fire to S only if they receive a sufficiently large drive input and an input from a UCS or a conditioned reinforcer. Otherwise, the network would emit persistent eating behavior in response to food in the absence of hunger. In this context, a conditioned reinforcer corresponding to a given drive is a cue whose S ---f a pathway to that drive locus has become relatively large owing to prior conditioning. T h e a + signals provide incentive motivation for modulating the activity in Uiz. If the net incentive motivation to Uiz is sufficiently positive and CSi is present, then signals from Ui2 to 3TZ can occur, sampling at the U,Z 3TZ synaptic knobs is initiated, and the habit strengths of these knobs are influenced by the patterns playing on 311 during the sampling interval. Given this basic network structure, several refinements must be made to guarantee that the network works well. These refinements eventually lead to mechanisms that are relevant to the phenomena described in the Introduction. T h e next three sections review refinements that have, at least partially, been discussed in previous papers.
s
-
s
111. The Rebound from Fear to Relief
How does the offset of shock a t a,+ gcncrate a transient rebound of activity a t a,-? A previous paper (Grossbcrg, 1972c) analyzes this question and constructs explicit rebound mechanisms. T h e simplest nonrecurrent (feed-forward) version of this mechanism is shown in Fig. 2, and is described mathematically in the Appendix. An internally generated, tonic (i.e., persistent) input I derives both the and the a,- channels. This input provides the activity that drives the rebound a t when the phasic shock-derived input J is shut off. T h e inputs I and J add up a t u1 and create signals along e13. A smaller signal is created by I in e24. At the synaptic knobs N13 and N24, transmitter is produced a t a fixed rate. This rate is inhibited by the transmitter end product at a rate proportional to transmitter concentration. Thc two processes taken together
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
J’
277
I
FIG. 2. Nonrecurrent rebound mechanism from fear to relief.
create the tendency for transmitter level in both N 1 3and Nzr to approach a fixed finite upper bound. Transmitter is also removed from N13 a t a rate jointly proportional to the signal strength in el3 and the amount of transmitter in N13. A similar release of transmitter occurs a t N N . Since the signal in e l 3 is larger than the signal in e21 whenever J > 0, less transmitter exists in N13 than in Nzi when J > 0. O n the other hand, it has been proved that the multiplicative coupling of signal strength to transmitter produces a larger output from N13 than from N z i when J > 0. Thus when shock is on, the signals emitted by u3 exceed those emitted by ua. Since these signals compete subtractively a t us and ug, only the output from u 5 is positive when shock is on. T h a t is, only the fear channel supplies incentive motivation to S.When shock is turned off, both u1 and u2 are driven by the equal tonic input I. T h e potentials at u1 and u2 rapidly equalize, as do the signals in e l 3 and e24. By contrast, the transmitters in Nla and Nz4 only slowly begin to readjust to the new input levels. The input to ug is, however, determined by a niultiplicative coupling of the signal in el3 with the transmitter level in N13. A similar coupling of e24 signal with N z , transmitter determines the input to u?. Since the transmitter level in Nzl exceeds that in N13, the input to u4 exceeds the input to 213, and hencc only uF)generates an output. Thus, after shock terminates, only the relief channel is active. Gradually the equal tonic input to u1 and u2 equalizes the amount of transmitter in N13 and N24. The two channels then annihilate cach other’s equal signals, and no outputs arise from either channel. The relief response is transient because the imbalance in transmitter accumulation caused by shock is gradually eliminated by the uniformly distributed tonic input. The identification of the accumulation-depletion-release substances in eI3 and e2%as transmitters is speculative at present. Grossberg ( 1 9 7 2 ~ cites ) compatible data. Any process with the same formal properties could do the job,
2 78
STEPHEN GROSSRERG
however; cf., accumulation of bound Ca2+and its participation in transmitter release.’ This rebound mechanism has technical properties that are relevant to the discussion below. These are the following : 1. Both fear and relief are inverted U functions of the tonic input level I . In other words, either underarousal or overarousal depresses emotional affect and incentive motivational feedback in the network. Overaroused depression is stable with respect to sensory inputs; the network is “indifferent” to emotionally charged cues. This is because sensory inputs to the fear or relief channels create only a small asymmetry in the pattern of inputs to these channels when the equal arousal inputs to these channels are large; the large equal arousal inputs tend to saturate the response of the two parallel channels. Thus, after subtractive competition between these channels, their net output is small. Underaroused depression is unstable in the sense that, after the system’s elevated thresholds are exceeded by external cues (i.e., there is not enough I input to exceed threshold in response to small J inputs), either aversive or rewarding cues can cause overreactive fear or relief responses ; network response is emotionally “irritable.” This is because sensory inputs to the fear or relief channels create an unusually large asymmetry in the total input to these channels, since the background arousal level is smaller than usual. This phenomenon formally illustrates the paradoxical fact that underarousal can be unusually aversive in some situations and unusually rewarding in others (Berlyne, 1969) . 2. There exist levels Z such that maximal relief is greater than maximal fear in response to a prolonged, but then abruptly terminated, fearful cue. I n fact, the ratio of maximal relief to maximal fear grows as I increases. By property (l),however, unduly small or large Ilevels create small absolute values of relief or fear. There exist intermediate I values, however, such that the (relief :fear) ratio is large, and the absolute size of these reactions is also large. This property is needed to make learned avoidance or escape behavior possible in the presence of fearful cues. One needs the guarantee that, although the fear channel is on, the relief channel can be so strongly activated by avoidance or escape dues that it can generate a positive net incentive motivational input to and thereby release motor activity that leads to avoidance or escape. 3. Once the transmitter levels have adjusted to a fixed input level, either a sudden decrease in arousal input or a sudden increase in fearful cue input will cause an increase in fear. Similarly, a sudden decrease in input from irrelevant cues (i.e., cues that send equal signals to the two channels) will
s,
‘ N o t e added in p r o o f : C. D. Wise, B. D. Berger, and I,. Stein, Biol. Psychiatry, 6, 1 (1973) present data suggesting that a norepinephrine reward system and a serotonin punishment system compete in parallel for relative dominance.
ATTENTION,
REINFORCEMENT, AND DISCRIMINATION
2 79
cause an increase in fear. By contrast, a sudden increase in arousal and/or irrelevant cue input, or a sudden decrease in fearful cue input, will tend to create a relief rebound. 4. More relief is generated by shutting off J units of shock than J / 2 units of shock. More relief is generated by shutting off J/2 units of shock than by cutting the shock level from J units to J/2 units. A relationship exists between the rewarding effect of cutting the shock level in half, the . size of the (re1ief:fear) ratio, and the arousal level (Grossberg, 1 9 7 2 ~ )This will be extended in Section XI to include the size of the partial reinforcement acquisition effect. This rebound mechanism is coupled to the learning mechanism in Fig. 3. Sampling by channels occurs at u3 and v4 for two reasons: (a) it must occur after the accumulation-depletion step to be able to sample the rebound; (b) it must occur before the subtractive stage in order to ensure that not both fear and relief control behavior a t any instant of time. Another reason is given in Section V. A noncurrent rebound mechanism is not capable of higher-order instrumental conditioning, i.e., of instrumentally motivated “chaining” (Kelleher, 1966). For example, the offset of a cue that was previously paired with shock could not be used to reward escape behavior (Maier et al., 1969). T o make higher-order instrumental conditioning possible, the network must be modified so that offset of activity in a conditioned + af+channel can drive a rebound from a,+ to a,-. I n the above example, a cue that was paired with shock has a strong --f @f+ pathway. Offset of this cue will reward escape behavior if it elicits a rebound at af-.Thus must send axons to a stage prior to the rebound, and these axom will have conditionable synaptic knobs.
s
s
s
s
I FIG.3. Interaction of learning and rebound mechanism.
280
S T E P H E N GROSSRERG
s
must also send its axons to a stage after the rebound, so that cues can sample the fear and relief reactions. Thus the anatomy of (Ctf+, @-) is recurrent to guarantee that the rebound occurs both before and after the stage where samples the rebound. See Fig. 4 for some recurrent anatomies. I n Fig. 4a, the cell sites u1 and u3 of Fig. 2 are identified, as are the sites u p and u4. Figure 4c is particularly interesting, since it permits the learning of a stable conditioned-avoidance response that is motivated on performance trials by relief rather than fear (Maier et al., 1969). The tonic source Z is moved downstream from the ---f @, sampling axons. There it can drive the rebound, which occurs still further downstream, but it does not countercondition patterns in the S + @, axons whenever sampling by occurs. The outputs from the rebound stage compete before they are fed back to be sampled by This feedback is positive only if one of the channels is stronger than the other. Thus the tonic input Z alone cannot generate any feedback, and therefore does not countercondition patterns encoded in + Ct synaptic knobs. Irrelevant cues in can, however, countercondition
s
s
s
s.
s
s
(C)
(d)
FIG. 4. Some recurrent rebound mechanisms coupled to the learning mechanism.
A T T E N T I O N , REINFOK(:EMENT, A N D DISCRIMINATION
28 1
conditioned S -+ @ pathways. Such cues send equal signals to a,+ and @f-, and these signals can extinguish the pattern in other active S + a pathways by contiguity. Even this source of extinction can be removed by a slight modification of network design : require that only the feedback signal along the recurrent loop can cause changes in the long-term memory of the S -+ a synaptic knobs. Then neither irrelevant cues nor the tonic input can countercondition + @ knobs. Section XX discusses this modification in greater detail. Section I V discusses ways to minimize the possibility of saturating the feedback loop with irrelevant cue and tonic inputs. Of course, if the avoidance or escape response does not remove 0 from sources of fearful cues, then the strong -+ a,+ connections which these cues control can countercondition the -+ @-, channels that motivate avoidance. In Grossberg (1972c), the inputs I and J add up a t u1, and their influence decays exponentially through time. Denote the response of u1 (its activity, stimulus trace, or short-term memory trace) by X I . There exists a signal threshold r in e13, such that the signal strength in el3 is zero if x l ( t ) 5 r and is a linear function of x l ( t ) - r if x I ( t ) > r. The inverted U in the relief channel, as a function of arousal level Z, does not depend on the threshold I‘, but the inverted U in the fear channel does; in fact, the amount of fear is a decreasing function of Z once x1 2 r. Zn vivo, signal functions are not always linear functions of activity above a threshold cutoff. Often they are sigmoid functions of activity (Kernell, 1965a,b; Rall, 1955). Section I V discusses the importance of this property for the processing of neural signals in noise when the network is recurrent. Figure 4 shows that rebound mechanisms often have a recurrent anatomy. We therefore consider how the rebound mechanism is altered by making the output signals in el:
s
s
s
S T E P H E N GROSSBERG
I2
I3
J
J
(C)
FIG. 5. Potentiation or compression of response to phasic input as arousal level varies.
The interaction of arousal level I and phasic increment J in Fig. 5 recalls Berlyne’s (1969) discussion of the interaction between arousal level and optimal arousal increment. In Section XV, the effect will be used to discuss how a novel light change can reinforce lever pressing.
IV. Short-Term Memory and Total Activity Normalization Several constraints must be placed on S + a signals. All these constraints can be guaranteed by the same mechanism, which is analyzed in Grossberg (1973). These constraints are listed below. They amount to predictions about the global behavior of recurrent on-center off-surround anatomies undergoing shunting interactions.
ATTENTION,
REINFORCI<MENT, A N D DISCRIMINATION
283
1. T h e total S + Q. output must have an upper bound that is independent of the number and intensity of --+ Q. channels that are active a t any time. Otherwise the S + @ channels, in the absence of drive inputs, could activate + S feedback. In such a network, persistent feeding could occur in the absence of hunger. Similarly, a single active S + @ channel, combined with a prepotent drive input, should be able to activate Q.+ S feedback, even though the drive input alone should not be able to do so. Thus the total activity of S, rather than just its upper bound, must be carefully regulated. Such regulation can also prevent the feedback loop in the rebound mechanism of Section 111 from being saturated by irrelevant cues. We call such regulation of total activity normalization. 2. Consider an experiment in which lever pressing preceded by a tone is rewarded, whereas lever pressing in the absence of the tone is punished. What mechanism keeps the internal representations of the tone and the lever press active in until the reward or punishment can alter the probability of lever pressing in response to thc tone-plus-cues of the lever, as opposed to cues of the lever alone? A mechanism for storing these representations in short-term memory (STM) is needed. 3. Consider the processing of a pattern of inputs delivered to an ensemble of noninteracting cell populations. Suppose that neural noise exists in these populations. In many neural systems, noise cannot be avoided, if only because they operate near the quantum range, as in the case of sensory systems. I n this context, if the input signals are too small, they can be lost in the noise. If they are too large, they can saturate their respective populations, thereby creating a uniform pattern of excitation across populations and destroying all information about the input pattern. To avoid these extremes in the noninteracting case, input intensities would have to be restricted to a narrow range, and the ability to process arbitrary patterns with fluctuating input intensities would be lost. The following mechanism regulates total network activity, is capable of STM, quenches network noise, and permits the effective processing of arbitrary input patterns without saturation. T h e simplest example of this mechanism will be reviewed below. I t is a network with a recurrent on-center off-surround anatomy whose interactions are of shunting type. The network is defined by the system
s
s
Xi =
-AX;
+ ( B - xi)f(x
i)
- ~ i Z+ k if(xk)
+ Zi
(1)
where i = i, 2, . . . , n, and u i ( < B ) is the mean activity of the ith cell, or cell population, ui of the network. Four effects determine this system: (a) exponential decay, via the term - h i ; (b) shunting self-excitation, via the term ( B - x i ) f ( x i ) ; (c) shunting inhibition of other populations, via the term if(^,^); and (d) externally applied inputs, via the term Ii.T h e
284
STEPHEN GROSSBERG
function f ( w ) describes the mean output signal of a given population as a function of its mean activity w. Zn uzuo, f(w)is often sigmoid function of w (Kernell, 1965a,b; Rall, 1955). This is an important property of the above model for the effective processing of signals in noise. T h e system in (a) can be formally motivated as follows. Consider n states ut whose responses x t ( t ) to inputs Z,(t) are linear, return to equilibrium (say 0) in the absence of inputs, and have a finite maximum (say B ) . Then i t= -Ax, ( B - xJZt. Term ( B - x J Z t says that inactive sitc’s become activatcd a t a rate jointly proportional to the number of inactive sites and the excitatory input size. At equilibrium, x z = BZ,(A which approaches B as I , becomes large. This system saturates and is not normalized. Both these difficulties vanish if an off-surround with shunting interaction exists. Then
+
+
X, =
-AX,
+ ( B - xZ)Zz- x , Z i g J k
The new term says that active sites become deactivated a t a rate jointly proportional to the number of active sites and the inhibitory input size. At equilibrium, x z = BZ,(A Z)-I, where Z = & Z k . Letting 8 , = ZJW1, we find x z = 8,BZ(A Z)-l. T h e activity x Lis proportional to 8, no matter how large Z becomes, and the total activity x = Z / x r never exceeds B. Such a system, however, is not capable of STM. The anatomy is made recurrent by replacing each Z,(t) byf(x,). External inputs are then added on to get ( u ) . The STM capabilities of recurrent networks carry with thern possible difficulties. If these networks can reverberate patterns imposed by external inputs, then why do they not also reverberate their own noise indefinitely, thereby flooding the network with its own noise? The answer is that they do, if the signal function f ( w ) is improperly chosen. For example, if f ( w ) is a linear function of w,or a function that grows slower than linearly, such as f ( w ) = w(1 w - l , then noise, even in the absence of signals, will be amplified and reverberated. If f ( w ) grows faster than linearly, such as f ( w ) = w2,then this problem is avoided. Sufficiently small noise values will dissipate through time. If a brief, but sufficiently intense, input pattern is imposed on the noise, however, then two things happen. First, all populations which receive the largest input in the pattern will suppress the activity in all other populations, including the noise. Second, if the function g ( w ) = za-’f(.i) is convex, then normalzzat~onoccurs: the total activity x ( t ) = 2 I l x l ( t ) of all the populations approaches a fixed positive limit through time. T h e first property shows that an extreme form of contrast enhancement occurs: only the peaks of the input pattern survive. If one population of the network receives more input than any other, then the network “chooses” this population and quenches all others. T h e second property shows that the system precisely regulates its total activity and can
+
+
+
ATTENTION,
R E I N F O R C E M E N T , AND DISCRIMINATION
285
store the activity of certain populations indefinitely in S T M by reverberating their activity through excitatory recurrent interneuronal loops. The first property is too strong: too much of the pattern is suppressed in the attempt to suppress the noise. How can this be avoided? The way is to choose f ( w ) so that it grows faster than linearly for small values of w,and (approximately) linearly at larger values of w. Then noise dissipates, and there exists a quenching threshold. This means that, given a sufficiently energetic pattern of inputs, the activities of populations which fall below the threshold are quenched (including noise) and those which fall above the threshold are contrast enhanced and stored in STM. Speaking intuitively, this is due to three interacting effects. The pattern is contrast enhanced for activities at which f ( w ) grows faster than linear, but is preserved for activities at which f ( w ) is linear. Normalization of actibity can drive the system from the first to the second region of f ( w ) growth. Thus, a pattern that is partially contrast-enhanced at small activity levels is preserved after it is normalized in the linear f ( w ) range. Various applications of this system, and its mathematical properties, are found in Grossberg (1973). An important property of the system is shown in Fig. 6. The function g ( w ) alluded in Fig. 6 is defined by g(w)= w-tf(w); g ( w ) tests how the signal f(w)deviates from a linear function as the activity level w increases. For example, if g ( w ) increases for small values of w and decreases for large values of w , then f(w)will be a sigmoid function of w.T h e arrows in Fig. 6 depict the direction in which the total activity x ( t ) = Z ; = ~ x n ( t ) will change if it falls in a given region between two adjacent vertical dotted lines. I n Fig. ba, g ( w ) is convex, and only one stable equilibrium point exists. Either x ( t ) ---f 0 as t + a if x ( 0 ) is small, or x ( t ) approaches the unique limit El. In Fig. bb, g ( w ) is not convex, and two positive limits El and Ez for x ( t ) exist. This property motivates thc following possibility, which is illustrated in Fig. 7. In Fig. 7, a nonspecific arousal input J A combines with a specific input J , at each population u z . Two important cases arise. I n case 1, J A and J , combine multiplicatively to influence the activity level x z . Input J A is said to shunt the activity level (Grossberg, 1973b). In case 2, J A and J , combine additively to influence the activity level x L . Consider case 1 for definiteness. Then the input J A does not change the relatzve input levels to the various populations. (In case 2, a large J,I tends to uniformize any pattern of Jt s.) Let J A be parametrically increased to ever higher levels. One hereby increases the number of populations that receive enough input to exceed the quenching threshold and are stored in S T M . Conversely, reducing J A decreases the number of populations that will be stored. Thus, given an input pattern in which many inputs are close to each other in relative size, one way to “make a choice” between populations is to lower the arousal level
286
STEPHEN GROSSBERG
Q
IE~~
G
~
E
,
FIG. 6. Stable limit points of total activity.
of the input until only one population exceeds the quenching threshold; in common parlance, put the network in a quiet place. By contrast, one way to make as many cues as possible relevant to further network processing is to substantially increase the arousal level. Thus, suppose that a “novel” stimulus excites the network’s nonspecific arousal source. Then all recently presented cues can have their network representations brought into STM 10 play a part in further network processing, including the sampling and subsequent learning of motor responses (Grossberg, 197213). I n this way, novel or unpredictable events can bring all possible information about presently available cues into an active state, to enhance the network’s ability to deal with the unexpected situation. A similar effect could be achieved if an increase in arousal lowered the quenching threshold by, say, decreasing the signal strength in the inhibitory
ATTENTION,
R E I N F O R C E M E N T , AND DISCRIMINATION
PATTERN BEFORE
PATTERN
PATTERN
PATTERN AFTER
BEFORE
287
AFTER
FIG. 7. Influence of arousal on final pattern stored in short-term memory.
off-surround using arousal-initiated inhibition of the inhibitory interneurons. Grossberg (1973) discusses how such a network can be thrown into “seizure” under abnormal circumstances. Thus, we suppose that the cell populations in are interconnected by a recurrent on-center off-surround anatomy whose interactions are of shunting type. This anatomy does not, of course, accomplish all the tasks that living sensory processors undertake. For example, it does not generate a sequential STM buffer (Atkinson and Shiffrin, 1968); it has no hierarchical structure. A later paper will investigate this extension in a setting that preserves properties (1)-(3) above.
s
288
STEPHEN GROSSBERG
V. Sensory-Drive Heterarchy
s
Several constraints must be placed on @ -+ feedback. These constraints are the following: 1. The total @ + ,'$, output must have an upper bound that is independent of the number and intensity of a --f ,'$, channels that are active a t any time. Otherwise, since the @ 4 channel is nonspecific, it could activate S --f 3 Z sampling by channels that receive no sensory inputs. Precise regulation of total Q. -+ s output would also provide a steady baseline of incentive motivation to activate compatible motor activity. 2. Consider the situation in which a student regularly eats meals in spite of the prolonged absence of a sexual partner. A positive, but nonprepotent, drive can control motor behavior in the presence of compatible sensory cues (e.g., eating food if hungry), if cues compatible with the prepotent drive are unavailable (e.g., absence of a sexual partner). 3 . Simultaneous Q. + feedback from two or more incompatible drives must be prevented to avoid the occurrence of incompatible motor commands. Property (3) can be achieved by any mechanism that can make a choice among n populations. For example, a recurrent on-center off-surround network with shunting interactions can achieve this if its quenching threshold is sufficiently large. A nonrecurrent on-center off-surround network with additive interactions can also achieve this (Grossberg, 1970). The system
s
s
s
xi =
-Ax;
+ zi -
&#ilk
with inputs Zi, i = 1, 2, . . . , n, is of this type. At any given time, only the population ui whose input Zi is maximal receives a net positive total input (Zi - & + i l k ) = 2 Z ( O ; - 56). All other populations receive negative inputs that drive their activity to subthreshold values. The nonrecurrent mechanism is incapable of STM. I t is driven, however, by signals from S which are capable of STM. Thus, the outputs from a nonrecurrent mechanism that chooses among @ -+ signals can be sustained by the S T M reverberation in S that drives 8 -+ @ signals. Property (2) requires that signals from sensory cues and from drive inputs combine before the choice mechanism of property ( 3 ) determines @ -+ S feedback. @-+ ,'$, feedback is then determined by the dominant combination of sensory-plus-drive cues (Fig. 8a) rather than by the dominant drive level (Fig. 8b). Figure 8a shows that a sensory-drive heterarchy can be established if the normalizer occurs after the stage at wbich sensory and drive inputs mix, after the stage at which -+ @ sampling occurs, after the stage at which a relief rebound occurs, and before @ + feedback signals can influence S. The normalizer determines a consensus between all the possible Q. + ,'$, feed-
s
s
ATTENTION,
REINFORCEMENT, A N D DISCRIMINATION
289
u3
FIG. 8. Sensory-drive heterarchy vs drive hierarchy.
back channels, and thus its inhibitory interactions cut across the channels controlled by different drives. Property (1) can be achieved by a recurrent on-center off-surround network with shunting interactions. This mechanism also regulates total activity of Q. --f S feedback, and therefore provides a steady baseline of incentive motivation to activate compatible motor activity. Given a nonrecurrent choice mechanism, total activity is bounded from above, but its precise regulation can be achieved only indirectly by the regulation of total S -+ Q. output.
290
STEPHEN GROSSBERG
VI. Conditionable C% + S Feedback and Psychological Set
The following unfortunate phenomenon can occur in the network thus far discussed. Suppose that a particular U C R (e.g., salivation) is encoded in S + 312 synaptic knobs when a given CS (e.g., ringing bell) and a UCS1 (e.g., smell of food) compatible with drive D1 (e.g., hunger) is active. Suppose on performance trials that a different drive DZ (e.g., sex) is stronger than D1; indeed, suppose that D1 has been satisfied, for definiteness. Then the U C R can be released if a UCSz (e.g., smell of mate) compatible with D Z is presented along with the CS. Such a network can release persistent eating behavior in the absence of hunger if it is sexually aroused by sexual cues other than a mate. This difficulty is due to the fact that (3, -+ S feedback in response to any given drive is nonspecific. O n performance trials, UCSZ supplies sensory input that combines with Dz drive input to release nonspecific a + feedback. In particular, the channels activated by the CS receive this feedback, and their 312 pathways are activated, in spite of the fact that drive D Z and the CS are unrelated. This difficulty can be cured by letting the a + pathways have conditionable synaptic knobs. Then the nonspecific 0,-+ S signals that are released by a given drive will become conditioned to the S representations that are compatible with (i.e., persistently contiguous in time with) this drive. See Fig. 9. In particular, if a cue compatible with drive Dj is activated and drive
s+
s
s
s
STRONG FEEDBACK
ui I
0
r FIG. 9. Conditionable
a+
---
feedback establishes psychological set.
29 1
ATTENTION, REINFORCI<MENT, AND DISCRIMINATION
s
Di is unsatisfied, then a-+ feedback controlled by this cue will activate the stages Ui3 of all representations that are compatible with Di. In this way, motor activity compatible with the cue can be released, and a “psychological set,” or predisposition to fire, is established in all the Uiz stages that are compatible with Dj. Making a + S synapses conditionable increases the symmetry of the network. Now both a + and S + a synapses are conditionable, and these pathways fire only in response to a specific cue plus a nonspecific arousal input. Do both types of cells have a common phylogenetic ancestor? We want conditioning of synaptic knobs to occur in both excitatory and inhibitory C i -+ $, pathways. Then inhibitory and excitatory psychological sets can be switched on separately by different cues, but an representation that often samples when both sets are on will control a mixture of excitatory and inhibitory feedback, and hence its net feedback can be small. The next section shows how to accomplish this without preventing attentiveness to fearful cues.
s
s
s
VII. The Persistence of learned Meanings
Now we are ready to begin a study of attentional factors in these networks. The networks have the following unfortunate property. Let CSi be a conditioned reinforcer for drive Di,i = 1,2. For example, let CS, be a roast turkey, CS, be one’s lover, D, be hunger, and D, be sex. Consider the situation of having roast turkey for dinner with one’s lover. At dinner, both CS, and CS, are scanned intermittently in rapid succession, or even simultaneously. I n daily life, we do not come away from the dinner table labeling our lover as a source of food and the turkey as a source of sex, as would happen if all contiguous cues were always mutually conditioned to their respective responses. Fortunately, the learned meanings of cues can endure in spite of parallel presentation of cues with different drive representations. Of course, if the turkey is persistently and consistently paired with all of our sexual encounters, then turkey can become a discriminative cue for sex, just as pairing turkey which shock can make us afraid of turkey. The above example distinguishes the forced pairing of events from the free reorganization of attention through time. How can persistence of learned meanings be achieved in these networks? Let the representations that are activated by CSi be denoted by S;,i = 1, 2. We want 0 to be able to “notice” each C S i as it is scanned. Thus each CSi should be able to activate its Si. We also want to prevent sustained simultaneous sampling of a by S1 and s2. Otherwise, 81 would activate the a1 channel, and $2 would sample this channel and strengthen its a1 connec-
s
292
STEPHEN GROSSBERG
tion. Simultaneously S z would activate &, and s1 would strengthen its a1 connection. I n effect, the turkey would become a cue for sex and the lover would become a cue for eating. Our task is to prevent sustained simultaneous sampling of B by S 1 and S Zif S1 and S Z project to incompatible drives. T o achieve this in the present context, at least three stages of processing are needed: 1. s1 and 5 2 send signals to 8 in order to test which B channels they control (e.g., do s1 and sz control incompatible drives?). 2. The a + 5 feedback measures which ai channel is stronger at any time, via the sensory-drive heterarchy. 3. Si-+ a sampling is shut off in the weaker channel. Stage 1 is accomplished when the CSi inputs activate Si + @i signals. Stage 2 is accomplished by the sensory-drive heterarchy. How is stage 3 accomplished? Suppose for definiteness that the 81 channel is stronger; that is, the strength of sensory-plus-drive inputs to B1 exceeds that to ( 3 2 , so that only a1 -+ S feedback is positive. Somehow this feedback must suppress SZ+ B sampling. By Section VI, al -+ feedback will be received only by the S1 representation. How does this feedback suppress $ 2 reverberation and sampling? An answer is suggested by Section IV. The total activity of the S representations is normalized, and a quenching threshold exists. We want strong S1 feedback to enhance the activity of the S1 representation and, as a consequence of normalization, to thereby, at least partially, suppress the activity, and hence the sampling, of the representation. T h e minimal way to accomplish this is to require that sFec$c Ui2 -+ Uil signals exist in each representation (see Fig. 10a). I n Fig. 10a, strong a1 + U I Z -+ U11 feedback increases the strength of activity in the Ui1 population relative to the activity in the Uzl population. The U Z Iactivity is thereupon suppressed by inhibitory signals from U11 to UZI. The above argument holds if the drives in question control positive a + S feedback. The case of drives, such as fear and frustration, which control negative feedback requires further argument. The problem is this. If the conditioned feedback is negative, then it will tend to differentially suppress activity in the controlling representation, rather than to enhance it. This would have the following maladaptive effect on behavior. Increasing the learned fearfulness of a given cue, in a fixed context of other cues, would decrease the attention paid to it. Jumping ahead in our discussion for a moment, we also would note that fearful cues could not overshadow or block learning in response to other cues, which is false (Kamin, 1968, 1969). Hence a distinction must be made between mechanisms for learned persistence of negative meanings and for negative incentive motivation. See Figs. 10b and 1Oc. The former feedback channel helps to focus attention on particular cues. The latter feedback channel suppresses motor activity. The
s
s
s
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
293
a,
(C
1
FIG. 10. Attentional feedback and motor arousal occupy different channels.
attentional feedback is always positive, even if the drive in question controls negative incentive motivation. The synapses of this feedback channel are, moreover, conditionable, so that the feedback can enhance the activity of particular representations on which attention will focus. Given that attentional feedback is conditionable, is it also necessary to make the a + incentive motivational synapses conditionable? At any given time, the conditionable attentional feedback will guarantee that only certain Uil stages will send signals to their respective Uiz stages. Can a n irrelevant drive release --f 3n sampling in the manner described by Section VI? If the irrelevant drive creates conditioned attentional feedback to its “psychological set” in S, then this set will tend to quench other sources in S, and therefore to prevent their firing. Thus the a + incentive motivational feedback is not necessarily conditionable, although making it conditionable could only improve network efficiency. T h e two kinds of feedback can be interpreted as slow potential shifts. The conditionable attentional feedback is reminiscent of the contingent negative variation, or CNV (Cohen, 1969). Such a wave has been associated
s
s
s
294
S T E P H E N GROSSBERG
with an animal’s expectancy, decision (Walter, 1964), motivation (Irwin et al., 1966; Cant and Bickford, 1967), volition (McAdam et al., 1966), preparatory set (Low et al., 1966), and arousal (McAdam, 1969). Walter (1964) hypothesized that the CNV is a conditionable shift in the average baseline of the cortex, acting to depolarize its apical dendritic potentials and to thereby prime the cortex for action by reducing its overt response threshold to other inputs. The incentive motivational feedback acts more as a form of motor arousal or suppression, since it controls whether or not the -+ m channels will fire. Thus, the above model suggests that at a stage following the sensorydrive heterarchy, feedback channels to sensory-motor areas should bifurcate; one channel, as in the case of the CNV, should be related to an animal’s attentional state, and is influenced by drives, motivation, arousal, etc. The second channel should be capable of enhancing or depressing motor output.
s
VIII. Overshadowing and the Triggering of Arousal by Unexpected Events
Adding the feedback connections Ut2-+ Uzl, or more generally from Q. to U,1 (cf. Fig. lOc), gives rise to phenomena like those reported in Sec-
tion 1,A. At the outset of Experiment 1 in that Section, neither CS1 nor CS2 projects to any particular drive representation in a. Thus both CS1 and CS2 can sample the fear representation when shock is on. Since the total S --+ output is normalized, the strength of St+ (3 signals depends on how many sLchannels are active at any time. Thus learning by 81 and s2 activated together will be slower than learning by $1 activated alone, unless there exists more than one limit point for x ( t ) , as in Fig. 6b. In experiment 2, CS1 fiist becomes conditioned to fear, which we will call the a1 channel. The channels S1+ a1 and a1+ both become conditioned during the first phase of this expcriment. When CS1 and CSt are Z U11 feedback presented in phase 2 of the experiment, Ull -+ a1 -+ U ~+ suppresses sampling in the U21-+ 61 channel before CS2 can become conditioned to fear. CS2 is hereby overshadowed by prior fear conditioning to CSI. Suppose in experiment 1 that CS1 is more salient than CS2. Then the sampling signals from S1 to a will initially be larger than those from SZto a. Consequently learning in 81’ 61 synaptic knobs will occur faster than learning in + a2 synaptic knobs. Similarly, learning in the feedback channel O1 -+ s1 will occur faster than learning in the a2-t sz channel. The U11+ 61-+ Ulz --+ U11 feedback therefore grows faster than the UZl+
s1
s2
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
295
a1-+ U Z Z-+ Uzl feedback. Sampling by U21is hereby gradually suppressed as learning trials proceed, and CS, is gradually overshadowed by CS1. Similarly, if a CS and UCS are simultaneously presented, then the UCS can overshadow the CS via S -+ a -+ feedback. If the CS occurs shortly before the UCS, then its sampling channels are active in the time interval after the UCS occurs and before S -+ a -+ feedback can quench their activity. Hence CS -+ U C R conditioning is possible in this latter case. Experiment 3 is not so easily approached. Somehow, the occurrence of an unexpected UCS must prevent CS1 from overshadowing CS2. Either the U11-+ a1 -+ Uzl-+ U11 feedback must be weakened, or an independent nonspecific (e.g., “arousal”) input to must keep activity at in the suprathreshold range. Weakening U11-+ Utl ---f U11 feedback does not seem to be a physically plausible way to overcome overshadowing. T o see this, change experiment 3 as follows: in phase 1 of the experiment ( C S l - + shock), use 40 units of shock, and in phase 2 of the experiment (CS1 C S 2 + shock), use 80 units of shock. T h e increase in shock level is unexpected, but it should surely be accompanied by an increase in a1 -+ Ul2 feedback. Indeed, the very survival of an animal can depend on its ability to process the reinforcing characteristics of unexpected events. The increase in @I-+ Ulz feedback would increase the overshadowing of CS2 by CSI, other things equal, but just the reverse occurs in vivo. Overshadowing can be eliminated, or a t least reduced, if unexpected events transiently increase the nonspecific arousal of and thus the number of S representations whose activity exceeds quenching threshold. This increase in overall arousal of competes with overshadowing tendencies -+ channels. Alternatively, it is possible controlled by motivational that unexpected events transiently decrease the quenching threshold of S. The latter effect could be achieved, say, by letting a n unexpected event trigger shunting inhibition of the inhibitory interneurons in the off-surround of each population in The triggering of arousal by unexpected events will be seen to be a basic feature of the model for dealing with a variety of phenomena (see Fig. 11). For example, the Appendix derives a formula showing that CS2 can become a learned source of relief, rather than of fear, if the shock level that follows CS1 CSr is lower than the shock level that follows CSI. This can be achieved using the increase in tonic arousal input to a that accompanies the unexpected change in shock level (cf Section IX). The increase in arousal a t enhances the tendency for a relief rebound to occur, whereas the increase of arousal at overcomes overshadowing and enables S -+ a sampling of this rebound to occur. By contributing an increase in irrelevant cue input to a, the increase in arousal at can also enhance the relief rebound at a.
s
s
s
s2
+
s,
s
s
s.
+
s
s
2 96
STEPHEN GROSSBERG
FIG. 1 1. Arousal-initiated inhibition of inhibitory interneurons.
Sec:tion XXI notes some possible clinical differences that would arise due to overactivity of nonspecific arousal to S vs overactivity of conditionable Q. + S feedback, especially with regard to the hypothesis that certain schizophrenic symptoms are due to imbalances in catecholamine production. The above conclusions can be phrased in a way that emphasizes the adaptability of a network to changing environmental demands as a fundamental principle of its design. The mechanism for preserving learned meanings of cues is an adaptive attentional mechanism that permits parallel processing of cues without spurious cross-conditioning of the learned meanings of all cues. Overshadowing is a consequence of this mechanism. Overshadowing can, however, yield maladaptive network performance if the environment changes, or is only partially understood, since then the cues that presently control network output will be imperfect predictors of environmental response to this output. The property of persistence, by itself, creates too rigid a network. Taken together with the liberating effect of unexpected events on nonspecific arousal (or the quenching threshold), it can achieve a stable, but adaptively changeable, attentional mechanism. The above discussion reduces the overshadowing problem to the problem of how arousal is triggered by an unexpected, but not by a n expected, event. This latter problem can be restated in an informative way: how does a network habituate (Grossman, 1967) to a repetitively presented, and therefore increasingly expected, event? A mechanism whereby network output
ATTENTION,
R E I N F O R C E M E N T , AND DISCRIMINATION
297
is regulated by the expectedness of an event is described in Grossberg (1972a), and will be applied to the present case in Section XIX.
IX. Pavlovian Fear Extinction vs Persistent learned Avoidance
T h e above results suggest a mechanism for the fact that classically conditioned fear can rapidly extinguish, whereas learned asymptotic avoidance behavior can be very stable. An explanation that uses the concept of expectation in a descriptive psychological theory has been given by Seligman and Johnston (1973). Our neural explanation will use the arousing effect of unexpected events on the fear-relief dipole of Section 111. Figure 4c illustrates a mechanism in which avoidance is stable, if it does not confront the network with a new source of fearful cues. T o approach the fear extinction problem we suppose that an unexpected event transiently increases not only the arousal input to but also the arousal input I to the fear-relief dipole; e.g., imagine that both regions receive arousal from a common source, such as reticular formation. Using this hypothesis, a mechanism of fear extinction is the following. Suppose that a CSI (e.g., bell) has persistently been paired with a shock UCS. Eventually will project strongly to the fear channel a,+, and will be capable of generating a conditioned emotional response (cf. Grossberg, 197213). If on a performance trial, the CS1 is not followed by the expected shock, then a transient increase in I occurs and causes a rebound a t a / . This The sl+ a,+ channel is hereby countercondirebound is sampled by channel, since tioned by the increase in relative strength of the $1’ the net positive feedback controlled by S1 decreases. If the fear has been suppressing consummatory activity based on a positive drive, then spontaneous recovery of this activity can occur; the $ + 3?Z synapses which encoded the activity were not counterconditioned by fear suppression, and the positive incentive motivation that originally activated these synapses is no longer inhibited by fear (Grossberg, 1972b). A similar rebound effect, triggered by arousal subsequent to an unexpected event, can be used to approach a neural mechanism of frustration.
s,
s1
s1.
X. Frustration
Let a CS1 (e.g., bell) supported by drive Dl (e.g., hunger) be conditioned to a response (e.g., lever press) to satisfy 9 1 (e.g., with food that appears after the lever is pressed). Suppose that the expected food does not appear. How does the network prevent itself from persistently responding to this
29%
STEPHEN GROSSBERG
CS1 with lever pressing for food? This problem can be phrased in a more general way as follows: How does an organism stop persistently performing learned motor acts which no longer satisfy its needs, and free itself to seek new sources of gratification before it suffers irreversible damage due to prolonged deprivation? T o accomplish this in the networks which have already been derived, we want the nonoccurrence of the expected event to create a negative incentive-motivational output that can be sampled by Thereafter, the occurrence of CS1 will create signals from both to the positive incentivemotivational source that used to support the motor act, as well as to the negative incentive-motivational source. The net incentive-motivation will decrease until CS1 no longer elicits the erroneous response. Clearly this situation is analogous to that involving fear and relief. This analogy is depicted graphically in Fig. 12. I n Figure 12a, a sudden reduction in shock or -+ @,f+ input, or a sudden increase in irrelevant -+ or I input tend to cause a rebound at a,-. I n Fig. 12b, suppose that the network is engaged in a sequence of behaviors compatible with hunger. Then a persistent, large -+ &+ input drives positive &+ -+ feedback that supports this behavior. Regulating this input through time is one of the tasks of the recurrent normalizer in S. Suppose that the nonoccurrence of the expected event follows this sequence of acts. Because the expected event does not occur, S + ah+ input can suddenly decrease. (Temporarily ignore the case in which an unexpected event projects to a h + . ) Simultaneously the nonoccurrence creates an increase in I. Both these factors conspire to create a negative incentive-motivational rebound at ah- that S1 can sample. Also note that an increase of arousal to ,$ can also create an increase in the total + Q. input due to cues that are “irrelevant” with respect to the ( a h + , a h - ) dipole. This input can also contribute to the rebound at a h - .
sl.
s1
s
s
s
s
(a 1
FIG. 12. Fear-relief
(b)
( a ) and hunger-frustration
(b) dipoles.
ATTENTION, REINFORCEMENT,
AND DISCRIMINATION
2 99
We hereby assume, as in the case of fear and relief, that euery drive capable of generating incentive motivational feedback has a complementary drive capable of generating incentive motivational feedback of the opposite sign. I n short, drives are organized in dipoles. The properties of the fear-relief dipole are assumed to occur also in all other dipoles; e.g., a sudden reduction in expected reward can cause frustration, just as a sudden reduction in expected shock can cause relief. We also imagine that all negative incentive motivational sources are grouped together anatomically, and that all positive incentive motivational sources are grouped together. The dipole organization then becomes a uniuersal feature of network design. The different behavioral meanings of particular dipoles are determined by the particular input sources that perturb them (e.g., shock at a given negative source, or metabolic levels parameterizing deprivation states at various positive sources) . This anatomical grouping into positive and negative sources is reminiscent of the organization of lateral hypothalamus as a reward center and of ventromedial hypothalamus as a punishment center (Grossman, 1967) (see Fig. 1 3 ) . I n Fig. 13, each dipole receives a tonic arousal input whose size through time is influenced by the unexpectedness of events. If the dipoles are arranged in a regular fashion (e.g., a row), then these arousal sources can also be regularly laid out in the network. A plausible candidate for these arousal sources is the reticular formation (Thompson, 1967). Perhaps the “poker chip” organization of the dipoles is one reason for the “poker chip” organization of reticular formation anatomy that has been so elegantly investigated by the Scheibels (1967). A
FIG. 13. Regular organization of dipoles and supportive arousal sources.
300
S T E P H E N GROS SBERG
' \
INCENTIVE MOTlVATl ON
LTM
L TM (HABIT STRENGTHS)
( CONDITION E D
RE1 NFORCERS)
FIG. 14. Influence of motivational feedback on transfer of short-term memory (STM) to long-term memory ( L T M ) .
The normalizer that determines the final sensory-drive heterarchical feedback to must come after the dipole stage. Various data suggest the plausibility of interpreting this normalizer as an idealized analog of hippocampus. I t can, for example, maintain a baseline of learned incentive motivational feedback (Olds, 1969), its elimination can prevent transfer of shortterm memory to long-term memory (Milner, 1958), and it is organized as a recurrent on-center off-surround anatomy (Anderson et al., 1969) (see Fig. 14). Thus rather simple behavioral considerations generate a subdivision of network anatomy into components that are, a t least in a broad qualitative way, suggestive of significant neuroanatornical structures.
s
XI. Partial Reinforcement Acquisition Effect
We analyze this effect by noting an analog in the fear-relief dipole. The relief rebound that occurs after termination of a fixed shock level typically has a maximum larger than the maximum fear levcl produced by the shock. This permits learned avoidance or escape cues to create positivc net incentive motivation in the presence of fearful cues, and to thereby activate avoidance or escape behavior. The (relief: fear) ratio is, moreover, an increasing function of arousal level. The same mathematical properties hold for thc hunger-frustration dipole, but with different labels. If a network is rewarded with food for running down an alley, the cue representations S ( R ) which control this behavior will become conditioned to the hunger arousal cells ah+. If the network is rewarded on a random schedule, then the nonoccurrence of expected food
ATTENTION, REINFORCEMENT,
30 1
AND DISCRIMINATION
can create a frustrative rebound at ah-, both by decreasing S + ah+input and by increasing nonspecific arousal due to the unexpected event. T h e S ( R ) and other S representations that are active during the a h - rebound will therefore be conditioned to a h - . T h e net feedback from ( a h + , a h - ) to activated by these representations will therefore be smaller than in the case of a continuously rewarded network, for which only S ( R ) -+ ah+ conditioning occurs. I n particular, early in training, while the S ( R ) + a h + connection and the a,- rebound develop, the partially rewarded network will find the goal box less attractive than the continuously rewarded network. As training proceeds, an ever stronger S ( R ) + elL-projection develops. The frequency of reward must be adjusted to prevent the S ( R )---f a h + channel from dominating the S ( R ) + a h + channel; otherwise the network will eventually stop seeking the goal. Suppose that the frequency of reward has been suitably adjusted. Then what happens on a reward trial late in the training of a randomly rewarded network? First, the usual boost in S + a h + connection strength will occur. Second, there will be a sudden reduction in activity of the cues that are conditioned to frustration. Third, owing to the partial unexpectedness of reward, there can be a transient increase in arousal. The second and third effects both tend to create a rebound from a h - to a h + due to “frustration reduction,” just as reduction of shock intensity tends to produce a relief rebound. This rebound combines with the usual rewarding effect of food to produce an enhancement of the desirability of the goal late in training. I n short, the enhancement can be analyzed as a double rebound, with conditioning in between, from a h + to a/ -, and then back to a h + . It is ironic that the persistence of this kind of frustrating behavior can be analyzed as a composite of two effects that have a manifestly adaptive biological value. The rebound from positive to negative allows an organism to countercondition erroneous expectations before they do irreversible damage. The rebound from negative to positive allows an organism to learn escape from or avoidance of noxious events. The rebound from positive to negative to positive can, however, generate a maladaptive persistence in seeking an unlikely goal; e.g., gambling. Moreover, while cues that are conditioned to frustration are active, they can negatively bias the interpretation of other cues as possible alternatiIPes, and can suppress exploratory behavior by inhibiting positive incentives. The network can “fall into a rut,” and might tenaciously await the elusive frustration reduction that can give it some relief.
s
XII. Generalization Gradients in Discrimination learning
O u r discussion of discrimination learning will try to show how various mechanisms fit together to qualitatively generate empirical effects. A more
302
STEPHEN GROSSBERG
quantitative analysis will appear in another place. The main new assumption is that a sensory input will excite not only its own internal representation in but also the representations of closely related inputs, e.g., the auditory system in which peripheral auditory cells have “tuning curves’’ that include a connected band of frequencies (see Fig. 15). Given this elementary fact, the model already a t our disposal generates various nontrival effects. will also be interconnected by a recurrent onThe populations in center off-surround field of shunting type for the reasons cited in Section IV. There exist variations on this theme which will not be considered here, but which are being studied. For example, let each input perturb only its own representation, but let the on-center and off-surround of each representation have a generalization gradient that includes related populations; or let the inputs, the on-center, and the off-surround have such a generalization gradient. Before considering particular experiments, we now note various general properties of this system, Let C S , be an X, cps tone for definiteness. Suppose that CS, activates s1 while a,+ is the dominant active arousal source. Then all representations in which are sufficiently excited by C S , to exceed the quenching threshold of S will sample a,+ with an intensity that increases as a function of input strength. If z = h, then s h eventually controls positive a -+ feedback. If i = f, then Sf eventually controls negative a .--f feedback (see Fig. 16). Let a test CS ( C S T ) have a n representation (ST)that lies within the generalization gradients of both S h and S,. ST will become conditioned to both ah+ during presentations of csh or with a relative strcngth that depends in part on how close ST lies to these foci of input activity. Thus the net a -+ feedback controlled by ST will be a mixture of positive and negative signals. How strong will the feedback be when CST is presented on test trials? CST also has a generalization gradient that excites a band of representations, including s h and Sf. Some of these representations lie in the generalization gradients of and/or s h , and will therefore be conditioned to
s,
s
s
s
s
s
s
csf,
s
s!
s
s
c si FIG. 15. Tuning curves underlying generalization gradients.
ATTENTION, REINFORCEMENT,
AND DISCRIMINATION
303
FIG. 16. Net feedback varies along generalization gradient.
a,+ and/or
a h + with a relative strength depending on how close they lie to these foci. Thus when C S T is presented, it will excite a band of S representations that project with differing patterns to a,+ and ah+. T h e net @-+ S feedback is a composite of all these patterns after they filtered through the sensory-drive heterarchy. Recall also, however, that there is conditionable a + S feedback, specific U,z+ [/,I feedback, and normalization within S to contend with. T o understand what happens i n a qualitative way, we first make an unsatisfactory approximation, and then improve it step-by-step. First, ignore the generalization gradient in S of CST and compute the net feedback that would occur in response to activating any ST when this feedback is just the resultant of the relative ST + a,,+and s7,+ a,+ path strengths, and the total path strength is normalized (see Fig. 17a).The boldface curve in Fig. 17a shows the resultant a t any S representation ST of the gradients centered a t & and &. Note that the resultant gradient always is less than the S h gradient, but that its slope is steeper than the SJ,gradient. What is the effect of normalization by thc on-center off-surround field? T h e normalized ST gradient is shown in Fig. 17b. Its maximum is higher than the S h gradient because the positive part of the &, gradient in Fig. 17a is narrower and steeper than the S h gradient. Thus normalization of the resultant gradient produces behavioral contrast. Also there is a peak shift away from &, and a steepening of the generalization gradient due to discrimination training. T h e need for normalization, in turn, can be traced back to the need to prevent S-t Q. signals from creating a + S feedback in the absence of supporting drives. T h e various other mechanisms a t work in the network can now be switched in without changing these qualitative conclusions. Why does a pronounced peak shift not occur if the training is errorless? In our networks, errorless training i m a m that there is no fear or frustration (Bower, 1966), hence no negative gradient to interact with the positive
301.
STEPHEN GROSSBERG
(b)
FIG. 17. Normalization yields behavioral contrast based on net generalization Sradient.
gradient to cause a shift. If, however, a shock is paired with the S- after errorless discrimination training, then S - t a,+ conditioning will occur (along with a,+ conditioning of all the representations in the S gradient) and a peak shift will develop (Grusec, 1968). Bloomfield’s (1969) remarks about “an unexpected change for the worse” can readily be interpreted in this context. A sudden reduction in reinforcement frequency generates a frustrative rebound, and hence sampling by of a negative feedback source, as does introduction of shock, etc. Honig’s (1962) suggestion that the peak shift occurs only if the S and S - are presented successively, but not if they are presented simultaneously, can also be discussed. One possible reason for the latter fact is that rewarding Sf gradually gives it control over a powerful positive @-+ feedback channel. This feedback enbles Sf to overshadow s-, so that as the animal’s expectation of reward develops, the tendency to approach S simultaneously dominates the tendency to approach S -. Moreover, before this expectation develops, responding to S- will not generate a large negative rebound. I n the successive paradigm, approach to s- is not inhibited by the presence of S + , so
s
s
s
+
s-t
s
+
ATTENTION, RI:INFOKCI:MI:NT,
AND DISCRIMINATION
305
that more occasions occur after the expectation of reward develops that generate a frustrative negative rebound. Consequently, in the simultaneous paradigm S- does not gain (as much) control over a negative feedback channel, and the peak shift does not develop. Normalization can also be used to interpret the Newrnan and Baron (1965) and Newman and Benefeld (Honig, 1970) studies. These studies suggest that the color dimension can mask the orientation dimension in pigeons, but that some conditioning is occurring in the orientation dimension nonetheless. I n the Newman and Benefeld study, a vertical line on a green background during training is replaced by a vertical line on a black background during testing, and a generalization gradient is demonstrated even in the absence of discrimination training with a n S - . We suppose that removal of the green background has three effects. It (1) eliminates the strong color + orientation inhibition due to the off-surround. Thus (2) the vertical orientation representation, and its generalization gradient, become more active owing to a shift of activity in the normalized field. This gradient can thereupon sample Q. and release its learned (but previously weak) Q. + S feedback. Removal of the green background also ( 3 ) causes surprise by changing the expected line-color combination used during training, and thereby enhances activity both of the orientation representation in and of the sampled drive representations in a by triggering a transient increase in nonspecific arousal.
s
XIII. Habituation and the Hippocampus
T h e Newman and Baron study shows that a white line on a green key as S + and a green key as S- produces a generalization gradient on the orientation dimension. Why does color not overshadow orientation in this situation? A phenomenon of this type exists in our networks. It is due to the interaction of several mechanisms, namely, (1) normalization and the quenching threshold, (2 j conditioned arousal, ( 3 ) conditionable net incentive motivation, and (4)feedback of net incentive motivation to the normaltrials, color will partially overshadow ization stage. O n the initial S orientation via the off-surround in S. Denote the relevant color representations in by s(C) and the orientation representations by S ( 0 ) . Both S(C) and s ( 0 ) will sample ElL+ on these trials, but s(C) will build up a stronger connection since its activity is greater. As this occurs, a,,++ S feedback paths will become conditioned to s(C) and s(0)with a similar difference in relative strength. I n the usual overshadowing experiment, this initial advantage of S(C) over s ( 0 )will be progressively enhanced as training continues until s(C) completely overshadows s ( 0 ) . I n the present context, however, S- trials occur. O n these trials, s(0) is inactive. s(C) is active,
+
306
STEPHEN GROSSBERG
but the expected reward does not occur. A frustrative rebound is therefore generated. S(C) thereupon samples a h + . Simultaneously, a h + samples the second stages Uiz of s ( C ) . Thus the net incentive motivation controlled by s(C) is progressively diminished by frustrative nonreward. O n S trials, the net feedback from (@+h, ah-) to S(C) is cut down, owing to competition between these two channels before they release incentive motivation, but the feedback to S(0) comes only from ah+, and increases through time. Asymptotically, the S ( 0 ) activity, bolstered by s ( 0 )-+ah+-+ s ( 0 ) feedback, can dominate the s(C) activity. I n this limited sense, the network has habituated to s(C), even as it grows ever more attentive to S ( 0 ) . This habituation mechanism has several interesting properties. First, the sensory channel itself does not habituate; habituation is an active process based on interpretive feedback of sensory information via the drive representations (Grossman, 1967; Sharpless and Jasper, 1956). Second, suppose that the normalizer which creates a sensory-drive heterarchy is indeed interpreted as a simplified hippocampus. Then the hippocampus becomes involved in attentional control and the habituation of attention, but only indirectly via its determination of which motivational channel will be active in response to particular cues. Section XX describes another habituation mechanism with the property that increasingly expected, and in particular repetitively presented, events elicit progressively smaller orienting reactions. Why does a generalization gradient not occur if the S- is a red key or if the S- is a vertical white line on a red key? Then the color dimension is not habituated by frustrative rebound. Indeed, in the latter case, the orientation dimension might habituate, although perhaps at a slow rate because it is overshadowed by the dominant color dimension. Note that expectation mechanisms can interact with habituation mechanisms in two opposing ways in the above experiment. First, they contribute to the frustrative rebound during S- trials, by altering the uniform (e.g., arousal) input to ah+ and ah- on these trials. Second, they work against habituation by creating nonspecific arousal in S that tends to overcome the reduction of S + a + S feedback in particular channels, and allows them to once again reverberate in STM.
+
XIV. Overshadowing vs Enhancement
We now interpret and contrast the Honig (1969) experiments with the
F. Freeman (unpublished master’s thesis, 1967) experiment. Honig used T D and PD training sessions, followed by dimensional acquisition and finally testing on an orientation dimension. I n the TD experiment, the pigeons
ATTENTION,
R E I N F O R C E M E N T , AND DISCRIMINATION
307
were trained to make a discrimination on a dominant (namely, color) dimension. In particular, lesser dimensions were overshadowed, and the pigeon acquired a strong expectation and a positive conditioned arousal path in response to the S+ color cues. O n dimensional acquisition trials, the color cues were not present, so that the orientation dimension, no longer overshadowed, could be trained, given that the pigeon still maintains general approach tendencies to the levcr. The PD training session, by contrast, frustrates the pigeon on the dominant color dimension. Yet the reinforcement schedule has been chosen so as to overcome frustration and yield a net approach tendency. The cues that elicit learned approach are not the frustrated cues of the color dimension. These cues develop their own powerful positive feedback paths. I t is reasonable to assume that these cues are also present on the orientation dimensional acquisition trials. If they are, they will ( a t least partially) overshadow the orientation dimension both on orientation training and testing trials. By contrast, Freeman trained pigeons to peck at a vertical line on a dark key (S +) but not to peck at a line tilted at 120' on the same dark background (S -). A generalization gradient is hereby established on the orientation dimension. Then dimensional acquisition occurs with the vertical line on a green background, and one finds a nontrivial chromatic generalization gradient on testing trials. Why does the orientation dimension not overshadow the green background during dimensional acquisition? One wants to say that surprise, and hence arousal, is triggered by changing the black background to green. Then the green s(C) representations will be able to sample the positive net incentive motivation controlled by the vertical S ( 0 ) representations. This explanation works, however, only if one first can answer the question: why does the orientation dimension not overshadow the dark background during discrimination training? And if the dark background is overshadowed, and therefore irrelevant, why is the pigeon surprised if it is removed on dimensional acquisition trials? T h e importance of these questions is perhaps better seen when they are phrased as follows: If the pigeon does get surprised when the dark background is replaced by green, then why does this not happen in all overshadowing experiments when the CSI is replaced by the C S , CS2, thereby preventing overshadowing from occurring? To answer these questions, we seek differences in how the expectation mechanism (and thus arousal) responds in the Freeman experiment as opposed to the usual overshadowing experiment. We want to say that introducing green in the Freeman experiment is more surprising, say, than introducing a tone as CS, after prior CER training with a flashing light as CS1 in a Kamin-type overshadowing experiment. A difference of degree is sought
+
308
STEPHEN GROSSBERG
in the two experiments, rather than the operation of different mechanisms. We suggest that this difference exists, in part a t least, because the pigeon can develop an expectation of a vertical line on a particular visual background more easily than it can develop an expectation of a flashing light in a prescribed combination of events involving nonvisual modalities. In other words, a learned expectation can be at least partially localized to a given cluster of features or events, and features which stream into the same modality in close physical continguity can be more easily grouped together as a coherent expectation than features which enter through different modalities, other things equal. If this is true, then it might be easier to eliminate overshadowing of CS, by CS, in a Kamin-type experiment if the CS, is a vertical line and the CS, is a green background, than if the two events involve different modalities. This kind of prediction is hard to analyze completely because inputs to two different modalities are hard to equate psychophysically, and can activate orienting reactions that need not be activated by two inputs to the same modality. The Freeman experiment demonstrates enhancement due to prior discrimination training. The closely related Mackintosh and Honig experiment (Honig, 1970) demonstrates blocking. We suggest that blocking occurs because the surprise that is triggered during redundant spectral discrimination training, after orientation discrimination training has been completed, only partly overcomes overshadowing. When no prior orientation discrimination exists, and only spectral discrimination training is given, there is no overshadowing to overcome. If the above analysis is accurate, then one might be able to create a transition from overshadowing to enhancement in a given experimental setup by varying the relative strength of the attentional and surprise channels, say by drugs.
XV. Novelty and Reinforcement
Berlyne (1969) showed that a novel light change, contingent on lever pressing, can reinforce lever pressing. We suggest that the novelty of the light change, as usual, triggers nonspecific arousal which, as usual, filters through all drive representations. If a positive incentive motivational source is active when arousal occurs, and this source dominates other drive representations at that time, then the arousal will enhance the amount of positive motivation. The lever press cues S ( L ) can become differentially conditioned to the positive source, which also supplies enough incentive to trigger S ( L ) + 311 sampling of the motor commands that control the lever press. We suggest that the source of positive incentive in this case is the motor
309
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
arousal source for exploratory approach and pressing of the lever, rather a specific drive representation. We can now provide an answer to a related question: Why is the approach incentive motivation not usually the motivational source for learned goal objects? One reason is that, unless the approach source is differentially strengthened by arousal enhancement or other means, then all meaningless objects in the environment can be approached, and none will be approached more frequently than any other, other things equal. A second reason is that, when a specific drive is rewarded, then the source of positive incentive tends to shift from general exploration and approach to the specific drive representation that was rewarded. T h e enhancing effect of arousal on the pattern of activity a t drive representations can also generate incentive motivational feedback to sensory representations in the absence of external sensory cues. For example, if the hunger drive is prepotent, and all drive representations are aroused, then a h f can generate feedback to its psychological set S h in S, leading, say, after further enhancement through the feedback loop S h @h+ + S h , to the motor output “I want food.” More generally, the network can ask itself << how it feels” by arousing its drive representations. T h e resulting motivational feedback from Q. to can establish a psychological set that is capable of generating compatible motor activity. This possibility is a special case of the “two-thirds rule” discussed in Section X X I I . ---f
s
XVI. Motivation and Generalization
How does increased drive flatten generalization gradients? A formal answer exists in the networks. Increasing the drive increases positive incentive motivational signals in Q. -+ { U i z )synapses. Increasing these signals has two effects. I t speeds up conditioning in the @ + (Ui2) synapses, and it increases the signals from { U i z ) to { U t l ) . At { Uil}, the increased input allows more S representations to exceed quenching threshold, and faster conditioning occurs in the + @ synapses of these representations. How does this mechanism affect generalization gradients? If a particular S representation is activated by external cues, its generalization gradient of S representations will also be activated, albeit to a lesser extent. Increasing Q. + { Uiz} --f { Uil} signals permits more of these representations to sample drive representations in @. Thus, on testing trials, more cues in the generalization gradient can generate the type of feedback that was elicited on training trials. T h e generalization gradient is hereby flattened. Section I V also shows that if the nonspecific feedback is additive a t S sites, then it will tend to flatten the gradient by uniformizing the pattern of activity in If two drives compete, then increasing one drive can steepen the general-
s
s.
310
STEPHEN GROSSRERG
ization gradient on the other drive (Hoffman, 1969). Let the two drives be hunger and fear for definiteness, and consider the generalization of fear conditioned to a 1000 cps tone. Let the representation S ( X ) of an X cps tone be activated by the tone on testing trials. Choose X in the generalization gradient of the 1000 cps tone. O n training trials, S ( X ) sampled a,+, and possibly also ah+. Thus the strength of ah++ feedback is increased relative to the strength of a,++ S feedback, which is driven by s ( X ) + a,+ signals. T h e suppressive effect of fear is hereby reduced by increasing the hunger level. Why does this mechanism steepen the fear generalization gradient? A formal reason is that a fixed increment in positive feedback can totally overcome the suppressive effect of a sufficiently small amount of negative feedback, but has only a small relative effect on large amounts of negative feedback. An X cps generalization gradient controls large amounts of negative feedback, but tones near the edge of the 1000 cps generalization gradient control only small amounts of negative feedback. Hence the increase in hunger narrows and steepens the fear gradient.
s
s
XVII. Predictability and Ulcers
If Weiss’s experiments (1971a,b,c) on the development of stomach ulcers in rats are performed on our networks, then the net incentive motivation in the networks is a monotone increasing function of the degree of ulceration in his experiments. This analysis does not give a physiological explanation of the ulcerogenic process, but it does suggest that the frustrative sources of negative incentive are also triggered at the same time as sources of ulcerinducing agents. Why do avoidance-escape networks develop less ulceration than yoked networks? Avoidance-escape networks have been trained to respond to cues which activate positive incentive motivation that supports avoidance and/or escape activity. The positive incentive competes with the negative incentive generated by shock, and thereby reduces the net negative incentive motivation. Why does a warning signal reduce the ulceration of both groups of networks? It can do so by reducing the novelty of the shock. By Section VIII, this will reduce the arousal level that accompanies the shock, and thus the net negative incentive that the shock produces. In the avoidance-escape networks, the warning signal can also be used as a discriminative cue for activating avoidance-scape cues that switch on positive incentive motivation. Why do the yoked networks develop less severe ulcers than the avoidance-escape networks if both groups receive a brief shock after each avoid-
ATTENTION, REINFORC:ICMENT,
AND DISCRIMINATION
31 1
ance-escape response? Three effects in the network conspire to produce this result. First, the network is motivated by positive incentive in making the avoidance-escape response ; this motivational source is abruptly terminated. Second, the network expects relief after performing the response, but does not get it; this unexpected event triggers nonspecific arousal. Third, a negative, or punishing, event occurs instead of the expected relief. The first effect tends to produce a positive-to-negative rebound. The third effect creates a second source of negativity. And the second effect enhances the total negative tendency. The first and second effects are absent, or a t least much weaker, in yoked networks. Why does little ulceration develop in avoidance-escape networks if each avoidance-escape response is followed by a feedback stimulus, such as a tone? Three effects are operating in our networks. First, the avoidanceescape response produces relief, as in Section 111. Second, the novel tone, of itself, produces nonspecific arousal. As in the analyses of the fear-relief dipole in Section 111, and of the Berlyne (1969) experiments in Section XV, this arousal enhances the relief rebound that is produced. Third, these effects speed u p the conditioning of avoidance-escape cues to the positive incentive motivational source, and therefore reduce the net negative incentive that is produced even before the coping response is made. Is this analysis compatible with Weiss’s idea that no ulcers can develop in the absence of a coping response? It is compatible with a weaker statement : that coping responses can enhance or suppress ulceration, but that any mechanism that produces negative incentive in the rebound mechanism creates a predisposition to ulcerate. A deeper analysis of the way in which positive and negative incentive actually regulate muscular contraction might refine this view at a later time.
XVIII. Orienting Reaction
We will show below that some properties of this reaction can formally be represented within the networks that are already at our disposal. We will invoke psychophysical examples to illustrate the formal meaning of the mechanisms, but do not presume that they are given a complete physiological explanation. Consider Fig. 18. I n Fig. 18, different paths P , arc differentially excited by different peripheral events, e.g., retinal loci, positions on the skin, auditory inputs. Suppose that U,z can fire only if orienting arousal combines with a signal from U , I .Let the axon collaterals from U,1 to 3n have relative strengths that determine a final orienting position for the muscles that they control. Different P, paths will determine different orienting positions by having dif-
312
STEP H E N GRO S SBERG
O R IENTING AROUSAL
FIG. 18. Orienting arousal activates position codes for motor control.
ferent patterns of axon strength. Let a recurrent normalizer interconnect the { Uil) populations. This establishes a “position code” in the field { Uil] of populations by normalizing the total activity of the field (e.g., normalizing the effect of variations in the total light energy hitting the retina), and letting the Uiz -+ 3?l axon strengths determine the terminal muscle positions. For example, if two Uil are simultaneously and equally active, then a position will be determined that lies between the positions determined by each Uill separately. As the relative activity of one Uil increases, the terminal position will approach the position controlled by this Uil alone. I n the case of vision, for example, if the arousal level of the field is tuned so that only one population can reverberate in STM, then only one retinal light source can attract the eyes. If arousal permits several populations simultaneously to reverberate, then weighted averages of the retinal positions can attract the eyes. Withdrawing orienting arousal prevents the release of signals from any Uiz. We assume that this arousal can be inhibited by activity from competing arousal sources, such as drive representations. Thus, before training, a loud noise in the direction of a subject’s left side can elicit an orienting reaction toward the left. Suppose, however, that the noise is used in learning trials as a discriminative cue for rapid lever pressing for food. Then on testing trials, the noise can differentially excite the ah+ representation, which can inhibit the source of orienting arousal via, say, the sensory-drive heterarchy. The source of orienting arousal is triggered by unexpected events. Minimality bids that we identify this arousal source with the arousal source, also triggered by unexpected events, that overcomes blocking and triggers enhancement or rebound in the various drive representations. A plausible candidate for this arousal source is the reticular formation.
ATTENTION,
REINFOKCl<MENT, AND DISCRIMINATION
313
There exist variations on this anatomical theme, such as an orienting arousal source supplying shunting excitation that permits the cells which carry the position code to fire. Such a n arousal source can also act at the synaptic knobs, or to inhibit tonic presynaptic inhibition of these knobs (disinhibition). In all the above anatomies, excitation and disinhibition can have similar functional effects. Disinhibition has the disadvantage of requiring an extra processing step, but it has the advantage that it permits sustained activity of cells, which prevents them from undergoing a chemical degradation due to disuse.
XIX. A learned Expectation Mechanism
An expectation mechanism is described below to help fix ideas in the above discussion. We wish to prevent orienting arousal if an expected event occurs, and to permit it if an unexpected event occurs. The first part of the construction synthesizes a network which can learn to expect a given event subsequent to the occurrence of another event. Several variations of this construction appear in Grossberg (1972a). This construction will be supplemented herein to guarantee additional properties of the expectation mechanism. The output cells U of the network will fire only if the learned expected event occurs. The construction in Grossberg (1972a) is reviewed below for completeness. T h e learned input pattern (or class of patterns) which can fire the cell (or cells) U is controlled by presetting cells P.T h e cells P send axons to the filtering mechanism (e.g., inhibitory interneurons and dendrites) that processes inputs to U. Each P cell can learn a particular pattern that will bias U s filter when P is active. For example, consider an animal 0 that learns to lever press for food. O n a testing trial, (7 “expects” food when it lever presses in response to hunger. We suppose that lever press cues also preset consummatory controls which can be released by expected sensory cues of the food reward. Similarly, suppose that one goes to the refrigerator expecting to find orange juice, which one loves, in a transparent container, but instead one finds tomato juice, to which one is indifferent. T h e same motor sequence of reaching, pouring, and drinking suffices for imbibing either the orange juice or the tomato juice. T h e orange fluid releases this sequence, but the red fluid does not; indeed, the red fluid can release a frustrative rebound. T h e consummatory controls have been preset by the expectation of a n orange fluid. How do the P cells learn the patterns on training trials that will bias the U-cell filter on testing trials? Consider the anatomy of Fig. 19, in which interacting signals combine additively. I n Fig. 19, the cells V I = id3):
314
STEPHEN GROSSBERG
"3
FIG. 19. Subtractive preset mechanism.
j = 1, 2, . . . , n f are P cells. These cells sample patterns playing on the cells V z = (ui: i = 1, 2, . . . , m ) when they are active. The cells V3 = ( u ; + ~ i: = 1, 2, 3, . . . , m ) receive the test patterns that will fire the cell U if they are expected by P. O n training trials, VQ-+ Vz signals reproduce these test patterns at Vz, where they can be sampled by V1. Signals in V3-f V4 axons, where V4 = (ui+zm: i = 1, 2 , . . . , m } ,also reproduce the test patterns at V4. O n testing trials, activity in a P cell generates a pattern 0 of activity in VZ,which is transferred to Vq as inhibitory signals by V2-) V4
ATTENTION, KEINFORCI<MI:NT, A N D DISCRIMINATION
315
axons. The test pattern 8 at V3 is also sent along to V4, where it can be compared with 6'. Simultaneously, 8 is sent to V z to be sampled by V1. At V Z ,8 is transferred to V4 as inhibitory signals, just as the V1+ Vz presetting signals are. The inhibitory V3+ V4 signals are chosen weaker than the excitatory V3 + V4 signals, so that the test pattern appears a t Vq with a net excitatory strength. The inhibitory preset pattern 6' and the excitatory test pattern 8 are thereupon compared at V4. The above transformations can be defined in greater detail as follows. Let the strength of the excitatory v,+,,+ u , + ~signal ~ be 8,Z, and of the inhibitory v,+, + v z + u,+?, signal be 8,vZ, 0 < ? < 1. Then the net signal to v,+z, from v,+, is 8,(l - v)Z, which is nonnegative. Let the v ( 3 ) + v,+, -+ v1+zm inhibitory signal from the jth preset cell be -0,K. If only v(1) in V1 is active, the total signal to v,+2, is e,(l - v)Z - 0,K. Under these circumstances, v,+Zm will fire only if
This constraint shows that all cells in Vq can fire only if all the relative pattern activities in 8 are not too much smaller than the relative pattern activities in 6'. Since Z k 0 k = & 8 k = 1, simultaneous firing in all channels is possible only if (1 - v)Z > K. Thus the total activities of 8 1 and of V Bmust be carefully regulated. Inequalities (2) do not suffice to prevent firing of a discriminative cell further downstream to patterns some of whose 8i are much larger than 0 , (Grossberg, 1970, 1972a). T o prevent this, the output signal from each v i + ’ ~in~ V4 excites both an excitatory and an inhibitory pathway. The inhibitory pathway (which can, in principle, be just a high threshold inhibitory ionic channel in the same axon pathway as the excitatory channel) overcomes the excitatory pathway if the signal from vi+zm is too large. When this happens, the net output from vi+Zrn to V sis negative, so that not all channels are simultaneously excitatory. Thus the net signal from u;+z, to V j is derived from two successive inhibitory mechanisms. I t is positive at Uz+Zm only if the relative pattern activity 8, is not too much smaller than the relative pattern activity 6'i. This positive activity is inhibited, however, if 8, is too much larger than 6'i. All channels in V4 contribute a positive signal to Vs only if the pattern 8 is close, in every component, to the pattern 8.The signal threshold of Vb is adjusted once and for all so that Vg will fire only if it receives (nearly) simultaneous positive signals from all V4 channels. Hence the cell U = VE,fires only in response to the expected pattern. Grossberg (1972a) shows that this anatomy has formal properties that are reminiscent of cerebellar anatomy, and thereby illustrates the anatomical pausibility of this expectation mechanism. The same principles have been used to synthesize a class of networks with a suggestive retinal analog ( Grossberg, 1972a). These networks are
316
STEPHEN GROS SBERG
capable of discriminating the relative figure-to-ground of spatial patterns (i.e., their Oi’s) but do not have a learnable expectation mechanism. Here also two successive inhibitory mechanisms are needed. If the receptor cells of this network are interpreted as light receptors, then the first inhibitory layer is reminiscent of retinal horizontal cells. Speaking functionally, this layer produces a form of light adaptation; cf. unicellular recordings in the mudpuppy retina (Werblin, 1971). The second layer is reminiscent of amacrine cells. The output cells (cf. ganglion cells) are then capable of hue constancy (including a lightness scale) , brightness constancy, velocity detection, etc., depending on which receptors are hooked into the network, and on how the anatomical connection coefficients are chosen. The expectation mechanisms defined above has two deficiencies : ( 1) I t does not automatically regulate the total activities of V , and V,; and ( 2 ) if no presetting cell in P is active, then every pattern presented to V , can fire V,, since no net inhibitory signal is produced at V4. The first deficiency can be overcome by introducing recurrent on-center off-surround anatomies with shunting interactions into V , and V , Section IV indicates the need for such mechanisms within sensory processors, so that their use here does not impose a new constraint. T h e second deficiency can be overcome by assuming that uniformly distributed tonic inhibitory signals are somehow generated from V , to I/, in the absence of presetting signals, and that the onset of presetting signals supplants the tonic inhibition with learned patterns of V , + V , inhibition. A simple way to do this is to assume that tonically active cells exist in V , and send uniformly distributed inputs to V , ; V,, in turn, generates inhibitory signals to V , that prevent inputs to V 3 from firing V,. When a presetting cell in V , becomes sufficiently active, it suppresses the activity in the tonic cells via the recurrent off-surround in V1, and substitutes its own patterned signals to V,. Tonically active cells that are suppressed by the onset of phasic afferents are known to exist in various neural structures; in the frog retina, for example, there are dimming cells whose tonic activity in the dark is suppressed by light (Chung et al., 1970). Note also that the distribution of tonic inputs to V , can be uniform even if no tonic cell is connected to all cells in V,; only the distribution of activity across all tonically active cells needs to be uniform, and this distribution can be suppressed uniformly by widely dispersed off-surround signals within V,.
XX. Regulation of Orienting Arousal
The output cells U fire only if their expected event occurs. If any unexpected event occurs, we want it to generate orienting arousal. I t seems very
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
317
unlikely that a brain contains internal models of the infinitely many events that are unexpected at any time, and that it generates orienting arousal whenever there is a match between one of these events and its internal model. By contrast, given the above construction, it is easy to devise a network that inhibits orienting arousal only if the expected event occurs. Thus, we assume that every event which is processed by the network’s sensory mechanisms can, in principle, activate orienting arousal using as a source the neural activity which it generates as it is processed. The output from the expectation mechanism can, however, inhibit orienting arousal (cf. Sokolov, 1960) (see Fig. 20). I n Fig. 20, the output from the cells U bifurcates. One channel inhibits orienting arousal and the other channel samples the drive representations in &. For example, suppose that the expected event is a loud noise to the left of the network, and that the noise has been trained as a discriminative cue for lever pressing. When the noise occurs, it generates activity that can drive the orienting reaction. This activity is, however, inhibited by the output from U. The U output also generates positive + feedback that elicits the lever press. T h e orienting reaction can be inhibited by this mechanism even if U controls no other motor reaction. The construction can be modified to change this conclusion. If the orienting arousal channel is included in the on-center off-surround anatomy of the
s+
s
AROUSAL
FIG. 20. Inhibition of orienting arousal by expected events.
318
S T E P H E N GROSSBERG
sensory-drive heterarchy, then the orienting arousal can occur unless it is supplanted by strong competing S ---f -+ S feedback in a specific drive channel. The relationship between presetting inputs and test inputs will be more completely studied in another place. I n particular, one must note that the events which excite P cells and those which excite V z cells need differ only in their onset times; the P events occur earlier than the V , events. Thus, V , events gain control over P cells as new events intervene. This shift in the spatial locus of an event’s internal representation can be subsumed under the study of sequential short-term memory buffers (Atkinson and Schiffrin, 1968). I t is schematically represented in Fig. 21. Given such a shift in representational locus, one can see how this network becomes habituated to a repetitively presented event. As the event is repeated, it serves as a source of P-cell activity in its “past” mode, and as a source of test inputs in its “present” mode. The event samples itself, in short. As the event is repeated, it samples itself repeatedly via P + V 2 axons which build up the strength of the expectation. As the event becomes more expected, the output from U increases and progressively inhibits orienting arousal, but does not prevent conditioned responses from occurring. For example, young foxes quickly habituate orienting reactions to the sound of mouse squeaking, but once they have eaten a mouse, the squeaks become conditioned stimuli and the orienting reactions do not readily habituate (Biryukov, 1958; Lynn, 1966). One can also see how the network can become habituated to a learned set J REPRESENTATION
PRESENT
PAST
FIG. 2 1. Habituation to repeated event as its past-representation samples its present-representation in the expectation mechanism.
ATTENTION, REINFORCEMENT,
AND DISCRIMINATION
319
of events via conditioning of the P + V , synapses of the P representations corresponding to this set. Indeed, if the sensory filter is capable of grouping peripheral events in classes that fall along unconditioned or conditioned generalization gradients, then these gradients will be transferred to the P cells via the sequential STM buffer. XXI. Hippocampal Feedback, Conditioning, and Dendritic Spines
This section provides a way to implement three formal requirements in the network using a common mechanism. The mechanism has a suggestive anatomical analog in terms of hippocampus, septum, hypothalamus, reticular formation, and neocortex. I n this analog, the hippocampus receives input from neocortex (in uiuo, via the entorhinal cortex) and septum (Raisman et al., 1966). T h e mock-hippocainpal output trifurcates and eventually feeds back to septum as signals conditionable a t a + Q. synapses, and to neocortex as nonspecific attentional or motor feedback, possibly via the anterior thalamic nuclei (Raisman et al., 1966). T h c mock-hypothalamus prepares drive inputs to this system, and the reticular formation provides nonspecific arousal, which can be triggered by specific events, and which is filtered through the sensory and drive representations to enhance or rebound their activity. The three formal requirements are these: 1. Consider Fig. 22. I n the figure, S1 is conditioned to a],but all Si, i Z 1, are irrelevant cues; i.e., thcy project equally to a1 and a,. Suppose that many of these irrelevant cues are active when is active. Then the az synapses will become progressively stronger and eventually SI will approach irrelevancy also. T h a t is, the act of performing in response to a relevant event can countercondition the event simply because irrelevant events exist. Part of this difficulty can be overcome by (3 + attentional feedback, which tends to quench irrelevant cues. This does not, however, prevent counterconditioning of the a1 4 S1 channel by irrelevant cues that are active before the feedback occurs, however, just as a CS that is presented
s1
s1+
s
a, FIG. 22. Coiinterconditioning by irrelevant cues.
320
S T E P H E N GROSSBERG
before a UCS can sample the UCS-controlled representations without being totally overshadowed by the UCS. T o prevent counterconditioning by irrelevant cues, the uniform part of the total input to a dipole’s channels must be inhibited before substantial + @ sampling of these channels occurs. This can occur only at a stage after the S --+ a synaptic knobs, because one only knows that the --f signals are uniform after they are emitted at their respective synapses. Also the resultant of this inhibition must feed back in a form that can be sampled by + a synaptic knobs. How can the bulk of the conditionable signal be due to @ -+Q. feedback, and not to the + @ signals that are in spatial contiguity with other + @ synapses? 2 . The existence of higher-order instrumental conditioning implies that S -+ sampling can occur both before and after the stage of drive rebounds; hence there exists a recurrent loop from sampling mechanism, to rebound mechanism, to choice-among-drives mechanism, and back to sampling mechanism. 3. What kind of feedback should be conditionable? Should the feedback be from the resultant of each dipole separately, or from the resultant of all competing drives? I n the latter case, conditioning is possible only with respect to the drive that supplies incentive motivation for regulating attention, motor performance, and the transfer of STM into LTM. We exhibit a system of the latter type for definiteness. Variations on the theme are then readily constructable. T o achieve ( 2 ) and (3), we use a mechanism as in Fig. 23. Note that the output of the sensory-drive heterarchy trifurcates: it is fed back to “neocortex” as attentional feedback and as motor arousal, and it is fed back to (1 septum” as conditionable signals. To achieve (l), we must somehow allow --+ @ signals to influence events further downstream without allowing these signals to be substantially conditioned to anything but sensory-drive heterarchical output. One way to dothis is suggested in Fig. 24. The S-+a signals reach “dendritic spines.” Here they produce local potentials that propagate to the cell body where they influence axonal firing. We assume that the resistances in spines are such that it is much harder for a signal to pass between spines than from a spine to the cell body. Alternatively, one can assume that the threshold for the post- to presynaptic signals that are needed to change transmitter levels in S + Q. synapses are too high for spine-to-spine interactions to overcome them. By contrast, heterarchical feedback from a, energized by nonspecific arousal (e.g., from reticular formation) causes a spike potential, or similar global potential change, throughout the dendritic column. This spike invades all the spines in its path and is sufficiently strong to induce transmitter level changes in active + @ channels. Thus a mechanism using
s
s
s
s
s
s
32 1
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
\
‘f
lAROUSAL
, , C O N D I T I O NA B L E S l G N A l- s
D
AROUSAL
FIG.23. Conditionable heterarchical feedback signals sampled by
S.
FIG.24. Heterarchical feedback causes global potential change that invades individual conditionable S channels.
dendritic spines and dendritic spike generators (or some formally analogous mechanism) can allow S + Q. signals to occur without major changes in S -+ a synaptic transmitter levels unless feedback invades the entire dendritic apparatus. XXII. Nervous Eating and Attentional Deficits Modulated by Arousal
Section I11 pointed out that the existence of higher-order instrumental conditioning implies the existence of feedback loops in the rebound mecha-
322
S T E P H E N GKOSSRERG
nism. Section I V showed that the signal function in a recurrent anatomy must be carefully chosen to avoid amplification of noise. Faster than linear growth of signals at small values of cell activity, followed by (approximately) linear growth at larger values, achieves this goal. Since saturation of signal response at very high activity levels is easily assured, it suffices to impose faster than linear growth of signals at small values of cell activity; by continuity, there will exist an approximately linear range between the faster-than-linear growth region and the slower-than-linear saturation region of signal response. The width of the linear region is an important parameter in determining the short-term memory characteristics of the system. Indeed, the slower-than-linear saturation region tends to create a uniform distribution of activity across cell populations, starting with any initial input pattern. Given such a nonlinear signal function, the Appendix shows that a n increase in arousal can potentiate the system’s response to phasic sensory inputs. Thus, a novel event can overcomc overshadowing, thereby freeing its S representation to send signals to 0.In a, these signals contribute to the sensory-plus-drive combinations that are struggling to gain control over a --+ S feedback via the sensory-drive heterarchy. Simultaneously, the arousal triggered a t a can cause a rebound and/or enhance the a+ S feedback from the dominant sensory-plus-drive Combination. Similarly, the novelty of an indifferent event can make it rewarding, if the network is engaged in approach behavior when the event occurs (Berlyne, 1969). In effect, enhancement by arousal produces an extra source of input to a. Using this new input source, the network can, in principle, generate 0 + S feedback in the absence of drive inputs, thereby yielding the following interesting possibilities: (1) Motor activity initiated by an internally grnerated arousal pulse in the absence of external sensory cues; e.g., by testing its drive states with an arousal pulse, the network can generate a hunger-related output, such as the statement “I want food,” if the hunger drive is dominant but no cues of food are present. (2) If the arousal level is high, it can elicit consummatory activity compatible with sensory cues or drives that are too low to otherwise initiate consummation; e.g., “nervous eating.” If, for example, there is a damming u p of motor activity in the absencc of a n appropriate goal, and this activity feeds into 0 as arousal, then potentiation by arousal can discharge the motor activity through the heterarchical channel that is dominant a t that time. (3) An unexpected event, even a frustrating one, can elicit transient motor activity via heterarchical feedback, even if the heterarchical Feedback is not related in a simple way to the unexpected event. These remarks illustrate a so-called “two-thirds rule.” Namely, a t least two channels from among 9, and are needed to elicit a -+ S feedback: D is the usual heterarchical constraint on -+ S firing; 9 a is illus-
s+
s,
+
ATTENTION,
REINFOHCI.:Ml
323
+
trated by the “ I want food” cxamplc; and $, Q. is illustrated by “nervous eating.” One can also imagine, in principle, thc perhaps pathological case in which intense $, 33 Q. activity allows more than one heterarchical channel to be active at a time, by driving the activity of more than one channel above quenching threshold ; cf. the remarks about quenching threshold relevant to determining the asymptotic “eye position” in Section X V I I I . T h e mixing of channel activities at the Q. cells has a n analog at $, cells. Here converge arousal inputs that are triggered by unexpected events, and which consequently tend to overcome overshadowing, as well as arousal inputs from Q. + $, channels that contribute motivational feedback, and can sustain overshadowing. LJnder pathological circumstances, either channel can become persistently overaroused. O n e possible consequence of overarousal is “seizurc activity” (cf. Grossberg, 1973). Another is the inability to pay attention (cf. Grossberg and Pepe, 1971). Either of the two arousal sources can cause such difficulties, in principle, but the overall “clinical syndrome” that the network would undergo could be quite different in the two situations. Q. + $, overarousal can cause, in addition, eniotional depression (Grossberg, 1972c), as well as pathological changes in the network’s “psychological sets” (Section VI). If, for example, one heterarchical Q. + S feedback channel became dominant, then it could bias all the network’s sensory processing in a direction that is compatible with the dominant drive, or more precisely, the dominant “psychological set.” Such an effect need not occur if the source of novelty-bound arousal is overaroused, since this arousal source is truly nonspecific. Previous work (Grossberg, 1972c) suggests than an analogy can be drawn between Q. + $, channels and midbrain channels influenced by catecholamine production. Thus, imbalances in catecholamine production might produce an overaroused syndrome (“simple schizophrenia”?) which is different from an overaroused syndrome produced by malfunction of the reticular formation, both in symptomatology and in proper treatment. SUBMITTED FOR PUBLICATION: March, 1973.
+ +
Appendix
CS, can become a learned source of relief, rather than of fear, if the CS, is sufficiently small compared to the shock level that follows CS, shock level that follows CS,. This happens if the increase in tonic input I that follows the unexpected change in shock level J is sufficiently great compared to the change in J . Moreover, for sufficiently small values of I and J , an increase in I can potentiate the fear produced by a fixed level of J , if the signal function is sigmoid.
+
324
STEPHEN GROSSRERG
Consider the following system for definiteness (see Grossberg, 1972c, for details). Xi
xz
= =
-a111 -ax2
+I +J +I
22
= P(Y - tl) - sf(xl(t - 7 ) ) Z l = P(r - z2) - 8 f ( x z ( t - 7))zz
i 3
=
-EX3
x4
= =
--EX4
il
li.5
-7X5
x 6 = -77x6
+ + { f ( x z ( t - 7))zz + JFf(Xl(t
7))21
K[X3(t
f
K[X?(t
U)
-
U)
Xq(t - U)] - xS(t - U)]
First, constant levels ZI and J 1 of tonic input and shock are switched on until after the potentials xi and transmitters zj ad,just to their new levels. Then new levels Iz and Jz are imposed. The potentials adjust much quicker than the transmitters. Hence a measure of the maximum response to the change in levels in computed by maintaining the transmitters at the steadystate levels determined by ZI and J1, and the potentials a t the new steadystate levels imposed by I2 and J z . Fear is produced if x5 > 0 > .%6; relief is produced if x g < 0 < xg. The functionf(w) computes the signal in response to the potential w . In Grossberg (1974), the functionf(w) = max (0, w - r) is used, where is a signal threshold. Fear is produced only if
where p = &?,and relief is produced only if the reveise inequality holds. For example, if f(w)= w,then fear is produced only if Jz/Ji
> Zz/(u
+
11)
where u = a/3/6. The steady-state fear response to constant Z and J is given by
where w = KJF~V-'-E-'. A sigmoid f yields potentiation of x g ( ~0 ) in response to an increase of Z, if I and J are sufficiently small. T o see this is the case that f" exists, we compute [ a x s ( oo ) ] / d I , and note that this function is positive if and only if the function
ATTENTION, REINFORCEMENT, AND DISCRIMINATION
325
is strictly monotone increasing in the desired region of Z and J values. If
f” exists, this condition becomes M w )
+ llf (w1
> 2df’(w)12
which is true for small values of w , since f(0) = f’(0) = 0
< f”(0).
REFERENCES Amsel, A. (1958). Psychol. Bull. 55, 102. Amsel, A. (1962). Psychol. R e v . 69, 306. Anderson, P., Gross, G. N., Lomo, T., and Sveen, 0. (1969). I n “The Interneuron” ( M . Brazier, ed.), p. 415. Univ. of California Press, Los Angeles. Atkinson, R. C., and Shiffrin, R. M. (1968). I n “The Psychology of Learning and Motivation,” (K. W. Spence and J. T. Spence, eds.), Vol. 2, p. 89. Academic Press, New York. Azrin, N. H., and Holz, W. C. (1966). I n “Operant Behavior” (W. K. Honig, ed.), p. 380. Appleton, New York. Berlyne, D. E. (1969). I n “Reinforcement and Behavior” (J. T . Tapp, ed.), p. 179. Academic Press, New York. Berlyne, D. E. (1970). I n “Attention: Contemporary Theory and Analysis” (D. E. Mostofsky, ed.), p. 25. Appleton, New York. Bersh, P. J., Notterman, J. M., and Schoenfeld, W. N. (1956). Air University, School of Aviation Medicine, U.S.A.F., Randolph AFB, Texas. Biryukov, D. A. (1958). The nature of orienting reactions. I n “The Orienting Reflex and Orienting-Investigating Activity” (I,. G. Voronin et al., eds.), Acad. Pedag. Sci., Moscow. Bitterman, M. E. (1965). Zn “Classical Conditioning” ( W . F. Prokasy, ed.), p. 1. Appleton, New York. B!oomfield, T. M. (1966). J . E x p . Anal. Behari. 9, 155. Bloomfield, T. M. (1969). I n “Animal Discrimination Learning” (R. M. Gilbert and N. S. Sutherland, eds.), p. 215. Academic Press, New York. Bower, G. H. (1966). Zn “Theories of Learning” (E. R. Hilgard and G. H. Bower, eds.) , Appleton, New York. Cant, B. R., and Bickford, R. G. ( 1967). Electroencephalogr. Clin. Neurophysiol. 23, 594. Chung, S.-H., Raymond, S. A,, and Lettvin, J. V. (1970). Brain Behav. Evol. 3, 72. Cohen, J. (1969). I n “Average Evoked Potentials” (E. Donchin and D. B. Lindsley, eds.), p. 143. Nat. Aeron. Space Adrnin., Washington, D.C. Goodrich, K. P. (1959). J . Exp. Psychol. 57, 57. Grastyan, E. (1959). I n “The Central Nervous System and Behavior” ( M . A. Brazier, ed.) Josiah Macy, Jr. Found., New York. Gray, J. A., and Smith, P. T . (1969). I n “Animal Discrimination Learning” (R. M. Gilbert and N. S. Sutherland, eds.), p. 243. Academic Press, New York. Grossberg, S. (1969a). J . Math. Psychol. 6, 209. Grossberg, S. (196913). M a t h . Biosci. 4, 201. Grossberg, S. (1970). J . Theor. Biol. 27, 291. Grossberg, S. (1971a). J . Theor. Biol. 33, 225. Grossberg, S. (1971b). Proc. h at. Acad. Sci. U S . 68, 828.
326
STEPHEN GROSSBERG
Grossberg, S. (1972a). Kybernetik 10, 49. Grossberg, S. (1972b). M a t h . Biosci. 15, 39. Grossberg, S. ( 1 9 7 2 ~ )M . a t h . Biosci. 15, 253. Grossberg, S. (1972d). I n “Delay and Functional Differential Equations and their Applications” ( K . Schmitt, ed.), p. 121. Academic Press, New York. Grossberg, S . (1973). S t u d . A p p l . M a t h . 52, 213. Grossberg, S. (1974). I n “Progress in Theoretical Biology” (F. M. Snell, ed.), p. 5 1. Academic Press, New York. Grossberg, S., and Pepe, J. (1971). 1.Statist. Phys. I, 319. Grossman, S. P. (1967). “A Textbook of Physiological Psychology.” Wiley, New York. Grusec, T. (1968). 1.Exp. Anal. Behav. 11, 239. Haggard, D. F. (1959). Psychol. R e c . 9, 11. Hanson, H. M. (1959). J . E x p . Psychol. 58, 321. Hoffman, H. S. (1969). I n “Animal Discrimination Learning” (R. M. Gilbert and N. S . Sutherland, eds.), p. 63. Academic Press, New York. Honig, W. K. (1962). 1.E x p . Psychol. 64, 239. Honig, W. K. (1969). I n “Animal Discrimination Learning” ( R . M. Gilbert and N. S. Sutherland, eds.), p. 35. Academic Press, New York. Honig, W. K. (1970). In “Attention: Contemporary Theory and Analysis” (D. I. Mostofsky, ed.), p. 193. Appleton, New York. Irwin, D. A,, Rebert, C. S . , McAdam, D. W., and Knott, J. R. (1966). Electroencephalogr. Clin. Neuroph ysiol. 2 1, 4 12. Jenkins, H. M., and Harrison, R. H. (1960). 1.E x p . Psychol. 59, 246. Jenkins, W. O., Pascal, G. R., and Walker, R. W., Jr. (1958). 1. E x p . Psychol. 56, 274. Kamin, L. J. (1968). I n “Miami Symposium on the Prediction of Behavior 1967: Aversive Stimulation” ( M . R. Jones, ed.), Univ. of Florida Press, p. 9. Coral Gables. Kamin, L. J. (1969). In “Punishment and Aversive Behavior” (B. A. Campbell and R. M. Church, eds.), p. 279. Appleton, New York. Kelleher, R. T. (1966). I n “Operant Behavior” (W. K. Honig, ed.), p. 160. Appleton, New York. Kernell, D. ( 1965a). A c t a Physiol. Scand. 65, 65. Kernell, D. ( 196513). A c t a Physiol. Scand. 65, 74. Kimble, G. A. (1961). “Conditioning and Learning.” Appleton, New York. Low, M. D., Borda, R. P., Frost, J. D., and Kellaway, P. (1966). Neurology 16, 771. Luria, A. R., and Homskaya, E. D. (1970). I n “Attention: Contemporary Theory and Analysis” (D. I. Mostofsky, ed.), p. 303. Appleton, New York. Lynn, R. (1966). “Attention, Arousal, and the Orientation Reaction.” Pergamon, Oxford. McAdam, D. W. (1969). Electroencephalogr. Clin. Neurophysiol. 26, 216. McAdam, D. W., Irwin, D. A,, Rebert, C. S., and Knott, J. R. (1966). Electroencephalogr. Clin. Neurophysiol. 21, 194. McAllister, W. R., and McAllister, D. E. (1971). I n “Aversive Conditioning and Learning” (F. R. Brush, e d . ) , p. 105. Academic Press, New York. Maier, S. F., Seligman, M. E. P., and Solomon, R. L. (1969). I n “Punishment and Aversive Behavior” (B. A. Campbell and R. M. Church, eds.), p. 299. Appleton, New York.
ATTENTION, REINFORC:lSMENT,
A N D DISCRIMINATION
327
Milner, B. (1958). I n “The Brain and Human Behavior” (H. C. Solomon, S. Cobb, and W. Penfield, eds.), p. 244. Williams & Wilkins, Baltimore, Maryland. Newman, F. L., and Baron, M. R. (1965). J . Comp. Physiol. Psychol. 60, 59. Newman, F. L., and Benefield, R. I,. (1968). J. C o m p . Physiol. Psychol. 66, 101. Olds, J. (1969). Amer. Psychol. 24, 114. Raisman, G., Cowman, W. M., and Powell, T. P. S. (1966). Brain 89, 83. Rall, W. (1955). 1. Cell. Comp. Physiol. 46, 413. Scheibel, M. E., and Scheibel, A. B. (1967). In “The Neurosciences: A Study Program” (G. C. Quarton, T. Melnechuk, and F. 0. Schmitt, eds.), p. 577. Rockefeller Univ. Press, New York. Seligman, M. E. P., and Johnston, J. C. (1973). I n “Contemporary Prospectives in Learning and Conditioning,” Scripta, Washington. Sharpless, S., and Jasper, H. (1956). Brain 79, 655. Sokolov, E. N. (1960). I n “The Central Nervous System and Behavior” (M. A. Brazier, ed. ) , Josiah Macy, Jr. Found., New York. Terrace, H. S. (1966). I n “Operant Behavior” (W. K. Honig, ed.), p. 271. Appleton, New York. Thompson, R. F. ( 1967). “Foundations of Physiological Psychology.” Harper, New York. Trabasso, T., and Bower, G . H. (1968). “Attention in Learning: Theory and Research.” Wiley, New York. Wagner, A. R. (1969a). I n “Punishment and Aversive Behavior” (B. A. Campbell and R. M. Church, eds.), p. 157. Appleton, New York. Wagner, A. R. (1969b). I n “Animal Discrimination Learning” ( R . M. Gilbert and N. S. Sutherland, eds.), p. 83. Academic Press, New York. Walter, W. G. Arch. Psychiat. Nerz enkr. 206, 309. Weiss, J. M. (1971a). J. Comp. Physiol. Psychol. 77, 1. Weiss, J. M. (1971b). J . Comp. Physiol. Psychol. 77, 14. Weiss, J. M. ( 1 9 7 1 ~ )J.. Comp. Physiol. Psychol. 77, 22. Werblin, F. S. (1971). 1.Neurophysiol. 34, 228.
This Page Intentionally Left Blank
MARIHUANA, LEARNING, AND MEMORY By Ernest 1. Abel
Research Institute on Alcoholism, Buffalo, N e w York
I. Introduction . 11. AnimalStudies A. Drug Effects on Performance B. Effects on Acquisition C. Post-Training Drug Administration D. State-Dependent Effect . 111. Human Studies . A. Retrieval . B. Acquisition IV. Summary and Further Considerations References .
.
.
.
. .
.
. . . . . . . . .
.
.
329 330 332 338 342 343 345 347 347 351 353
I. Introduction
A great deal of attention has recently been focused on the effects of marihuana' (Cannabis sativa L ) and its derivatives and homologs' on cognitive functions, particularly as regards the influence of these substances on learning and memory. Although there is general agreement among investigators that in man marihuana markedly impairs the acquisition rather than T h e terms cannabis, marihuana, and hashish will be used interchangeably without distinction. T h e active principles in cannabis are the cannabinoids of which trans-A'-tetrahydrocannabinol (A"-THC = A'-THC) and trans-AH-tetrahydrocannabinol (A*-THC = A"-THC) are presently considered to be the most potent (Mechoulam, 1970). Although some of these other cannabinoids, such as cannabidiol, cannabinol, and cannabigerol, are pharmacologically inactive themselves (Edery e t al., 1971 ), evidence has recently been accumulating that these otherwise inactive cannabinoids can affect the activity of AO-THC (Jones and Pertwee, 1972; Fernandes et al., 1973; Karniol and Carlini, 1973) so that on a A9-THC milligram per kilogram basis, crude cannabis extracts can be expected to be different from As-THC per se. I n addition to these naturally occurring substances, two homologs derived from d,l-synthetic tetrahydrocannabinol have been widely used in psychopharmacological investigations. These homologs are the n-hexyl derivative pyrahexyl (also known as synhexyl), and the demethylheptyl derivative DMHP. 329
330
E R N E S T L. ABEL
the retrieval of information (e.g., Abel, 1971b; Darley et al., 1973a; Miller et al., 1972), studies with animals have not produced the same degree of consensus as to the nature of marihuana’s actions on these cognitive processes, The purpose of this paper is both to review and to examine these studies in animals and man with regard to some of the experimental problems which may have contributed to the inconsistencies in the literature. Among the various models of memory that are currently available, one of the most elegant is that proposed by Atkinson and Shiffrin (1968) which basically posits a relatively long-term store and a relatively limited capacity short-term store. Newly acquired information first passes to a labile shortterm store and is held there for a brief time. I t then either decays and is replaced by new information or else it is transferred to the long-term store, where it is permanently encoded in the neural tissue for subsequent recall at a later time. I n addition to this two component system, some investigators such as McGaugh (1968) have suggested that there may be other intermediate mechanisms interposed between the short-term and long-term stores. I n light of such a multiphasic concept of memory, it is possible that different neurophysiological and biochemical mechanisms are operative at different phases and hence it is likely that a drug may affect some processes and not others. With this concept or one similar to it in mind, a number of investigators have employed various methods to try to demonstrate and tease apart the effects of marihuana on short- and long-term memory function. The basic consideration in these studies has been the attempt not only to demonstrate the presence or the absence of drug-induced impairment on memory, but also to separate drug effects upon storage and retrieval processes where such differences are found. II. Animal Studies
In general, studies of the effects of drugs on animal learning and/or memory involve the training of animals in such tasks as conditioned avoidance or maze learning and the measurement of drug-induced changes in performance. One of the main problems inherent in such studies, however, is that these cognitive processes cannot be observed directly but must be inferred from changes in the subject’s behavior. Moreover, there is the additional problem of distinguishing between the concepts of learning and memory per se since each encompasses the other; for example, the acquisition of information requires registration, retention, and retrieval if there is to be a change in behavior resulting from experience. Evidence that learning has occurred at all can only be inferred from some change in behavior that is not due to maturation or some temporal
MARIHUANA, I.EARNING, AND MEMORY
331
state such as fatigue. Drugs can either have no effect or can reduce or increase the number of trials needed by an animal to reach some arbitrary criterion, e.g., 8 correct trials out of 10. Many drugs are capable of interfering with acquisition solely on the basis of their depressant effects on motor activity. An example of drug-induced facilitation of learning is the enhancement of maze and avoidance learning in various animals following injection of analeptic drugs, such as strychnine, picrotoxin, and pentylenetetrazol prior to training (McGaugh and Petrinovich, 1959; Kelemen and Bovet, 1961). A second point of attack in studying learning/memory processes in animals is to administer drugs after the training session. The animals are then retested, and the effects of the drugs on information-consolidation processes are inferred from the animal’s retest behavior. The advantages of such a paradigm are that the animals are neither trained nor tested under the influence of the drug so that variables such as attention, motivation, or locomotor activity are not relevant, and the effects of the drugs on memory storage can then be studied independent of original performance. The results of experiments of this nature have demonstrated that consolidation, i.e., the formation of a permanent memory trace following a given experience, is a rather labile process in which analeptic drugs such as strychnine enhance consolidation (McGaugh et al., 1962) whereas barbiturate drugs such as pentobarbital appear to impair such processes (Garg and Holland, 1969). However, because of the time-dependent nature of the consolidation period, the implementation of this design is not feasible in the case of drugs that have a slow onset of action. Thus, marihuana cannot be used to study this process in animals unless it is administered intravenously (cf. Abel et al., 1974). The third phase of memory is retrieval of information that has already been learned and stored. Once this has happened, there is no reason to believe that information is subsequently forgotten simply because the response does not occur under a given set of circumstances. Other possibilities are that the retrieval or search process by which the memory trace is to be located becomes impaired because of a drug-induced effect on the “trace scanning” mechanism. This may occur because the drug directly interferes with the “search” process by altering neural activity, or because the cues which might normally aid in the memory search become unrecognizable due to state-dependent learning. For example, Overton ( 1964) has demonstrated that animals that are trained under the influence of barbiturate drugs, such as pentobarbital, will show evidence of such learning only if they are retested under the drug. This dissociation of performance between drug and nondrug conditions has been attributed to drug-induced changes in neural firing patterns associated with the storage of learned material. In order for this mate-
332
ERNEST L. ABEL
rial to be recalled, it is contended that the drug-induced changes in neural firing patterns must be reconstituted (John, 1967). Having now reviewed some of the basic designs and considerations that must enter into any discussion of drug-induced changes in learning/’memory, the following survey will examine the literature with respect to marihuana in this kind of framework. However, since acquisition also involves retrieval whereas retrieval can be studied independently of acquisition, we will first consider drug effects on retrieval processes. Thus, if it can be shown that retrieval is not affected by marihuana, it would not be unreasonable to assume that any effects of marihuana on learning and memory are probably due to interference with acquisitional or consolidation processes. I n this regard, the following material is organized in terms of marihuana’s effects on behavior that is motivated by aversive conditions, e.g., shock, followed by its effects on appetitively motivated conditions, e.g., food reward. These experiments are summarized in Table I . I t is assumed that the reader is already familiar with the basic paradigms of shock avoidance, conditioned emotional response and maze learning, and hence, detailed descriptions of procedures have been omitted.
A. DRUGEFFECTSO N PERFORMANCE Assessing the effects of marihuana on performance in the context of learning and memory is exceedingly problematical since, for the most part, many of the experiments discussed below were not designed to examine such processes. Criticisms in methodology vis-A-vis learning and memory are thus often not pertinent to the original intentions of many of the investigators cited below. Nevertheless, since these experiments are somewhat similar to those that have been designed for this purpose, they have been included in the present discussion. I n the conditioned shock avoidance, a stimulus such as a tone or light (the conditioned stimulus, CS) precedes the onset of shock (the unconditioned stimulus, UCS). At first the animal responds only to the shock by jumping a barrier to a safe compartment of the apparatus or by pressing a bar, which terminates shock (escape). After a number of such paired presentations, the animal eventually avoids being shocked by making the appropriate response during the warning signal. I n one of the early studies of the effects of AS-THC (1-10 mglkg, i.p.) on previously acquired shock avoidance behavior, Grunfeld and Edery (1969) reported a dose-dependent suppression of performance in rats. Although the animals were able to make a n escape response to the UCS, the effect of the drug was still attributed to its cataleptic action rather than to impairment of memory retrieval.
333
MARIHUANA, LEAKNING, AND MEMORY
TABLE I EFFECTS OF MARIHUANA ON LEARNING AND MEMORY I N ANIMALS
Taska
Testb
Dosec (mg/kg)
ReRoutec Species sultsd
C.A.
P
(1-10) As-THC
i.p.
Rat
4
C.A.
P
(50) A’-THC
i.p.
Rat
1
C.A. C.A. C.A. S.A.
P P P P
(10) DMHP (30) A8-THC (1-20) A’-THC (16) As-THC
i.p. i.p. i.p. i.p.
Rat Rat Rat Rat
4 1
S.A.
P
(1) As-THC
i.p.
Monkey
S.A.
P
(8) As-THC (c)
i.p.
Rat
S.A.
P
(4) Ag-THC
i.p.
Rat
T
S.A.
P
(4-64) As-THC
i.p.
Monkey
T
-
1 1
References Grunfeld and Edery (1969) Orsingher and Fulginiti (1970) Delini-Stula (1973) Delini-Stula (1973) Pradhan ~t al. (1972) Barry and Kubena (1970) Barry and Kubena (1970) Barry and Kubena (1971) Sodetz (1972)
1 Scheckel et al. (1968)
I
1
i.p. i.p. i.p.
Rat Rat Rat
-
1
Webster et al. (1971) Herring (1972) Gonzalez et al. (1972)
i.p.
Rat
1
Gonzalez et al. (1972)
P P
(0.75-9.0) As-THC (4) Ag-THC (10) Cannabis extract (10) Cannabis extract (c) (2.5) AO-THC (0.25-4) As-THC
i.p. i.p.
Rat Monkey
1 -
Carlini et al. (1970) Scheckel et al. (1968)
M.t.S.
P
(0.5-2.0) A’-THC
Oral
Monkey
M.t.S. M.t.S. M.t.S. C.A.
P
P P Pt
(1.0) Ag-THC (1.0) Ag-THC (c) (4.0) AO-THC (c) (8) A’-THC
Oral Monkey Oral Monkey Oral Monkey i.p. Rat
C.A. Maze
Pt pt
i.p. i.p.
Mouse Rat
-
c..4.
A
(20-40) A’-THC (10) Cannabis extract (3.2) Pyrahexyl
i.p.
Gerbil
-
C.A.
A
i.p.
Rat
-
C.A.
A
i.p.
Rat
1
C.A.
A
(10) Cannabis extract (10) Cannabis extract (7.5) As-THC
i.p.
Rat
1
S.A. S.A. C.E.R.
P P P
C.E.R.
P
Maze M.t.S.
1
1 1 1 -
-
Zirnmerrnan et al. (1971) Ferraro (1972) Ferraro (1972) Ferraro (1972) Barry and Kubena (1971) Goldberg et al. (1973) Carlini and Kranier (1965) Walters and Abel (1970) Orsingher and Fulginiti (1970) Orsingher and Fulginiti (1970) Henriksson and Jarbe (1971) (Continued)
334
ERNEST L. ABEL
TABLE I (Continued) ~~
~~~
Task"
Dose" (mg/kd
Testb
ReRoutee Species sultsd
References
C.A.
A
(15) A*-THC
i.p.
Rat
C.A.
A
(4) AS-THC
i.p.
Rat
Henriksson and Jarbe (1971) Herring (1972)
C.A.
A
(1.25-80) A9-THC
i.p.
Mouse
Goldberg et al. (1973)
C.E.R.
A
i.p.
Rat
Gonzalez et al. (1973)
C.E.R.
A
i.p.
Rat
Gonzalez et al. (1972)
P.A. P.A. P.A. P.A.
A A A A
(10) Cannabis extract (10) Cannabis extract (c) (5,15) As-THC (10) DMPH (10) A*-THC (2-8) AS-THC
i.p. 1.p. i.p. i.p.
Rat Rat Rat Mouse
Water escape
A
(10-20) As-THC
i.p.
Rat
Water escape
A
(5) As-THC
i.p.
Rat
C.E.R. C.A. (extinction) Maze
A A
i.p. i.p.
Rat Rat
i.p.
Rat
Maze
A
i.p.
Rat
Maze
A
i.p.
Rat
Latent learn Habituation
A A
(15) Pyrahexyl (62.5,125) Hashish resin (10) Cannabis extract (10) Cannabis extract (10) Cannabis extract (3) As-THC (1, 3.2) As-THC
Miller et al. (1973) Delini-Stula (1973) Delini-Dtula (1973) Glick and Milloy (1972) Jarbe and Henriksson (1973) Jarbe and Henriksson (1973) Abel (1969) Jaffe and Baum (1971)
i.p. i.p.
Rat Mouse
A
Carlini and Kramer (1965) Orsingher and Fulginiti (1970) Gonazlez and Carlini (1971) Miller and Drew (1973) Brown (1971)
0 C.A., conditioned avoidance: S.A., Sidman avoidance: C.E.R., conditioned emotional response: M.t.S., matching-to-sample: P.A., passive avoidance. * P, performance: pt, post-trial: A, acquisition. CAll experiments are acute except for those followed by (c), which are chronic, A8-THC and As-THC, respectively, As-trans- and As-trans-tetrahydrocannabinol: DMP1-I. dimethylheptyl derivative. d t, facilitation: impairment: -, no change.
r,
A similar suppression of conditioned avoidance behavior following administration of a cannabis extract (50 mg/kg, i.p.) in previously trained rats was reported by Orsingher and Fulginiti ( 1970). Interestingly, the effect was greater in animals trained with light rather than noise as the CS. However, since no information was reported concerning the effects of the drug
MARIHUANA,
LEARNING, AND MEMORY
335
on intertrial activity or escape responding, it is not possible to exclude the possibility that the observed effects were due to the depressant actions of the drug on locomotor activity. Likewise, Domino (1971) observed a doserelated suppression of avoidance behavior in rats and, although he reported that escape responding was only minimally affected, no intertrial data were given so that here too an effect on locomotor activity cannot be excluded (cf. also Domino et al., 1971). Impairment of performance of conditioned avoidance has also been observed in rats by Delini-Stula (1973) following administration of D M H P and A’THC (10 and 30 mg/kg, i.p.). However, a depression of spontaneous motor activity, muscle coordination, and inhibition of unconditioned escape responding were also observed at these doses. In contrast to these reports, no disruption of bar press avoidance was reported by Pradhan et al. (1972) although these investigators did note a significant effect of the drug on response latencies. I n the continuous Sidman avoidance procedure no exteroceptive warning signal (CS) precedes shock ( U C S ) . Instead, shock is scheduled every 20 or 30 seconds by a timing device which is reset by each lever rcsponse. Using an avoidance procedure of this type, Barry and Kubena (1970) reported an increase in the number of shocks received by rats and monkeys following injection of AS-THC (16 and 1 mg/kg, i.p., respectively) that were previously trained in the task. Since no measurements were made of the drug’s effect on locomotor activity, general motor depression cannot be eliminated as the most likely explanation for these results. I n a subsequent study, Barry and Kubena (1971) reported an iniprovement (less shocks received) in rats previously trained in this task following repeated injections of AS-THC (8 mg/kg, i.p.). The effect was not evident on the first few days of drug treatment, however, and significance was determined by analyzing the trend for the 8 days of drug treatment which showed a decrease in shocks received as testing continued. The behavior of control subjects receiving the drug after 8 prior days of training was not affected, but when chronically treated drug animals were treated on days 11 and 12 without drug, there was a large increase in the nuniber of shocks received. The authors state that the average number of lever presses per group did not appear to be different, but do not report this data nor indicate to what tests the data were subjected. Since several days of drug treatment were required for this effect to occur, it is possible that the result was due to tolerance to the depressanteffects of the drug on motor activity, which resulted in an “unmasking” of excitatory activity (cf. Carlini et al., 1972) and improved avoidance on the basis of this effect. In an experiment reported by Sodetz (1972) an improvement in the avoidance behavior in some rats but a decrement in the performance of others (number of shocks received) was observed following AS-THC (4 mg/kg,
336
E R N E S T L. ABEL
i.p.). Interestingly, three out of five of his animals died during the experiment. These results, including the mortality rate, are similar to those reported by Scheckel et al. (1968) for monkeys, in which it was found that i.p. doses of A!’-THC (4-8 mg/kg) decreased response rate whereas doses of 16, 32, and 64 mg/kg increased response rate relative to control rates (nine out of eleven subjects subsequently died). Animals receiving the low doses appeared to be depressed whereas those receiving high doses appeared to be stimulated. Rather than attributing the effects on avoidance behavior to drug-induced impairment of memory, Scheckel and co-workers imply that these results are more likely due to an effect on motor activity. I n a variant of this procedure, called discriminated Sidman avoidance, an exteroceptive signal precedes the shock stimulus, but a response at any time resets the timing device, so that it is possible to avoid receiving shock by responding either during the interval before the onset of the warning signal or during the warning signal itself. Using various doses of a9-THC (0.75-9.0 mg/kg, i.p.), Webster et al. (1971) found that there was a significant increase in the number of shocks received in drug-treated animals previously trained to a high criterion on this task. However, it is not apparent whether performance was disrupted because of some effect of the drug on timing behavior, discrimination, or memory. Using a procedure similar to this, however, a member of this team (Herring, 1972) failed to replicate this effect on performance in rats given h9--THC ( 4 mg/kg, Lp.). Another type of procedure employing aversive stimuli to motivate behavior is the conditioned emotional response ( C E R ) paradigm. Although there are a number of variations of the procedure, the basic method involves first training an animal to perform some response, such as bar pressing for food or water reinforcement, until a stable rate of responding is attained. The animal is then subjected to a number of tone-shock pairings. Next, operant behavior is reinstated and when the animal is once again responding at constant rate, the tone previously paired with shock is introduced and the number of responses emitted during presentation of this tone is compared with the number of responses emitted prior to the stimulus. Using the CER paradigm, Gonzalez and co-workers (1972) injected animals with cannabis extract (10 mg/kg, i.p.) 24 hours after exposure to the tone-shock presentations. A second group was given 20 injections of drug or placebo prior to the shock treatment but were not given drug during exposure to the shock. Twenty-four hours later they were again injected with drug or placebo as before. The investigators found that in both the acute and the chronic conditions, the latency to approach the tube during testing was significantly shorter for drug-treated animals than for control subjects. However, since no data regarding drug effects on general activity
M A R I H U A N A , LEARNING, A N D M E M O R Y
337
were presented, it is not possible to rule out some motor related effects as contributing to these results. Several studies of marihuana and its homologs on performance of tasks motivated by positive reinforcement have also been reported. Carlini et al. (1970) administered AS-THC (i.p.) to female rats after they had received extensive training in negotiating a Lashley I11 maze. Doses of 2.5 mg/kg and higher were reported as disrupting performance. However, the investigators scored any failure to complete the maze within a 5-minute period as an error, thus, confounding the time measure with error rate. To evaluate this result in terms of the drug's effect on memory would thus not be appropriate since the data were scored in terms of drug's depressant effects on motor activity. Another appetitively motivated learning situation in which the cannabinoids have been studied is the delayed matching-to-sample task, in which monkeys are presented with a stimulus such as a colored light; then, after a delay, they are presented with a series of lights and must match the sample by touching the color previously presented to receive a reward. Using this basic procedure, Scheckel and co-workers (1968) found that with A'-THC ( 4 mg/kg, i.p.) one squirrel monkey would not perform in this task until a day after injection, and then even though it made more correct than incorrect responses, it failed to respond at all in the majority of trials. Siniilar effects were found after injections of 1 and 2 mg/kg. At 0.25 mg/kg, responding was not markedly affected, but at that dose the drug appeared to have little effect on performance. In an analogous experiment reported by Zimmerman et al. (1971), rhesus monkeys that had inhaled cigarettes containing AS-THC made significantly more errors than did control animals if the delay between the stimulus presentation and the test was 30 seconds. Performance was not significantly different from controls; however, at delays of zero and 5 seconds, performance was impaired. I n a second study in which the animals had to select the light that was different from the sample light, all oral doses of A"-THC (0.5-2.0 mg/kg) significantly irnpaired accuracy as before, with the longer interval between the sample and the test presentations. However, with the 2.0 mg/kg dose, accuracy was significantly impaired at the other intervals as well. Results similar to these have also been reported by Ferraro (1972) for chimpanzees given oral doses of 1.0 mg/kg Ag-THC. Although accuracy decreased for both drug and control animals as the interval between sample presentation and testing lengthened, the number of correct responses by drug-treated subjects was significantly less than for control subjects. I n a subsequent experiment (Ferraro, 1972) tolerance to this effect on perfor-
338
ERNEST L. ABEL
mance was not detected even after 21 consecutive daily administrations of the 1.0 mg/kg or 42 consecutive administrations of 4.0 mg/kg.
B. EFFECTSO N ACQUISITION Surprisingly, there have not been many reports of drug effects on the acquisition of behavior. Jarvik (1972) cites time, expense, and difficulty as some of the general reasons for the lack of experimentation in this area. Nevertheless, a certain amount of work has been published regarding the effects of cannabis and its derivatives on acquisition, and this literature will comprise this section of this review. Using the conditioned avoidance procedure to study learning, Walters and Abel (1970) failed to observe any effects on acquisition (trials to criterion) of pyrahexyl (3.2 mg/kg, i.p.) in gerbils, although the drug did significantly shorten the latency for these animals to jump a hurdle in response to the CS. Likewise, Orsingher and Fulginiti (1970) failed to observe any effect of a single dose of cannabis extract (10 mg/kg, i.p.) on acquisition of a conditioned avoidance response in rats. However, 23 daily injections up to, but not including training, markedly impaired acquisition. Although no mention of the physical condition of these animals was stated in the study, this latter result may have been due to the poor health of these animals following such chronic treatment (cf. Manning et al., 1971). A significant impairment in acquisition of conditioned avoidance behavior in rats has also been observed following administration of aS-TI-IC (15.0 mg/kg, i.p.) (Henriksson and Jarbe, 1971) . Although the investigators reported that escape behavior was not suppressed by either compound, no data were given as to the effects of these drugs on intertrial responding (a measure of nonspecific responding) . However, the investigators did note the presence of ataxia in some of the animals in both drug groups, suggesting that the effect on avoidance behavior may have resulted from the depressant effects on motor activity of these drugs. I n contrast to the previous experiments, Herring (1972) reported facilitated acquisition of a bar-press avoidance response in rats after administration of AS-THC (4 mg/kg, i.p.). Although Herring stated that there were no significant differences in the total number of bar presses, she did not indicate whether this analysis included the number of responses during shock. If so, this might mean that while the total number of bar presses in the two groups was similar, one group was responding prior to shock (avoidance) whereas the other was responding after the onset of shock (escape) . Goldberg et al. ( 1973) administered various doses of AS-THC (1.25-80.0 mg/kg, Lp.) to mice for five consecutive days but reported the effects of the drug on acquisition of shock avoidance behavior for the last training
MARIHUANA, LZARNING, A N D MEMORY
3 39
day only. The 5 mg/kg dose depressed both avoidance and intertrial responding. Doses of 10-20 mg/kg day had no significant effect on either behavior whereas animals which received 40-80 mg/kg made significantly more avoidance and intertrial responses than did controls. However, because of the drug-induced increase in intertrial responding, the investigators declined to attribute the changes in avoidance behavior to any drug-related effect learning or memory processes. A series of experiments by Gonzalez et al. (1972) examined the effect of cannabis extract on acquisition and retention using the CER paradigm. These investigators first trained rats to lick a spout in order to receive water. The animals were then injected (i.p.) with 10 mg/kg cannabis extract equivalent to approximately 2.8 mg/kg AQ-THCor with placebo and then subsequently subjected to five tone-shock pairings in the apparatus in which they had been trained to drink. Twenty-four hours later they were retested without shock or drug. The retesting was repeated 48, 72, 120, and 168 hours after the original shock treatment. Gonzalez and co-workers (1972) found the latency to approach the drinking tubes in the apparatus even in absence of the tone, was significantly shorter in the drug-treated animals than in placebo-treated subjects for the test periods. These findings indicate that not only the tone, but the box itself, had become associated with shock for control but not for drug-treated animals. During presentation of the tone, the previously drug-treated animals also drank more water during the 3-minute test period than did the controls. However, this latter measure was confounded with the latency measure since the tones were presented even if the animals were not drinking prior to the onset of the tone (if the animals were not drinking, the tones could not suppress drinking behavior). Consequently, only the approach data are relevant and these indicate that the cues associated with the box in which the shock was presented tended to suppress behavior to a much lesser extent in the drug-treated animals. In a second experiment, rats were exposed to the tone-shock pairings after they had received 20 preinjections of the drug or placebo. No significant effect of the drug on either latency to approach the tube or amount consumed was observed suggesting tolerance to the effect as a consequence of the prior series of chronic injections. Gonzalez and co-workers (1972) attributed these results to an affect of the drug on the processes underlying acquisition and retention of the CER, viz, reduction of fear associated with the box and the tone during the shock treatment. This interpretation, they feel, is supported by the significant decrease in defecation scores also observed in the drug-treated subjects ; a decrease in open-field defecation being accepted by many investigators as an index of fear in rats (cf. Henderson, 1970).
340
ERNEST L. AREL
I n passive avoidance, the subject must learn to refrain from making some response he would normally make (inhibition), typically that of entering some part of an apparatus. If inhibition does not occur, the animal receives punishment, e.g., electric shock, upon entry into the prohibited area. Using this kind of paradigm, Miller et al. (1973) found that AS-THC (5 and 15 mg/kg, Lp.) did not affect acquisition of this inhibitory response in rats. A similar lack of effect of DMHP and A‘-THC (10 and 30 mg/kg, i.p.) on passive avoidance in rats has also been reported by Delini-Stula (1973). Significant impairment of passive avoidance in female mice was observed, however, by Click and Milloy (1972) with A!’-THC (2-8 mg/kg, i.p.) when animals were retested as long as 4 weeks after a single learning trial. Using water instead of shock as the aversive motivation for learning, Jarbe and Henriksson (1973) trained rats to swim to one side of a water T-maze to achieve escape from the water. Animals injected (i.p.) with either AR-THC (1-20 mg/kg) or A”-THC (5 mg/kg) made significantly more errors (slower rate of acquisition than did control subjects). By the sixth day of training, however, most of the animals had reached the criterion of eight correct responses out of ten. Since the dependent variable in this experiment was the number of errors made by the animals rather than swimming speed, this result cannot be accounted for on the basis of a depression of locomotor activity. Some investigators have also studied the effects of marihuana administered during the extinction phase of an experiment (the repeated presentation of test conditions without presentation of the UCS). Since testing animals during extinction can also be thought of as another method for studying acquisition (learning that reinforcement is no longer associated with a particular response), these experiments have been dealt with as such. However, these studies themselves add little to the overall picture. For instance, in a study by Abel (1969) pyrahexyl (15 mg/kg, i.p.) decreased the latency for rats to contact a lever during extinction of a conditioned emotional response (presentation of the CS without the UCS) . However, no observations were made of the drug’s effect on locomotor activity and, therefore, it is possible that the result was due to some excitatory effect on locomotor activity which caused the animals to move about the Skinner box thereby causing them to come in contact with the lever. Jaffe and Baum (1971) studied the effect of hashish resin given to female rats during the extinction period (no shock) of an active avoidance task. After reaching a criterion of 10 consecutive avoidance responses, experimental animals were injected (i.p.) with either 62.5 mg/kg or 125 mg/kg of the hashish resin or with placebo. Both drug groups made significantly more responses before reaching the criterion (no responding) than did the placebo group. Although this result could be interpreted as a demonstration of re-
MAHIHUANA, LEARNING, A N D MEMORY
34 1
tarded learning, the authors give no indication of the effects of the drug on general activity, and, hence, it is also possible that these doses of hashish resin acted to increase general activity and in so doing contributed to the less efficient performance of the animals. T h e effects of marihuana have also been studied on maze-learning that is motivated by appetite reinforcement such as food and water. Again, however, no clear patterns in results are discernible. For example, although Carlini and Kramer (1965) reported that 10 mg/kg of cannabis extract (i.p.) improved learning to negotiate a Lashley I11 maze for food reward as indicated by fewer errors by drug-treated animals, a direct replication of this study by Orsingher and Fulginiti (1970) using the same strain of rats, a similar Lashley I11 maze, and the same amount of cannabis extract obtained from the same geographic area, found an opposite effect, viz, an increase in the number of errors during acquisition. In view of the similarity of experimental conditions, this discrepancy in findings is difficult to account for. Gonzalez and Carlini (1971) studied the effects of cannabis extract (10 mg/kg, i.p.) injected during extinction in rats previously trained in T- and Lashley I11 mazes. In neither situation did the drug affect the overall rate of extinction (number of errors). Following a suggestion by Abel (1971a) that the effects of marihuana on human memory might be incidental to its effects on attentional processes. Miller and Drew (1973) used a latent learning paradigm to test this hypothesis in rats. In this type of paradigm, animals are initially allowed to explore the experimental apparatus but are not rewarded therein. Animals are then food deprived and are trained in the apparatus as usual to determine whether they derived any benefit from this preexposure as indicated by the fewer number of errors they might subsequently make compared to control animals. The purpose of the Miller and Drew study was to test whether preexposure to the maze would reduce the innate curiosity of these animals during subsequent testing, and hence cause them to enter fewer culs. The results showed that animals given preexposure under the influence of h9-THC ( 3 mg/kg, i.p.) made significantly more errors than did preexposed control animals and did not differ from animals receiving no preexposure at all, suggesting that attentional processes may have been affected as suggested. However, no separate analysis was made of the exploratory behavior of drug and control animals during the preexposure period and, therefore, it is possible that AS-THC treated animals simply explored less of the maze, owing to possible depressant effects of the drug. Hence, their curiosity was not satisfied, and as a result they made more errors during testing. Alternatively, on the basis of an experiment by Brown (1971), Miller and Drew also suggested that the effect of the drug might have been to block habitua-
342
E R N E S T L. ABEL
tion. If so, the tendency of the animals to enter previously explored but incorrect culs would still remain high, and therefore, they would tend to make more errors than control subjects upon retesting. The experiment by Brown (1971) just referred to involved exposing mice under either drug or placebo conditions to an empty box for 15 minutes, and then returning them to their home cage. Two other groups of mice were injected with either drug or placebo but were not placed in the apparatus. Seventy-two hours later, these animals (which had been deprived of water for 24 hours) were reintroduced into the test box, which now contained a water bottle. The rationale was that previously exposed animals would have become habituated to the apparatus and, therefore, would spend less time in exploration. The data showed that animals previously exposed to the apparatus under the influence of As-THC ( 1 or 3.2 mg/kg, i.p.) did not manifest any effect of such prior treatment whereas their placebo counterparts exhibited significantly shorter latencies in contacting the water bottle. These data thus suggest that the drug inhibited the habituation process. Since habituation can be thought of as a simple form of learning [not to respond to stimuli which have no biological consequences (cf. Thorpe, 1956)], Brown’s data can be used as evidence to support the hypothesis that As-THC suppresses acquisitional processes in learning, However, since no independent observations were made of the general activity of the treated animals, it is possible that animals receiving the drug were either depressed or overly excited during the preexposure condition, in which case they would not have attended to the culs of the apparatus to the same extent as conrol subjects. Hence, the observed effects on habituation may have been secondary to the drug’s action on motor activity, or attentional processes.
C. POST-TRAINING DRUGADMINISTRATION Very little experimental work has been devoted to examining the effects of marihuana on consolidation processes using posttreatment drug administration. In an experiment by Barry and Kubena (1971) in which the effects of chronic administration of AS-THC (8 mg/kg, i.p.) on shock avoidance learning in rats was investigated, animals injected 1-3 hours after training did not differ from controls on retesting. Similarly, A9-THC (20-40 mg/kg, i.p.), administered 1 hour after training did not affect avoidance behavior in the mouse (Goldberg et al., 1973). In a maze learning experiment reported by Carlini and Kramer (1965), rats were injected with cannabis extract (10 mg/kg, i.p.) 30 seconds after they had successfully negotiated the maze and had received food reward. I n agreement with the two previous reports, no effect of this post injection procedure was apparent. One would not expect to find any posttrial effects with AS-THC in ani-
M A R I H U A N A , LEARNING, A N D MEMORY
343
ma1 learning experiments, however, owing to its rather slow rate of onset when administered by routes other than intravenous (cf. Abel et al., 1974).
EFFECTS D. STATE-DEPENDENT As noted in the introductory section of this review, some centrally acting drugs act like discriminative stimuli in the control of behavior. Accordingly, behavior that is learned under the influence of a drug may not transfer, or may only partly transfer to nondrug or other drugs states and vice versa for behavior originally learned under a nondrug state. The possibility that marihuana produces such “state dependent” learning is thus of no small importance to the present discussion, Since drugs that produce state-dependent learning are usually effective as discriminative stimuli, investigators have also examined whether the marihuana-state can be used as a discriminative internal stimulus for the learning of differential responses. For example, Kubena and Barry (1972) trained food-deprived rats in a bar-pressing situation where either food reward or shock was delivered after every fifth response ( F R 5 ) . During training, half the animals were injected with AS-THC ( 4 mg/kg, i.p.) while the other half were injected with placebo 30 minutes prior to testing and each of these conditions was associated with the delivery of either food or shock. The results showed that with the exception of one animal, the subjects averaged more than 90% correct responses. A similar effect has been reported by Henriksson and Jarbe (1972). A group of rats were trained to swim to either of the two arms of a T-shaped water maze depending on whether they were treated with Ag-THC (5-10 mg/kg, i.p.) or with placebo. After 11-13 sessions the animals were responding at the 100% level, i.e., they differentially swam to one of the two arms of the maze depending on whether they had been given the drug or placebo. A very interesting example of this phenomenon was reported by Bueno and Carlini (1972), in which they showed that even after the development of tolerance, a drug may still retain its internal cue properties. Rats were first trained to press one of two levers for food reinforcement until a stable baseline of behavior was achieved. The animals were next trained to climb a rope for food reward and then they were injected daily with either cannabis extract (10 mg/kg, i.p.) or placebo. After 14 such treatments, the drugtreated animals exhibited tolerance as indicated by their performance on the rope climbing task. I n the final phase of the experiment, they were returned to the two lever chambers and were given an additional session without drug. Thereafter, the animals were given 30 sessions in which they were
344
E R N E S T I,. AI3ISL
submitted to both rope climbing and bar pressing. The marihuana-tolerant animals received the extract and control solution on alternate days. O n marihuana days, only the left bar was activated whereas on placebo days the right bar delivered food reinforcement. Thus, each bar was associated with a particular drug state. The data showed that even though the animals were tolerant to the drug as indicated by their rope climbing behavior, they were still able to use the drug no-drug cue to respond correctly in the two choice situation. T o demonstrate state-dependent learning effects, four groups of animals must be studied. Two of the groups are trained under either drug or placebo conditions and are then tested under these same conditions (designated D-D and C-C, respectively). T h e other two groups are switched such that those originally trained under drug conditions are tested under placebo (D-C) and vice versa for those originally trained under placebo conditions (C-D) . Using this basic paradigm, Henriksson and Jarbe ( 1971 ) trained rats that had been injected daily with AS-THC (7.5 mg/kg, i.p.) A*-THC (15 mg/kg, i.p.) or placebo for 6 days in a shock avoidance task. O n day 8, the animals were tested under either drug or placebo as described above. T h e data showed a complete disruption of avoidance behavior when animals were switched from the drugged to the nondrugged condition (D-C) and vice versa for the nondrugged to drugged condition (C-D), for both drugs. Animals in the D-D and C-C conditions either showed continued improvement or performance that did not differ from previous behavior. In a second experiment of this type Jarbe and Henriksson (1973) trained rats for 6 consecutive days under either AS-THC ( 5 mg/kg, i.p.), AS-THC (10-20 mg/kg, i.p.) or placebo to swim to one of two arms of a T-shaped water maze. O n the seventh day, reward training was begun. These conditions were similar to those in original training except that the correct side was reversed for all subjects. The rationale for this experiment was that if there was state-dependence associated with these drugs, animals in the D-D and C-C conditions should have more difficulty in making the switch because of the association of a particular response, e.g., right turn, with the particular drug state, and hence they should make more errors than subjects in the D-C and C-D conditions for whom the association of drugright turn no longer holds. The data indicated that this was indeed the case for the a R - T H Cgroups, but not for the animals treated with Ag-THC. Glick and Milloy (1972) used the passive avoidance shock paradigm to study state-dependent learning with Ag-THC ( 2 mg/kg) in female mice. The animals were injected (i.p.) with the drug prior to a single learning trial. They were retested on days 1 and 7 after this experience under either placebo or drug conditions. Only the animals receiving the drug prior to the initial learning experience showed impairment of retention (i.e., reentry
M A R I H U A N A , I,ISAKNING, AND M E M O R Y
345
into the shock area). Neither the placebo or the animals receiving drug prior to training and again prior to retesting showed impairment. Thus, these experiments demonstrate that state-dependent learning can occur under the influence of marihuana. The existence of this phenomenon thus represents another source of difficulty in trying to assess the effects of this substance on learning and retrieval processes.
Ill. Human Studies
In human learning/memory experiments, four basic tasks have generally been employed by investigators to assess marihuana’s effects on these cognitive processes. These are digit span ( D S ) , serial subtraction (SS), goal directed serial alternation (GDSA), and free recall. Since these tasks are also widely used, they will be described only briefly. I n the DS task, the subject hears a number of digits and is required to reproduce them accurately in the same or reverse order. The largest number of digits reproduced by the subject without an error constitutes the measure of the memory. The SS task requires the subject to repeatedly subtract some number such as 7 from some assigned number until zero is reached. I n the third numerical problem, the GDSA task, the subject is given a starting number such as 110 and he is then asked to subtract 7 and then add either 1, 2 , or 3 and to continue such alternate subtraction and additions until he reaches a particular numerical goal. I t is assumed that the DS task reflects short-term rote memory, the SS reflects simple arithmetic operations stored in long-term memory, and the GDSA reflects both short and long-term memory functions since the subject must simultaneoulsy hold in mind and coordinate information as well as perform cognitive operations relevant to pursuing a goal. In general, even with simple memory tasks such as the DS, the results have been inconsistent. For example, while a number of investigators have reported that marihuana impaired performance on this task (Clark and Nakashima, 1968; Galanter et al., 1973; Melges et al., 1970; Tinklenberg et al., 1970), other investigators have not been able to detect any differences between marihuana and control conditions (Caswell and Marks, 1973b; LaGuardia Report, 1944; Rafaelsen et al., 1973; Tinklenberg et al., 1972; Waskow et al., 1970). As indicated by Table I1 this discrepancy cannot be attributed to either dose or route of administration. With respect to serial subtraction tasks, several investigators have observed poorer performance under drug compared with placebo conditions (Caswell and Marks, 1973b; Manno et al., 1970; Rafaelsen et al., 1973; Waskow et al., 1970) although no differences have also been reported (Melges et al., 1970). I n evaluating
346
ERNEST L. ABEL
TABLE I1 EFFECTS OF MARIHUANA ON SIMPLE HUMAN MEMORY TASKS Task
DS DS DS DS DS
DS DS DS
ss ss ss ss
GSDA GSDA GSDA
Dose
Route
1-30 mg/16 26 mg Ag-THC 10 mg Ag-THC 20, 40, 60 mg Ag-THC 8, 12, 16 mg A9-THC 3.3, 6.6 mg A9-THC 20 mg/kg Ag-THC 40, 60 mg A9-THC 10 mg A9-THC 3.3, 6.6 mg Ag-THC 40, 60 mg A9-THC 300, 400 mg AQ-THC 26 mg AS-THC 3.3, 6.6 mg A9-THC 40, 60 mg Ag-THC
Oral Oral Smoke Oral Oral Smoke Oral Oral Smoke Smoke Oral Oral Oral Smoke Oral
Effect
References
1
Clark and Nakashima (1968) Tinklenberg et al. (1972) Galanter et 01. (1973) Tinklenberg et al. (1970) Rafaelsen et al. (1973) Caswell and Marks (1937a) Waskow el al. (1970) Melges et 01. (1970) Manno et al. (1973) Caswell and Marks (1973a) Melges et al. (1970) Kafaelsen ct al. (1973) Tinklenberg et al. (1972) Caswell and Marks (1973a) Melges et al. (1970)
-
4 -
-
1 1 1 -
1 -
1 1
these data, most of these studies have given equal weight to both the time taken and the number of errors made. When errors alone have been examined, however, no differences have been observed between drug and control conditions (Rafaelsen et al., 1973). This same confounding of reaction time and error rate has also characterized GDSA task. Hence, the poor performance of subjects in the task under drug condition in some experiments (Caswell and Marks, 1973b; Melges et al., 1970), though not all (Tinklenberg et al., 1972) must be interpreted with caution, since it is not clear whether these effects are due to drug interference with memory or with arithmetic calculation processes or both. Accordingly, other kinds of task have been used that are more complex than the simple DS and require the subject to learn and repeat passages of prose or learn lists of disconnected words or syllables. For example, in a study by Abel (1970b), subjects were required to read a story through twice under either marihuana (Ag-THC content not specified) or control conditions. Fifteen minutes later they were asked to write out as much of the story using as many exact words as possible. One month later the procedure was repeated with the exception that subjects previously tested under marihuana were now tested under no drug and vice versa. Analysis of the protocols indicated that in the marihuana condition subjects remembered significantly fewer ideas and fewer context words from the story than in the nondrug condition. Similar results using prose material has also been reported by Miller et al. (1972), who also found significantly more distortions in the marihuana (25 mg/kg) protocols (cf. also Drew et al., 1972).
M A R I H U A N A , LEARNING, A N D MEMORY
347
The data arising from these latter procedures are largely taken from the growing study of human verbal learning and memory (e.g., Cermak, 1972; Dixon and Horton, 1968; Hall, 1971) and can be interpreted within the theoretical models of short- and long-term memory processes that have been described by Atkinson and Shiffrin (1968). I t is in the context of this model that the following studies will be discussed.
A. RETRIEVAL One plan of attack has been to read lists of words to subjects prior to giving them any drug treatments. The subjects are then asked to repeat as many words from each list as they can without regard for sequence (called free recall, in contrast to asking them to repeat the words in the order originally read to them). These subjects are matched on their predrug performance, then are given the drug or placebo. (This corresponds to the training of animals under a no-drug condition and then testing them while they are under the influence of the drug. It avoids many of the problems of such animal studies, however, because memory function is inferred from verbal rather than motor activity.) If the two groups do not differ on their predrug scores, any postdrug differences in their recall of this material can only be due to a drug-induced impairment in the retrieval of material already in long-term storage. A failure to observe any such effect of marihuana (Ag-THC content not specified) on retrieval processes was initially reported by Abel (1971a,b) and has been confirmed by Darley, Tinklenberg, Roth, Hollister, and Atkinson (197313) and Dornbush (1974) using doses of 20 mg of AS-THC (oral and smoked, respectively). The absence of any significant differences between marihuana and control groups in these delayed-recall tasks has been interpreted as evidence suggesting that marihuana does not affect the retrieval of information that has already been stored in memory before drug administration.
B. ACQUISITION
If a subject is required to recall material to which he has been exposed while under the influence of the drug, it is reasonable to assume that an inability to do so is likely due to either ( 1 ) the information not being registered at the sensory receptor level, or ( 2 ) it enters the short-term memory store, but is not passed on to long-term memory or, ( 3 ) it is passed on to the long-term memory store but does not become permanently encoded, or, (4) it is encoded in long-term memory, but is stored in such a manner that it cannot be located except when the original drug condition is reinstated (state-dependent learning).
348
E R N E S T L. ABEL
If the total number of words a subject recalls are plotted on the ordinate against their original serial position on the abscissa, it is then possible to use the model described by Atkinson and Shiffrin (1968) to determine which of the above-mentioned possibilities are involved in the drug-induced impairment. This analysis is based on experimental findings which indicate that the shape of the serial position curve in free recall can be altered by different procedures. Typically, the curve is U-shaped with the right side of the U slightly higher than the left, indicating that, with no delay between presentation of words and their recall, more words are recalled from the last items of the list than from the early part, whereas the fewest items are recalled from middle. This is because the subject typically repeats the words at the end of the list first while they are still ‘‘echoing” around in his head. This is called the recency effect. Items in this very end position are assumed to be those recalled from the sensory register component of memory which receives information from the sense organs (the “echo” box) . This information decays very rapidly if not passed to short-term memory. Next to these items on the curve are those assumed to be recalled from the short-term store. Loss of material from this store occurs by spontaneous decay but such loss can be prevented by rehearsing items (i.e., by repeating to oneself those items) that one wishes to retain. Thus, the longer an item is rehearsed, the longer it stays in the short-term store and the longer it remains there, the greater the probability that it will be transferred to the long-term store. T h e left-hand side of the serial position curve is assumed to represent items that have been transferred to this long-term store. Finally, the middle of the curve reflects those items which have decayed and hence have not passed to the long-term store. Following the assumptions of this model, Abel (1971a) reported that while marihuana did not affect the very end position (sensory register) or the middle of the serial position curve, it did affect the very beginning (longterm store) and the items near the end (short-term store). Thus, since items in the sensory register were not affected, subjects under the influence of marihuana had received the same amount of information as in the control condition. On the other hand, inspection of the curves indicated that items in the short-term store were adversely affected as were items in the long-term memory. Results similar to these have recently been reported by Darley and his associates (1973a) and Dornbush (1974) with similar procedures using 20 mg of Ag-THC (oral and smoking, respectively). These results, thus, suggest that marihuana has a detrimental effect on short-term storage processes in human memory. In another experiment of this type, Dornbush et al. (1971) presented a consonant trigram, e.g., DKF, to the subjects and had them recall it either immediately or after a delay of 6, 12, or 18 seconds during which they were
MARIHUANA, LEARNING, A N D MEMORY
349
required to perform some interpolated task. (Although not stated, subjects usually are asked to count backward out loud by 3's from some given number until the time interval is over. I t is assumed that the subject cannot count backward and rehearse material at the same time. Hence, the longer the time interval from presentation to recall, the greater the opportunity for spontaneous decay of information from the short-term store.) Dornbush and her co-workers (1971) found that the longer the delay, the worse the performance by the subject in the marihuana (3.75 and 11.25 mg of A'-THC, oral injestion) condition. A similar experiment was reported by Galanter and his associates (1973) in which they presented a list of numbers to subjects and then asked the subjects to recall it immediately or after a 4-second delay in which they had to recite letters of the alphabet. Subjects in the marihuana condition (10 mg of h'-THC, smoking) did worse after both the zero and the 4-second delay. Since even at zero delay, marihuana impaired performance, it would appear that marihuana also affects the encoding of information in the short-term memory store. An effect on encoding mechanisms has also been proposed by Darley et al. (1973a) as a possible effect of marihuana on the basis of the following study: Subjects were presented a series of items followed by a test stimulus which had or had not been in the preceding series. The subject had to press one key if the test item was part of the list and another key if it were a foreign item. Since subjects are nearly always correct in this task, the main measure of performance was reaction time, i.e., the time between presentation of the test stimulus and the response. I t is assumed that three independent processes contribute to this measure. The first is the encoding time, the second is the time for comparison of the test stimulus with the memory set, and the third is the time for selection of the response (yes or no) and execution. On the basis of mathematical analysis of the data, it is possible to distinguish the first and third process from the second, but not the first from the third. With this procedure and analysis, it was found that under marihuana (20 mg of AS-THC, oral) there was an increase in encoding and/or response times in short-term memory that was not due to any effect on search or comparison mechanism. However, since marihuana does affect motor activity, it is possible that the greatest effect was on response time rather than on encoding. Abel ( 1971a ) has suggested that information does not become stored in long-term memory in subjects under the influence of marihuana because they do not rehearse material in short-term memory and hence it does not remain in the latter long enough for it to be transferred. I n order to rehearse, the subject must constantly attend to the task at hand. Such lacks of attention following marihuana have been reported anecdotally as well as experimentally. For example, in a study reported by Caswell and Marks (197313)
350
ERNEST L. ABEL
subjects faced a panel containing a control light surrounded by a number of peripheral lights. The central light flashed at a rate of once per second with randomly programmed breaks interspersed, and the subject had to press a key each time a break in the sequence was detected. I n addition, the peripheral lights were also programmed to flash randomly and the subject had to press a key in response to the peripheral light as well. Caswell and Marks found that significantly more lights were missed by subjects under marihuana (3.3 and 6.6 mg of AS-THC, smoking) compared to control conditions suggesting that the drug adversely affected attention (although the contribution of motor impairment per se was not examined). I t would thus appear that the effects of marihuana on human memory are due both to a loss of material in short-term memory as a result of spontaneous decay, and to an impairment of attentional mechanisms necessary for rehearsal to occur. As noted earlier, subjects may also perform worse in memory tasks under the influence of marihuana due to a state-dependent effect. If this were so, however, one would expect differences between drug and control subjects on retrieval of information previously learned under nondrug conditions. However, there are no experimental data to support this. On the other hand, there is some evidence reported by Rickles et al. (1973) indicating that material originally learned under the influence of marihuana (14 mg of A9-THC) is recalled better under marihuana than under placebo conditions. In this experiment, subjects were divided into four groups and were tested twice, 10 days apart. Testing was performed under either marihuana or placebo. Subjects were first required to learn paired associates consisting of a three consonant trigram and a common English word until they could anticipate each English word associated with the trigram given only the latter. This was then followed by 100% overlearning in which the subject was given as many additional practice trials as he originally required to master the list. For example, if he needed five trials to learn the list, he was given five more practice trials. Ten days later subjects were required to associate the correct English word with each trigram as before. Rickles and associates found that significantly more pairs were remembered by subjects who were both trained and tested under marihuana or placebo than did subjects trained under one condition and then switched to another. A similar effect was reported by Hill et al. (1973) using a task involving sequential memory (recall of ordered objects) but not in a task involving transfer of learning (subjects were taught to press one of two switches in response to a light on day 1, and then the other switch on day 2 ) . The discrepancy between the tasks was attributed to greater difficulty in the sequential memory task, making it more sensitive to dissociative effects.
MARIHUANA, LEARNING, A N D MEMORY
35 1
IV. Summary and Further Considerations
O n the basis of the results of the various studies just reviewed, one is left with the impression that the animal and the human data are at variance with one another. For instance, while in animals marihuana appears to impair performance in previously learned tasks, it has no such effect in humans. Conversely, whereas marihuana adversely affects human performance in acquisition, it has no such clear-cut effect in animals. However, when we take into account the procedural difficulties of many of the animal studies, then one is left with the impression that most of the studies that have been conducted with regard to the effects of cannabis on learning and memory are inadequate to their intended purposes. If we subscribe to the law of parsimony, then nearly all of the animal experiments discussed with respect to acquisition are more readily explicable in terms of drug effects on motor activity, motivational levels, attentional mechanisms, etc. As stated at the outset of this review, one of the main problems in the area of learning and memory is that these processes are not directly observable but must be inferred from a subject’s behavior. Before any effect on learning and/or memory can be accepted then, all other possibilities must be eliminated. For example, if a drug is found to either depress or facilitate shock avoidance behavior it must be demonstrated that the drug has not affected general motor activity before any direct effect of the drug on learning and/or memory can be accepted. In such cases it is not merely adequate to show an effect on avoidance and no depression of escape, it is also important to determine the latency for such escape responses because of the possibility of analgesic and cataleptic effects associated with AS-THC. The effects of shock level per se, which are usually totally ignored in drug studies, have also been shown to be quite important insofar as behavior is concerned. For example, the speed of running from a shock source is directly related to the intensity of the shock (Amsel, 1950; Campbell and Kraeling, 1953) as is the latency of the bar-pressing response to escape shock (Borren et al., 1959). If a drug affects the sensitivity of an animal to foot shock, then any effects of that drug on learning and/or memory would be secondary to its affects on performance as a result of a change in the level of reinforcement. This has been shown to be the case for p-chlorophenylalanine (pCPA) by Tenen (1967), who observed that this compound had a facilitatory effect on shock avoidance learning in rats at a dose that produced no overt behavioral effects. Further experimentation, however, revealed that the drug had increased the sensitivity of the rats to the foot shock which was tantamount to increasing the shock intensity for drug-treated animals compared
352
E R N E S T L. ABEL
to controls. Since A“-THC has been observed to have analgesic effects (Bicher and Mechoulam, 1968; Bauxbaum, 1972), differences in performance may possibly be due to a perceived difference in the shock level being used to motivate behavior. Studies of the effects of marihuana on avoidance should thus require simultaneous measurement not only of the number of conditioned (avoidance) and unconditioned responses (escape) made by the animal, but also the number of intertrial responses and the latency of response to the CS and UCS. These data are important because only with such information is it possible to draw conclusions regarding a drug’s effects on conditioned versus unconditioned behavior, general motor activity, and sensitization (overall responsiveness) . A second possibility to be considered arises from the nature of the conditioned avoidance situation itself. According to the theory first proposed by Mowrer (1947) and later elaborated on the basis of further experimental evidence by Solomon and Wynne (1954), the pairing of the CS and UCS results in establishment of a conditioned emotional response (“fear”) which serve as the motivation for performing the avoidance response. If this theory is accepted as a reasonable explanation of the avoidance phenomenon, then it behooves any investigator who uses this paradigm to explore the effects of a drug on the learning or retention of an avoidance response, to demonstrate that the drug has no effect on “emotionality.” In some cases this can be done by independently assessing the animal’s behavior in a situation such as the “open-field’’ (cf. Henderson, 1970). I n the case of learning/memory tasks involving food reinforcement as the motivating variable, the clearly depressant effects of marihuana and its homologs on food intake (e.g., Abel and Schiff, 1969; Manning et al., 1971; Scheckel et al., 1968) must be taken into account in evaluating the data from such experiments. In addition, in light of the motor effects of this compound, speed scores are clearly inappropriate in monitoring the results of maze learning performance. Finally, the evidence with respect to marihuana-induced state-dependent learning would seem to preclude any easy solution to the problems of this drug’s effects on retrieval mechanisms in either maze or shock avoidance learning paradigms. A few comments with regard to cannabis itself also seem to be in order. First and foremost, as with all drug experiments, single-dose studies are only suggestive of effects. Higher or lower doses of drug may give just the opposite effect as that observed for any single dose. This is especially likely in light of the biphasic effects connected with A9-THC such as increases in motor activity following small doses and decreases in motor activity following large doses (Abel, 1970a; Carlini et al., 1970; Davis et al., 1972). The importance of the time variable should also not be underestimated since a period
MAKIHUANA, LEARNING,
AND M E M O R Y
353
of initial stimulation of activity followed by depression is often observed with the same dose of this drug (Garriott et al., 1967; cf. Domino, 1971). I n the case of repeated training and/or testing of aimals under drug conditions, the rapid development of tolerance to cannabis in most animals (Abel et al., 1974; McMillan et al., 1970) must be clearly taken into account and may actually preclude assessment of the acute effects of the drug on learning and/or memory. I n light of the comments regarding the role of emotionality in avoidance behavior, it should also be noted that a n increase in open-field defecation, indicative of increasing fear (Henderson, 1970) has been reported following chronic treatment with cannabis (Masur et al., 1971 ) . Likewise, Carlini and co-workers (1972) have presented data which suggest that following the development of tolerance to the depressant effects of cannabis, excitatory effects of the drug are likely to become “unmasked.” I n summary, the data with respect to the effects of marihuana on learning and/or memory are such that it is not possible at present to conclude that this material has or has not any effect on these cognitive processes in animals. With respect to the human data, most of the pitfalls vis-A-vis interpretation of results have been circumvented by use of tasks that do not involve measurements of the rate or speed of locomotor activity. While the problem of motivation must still be considered, the nature of the problem does not involve the difficulties inherent in the animal data such as drug effects on emotionality or hunger which must certainly enter into discussion of animal behavior studies in this area. Although previous tolerance to marihuana must also be taken into account in evaluating the human data, for the most part this possibility has not been studied. However, if tolerance were an important consideration, then we ought to expect equal impairment on both retrieval and acquisition when the same subjects are used. Since, this is not the case, it would suggest that tolerance to the adverse effects of marihuana on human learning does not occur. ACKNOWLEDGMENTS T h e comments and advice of Dr. Allen Barnett and the secretarial fortitude and assistance of Peggy Bielawicz are gratefully acknowledged.
REFERENCES Abel, E. L. (1969). Psychon. Sci. 16,44. Abel, E. L. (1970a). /. Pharm. Pharnzacol. 22, 785. Abel, E. L. (1970b). Nature ( L o n d o n ) 227, 1151-1152. Abel, E. L. (1971a). Science 173, 1038-1040. Abel, E. I,. (1971b). Nature ( L o n d o n ) 231, 58. Abel, E. L., and Schiff, B. B. (1969). Psychon. Sci. 16,38. Abel, E. L., McMillan, D. E., and Harris, I,. S. (1974). Psychopharmacologia 35, 29-38.
354
ERNEST L. ABEL
Amsel, A. (1950). J. Exp. Psychol. 40, 1-14. Atkinson, R. C., and Shiffrin, R. M. (1968). I n “The Psychology of Learning and Motivation” (K. W. Spence and J. T. Spence, eds.), Vol. 2, pp. 89-195. Academic Press, New York. Barry, H., and Kubena, R. K. (1970). Proc. 78th Annu. Conv. Amer. Psychol. Ass. Vol. 5, pp. 805-806. Barry, H., and Kubena, R. K. (1971). Proc. 79th Annu. Conv. Amer. Psychol. ASS.V O ~6,. pp. 747-748. Bauxbaum, D. M. (1972). Psychopharmacologia 25, 275. Bicher, H. I., and Mechoulam, R. (1968). Arch. Znt. Pharmacodyn. Ther. 172, 1. Borren, J. J., Sidman, M., and Herenstein, R. J. (1959). /. Cornp. Physiol. Psychol. 52, 420-425. Brown, H. (1971). Psychopharmacologia 21, 294-301. Bueno, 0. F., and Carlini, E. A. (1972). Psychopharmacologia 25, 49-56. Campbell, B. A., and Kraeling, D. (1953). 1.Ex#. Psychol. 45, 97-101. Carlini, E. A,, and Kramer, C. (1965). Psychopharmacologia 7, 175-181. Carlini, E. A., Hamaoui, A., Bieniek, D., and Korte, F. (1970). Pharmacology 4, 359-368. Carlini, E. A., Masur, J., Karniol, I. G., and Leite, J. R. (1972). I n “Cannabis and its Derivatives” (W. D. M. Paton and J. Crown, eds.), p. 154-173. Oxford Univ. Press, London and New York. Caswell, S . , and Marks, D. F. (1973a). Science 179,803-805. Caswell, S., and Marks, D. (197313). Nature ( L o n d o n ) 241, 60-61. Cermak, L. S. (1972). “Human Memory.” Ronald Press, New York. Clark, L. D., and Nakashima, E. N. (1968). Amer. J. Psychiat. 125, 379-384. Darley, C. F., Tinklenberg, J. R., Hollister, L. E., and Atkinson, R. C. (1973a). Psychopharrnacologia 29, 231-238. Darley, C. F., Tinklenberg, J. R., Roth, W. T., Hollister, L. E., and Atkinson, R. C. (1973b). Mem. Cognit. 1, 196-200. Davis, W. M., Moreton, J. E., King, W. T., and Pace, H. B. (1972). Res. Commun. Chern. Pathol. Pharmacol 3 , 29-35. Delini-Stula, A. (1973). Pharmakopsychiat./Neuro-Psychopharmakol. 6, 189-197. Dixon, T. R., and Horton, D. L., eds. (1968). “Verbal Behavior and General Behavior Theory.” Prentice-Hall, Englewood, New Jersey. Domino, E. (1971). Ann. N . Y . Acad.Sci. 191, 166-191. Domino, E. F., Hardman, H. F., and Seevers, M. H. (1971). Pharmacol. Rev. 23, 317-366. Dornbush, R. L. (1974). Trans. N . Y . Acad. Sci. [2] 36, 94-100. Dornbush, R. L., Fink, M., and Freedman, A. M. (1971). Amer. J. Psychiat. 128, 194-197. Drew, W. G., Kiplinger, G. F., Miller, L. L., and Marx, M. (1972). Clin. Pharmacol. Ther. 13, 526-533. Edery, H., Grunfeld, Y., Ben-Zivi, Z., and Mechoulam, R. (1971). Ann. N . Y . Acad. Sci. 191, 40-53. Fernandes, M., Warning, N., Christ, W., and Hill, R. (1973). Biochem. Pharmacol. 22, 2981-2987. Ferraro, D. P. (1972). I n “Current Research in Marihuana” (M. F. Lewis, ed.), pp. 49-95. Academic Press, New York. Galanter, M., Weingarten, H., Vaughan, T. B., Roth, W. T., and Wyatt, R. J. (1973). Arch. Gen. Psychia. 28, 278-281.
MARIHUANA,
LEARNING, A N D MEMORY
355
Garg, M., and Holland, H. C. (1969). Psychopharmacologia 14, 426-431. Garriott, J. C., King, L. J,, Forney, R. G., and Hughes, F. W. (1967). Life Sci. 6, 2119-2128. Glick, S . D., and Milloy, S. (1972). I n “Current Research in Marijuana” (M. F. Lewis, e d . ) , pp. 1-24. Academic Press, New York. Goldberg, M. E., Hefner, M. A., Robichaud, R. C., and Dubinsky, B. (1973). Psychopharmacologia 30, 173-184. Gonzalez, S. C., and Carlini, E. A. (1971). Psychon. Sci. 24, 203-204. Gonzalez, S. C., Karniol, I. G., and Carlini, E. A. (1972). Behau. B i d . 7, 83-94. Grunfeld, Y., and Edery, H. ( 1969). Psychopharmacologia 14, 200-210. Hall, J. F. ( 1971 ) . “Verbal Learning and Retention.” Lippincott, Philadelphia, Pennsylvania. Henderson, N. D. (1970). J. Psychol. 75, 19-34. Henriksson, B. G., and Jarbe, T. (1971). Psychopharmacologia 22, 23-30. Henriksson, B. G., and Jarbe, T. (1972). Psychon. Sci. 27, 25-26. Herring, B. (1972). Psychopharmacologia 26, 401-406. Hill, S. Y., Schwin, R., Powell, B., and Goodwin, D. W. (1973). Nature ( L o n d o n ) 243, 241-242. Jaffe, P. G., and Baum, M. (1971). Psychopharmacoolgia 20, 97-102. Jarbe, T. U. C., and Henriksson, B. G. (1973). Psychopharmacologia 31, 321-332. Jarvik, M. E. (1972). A n n . R e v . Psychol. 23, 457-486. John, E. R. (1967). “Mechanisms of Memory.” Academic Press, New York. Jones, G., and Pertwee, R. G. (1972). Brit. 1. Pharmacol. 45, 375-377. Karniol, I. G., and Carlini, E. A. (1973). Psychopharmacologia 33, 53-70. Keleman, K., and Bovet, D. (1961). Acta Physiol. 19, 143-154. Kubena, R. K., and Barry, H. (1972). Nature ( L o n d o n ) 235, 397-398. La Guardia Report (1944). “The Marihuana Problem in the City of New York.” Jacques Cattell Press, Lancaster, Pa. McGaugh, J. L. ( 1968). Atti. Accad. Nar. Lincei 109, 13-24. McGaugh, J. L., and Petrinovich, I,. F. (1959). A m e r . J. Psychol. 72, 99-102. McGaugh, J. L., Thompson, C. W., Westbrook, W. H., and Hudspeth, W. (1962). Psychopharmacologia 3, 352-360. McMillan, D. E., Harris, L. S., Frankenheim, J. M., and Kennedy, J. S. (1970). Science 169, 501-503. Manning, F. J., McDonough, J. H., Elsmore, T. F., Saller, C., and Sodetz, F. J. (1971). Science 174, 424-426. Manno, J. E., Kiplinger, G. F., Haine, S. E., Bennett, I . F., and Forney, R. B. (1970). Clin. Pharmacol. T h e r . 11, 808-815. Manno, J. E., Kiplinger, G. F., Haine, S. E., Roth, W. T., and Wyatt, R. J. (1973). Arch. Gen. Psychiat. 28, 278-281. Masur, J., Martz, R. M. W., and Carlini, E. A. (1971). Psychopharmacologia 19, 388-392. Mechoulam, R. (1970). Science 168, 1159-1 166. Melges, F. T., Tinklenberg, J. R., Hollister, L. E., and Gillespie, H. K. (1970). Science 168, 1118-1 120. Miller, L. L., and Drew, W. G. (1973). Nature ( L o n d o n ) 243, 473-474. Miller, L. L., Drew, W. G., and Kiplinger, G. F. (1972). Nature ( L o n d o n ) 237, 172-173. Miller, L. L., Drew, W. G., and Joyce, P. (1973). Behav. B i d . 8, 421-426. Mowrer, 0. H. (1947). Haruard Educ. Rev. 17, 102-148.
356
E R N E S T L. ABEL
Orsingher, 0. A,, and Fulginiti, S. ( 1970). Pharmacology 3, 337-344. Overton, D. A. (1964). 1.C o m p . Physiol. Psychol. 57, 3-12. Pradhan, S. N., Bailey, P. T., and Ghosh, B. (1972). Res. C o m m u n . C h e m . Pathol. Pharmacol. 3, 197-204. Rafaelsen, L., Christrup, H., Bech, P., and Rafaelsen, 0. J. (1973). Nature ( L o n d o n ) 242, 117-118. Rickles, W. H., Cohen, M. J., Whitaker, C. A., and McIntyre, K. E. (1973). Psychopharmacologia 30, 349-354. Schneckel, C. L., Boff, E., Dahlen, P., and Smart, T. (1968). Science 160, 1467-1469. Sodetz, F. J. (1972). In “Current Research in Marijuana” (M. F. Lewis, ed.), pp. 25-48. Academic Press, New York. Solomon, R. L. and Wynne, L. C. (1954). Psychol. Rev. 61, 353-385. Tenen, S. S. (1967). Psychopharmacologia 10, 204-219. Thorpe, W. H. (1956). “Learning and Instinct in Animals.” Methuen, London. Tinklenberg, J. R., Melges, F. ’I’.Hollister, , L. E., and Gillespie, H. K. (1970). Nature ( L o n d o n ) 226, 1171-1172. Tinklenberg, J. R., Kopell, B. S., Melges, F. T., and Hollister, L. E. (1972). Arch. Gen. Psychiat. 27, 812-815. Walters, G. C., and Abel, E. L. (1970). I . Pharm. Pharmacol. 22, 310-312. Waskow, I. E., Olsson, J. E., Salzman, C., and Katz, M. M. (1970). Arch. Gen. Psychiat. 22, 97-107. Webster, C . D., Willinsky, M. D., Herring, B. S., and Walters, G. C. (1971). Nature ( L o n d o n ) 232, 498-501. Zimmerman, B., Glick, S. D., and Jarvik, M. E. (1971). Nature ( L o n d o n ) 233, 343-345.
NEUROCHEMICAL AND NEUROPHARMACOLOGICAL ASPECTS OF DEPRESSION By B. E. Leonard’
Pharmacology Department, Organon International B.V., Oss, The Netherlands
I. Introduction
.
.
11. Characteristics of the Affective Disorders 111. The Biogenic Amine Hypothesis of Affective Disorders
.
IV. Cyclic AMP and Possible Connection with Affective Disorders . V. Some Biochemical Effects of Drugs Used in the Treatment of Affective Disorders A. Tricyclic Antidepressants . B. Monoamine Oxidase Inhibitors C. Amphetamines. D. Lithium. E. Electroconvulsive Shock (Em) F. Reserpine and Related Alkaloids . G. Steroids. VI. Conclusion . References .
.
. .
. . .
.
. *
.
. . . .
. . .
357 359 360 367 368 368 372 374 375 376 377 377 380 381
I. Introduction
Some 90 years ago Thudicum, the father of neurochemistry, speculated : “Many forms of insanity are unquestionably the external manifestations of the effects upon the brain substance of poisons fermented within the body, just as mental aberrations accompanying chronic alcoholic intoxication are the accumulated effects of a relatively simple poison fermented out of the body. These poisons we shall, I have no doubt, be able to isolate after we know the normal chemistry in uttermost detail. And then will come in their turn the crowning discoveries to which our efforts must ultimately be directed, namely, the discoveries of the antidotes to the poisons and to the fermenting causes and processes which produce them.” Until the 1950s many people would have questioned the validity of such a prophesy. However, since that time advances in pharmacological, neurochemical, and clinical
’
Present address : Department of Pharmacology, University College, Galway, Republic of Ireland.
357
358
B. E. LEONARD
techniques have changed the very fabric of psychiatric medicine in most industrialized countries of the world, thereby giving credence to Thudicum’s speculations. From the philosophical point of view, mental illness can be considered either as a manifestation of an inherent pathological abnormality which may be exacerbated by the external environment, or as an abnormality which is caused by external environmental influences a t some stage of development. Genetic, biochemical, and pharmacological evidence, which I would prejudicedly call “objective evidence,” strongly implicates a biochemical lesion as the underlying cause of severe mental disease. In contrast, the psychoanalytical approach strongly implicates the external environment as the fundamental cause. Thus the way in which mental disease is treated depends to a large extent upon the philosophical approach of the clinician. There is little doubt that the advances in psychiatric medicine over the last decade or so can be attributed to the efficacy of the phenothazines and the tricyclic antidepressants. And yet such an advance was not the result of the application of a carefully considered hypothesis concerning the pathological basis of mental disease. Serendipity has almost invariably been the only means whereby therapeutically useful drugs have been discovered. For example, lithium salts, which are now quite widely used in the treatment of mania, were discovered by Cade (1949) following the observation that high and undoubtedly toxic doses of these salts produced behavioral depression when they were injected into rodents. He speculated that this “depressive” effect would be beneficial in the treatment of manic patients and found this to be the case. Because of its structural similarity to isoniazid, iproniazid was first used in the treatment of tuberculosis, where it was found to have a mood-elevating effect. This led the investigators to explore its actions in depressed patients with beneficial results (Crane, 1956, 1957 ; Loomer et al., 1957). Similarly, imipramine was first tested in schizophrenic patients because of its structural similarity to chlorpromazine, a well established antischizophrenic drug whose therapeutic properties had also been discovered by chance some years earlier. Kuhn (1958) found that imipramine was ineffective as an antipsychotic agent but that it did possess useful antidepressant properties. During this period when rapid advances were being made in the treatment of mental disease, clinicians were also discovering means whereby abnormal mental states could be exacerbated by drugs. Thus Bunney and Davis (1965) reported that up to 15% of patients being treated with reserpine for severe hypertension developed depressive symptoms, which abated only after the therapy was terminated. This was yet another observation which, together with the earlier discovery that reserpine caused marked behavioral depression, ptosis, and hypothermia in rodents, helped to lay the basis for a biochemical theory of mental disease. The existence of drugs that either
NEUROCHEMICAL ASPECTS OF DEPRESSION
359
caused or relieved mania and depression raised the question of whether a common denominator could be found in the action of these drugs. If so, this might provide the essential clue to the etiology of these syndromes. The essential link appeared to be provided by the biogenic amines, in particular norepinephrine, dopamine, and serotonin. I t now seems likely that the amine theory of mental disease provides a comprehensive basis not only for understanding the disease process, but, more importantly, for enabling therapeutically efficacious drugs to be discovered in a rational way.
II. Characteristics of the Affective Disorders The term “depression” generally refers to the symptoms of sadness of affect or to one of the various psychiatric syndromes in which the depressed mood is a prominent feature. The clinical syndromes of depression include different combinations of the following symptoms : depressed mood which is of greater severity and duration than may be regarded as normal in the situation; crying, feelings of inadequacy ; guilt ; hopelessness ; suicidal preoccupation; loss of drive and ambition; mental and motor retardation or agitation ; anxiety and sleep disturbances. There are also disturbances of vegetative functions, such as appetite, constipation, weight loss, loss of libido, which are associated with these psychological disorders. Distinguishing and classifying the separate syndromes of depression on the basis of these symptoms has been a long-standing problem in psychiatry, so that a number of overlapping systems have been developed for classifying these disorders. These have been reviewed by Schildkraut et al. ( 1968), Ollerenshaw ( 1973), and Sartorius (1974) and have been the subject of a monograph by the American Psychiatric Association (Davis et al., 1968a). The depressions for which the antidepressant drug and electroconvulsive therapy (ECT) are of most clinical value are those disorders which have been designated “endogenous depression” in many systems of classification (Davis et al., 196813; Kiloh et al., 1962). These depressions characteristically show autonomy of the depressive symptoms once the illness is established and lack of reactivity of the symptoms either to day-to-day changes in the patient’s environment or to social interactions. Endogenous depression and involutional melancholia are generally included under the heading of manicdepressive disorders in many systems of classification. It must be emphasized that the diagnostic classification of the depressions is based largely on the clinical history and symptomatology; as yet there is no biochemical or physiological criterion for distinguishing among the various depressive disorders, although there is some evidence that they are genetically distinct entities (Winokur and Clayton, 1967; Perris, 1966; Rainer, 1966; Winokur, 1974).
360
B . E. LEONARD
Manias and hypomanias, which are milder manic states, are psychiatric syndromes which include elation alternating with irritability and depression, grandiosity, rapid speech, flights of ideas, and increased motor activity. There is a loss of discretion and judgment in these patients so that grandiose and unrealistic plans and commitments are often undertaken while the patient is in the manic state. I n many cases there is a fluctuation between the depressive and manic state; this may occur quite abruptly overnight. This brief summary of the clinical features of affective disorders can provide a useful background for considering the neuropharmacology of some of the drugs used in the treatment.
111. The Biogenic Amine Hypothesis of Affective Disorders
A detailed account of the synthesis, metabolism, subcellular distribution, and the various factors concerned in the physiological control of brain amines has been admirably covered by Schumann and Kroneberg (1970), Baldessarini (1972), Fuxe et al. (1970), and Costa and Meek (1974). The present account will therefore be restricted to a survey of the salient features which are important for understanding the actions of the drugs used in the treatment of these disorders. Norepinephrine (NE ) is a catecholamine located in specific storage vesicles contained in the terminals of postganglionic sympathetic nerve fibers, where it functions as a neurotransmitter (Thierry e t al., 1973). I t is also localized in specific anatomical regions of the brain, where it is presumed to have a similar function. The synthesis of NE involves the hydroxylation of tyrosine to 3,4-dihydroxyphenylalanine (DOPA) which is then decarboxylated by DOPA decarboxylase to dopamine. Dopamine (DA) is then hydroxylated in the side chain to NE by the action of dopamine p-oxidase. This latter step takes place in the synaptic vesicles. I t is now well established that the synthesis rate under normal physiological conditions is controlled by a feedback mechanism whereby an excess concentration of physiologically active amine reduces the activity of tyrosine hydroxylase, the rate-limiting enzyme, either by acting on an allosteric site on the enzyme surface or by compexing with the pteridine cofactor for this enzyme. This subject has been admirably reviewed by Glowinski and Baldessarini ( 1966), Bloom and Giarman ( 1968), Baldessarini ( 1972), and Schildkraut (1974a). The breakdown of NE involves two main pathways. I t may be initially deaminated by monoamine oxidase (MAO) and then 0-methylated by catechol-0-methyltransferase ( C O M T ) to form 3-hydroxy-4-methoxymandelic acid (VMA) or 3-methoxy-4-hydroxyphenylglycol(MHPG) . Alternatively, circulating or newly released NE may be initially methylated by
NEUROCHEMICA13 ASPECTS OF DEPRESSION
361
C O M T to form normetanephrin, which can then be deaminated by MAO. The major process leading to the inactivation of physiologically released NE and serotonin (HT) appears to be by a specific reuptake mechanism whereby the amines are transported by an active transport (i.e., energy dependent) mechanism back into the preganglionic nerve terminal, where they enter specific storage granules. In peripheral nervous tissue, it would appear that a second active transport mechanisms exists whereby the NE can be taken into the postsynaptic terminal and inactivated by rebinding ( Iversen,
1974). Until recently it was generally assumed that the primary function of dopamine in the brain was to act as a precursor of NE. I t is now believed that this amine has a transmitter function in its own right. It is known to be concentrated in the striatal region of the brain; a phylogenetically old region which is functionally concerned with locomotion and, in nonprimates, in instinctive, stereotyped behavior. I n patients suffering from Parkinsonism there is a degeneration of the dopaminergic fibers in this region, the extent of the degeneration bcing correlated with the severity of the disease. When DOPA, the amino acid precursor of this amine, is given to such patients (particularly when it is combined with a peripheral decarboxylase inhibitor to prevent any breakdown of the amino acid before it enters the brain), then many of the symptoms of the disease are reduced. Such evidence suggests that dopamine has a transmitter role in its own right. Further evidence is provided by the finding that most neuroleptics, which have been shown experimentally to act on DA receptors, also produce symptoms of Parkinsonism when they are administered for long periods to man. Dopamine is also metabolized by M A 0 and C O M T to yield a number of metabolites, the most important of which is probably homovanillic acid (HVA) . Serotonin ( H T ) is localized in those areas of the brain which are concerned with the regulation of emotion (the limbic system) and sleep (the raphe system in the upper part of the brainstem). There is a considerable body of evidence to implicate this amine either as a transmitter in its own right or as a modulator of transmission in the brain. HT is synthesized from tryptophan by hydroxylation to 5-hydroxytryptophan ( H T P ) ; this is catalyzed by tryptophan hydroxylase and forms the rate-limiting step in the synthesis of HT. H T P is then decarboxylated by L-aromatic amino acid decarboxylase to H T ; there is some dispute over whether this is the same enzyme as DOPA decarboxylase. The amine can be inactivated either by reuptake from the postsynaptic site into presynaptic terminal, where it can reenter the storage vesicle, or be catabolized by M A 0 intraneuronally to form 5hydroxyindoleacetic acid (HIAA) . There is now evidence to suggest that the rate of synthesis of this amine is controlled by a feedback mechanism from the postsynaptic receptor site (Carlsson et al., 1972).
362
B. E. LEONARD
Although it seems to be generally agreed that the synthesis of serotonin is controlled by a feedback mechanism, it appears that in some regions of the brain (cortex, septum, caudate and lumbosacral spinal cord) which are rich in serotonin-containing nerve endings, the rate of serotonin synthesis is controlled by the rate of uptake of tryptophan into the nerve ending (Knapp and Mandell, 1972; Mandell et al., 1974). Stated in its simplest form, the amine theory of affective disorders suggests that depression results from the reduction in biogenic amines at the receptor sites within the brain. Thus any drug (such as reserpine, which depletes brain NE, and H T or a-methyl-p-tyrosine, which reduces brain catecholamine concentrations by inhibiting tyrosine hydroxylase activity) that reduces the effective functional concentration of these amines at the receptor site will produce symptoms of behavioral depression. This indeed is the case in man; reserpine can cause symptoms that resemble those of endogenous depression (Lemieux et al., 1956; Harris, 1957), and a-methyltyrosine has been reported to cause depression in patients who were treated with the drug (Engelman and Sjoerdsma, 1966). Conversely, drugs that raise the effective concentration of these transmitters at the central receptor sites can produce behavioral stimulation. This is verified by the stimulant effects of some M A 0 inhibitors and the amphetamines in man. However, despite the support which such findings give to the amine theory caution must be exercised in extrapolating from such drug studies to specific biochemical lesions. Schildkraut (1973) has assessed many of these problems in his excellent review of the subject. All the drugs used in these clinical studies have effects on other systems as well as the monoamine pathways. More conclusive evidence has been obtained using high concentrations of the amino acid precursors of NE and HT. Thus Goodwin and co-workers (1970) at the NIMH found that doses of drugs greater than 3-4 gm a day were essential if an improvement in mood was to be obtained; they also found that one-third of their patients who did not respond to the DOPA therapy exhibited severe anger and another one-third showed signs of mania. Matussek and associates (1970) have reported a definite improvement in mood after the administration of lower doses of DOPA in combination with a peripheral decarboxylase inhibitor. In spite of these and the results of similar controlled studies, Carroll (1971 ) has concluded a critical survey of the literature by stating that L-DOPA appears to be ineffective in most cases of clinical depression. I t is well established that DOPA interferes with the metabolism of H T by completing with the uptake of H T P into the brain, by competitively inhibiting the synthesis of HT from HTP, and also by displacing H T from serotonergic sites, the DA formed from DOPA thereby acting as a false transmitter substance.
NEUROCHEMICAI, ASPECTS OF DEPRESSION
363
Thus, while it may be argued that the studies with dopa provide some support to the amine theory of affective disorders, evidence provided is by no means conclusive. More enthusiastic claims have been made for the use of serotonin precursors in depression-in particular, tryptophan and 5-hydroxytryptophan. None of the controlled trials with H T P have shown that the precursor has a beneficial effect (see, for example, Kline et al., 1964) even though the investigators found improvement in the mood of their patients in a previous uncontrolled trail. More recently, van Praag et al. (1972) found that 3 out of 5 depressed patients given up to 3 gm of HTP daily showed significant improvement compared with a placebo-treated group of depressives. The main criticism that can be made of such studies is that all the investigators used only relatively low doses of the precursor. Furthermore, it must be emphasized that although endogenous depression may be classified clinically as homogeneous, it may well be a pathogenetically heterogenous entity. This suggests that a patient would respond beneficially to H T P therapy only if he or she had a depression which involved a disturbance in brain HT metabolism (van Pragg et al., 1973; van Praag and Korf, 1971; Asberg et al., 1973). The situation regarding to use of tryptophan in the treatment of depression appears to be confusing at the present time. The first trial in which this amino acid was used was conducted by Coppen and co-workers ( 1963), who reported that it potentiated the action of a monoamine oxidase inhibitor ( M A O I ) ; a similar effect was reported by Pare (1963). Later Coppen and his group (1967) compared the effects of tryptophan alone and together with a M A 0 1 against electroconvulsive shock treatment ( E C T ) and concluded that tryptophan therapy was as efficacious in the treatment of depression as ECT. I n a double-blind study using tryptophan and amitriptyline in the treatment of depression, AliFio et al. (1973) found that patients receiving the amino acid and amitriptyline improved significantly better than these given amitriptyline alone. Prange et al. (1974) found that tryptophan was slightly superior to chlorpromazine in the treatment of mania. These authors suggested that mania and depression may be linked by a central indoleamine deficit. However, such important findings have been critically challenged by several investigators. Thus Carol1 et al. (1970) in a carefully controlled trial found that whereas only 1 out of 12 depresed patients improved after tryptophan therapy, all 12 of the patients on E C T improved. I n this study all the patients were carefully selected because of the severity of their depression, and it can be argued that tryptophan is efficacious only in the treatment of less severe forms of the disease. Other investigators have also been unable to demonstrate that tryptophan has an antidepressant
364
B. E. LEONARD
effect, even though there was biochemical evidence of increased platelet HT and cerebrospinal fluid HIAA levels (Murphy et al., 1971; Dunner and Goodwin, 1972). From the biochemical point of view there are several criticisms to the administration of high tryptophan loads. I t is assumed that by increasing the blood concentration of the amino acid, there will be an increase in the brain serotonin concentration and, as a consequence of this, the quantity of physiologically active amine which is believed to be reduced in depressives, will increase. However, it is well established that the major pathway for the catabolism is by the enzyme tryptophan pyrrolase, a liver enzyme that converts and amino acid to formylkynurenine. I t is well established that tryptophan loading can induce a rapid synthesis of this enzyme; stress is another important factor with the same effect. Thus, after an increased enzyme activity, the blood and tissue concentration of the tryptophan metabolites formed by this pathway (kynurenine, 3-hydroxykynurenine, 3-hydroxyanthranillic acid) increase and there is a reduction in the concentration of brain serotonin due to impaired transport of the precursor into the brain (Green and Cruzon, 1970). The situation is further complicated by the fact there is generally an increased adrenocortical activity associated with depression, and it is well established that a rise in the serum concentration of the corticosteroids causes an increased tryptophan pyrrolase activity. I t has been reported that, although in nondepressed patients tryptophan loading may increase HT synthesis in the brain (Eccleston et al., 1970a), as indicated by raised HIAA levels in the cerebrospinal fluid (CSF) , there is no increase in the concentration of this metabolite in depressed patients after a tryptophan load (Bowers, 1970). Nevertheless, despite these criticisms of the tryptophan loading tests, Fraser et al. (1973) found no evidence to suggest that there is an increased metabolism of tryoptophan along the kynurenine pathway in depression. I t has also been clearly established (Korf and van Praag, 1970 ; van Praag et al., 1970; Roos and Sjostrom, 1969) that the rate of increase of HIAA in the CSF of depressed patients is lower than that of nondepressed controls in an investigation in which the efflux of HIAA is blocked by probenecid. The probable explanation for these findings is that the activity of tryptophan hydroxylase is decreased in depressed patients; this causes a reduction in the rate of synthesis of H T and HIAA and an apparent decrease in the turnover of H T in the brains of depressed patients. If this is the explanation, then tryptophan loading is unlikely to have any beneficial effect on depression. Ashcroft and colleagues (1973b), in their study of the effect of a tryptophan load on brain serotonin metabolism in endogenously depressed patients, suggested that there is a reduced activity in the serotonergic system of these patients rather than a reduction in tryptophan hydroxylase activity.
NEUROCHEMICAL ASPECTS OF DEPRESSION
365
O n biochemical grounds it would seem more logical to administer the immediate precursor of HT, 5-hydroxytryptophan ( H T P ), to depressed patients, and thereby circumvent the suspected hydroxylase block. However, from experimental studies it would seem that H T P administration leads to the synthesis of HT at a number of nonserotonergic sites in the brain. I t appears that the HT formed at these sites is rapidly destroyed by MAO. Furthermore, any HT formed in NE or DA storage granules probably acts as a false transmitter substance and displaces the catecholamines from their storage sites (Carlsson, 1964; Fuxe and Ungerstedt, 1967; Lichtensteiger et al., 1967; Shaskan and Snyder, 1970). Nevertheless, there is some equivocal evidence which suggests that H T P has an antidepressant effect (see page
363). The most direct verification of the amine theory of mental disease comes from the determination of the amines in the human brain. Shaw and coworkers (1967) found that brains from suicide patients contained a lower concentration of HT than did those from patients dying from cardiac infarction. Brains from suicides were also found to have a lower HIAA concentration than those from nonsuicides (Bourne et al., 1968). These results have been confirmed by at least two groups of investigators who reported that depressed patients have a lower HIAA concentration than controls (Dencker et al., 1966; Ashcroft et al., 1966). However, in another study it was found that, although there was a reduction in the brainstem HT of suicides, most of brains were obtained from people who were suffering from inactive depression. Moreover these investigators found that there was a positive correlation between age and HT concentration so that the decrease in brainstem HT was offset to some extent by the age difference between the suicide and the control group (Pare et al., 1969). Dencker and co-workers (1966) could not find any difference in NE levels in their investigation of depressed patients. Most clinical studies have been restricted to an analysis of amine metabolites in body fluid of depressed patients. Apart from the CSF, changes in the blood and urine do not clearly reflect any change in amine metabolism in the brain as none of the amines pass through the blood-brain barrier in any significant amount. Even an analysis of the CSF may not directly reflect changes in brain amine metabolism; changes in HVA and HIAA levels may merely be a reflection of DA and HT catabolism in the capillary wall (Mendels et al., 1972). Not all investigators agree with such a pessimistic view however. Thus Weir et al. (1973) in their study of the origin of HIAA in the CSF of cats found that approximately 40% of the HIAA in the lumbar CSF was obtained from the brain, most of the remainder being contributed from the spinal cord. Garelis and Sourkes (1973) have come to a similar conclusions in their investigations of human CSF.
366
B. E. LEONARD
Although it is necessary to interpret any data obtained from an analysis of body fluids with caution, there would seem to be some relevance in undertaking such studies. Thus in patients with manic-depressive illness, urinary epinephrine and NE excretion was found to be greater during the depressive phase of the illness (Strom-Olsen and Weil-Malherbe, 1958) . I n patients with endogenous depression, Sloane and colleagues (1966) found that catecholamine excretion was reduced ; in contrast, schizophrenics with depressive episodes excreted a larger quantity of catecholamines and their metabolites than did neurotic depressives (Bunney et al., 1967). Schildkraut (1965) and colleagues reported that the excretion of normetanephrin and metanephrin were markedly increased during the psychotic delusion phase of the patients they studied. Since it has been postulated that depression is associated with a lower steady-state concentration of brain NE, and as significant quantities of this amine in the brain are metabolized to MHPG, it has been reasoned that the amount of MHPG excreted in the urine may reflect brain NE metabolism more closely than that of the other catecholamine metabolites which are conventionally measured in the urine (Maas and Landis, 1968; Schanberg et al., 1968). Thus Maas and co-workers (1968) found that the urinary levels of MHPG were significantly lower in a group of seriously depressed patients than in the nondepressed controls. I n a recent study, this group of investigators found that those depressed patients who had a low MHPG excretion prior to drug therapy showed the best response to tricyclic antidepressant (imipramine ) therapy ; no correlation was found between the improved clinical state and the excretion of metanephrine, normetanephrine, and vanillylmandelic acid (VMA) ( Maas et al., 1972). Shaw et al. (1973), however, found that there was no correlation between urinary excretion of MHPG and the concentration of this metabolite in the CSF of depressed patients; the MHPG concentration in the urine increased after recovery of the patients. The disparity between clinical improvement and lack of change of urinary or CSF-MHPG could be due to poor correlation between MHPG and NE metabolism in the brain. I n support of this view, Meek and Neff (1973a) have found that MHPG sulfate reflects brain NE metabolism more adequately than MHPG alone. Bond et al. (1972) investigated changes in urinary MHPG in 2 patients suffering from manic-depressive psychoses and found that the levels of this metabolite were elevated during the manic phase and depressed during the depressive phase of the illness. From these studies it would appear that there is a close correlation between changes in the metabolism of NE in the brain and the severity of the disease. Indole metabolism is also affected in some patients during the depression; Rodnight (1961) was one of the first investigators to show that the urinary excretion of tryptamine was significantly decreased during depression, a finding subsequently confirmed by Coppen and colleagues ( 1965) and Coppen (1972).
NEUROCHEMICAL ASPECTS OF DEPRESSION
367
From all the clinical studies it can be concluded that both NE and HT metabolism can be correlated with the severity of the depression. The function of NE in the brain is thought to be related to the drive and the motivation of the individual (Stein, 1967; Poschel and Nintemann, 1963). Some investigators have suggested that the serotoninergic system functions primarily in the control of mood (Kielholz, 1968). I t seems clear from the symptoms affected, i.e., both drive and mood are decreased, and therefore the amine theory of affective disorders seems reasonable.
IV. Cyclic AMP and Possible Connection with Affective Disorders
Bunney and co-workers ( 1970) have presented evidence indicating that changes in brain catecholamine metabolism may precede and also accompany the profound changes in mood shown by manic-depressive patients. I n their study they found a marked increase in the concentration of urinary NE on the day preceding the manic episode; this period was also associated with a decrease in the total amount of sleep, in particular in REM (rapid eye movement) sleep. These investigators then proceeded to implicate changes in the excretion of cyclic AMP (CAMP) in the etiology of manic depressive illness. The connection between disturbed amine metabolism and the cAMP system is not surprisingly because it is well established from experimental studies that this substance acts as a secondary hormone in nervous tissue in many species of vertebrate and invertebrate, its synthesis and physiological activity being governed by the action of the neurotransmitters (HT, NE, ACh, DA, Hist). The physiological role of cAMP has been the subject of an excellent monograph (Robison et al., 197 1 ) and reviewed by Greengard et al. (1972). Paul et al. (1970) found that during mania, cAMP is elevated in the urine during the manic phase but reduced during the depressive phase of the disease. Although this has been confirmed by Abdullah and Hamadeh (1970), and more recently reconfirmed in a study by Hamadeh et al. (1974) in which clinical improvement was associated with a significant rise in urinary cAMP in a group of endogenous depressives but did not change in urine of reactive depressives, not all investigators have been able to confirm these findings. Thus, Robison et al. (1970) and Brown and co-workers (1972) were unable to find a relationship between the mood changes found in manic-depression and cAMP excretion. Furthermore, Eccleston and colleagues ( 1970b) showed that changes in the urinary concentration of cAMP were closely related to the total amount of exercise of the patient. This suggests that the depressed patient has a low cAMP excretion because of behavioral depression and lack of exercise; the converse is the case in mania. Urinary CAMP levels are unchanged in periodic catatonia (Perry et al., 1973) ; these
368
I).
E. LEONARD
investigators found that, irrespective of whether the patients exhibited stupor or excitement during the psychotic phase, the urinary cAMP level was unchanged. Nevertheless, some investigators still insist that a meaningful relationship exists between the symptomatology of affective disorders and the excretion of cAMP (Paul et al., 1971). The interest of pharmacologists in the possible role of cAMP in depressive illness has recently been aroused by the discovery that the clinical efficacy of a series of phenothiazines is correlated with their ability to inhibit the adenylcyclase system, which leads to the production cAMP (Uzunov and Weiss, 1971, 1972). Furthermore, in a study of the effects of a series of antianxiety and other centrally acting drugs on the activity of CAMP-phosphodiesterase in the brain it was found that a correlation existed between the anxiety-reducing properties of the drugs and their ability to inhibit the destruction of cAMP by phosphodiesterase (Beer et al., 1972). Amer and McKinney ( 1973) have reviewed the possibilities for drug development based on the cAMP system. I t would thus seem necessary to give further consideration to the possible involvement of the cAMP system in mental disease, in particular to define more precisely the involvement of clinically efficacious drugs in affecting the adenyl cyclase system in uiuo.
V. Some Biochemical Effects of Drugs Used in the Treatment of Affective Disorders
A. TRICYCLIC ANTIDEPRESSANTS The therapeutic efficacy of these drugs has been attributed to their ability to potentiate the physiological effects of NE and H T in both the central and peripheral nervous system (Sigg, 1959; Thoenen et al., 1964; Glowinski and Axelrod, 1964). Such effects have been explained as the ability of drugs such as imipramine to block the uptake of NE and H T into the nerve ending, thereby allowing the transmitters to remain at the receptor sites for a longer period (Dengler et al., 1961). The relationship between the structure of the tricyclic antidepressants, their effects on amine uptake mechanisms, and their clinical efficacy has been considered recently by Bopp and Biel (1974). I t has been shown that synaptosomes will concentrate both NE and HT against a concentration gradient (Davis et al., 1968a), and such a finding provides evidence in favor of an active transport mechanism for the reuptake of NE and H T from the postsynaptic receptor site. The tricyclic antidepressants apparently inhibit the uptake of these amines in a competitive manner (J. M. Davis, R. W. Colburg, and D. Robinson, un-
NEUROCHEMICAL ASPECTS OF DEPRESSION
369
published data, 1968, cited by Davis, 1970; Kannengiesser et al., 1973). Such a view is supported by the in viuo studies in which it has been shown that drugs such as imipramine increased the concentration of labeled normetanephrin after a pulse injection of labeled NE (Glowinski and Axelrod, 1965, 1966). I t has also been found that the urinary excretion of O-methylated amine metabolites is increased following the administration of imipramine (Schildkraut, 1965), which is additional support for the view that the drug blocks the reuptake of NE into the presynaptic nerve terminal. Further evidence has been provided by the study of Rosenblatt and Chanley (1974), who found a correlation between the inhibition of NE uptake into peripheral sympathetic nerve endings and improvement in the clinical state of depressed patients after the administration of imipramine. However, there is some dispute as to whether amitriptyline owes its therapeutic effect primarily to an ability to inhibit the reuptake of NE. Thus Schildkraut et al. (1969a, 1972) and Stille (1968) could find no evidence to suggest that this drug inhibits the NE reuptake mechanism. Schildkraut and co-workers (1972) suggested that amitriptyline acted primarily by decreasing the deamination of the amine and also its synthesis. Serotonin metabolism appears to be decreased in man following the administration of imipramine or amitriptyline. Thus Post and Goodwin (1974) found that these tricyclics reduced the accumulation of HIAA in the CSF of depressed patients after the blockade of the efflux of the metabolite by probenecid; HVA accumulation has unaffected. Bowers (1974) also found that amitriptyline decreased serotonin turnover in endogenous depressives; this effect was not due to a decreased central availability of tryptophan. I t has been suggested that, as the blood platelet membrane is structurally and functionally similar to the nerve membrane, it should be possible to use the platelet as an in uitro model in studying the effect of antidepressant drug. Thus some investigators have shown that the uptake of HT into the platelets of patients who have been treated with imipramine is reduced compared to nonimipramine-treated patients (Davis et al., 1968b). Recently, Hamberger and Tuck (1973) showed that the uptake of serotonin and NE into rat brain slices was inhibited when the slices were incubated in plasma obtained from patients who had been treated with antidepressants. This appears to be a particularly useful method for studying the action of antidepressants in an in uitro system. However, while this may be evidence that imipramine and related antidepressants has a generalized effect on membranes through which amines can pass by an active transport mechanism, it has now been realized that great caution needs to be exercised in extrapolating results obtained from the in vitro platelet studies to in viuo effects in the brain. The steady-state concentrations of the biogenic amines do not appear
370
B. E. LEONARD
to be appreciably altered following the acute administration of tricyclic antidepressants. However, there is good evidence that the turnover of these amines is affected. Thus Glowinski and Axelrod (1966) and Schanberg et al. (1967) showed that the disappearance of 3H-labeled NE from the brain following its intracisternal administration is reduced after the acute administration of imipramine. However, the effects of the tricyclic compounds on the turnover varies depending upon their structure. Thus imipramine was found to reduce the rate of disappearance of labeled HT whereas desmethylimipramine ( D M I ) had no effect. In a detailed study of the effects of imipramine, DMI, amitriptyline, and nortriptyline on the rate of synthesis of DA, NE and HT from labeled tyrosine and tryptophan, respectively, Schubert et al. (1970) found that the monomethylated compounds ( D M I and nortriptyline) primarily reduced the rate of synthesis of NE; this suggested that these drugs selectively reduce the turnover of this amine, whereas the dimethylated antidepressants (imipramine and amitriptyline) reduced the turnover of serotonin. I n an attempt to define the mode of action of the tricyclic antidepressants more exactly, Carlsson and co-workers (1969) investigated the effects of a number of antidepressant drugs on the reserpine-resistant uptake mechanism in both central and peripheral neurons. I n this study it was found that the compound 4a-dimethyl-m-tyramine (H77/77) and 4-methyl-a-ethyl-mtyramine (H75/ 12) were fairly specific in depleting catecholamines and H T from their respective neurons in the central nervous system. Furthermore, Carlsson and co-workers found that these compounds were taken up into the nerve ending by the amine active transport mechanism and then acted on those storage compartments that are normally resistant to the depleting action of the reserpine; presumably the reduction of the amine content of these storage vesicles was due to a “false-transmitter” type of displacement. Carlsson and colleagues (1969) found that DMI and nortriptyline were more potent in blocking the depleting action of H77/77 than either imipramine or amitriptyline whereas the converse was the case regarding the blockade of the depletion of HT by H75/12. It seems likely that these antidepressants produce their effects by reducing the uptake of the tyramine derivative into the neuron. These results therefore substantiate the findings of Schubert et al. (1970) and suggest that the therapeutically useful antidepressants have a fairly specific action on the physiologically active pool of NA and HT. But despite the general agreement which seems to have been reached over the mode of action of most tricyclic antidepressants, the tricyclic compound iprindole does not apparently affect the amine reuptake mechanism (Gluckman and Baum, 1969; Lemberger et al., 1970; Lahti and Maickel, 1971), but is an effective antidepressant (Hicks, 1965; McClatchey et al., 1967; Ayd, 1969; Rickels et al., 1973). So far, the mechanism of action
NEUROCHEMICAL ASPECTS OF DEPRESSION
37 1
of iprindole is obscure; there is no evidence that, after acute administration, this drug has any effect as the in vivo synthesis of brain monoamines from their tritiated amino acid precursors (Leonard and Kafoe, 1975). Recently, the tetracyclic compound mianserine, which can be considered to be structurally related to the tricyclic antidepressants, has been shown to have antidepressant properties (Fell et al., 1973; Itil et al., 1972). Mianserine does not affect the uptake mechanism for NE and serotonin (Leonard, 1974) and unlike the tricyclic antidepressants of the imipramine type, it increases the turnover of brain NE and to a lesser extent dopamine and serotonin, following acute administration (Kafoe and Leonard, 1973). The discovery of these antidepressants with a different mechanism of action to antidepressants of the imipramine type serve to emphasize the need for caution in using any one parameter as an index of potential antidepressant activity in this series of compounds. Lahti and Maickel (1971 ) have vindicated this by showing that the blockade of the uptake of tritiated NE by the mouse heart does not correlate either with the NE potentiating effect of the drug or its clinical efficacy as an antidepressant. The possibility remains that the tricyclic antidepressants act as monoamine oxidase inhibitors (MAOI’s), thereby producing their mood-elevating effects by causing a rise in the amine concentration at the postsynaptic receptor site. However, it is now well established that, in doses that have a profound effect on the turnover of the monoamines, none of the tricyclic antidepressants so far investigated have any significant M A 0 1 activity. The main criticism of all the experimental studies which have been undertaken so far is that they have been made following the acute administration of the drug. It is well established that the clinical efficacy of the tricyclic antidepressants takes up to 14 days to become apparent and therefore the significance of extrapolating from the acute experiments to clinical effects, may be of limited relevance. Alpers and Himwich ( 1972) administered imipramine to rats for up to 10 days and this resulted in considerable changes in brain monoamine metabolism; the steady-state concentration of HIAA in the pons-medulla region and the H T concentration in the midbrain and pons-medulla region were reduced. Furthermore, the concentration of DA in the striatum was reduced after chronic administration of the drug while that of NE was unaffected. This study confirms the results of the investigation of the chronic effect of imipramine by Schildkraut et al. (1970). These results are in marked contrast to the acute effects of this drug, where several investigators have established that imipramine affects the steady-state levels of the monoamine only when administered in higher doses than those used by Alpers and Hirnwich (1972). Furthermore, after acute administration no effect on brain DA metabolism has been reported. This study helps to stress the need for investigations to be undertaken into the long-term effects
372
B. E. LEONARD
of centrally acting drugs on amine metabolism particularly when it is intended that the drugs should be administered to man for a prolonged period. Whenever possible, such studies should be undertaken in vivo, particularly at this time when there appear to be a growing acceptance of the view that antidepressant drugs, by definition, act by inhibiting the reuptake of monoamines into nervous tissue. With the discovery of tricyclic and tetracyclic antidepressants such as iprindole and mianserine, respectively, which do not affect amine uptake mechanisms in viuo, such a restricted view of the mechanism of action of antidepressants must be challenged. B. MONOAMINE OXIDASE INHIBITORS A change in the activity of the enzymes in the metabolism of the monoamines could play some role in the disturbances in brain amine metabolism which may underlie the different types of depression. This hypothesis has some support from studies of manic depressives and endogenous (unipolar) depressives. Thus, Dunner et al. (1971) found that C O M T activity was reduced in the erythrocytes of unipolar depressives and only slightly reduced with erythrocytes of bipolar depressives. There is no evidence that dopamine p-hydroxylase (which converts dopamine to NE) activity is altered in patients with endogenous depression (Shopsin et d., 1972). I n contrast, bipolar (manic depressive) patients showed a greater reduction in platelet M A 0 activity than monopolar depressives (Biegel and Murphy, 1971) . Clearly this cannot be an explanation for the underlying biochemical lesion as one would postulate that an increased activity of these enzymes in the brain would be required for there to be an effective reduction in the concentration of the transmitters. The findings of Nies and co-workers (1971), that the platelet M A 0 activity was significantly higher in a large heterogeneous group of depressed patients than in a group of normal subjects matched for age, offers some clinical evidence in support of the hypothesis that M A 0 activity is connected with the etiology of the disease. However, as yet there is no evidence to suggest that brain monoamine oxidase activity is abnormal in patients suffering from depression. As the name implies, the MAOI’s act by inhibiting MAO, the enzyme concerned in the oxidative deamination of the catecholamines and HT intraneuronally. Inhibitors of this type have been the subject of a major review recently (Sandler and Youdim, 1972). After the MAOI’s have been administered, the steady state levels of NE, DA, and HT in the brain increase and the concentrations of the deaminated metabolites decrease ; the O-methylated metabolites of NE also increase as a result of the increased concentration of the amine at the postsynaptic receptor site and its subsequent destruction by COMT. I n clinical studies, MAOI’s have been found to decrease
NEUROCHEMICAL ASPECTS O F DEPRESSION
373
the urinary excretion of deaminated metabolites (Sjoerdsma et al., 1958). The turnover and the rate of synthesis of both NE and HT has been found to decrease after the administration of a MAOI; this has been explained in terms of end-product inhibition of the rate-limiting hydroxylation reaction (Neff and Costa, 1968). The clinical use of the MAOI’s has been limited by their detrimental side effects which result when the patient ingests foods rich in monoamines, particularly tyramine. This may result in a fatal hypertensive crisis. However, recent studies have shown that the distribution of the six isoenzymes of M A 0 in different types of tissue varies (Sandler and Youdim, 1974). This has led to the suggestion that a drug which inhibits the activity of the isoenzymes which are specific for different regions of the brain may provide a therapeutic agent which does not suffer the usual disadvantages found with the conventional MAOI’s. There is now clear evidence that there are two distinct forms of M A 0 in mammalian brain (Johnson, 1968; Fuller, 1972; Squires, 1972). Type A is sensitive to the inhibitor clorgyline and oxidatively deaminates HT, NE, and tyramine, but not phenylethylamine. I n contrast type B is resistant to clorgyline, oxidizes phenylethylamine and tyramine, but not H T and NA, and is inhibited by deprenyl. These findings suggest that the development of specific inhibitors of brain M A 0 may revive the interest of clinicians in a pharmacological approach to the treatment of depression, which was formerly limited by the discovery of a relatively few, though serious, side effects that resulted from the patients eating amine-rich foods. In their stimulating review, Sandler and Youdim (1972) suggested that M A 0 could be an important factor in the control of the action of the monoamines at the receptor sites, not merely a means whereby an excessive intraneural amine concentration is destroyed. This concept therefore poses the possibility that M A 0 is more than a crude intracellular disposer of waste monoamine, the reuptake mechanism acting as the primary system of defense against amine excess at the receptor site. Attempts to relate the behavioral changes produced by the MAOI’s to alterations in the metabolism of specific amines have been inconclusive, all three amines having been cited at some time as the causative factor in the elevation of mood. Furthermore, it is still uncertain whether these amines alone are responsible for the moodelevating effects of the MAOI’s because it has been found that such derivatives of tyrosine as octopamine are increased, in peripheral nervous tissue at least, following M A 0 inhibition (Murphy, 1972) . The possibility thus remains that amines other than the catecholamines and H T may play a role in the beneficial effects of MAOI’s in the treatment of depression. I t is not without interest that phenylethylamine levels in the urine have been found to decrease during the depressive phase and increase during the manic phase in a group of bipolar depressives (Fischer et al., 1972).
374
B. E. LEONARD
C . AMPHETAMINES Because of the rapid stimulant and euphoriant effects which they produce, the amphetamines D-amphetamine and methylamphetamine have been used in the treatment of relatively minor forms of depression (reactive depression). These drugs have only limited clinical effectiveness in the treatment of most types of depression. Indeed some depressed patients were found to experience a dysphoria when given the drug (Klerman, 1972). I t is now generally agreed that these drugs produce their stimulant effects by releasing catecholamines and inhibiting their reuptake into the neuron (Stein, 1964; Glowinski and Axelrod, 1965), thereby reducing the steady-state concentration of the amines and increasing their concentration at the receptor site. The effects of these amphetamines on the steady-state levels of H T are much less marked (Leonard and Shallice, 1971) ; the release of tritiated serotonin from serotoninergic neurons also appears to be quite insensitive to the effect of amphetamine (Azzaro and Rutledge, 1973). However, some investigators have found that D-amphetamine decreased the uptake of H T into brain slices (Roos and Renyi, 1967). The effect of the amphetamines on the concentration of catecholamines appears to be dose dependent; low doses (below 1 mg,/kg) in the rat raise the amine concentration whereas higher doses, which produce all the signs of behavioral stimulation, reduce the concentration of the catecholamines (Leonard and Shallice, 1971; Leonard, 1973) . The increased release and decreased reuptake of the amines leads to an increased tissue and urinary concentrations of normetanephrin. I t has been postulated by some investigators that amphetamine produces its stimulant effect by acting directly on noradrenergic receptors in the brain. This seems unlikely, however, as it is well established that if catecholamine synthesis is blocked by the administration of a-methyl-p-tyrosine then amphetamine no longer has a stimulant effect (Weisman et al., 1966; Hanson, 1967; Randrup and Munkrad, 1967). I t has been suggested that the stimulant and antidepressant effects of this drug can therefore be entirely attributed to the ability to release the catecholamines from the physiologically labile amine pool in the brain. This account would be incomplete without mention of the introduction of the chloramphetamines p-chloro- and p-chloromethylamphetamine as antidepressants (see van Praag e t al., 1968; Korf and van Praag, 1973). The precise mechanism of action of these drugs still await elucidation, but it would appear that their main effect is to reduce the concentration of serotonin at the receptor site (Sanders-Bush and Sulser, 1970; Fuller et al., 1973; Wong et al., 1973) possibly by inhibiting tryptophan hydroxylase activity (Costa et al., 1971). Snyder and colleagues (1970) have suggested that the L-isomer of am-
NEUROCHEMICAL ASPECTS OF DEPRESSION
375
phetamine might be useful in the treatment of Parkinson’s disease. Thus it has been found that, unlike the D-isomer which causes pronounced stimulation, L-amphetamine has only a weak NE-releasing effect but is equiactive with the D-isomer in reducing the reuptake of DA into striatal neurons. An in vitro system was used in these studies by Snyder et al. (1970) ; in uivo, it has been shown that D-amphetamine is more effective in increasing the concentration of striatal HVA than the L-isomer (Jori et al., 1973). This, once again, emphasizes the disparity between results obtained in viuo from those obtained in vitro. Nevertheless, the possibility arises that while the amphetamines only have a very limited use in psychiatric medicine, some may be useful in the treatment of Parkinsonism. D. LITHIUM Many investigators consider lithium to be the drug of choice for manic depression, but, owing to the high prevalence of toxic symptoms that can occur, it is generally administered only to hospitalized patients. I n a doubleblind trial, lithium was shown to produce an 80% improvement in bipolar manic depressives, but only an improvement in 30% of the patients suffering from unipolar depression (Murphy et al., 1971) . The clinical effectiveness of the lithium salts in the treatment of manic depression stimulated investigations into the effects of this substance on brain monoamine metabolism. I t was found that lithium salts increased the rate of decrease of NE from the brain after the synthesis of this amine had been blocked by a-methyl-ptyrosine (Corrodi et al., 1967) and increased the rate of disappearance of decreased following lithium therapy (HaSkovec and Rysinek, 1967). I t is intracisternally administered 3H-labeled NE from the brain (Schildkraut et al., 1969b). Such findings suggest that lithium increases the rate of NE turnover in the brain. However, the rate of decrease of H T after synthesis blockade is apparently not affected by this cation, which suggests that it was a fairly specific effect on NE metabolism. Some studies have also shown that lithium increases the uptake of 3H-labeled NE into the synaptosomal fraction of rats which have been pretreated with the drug (Colburg et al., 1968), but not all investigators have found evidence to suggest that it increases NE uptake into neurons (Schanberg et al., 1967). Nevertheless, there is circumstantial evidence which suggests that lithium does enhance the reuptake of NE into neurons. Thus the concentrations of tritiated, deaminated metabolites of 3H-labeled NE increase while that of the extraneuronal metabolite normetanephrin decreases after the administration of lithium salts to animals (Schildkraut et al., 1966). I n man, the urinary excretion of vanillylmandelic acid is increased and that of normetanephrin and metanephrin decreased following lithium therapy (HaSkovec and RySinek, 1967). It is
376
B. E. LEONARD
therefore suggested that this cation acts by potentiating the NE reuptake mechanism. I t should be emphasized that most of the studies which show that lithium has a specific effect on NE metabolism have been carried out after the acute administration of the drug. Recently, Schildkraut ( 1974b) has shown that the long-term effects of lithium treatment differ from the immediate effects; in patients the increase in NE turnover is found only at the start of treatment. I t is well established that lithium salts alter the transport of acidic substances in both the kidney and the brain (Anumonye et al., 1968), so it remains a possibility that the increased concentration of deaminated metabolites found after the intracisternal administration of NE to lithium-treated animals could be due to an effect on the reuptake of the metabolites, not on the uptake of NE. There is some evidence that rubidium is clinically effective in treatment of certain types of depression (Meltzer et al., 1969; Stolk et al., 1970). The limited effectiveness of rubidium in the treatment of manic depression could be due to the fact that this ion reduces the uptake of NE into presynaptic sites (Meltzer et al., 1969; Stolk et al., 1970). E. ELECTROCONVULSIVE SHOCK(ECT) As this is still one of the most effective treatments for severe endogenous depression (Davis, 1965; Davis et al., 1968a), it is of importance to consider what effects, if any, such treatment has on brain amine metabolism. Unfortunately, few intensive studies have been made of this form of treatment on amine metabolism. Schildkraut et al. (1967) found that ECT lowered the steady-state concentrations of NE and increased those of normetanephrin; this finding suggests that there is an increased release of NE onto the receptor sites. Kety and co-workers (1967) also found evidence to suggest that ECT increases both the synthesis and utilization of NA in the brain, which provides further support for the view that the beneficial effects of this therapy are due to an increased turnover of brain NE. Recently, Essman (1972) has also suggested that H T may be involved as a causative factor in the beneficial effects of ECT. There is always the possibility that the efficacy of ECT may be due to an effect on a disturbed balance of water and electrolyte. However, despite the detailed studies which have been carried out in depressed patients in recent years (StGm-Olsen and WeilMalherbe, 1958; see Davis, 1970), no definite conclusions can be drawn regarding the involvement of the electrolytes in the etiology of the disease. Although Coppen and his colleagues (1966) found evidence to suggest that sodium was retained to a greater extent than normal in depressed patients, it is possible to explain such changes as a consequence of the elevated cortisol levels that are generally associated with the disease.
NEUROCHEMICAL ASPECTS OF DEPRESSION
377
F. RESERPINE AND RELATED ALKALOIDS Both clinical and experimental interest in this class of drugs arises from the well documented cases in which severe depression has been found in some patients who have been given reserpine or related alkaloids for the treatment of hypertension. Such an effect appears to be dose related, the onset of depression occurring from 1 week to up to 14 months after the commencement of therapy (Lemieuz et al., 1956; Lingjaerde, 1963). The depression usually subsides after the withdrawal of the drug. I t is well established that these drugs deplete NE and HT from both central and peripheral stores. This effect seems to be due to a specific action on the storage vesicle membrane, allowing the amines to leak into the cytosol, where they are catabolized by MAO. In rodents reserpine causes marked sedation, ptosis, and hypothermia; these symptoms are often taken to represent an animal model for endogenous depression. Enthusiasm for such a model has waned recently, however, for although it has been found that most clinically efficacious tricyclic antidepressant drugs will reverse these symptoms so do a large number of drugs which are ineffective in the treatment of endogenous depression. I t is still uncertain whether it is the depletion of the catecholamines or of HT by reserpine which results in depression. Experimental evidence would support the suggestion that it is the depletion of NE which causes the symptoms; p-chlorophenylalanine treatment does not cause depression, whereas blockade of catecholamine synthesis by a-methyl-p-tyrosine does. Some investigators have also suggested that the depletion of DA may be responsible for the behavioral effects of reserpine (Everett and Wiegand, 1963; Creveling et al., 1968). Recently a number of experimental drugs have been discovered which appear to be effective antidepressants but which inhibit tryptophan hydroxylase and thereby reduce the concentrations of H T and HIAA. The compound Ro 4-6861 (van Praag et al., 1968) p-chloroand p-chloromethamphetamine fall into this category (Costa et al., 1971 ; Miller et al., 1970). The possibility thus arises that a drug need not necessarily cause a rise in one or other of the biogenic amines at the receptor sites in order to be considered as a candidate for antidepressant drug action. They could act by reducing the concentration of HT while leaving the metabolism of NE relatively unchanged.
G. STEROIDS A significant proportion of patients with Cushing’s syndrome, or those under long-term ACTH or corticosterone therapy, show mental changes such as euphoria, depression, suicidal tendencies, and even overt psychosis (Clark et al., 1952; Spillane, 1951; Trethowen and Cobb, 1952; Glaser, 1953).
378
B. E. LEONARD
Addison’s disease is also associated with symptoms of depression, anxiety, apathy, irritability, and sleep disturbances (Cleghorn, 1951) . The occurrence of such mental changes coinciding with abnormal adrenal function raises the question of the possible role of the adrenocorticoids in affective disorders. This view is substantiated by the finding that there is a correlation between changes in plasma glucocorticoid levels and the abnormal mental state. Thus 17-hydroxycorticosteroids are raised in depressed patients (Board et al., 1956, 1957), and these steroids return to normal levels as the clinical condition of the patient improves. However, other investigators could detect no change in the urinary concentrations of 17-hydroxycorticosteroids and 17-ketosteroids in depressed patients (Balfour Sclare and Grant, 1971). King (1973) determined the plasma cortisol-binding capacity of unipolar depressive patients and found that the level was significantly lower than that found in patients with the bipolar illness or in the controls. Thus the elevated total plasma cortisol levels found in some depressives are probably associated with an increased concentration of unbound cortisol. Krieger (1974) found that there was a relationship between suicidal risk in depressed patients and a high 08:30-hour plasma cortisol level. I n manic depressive patients there have been conflicting reports of the changes in plasma corticosteroid levels, some studies reporting a reduction whereas others report no change in the plasma concentration of these substances during the manic phase. However, although it has been shown that dexamethasone, which decreases the plasma corticosteroid concentration during severe depression, may have a marginal effect on the severity of the symptoms, it is generally assumed that the steroid changes are secondary to, not causative of, the depressed state. Besides mental abberations that can result from pathological changes in the pituitary-adrenal system, reports have been made of an increased incidence of depression and psychotic episodes during the premenstrual period (Dalton, 1964), in the first month after parturition (Paffenberger, 1962), and during the menopause (Bigelow, 1960). There is also the more controversial finding that severe depression can be precipitated by oral contraceptive agents (Kane, 1968; Glick and Bennett, 1972). The major difficulty which occurs in trying to assess the relationship between oral contraceptive agents and depression arises from the scarcity of carefully controlled studies using a large number of patients. This is further complicated by the difficulty in obtaining a suitable control group and also in carrying out placebo double-blind studies in order to differentiate psychological from the pharmacological effects of the drugs. These problems have been critically discussed by Weissman and Slaby (1973). Recently Adams and co-workers (1973) studied a group of 22 depressed women whose symptoms were judged to be due to the effects of oral contraceptives. I t was found that 11 of these
NEUROCHEMICAI. ASPECTS OF DEPRESSION
379
women showed biochemical evidence of an absolute deficiency of pyridoxine, the remainder had a functional insufficiency as assessed by the reduction in the activities of several pyridoxine-dependent enzymes. These investigators found that only those women with the absolute pyridoxine deficiency responded to treatment with a large dose (20 mg) of pyridoxine; placebo administration was without effect. These findings suggest that the orally administered estrogens could cause the symptoms of depression by increasing the metabolism of tryptophan through the kynurenine pathway, possibly as a consequence of an increased liver tryptophan oxygenase activity (Curzon, 1969; Rose and McGinty, 1970; Winston, 1973). A reduction in brain serotonin synthesis could therefore result from a decreased uptake of tryptophan and a reduction in L-aromatic amino acid decarboxylase activity due to the reduced pyridoxine levels. Wolf (1974) has reviewed the effect of oral contraceptive agents on tryptophan metabolism. I t thus seems that a connection may exist between the fluctuating levels of estrogens and progesterones during certain parts of the female life cycle and mental abnormalities. This has been the basis of a study into the possible connection between changes in sex hormone levels and changes in brain monoamine metabolism. Greengrass and Tonge ( 1971 ) found that the steady-state concentration of DA, NE, and HT was maximal at the time of proestrus in mice when the levels of progestrogens and estrogens are minimal, but became minimal at estrus, when the estrogen level was maximal. These investigators also found estrogenic and progestational hormones both elevate NE and DA concentrations in the cerebral cortex but decrease the concentration of these amines in the midbrain region. Only progesterone had an effect on HT metabolism; it caused a rise in the concentration of this amine in the mid- and hindbrain region (Tonge and Greengrass, 1971). Other investigators have shown that a fluctuation in the concentration of the dopamine metabolite HVA occurs during the estrus cycle of the rat, the minimum concentration occurring during the proestrus and the maximum in dioestrus (Jori and Cecchetti, 1973). Zschaeck and Wurtman (1973), in a study in which the rate of accumulation of tritiated catechols formed from labeled tyrosine was determined during different phases of the estrus cycle in the rat, found that rats killed during proestrus showed accumulation rates which were 4 times as rapid as during diestrus and more than twice as rapid as during estrus. The mechanisms whereby the sex hormones affect brain amine metabolism is uncertain. However, it has been shown that tritiated estradiol is localized in the amygdala region and in other limbic structures (Stumpf and Sar, 1971). These structures are implicated in the control of emotion, so that it is possible that the estrogens act on specific suprahypothalamic receptors in the brain and thereby specifically affect transmitter function in these areas.
380
B. E. LEONARD
I t is necessary to be cautious about extrapolating from the rodent to man, but these studies suggest that there may be a connection between the changes in the sex homone concentration during the menstrual cycle, parturition, and menopause and the mental abnormalities that may be a consequence of an altered brain amine metabolism. VI. Conclusion
The major problem facing the clinicians and neurochemists who are searching for the pathological basis of the affective disorders is the apparent heterogeneity of such disorders. This is reflected in the symptomatology of the disease, so that there appears to be little agreement among psychiatrists of the range of symptoms which should be diagnosed as “depression.” Thus although it is generally agreed that several types of depression exist, there is still no conclusion as to the classification to be employed in defining them. Inevitably the changes in the criteria over the past years for diagnosing depression, together with the introduction of effective drugs for the treatment of the disease, has increased the proportion of patients receiving outpatient treatment. This has lead to only the more severe cases being hospitalized, such as those who are refractive to drug treatment, thereby leading to somewhat atypical samples of depressed patients being considered for clinical and biochemical studies. From the neurochemical point of view, data from patients are necessarily restricted to a limited assay of biological fluids (urine, blood, CSF) , which may well give only an inexact assessment of amine metabolism in the brain. Indeed, it has recently been suggested that HIAA is primarily removed from the brain by direct diffusion into the cerebral vasculature and only a small proportion of this metabolite is removed via the CSF (Meek and Neff, 197313). Should this finding be verified, then it may be necessary to reconsider the relevance of CSF studies with regard to their usefulness in understanding brain amine metabolism. I t is also disturbing to find that the reduction in the concentrations of CSF-HVA and HIAA which occur during endogenous depression do not return to normal on recovery of the patients (Ashcroft et al., 1973a). However, in view of the differences of opinion which at present exist regarding the relevance of amine metabolites in the CSF to brain amine metabolism, it is difficult to agree with these investigators that “such findings do not support the view that depression results from a reduction in the concentration of the amines at the synaptic function.” The other principle neurochemical approach is to analyze postmortem material. Such studies are limited to autopsy material, which is generally available for analysis hours, if not days, after death. Furthermore, many
NEUROCHEMICAL ASPECTS OF DEPRESSION
38 1
parameters which must be taken into account (for example, the quantity and nature of any drugs taken before death and the precise clinical diagnosis) are often unknown factors, particularly in suicide cases. Superimposed upon these difficulties is the heterogeneity of the disease itself. This heterogeneity has been exemplified by Leonhard (1966) and Winokut and Clayton ( 1967), who clearly distinguished between bipolar and monopolar depression. They found that virtually all bipolar depressives had a case history of mania and were also particularly responsive to lithium therapy. It also appears that this heterogeneity exists not only within the field of clinical diagnosis; van Praag and Korf ( 1971) and Asberg and colleagues (1973) have shown that there are subgroups of endogenously depressed patients who show a disturbance of indoleamine metabolism while other subgroups do not. I n one survey, mania has been observed in about 5% of all the affective disorders seen in community surveys (Klerman, 1972). I n more restricted studies, however, up to 15% of hospitalized depressives had a history of mania (Murphy et al., 1971). There is now evidence that the unipolar and bipolar types of depression have a separate genetic basis (Perris, 1966). It has been found, for example, that close relatives of bipolar depressives have cyclothymic personalities and that the patients themselves have an earlier onset of the illness and a higher suicide incidence than those with the unipolar illness (Murphy et al., 1971) . These facts serve to emphasize the difficulties encountered in proposing a comprehensive theory of the affective disorders. Yet despite these difficulties, it seems reasonable to conclude that the hypothesis implicating an abnormality in brain monoamine metabolism in the causation of such diseases is the most satisfactory to date. REFERENCES Abdullah, Y . H., and Harnadeh, K. (1970). Lancet 1,378. Adams, P. W., Rose, D. P., Folkard, J., Wijnn, V., Seed M., and Strong, R. (1973). Lancet 1, 897. Aliiio, J. J. L. I., Gutierrez, J. L. A., and Iglesias, M. L. M. (1973). Int. Pharmacopsychiat. 8, 145. Alpers, H. S., and Himwich, H. E. (1972). J . Pharmacol. Exp. T h e r . 180, 531 Amer, M. S., and McKinney, G. R. (1973). Life Sci. 13, 753. Anumonye, A,, Reading, H. W., Knight, F., and Ashcroft, G. W. (1968). Lancet 2, 1290. Asberg, M., Bertilsson, L., Tuck, D., Cronholm, B., and Sjokvist, F. (1973 . Clin. Pharmacol. Ther. 14, 277. Ashcroft, G. W., Crawford, T . B. B., Eccleston, D., Sharman, 0. F., MacDougall, E. J., Stanton, J. B., and Binns, J. K. (1966). Lancet 2, 1049. Ashcroft, G. W., Blackburn, I. M., Eccleston, D., Glen, A. I. M., Hartley, W., Kinloch, N. E., Lonergan, M., Murray, L. G., and Pullar, I. A. (1973a). Psychol. M e d . 3, 319. Ashcroft, G. W., Crawford, T. B. B., Cundall, R. L., Davidson, D. L., Dobson,
382
B. E. LEONARD
J., Dow, R. C., Eccleston, D., Loose, R. W., and Pullar, I. A. (1973b). Psychol. M e d . 3, 326. Ayd, F. J. (1969). Dis. N e r v . Syst. 30, 818. Azzaro, A. J., and Rutledge, C. D. (1973). Biochem. Pharmacol. 22, 2801. Baldessarini, R. J. (1972). Annu. R e v . M e d . 23, 343. Balfour Sclare, A., and Grant, J. K. (1971). Scot. M e d . J . 16, 224. Beer, B., Chassin, M., Clody, D. E., Vogel, J. R., and Horovitz, Z. P. (1972). Science 176, 428. Biegel, A., and Murphy, D. L. (1971). Amer. 1.Psychiat. 128, 688. Bigelow, N. (1960). I n “American Handbook of Psychiatry,” p. 540. Basic Books, New York. Bloom, F. E., and Giarman, N. J. (1968). Annu. Rev. Pharmacol. 8,229. Board, F., Persky, H., and Hamburg, D. A. (1956). Psychosom. M e d . 18, 342. Board, F., Wadeson, R., and Persky, H. (1957). A M A Arch. Neurol. Psychiat. 78, 612. Bond, P. A., Jenner, F. A,, and Sampson, G. A. (1972). Psychol. M e d . 2, 81. Bopp, B., and Biel, J. H. (1974). Life Sci. 14, 415. Bourne, H. R., Bunney, W. E., Colburn, R. W., Davis, R. W., Davis, J. N., Shaw, J. M., Shaw, D. M., and Coppen, A,, ( 1968). Lancet, 2,805. Bowers, M. B. (1970). Neuropharmacology 9, 599. Bowers, M. B. (1974). Clin. Pharmacol. Ther. 15, 167. Brown, B. L., Salway, T. G., Albano, T. D. M., Hullin, R. P., and Ekins, R. P. (1972). Brit. J . Psychiat. 120, 405. Bunney, W. E., and Davis, J. M. (1965). Arch. Gen. Psychiat. 13, 483. Bunney, W. E., Davis, J. M., Weil-Malherbe, H., and Smith, R. R. B. (1967). Arch. Gen. Psychiat. 16, ,448. Bunney, W. E., Murphy, D. L., Brodie, H. K. H., and Goodwin, F. K. (1970). Lancet 1, 352. Cade, J. F. J. (1949). M e d . J . Aust. 2, 349. Carlsson, A. (1964). Progr. Brain Res. 8, 9. Carlsson, A., Fuxe, K., Hamberger, B., and Malmfors, T. (1964). Brit. J . Pharmacol. 36, 18. Carlsson, A,, Corrodi, H., Fuxe, K., and Hokfelt, T. (1969). Eur. J . Pharmacol. 5, 357. Carlsson, A., BCdard, P., Lindqvist, M., and Magnusson, T. (1972). Biochem SOC. Symp. No. 36, Ed. R. M. S. Smellie, p. 17. Carroll, B. J. (1971). Clin. Pharmacol. Ther. 12, 743. Carroll, B. J., Mowbray, R. M., and Davies, B. M. (1970). Lancet 1,967. Clark, L. D., Bauer, W., and Cobbs, S . (1952). N . Engl. J . M e d . 246, 205. Cleghorn, R. A. (1951). Can. M e d . Ass. J . 65,449. Colburg, R. W., Goodwin, F. K., Murphy, D. L., Bunney, W. E., and Davis, J. M. (1968). Biochem. Pharmacol. 17, 957. Coppen, A. (1972). J . Psychiat. Res. 9, 163. Coppen, A., Shaw, D. M., and Farrell, J. P. (1963). Lancet 1, 79. Coppen, A,, Shaw, D. M., and Malleson, A. ( 1965). Brit. J . Psychiat. 111, 105. Coppen, A., Shaw, D. M., Malleson, A,, and Costain, R. (1966). Brit. M e d . J . 1, 71. Coppen, A., Shaw, D. M., Herzberg, B., and Maggs, R. ( 1967). Lancet 2, 1178. Corrodi, H., Fuxe, K., Hokfelt, T., and Schou, M. (1967). Psychopharmacologia 11, 345. Costa, E., and Meek, J. L. (1974). Annu. R e v . Pharmacol. 14, 491.
NEUROCHEMICAL ASPECTS OF DEPRESSION
383
Costa, E., Naimzada, K. M., and Revuelta, A. (1971). Brit. J . Pharmacol. 43, 570. Crane, G. E. ( 1 9 5 6 ) . J . Neru. M e n t . Dis. 124, 322. Crane, G. E. ( 1 9 5 7 ) . Psychiat. Res. R e p . A m e r . Psychiat. Ass. 8, 142. Creveling, C. R., Daly, J. W., Tokuyama, T., and Witkop, B. ( 1 9 6 8 ) . Biochem. Pharmacol. 17, 65. Curzon, G. ( 1 9 6 9 ) . Brit. J . Psychiat. 115, 1367. Dalton, K. ( 1964). “The Pre-Menstrual Syndrome.” Thomas, Springfield, Illinois. Davis, J. M. ( 1 9 6 5 ) . Arch. G e n . Psychiat. 13, 552. Davis, J. M. ( 1 9 7 0 ) . Znt. J . Neurobiol. 12, 145. Davis, J. M., Colburg, R. W., Murphy, D. L., and Bunney, W. E. (1968a). Sci. Proc. 124th A n n u . G e n . M e e t . A m e r . Psychiat. Ass. p. 228. Davis, J. M., Klerman, G. L., and Schildkraut, J. J. (1968b). Zn “Psychopharmacology: A Review of Progress” (D. H. Efron, ed.), p. 719. Dencker, S . J., Malm, U., and Haggendal, J. (1966). Lancet 2, 754. Dengler, H. G., Spiegel, H. E., and Titus, E. 0. (1961). Nature ( L o n d o n ) 191, 816. Dunner, D. L., and Goodwin, F. K. ( 1 9 7 2 ) . A r c h . G e n . Psychiat. 26, 364. Dunner, D. L., Cohn, C. K., and Gershon, E. S. (1971). Arch. G e n . Psychiat. 25, 34. Eccleston, D., Loose, R., Pullar, I. A., and Sugden, R. F. (1970a). Lancet 1, 612. Eccleston, D., Ashcroft, G. W., Crawford, T. B. B., Stanton, J. B., Wood, D., and McTurk, P. H. (1970b). 1.Neurol., Neurosurg. Psychiat. 33, 269. Engelman, K., and Sjoerdsma, A. ( 1 9 6 6 ) . Circ. Res. 18, 1104. Essman, W. B. ( 1 9 7 2 ) . Zn “Workshop on ‘Recent Advances in the Psychobiology of ECT.’ ” Puerto Rico. Everett, G. M., and Wiegand, R. G. ( 1 9 6 3 ) . Proc. Znt. Pharmacol. M e e t . , l s t , 2962 Vol. 8, p. 85. Fell, P. I., Quantock, D. C., and van der Burg, W. J. (1973). Eur. 1. Clin. Pharmacol. 5, 166. Fischer, E., Spatz, H., Saavedra, J. M., Reggiani, H., Miro, A. H., and Heller, B. ( 1 9 7 2 ) . Biol. Psychiat. 5, 139. Fraser, A,, Pandy, G. N., and Mendels, J. ( 1 9 7 3 ) . A r c h . G e n . Psychiat. 29, 528. Fuller, R. W. ( 1 9 7 2 ) . A d o a n . Biochern. Psychopharmacol. 5, 339. Fuller, R. W., Snoddy, H. D., Roush, B. W., and Mollowy, B. B. (1973). Neuropharniacology 12, 33. Fuxe, K., and Ungerstedt, V. (1967). J . Pharm. Pharmacol. 19, 335. Fuxe, K., Hokfelt, T., and Ungerstedt, V. ( 1 9 7 0 ) . I n t . Rev. Neurobiol. 13, 93. Garelis, E., and Sourkes, T. L. ( 1 9 7 3 ) . J . Neurol., Neurosurg. Psychiat. 36, 625. Glaser, G. H. ( 1 9 5 3 ) . Psychosom. M e d . 15, 280. Glick, I. D., and Bennett, S. E. (1972). I n “Psychiatric Complications of Medical Drugs” (R. I. Shader, ed.), p. 295. Raven Press, New York. Glowinski, J., and Axelrod, J. (1964). Nature ( L o n d o n ) 204, 1318. Glowinski, J., and Axelrod, J. ( 1 9 6 5 ) . J . Pharmacol. E x p . T h e r . 149, 43. Glowinski, J., and Axelrod, J. ( 1 9 6 6 ) . Pharmacol. R e v . 18, 775. Glowinski, J., and Baldessarini, R. J. ( 1 9 6 6 ) . Pharmacol. R e v . 18, 1201. Gluckman, M. I., and Baum, T. (1969). Psychopharmacologia 15, 169. Goodwin, F. K., Brodie, H. K. H., Murphy, D. I,., and Bunney, W. E. ( 1 9 7 0 ) . Lancet, 1, 908. Green, A. R., and Curzon, G. ( 1 9 7 0 ) . Riochetrz. Pharmacol. 19, 2061.
384
n.
E. LEONARD
Greengard, P., McAfee, P. A., and Kebabian, J. W. (1972). Advan. Cyclic Nucleotide Res. 1, 337-355. Greengrass, P. M., and Tonge, S. R. (1971). J. Pharrn. Pharmacol. 23, 897. Hamadeh, K., Holmes, H., Hartman, G. C., and Parke, D. V. (1974). Biochem. Soc. Trans. 2, 459. Hamberger, B., and Tuck, J. R. (1973). Eur. J. Clin. Pharmacol. 5, 229. Hanson, L. C. F. (1967). Psychopharrnacologia 10, 289. Harris, T. H. (1957). Arner. J . Psychiat. 113, 950. HaSkovec, L., and RyHinek, K., (1967). Psychopharrnacologia 11, 18. Hicks, J. T. (1965). Ill. M e d . J . 128, 622. Itil, T. M., Polvan, N., and Hsu, W. (1972). Curr. Ther. Res. 14, 395. Iversen, L. L. (1974). Biochern. Pharrnacol. 23, 1927. Johnson, J. P. (1968). Biochem. Pharrnacol. 17, 1285. Jori, A., and Cecchetti, G. (1973). J . Endocrinol. 58, 341. Jori, A., Dolfini, E., Tognoni, G., and Garattini, S. (1973). J. Pharrn. Pharrnacol. 25, 315. Kafoe, W. F., and Leonard, B. E. (1973). Arch. I n t . Pharrnacodyn. Ther. 206, 389. Kane, F. J. (1968). Arner. J . Obstet. Gynecol. 102, 1053. Kannengiesser, M. H., Hunt, P., and Raynaud, J. P. (1973). Biochern. Pharrnacol. 22, 73. Kety, S. S.,Javoy, F., Thierry A.-M., Julou, L., Glowinski, J. (1967). Proc. Natl. Acad. Sci. U.S. 58, 1249. Kielholz, P. (1968). Deut. M e d . Wochenschr. 93, 701. Kiloh, L. G., Ball, J. R., and Garside, R. F. (1962). Brit. M e d . J . 1, 1225. King, D. J. (1973). Psychol. Med. 3, 53. Klerman, G. L. (1972). J . Psychiat. R e f . 9, 253. Kline, N. S., Sacks, W., and Simpson, G. M. (1964). Amer. J . Psychiat. 121, 379. Knapp, S., and Mandell, A. J. (1973). In “Serotonin and Behaviour” (J. Barchas and E. Bsdui, eds.), p. 61. Academic Press, New York. Korf, J., and van Praag, H. M. (1970). Psychopharrnacologia 18, 129. Korf, J., and van Praag, H. M. (1973). J. Clin.Pharrnacol. Krieger, G. (1974). Dis. Nerv. Syst. 35, 237. Kuhn, R. (1958). Arner. J. Psychiat. 115, 459. Lahti, R. A., and Maickel, R. P. (1971). Biochern. Pharmacol. 20, 482. Lemberger, L., Sernatinger, E., and Kuntzrnan, R. ( 1970). Biochern. Pharmacol. 19, 3021. Lemieux, G., Davignon, A., and Genest, J. (1956). Can. M e d . Ass. J . 74, 522. Leonard, B. E. (1973). Proc. Congr. Hung. Pharmacol. Soc., Ist, p. 37. AkadCmiai Kiad6, Budapest. Leonard, B. E. (1974). Psychopharrnacologia 36,221-236. Leonard, B. E., and Kafoe, W. F. (1975). Biochern. Pharrnacol. (in press). Leonard, B. E., and Shallice, S. A. (1971). Brit. J . Pharmacol. 41, 198. Leonhard, K. ( 1966). “Aufteilung des Endogenen Psychosen,” 3rd ed. Berlin. Lichtensteiger, W., Hutzner, U., and Langemann, H. (1967). J . Neurochern. 14, 489. Lingjaerde, P. S. (1963). Acta Psychiat. Scand. 39, Suppl. 170, 1. Loomer, H. P., Saunders, T . C., and Kline, N. S. (1957). Psychiat. Res. Rep. 8, 129. Maas, J. W., and Landis, D. H . (1968). J . Pharrnacol. E x p . Thera. 163, 147.
NEUROCHEMICAL ASPECTS OF DEPRESSION
385
Maas, J. W., Fawcett, J. A,, and Dekirmenjian, H. (1968). Arch. Gen. Psychiat. 19, 129. Maas, J. W., Fawcett, J. A., and Dekirmenjian, J. (1972). Arch. Gen. Psychiat. 26, 252. McClatchey, W. T., Moffat, T., and Irvine, G. M. (1967). J. T h e r . Clin. Res. 1, 13. Mandell, A. J., Knapp, S., and Hsu, L. L. (1974). Life Sci. 14, 1. Matussek, N., Benkert, O., Schneider, K., Otten, H., and Pohlmeier, H. (1970). Arneim.-Forsch. 20, 934. Meek, J. L., and Neff, N. H. (1973a). Neuropharmacology 12, 497. Meek, J. I,., and Neff, N. H. (1973b). J. Pharmacol. E x p . T h e r . 184, 570. Meltzer, H. L., Taylor, R. M., Platman, S. R., and Fieve, R. E. (1969). Nature ( L o n d o n ) 223, 321. Mendels, J., Frazer, A,, Fitzgerald, R. G., Rarnsey, T . A., and Stokes, J. (1972). Science 175, 1380. Miller, F. P., Cox, R. H., Snodgrass, W. D., and Maickel, R. P. (1970). Biochem. Pharmacol. 19, 435. Murphy, D. L. (1972). Amer. J . Psychiat. 129, 5 5 . Murphy, D. L., Goodwin, F. K., and Bunney, W. E. (1971). Znt. Pharmacopsychiat. 3, 137. Neff, N. H., and Costa, E. (1968). J . Pharniacol. Exp. T h e r . 160, 40. Nies, A,, Robinson, D. S., Ravaris, C. Id., and Davis, J. M. (1971). Psychosom. M e d . 33, 470. Ollerenshaw, D. P. (1973). Brit. J . Psychiat. 122, 517. Paffenberger, R. S. (1961). J. Chronic Dis. 13, 161. Pare, C. M. B. (1963). Lancet 2, 527. Pare, C. M. B., Yeung, D. P. H., Price, K., and Stacey, R. S. (1969). Lancet 2, 133. Paul, M. I., Ditzion, B. R., Pauk, G. L., and Tanowsky, D. S. (1970). Amer. J. Psychiat. 126, 1943. Paul, M . I., Cramer, H., and Bunney, W. E. (1971). Science 171, 300. Perris, C. (1966). Acta Psychiat. Scand., Suppl. 194, 42. Perry, T . L., Hernrnings, S., Drummond, G. I., Hansen, S., and Gjessing, L. R. (1973). Amer. J. Psychiat. 130, 927. Poschel, B. P. H., and Ninternann, F. W. (1963). Life Sci. 2, 782. Post, R. M., and Goodwin, F. K. (1974). Arch. Gen. Psych&. 30, 234. Prange, A. J., Wilson, I. C., Lynn, C. W., Alltop, L. B., Stikeleather, R. A., and Raleigh, N. C. (1974). Arch. G e n . Psychiat. 32, 56. Rainer, J. D. (1966). Can. Psychiat. Ass. J . 11, GWAN Suppl., S29. Randrup, A,, and Munkrad, I. (1967). Psychopharmacologia 11, 300. Rickels, K., Chung, H. R., Csanalosi, I., Sablosky, L., and Simon, J. H. (1973). Brit. J . Psychiat. 123, 329. Robison, G. A., Coppen, A., Whyhurn, P. C., and Prange, A. J. (1970). Lancet 2, 1038. Robison, G. A,, Butcher, R. W., and Sutherland, E. W. (1971). “Cyclic AMP.” Academic Press, New York. Rodnight, R. (1961). Znt. Reri. Neurobiol. 3, 251. Roos, B. E., and Renyi, A. L. ( 1967). Life Sci. 6, 1407. Roos, B. E., and Sostrorn, R. (1969). Pharmacol. Clin. 1, 153.
386
B. E. LEONARD
Rose, D. P., and McGinty, F. (1970). Advan. Steroid Biochem. Pharmacol. 1, 97. Rosenblatt, S., and Chanley, J. D. (1974). Arch. Gen. Psychiat. 30, 456. Sanders-Bush, E., and Sulser, F. (1970). J . Pharmacol. Exp. T h e r . 175, 419. Sandler, M., and Youdim, M. B. H. (1972). Pharmacol. R e v . 24, 331. Sandler, M., and Youdim, M. B. H. (1974). I n t . Pharmacopsychiat. 9, 27. Sartorius, N. ( 1974). Pharmakopsychiat. Neuro-Psychopharmakol. 7, 76. Schanberg, S. M., Schildkraut, J. J., and Kopin, I. J. (1967). Biochem. Pharmacol. 16. 393. Schanberg, S. M., Schildkraut, J. J., Breese, G. R., and Kopin, I. J. (1968). Biochem. Pharmacol. 17, 247. Schildkraut, J. J. (1965). A m e r . J . Psychiat. 122, 509. Schildkraut, J. J. (1973). A n n u . R e v . Pharmacol. 13, 427. Schildkraut, J. J. (1974a). A n n u . Rev. M e d . 25, 333. Schildkraut, J. J. (197413). J. Nerv. M e n t . Dis. 158, 348. Schildkraut, J. J., Green, R., Gordon, E. K., and Durell, J. (1966). Amer. J. Psychiat. 123, 690. Schildkraut, J. J., Schanberg, S. M., Breese, G. R., and Kopin, I. J. (1967). Amer. J . Psychiat. 124, 600. Schildkraut, J . J., Davis, J. M., and Klerman, G. L. (1968). In “Psychopharmacology-A Review of Progress” (D. H. Efron, ed.), p. 625. Schildkraut, J. J., Dodge, G. A., and Logue, M. A. (1969a). 1. Psychiat. Res. 7, 29. Schildkraut, J. J., Logue, M. A,, and Dodge, G. A. (196913). Psychopharmacology 14, 135. Schildkraut, J. J., Winokur, A., and Applegate, C. W. (1970). Science 168, 867. Schildkraut, J. J., Draskoczy, P. R., Gershon, E. S., Reich, D., and Grab, E. L. (1972). J . Psychiat. Res. 7, 29. Schubert, J., Nyback, H., and Sedvall, G. (1970). J . Pharm. Pharmacol. 22, 136. Schiimann, H. J., and Kroneberg, G., eds. (1970). “New Aspect of Storage and Release Mechanisms of Catecholamines.” Springer-Verlag, Berlin and New York. Shaskan, E. G., and Snyder, S. H. (1970). J . Pharmacol. Exp. T h e r . 175, 404. Shaw, D. M., Camps, F. E., and Ecclestm, E. G., (1967). Br. J. Psychiat. 113, 1047. Shaw, D. M., O’Keeffe, R., MacSweeney, D. A., Brooksbank, B. W. L., Noguera, R. and Coppen A. (1973). Psychol. M e d . 3, 333. Shopsin, B., Freedman, L. S., Goldstein, M., and Gershon, S. (1972). Psychopharmacologia 27, 11. Sigg, E. B. (1959). Can. Psychiat. Ass. J. 4, 575. Sjoerdsma, A., Gillespie, L., and Udenfriend, S. ( 1958). Lancet ii, 159. Sloane, R. B., Hughes, W. W., and Haust, H. L. (1966). Can. Psychiat. Ass. 1. 11., 6. Snyder, S. H., Kuhar, M. J., Green, A. I., Coyle, J. T., and Shaskan, E. G., (1970). I n t . R e v . Neurobiol. 13, 127. Spillane, J. D. (1951). Brain 74, 72. Squires, R. F. (1972). Advan. Biochem. Psychopharmacol. 5, 355. Stein, L. (1964). Fed. Proc., Fed. Amer. Soc. Exp. Biol. 23, 836. Stein, L. (1967). Antidepressant Drugs, Proc. I n t . Symp., Ist, 2966 p. 130. Stille, G. ( 1968). Pharmakopsychiat./Neuro-Psychopharmakol. 1, 92. Stolk, J. M., Nowack, W. J., Barchas, J. D., and Platman, S. R. (1970). Science 168, 501.
NEUROCHEIMICAI, ASPECTS OF DEPRESSION
387
Strorn-Olsen, R., and Weil-Malherbe, H . ( 1 9 5 8 ) . J. M e n t . Sci. 104, 696. Stumpf, W. E., and Sar, M. (1971 ) . Proc. SOC.Exp. Biol. M e d . 136, 102. 'I'hierry, A. M., Blanc, G., and Glowinski, J. (1973). Naunyn-Schmiedebergs Arch. Pharmacol. 279, 255. Thoenen, H., Hiirlimann, A,, and Haefely, W. (1964). J. Pharmacol. Ex$. T h e r . 144, 405. longe, S. R., and Greengrass, P. M. ( 1 9 7 1 ) . Psychopharmacology 21, 374. Trcthuwen, W. H., and Cobb, S. (1952). A M A Arch. Neurol. Psychiat. 67, 283. Uzunov, P., and Weiss, B. (1971). Neuropharmacology 10, 697. Uzunov, P., and Weiss, B. (1972). Aduan. Cyclic Nucleotide Res. 1, 435. van Praag, H . M., Korf, J., van Woudenberg, F., and Kits, T . P. ( 1 9 6 8 ) . Psychopharmacologia 13, 145. van Praag, H . M., Korf, J., and Puite, J. ( 1 9 7 0 ) . Nature ( L o n d o n ) 225, 1259. van Praag, H. M., Korf, J., Dols, I,. W. C., and Schut, T. ( 1 9 7 2 ) . Psychopharmacologia 25, 14. van Praag, H . M., Korf, J., and Schut, D. ( 1 9 7 3 ) . Arch. Gen. Psychiat. 28, 827. Weir, R. I,., Chase, T. N., Ng, L. K. Y . , and Kopin, I. J. (1973). Brain Res. 52, 409. Weisrnan, A,, Koe, K. N., and Tenen, S. S. ( 1 9 6 6 ) . J. Pharmacol. Exp. T h e r . 151, 339. Weissrnan, M. M., and Slaby, A. E. (1973). Brit. J. Psychiat. 123, 513. Winokur, G. ( 1 9 7 4 ) . I n t . Pharmacopsychiat. 9, 5 . Winokur, G., and Clayton, P. ( 1 9 6 7 ) . Recent Aduan. Biol. Psychiat. 9, 35. Winston, F. ( 1 9 7 3 ) . A m e r . J. Psychiat. 130, 1217. Wolf, H. ( 1 9 7 4 ) . Scand. J. Clin. Lab. Inuest. 33, Suppl. 136. Wong, D. T., Horng, J. S., and Fuller, R. W. ( 1 9 7 3 ) . Biochem. Pharmacol. 22, 311. Zschaeck, L. L., and Wurtrnan, R. J. (1973). Neuroendocrinology 11, 144.
,.
This Page Intentionally Left Blank
SUBJECT INDEX A
Acetylcholine metabolism in cholinergic neuron, 69-140 receptor for, acetylcholinesterase and, 130-1 3 3 Acetylcholinesterase acetylcholine receptor and, 130- 133 in axon, 82-84 cytochemistry of, 71-72 in perikaryon, 77-81 in Renshaw e!ements, 108-1 12 in spinal motoneuron, 77-1 12 in synaptic transmission, 75-77 N-Acetylserotonin, assays of, 57-59 Acquisition of behavior, marihuana effects on, 338-342 S-Adenosylmethionine assay of, 44-46 biological transmethylation by, 4 1-67 clinical aspects of, 61-63 levels of, substrate effects on, 47-51 methionine loading effects on, 51-55 tissue concentrations of, 47 turnover of, 46-47 Affective disorders biogenic amines effects on, 360-367 characteristics of, 359-360 cyclic AMP affects on, 367-368 drug therapy of, 368-381 Aggression, 213-262 by cerebral chemostimulation. 232-237 cholinergic stimulation of, 232-237 factors influencing, 2 16-2 19 isolation-induced, 229-23 1, 237-241, 249, 251 neuroanatomical correlates of, 2 17-2 19 neurochemical correlation of, 220-23 1 389
neuropharmacological stimulation of, 23 7-253 neurotransmitters and, 2 13-262 shock-induced, 231, 241, 249, 251-252 types of, 2 14-2 16 Amantadine, effect on aggression, 248 Amino acids, in brain, 170-1 72 AMP, cyclic, see Cyclic AMP Amphetamines in affective disorder therapy, 374-375 effect on aggression, 247 Amygdala, aggression sites in, 223 Apomorphine, effect on aggression, 247-248 ATP: L-methionine adenosyltransferase, assay of, 59 Attention, neural model of, 263-327 Axon (s) acetylcholinesterase in, 82-84 in terminals, 84-108 as delay lines, 4-6 demyelination of, 33-34 diameter spectra of, 26-28 electrotonic coupling in pathways of, 19-20 external effects on, 18-19 as filtering systems, 6-18 functions of, 32-33 myelination in, critical diameter of, 2 9-3 2 nodes and internode spacing in, 20-26 properties and design principles of, 1-40 structure-function relations for, 20-32 as transmission line, 2-3 B
Biogenic amines, affective disorders and, 360-367
390
SUBJECT INDEX
Brain amino acids in, 170-172 cell types in, 142-144 metabolic studies, 145-147 cellular membrane potentials in, 186-1 9 1 compartmentation in, 150-151 complexity of, 142-151 enzymes in, 174-1 76 glycolysis in, 173-1 74 high-energy phosphates in, 160-1 66 intermediary metabolism in, 166-176 interstitial space of, 144-145 ion concentrations in, 147-148 ion and energy metabolism of, 141-21 1 metabolic compartmentation in, 169-170 ontogenetic and phylogenetic development of, 148-150 oxygen consumption in, 151-1 60 water metabolism in, 176-1 91 C
Central nervous system, space-time transformations in, 6-7 Cholinergic neuron acetylcholine metabolism in, 69-140 indirect studies on, 112-1 19 transmitter release of, 119-133 Cholinesterase, in acetylcholine metabolism, 69-77 Conduction, intermittent, in vertebrates and invertebrates, 7-13 Cyclic AMP, affective disorders and, 36 7-3 68 D
Demyelination, of axons, 33-34 Depression characteristics of, 359-360 drug therapy of, 368-381 neurochemistry of, 357-387 Diencephalon, aggression sites in, 226-227 Discrimination learning generalization gradients in, 301-305 neural model of, 263-327 expectation mechanism of, 3 13-3 16
DOPA, in aggression mechanism, 242-247 Dopaminergic drugs, aggressive behavior from, 242-249 E
Eating, nervous, attentional deficits and, 321-323 Electroconvulsive shock, in affective disorder therapy, 376 Electrotonic coupling, in axonal pathways, 19-20 Energy metabolism, of brain, 141-21 1 Enzymes, in brain, 174-176 Exocytosis, in synaptic vesicle discharge, 125-128 F
Fear Pavlovian extinction of, vs. learned avoidance, 297 rebound to relief of, 276-282 Frustration, in discrimination learning, 297-300 G
Glia cells, cerebral, energy metabolism in, 155-157 Glycolysis, in brain, 173-174 H
Habits, drives, rewards, motivation, and, 274-276 Habituation, hippocampus and, 305-306 Hemicholinium, autoradiography of, 112-117 Hippocampus feedback from, conditioning and, 3 19-32 1 habituation and, 305-306 Histamine, assays of, 57-59 6-Hydroxydopamine, effect on aggression, 246-247 I
Ion metabolism, of brain, 141-21 1
391
SUBJECT INDEX 1
Learning, marihuana effects on, 329-356 Lithium, in affective disorder therapy, 3 75-3 76 M
M A 0 inhibitors, effect on aggression, 244-246 Marihuana learning, memory, and, 329-356 animal studies, 330-345 human studies, 345-353 Meanings, learned, persistence of, 291-294 Memory marihuana effects on, 329-356 reinforcement acquisition effect on, 300-30 1 short-term, activity normalization and, 282-289 Mesencephalon, aggression sites in, 22 7-228 Methionine assay of, 59-61 effects on S-adenosylmethionine levels, 51-55 role in schizophrenia, 61-63 Methyl acceptors, assays of, 57-61 Methyl donor, 5'-adenosylmethionine as, 41-67 Methyl tetrahydrofolate, in transmethylation, 55-57 Monoamine oxidase inhibitors, in affective disorder therapy, 372-373 Morphine, withdrawal from, aggression from, 248 Motivation, generalization and, 309-3 10 Muricide, aggression inducement and, 231,241-242, 249-250, 252 Myelination, in axons, critical diameter of. 29-32 N
Na+-K+-ATPase,in brain, 161-163 Neurons, cerebral, energy metabolism in, 155
Neuropsychiatry, transmethylation studies relating to, 41-67 Nodes, in axons, spacing of, 20-26 Nonadrenergic drugs, aggressive behavior from, 237-242 0
Olfactory bulb, aggression sites in, 223, 226 Orienting reactions in discrimination learning, 3 11-3 13 regulation of, 316-319 P
Perikaryon, acetylcholinestase in, 77-81 Phosphates, high-energy, in brain, 160- 166 Pimozide, aggression and, 248 R
Reinforcement neural model of, 263-327 novelty and, 308-309 Relief, rebound from fear to, 276-282 Kenshaw elements, acetylcholinesterase in, 108-112 Reserpine, in affective disorder therapy, 377 S
Schizophrenia, transmethylation in, 6 1-63 Serotonereic drugs, aggressive behavior from, 251-253 Septum, aggression sites in, 226 Steroids, in affective disorder therapy, 377-380 Synaptic vesicles charging of, 121-123 discharging of, 123-128 origin of, 119-121 Synaptochemistry, of acetylcholine metabolism, 69-140
T Telencephalon, aggression sites in, 223 'Thiamine pyrophosphatase, in synapses, 117-119
392
SUBJECT INDEX
Transmethylation by S-adenosylmethionine, 41-67 assays for, 57-63 biochemical assays of, 44-57 Transmission, by axons, 2-4 Tricyclic antidepressants, in affective disorder therapy, 368-372
U Ulcers, predictability and, 27 1-272, 310-311 W
Water metabolism, in brain, 176-191
CONTENTS OF PREVIOUS VOLUMES Volume 1 Recent Studies of the Rhinencephalon in Relation to Temporal Lobe Epilepsy and Behavioral Disorders W . R. Adey Nature of Electrocortical Potentials and Synaptic Organizations in Cerebral and Cerebellar Cortex Dominick P. Purpura Chemical Agents of the Nervous System Catherine 0. Hebb Parasympathetic Neurohumors ; Possible Precursors and Effect on Behavior Carl C. Pfeiffer Psychophysiology of Vision
G . W . Granger Physiological and Biochemical Studies in Schizophrenia with Particular Emphasis on Mind-Brain Relationships Robert G. Heath
The Mechanism Hemicholiniums F . w- Schueler
of
Action
of
the
The Role of Phosphatidic Acid and Phosphoinositide in Transmembrane Transport Elicited by Acetylcholine and Other Humoral Agents Lowell E . Hokin and Mabel R. Hokin Brain Neurohormones and Cortical Epinephrine Pressor Responses as Affected by Schizophrenic Serum Edward J . Walastek The Role of Serotonin in Neurobiology
Erminio Costa Drugs and the Conditioned Avoidance Response Albert Hertz Metabolic and Neurophysiological Roles of y-Aminobutyric Acid Eugene Roberts and Eduardo Eidelberg
Studies on the Role of Ceruloplasmin in Schizophrenia S. Miirtens, S. Yallbo, and B. Melander
Objective Psychological Tests and the Assessment of Drug Effects H . J . Eysenck
Investigations in Protein Metabolism in Nervous and Mental Diseases with Special Reference to the Metabolism of Amines F . Georgi, C . G . Honegger, D. Jordan, H . P. Rieder, and M . Rottenberg
AUTHOR INDEX-SUBJECT INDEX
A UT HOR INDEX-SUB JECT INDEX
Volume 2 Regeneration of Amphibia R. M . Gaze
the Optic Nerve in
Volume 3 Submicroscopic Morphology and Function of Glial Cells Eduardo De Robertis and H . M . Gerschenfeld Microelectrode Studies of the Cerebral Cortex Vahe E . Amassian Epilepsy Arthur A. Ward, Jr.
Experimentally Induced Changes in the Functional Organization of Somatic Areas Free Selection of Ethanol of the Cerebral Cortex Jorge Mardones Hiroshi Nakahama 393
394
CONTENTS OF PREVIOUS VOLUMES
Body Fluid Indoles in Mental Illness
Volume 5
R. Rodnight Some Aspects of Lipid Metabolism in Nervous Tissue
G. R. Webster Convulsive Effect of Hydrazides: Relationship to Pyridoxine
Harry L. Williams and James A . Bain T h e Physiology of the Insect Nervous System
D . M . Vowles AUTHOR INDEX-SUB JECT INDEX
Volume 4 T h e Nature of Spreading Depression in Neural Networks
Sidney Ochs Organizational Aspects of Some Subcortical Motor Areas
Werner P. Koella Biochemical and Neurophysiological Development of the Brain in the Neonatal Period
T h e Behavior of Adult Mammalian Brain Cells in Culture R u t h S. Geiger T h e Electrical Activity of a Primary Sensory Cortex: Analysis of EEG Waves
Wa!ter J. Freeman Mechanisms for the Transfer of Information along the Visual Pathways
Koiti Motokawa Ion Fluxes in the Central Nervous System F. J . Brinley, J r . Interrelationships between the Endocrine System and Neuropsychiatry
Richard P. Michael and James L. Gibbons Neurological Factors in the Control of the Appetite
AndrB Soulairac Some Biosynthetic Activities of Central Nervous Tissue R. V . Coxon Biological Aspects of Electroconvulsive Therapy
Gunnar Holrnberg AUTHOR INDEX-SUBJECT
INDEX
Williarnina A . Himwich Substance P: A Polypeptide of Possible Physiological Significance, Especially within the Nervous System
F . Lembeck and G . Zelter Anticholinergic Psychotomimetic Agents
L. G. Abood and J. H . Biel Benzoquinolizine Derivatives : A New Class of Monamine Decreasing Drugs with Psychotropic Action A. Pletscher, A . Brossi, and K . F. Gey T h e Effect of Adrenochrome and Adrenolutin on the Behavior of Animals and the Psychology of Man
A. Hoffer AUTHOR INDEX-SUB JECT INDEX
Volume 6 Protein System
Metabolism
of
the
Nervous
Abel Lajtha Patterns of Muscular Innervation in the Lower Chordates
Quentin Bone T h e Neural Organization of the Visual Pathways in the Cat Thomas H . Meikle, Jr. and
James M . Sfirague Properties of Afferent Synapses and Sensory Neurons in the Lateral Geniculate Nucleus
P. C. Bishop
395
CONTENTS OF PREVIOUS VOLUMES
Regeneration in the Vertebrate Central Nervous System Carmine D. Clemente Neurobiology of Phencyclidine (Sernyl ) , a Drug with an Unusual Spectrum of Pharmacological Activity Edward F . Domino Free Behavior and Brain Stimulation JosB M . R . Delgado AUTHOR IXDEX-SUBJECT
INDEX
T h e Anatomophysical Basis of Somatosensory Discrimination David Bowsher, with the collaboration of Denise Albe-Fessard Drug Action on the Electrical Activity of the Hippocampus Ch. Stumpf Effects of Drugs on Learning and Memory James L. McGaugh and Lewis F . Petrinouich Biogenic Amines in Mental Illness Giinter G. Brune
Volume 7 Alteration and Pathology of Cerebral Protein Metabolism Abel Lajtha Micro-Iontophoretic Studies on Cortical Neurons K . KrnjeviC Responses from the Visual Cortex of Unanesthetized Monkeys John R . Hughes
T h e Evolution of the Butyrophenones, Haloperidol and Trifluperidol, from Meperidine-like 4-Phenylpiperidines Paul A . J , Janssen Amplitude Analysis of the Electroencephalogram (Review of the Information Obtained with the Integrative Method) Leonide Goldstein and Raymond A . Beck
Recent Development of the Blood-Brain Barrier Concept Ricardo Edstrom
A U T H O R INDEX-SUBJECT
Monoamine Oxidase Inhibitors Gordon R . Pscheidt
Volume 9
The Phenothiazine Tranquilizers: Biochemical and Biophysical Actions Paul S. Guth and Morris A . Spirtes
Development of “Organotypic” Bioelectric Activities in Central Nervous Tissues during Maturation in Culture Stanley M . Crain
Comments on the Selection and Use of Symptom Rating Scales for Research in Pharmacotherapy J . B. Wittenborn Multip!e Molecular Forms of Brain Hydrolases Joseph Bernsohn and Kevin D . Barron A U T H O R IXDEX-SUBJECT
INDEX
T h e Unspecific Intralaminary Modulating System of the Thalamus P. Krupp and M . Monnier T h e Pharmacology of Imipramime and Related Antidepressants Laszlo Gyermek
INDEX
Volume 8
Membrane Stabilization by Drugs: Tranquilizers, Steroids, and Anesthetics Philip M . Seeman
A Morphologic Concept of the Limbic Lobe Lowell E. White, J r .
Interrelationships between Phosphates and Calcium in Bioelectric Phenomena L. G. Abood
396
CONTENTS OF PREVIOUS VOLUMES
The Periventricular Hypothalamus Jerome Sutin
Stratum
of
the
Neural Mechanisms of Facial Sensation I . Darian-Smith A U T H O R INDEX-SUBJECT
Exopeptidases of the Nervous System Neville Marks Biochemical Responses to Narcotic Drugs in the Nervous System and in Other Tissues Doris H . Clouet
INDEX
Periodic Psychoses in the Light of Biological Rhythm Research F . A . Jenner
Volume 10 A Critique of Iontophoretic Studies of Central Nervous System Neurons G. C . Salmoiraghi and C. N . Stefanis Extra-Blood-Brain-Barrier Brain Structures Werner P . Koella and Jerome Sutin Cholinesterases of the Central Nervous System with Special Reference to the Cerebellum Ann Silver Nonprimary Sensory Projections on the Cat Neocortex P . Buser and K . E. Bignall Drugs and Retrograde Amnesia Albert Weissman Neurobiological Action of Some Pyrimidine Analogs Harold Koenig A Comparative Histochemical Mapping of the Distribution of Acetylcholinesterase and Nicotinamide Adenine Dinucleotide -Diaphorase Activities in the Human Brain T . Zshii and R . L . Friede
Endocrine and Neurochemical Aspects of Pineal Function Bila Mess The Biochemical Investigation of Schizophrenia in the USSR D . V. Lorousky Results and Trends of Studies in Schizophrenia J . Saarma
Conditioning
Carbohydrate Metabolism in Schizophrenia Per S . Lingjaerde The Study of Autoimmune Processes in a Psychiatric Clinic S. F. Semenou Physiological Foundations of Mental Activity N . P. Bechtereua and V . B . Gretchin A U T H O R INDEX-SUBJECT
INDEX
CUMULATIVE TOPICAL INDEXFOR VOLUMES1-10
Behavioral Studies of Animal Vision and Drug Action Hugh Brown
Volume 12
The Biochemistry of Dyskinesias C . Curzon
Pathobiology of Acute Triethyltin Intoxication R . Torack, J . Gordon, and J . Prokop
A U T H O R INDEX-SUBJECT
INDEX
Volume 11 Synaptic Transmission in the Central Nervous System and Its Relevance for Drug Action Philip B. Bradley
Drugs and Body Temperature Peter Lomax
Ascending Control of Thalamic and Cortical Responsiveness M . Steriade Theories of Biological Etiology of Affective Disorders John M . Davis
CONTENTS OF
Cerebral Protein Synthesis Block Long-Term Memory Samuel H . Barondes
PnEvIous
397
VOLUMES
Inhibitors
Molecular Mechanisms in Information Processing Georges Ungar
The Mechanism of Action of Hallucinogenic Drugs on a Possible Serotonin Receptor in the Brain J . R . Smythies, F. Benington, and R. D. Morin
The Effect of Increased Functional Activity on the Protein Metabolism of the Nervous System B. Jakoubek and B. Semiginouskj
Simple Peptides in Brain Isamu Sano The Activating Effect of Histamine on the Central Nervous System M . Monnier, R . Sauer, and A . M . Hatt Mode of Action of Psychomotor Stimulant Drugs Jacques M . van Rossum
Protein Transport in Neurons Raymond J . Lasek Neurochemical Correlates of Behavior M . H . Aprison and J . N . Hingtgen Some Guidelines from System Science for Studying Neural Information Processing Donald 0. Walter and Martin F. Gardiner AUTHOR INDEX-SUB JECT INDEX
AUTHOR INDEX-SUBJECT
INDEX
Volume 13 Of Pattern and Place in Dendrites Madge E. Scheibel and Arnold B. Scheibel The Fine Structural Localization of Biogenic Monoamines in Nervous Tissue Floyd E. Bloom Brain Lesions and Amine Metabolism Robert Y . Moore Morphological and Functional Aspects of Central Monoamine Neurons Kjell Fuxe, Tomas Hokfelt, and Urban Ungerstedt Uptake and Subcellular Localization of Neurotransmitters in the Brain Solomon H . Snyder, Michael J . Kuhar, Alan I . Green, Joseph T . Coyle, and Edward G. Shaskan Chemical Mechanisms of TransmitterReceptor Interaction John T . Garland and Jack D w e l l The Chemical Nature of the Receptor Site-A Study in the Stereochemistry of Synaptic Mechanisms J . R . Smythies
Volume 14 T h e Pharmacology Geniculate Neurons J . W . Phillis
of
Thalamic
and
The Axon Reaction: A Review of the Principal Features of Perikaryal Responses to Axon Inquiry A . R. Lieberman CO? Fixation in the Nervous Tissue Sre-Chuh Cheng Reflections on the Role of Receptor Systems for Taste and Smell John G. Sinclair Central Cholinergic Mechanism and Behavior S.N . Pradhan and S. N . Dutta The Chemical Anatomy Mechanisms: Receptors J . R. Smythies
of
Synaptic
AUTHOR INDEX-SUB JECT INDEX
Volume 15 Projection of Forelimb Group I Muscle Afferents to the Cat Cerebral Cortex Ingmar R o s i n
398
CONTENTS OF PREVIOUS VOLUMES
Physiological Pathways through the Vestibular Nuclei Victor J. Wilson
A Comparison of Cortical Functions in Man and the Other Primates R. E. Passingham and G. Ettlinger
Tetrodotoxin, Saxitoxin, and Related Substances: Their Applications in Neurobiology Martin H . Evans
Porphyria: Theories of Etiology and Treatment H . A. Peters, D. J. Cripps, and H . H. Reese
The Inhibitory Action of y-Aminobutyric Acid, A Prohahle Synaptic Transmitter Kunihiko Obata Some Aspects of Protein Metabolism of the Neuron M e i Satake Chemistry and Biology of Two Proteins, S-100 and 14-3-2, Specific to the Nervous System Blake W. Moore The Genesis of the EEG Rafael Elul Mathematical Identification of Brain States Applied to Classification of Drugs E. R. John, P. Walker, D. Cawood, M . Rush, and J. Gehrmann A U T H O R INDEX-SUBJECT
INDEX
Volume 16 Model of Molecular Mechanism Able to Generate a Depolarization-Hyperpolarization Cycle Clara Torda Antiacetylcholine Drugs : Chemistry, Stereochemistry, and Pharmacology T . D. Inch and R. W . Brimblecombe Kryptopyrrole and Other Monopyrroles in Molecular Neurobiology Donald G. Iruine RNA Metabolism in the Brain Victor E. Shashoua
S U B J E C T INDEX
Volume 17 Epilepsy and y-Aminobutyric Acid-Mediated Inhibition B. S . Meldrum Peptides and Behavior Georges Ungar Biochemical Transfer of Acquired Information S. R. Mitchell, J. M . Bsaton, and R. J . Bradley Aminotransferase Activity in Brain M . Benuck and A . Lajtha The Molecular Structure of Acetylcholine and Adrenergic Receptors: An AllProtein Model J . R. Smythies Structural Integration of Neuroprotease Activity Elena Gabrielescu O n Axoplasmic Flow Liliana L u b i h k a Schizophrenia: Perchance a Dream? 1.Christian Gillin and Richard J . Wyatt
INDEX SUBJECT
A 6 c D E
5 6 7 8 9
F O G 1
H 2 1 3 J 4
