ARF Family GTPases (Proteins and Cell Regulation)

ARF Family GTPases

PROTEINS AND CELL REGULATION Volume 1

Series Editors:

Professor Anne Ridley Ludwig Institute for Cancer Research and Department of Biochemistry and Molecular Biology University College London London United Kingdom

Dr. Jon Frampton Institute for Biomedical Research, Birmingham University Medical School, Division of Immunity and Infection Birmingham United Kingdom

Aims and Scope

Our knowledge of the ways in which a cell communicates with its environment and how it responds to information received has reached a level of almost bewildering complexity. The large diagrams of cells to be found on the walls of many a biologist’s office are usually adorned with parallel and interconnecting pathways linking the multitude of components and suggest a clear logic and understanding of the role played by each protein. Of course this two-dimensional, albeit often colourful representation takes no account of the three-dimensional structure of a cell, the nature of the external and internal milieu, the dynamics of changes in protein levels and interactions, or the variations between cells in different tissues.

Each book in this series, entitled “Proteins and Cell Regulation”, will seek to explore specific protein families or categories of proteins from the viewpoint of the general and specific functions they provide and their involvement in the dynamic behaviour of a cell. Content will range from basic protein structure and function to consideration of cell type-specific features and the consequences of disease-associated changes and potential therapeutic intervention. So that the books represent the most up-to-date understanding, contributors will be prominent researchers in each particular area. Although aimed at graduate, postgraduate and principle investigators, the books will also be of use to science and medical undergraduates and to those wishing to understand basic cellular processes and develop novel therapeutic interventions for specific diseases.

ARF Family GTPases Edited by

RICHARD A. KAHN Emory University School of Medicine, Atlanta, Georgia, U.S.A.

KLUWER ACADEMIC PUBLISHERS NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW

eBook ISBN: Print ISBN:

1-4020-2593-9 1-4020-1719-7

©2004 Springer Science + Business Media, Inc. Print ©2003 Kluwer Academic Publishers Dordrecht All rights reserved No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher Created in the United States of America

Visit Springer's eBookstore at: and the Springer Global Website Online at:

http://www.ebooks.kluweronline.com http://www.springeronline.com

Dedicated to the memory of Annette L. Boman, Ph.D. Mother, scholar, wife, mentor, athlete, cook, and much more She loved life and shared that joy with everyone We were all made better by having known her

TABLE OF CONTENTS Preface: The Arf Family Richard A. Kahn

vii

List of Contributors

xxv

1. The Arf Family Tree John M. Logsdon Jr., Richard A. Kahn

1

2. The GDP/GTP Cycle of Arf Proteins Sebastiano Pasqualato, Louis Renault, Jaqueline Cherfils

23

3. Arf and Phospholipids Paula A. Randazzo, Zhongzhen Nie, Dianne S. Hirsch

49

4. The Sec7 Family of Arf Guanine Nucleotide Exchange Factors Catherine L. Jackson

71

5. Large Arf GEFs of the Golgi Complex Paul Melançon, Xinhua Zhao, Troy K.R. Lasell

101

6. BIG1 and BIG2: Brefeldin A-Inhibited Exchange Factors for Arfs G. Pacheco-Rodriguez, J. Moss, M. Vaughan

121

7. Arf GTPase-Activating Protein 1 Dan Cassel

137

8. GIT Proteins: Arf Gaps and Signaling Scaffolds Robert Schmalzigaug, Richard Premont

159

9. Paxillin-Associated Arf Gaps Hisataka Sabe

185

10. Activation of Cholera Toxin and E. Coli Heat-Labile Enterotoxin (LT) by Arf G. Pacheco-Rodriguez, Naoko Morinaga, Masatoshi Noda, J. Moss, M. Vaughan

209

vi

11. Phospholipase D, Arfaptins and Arfophilin John H. Exton

223

12. Coat Proteins Annette Boman, Tommy Nilsson

241

13. Heterotetrameric Coat Protein-Arf Interactions M.L. Styers, V. Faundez

259

14. Arf6 James E. Casanova

283

15. A Role for Arf6 in G Protein-Coupled Receptor Desensitization Mary Hunzicker-Dunn

305

16. Arf-Like Proteins Annette Schürmann, Hans-Georg Joost

325

Preface THE ARF FAMILY RICHARD A. KAHN Department of Biochemistry, Emory University School of Medicine, Atlanta, USA

1.

INTRODUCTION TO GTPASES

What questions drive biomedical research today? Much publicity has surrounded the publication of the human genome in recent years. Certainly this marks a milestone and a database from which a tremendous amount of new insights will be learned. But the genome is linear, i.e., a onedimensional feature of all cells. I believe that what drives the majority of research today is the goal of understanding the complete wiring diagram of the eukaryotic cell. This will require much more complicated databases, in large part because the solution is four-dimensional. We need to learn not only all the pathways of metabolism and signaling, but how they are integrated. How are nucleocytoplasmic transport, protein and lipid metabolism, the cell cycle, organellogenesis, and the cytoskeleton all regulated; individually and collectively? GTPases are at least one of the keys to unraveling the layers of cell regulation required. I remember my first biochemistry course and the expectation that the cell is a bag of enzymes that we can decipher simply by knowing the concentrations of reactants, their affinities for enzymes, and the rates of the reactions. Perhaps for understanding intermediary metabolism this was useful, though today even that is doubtful. Instead, as each metabolic or signaling pathway is described we gain a better appreciation for the need to understand the connections between pathways. All cells are changing, modifying and responding to their environment and their neighboring cells, be they friend or foe. A cell’s ability to mount an appropriate response to whatever it encounters depends in large part on its ability to coordinate many different systems or pathways and synchronize them in time. Thus, cell vii

viii

KAHN

regulation and the integration of signaling pathways can be viewed as a second order system that increases a cell’s fitness. It is at these regulatory points of intersection between critical cellular processes that we usually encounter the regulatory GTPases. They have been described as molecular switches. This is unfortunate because they are so much more sophisticated and interesting than that. With multiple inputs and outputs, they are never simply on or off. Instead, they sit at junction points between essential cellular processes and direct the flow of information that is required for all the parts to work together. Just like computer software that became much more powerful when programmers learned to introduce conditional points in a program, cells became much more “fit” when they acquired the ability to use conditional regulators and effectively integrate multiple essential processes. This type of regulation is distinct from that used on a single enzyme or pathway. The rate at which phosphofructokinase converts fructose 6phosphate to fructose 1,6-bisphosphate is subject to regulation by multiple modulators (e.g., citrate, pH, ATP, ADP, fructose 2,6-bisphosphate) because this is the commitment point in glycolysis and the substrate may be required for other uses. But in this case, the product is known and only the rate is varied. In contrast, several GTPases have more than a dozen different effectors and may have a similar number of activators. Thus, not only is their level of activity (percentage bound to GTP vs. GDP) varied but so also is the response itself. It could be an alteration in lipid metabolism, cytoskeletal organization, or macromolecular complex assembly. Thus, to fully understand the functions of even one family of regulatory GTPases we will likely need to understand the means of integration used by cells to coordinate multiple aspects of cell biology.

2.

A BRIEF INTRODUCTION TO ARFS

Although Ras was known to bind guanine nucleotides, it was the discovery that Ras is a GTPase that heralded the start of what has grown into an incredibly diverse and important area of cell regulation research (Gibbs et al., 1984; Sweet et al., 1984). One is hard pressed to name an aspect of cell biology that is not directly or indirectly affected by one or more regulatory GTPase today. The Ras superfamily owes much to the heterotrimeric G proteins. Many of the original models for GTPases as signal transducers and techniques used in the field (e.g., nucleotide-binding, filter trapping assay, GEF assays, and the use of slowly hydrolyzing GTP analogs) came from work on G proteins. The three-component model (activator, G protein, and effector) for cell signaling arose from studies of hormone- and light-

THE ARF FAMILY OF REGULATORY GTPASES

ix

activated second messenger systems, adenylyl cyclase and phosphodiesterase, respectively. This debt to G protein research is nowhere more evident than for the Arfs. Before any G protein had been purified, the guanine nucleotide-binding α subunit of Gs, the direct activator of adenylyl cyclase, had been identified as a substrate for the ADP-ribosyltransferase activity of cholera toxin (Cassel and Pfeuffer, 1978) (see Chapter 10). Membranes treated with cholera toxin and [α-32P]NAD covalently incorporated radionucleotide (ADP-ribose) into a band at 45 kDa (and sometimes 52 kDa) that was later proven to be the α subunit. During the purification of Gs from mammalian sources (Northup et al., 1980) it was discovered that it lost the ability to serve as substrate for cholera toxin. It was soon realized that an essential cofactor in the reaction (the ADP-ribosylation factor (Arf)) had been lost during purification. Thus, the search was begun to identify Arf on the grounds that it was likely to hold a key to other regulatory features of Gs and adenylyl cyclase. A few years later the first Arf preparation was pure (Kahn and Gilman, 1984, 1986), but 20 years later we still lack any evidence for a role for Arf in the endogenous regulation of Gs or adenylyl cyclase. With the discovery around that time that the human oncogene product, p21 Ras, was a GTPase there was even a short period where we believed they could be one and the same. They are not. And without the glamour of oncogenic potential, Arf remained something of a curiosity without a clearly defined cellular function for several years. Arfs have been implicated in a large number of cellular processes, including vesicular membrane traffic, Golgi morphology, lipid metabolism, actin cytoskeleton, endocytosis, cytokinesis, and exocytosis. The situation today, like other GTPase families, is that Arf has perhaps too many functions and effectors (>12 so far). While working to describe actions of Arf on specific signaling systems, we also hope to understand the broader question of why nature has evolved to use these proteins in so many ways. It is my premise that it is this diversity in functions that is the real key to Arfs, and probably other families of GTPases. Because in sharing a common regulatory factor these different cellular systems are provided with a way to communicate with one another and systems integration is achieved (see below). Though Arf was identified, purified, and cloned in the 1980s (Kahn and Gilman, 1984, 1986; Kahn et al., 1988; Sewell and Kahn, 1988) that work would not have been possible without the groundwork performed earlier by Dany Cassel and Thomas Pfeuffer (Cassel and Pfeuffer, 1978), whose work established the mechanism of activation of adenylyl cyclase by cholera toxin; by Joel Moss, Martha Vaughan, and colleagues (Moss et al., 1976; Moss and Vaughan, 1977) who worked out the enzymology of cholera toxin

x

KAHN

and the role of NAD as substrate and ADP-ribosyl donor; and of Mike Gill and co-workers (Enomoto and Gill, 1979; Gill, 1975, 1976, 1977; Gill and Meren, 1978; Gill and Rappaport, 1979) whose earlier work on ADPribosylation by diphtheria toxin and movement into the cholera toxin field contributed so much, along with his clever means of identifying the soluble factor, later termed Arf. Mike died suddenly and far too prematurely. His enthusiasm and inventiveness are still missed. With the discovery of the conservation of Arf in eukaryotes (Kahn et al., 1991; Sewell and Kahn, 1988) and its linkage both spatially and genetically to the Golgi (Stearns et al., 1990), Arf functionality began to have a home. The pace for modeling Arf functions quickened enormously with the observation that Arfs, or more precisely the Arf activating proteins, were targets of the drug brefeldin A, a fungal metabolite with profound effects on Golgi morphology and membrane traffic (Donaldson et al., 1992a; Donaldson et al., 1992b; Helms and Rothman, 1992; Lippincott-Schwartz et al., 1989; Randazzo et al., 1993). A role for Arf in the initiation and composition of vesicle coating became the first molecular mechanism for an Arf action in mammalian cells (Orci et al., 1993; Rothman and Orci, 1992; Serafini et al., 1991). However, with the identification of Arf as the cytosolic factor that conferred sensitivity to in vitro assays of phospholipase D (Brown et al., 1993; Cockcroft et al., 1994) and later phosphatidylinositol 4-phosphate 5-kinase (PI(4)P 5-kinase) (Honda et al., 1999), roles in lipid metabolism became evident. Attempts to integrate the vesicle coating actions of Arfs and those on lipid metabolisms are ongoing. Though purified Arf preparations were heterogeneous (multiple bands in SDS gels) it was molecular cloning techniques that defined the true complexity of the Arf family, a process that is still continuing today. Work from Joel Moss and co-workers (Bobak et al., 1989; Lee et al., 1992; Monaco et al., 1990; Price et al., 1988; Tsuchiya et al., 1989) as well as members of my own lab (Clark et al., 1993; Kahn et al., 1991; Sewell and Kahn, 1988; Tamkun et al., 1991) and that of Hans Joost (Breiner et al., 1996; Schurmann et al., 1994; Schurmann et al., 1995) described six mammalian Arfs and most of the Arf-like (Arl) proteins. The high degree of structural and functional conservation between Arf and Arl orthologs is remarkable and has even led to the proposal that the Arf family may be the predecessor of the Ras and perhaps also the heterotrimeric G protein families (Lee et al., 1992; Murtagh et al., 1992). These and other aspects of the evolution of the Arf family are considered in the chapter from John Logsdon (see Chapter 1). The overwhelming majority of studies to date have focused on roles and activities of Arf1 and or Arf3. The most diverse Arf in sequence and function is Arf6, which acts primarily at the plasma membrane and impacts

THE ARF FAMILY OF REGULATORY GTPASES

xi

both GPCR signaling (the long sought connection between Arf and G proteins?) and the cytoskeleton. And we still lack functions for Arf4 and Arf5. While the six Arfs are difficult to distinguish biochemically, recent results reveal that each can be shown to possess distinctive functions in live cells (K. Hill and R. A. Kahn, unpublished observations). Jim Casanova (see Chapter 14) has undertaken the Herculean task of making the Arf6 literature comprehensible and he has succeeded beyond my wildest hopes. Together with the exciting work and novel model presented by Mary Hunzicker-Dunn (see Chapter 15), we all have a much better appreciation for the roles of Arf6 in the plasma membrane. The Arf family is divided into the Arfs, sharing very high percent identities (>60%) and defining activities, and Arf-likes (Arls) of lesser but still quite high (40-60%) identity and devoid of those activities1. It was Hans Joost and Annette Schurmann (Breiner et al., 1996; Schurmann et al., 1994; Schurmann et al., 1995) who continued to increase the number of Arf-like proteins in the database through screens and their studies of differentiation of 3T3-L1 cells and adipocytes. They describe their latest work, including knockouts in mice, and provide the first substantive evidence for Arl functions in development (see Chapter 16). The Arls are proving to be much more diverse in function, as expected from their lower percent identities. With further definition of Arl functions and their partners we will certainly have a richer understanding of the Arf family as a whole.

3.

THE GTPASE CYCLE

The reason these GTPases are often referred to as molecular switches is that they exist in the cell in two forms, one bound to GDP and the other to GTP. These are traditionally referred to as “off” and “on”, or “inactive” and “active”, respectively, though this is clearly an over-simplification. It is true that in the active conformation a GTPase may bind and allosterically activate an enzymatic activity; as is the case with Arfs and phospholipase D (see Chapter 11), PI(4)P 5-kinase, or the ADP-ribosyltransferase activity of cholera toxin (see Chapter 10). However, many of the binding partners of GTPases are not enzymes. In this case, the binding of the GTPase may serve 1

Because the original Arf assay (Kahn, 1991; Kahn and Gilman, 1984; Weiss et al., 1989) is technically demanding and requires a source of purified Gs, other assays have been used to characterize Arfs. These include the simple substitution of agmatine for Gs as ADPribose acceptor (Tsai et al., 1987) to the completely independent assays of Arf dependent PLD (Brown et al., 1993; Kuai and Kahn, 2002) or PI(4)P 5-kinase assays (Honda et al., 1999). Perhaps the most stringent assay for Arf function, because it is a live cell assay and tests for the essential activities of Arf in eukaryotes, is that of rescue of the double deletion arf1- arf2- in the yeast S. cerevisiae (Kahn et al., 1991; Lee et al., 1992).

xii

KAHN

to alter the cellular location of the effector, promote further protein interactions and the assembly of a multisubunit complex, or perhaps prevent the binding of another protein and thereby function indirectly. It is therefore more accurate to simply state that the two forms of a GTPase interact with distinct sets of binding partners. The best example of this paradigm is Ran, which coordinates transport in and out of the nucleus depending on whether it is liganded to GDP or GTP. Yet each form is active and must bind specific partners, the functionally important difference is only that of direction. The two interconvertible states of any regulatory GTPase result from nucleotide-sensitive conformational changes in two regions of the protein referred to as switches I and II. Arf family members have two additional regions that are important to their functions, referred to as the N-terminal helix and the interswitch region. The story of how these are integrated in signaling is beautifully told by Jacqueline Cherfils (see Chapter 2). The Nterminal helix is central to the ability of Arfs to translocate from cytosol to membranes in a fashion that coordinates activation with membrane binding (Antonny et al., 1997; Franco et al., 1996; Paris et al., 1997). The presence of a permanent, covalently attached, saturated, 14-carbon fatty acid (myristate) to the N-terminal glycine of all Arfs (Kahn et al., 1988) (but apparently not all Arls (Sharer et al., 2001)) is essential to both membrane association and biological effects. The ability of Arfs to both directly bind phospholipid membranes and modify their composition is expected to play key roles in Arf biology, as summarized by Paul Randazzo (see Chapter 3). The interconversion between the two states of a GTPase is promoted by two types of activities: guanine nucleotide exchange factors (GEFs) increase the level of activated (GTP-bound) GTPase and GTPase activating proteins (GAPs) speed the intrinsic hydrolysis of the bound GTP. It is generally assumed, but almost never documented, that the release of GDP is the ratelimiting step in the GTPase cycle. The term GEF has tended to supersede two more precise terms; guanine nucleotide dissociation inhibitor (GDI) or stimulator (GDS). This is unfortunate because with so many different GEFs for each GTPase there is ample opportunity for distinctive mechanisms of activation that may prove to have important functional repercussions. Unlike the heterotrimeric G proteins, which are activated by the G protein coupled receptors (GPCRs) upon their binding ligand, we do not yet know what the stimulus is for Arf GEFs to act or indeed if one exists. With many of the processes effected by Arfs being constitutive in nature it is possible that Arf GEFs function all the time and their activities are controlled not by binding specific signaling ligands but perhaps by secondary factors such as the presence of specific lipids or location in the cell. The diversity and

THE ARF FAMILY OF REGULATORY GTPASES

xiii

complexity in Arf activation is revealed clearly by the three chapters on the Arf GEFs (see Chapters 4-6). The importance of diversity in mechanism is even more evident for the Arf GAPs. The elegant structure-based studies of Jonathan Goldberg (Goldberg, 1998, 1999, 2000) have led him to propose a model in which an Arf GAP and an effector (COPI) may bind the two switches of Arf•GTP simultaneously. This model fits beautifully with a proofreading mechanism for Arfs and Arf GAPs in cargo selection during vesicle biogenesis but is unlikely to hold true for all Arf GAPs (Jacques et al., 2002; Nie et al., 2002). The presence of the Arf GAP motif in >20 (predicted) human proteins certainly holds the promise of great diversity in functions as well as mechanisms. While many of these have yet to be shown to possess Arf GAP activity they beg the question: does the presence of an Arf GAP in a complex or pathway provide evidence for a role of an Arf? The first Arf GAP discovered, by Dany Cassel (Cukierman et al., 1995; Makler et al., 1995), now appears streamlined and specialized (see Chapter 7). In contrast, many of the Arf GAPs contain multiple protein-interaction domains and the promise of more complicated and diverse actions. Roles for Arf GAPs (and Arfs?) in G-protein coupled receptor (see Chapter 8) and paxillin (see Chapter 9) actions seem more than justified from the original work of Richard Premont, Hisataka Sabe (respectively) and their colleagues.

4.

NOMENCLATURE

After a section on GEFs and GAPs seems a very appropriate place to inject a note on nomenclature. Like any area of signaling today, we are dealing in each case with gene families and key discoveries coming from a number of different labs, using model genetic systems, each using different systems of naming genes and proteins. While it is important to honor priority in discovery, names can quickly become outdated, confusing, or factually incorrect. Change is happening so fast that it can even be the curators of public databases, in well-intentioned efforts at consistency, which exacerbate or perpetuate nomenclature problems. The discovery of the Arf GAP domain (and to only a slightly lesser extent this is true of the Arf GEF/Sec7 domain as well) occurred at a time when genome sequence data was pouring into databases around the world. Workers in several countries started finding more proteins than we had functions. The presence of additional protein interaction domains can help researchers identify biological roles for a protein, but can also foster confusion in naming it. This has been just as true in the Arf GAP and Arf GEF field as in any.

N-term Arf GAP N-term Arf GAP

NM_021940; NP_068759 (440aa) XM_029830; XP_029830 (429aa) Arf GAP7 Arf GAP8

SMAP1a SMAP1b

Arf GAP7

N-term Arf GAP, FG repeats N-term Arf GAP, FG repeats

NM_004504; NP_004495 (562aa) NM_006076; NP_006067 (481aa)

Arf GAP5 Arf GAP6

RIP/RAB/HRB RAB-R/HRB1

Arf GAP5

Arf GAP, ankyrin repeats (2), PH (2) Arf GAP, ankyrin repeats (2), PH (2)

NM_006869/NP_006860 (374aa) NM_018404; NP_060874 (381aa)

AFAP1 AFAP2

PIPBP Cent. Alpha2

Centaurin (alpha1) Centaurin (beta)

AFAP

PH, Arf GAP, PH (2), RhoGAP, PH PH, Arf GAP, PH (2), RhoGAP, PH

Note: 48% identical to SMAP1a

p42IP4

KIAA0580

NM_139181 (1204aa); NM_015242 (1210aa) NM_015230 (1704aa) NM_022481; NP_071926 (1544aa)

FLJ13675

KIAA0167 KIAA1099

ARAP1 ARAP2 ARAP3

Cent. Delta2 Cent. Delta1

ARAP1 ARAP2 ARAP3

ARAP

MRIP1

ACAP1 ACAP2 ACAP3

KIAA0050 KIAA0041 KIAA1716

GTPase, PH, Arf GAP, ankyrin repeats (2) GTPase, PH, Arf GAP, ankyrin repeats (2) GTPase, PH, Arf GAP, ankyrin repeats (2)

PIKE

AMAP1 AMAP2 AMAP3

KIAA1249 KIAA0400

N-term. Arf GAP, ankyrin repeats (3) N-term. Arf GAP, ankyrin repeats (3)

NM_014030; NP_054749 (761 aa) NP_476510 (759aa)

APAP1 APAP2

NM_014770/NM_023017(836aa) NM_014914 (804aa) NM_031946; NP_114152 (875aa)

Cent. Gamma1 Cent. Gamma2 Cent. Gamma3

AGAP1 AGAP2 AGAP3

AGAP

Cent. Beta4 Cent. Beta3 Cent. Beta6

KIAA0148

N-term. Arf GAP N-term. Arf GAP N-term. Arf GAP

NM_018209;NP_060679 (406aa) NM_014570 (516aa) NM_032389; NP_115765 (521 aa)

Arf GAP1 Arf GAP3 Arf GAP4

AGAP1 AGAP2 AGAP3

Cent. Beta1 Cent. Beta2 Cent. Beta5

ACAP1 ACAP2 ACAP3

ACAP

DDEF1/SHAG1 SHAG2/PAG3

p95PKL

Domain organization

Accession number (length)

New names

PH, Arf GAP, ankyrin repeats (2) PH, Arf GAP, ankyrin repeats (2) PH, Arf GAP, ankyrin repeats (2)

DEF1 DDEF2

ASAP1 PAP(alpha) UPLC1

AMAP

p95APP1 p95APP2

KIAA#

NM_014716 (740aa) NM_012287 (778aa) NM_030649 (759aa)

Cat1 Cat2

Git1 Git2

APAP

Arf GAP

Name #4

PH, Arf GAP, ankyrin repeats (3), SH3 PH, Arf GAP, ankyrin repeats (3), SH3 PH, Arf GAP, ankyrin repeats (3)

Arf1GAP ArfGAP1

Arf GAP1 Arf GAP3 Zfp289

Arf GAP1

Name #3

XP_005169(483aa); BAA86563(949aa) NM_003887(1006aa) XP_031298 (EST only)

Name #2

Name #1

Sub-family

Table I: Current and proposed names for human Arf GAPs and their orthologs

xiv KAHN

ACAP = Arf GAP, coiled-coil, ankyrin repeat, PH protein AGAP = Arf GAP, GTPase, ankyrin repeat, PH protein APP = Arf GAP putative, PIX1-interacting, paxillin bind ARAP = Arf GAP, Rho GAP, ankyrin repeat, PH protein ASAP = Arf GAP, SH3, ankyrin repeat, PH protein CAT = Cool-associated tyrosine phosphorylated protein DEF = Differentiation enhancing factor GIT = G protein receptor kinase (GRK) i nteracting GT Pase activating protein HRB = HIV Rev binding protein PAG = paxillin associated Arf GAP PAP = Pyk2 associated protein PKL = paxillin-kinase linker RAB = HIV Rev-associated binding protein RIP = HIV Rev-interacting protein SMAP = stromal membrane associated protein

List of acronyms

List of new acronyms

NOTES: (1) Some human protein sequences are incomplete (based on the presence of orthologs in other mammals) (2) Several messages are alternatively spliced; only the longest protein sequence is indicated (3) Arf GAP 3 = ARG1 (Arf GAP1) locus in human (4) APAP1 and APAP2 named GIT1 and GIT2 loci in mouse and human databases

APAP = Arf GAP with pix and paxillin binding AMAP = A multi-domain Arf GAP protein ACAP = Arf GAP, coiled-coil, ankyrin repeat, PH protein AGAP = Arf GAP, GTPase, ankyrin repeat, PH protein ARAP = Arf GAP, Rho GAP, ankyrin repeat, PH protein AFAP = Arf GAP with phosphoinositide binding and PH domain

THE ARF FAMILY OF REGULATORY GTPASES xv

C. elegans C. elegans C. elegans C. elegans C. elegans

D. melanogaster D. melanogaster D. melanogaster D. melanogaster D. melanogaster D. melanogaster

Species Human Human Human Human Human Human Human Human Human Human Human Human Human Human Human Human

ARNO BRAG EFA6 BIG1 GBF1

BIG1 BRAG ARNO CG11628 CG8487 EFA6

Name 1 BIG1 BIG2 BIG3 Cytohesin 1 ARNO ARNO3 Cytohesin 4 GBF1 EFA6B EFA6A EFA6D BRAG3 Arf GEP100 BRAG1 EFA6C FBS

KIAA1244

Name 3 p200 Arf GEP1

PSD KIAA0942 KIAA1110 KIA0763 KIAA0522

KIAA0248

K06H7.4 M02B7.5 Y55D9A.1B Y6B3A.1 C24H11.7

GBF1 CG6941 CE26942 CE12328 CE19234 CAA21704 CE19701

DS00797.7 CG7578 CG10577 CG11628

DKFZ FBX8

BRAG2

TIC TYL HCA67

Cytohesin 2 Cytohesin 3 GRP1

Name 2 Arf GEF1 Arf GEF2

Table 2: Verified and predicted human and other Arf GEFs

377/393 614 816 1628 1820

1643/1653 1210/1082 348 410 2040 900

Length 1843 1785 1770 398 399 399 394 1859 1056 645 534(F) 740/701 841 1560(F) 771 319

53-239 209-343 356-532 524-655 672-799

57-241 195-391 357-534 510-697 658-844

/579-767 /520-711 12-197 75-260 640-830 399-575

277-464 133-313

317-422 146-276 /593-724 /523-709 26-155 71-258 654-785 421-533

59-244 58-243 63-248 58-243 696-884 549-738 142-329 36-222 /167-358 399-590

(PF01369) SEC7 dom 695-882 640-827

(PS50190) SEC7 dom 709-840 654-785 179-391 73-202 72-201 77-206 72-201 542-671/710-839 578-696 173-287 74-179 /181-314 413-546

NP_498764 NP_500420 NP_502416 NP_493386 AL032627

Q9NKE8 Q95U36 Q9V9Q6 Q8SYT7

AAF53331 AAF51675 AAF57230 NP_610120 Q9V696 AAF56027

P34512 Q94287 Q95Q14 Q9XWG5 Q9XTF0

Q9VCX8

Alternate Accession# XM_037726 NM_006420 AAL04174 NM_004762 NM_017457 O43739 NM_013385 NM_004194 NP_036587 NP_002770 Q9NYI0 Q9UPP2 NP_055684 T00080 NP_115665 Q9NRD0

Protein Accession # XP_037726 NP_006411 XP_050424 NP_004753 NP_059431 NP_004218 NP_037517 Q92538 O95621 Q15673 NP_056125 BAA83062 O94863 O60275 CAB66494 NP_036312

xvi KAHN

YDG1 YDYB YEE1 YEP1 GEA1 YE51 BIG1 BIG2

GNOM1 BIG1 O65490 F22M8.9 F7K15.150 T4C21.270 GNL1 GNL2

S. pombe S. pombe S. pombe S. pombe S. pombe S. pombe S. pombe S. pombe

A. thaliana A. thaliana A. thaliana A. thaliana A. thaliana A. thaliana A. thaliana A. thaliana

SEC7

RP374

Arf GEF

Sec7

Arf GEF

Pichia Pastoris

Rickettsia prowazekii

Plasmodium falciparum

Paramecium tetraurelia

Encephalitozoon cuniculi

Dictyostellum Discoideum Cytohesin3

SEC7 Gea1 Gea2 Syt1 YBGO

S. cerev. S. cerev. S. cerev. S. cerev. S. cerev.

EMB30 F20D10.320 BIG4 BIG3 BIG5 BIG2 Patt. Form.

C26F1.01 C11E3.11C C19A8.01C C23H3.01 BC211.03C C4D7.01C C30.01C P8A3.15C

YDR170C YJR031C YEL022W YPR095C YBL060W

Put prot 206.9 kDa Pr

Put. GEF

Arf GEF5

F=fragment

1063

1135

3384

462

295-481

573-759

3-194

676-862

256-443

558-745 /547-732 564-749 605-790 565-726 614-799 558-745

1-84 545-732 696-882 705-891 696-882

252-422 300-486

824-1010 538-748 556-756 426-622 61-266

Q8SR27

Q9NJQ4

CAA14833.1

AAK40234

AC115599.1

Q42510 NP_195533 NP_195264 NP_171698 NP_189916 NP_191645 NP_198766 NP_197462

Q10491 NP_594936 O13817 O13937 NP_596613 O14168 NP_594555 NP_594954

P11075 P47102 P39993 Q06836 P34225

Q9ZDF5

Q96X17

Q8T2K1

S65571 Q9SZM0 O65490 Q9LPC5 Q9LXK4 Q9LZX8 Q9FLY5

O13690 CAB11637.1 CAB16232.1 Q9P7R8 CAB11286.1 Q9P7V5 Q9UT02

NP_015420 NP_009493

AAB04031

InterPro#s used when others not available (e.g., Q9LZX8)

303-479

587-717

17-150

672-860

252-441

931 1772

572-703 /561-690 578-707 619-748 579-684 628-757 572-703

263-420 325-446 338-428 1-46 559-688 710-840 719-849 710-840

928 942 428 660(F?) 1462 1571 1822 1811 1451 1698/1643 1711 1750 1669 1793 1443 1375

838-968 552-706 570-714 477-578 105-208

2009 1408 1459 1226 687

THE ARF FAMILY OF REGULATORY GTPASES xvii

xviii

KAHN

The occasion of the publication of this book led to an unusual effort by researchers in the field to reach a consensus on nomenclature for the Arf GAPs. The willingness of these people to put aside personal claims of priority in the spirit of collegiality and to bring order to what is already a complex field is laudable and is to the betterment of the field. The results of this altruism are presented in Table 1. There you will find a compilation of the names of the predicted human Arf GAPs, with accession numbers, lengths of predicted proteins, domain structure, and the currently “preferred” name settled on by workers in the field. While not all parties have adopted this “system” of naming the Arf GAPs, a clear majority has and with consistent use in the literature all have indicated their willingness to adopt this system. It is hoped that as a result we can all spend more time on documenting activities and functions for these proteins and less time worrying about the names. Input from close to 20 investigators was important to the construction of this Table. A similar effort was undertaken with the Arf GEFs, though much less exhaustive in terms of input from investigators. Here we tried to focus on proteins containing Sec7 domains from multiple model organisms as help in defining orthologous proteins. As a result, no effort was made to develop a new system for naming these proteins/genes. I would like to thank Cathy Jackson for her help in putting together Table 2.

5.

ARF EFFECTORS

Despite the discussion above, that might lead one to believe Arfs do everything, it now appears clear that at least a primary function of Arfs is the regulation of aspects of vesicular membrane traffic. More specifically, Arfs are involved in the formation of coated vesicles through the recruitment of specific coat proteins and initiation of large protein complex assemblies. Whether changes in phospholipid metabolism play an obligate or regulatory role in this process remains uncertain. Whether other nucleotide-sensitive binding partners of less well-defined function (including mitotic kinesin like protein 1 (MKLP1), Arfaptin1 and Arfaptin2/POR1, and Arfophilin (see Chapter 11)) are also involved in vesicle traffic is also not known. Twohybrid and genetic screens have identified other partners of Arfs, which lack obvious ties to vesicle traffic (A. L. Boman, C. Zhang, and R. A. Kahn, unpublished observations). Results with Arf6 strongly suggest actions distinct from vesicle traffic, though changes in the cytoskeleton may be part of the remodeling required for vesiculation. Whether all these effectors end up being involved in vesicle biogenesis seems doubtful, but it is becoming clear that it is there that we will first gain a clear understanding of a

THE ARF FAMILY OF REGULATORY GTPASES

xix

molecular model for a function of Arf and Arf GAPs in cells. For this reason, we have two chapters devoted to aspects of coat protein biology, those from Annette Boman and Tommy Nilsson (see Chapter 12) and Victor Faundez (see Chapter 13). Some people may be surprised to see a discussion of Arf GAPs in the section on effectors. I believe this is because GAPs for several GTPases were first discovered as negative regulators of the GTPase signaling. However, GAPs share the biochemical feature of binding preferentially to the GTP-bound form of the GTPase they act upon and necessarily localize to the same place in cells. By using the same switch domains as effectors GAPs have the potential to alter cell signaling both through the increased hydrolysis of GTP bound to the GTPase but also by competing with effectors for the binding of the GTP-bound GTPase. In addition, the presence of other protein interaction domains on Arf GAPs is expected to facilitate the assembly of multi-subunit complexes, perhaps analogous to the action of Arfs to recruit coat proteins to budding vesicles. Whether a “pure” GAP exists, whose function is solely to terminate GTPase signaling, remains to be seen, but I doubt it. Rather, I believe that the vast majority of GAPs, particularly the Arf GAPs with their multiple interaction domains, will prove to be effectors in every sense of the word and transduce positive signals between the GTPase and downstream effects. The fact that GTPase stimulation is a part of their action lends some support to the basic “proofreading” model proposed by Goldberg for Arf GAP1 (Goldberg, 1999, 2000), though modifications are required to accommodate many of the other Arf GAPs. If we add Arf GAPs to the list of effectors, we currently have >30 effectors mediating the actions of 6 mammalian Arfs, which were activated by ≥16 Arf GEFs. And so far there appears to be little or no cross reactivity of Arf GEFs and GAPs with the Arls. If Arfs are the earliest regulatory GTPase they will have the most layers of complexity to untangle as we develop models of their cellular functions. But they also offer the most promise of teaching us some fundamental principles of cell regulation and the evolution of cell regulation.

6.

THE FUTURE OF ARF

In the next few years we will develop much more complete models of how Arfs and their partners regulate vesicle biogenesis and integrate protein and lipid signaling in eukaryotes. Efforts recently underway to investigate global changes in lipid content of cells in response to agonist stimulation or differentiation will give us a much more complete view of cell signaling and

xx

KAHN

this will likely be very important in developing a broader view of Arf biology. Though we have plenty of work ahead to complete the molecular description of Arf actions in eukaryotes, studies of Arls have begun to yield answers that broaden the functions of Arf family members considerably (see 16). The percent identity between Arfs and Arls is comparable to that between Gs and Gi or between Rac and Rho. Thus, we expect both differences and overlap in functions of Arls and Arfs. There is still no Arl GEF or Arl GAP described. I expect this will change soon and with that change will come another quantum jump forward in our understanding of the actions of this family; along with many more surprises and fascinating features and functions for members of the Arf family and their acquaintances. Perhaps one or more Arl will serve as a simpler model for Arf actions and allow us to move faster in developing these models and in furthering out understanding of cell regulation by members of the Arf family.

7.

ACKNOWLEDGEMENTS

I would like to thank Judith Karl for her assistance in putting together and proof reading each of the chapters in this book, including this preface. Her attention to detail and organizational skills made this a much easier task for me. The patience and support of other members of the department of Biochemistry at Emory University is also appreciated. I also thank Paul Randazzo for his critical reading of this preface.

REFERENCES Antonny, B., Beraud-Dufour, S., Chardin, P., and Chabre, M. (1997). N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry, 36, 4675-4684. Bobak, D. A., Nightingale, M. S., Murtagh, J. J., Price, S. R., Moss, J., and Vaughan, M. (1989). Molecular cloning, characterization, and expression of human ADP-ribosylation factors: two guanine nucleotide-dependent activators of cholera toxin. Proc. Natl. Acad. Sci., USA, 86, 6101-6105. Breiner, M., Schurmann, A., Becker, W., and Joost, H. G. (1996). Cloning of a novel member (Arl5) of the Arf-family of Ras-related GTPases. Biochim. Biophys. Acta, 1308, 1-6. Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity [see comments]. Cell, 75, 1137-1144.

THE ARF FAMILY OF REGULATORY GTPASES

xxi

Cassel, D., and Pfeuffer, T. (1978). Mechanism of cholera toxin action: covalent modification of the guanyl nucleotide-binding protein of the adenylate cyclase system. Proc. Natl. Acad. Sci., USA, 75, 2669-2673. Clark, J., Moore, L., Krasinskas, A., Way, J., Battey, J., Tamkun, J., and Kahn, R. A. (1993). Selective amplification of additional members of the ADP-ribosylation factor (Arf) family: cloning of additional human and Drosophila Arf-like genes. Proc. Natl. Acad. Sci., USA, 90, 8952-8956. Cockcroft, S., Thomas, G. M., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of Arf in granulocytes. Science, 263, 523-526. Cukierman, E., Huber, I., Rotman, M., and Cassel, D. (1995). The Arf1 GTPase-activating protein: zinc finger motif and Golgi complex localization. Science, 270, 1999-2002. Donaldson, J. G., Cassel, D., Kahn, R. A., and Klausner, R. D. (1992a). ADP-ribosylation factor, a small GTP-binding protein, is required for binding of the coatomer protein betaCOP to Golgi membranes. Proc. Natl. Acad. Sci., USA, 89, 6408-6412. Donaldson, J. G., Finazzi, D., and Klausner, R. D. (1992b). Brefeldin A inhibits Golgi membrane-catalysed exchange of guanine nucleotide onto Arf protein. Nature, 360, 350352. Enomoto, K., and Gill, D. M. (1979). Requirement for guanosine triphosphate in the activation of adenylate cyclase by cholera toxin. J. Supramol. Struct., 10, 51-60. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1996). Myristoylation-facilitated binding of the G protein Arf1GDP to membrane phospholipids is required for its activation by a soluble nucleotide exchange factor. J. Biol. Chem., 271, 1573-1578. Gibbs, J. B., Sigal, I. S., Poe, M., and Scolnick, E. M. (1984). Intrinsic GTPase activity distinguishes normal and oncogenic Ras p21 molecules. Proc. Natl. Acad. Sci., USA, 81, 5704-5708. Gill, D. M. (1975). Involvement of nicotinamide adenine dinucleotide in the action of cholera toxin in vitro. Proc. Natl. Acad. Sci., USA, 72, 2064-2068. Gill, D. M. (1976). The arrangement of subunits in cholera toxin. Biochemistry, 15, 12421248. Gill, D. M. (1977). Mechanism of action of cholera toxin. Adv. Cyclic Nucleotide Res., 8, 85118. Gill, D. M., and Meren, R. (1978). ADP-ribosylation of membrane proteins catalyzed by cholera toxin: basis of the activation of adenylate cyclase. Proc. Natl. Acad. Sci., USA, 75, 3050-3054. Gill, D. M., and Rappaport, R. S. (1979). Origin of the enzymatically active A1 fragment of cholera toxin. J. Infect. Dis., 139, 674-680. Goldberg, J. (1998). Structural basis for activation of Arf GTPase: mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Goldberg, J. (1999). Structural and functional analysis of the Arf1-ArfGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Goldberg, J. (2000). Decoding of sorting signals by coatomer through a GTPase switch in the COPI coat complex. Cell, 100, 671-679. Helms, J. B., and Rothman, J. E. (1992). Inhibition by brefeldin A of a Golgi membrane enzyme that catalyses exchange of guanine nucleotide bound to Arf. Nature, 360, 352354. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase alpha is a downstream effector of the small G protein Arf6 in membrane ruffle formation. Cell, 99, 521-532.

xxii

KAHN

Jacques, K. M., Nie, Z., Stauffer, S., Hirsch, D. S., Chen, L. X., Stanley, K. T., and Randazzo, P. A. (2002). Arf1 dissociates from the clathrin adaptor GGA prior to being inactivated by Arf GTPase-activating proteins. J. Biol. Chem., 277, 47235-47241. Kahn, R. A. (1991). Quantitation and purification of ADP-ribosylation factor. Methods Enzymol., 195, 233-242. Kahn, R. A., and Gilman, A. G. (1984). Purification of a protein cofactor required for ADPribosylation of the stimulatory regulatory component of adenylate cyclase by cholera toxin. J. Biol. Chem., 259, 6228-6234. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kahn, R. A., Goddard, C., and Newkirk, M. (1988). Chemical and immunological characterization of the 21-kDa ADP-ribosylation factor of adenylate cyclase. J. Biol. Chem., 263, 8282-8287. Kahn, R. A., Kern, F. G., Clark, J., Gelmann, E. P., and Rulka, C. (1991). Human ADPribosylation factors. A functionally conserved family of GTP- binding proteins. J. Biol. Chem., 266, 2606-2614. Kuai, J., and Kahn, R. A. (2002). Assays of Arf function. Methods Enzymol., 345. Lee, F. J., Moss, J., and Vaughan, M. (1992). Human and Giardia ADP-ribosylation factors (Arfs) complement Arf function in Saccharomyces cerevisiae. J. Biol. Chem., 267, 2444124445. Lippincott-Schwartz, J., Yuan, L. C., Bonifacino, J. S., and Klausner, R. D. (1989). Rapid redistribution of Golgi proteins into the ER in cells treated with brefeldin A: evidence for membrane cycling from Golgi to ER. Cell, 56, 801-813. Makler, V., Cukierman, E., Rotman, M., Admon, A., and Cassel, D. (1995). ADP-ribosylation factor-directed GTPase-activating protein. Purification and partial characterization. J. Biol. Chem., 270, 5232-5237. Monaco, L., Murtagh, J. J., Newman, K. B., Tsai, S. C., Moss, J., and Vaughan, M. (1990). Selective amplification of an mRNA and related pseudogene for a human ADPribosylation factor, a guanine nucleotide-dependent protein activator of cholera toxin. Proc. Natl. Acad. Sci., USA, 87, 2206-2210. Moss, J., Manganiello, V. C., and Vaughan, M. (1976). Hydrolysis of nicotinamide adenine dinucleotide by choleragen and its A protomer: possible role in the activation of adenylate cyclase. Proc. Natl. Acad. Sci., USA, 73, 4424-4427. Moss, J., and Vaughan, M. (1977). Choleragen activation of solubilized adenylate cyclase: requirement for GTP and protein activator for demonstration of enzymatic activity. Proc. Natl. Acad. Sci., USA, 74, 4396-4400. Murtagh, J. J., Jr., Mowatt, M. R., Lee, C. M., Lee, F. J., Mishima, K., Nash, T. E., Moss, J., and Vaughan, M. (1992). Guanine nucleotide-binding proteins in the intestinal parasite Giardia lamblia. Isolation of a gene encoding an approximately 20-kDa ADP- ribosylation factor. J. Biol. Chem., 267, 9654-9662. Nie, Z., Stanley, K. T., Stauffer, S., Jacques, K. M., Hirsch, D. S., Takei, J., and Randazzo, P. A. (2002). AGAP1, an endosome-associated, phosphoinositide-dependent ADPribosylation factor GTPase-activating protein that affects actin cytoskeleton. J. Biol. Chem., 277, 48965-48975. Northup, J. K., Sternweis, P. C., Smigel, M. D., Schleifer, L. S., Ross, E. M., and Gilman, A. G. (1980). Purification of the regulatory component of adenylate cyclase. Proc. Natl. Acad. Sci., USA, 77, 6516-6520. Orci, L., Palmer, D. J., Amherdt, M., and Rothman, J. E. (1993). Coated vesicle assembly in the Golgi requires only coatomer and Arf proteins from the cytosol. Nature, 364, 732-734.

THE ARF FAMILY OF REGULATORY GTPASES

xxiii

Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997). Role of protein-phospholipid interactions in the activation of Arf1 by the guanine nucleotide exchange factor Arno. J. Biol. Chem., 272, 22221-22226. Price, S. R., Nightingale, M., Tsai, S. C., Williamson, K. C., Adamik, R., Chen, H. C., Moss, J., and Vaughan, M. (1988). Guanine nucleotide-binding proteins that enhance choleragen ADP-ribosyltransferase activity: nucleotide and deduced amino acid sequence of an ADPribosylation factor cDNA. Proc. Natl. Acad. Sci., USA, 85, 5488-5491. Randazzo, P. A., Yang, Y. C., Rulka, C., and Kahn, R. A. (1993). Activation of ADPribosylation factor by Golgi membranes. Evidence for a brefeldin A- and proteasesensitive activating factor on Golgi membranes. J. Biol. Chem., 268, 9555-9563. Rothman, J. E., and Orci, L. (1992). Molecular dissection of the secretory pathway. Nature, 355, 409-415. Schurmann, A., Breiner, M., Becker, W., Huppertz, C., Kainulainen, H., Kentrup, H., and Joost, H. G. (1994). Cloning of two novel ADP-ribosylation factor-like proteins and characterization of their differential expression in 3T3-L1 cells. J. Biol. Chem., 269, 15683-15688. Schurmann, A., Massmann, S., and Joost, H. G. (1995). ARP is a plasma membraneassociated Ras-related GTPase with remote similarity to the family of ADP-ribosylation factors. J. Biol. Chem., 270, 30657-30663. Serafini, T., Orci, L., Amherdt, M., Brunner, M., Kahn, R. A., and Rothman, J. E. (1991). ADP-ribosylation factor is a subunit of the coat of Golgi-derived COP-coated vesicles: a novel role for a GTP-binding protein. Cell, 67, 239-253. Sewell, J. L., and Kahn, R. A. (1988). Sequences of the bovine and yeast ADP-ribosylation factor and comparison to other GTP-binding proteins. Proc. Natl. Acad. Sci., USA, 85, 4620-4624. Sharer, J. D., Shern, J. F., Van Valkenburgh, H., Wallace, D. C., and Kahn, R. A. (2001). Arl2 and BART enter mitochondria and bind the adenine nucleotide transporter (ANT). Mol. Biol. Cell, 2002, 71-83. Stearns, T., Willingham, M. C., Botstein, D., and Kahn, R. A. (1990). ADP-ribosylation factor is functionally and physically associated with the Golgi complex. Proc. Natl. Acad. Sci., USA, 87, 1238-1242. Sweet, R. W., Yokoyama, S., Kamata, T., Feramisco, J. R., Rosenberg, M., and Gross, M. (1984). The product of Ras is a GTPase and the T24 oncogenic mutant is deficient in this activity. Nature, 311, 273-275. Tamkun, J. W., Kahn, R. A., Kissinger, M., Brizuela, B. J., Rulka, C., Scott, M. P., and Kennison, J. A. (1991). The Arflike gene encodes an essential GTP-binding protein in Drosophila. Proc. Natl. Acad. Sci., USA, 88, 3120-3124. Tsai, S. C., Noda, M., Adamik, R., Moss, J., and Vaughan, M. (1987). Enhancement of choleragen ADP-ribosyltransferase activities by guanyl nucleotides and a 19-kDa membrane protein. Proc. Natl. Acad. Sci., USA, 84, 5139-5142. Tsuchiya, M., Price, S. R., Nightingale, M. S., Moss, J., and Vaughan, M. (1989). Tissue and species distribution of mRNA encoding two ADP-ribosylation factors, 20-kDa guanine nucleotide binding proteins. Biochemistry, 28, 9668-9673. Weiss, O., Holden, J., Rulka, C., and Kahn, R. A. (1989). Nucleotide binding and cofactor activities of purified bovine brain and bacterially expressed ADP-ribosylation factor. J. Biol. Chem., 264, 21066-21072.

Contributors Annette Boman, Ph.D./ Department of Biochemistry and Molecular Biology/ University of Minnesota-Duluth School of Medicine/ 10 University Drive/Duluth, MN 55812/USA/Phone 218 726 7036/Fax 218 726 8014/Email [email protected] James Casanova, Ph.D./Department of Cell Biology/Box 800732/University of Virginia Health Sciences Center/1300 Jefferson Park Avenue/Charlottesville, VA 22908-0732/USA/Phone 434 243 4821/Fax 434 982 3912/Email [email protected] Dan Cassel, Ph.D./Department of Biology/Technion-Israeli Institute of Technology/Haifa 32000/Israel/Phone 972 4829 3408/Fax 972 4822 5153/Email [email protected] Jacqueline Cherfils/Laboratoire d'Enzymologie et Biochimie Structurales/CNRS, Avenue de la Terrasse/91198 Gif-sur-Yvette cedex/France/Phone 33 1 69 823 492/Fax 33 1 69 823 129/Email [email protected] John H. Exton, M.D., Ph.D./Department of Molecular Physiology/Biophysics and Pharmacology/Howard Hughes Medical Institute/Vanderbilt University/702 Light Hall/Nashville, TN 372320615/USA/Phone 615 322 6494/Fax 615 322 4381/Email [email protected] Victor Faundez, M.D., Ph.D./Department of Cell Biology and Center for Neurodegenerative Diseases/446 Whitehead Biomedical Research Building/615 Michael Street/Emory University School of Medicine/Atlanta, GA 30322/USA/Phone 404 727 3900/Fax 404 727 6256/Email [email protected] Dianne S. Hirsch, Ph.D./Laboratory of Cellular Oncology CCR, NCI/Bldg 37 Room 6032/Bethesda, MD 20892 USA/Phone 301 496 6808/Fax 301 480 1260/Email [email protected] Mary Hunzicker-Dunn/Department of Cell and Molecular Biology/Northwestern University Medical School/Tarry Building 8xxv

xxvi

721/303 East Chicago Avenue/Chicago, IL 60611/USA/Phone 312 503 7459/Fax 312 503 0566 or 312 503 7912/Email [email protected] Catherine L. Jackson, Ph.D./Cell Biology and Metabolism Branch/NICHD, NIH/Building 18T, Room 101/18 Library Drive MSC 5430/Bethesda, MD 20892-5430/USA/Phone 301 594 3944 Fax 301 402 0078/Email [email protected] Prof. Dr. H.-G. Joost/German Institute of Human Nutrition/ArthurScheunert Allee 114-116D-14558 PotsdamRehbrücke/Germany/Phone 49 33200 88216/Fax 49 33200 88555/Email [email protected] Mr. Troy Lasell/Department of Cell Biology/Medical Sciences Building Room 5-14/University of Alberta/Edmonton AB/Canada T6G 2H7/Phone 780 492 6309/Fax 780 492 0450/Email [email protected] John Logsdon, Ph.D./Department of Biology/Emory University/1510 Clifton Road/Atlanta, GA 30322/USA/Phone 404 727 9516/Fax 404 727 2880/Email [email protected] Paul Melancon, Ph.D./Department of Cell Biology/Medical Sciences Building Room 5-14/University of Alberta/Edmonton AB/Canada T6G 2H7/Phone 780 492 6183/Fax 780 492 0450/Email [email protected] Naoko Morinaga, Ph.D./2nd Department of Microbiology/School of Medicine/Chiba University/Chiba 280/Japan/Phone 81 043 226 2048/Fax 81 043 226 2049/Email [email protected] Joel Moss, M.D., Ph.D./Pulmonary-Critical Care Medicine Branch/HLBI, NIH/Building 10, Room 6D03/9000 Rockville Pike, 10 Center Drive/Bethesda, MD 20892-1590/USA/Phone 301 496 1597/Fax 301 496 2363/Email [email protected] Zhongzhen Nie, Ph.D./Laboratory of Cellular Oncology/CCR, NCI/Bldg 37 Room 6032/Bethesda, MD 20892/USA/Phone 301 496 6833/Fax 301 480 1260/Email [email protected]

xxvii

Professor Tommy Nilsson/Department of Medical Biochemistry/ Goteborg University/Medicinaregatan 9A/413 90 Gothenburg/Sweden and until December 2003/Cell Biology and Biophysics Programe/ EMBL/Meyerhofstrasse 1/D-69017 Heidelberg/Germany/Phone 49 6221 387 294 (or 408)/Fax 49 6221 387 306 (or 512)/Email [email protected] Mastoshi Noda, Ph.D./2nd Department of Microbiology/School of Medicine/Chiba University/Chiba 280/Japan/Phone 81 043 226 2046/Fax 81 043 226 2049/Email [email protected] Gustavo Pacheco-Rodriguez, Ph.D./Pulmonary-Critical Care Medicine Branch/NHLBI, NIH/Building. 10, Room 5N314/9000 Rockville Pike, 10 Center Drive/Bethesda, MD 20892-1434/ USA/Phone 301 496 1454/Fax 301 402 1610/Email [email protected] Sebastiano Pasqualato, Ph.D./Laboratoire d'Enzymologie et Biochimie Structurales/CNRS, Avenue de la Terrasse/91198 Gif-surYvette cedex/France/Phone 33 1 69 823 492/Fax 33 1 69 823 129/Email [email protected] Current Address: Dipartimento di Oncologia Sperimentale/Istituto Europeo de Oncologia/Vai Ripamonti 435/20141 Milano, Italy Richard T. Premont, Ph.D./Assistant Research Professor/Liver Center - 329 Sands Building/Department of Medicine (Gastroenterology)/Duke University Medical Center/Campus Box 3083/Durham, NC 27710/USA/Phone 919 684 5620/Fax 919 684 4983/Email [email protected] Paul Randazzo, M.D., Ph.D./Laboratory of Cellular Oncology/CCR, NCI/Bldg 37 Rm 6032/Bethesda, MD 20892/USA/Phone 301 496 3788/Fax 301 480 1260/Email [email protected] Louis Renault, Ph.D./Laboratoire d'Enzymologie et Biochimie Structurales/CNRS, Avenue de la Terrasse/91198 Gif-sur-Yvette cedex/France/Phone 33 1 69 823 492/Fax 33 1 69 823 129/Email [email protected]

xxviii

Hisataka Sabe, Ph.D./Department of Molecular Biology/Osaka Bioscience Institute/6-2-4 Furuedai, Suita City/Osaka 5650874/Japan/Phone 81 6 6872 4814, 4850/Fax 81 6 6871 6686/Email [email protected] Robert Schmalzigaug, Ph.D./Liver Center - 334 Sands Building/Department of Medicine (Gastroenterology)/Duke University Medical Center/Campus Box 3083/Durham, NC 27710/ USA/Phone 919 684 8647/Fax 919 684 4983/Email [email protected] Prof. Dr. Annette Schürmann/German Institute of Human Nutrition/Department of Pharmacology/Arthur-Scheunert Allee 114116/D-14558 Potsdam-Rehbrücke/Germany/Phone 49 33200 88368/Fax 49 33200 88334/Email [email protected] M. L. Styers/Department of Cell Biology and theBiochemistry, Cell, and Developmental Biology/Graduate Division of Biological and Biomedical Sciences/455 Whitehead Biomedical Research Building/615 Michael Street/Emory University School of Medicine/Atlanta, GA 30322/USA/Phone 404 727 3945/Fax 404 727 6256/Email [email protected] Martha Vaughan, M.D./Pulmonary-Critical Care Medicine Branch/ NHLBI, NIH/Building. 10, Room 5N307/9000 Rockville Pike, 10 Center Drive/Bethesda, MD 20892-1434/USA/Phone 301 496 4554/ Fax 301 402 1610/Email [email protected] Ms. Zinhua Zhao/Department of Cell Biology/Medical Sciences Building Room 5-14/University of Alberta/Edmonton AB/Canada T6G 2H7/Phone 780 492 6309/Fax 780 492 0450/Email [email protected]

Chapter 1 THE ARF FAMILY TREE JOHN M. LOGSDON, JR.,1 AND RICHARD A. KAHN2

Departments of Biology1 and Biochemistry2, Emory University School of Medicine, Atlanta, USA

Abstract:

1.

A phylogeny of the Arf family of proteins has been constructed using sequence data available from the complete (or nearly complete) genomes of six model eukaryotic species; including three animals, two fungi and a protist. Our analyses reveal distinct groupings of three Arf and ten Arl (Arf-like) subfamilies. These Arf and Arl subfamilies have arisen by numerous gene duplications that took place at different times during eukaryotic evolution. Some apparently young subfamilies are comprised of proteins found only in animals, while older families include members from all of the eukaryotic lineages considered. Available functional information for these proteins is considered in light of their evolutionary relationships.

INTRODUCTION

ADP-ribosylation factor(s) (Arf(s)) were first described as, and named after, a biochemical activity: cofactor in the ADP-ribosylation of Gs by cholera toxin (Kahn and Gilman, 1984, 1986; Schleifer et al., 1982). The original cloned cDNAs and their proteins that possess that activity were named accordingly (Sewell and Kahn, 1988). Unfortunately, that first Arf activity is now understood to have little or nothing to do with the cellular functions of Arfs. This disconnect between name and function is even more severe for the Arf-like (Arl) proteins, which lack Arf activities and were named on the basis of protein sequence homology (Cavenagh et al., 1994; Clark et al., 1993; Tamkun et al., 1991). As the size of the Arf family expanded, so did the number of activities and binding partners found associated with Arfs and Arls. Other activities emerged that were predicted to be better 1

2

LOGSDON AND KAHN

discriminators between functionally different groups of proteins. These included direct activation of phospholipase D (Brown et al., 1993; Cockcroft et al., 1994) and the ability to rescue the lethality of the Saccharomyces cerevisiae double null mutant arf1-arf2- (Kahn et al., 1991), activities that, in fact, have correlated very well with the original Arf activity. Nonetheless, with more than a dozen effectors now known for Arfs alone and evidence of complicated networks of cross reactivity between Arf family GTPases and effectors (e.g., see (Van Valkenburgh et al., 2001)), it is unlikely that a single activity will be found with the ability to appropriately discriminate between them to allow unambiguous classification on the basis of function. Sequencing of entire genomes has allowed a complete, or very nearly so, cataloging of Arf family members and allows the tracing of their evolutionary relationships. Because functional redundancy is a common complication in interpretation of genetic and biochemical data, this genealogical framework of Arfs and Arls is important as we focus on functions of these proteins. In this chapter we present the results of recent efforts at identifying and reconstructing the evolutionary relationships among all Arf family members in a number of model eukaryotes: human (Homo sapiens), fly (Drosophila melanogaster), worm (Caenorhabditis elegans), two fungi (the yeasts S. cerevisiae, and Schizosaccharomyces pombe), and a protist (the diplomonad Giardia lamblia). We describe evidence for the existence of three Arf classes (the previously described Arf classes I-III) and ten Arls, including Arls1-5, Arl8, Arl9, Arfrp and Sar1. A simplified nomenclature is proposed based on these phylogenetic classifications. We then discuss unique features of each group, their predicted evolutionary origins and functional implications. A phylogenetic analysis of the human Arf family was presented in Jacobs, et al. (Jacobs et al., 1999) and reached many of the same conclusions regarding the naming and number of Arl groups.

2.

THE FAMILY

An iterative bioinformatic process of probing both protein and nucleotide databases with Arf and Arl sequences with BLAST (Altschul et al., 1990; Altschul et al., 1997), followed by construction of protein alignments, phylogenetic analyses, and re-probing for potentially “missing” members led to a final collection of 64 Arf family members, listed in Table 1. These include the protein-coding sequences of 23 human, 12 D. melanogaster, 12 C. elegans, 7 S. cerevisiae, 4 S. pombe, and 6 G. lamblia genes. Protein sequences were aligned as shown in Figures 1 and 2. The alignment was then analyzed by a variety of methods to construct phylogenetic trees,

The ARF Family Tree

3

culminating in the Bayesian maximum likelihood analyses shown in Figure 3 and the simplifed view shown in Figure 4. The aligned and un-edited Nterminal sequences are shown in Figure 1. This serves to illustrate the editing process used and also reveals in many cases the conservation of lengths in regions within groups. For example, note that all Arl2 and Arl3 group members are shorter at their N-termini than the other proteins. This is also true for ArfIIIs vs Arf I or ArfIIs. The differences in lengths in the loop that connects the N-terminal α-helix to the rest of the protein are also likely to have functional significance as this essential domain has previously been shown to be important in coupling nucleotide exchange to protein hydrophobicity (and hence membrane association) and to interact with specific lipids. We focused on these six completely sequenced genomes to compose a list of predicted Arf family members, though we also made use of data from other species to help resolve issues (most of which involved incorrect assignment of intron/exon boundaries by gene prediction software). The Arfs themselves are so well conserved (>60% identity) that there is little ambiguity or difficulty in finding them with simple BLAST searches of proteomes or genomes. Researchers have been inconsistent previously in grouping Arfrp and Sar proteins in the Arf family, as they are the most divergent. Our analysis clearly puts Arfrp in the Arf family and we argue for inclusion of Sar on the basis of both sequence and functional conservation. After Sar, the next proteins appearing on BLAST results (using Arf or Arl sequence probes from any organism) were members of the Rab family of GTPases. This suggests that by including Sar in the Arf family we have found a clear demarcation between these two families of GTPases. Arfs are small proteins at ~20 kDa and therefore smaller or similar in size to several protein domains, including the Sec7 (Arf GEF) and PTB domains, each ~200 residues. In fact, there exists at least one protein that uses the Arf sequence as a domain within a larger context. These are the Arf-domain proteins; e.g., the human 64 kDa Ard1 protein (569 residues) contains an Arf domain at the C-terminus and an Arf GAP domain (Mishima et al., 1993; Vitale et al., 1998). Rodents (rat and mouse; 97% identical) and C. elegans (~37% identical) express orthologs of Hs Ard1. These proteins are not included in our analyses, despite the fact that the Arf domain of Ard1 is ~60% identical to Arfs. Table 1. Summary of Arf family members in H. sapiens (Hs), D. melanogaster (Dm), C. elegans (Ce), S. pombe (Sp), S. cerevisiae (Sc), and G lamblia (Gl). Information from a variety of sources is summarized here to help identify the different proteins. The names used in this chapter are shown in the first column and common names from the different species are shown in column two. Accession numbers are in column three to help locate full-length cDNA and protein sequences. Locus identifiers are also included to aid in gene identification. Information on the chromosomal location for each gene and the length of predicted proteins is included, as are other names that appear in the literature or databases. Finally, the first 12 residues are shown as another means of resolving any confusion between protein names and identity.

12

Proposed name H. sapiens HsArf1 (MmArf2) HsArf3 HsArf4 HsArf5 HsArf6 HsArl1 HsArl2 HsArl3 HsArl4 HsArl4B HsArl5 HsArl5B HsArl6 HsArl7 HsArl8A HsArl8B HsArl9 HsArl10 HsArl11 HsArfrp HsSar1 HsSara 23 C. elegans CeArf1 CeArf3 CeArf6 CeArl1 CeArl2 CeArl3 CeArl5 CeArl6 CeArf CeArl8 CeArfrp CeSar1

Accession number

NM_001658/GI:6995997 NM_007477/GI:6671570 NM_001659/GI:4502202 NM_001660/GI:6995998 NM_001662/GI:6995999 NM_001663/GI:6996000 NM_001177/GI:4755126 NM_001667/GI:4502228 NM_004311/GI:4757773 NM_005738/GI:5031602 NT_033984/XP_166703.1 NM_012097/GI:6912243 XM_167671/GI:20473688 NM_032146/GI:21314750 NM_005737/GI:6552295 NM_138795/GI:15929963 NM_018184/GI:8922600 NM_001661/GI:4502206 NM_025047/GI:13376573 NM_138450/GI:27764878 NM_003224/GI:4507448 NM_020150/GI:21361614 NM_016103/GI:7705826

NM_065834/GI:17551729 NM_068935/GI:17538183 NM_070610/GI:17538185 NM_063415/GI:17531196 NM_063378/GI:17532860 NM_064636/GI:17531200 NM_066777/GI:17551731 NM_065592/GI:17551733 NM_068841/GI:17538189 NM_070390/GI:17543765 NM_058693/GI:17510340 NM_068181/GI:17544539

Other name(s)

Arf1 Arf2 Arf3 Arf4 Arf5 Arf6 Arl1 Arl2 Arl3 Arl4 Arl4B Arl5 Arl5B Arl6 Arl7 Arl8A Arl8B Arf4L/Arl6 Arl10 Arl11 Arp Sar1 Sara

arf-1 arf-3 arf-6 arl-1 evl-20/arl-2 arl-3 arl-5 arl-6 arf arl-8 arfrp sar-1

Gene Location

1q42 11 E1(Mm) 12q13 3p21.2-p21.1 7q31.3 7q22.1 12q23.3 11q13 10q23.3 7p21-p15.3 2q23.3 10p12.33 3q12.1 2q37.2 1q31.3 3p26.1 17q12-q21 3q26.1 13q14.11 20q13.3 10q22.1 5q31.1

III IV IV II II II III III IV IV I IV

Locus identifier

LocusID: 375 LocusID: 11841 LocusID: 377 LocusID: 378 LocusID: 381 LocusID: 382 LocusID: 400 LocusID: 402 LocusID: 403 LocusID: 10124 LocusID: 26225 LocusID: 221079 LocusID: 84100 LocusID: 10123 LocusID: 127829 LocusID: 55207 LocusID: 379 LocusID: 80117 LocusID: 115761 LocusID: 10139 LocusID: 56909 LocusID: 51128

B0336.2 F57H12.1 Y116A8C.12 F54C9.10 F22B5.1 F19H8.3 ZK632.8 C38D4.8 F45E4.1 Y57G11C.13 Y54E10BR.2 ZK180.4

181 180 175 180 184 184 178 190 179 185 191 193

181 181 181 180 180 175 181 184 182 200 234 179 179 186 192 186 186 201 192 196 201 198 198

Length (a.a.)

N-terminal sequence

arf-3 arf-6 arl-3 evl-20 arl-3C arl-2 arl-1, arl-query arl-6; arf-3

MGNVFGSLFKGL MGLTISSLFNRL MGKFLSKIFGKK MGGVMSYFRGLF MGFLKILRKQRA MGLIDVLKSFKS MGLIMAKLFQSW MGFFSSLSSLFG MGLFFSKISSFM MLAMVNKVLDWI MYTLSAGIWEKF MSFLWDWFNGVL

MGNIFANLFKGL MGNVFEKLFKSL MGNIFGNLLKSL MGLTISSLFSRL MGLTVSALFSRI MGKVLSKIFGNK Arfl1 MGGFFSSIFSSL Arfl2 MGLLTILKKMKQ Arfl3 MGLLSILRKLKS MGNGLSDQTSILSN MGNGLSDQTSILSS MGILFTRIWRLF Sim. to Arl5 MGLIFAKLWSLF MGLLDRLSVLLG LAK MGNISSNISAFQ Hypo. Prot BC015408 MIALFNKLLDWF FLJ10702 MLALISRLLDWF Arl6 MGNHLTEMAPTA FLJ22595 MGSLGSKNPQTK MGSVNSRGHKAE MGC17429 Arfrp1 MYTLLSGLYKYM MSFIFEWIYNGF MSFIFDWIYSGF

Other names

Table I: Summary of Arf family members in H. sapiens , D. melanogaster , C. elegans , S. cerevisiae , S. pombe , and G. lamblia

4 LOGSDON AND KAHN

D. melanogaster DmArf1 DmArf2 DmArf3 DmArl1 DmArl2 DmArl3 DmArl4 DmArl5 DmArl6 DmArl8 DmArfrp DmSar1 12 S. cerevisiae ScArf1 ScArf2 ScArf3 ScArl1 ScArl2 ScArfrp ScSar1 7 S. pombe SpArf1 SpArf2 SpArl2 SpSar1 4 G. lamblia GlArf GlArl2 GlArl3 GlArl GlArl8/Arfrp GlSar1 6

P11076/GI:114121 NP_010144/GI:6320064 NP_014737/GI:6324668 NP_009723/GI:6319641 NP_013858/GI:6323787 NP_015274/GI:6325206 NP_015106/GI:6325038

P36579 /GI:543843 Q9Y7Z2 /GI:20137584 Q09767 /GI:1168499 Q01475 /GI:266990

P26991 /GI:114131

Arf1 Arf2 Arf3 Arl1 Cin4 Arl3 Sar1

Arf1 Arf2 alp41/Arl2 Sar1

GlArf Gl4192 GlArlB Gl7562 GlArlA GlSar1

BAB58894/GI:14275896 AY125726/GI:22035408

BAB58895/GI:14275898

NM_057607/GI:24668765 NM_079892/GI:24638807 NM_079027/GI:24653826 NP_524098/GI:24664933 NM_057538/GI:19549394 NM_142738/GI:24648859 NP_651905/GI:24638544 NM_139944/GI:24660780 NM_137577/GI:24655720 NP_649769/GI:21355085 NM_132298/GI:24640728 NP_732717/GI:24648946

DmArf79F DmArf102F DmArf51F DmArf72A DmArl84F DmCG6560 DmCG2219 DmCG7197 DmArl6 DmArl8 DmArfrp DmSar1

Table I (cont.)

3L 2 3 1 3

IV IV XV II XIII XVI XVI

II II II II

YDL192W YDL137W YOR094W YBR164C YMR138W YPL051W YPL218W

SPBC4F6.18c SPBC1539.08 SPAC22F3.05c SPBC31F10.06c

3L 4 2R 3L 3R 3R

Arf79F Arf102F Arf51F Arf72A Arl84F CG6560 CG2219 CG7197 CG7735 CG7891 CG7039 CG7073

191 197 149 168 157 191

180 184 186 190

181 181 183 183 191 198 190

182 180 175 180 184 179 175 179 202 186 200 193

ArlA

ArlB

alp41

GTP1, UGX1 ARL3

ARL2

Darf1, dARFI, CG8385 dARFII, ARFII dArf3, ARFIII arflike, arl, dArl1 arf-like2, arl2 CG6560 CG2219 CG7197 CG7735 CG7891 CG7039 CG7073, sar-1

MGQGASKIFGKL MGFLSVLRAIKA ...TILRKLCEEDTS MGAASSRPARHK ...TTILHQMAYGMT MGFGDWFKSALS

MGLSISKLFQSL MGNSLFKGFSKP MGLLTILRQQKL MFIINWFYDALA

MGLFASKLFSNL MGLYASKLFSNL MGNSISKVLGKL MGNIFSSMFDKL MGLLSIIRKQKL MFHLVKGLYNNW MAGWDIFGWFRD

MGNVFANLFKGL MGLTISSLLTRL MGKLLSKIFGNK MGGVLSYFRGLL MGFLTVLKKMRQ MGLLSLLRKLRP MGATMVKPLVKN MGLLLSRLWRMF MGMLHNLADLIK MLALINRILEWF MYTLMHGFYKYM MFIWDWFTGVLG

The ARF Family Tree 5

6

2.1

LOGSDON AND KAHN

The Arf sub-family

The Arf proteins have been previously divided into three classes (I, II, III), based on their gene sequences and organization (Lee et al., 1994a). This classification is further supported by our analyses. Flies and worms have single representatives for each of these three classes, while mammals have either five or six genes (see below), S. cerevisiae has three Arfs, S. pombe has two and Giardia has only one. The class I and II Arfs, found only among animals, apparently diverged from a common ancestor more recently (after the animal-fungal split) than did class III Arfs, which are found in both animals and fungi. The two yeasts have single representatives for class I/II and class III (excepting that in S. cerevisiae, a very recent duplication of a chromosome IV region resulted in two very similar Arf I/II genes, Arf1 and Arf2). The fact that Arf classes I, II and III form a well-supported group including representatives in animals, fungi and a single Arf gene from Giardia supports our conclusion that all Arfs arose from a common ancestor. The fact that the single Giardia Arf groups (albeit weakly) with class I/II Arfs, suggests that it represents the class I/II ancestral type (as do Sc Arf 1 and 2). It is also possible that the single Giardia Arf represents the class I/II/III ancestral type, which would indicate that the I/II vs. III duplication occurred after the divergence of Giardia, but before the animal-fungal split. In C. elegans a fourth Arf gene (Ce Arf) has been found. It is very likely a pseudogene, based upon its proximity to another (highly expressed) gene and lack of any evidence of its expression. This makes its phylogenetic placement dubious due to its rapid evolutionary rate (and resulting long branch on the phylogenetic trees). The human Arf III class gene, Arf6, has activity as cofactor for cholera toxin, activation of phospholipase D and rescue of the double arf- null in S. cerevisiae (Lee et al., 1992; Massenburg et al., 1994). It will be interesting to see if the class III representatives from flies, worms and yeasts share these activities. Sc Arf3 has been reported to support cholera toxin activity, albeit weakly, but it cannot substitute for Arf1 in yeast cells (Lee et al., 1994b).

The ARF Family Tree

7

1 26 47 71 HsArf1 MGNIFANLFK-------------GLFGK-KEMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK MmArf2 MGNVFEKLFK-------------SLFGK-KEMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK HsArf3 MGNIFGNLLK-------------SLIGK-KEMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK DmArf1 MGNVFANLFK-------------GLFGK-KEMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK CeArf1 MGNVFGSLFK-------------GLFGK-REMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK SpArf1 MGLSISKLFQ-------------SLFGK-REMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YRNIS-FT--VWDVGGQDK HsArf4 MGLTISSLFS-------------RLFGK-KQMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIC-FT--VWDVGGQDR HsArf5 MGLTVSALFS-------------RIFGK-KQMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIC-FT--VWDVGGQDK DmArf2 MGLTISSLLT-------------RLFGK-KQMRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIC-FT--VWDVGGQDK CeArf3 MGLTISSLFN-------------RLFGK-RQVRILMVGLDAAGKTTILYKLKLGEIV--------TTIPTIGFNVETVE-----YKNIS-FT--VWDVGGQDK ScArf1 MGLFASKLFS-------------NLFGN-KEMRILMVGLDGAGKTTVLYKLKLGEVI--------TTIPTIGFNVETVQ-----YKNIS-FT--VWDVGGQDR ScArf2 MGLYASKLFS-------------NLFGN-KEMRILMVGLDGAGKTTVLYKLKLGEVI--------TTIPTIGFNVETVQ-----YKNIS-FT--VWDVGGQDR CeArf MGLFFSKISS-------------FMFPN-IECRTLMLGLDGAGKTTILYKLKLNETV--------NTIPTIGFNVETVT-----FQKIT-LT--VWDVGGQKK GlArf MGQGASKIFG-------------KLFSK-KEVRILMVGLDAAGKTTILYKLMLGEVV--------TTVPTIGFNVETVE-----YKNIN-FT--VWDVGGQDS HsArf6 MGKVLSKIF-----------------GN-KEMRILMLGLDAAGKTTILYKLKLGQSV--------TTIPTVGFNVETVT-----YKNVK-FN--VWDVGGQDK DmArf3 MGKLLSKIF-----------------GN-KEMRILMLGLDAAGKTTILYKLKLGQSV--------TTIPTVGFNVETVT-----YKNVK-FN--VWDVGGQDK CeArf6 MGKFLSKIF-----------------GK-KELRILMLGLDAAGKTTILYKLKLGQSV--------TTIPTVGFNVETVT-----YKNIK-FN--VWDVGGQDK SpArf2 MGNSLFKGFSKPFS---------RLFSN-KEMRILMLGLDAAGKTTILYKLKLNQSV--------VTIPTVGFNVETVT-----YKNIK-FN--VWDVGGQDK ScArf3 MGNSISKVLG-------------KLFGS-KEMKILMLGLDKAGKTTILYKLKLNKIK--------TSTPTVGFNVETVT-----YKNVK-FN--MWDVGGQQR HsArl1 MGGFFSSIFS-------------SLFGT-REMRILILGLDGAGKTTILYRLQVGEVV--------TTIPTIGFNVETVT-----YKNLK-FQ--VWDLGGQTS DmArl1 MGGVLSY-FR-------------GLLGS-REMRILILGLDGAGKTTILYRLQVGEVV--------TTIPTIGFNVEQVT-----YKNLK-FQ--VWDLGGQTS CeArl1 MGGVMSY-FR-------------GLFGA-REMRILILGLDGAGKTTILYRLQVGEVV--------TTIPTIGFNVEQVE-----YKNLK-FQ--VWDLGGQTS ScArl1 MGNIFSSMFD-------------KLWGSNKELRILILGLDGAGKTTILYRLQIGEVV--------TTKPTIGFNVETLS-----YKNLK-LN--VWDLGGQTS HsArl5 MGILFTRIW--------------RLFNH-QEHKVIIVGLDNAGKTTILYQFSMNEVV--------HTSPTIGSNVEEIV-----INNTR-FL--MWDIGGQES HsArl5B MGLIFAKLW--------------SLFCN-QEHKVIIVGLDNAGKTTILYQFLMNEVV--------HTSPTIGSNVEEIV-----VKNTH-FL--MWDIGGQES DmArl5 MGLLLSRLW--------------RMFGN-EEHKLVMVGLDNAGKTTILYQFLMNEVV--------HTSPTIGSNVEEVV-----WRNIH-FL--VWDLGGQQS CeArl5 MGLIMAKLFQ-------------SWWIG-KKYKIIVVGLDNAGKTTILYNYVTKDQV--------ETKPTIGSNVEEVS-----YRNLD-FV--IWDIGGQES HsArl4 MGNGLSDQTSILS--------NLPSFQS---FHIVILGLDCAGKTTVLYRLQFNEFV--------NTVPTKGFNTEKIKVTLGNSKTVT-FH--FWDVGGQEK HsArl4B MGNGLSDQTSILS--------SLPSFQS---FHIVMLGLDCAGKTTVLYRLQFNEFV--------NTVPTKAFNTEKIKVNLRNSKTVT-FH--FGDVGGQEK HsArl7 MGNISS---------------NISAFQS---LHIVMLGLDSAGKTTVLYRLKFNEFV--------NTVPTIGFNTEKIKLSNGTAKGIS-CH--FWDVGGQEK HsArl9 MGNHLTEMAPTASS-------FLPHFQA---LHVVVIGLDSAGKTSLLYRLKFKEFV--------QSVPTKGFNTEKIRVPLGGSRGIT-FQ--VWDVGGQEK DmArl4 MGATMVKPLVKNGNI------LDALPSQ-ATLHVVMLGLDSAGKTTALYRLKFDQYL--------NTVPTIGFNCEKVQCTLGKAKGVH-FL--VWDVGGQEK HsArl10 MGSLGSKNPQT------------------KQAQVLLLGLDSAGKSTLLYKLKLAKDI--------TTIPTIGFNVEMIEL----ERNLS-LT--VWDVGGQEK HsArl11 MGSVNSRGHK-------------------AEAQVVMMGLDSAGKTTLLYKLKGHQLV--------ETLPTVGFNVEPLKA----PGHVS-LT--LWDVGGQAP HsArl2 MGLLTILKKMKQ--------------KE-RELRLLMLGLDNAGKTTILKKFNGEDID--------TISPTLGFNIKTLE-----HRGFK-LN--IWDVGGQKS DmArl2 MGFLTVLKKMRQ--------------KE-REMRILLLGLDNAGKTTILKRFNGEPID--------TISPTLGFNIKTLE-----HNGYT-LN--MWDVGGQKS CeArl2 MGFLKILRKQRA--------------RE-REMRILILGLDNAGKTTLMKKFLDEPTD--------TIEPTLGFDIKTVH-----FKDFQ-LN--LWDVGGQKS SpArl2 MGLLTILRQQKL--------------KE-REVRVLLLGLDNAGKTTILKCLLNEDVN--------EVSPTFGFQIRTLE-----VEGLR-FT--IWDIGGQKT ScArl2 MGLLSIIRKQKL--------------RD-KEIRCLILGLDNSGKSTIVNKLLPKDEQ-----NNDGIMPTVGFQIHSLM-----IKDVT-IS--LWDIGGQRT GlArl2 MGFLSVLRAIKA--------------SE-NEARVLVVGLDCSGKTTIVSSLLGKPLT--------EIAPTLGFKISTWR-----PQNTS-MAVALWDVGGQQT HsArl3 MGLLSILRKLKSA-------------PD-QEVRILLLGLDNAGKTTLLKQLASEDIS--------HITPTQGFNIKSVQS-----QGFK-LN--VWDIGGQRK DmArl3 MGLLSLLRKLRPN-------------PE-KEARILLLGLDNAGKTTILKQLASEDIT--------TVTPTAGFNIKSVAA-----DGFK-LN--VWDIGGQWK CeArl3 MGLIDVLKSFKSP-------------SG-REIRILLLGLDNAGKTTILKQLSSEDVQ--------HVTPTKGFNVKTVAA----MGDIR-LN--VWDIGGQRS GlArl3 TILRKLCEEDTS--------VIMPTQGFNAKTIKG----PKGVS-LN--CWDVGGQKA HsArl8A MIALFNKLLDW----------FKALFWK-EEMELTLVGLQYSGKTTFVNVIASGQFN-------EDMIPTVGFNMRKIT-----KGNVT-IK--LWDIGGQPR HsArl8B MLALISRLLDW----------FRSLFWK-EEMELTLVGLQYSGKTTFVNVIASGQFS-------EDMIPTVGFNMRKVT-----KGNVT-IK--IWDIGGQPR DmArl8 MLALINRILEW----------FKSIFWK-EEMELTLVGLQFSGKTTFVNVIASGQFA-------EDMIPTVGFNMRKIT-----RGNVT-IK--VWDIGGQPR CeArl8 MLAMVNKVLDW----------IRSLFWK-EEMELTLVGLQNSGKTTFVNVIASGQFT-------EDMIPTVGFNMRKIT-----KGNVT-IK--LWDIGGQPR GlArfrp TTILHQMAYGMT--------VETIPTMGFTLQTVK-----KGKLE-LD--VWDIGGQSE HsArfrp MYTLLSGL-------------YKYMFQK-DEYCILILGLDNAGKTTFLEQSKTRFNKNYKGMSLSKITTTVGLNIGTVD-----VGKAR-LM--FWDLGGQEE DmArfrp MYTLMHGF-------------YKYMTQK-DDYCVVILGLDNAGKTTYLEAAKTTFTRNYKGLNPSKITTTVGLNIGTID-----VQGVR-LN--FWDLGGQQE CeArfrp MYTLSAGI-------------WEKFFKK-KDYYVVIVGLDNAGKTTFLEQTKSHFVKDYGVLNPSKITATVGLNTGNVE-----LNNTC-LH--FWDLGGQES ScArfrp MFHLVKGL-------------YNNWNKK-EQYSILILGLDNAGKTTFLETLKKEYSL--AFKALEKIQPTVGQNVATIP-----VDSKQILK--FWDVGGQES HsArl6 MGLLDRLSV------------LLGL-KK-KEVHVLCLGLDNSGKTTIINKLKPSNAQ------SQNILPTIGFSIEKFK-----SSSLS-FT--VFDMSGQGR DmArl6 MGMLHNLAD------------LIKI-KK-DKMTILVLGLNNSGKSSIINHFKKSSEQ------TSIVVPTVGFMVEQFYI---GMSGVS-IK--AIDMSGATR CeArl6 MGFFSSLSS------------LFGL-GK-KDVNIVVVGLDNSGKTTILNQLKTPETR------SQQIVPTVGHVVTNFS-----TQNLS-FH--AFDMAGQMK GlArl MGAASSRPA-----------------RH-KTAKLLILGRQFSGKSTLINYLKTNSFT--------VVNPTINFTAEKYN---------R-YE--VIDVGGSAE HsSar1 MSFIFEWIYNG----FSSVLQFLGL-YK-KSGKLVFLGLDNAGKTTLLHMLKDDRLG--------QHVPTLHPTSEELT-----IAGMT-FT--TFDLGGHEQ HsSara MSFIFDWIYSG----FSSVLQFLGL-YK-KTGKLVFLGLDNAGKTTLLHMLKDDRLG--------QHVPTLHPTSEELT-----IAGMT-FT--TFDLGGHVQ CeSar1 MSFLWDW--------FNGVLNMLGL-AN-KKGKLVFLGLDNAGKTTLLHMLKDDRIA--------QHVPTLHPTSEQMS-----LGGIS-FT--TYDLGGHAQ DmSar1 M-FIWDW--------FTGVLGYLGL-WK-KSGKLLFLGLDNAGKTTLLHMLKDDKLA--------QHVPTLHPTSEELS-----IGNMR-FT--TFDLGGHTQ ScSar1 M-AGWDIFGW-----FRDVLASLGL-WN-KHGKLLFLGLDNAGKTTLLHMLKNDRLA--------TLQPTWHPTSEELA-----IGNIK-FT--TFDLGGHIQ SpSar1 M-FIINW--------FYDALAMLGL-VN-KHAKMLFLGLDNAGKTTLLHMLKNDRLA--------VMQPTLHPTSEELA-----IGNVR-FT--TFDLGGHQQ GlSar1 M-GFGDW--------FKSALSFLGL-YK-KKATIVFVGLDNAGKSTLLAMLKNSATT--------TVAPTQQPTSQELV-----MGSIR-FK--TFDLGGHEV EXCLUDED XXXXXXXXXXXXXXXXXXX X XXXXXXXX XXXXX X XX

Figure 1. Protein alignment of the N-terminal region of the 64 members of the Arf family described in this chapter. A dash (-) indicates a gap introduced for alignment; white space indicates that two sequences are incomplete. Numbers above indicate residues in the unedited hArf1 and indicate critical residues and/or simply regular spacing. Positions excluded from the alignment used for phylogenetic analyses (Figure 2) are indicated below.

HsArf1 MmArf2 HsArf3 DmArf1 CeArf1 SpArf1 HsArf4 HsArf5 DmArf2 CeArf3 ScArf1 ScArf2 CeArf GlArf HsArf6 DmArf3 CeArf6 SpArf2 ScArf3 HsArl1 DmArl1 CeArl1 ScArl1 HsArl5 HsArl5B DmArl5 CeArl5 HsArl4 HsArl4B HsArl7 HsArl9 DmArl4 HsArl10 HsArl11 HsArl2 DmArl2 CeArl2 SpArl2 ScArl2 GlArl2 HsArl3 DmArl3 CeArl3 GlArl3 HsArl8A HsArl8B DmArl8 CeArl8 GlArfrp HsArfrp DmArfrp CeArfrp ScArfrp HsArl6 DmArl6 CeArl6 GlArl HsSar1 HsSara CeSar1 DmSar1 ScSar1 SpSar1 GlSar1

126 160 178 1 26 47 71 101 MGNIFANGKKEMRILMVGLDAAGKTTILYKLKLGEIVTTIPTIGFNVETVEYKNISFTVWDVGGQDKIRPLWRHYFQNTQGLIFVVDSNDRERVNEAREELMRMLAEDELRDAVLLVFANKQDLPNAMNAAEITDKLGLHSLRHRNWYIQATCATSGDGLYEGLDWLSNQLR ...V.EK..............................................................................................T................V........................Q...........................K .....G.................................................................................................................................................................A...K ...V.........................................................................................IG.....................I..........................N...........................K ...V.GS..R....................................................................................G..............................Q......V..........N.S.........................K ..LSISK..R..........................................R............................I...........IS..H...Q...N.......L...............................Q..................E...TN.K ..LTISS...Q............................................C..........R.....K....................IQ.VAD..QK..LV.........L...........AIS.M......Q...N.T..V......Q.T...........E.S ..LTVSA...Q............................................C......................................Q.SAD..QK..Q.................M....PVS.L......QH..S.T..V......Q.T...D......HE.S ..LTISS...Q............................................C...................................D.IT..ER..QN..Q......................T...L....R.NQ..N.H.F..S....Q.H..........AE.A ..LTISS..RQV..............................................................................K..IE.S....HK..N.......T..............T...L.......N..S.Q.........Q.H.............S ..LFASK.N...........G.....V........VI.............Q...............R..S.....YR..E.V.........S.IG....VMQ...N.....N.AW..........E..S.....E......I.N.P.F.........E......E....S.K ..LYASK.N...........G.....V........VI.............Q...............R..S.....YR..E.V...I.....S.IG....VMQ...N.....N..W..........E..S.....E......I.N.P.F..S......E......E....N.K ..LF.SKPNI.C.T..L...G............N.T.N............TFQK.TL........K...A..KY..P..TT.V.....S.I..IP..K...FSL...P..A.SH.........M...RSP..L.QL.D.G..KN.E.F.CG.N.H..Q......M.VKK.MK ..QGASKS...V...................M...V...V...............N..........S........Y...DA..Y.I..A.L..IED.KN..HTL.G.......A...........K..STTDL.ER...QE.KK.D....P...R......Q......DYIF ..KVLSK.N.......L.................QS......V.......T...VK.N.................YTG.........CA..D.ID...Q..H.IINDR.M...II.I........D..KPH..QE....TRI.D....V.PS............T..TSNYK ..KLLSK.N.......L.................QS......V.......T...VK.N.................YTG.........CA..D.ID...T..H.IINDR.M...II.I........D..KPH..QE....TRI.D....V.PS........S...I..TSNHK ..KFLSK....L....L.................QS......V.......T....K.N.................YTG..A....M.AA..D..D...M..H.IINDR.MKE.II.........AD..KPH..Q.....TRI.D....V.PS..ST....H...T...QNCK ...SLFKSN.......L................NQS.V....V.......T....K.N..................TG.K........A.SN.IS...Q..H.IISDR.M..CL...L.......G.LSP.Q...V.Q.DK.KD.L.NV.P...LT....L...A...QNAK ...SISK.S...K...L...K............NK.K.ST..V.......T...VK.NM......QRL........PA.TA....I..SA.N.ME..K...YSIIG.K.MENV....W......KD..KPQ.VS.F.E.EN.KNQP.CVIGSN.L..Q..V...S.I..NTN ..GF.SS.TR.....IL...G........R.QV..V..............T...LK.Q...L...TS...Y..C.YS..DAV.Y....C..D.IGISKS..VA..E.E...K.I.V.......MEQ..TSS.MANS...PA.KD.K.Q.FK.S..K.T..D.AME..VET.K ..GVLSY.SR.....IL...G........R.QV..V............Q.T...LK.Q...L...TS...Y..C.YS..DAI.Y....A..D.IGISKD..LY..R.E..AG.I.V.L.....MDGC.TV..VHHA...EN.KN.TFQ.FK.S..K.E..DQAM.....T.Q ..GVMSY.AR.....IL...G........R.QV..V............Q.....LK.Q...L...TS...Y..C.YA..DAI.Y....A..D..GIS.Q..AT..Q....QG...A.L.....IAGCLTET.VYKA...DA..N.TIQ.FK.S.SK.E..DPAM...A...Q .....SS.S..L...IL...G........R.QI..V...K.........LS...LKLN...L...TS...Y..C.YAD.AAV......T.KD.MST.SK..HL..Q.E..Q..A.........Q.G.LS.S.VSKE.N.VE.KD.S.S.V.SS.IK.E.IT......IDVIK ..IL.TRNHQ.HKVII....N........QFSMN.V.H.S....S...EIVIN.TR.LM..I...ESL.SS.NT.YT..EFV.V....T....ISVT....YK...HED..K.G..I......VKEC.TV...SQF.K.T.IKDHQ.H...C..LT.E..CQ..E.MMSR.K ..L...KCNQ.HKVII....N........QFLMN.V.H.S....S...EIVV..TH.LM..I...ESL.SS.NT.YS..EFI.L....I....LAITK...Y....HED..K.AV.I......MKGC.T....SKY.T.S.IKDHP.H..SC..LT.E..CQ..E.MTSRIG ..LLLSR.NE.HKLV.....N........QFLMN.V.H.S....S...E.VWR..H.L...L...QSL.AA.ST.YT..ELV.M.I..T....LAVT....Y...QHED.SK.S...Y......KGS.S....SRQ.D.T.IKKHQ.H...C..LT.E...Q..E.IVQRIK ..L.M.KIG.KYK.IV....N........NYVTKDQ.E.K....S...E.S.R.LD.VI..I...ESL.KS.ST.YVQ.DVV.V.I..S.TT.IPIMK.Q.HN..QHED.AR.HI..L.......G...P..VSTQ...QT..A.K.Q.NGC..VK.E..P.A.E.IA.N....GLSDQS--FH.VIL...C.....V..R.QFN.F.N.V..K...T.KIKS.TVT.HF......E.L....KS.TRC.D.IV.....V.V..ME..KT..HKITRIS.NQGVPV.IV......R.SLSLS..EKL.AMGE.SSTP.HL.P...II....K...EK.HDMII ...GLSDQS--FH.V.L...C.....V..R.QFN.F.N.V..KA..T.KIKS.TVT.HFG.....E.LM...KS.TRC.D.IL.LM..V.I..ME..KT..HKITRLS.NQGVPV.TV......E.SLSLSG.EKL.ATGE.SSTP.HL.P...II....K...EK.HDMII ....SS-QS--LH.V.L...S.....V..R..FN.F.N.V......T.KIKA.G..CHF......E.L....KS.SRC.D.I.Y....V.VD.LE..KT..HKVTKFA.NQGTP...I.......KSLPV...EKQ.A..E.IATTYHV.PA..II.E..T..M.K.YEMIL ...HLTEQA--LHVVVI...S....SL..R..FK.F.QSV..K...T.KIRSRG.T.Q.......E.L.....S.NRR.D..V....AAEA..LE..KV..H.ISRASDNQGVPV..L.....Q.G.LS...VEKR.AVRE.AATLTHV.GCS.VD.L..QQ..ER.YEMIL ..ATMVKSQATLHVV.L...S.....A..R..FDQYLN.V......C.K.QA.GVH.L.......E.L.....S.TRC.D.IL..I..V.T..ME..KM....TAKCPDNQGVPV.IL.........CG.M.LEKL...NE.--.G....P...IT.E..Q....A.YDMIL ..SLGSK--.QAQV.LL...S...S.L......AKDI...........MI.ER.L.L........E.M.TV.GC.CE..D..VY....T.KQ.LE.SQRQFEHI.KNEHIKNVPVVLL.....M.G.LT.ED..RMFKVKK.-D....V.PC..LT.E..AQ.FRK.TGFVK ..SVNSR--A.AQVV.M...S.....L.....GHQL.E.L..V.....PLKPGHV.L.L......APL.AS.KD.LEG.DI.VY.L..T.EA.LP.SAA..TEV.NDPNMAGVPF..L....EA.D.LPLLK.RNR.S.ERFQDHC.ELRGCS.LT.E..P.A.QS.WSL.K ..LLTILKER.L.L..L...N.......K.FNGED.D.IS..L...IK.L.HRGFKLNI......KSL.SY..N..ES.D...W....A..Q.MQDCQR..QSL.V.ER.AG.T..I........G.LSSNA.REA.E.D.I.SHH.C..GCS.VT.EN.LP.I...LDDIS ..FLTVLKER.....LL...N.......KRFNGEP.D.IS..L...IK.L.HNGYTLNM......KSL.SY..N..ES.D..VW....A..M.LESCGQ..QVL.Q.ER.AG.T...LC......G.LSSN..KEI.H.EDI-THH.LVAGVS.VT.EK.LSSM...IADIA ..FLKILRER.....IL...N.....LMK.FLDEPTD.IE..L..DIK..HF.DFQLNL......KSL.SY.KN..ES.DA..W....S....LLQCS...KKL.G.ER.AG.S...L...S...G.IDVNS.AQV.D...I-SHH.K.FSC..L...R.VQAMT..CDDVG ..LLTILKER.V.V.LL...N.......KC.LNEDVNEVS..F..QIR.L.VEGLR..I..I...KTL.NF.KN..ES.EAI.W....L.DL.LE.C.NT.QEL.V.EK.LFTSI..L...S.VSG.LSSE..SKI.NISKYKSSH.R.FSVS.LT.LNIKDAIS..A.D.K ..LLSIIRD..I.C.IL...NS..S..VN..LPKDEQGIM..V..QIHSLMI.DVTISL..I...RTL..F.DN..DK..AM.WCI.VSLSM.FD.TLQ..KELINR..N.ECAVI.VL..I..VEDKSELH-RRC.LVEE.KDIRIELVKCSGVT.E.ID----N.RDR.V ..FLSVLSEN.A.V.V....CS.....VSS.LGKPLTEIA..L..KIS.WRPQ.T.MAL......QT..TY..N..SS.DA..W....T..R.MDTC.KA.SEV.HAER.QGCPVMILC....I.S.LSVE..SAAIT.DA.-N.K.AVLPCS.M.RQ..D.AFS..LSE.L ..LLSILPDQ.V...LL...N.....L.KQ.ASED.SHIT..Q...IKS.Q-QGFKLN...I...R....Y.KN..E..DI..Y.I..A..K.FE.TGQ..AEL.E.EK.SCVPV.I.......LT.AP.S..AEG.N..TI.D.V.Q..SCS.LT.E.VQD.MN.VCKNVN ..LLSLLPE..A...LL...N.......KQ.ASED.T.VT..A...IKS.A-DGFKLN...I...W....Y.KN..A..DV..Y.I.CT..T.LP..GS..FE..MDNR.KQVPV.I......M.D..S...VAE.MS.VQ.QG.T.E.K.CT.VD.T..K..M..VCKNMK ..L.DVLSGR.I...LL...N.......KQ.SSEDVQHVT..K....K..AMGD.RLN...I...RS...Y.SN.YE.IDT....I....KK.FD.MNI..GEL.D.EK..KVPV.I.......VT.ASSE...R..N.DL..D.T.H...CS.LKNE.IND.IT.VASN.K -------------------------...R..CEEDTSVIM..Q...AK.IKP.GV.LNC......KA..TY.QN.YAG.TC..W...AA.VK.MD.VAL.VSLA.EDCR.AGVPI.........AT.LKPSDLC..I..WA..D.V.H..GCS.LT.E..E..IL.MVSKTD .IAL.NKW.E..ELTL...QYS....FVNVIAS.QFNDM...V...MRKITKG.VTIKL..I...PRF.SM.ER.CRGVSAIVYM..AA.Q.KIEASKN..HNL.DKPQ.QGIPV..LG..R...G.LDEK.LIE.MN.SAIQD.EICCYSISCKEK.NIDIT.Q..IQHSK .LALISRW.E..ELTL...QYS....FVNVIAS.QFSDM...V...MRK.TKG.VTIKI..I...PRF.SM.ER.CRGVNAIVYMI.AA...KIEAS.N..HNL.DKPQ.QGIPV..LG..R.....LDEKQLIE.MN.SAIQD.EICCYSISCKEK.NIDIT.Q..IQHSK .LALINRW.E..ELTL...QFS....FVNVIAS.QFADM...V...MRKITRG.VTIK...I...PRF.SM.ER.CRGVNAIVYM..AA.LDKLEAS.N..HSL.DKPQ.AGIPV..LG..R...G.LDETGLIERMN.S.IQD.EICCYSISCKEK.NIDIT.Q..IQHSK .LAMVNKW.E..ELTL...QNS....FVNVIAS.QFTDM...V...MRKITKG.VTIKL..I...PRF.SM.ER.CRGVNAIV.M..AA.E.KLEAS.N...QL.DKPQ.DAIPV..LG..K...G.LDERQLIERMN.S.IQN.EICCYSISCKEKENIDIT.Q..IDHSK ------------------------....HQMAY.MT-E....M..TLQ..KKGKLELD...I...SEF.NI.V..YVDKHAA.....AA.KA.ME...TA.EGV.TAP..SGVPI.IL.....IDG..SGDAVASM.D.QNVKDHEVKVVE.VG.T.K..DDA.K...TAIE .YTLLSGQ.D.YC..IL...N.....F.EQS.TRFNKKITT.V.L.IG..DVGKARLMF..L...EELQS..DK.YAECH.V.Y.I..T.E..LA.SKQAFEKVVTSEA.CGVPV..L.....VETCLSIPD.KTAFSDSKIGR.DCLT..CS.LT.K.VR..IE.MVKCVV .YTLMHGQ.DDYCVVIL...N.....Y.EAA.TTFTRKITT.V.L.IG.IDVQGVRLNF..L...QELQS..DK.Y.ESH.V.Y.I.......ME.SKAIFDK.IKNEL.SGVP..IL.......DV.GVR..KPVFQQALIGR.DCLTIPVS.LH.E.VD..IK..VEAIK .YTLS.GK..DYYVVI....N.....F.EQT.SHFVKKITA.V.L.TGN..LN.TCLHF..L...ESL.E..AT.YDDANAM.....ATRSDLFPIVASQFKEVM.NEIVQNIPV..AV..SEMEG.AA...VRML.EDDNH.-SDLAVLPVS.LE.TNIERCVH.IVRS.A .FHLVKGK.EQYS..IL...N.....F.ET..KEYSLKIQ..V.Q..A.IPVDSKQLKF......ESL.SM.SE.YSLCH.I..I...S....LD.CSTT.QSVVMDE.IEGVPI.ML.....RQDR.EVQD.KEVFNKEHISA.DSRVLPIS.LT.E.VKDAIE.MIVR.E ..LLDRLK...VHV.CL...NS.....IN...PSNAQNIL.....SI.KFKSSSL....F.MS..GRY.N..E..YKEG.AI...I..S..L.MVV.K...DTL.NHPDIKRIPI.F....M..RD.VTSVKVSQL.C.ENIKDKP.H.C.SD.IK.E..Q..V...QD.IQ ..MLHNLK.DK.T..VL..NNS..SS.INHF.KSSEQIVV..V..M..QFYMSGV.IKAI.MS.ATRY.N..E.Q.K.CH.I.Y.I..S..M.FVVVKD..DLV.QHPD.CIVPI.FYG..M.MEDSLSSVK.AAA.R.ENIKDKP.H.CSSS.I..E..G..VQ..IQ.M. ..FFSSL...DVN.VV....NS......NQ..TP.TRQIV..V.HV.TNFSTQ.L..HAF.MA..M.Y.ST.ES..HSS..V...L..S..L.MELLKD...MVMEHKDVVGIPIVIL...M.I.G..T.SD..VA...NLY.SGT.S.HS...LT....DKAMQQ..AEIT ..AASSRRH.TAKL.IL.RQFS..S.LINY..TNSFTVVN...N.TA.KYN----RYE.I....SAEMMQ..SEHYSG..TCL....GTS--PLDQ..DLIKDI.GHTDM.ASRFA.VVT.I.Q.GVSQK.DYYRT-T.QLDKD.VQEVFMF.GLD.Q.CDKIAQ..--DVP .SF..EWY..SGKLVFL...N.....L.HM..DDRLGQHV..LHPTS.ELTIAGMT..TF.L..HEQA.RV.KN.LPAIN.IV.L..CA.HS.LV.SKV..NALMTDETISNVPI.ILG..I.RTD.ISEEKLREIF..YGQTA.PMEVFMCSVLKRQ.YG..FR...QYID .SF..DWY..TGKLVFL...N.....L.HM..DDRLGQHV..LHPTS.ELTIAGMT..TF.L..HVQA.RV.KN.LPAIN.IV.L..CA.H..LL.SK...DSLMTDETIANVPI.ILG..I.R.E.ISEERLREMF..YGQTA.PLEVFMCSVLKRQ.YG..FR.MAQYID .SFLWDWAN.KGKLVFL...N.....L.HM..DDR.AQHV..LHPTS.QMSLGG....TY.L..HAQA.RV.KD..PAVDAVV.LI.VA.A..MQ.S.V..ESL.QDEQIASVPV.ILG..I.K.G.LSEDQLKWQ.NIQHMCS.PMEVFMCSVLQRQ.YG..IR..GQY..-F.WDWW..SGKL.FL...N.....L.HM..DDKLAQHV..LHPTS.ELSIG.MR..TF.L..HTQA.RV.KD..PAVDAIV.LI.AW..G.FQ.SKN..DSL.TDEA.SNCPV.ILG..I.K.G.ASED.LRNVF..YQ.TG.PLELFMCSVLKRQ.YG..FR..AQYID .-AGWDIWN.HGKL.FL...N.....L.HM..NDRLA.LQ..WHPTS.ELAIG..K..TF.L..HIQA.R..KD..PEVN.IV.L..AA.P..FD...V..DALFNIA..K.VPFVILG..I.A...VSE..LRSA...LNTTQ.PVEVFMCSVVMRN.YL.AFQ...QYI.-F.INWVN.HAKM.FL...N.....L.HM..NDRLAVMQ..LHPTS.ELAIG.VR..TF.L..HQQA.R...D..PEVN.IVYL..CC.F..LS.SKA..DAL..ME..ARVPF.ILG..I.A.G.ISED.LKAA...YQTTI.PIEVFMCSVVLRQ.YG..FK..AQYV.-GFGDWY..KAT.VF....N...S.L.AM..NSATT.VA..QQPTSQELVMGS.R.KTF.L..HEVA.Q..EQ.VT.SD.IV.L...A.PS.FE.S.RT.QEL.DNHD.ATTPI.ILS..V.IQT.VSMETMVQSF.IQH.LQ.PLEVFPCSVINRF.YTD.FK...KY.-

8 LOGSDON AND KAHN

Figure 2. Protein alignment of the 64 members of the Arf family used for the phylogenetic analyses. Protein sequences were aligned and edited to remove non-conserved insertions. Lines are shown to delineate between sub-families. A dot (.) indicates identity with the top sequence (human Arf1), while other conventions are as in Figure 1.

The ARF Family Tree

9

Figure 3. Phylogenetic tree of Arf family homologs from six model eukaryotic organisms. The consensus Bayesian maximum likelihood tree is shown. The tree is based on the alignment of Arf/Arl protein sequences (64 sequences, 172 aligned amino-acid sites) shown in Figure 2. The analysis used MrBayes 2.01 (Huelsenbeck and Ronquist, 2001), using a JTT model with nine gamma-distributed rates for 80400 generations (including 20,000 for burnin) and sampling frequency of 1000 generations. The resulting parameter estimates (and their 95% confidence intervals) are: LnL –13164.546 (80.502), alpha 1.585 (0.0261). The numbers on the branches are the consensus percentages (clade confidences) among the 785 “good” (post burn-in) trees; bullets represent 99-100%. The scale bar represents amino acid substitutions per site.

10

LOGSDON AND KAHN

This further demonstrates the limitations of functional data as the sole criterion of classification and naming these genes/proteins. Results from studies of mammalian Arf6 support the conclusions that class III Arfs share a degree of functional overlap with class I/II Arfs but have clearly distinctive features as well (see Chapters 14 and 15 for details). Despite the fact that cows, mice, and rats all have six Arf paralogs, the Arf2 gene appears to be missing from humans. Given the current state of completeness of human genome sequencing, combined with numerous attempts to isolate Arf2 from humans (both experimentally and using bioinformatics), it may be time to conclude that Arf2 has been lost from the human lineage. Because mammalian Arf1-3 protein sequences are identical between mammals, all are >96% identical to one another, and are biochemically indistinguishable, there is no obvious need for this redundancy. However, recent results (K. Hill and R. A. Kahn; unpublished observations) establish functional differences in human cells between Arf1 and Arf3 and are supportive of distinct functions for each Arf. It is also very likely that functional differences exist between the class I and II Arfs but these have only begun to be described.

2.2

The Arl sub-families

In addition to the three Arf classes, our analysis helps to delineate ten orthologous groups that collectively form the Arls: Arls1-5, Arl8, Arl9, Arfrp, and Sar1. Each of these groups is discussed below. For more details on the biology of Arl proteins, see Chapter 16 by Schurmann and Joost. 2.2.1

Arl1

Single representatives of the Arl1 proteins were found in human, fly, worm, and S. cerevisiae proteomes. Mammalian Arl1s have previously been shown to have slightly higher percent identity (~55%) with Arfs than do other Arls and to share some binding partners (Van Valkenburgh et al., 2001). The phylogenetic placement of Arl1 as the closest to the Arfs is consistent with the idea that they are close relatives, both functionally and evolutionarily. The first of these Arls described was the fly arflike, which is essential in flies (Tamkun et al., 1991). The lack of Arf activities and essential nature of the fly Arl1 revealed for the first time that Arls perform essential functions that can be clearly distinguished from those of Arfs. Arl1 has been localized to Golgi membranes (Lowe et al., 1996) and expression of dominant activating or inactivating mutants in mammalian cells results in changes in Golgi morphology that are highly reminiscent of, but of lesser magnitude than, those resulting from expression of analogous Arf mutants (Lu et al., 2001;

The ARF Family Tree

11

Van Valkenburgh et al., 2001). Studies of S. cerevisiae ARL1 are also consistent with the idea that it regulates aspects of vesicular traffic but cannot substitute for Arfs (Lee et al., 1997; Rosenwald et al., 2002). All Arl1 proteins contain a glycine at position 2 and have either been shown to be N-myristoylated or are predicted to be so modified (Lee et al., 1997; Randazzo and Kahn, 1995). A structure has been solved for ScArl1 that revealed differences in both the electrostatics and in the structure of the N-terminal region when compared to Arfs (Amor et al., 2001), though the biological significance of these observations is uncertain. 2.2.2

Arl2

Arl2 proteins were identified in all six organisms included in this study, including Giardia. This argues for their early origin in eukaryotic evolution. In mammals, S. cerevisiae (Arl2/CIN4), S. pombe (Arl2/alp41), C. elegans (Arl2/evl-20), and A. thaliana (Arl2/TITAN5), there is genetic evidence supporting roles for Arl2 in tubulin assembly and microtubule dynamics (Antoshechkin and Han, 2002; Bhamidipati et al., 2000; Hoyt et al., 1997; Hoyt et al., 1990; McElver et al., 2000; Radcliffe et al., 2000). Arl2 mutations result in cytokinesis defects and death in most of these studies, though Arl2/CIN4 is not essential in S. cerevisiae. A model for Arl2 acting in tubulin folding has been proposed by Bhamidipati, et al. (Bhamidipati et al., 2000), based on the evidence for the direct binding of the tubulin-specific chaperone, cofactor D, by Arl2•GDP. Based on all of the phylogenetic analyses we have carried out to date, the Arl2 and Arl3 proteins group strongly together, making it very likely that they diverged from a common ancestor. Indeed, this must have been an early event in eukaryotic evolution as Giardia has representatives of each group. In addition, each of the yeasts in our study appears to have lost one of the Arl2/Arl3 group proteins. This appears to have been Arl3, though the long branch lengths (e.g., see Sc Arl2 in Figure 3) make this conclusion somewhat tenuous. There is also biochemical support for the conclusion that Arl2 and Arl3 are more closely related to each other than to other Arls. In particular, guanine nucleotide exchange is much more rapid, proceeds to higher stoichiometries in vitro, and is independent of lipids or detergents; properties that are distinct from those of Arfs and Arl1 (Cavenagh et al., 1994; Clark et al., 1993; Hillig et al., 2000; Sharer and Kahn, 1999). Structural data explain these differences in divalent metal and nucleotide binding by Arl2 and Arl3, compared to other members of the Arf family (Hillig et al., 2000), and further support the conclusion that Arl2 and Arl3 are more closely related to each other than they are to other Arls.

12

LOGSDON AND KAHN

Figure 4. Schematic phylogenetic tree depicting the relationships between the three Arf and ten Arl groups. The presence of orthologs of each group in animals (A, in triangle), fungi (F, in box), or protists (P, in circle) is indicated on the right. In two cases, a Giardia protein could not be unambiguously assigned between groups so arrows to each are shown with a question mark.

These biochemical properties allowed the isolation of a specific Arl2•GTP binding partner (binder of Arl2, BART) that does not bind Arfs or Arl1 (Sharer and Kahn, 1999). Arl2•GTP also binds the delta subunit of phosphodiesterase 6, and its homolog HRG4, and has been crystallized as a co-complex (Hanzal-Bayer et al., 2002; Kobayashi et al., 2003; Renault et

The ARF Family Tree

13

al., 2001; Van Valkenburgh et al., 2001). Arl2 was localized to mitochondria (Sharer et al., 2001), cytosol, and plasma membranes (J. Shern and R. A. Kahn; unpublished observations), suggesting roles for Arl2 in addition to tubulin assembly. A future challenge is to define how these different binding partners and localization patterns relate to the role for Arl2 in tubulin dynamics or other cellular functions. The most prevalent covalent modification to members of the Arf family, N-myristoylation, appears to be absent in both Arl2 and Arl3 (Randazzo and Kahn, 1995; Sharer et al., 2001; Van Valkenburgh and Kahn, 2002). Because of the central role of the N-myristate in the rapidly reversible association of Arfs with membranes, we infer that Arl2 and Arl3 lack this feature as well. This is consistent with the findings that a subset of Arl2 is imported into mitochondria (Sharer et al., 2001) and that the majority of Arl2 is in a high molecular weight, cytosolic complex (J. Shern and R. A. Kahn; unpublished observation). 2.2.3

Arl3

The presence of orthologs of Arl3 in Giardia and the three animals suggest that the two yeasts have lost the Arl3 gene. Whether this is the result of the lack of a particular pathway or specific function (e.g., cell motility) in yeast cannot be determined without more functional data on Arl3. Like Arl2, Arl3 probably is not N-myristoylated. Despite sharing a number of binding partners (Van Valkenburgh et al., 2001) and biochemical features (Hillig et al., 2000) with Arl2, Arl3 appears to lack the inferred functions of Arl2 in tubulin assembly and in mitochondrial localization (C. Snelson, J. Shern, and R. A. Kahn, unpublished observations). 2.2.4

Arl4

The Arl4 group is unusual in having four paralogs in human, a single ortholog in the fly, and none in any of the other genomes searched. Arl4, Arl4B, and Arl7 appear to have diverged from a common ancestor more recently than the divergence of this family from flies, and may be more closely related functionally as well. This is supported by the observations that these three Arls all exchange guanine nucleotides more rapidly than does Arf1 and all localize to the nucleus (Jacobs et al., 1999). Mouse knockouts of Arl4 and Arfrp (see below) are the first reported deletions for any Arf family members in mammals. The impact of those deletions on development of the animal and other aspects of Arl4 biology are described in detail in Chapter 16. These roles in mammalian development are consistent with the recent divergence and expansion of the Arl4 genes.

14 2.2.5

LOGSDON AND KAHN Arl5

Representatives of the Arl5 group were identified in all three animals, with two closely related paralogs in humans. Wormbase (http://www.wormbase.org/) reports no phenotype resulting from RNAi of Ce Arl5, and there is no functional information available on any of these genes or proteins to date. 2.2.6

Arl6

The Arl6 proteins were found in each of the animal proteomes and that of Giardia, but in neither fungus. Each of the sequences result from searches of genomic, cDNA, or EST databases and nothing is known about the activities or functions in cells of the proteins. The fact that the Giardia Arl6 has three (corresponding to Arf1 residues Asp26, Gln71, and Asn126) and Drosophila Arl6 has two (Asp26 and Gln71) deviations in sequence at critical, and otherwise very highly conserved, positions leads us to predict that novel means of controlling guanine nucleotide binding and hydrolysis are used by the Arl6 proteins. 2.2.7

Arl8

The human Arl8 sequences were only very recently reported (Pasqualato et al., 2002) and our analysis identified single orthologs in the fly and worm proteomes. One of the six Arf family members we have identified in Giardia groups with Arl8, but in other analyses groups with the Arfrp. Indeed, a weak, but consistently observed relationship is recovered between the Arl8 and Arfrp subgroups. No functional or expression data are available for any Arl8. However, we note (e.g., see Figures 1 and 2) that each of the Arl8 proteins (and two of the Arl6’s) are the only members of the Arf family that deviate from the aspartate at residue 26 of Arf1, and that corresponds to glycine 12 of Ras proteins. This residue is critical to the hydrolysis of GTP and so its change in Arl8’s may indicate a distinctive means of hydrolysis or interaction with its GAP(s). 2.2.8

Arl 10/11

Human Arl10 and Arl11 are described here for the first time as members of the Arf family, and specifically as possible close relatives of the Arl4 group. Arl10 and Arl11 are their own closest relatives; their inferred protein sequences are ~57% identical, and they also share N-terminal spacing (see Figures 1 and 2). Arl10, Arl11, and DmArl4 were all found from genome

The ARF Family Tree

15

sequence analyses and so no information is yet available on their functions or biochemical properties. 2.2.9

Arfrp

Human and rat orthologs of the Arf-related protein, Arfrp1, were first described in 1995 by Schurmann, et al. (Schurmann et al., 1995). Orthologs are present also in D. melanogaster, C. elegans, S. cerevisiae, and, possibly G. lamblia. This gene appears to have been lost from S. pombe. The genetic loci and predicted proteins in human, rat, and mouse have been assigned the name Arfrp1, though the protein has previously been named Arp. Arfrp proteins all lack the glycine at position 2 and therefore cannot be Nmyristoylated. Arfrps are also unique in that three of the five proteins in our analysis deviated from the conserved residue, corresponding to proline 41 in Arf1 (see Figure 1). This proline is critical to the formation of the extra βsheet found in Arfs, in comparison to Ras family proteins, and its change may indicate a novel structure in this critical domain of the GTPase. The gene has been knocked out in mice and results in embryonic lethality during gastrulation (Mueller et al., 2002). A much more detailed description of this and other functional aspects of Arfrp biology are presented in Chapter 16. 2.2.10

Sar1

The Sar proteins share the lowest percent identity to Arfs (e.g. HsArf1 vs. HsSar1, ~29% identity) or Arls and for that reason are often not included in descriptions of the Arf family. Alignments with other families of GTPases (e.g., Ras, Rab, Rho) reveal that Sar proteins are clearly more closely related to Arfs than these other GTPases. The strongest argument in favor of including Sar proteins in the Arf family is functional. Sar1 is thought to play a highly analogous role to those of class I Arfs in the recruitment of coat proteins to budding vesicles (Barlowe et al., 1993; Kimura et al., 1995; Kuge et al., 1994; Nakano and Muramatsu, 1989; Nishikawa and Nakano, 1991; Oka et al., 1991). But while Arfs act at the Golgi/TGN, Sar1 acts at the ER. Interestingly, the Sar1 proteins have been highly conserved throughout eukaryotes and Sar1 orthologs are present in each of the six organisms in our study. Like the Arl8 and Arfrp proteins, Sar1’s lack an N-terminal glycine and therefore are not N-myristoylated. Additional information on Sar1 can be found in Chapter 16.

16

3.

LOGSDON AND KAHN

NOMENCLATURE ISSUES IN THE ARF/ARL FAMILY

Because this chapter focuses on the evolutionary relationships between highly conserved proteins, we have adopted a simpler system of naming the Arf family members in the different species that emphasizes the family relationships. Any time you deal with gene families in multiple species today you are likely to encounter issues regarding how proteins and genes are named or referred to by researchers and the databases. And while it is important to maintain the evidence of priority in discovery that may result from deposition into databases, it does not make sense to continue using names that are misleading or incorrect. Hence, Table 1 is included to help with the correct identification of the different Arf family members and includes other names found in databases or the literature. In some cases, simultaneous discovery of cDNAs in two or more labs resulted in more than one name for the same cDNA or protein. For example, two different cDNAs/proteins were identified and published in 1999 under the name Arl6, as the database at that time had Arl1-5. Rather than change both to new names, we made the minimal changes in names that would be consistent with the numbering of Arl groups, as shown in Figures 3 and 4. We have introduced new names in the following cases: 1) The protein previously termed Arf4L/Arl6 has been named Arl4B because it is 60% identical to Arl4, nearly identical in length, and because another protein has also been named Arl6. In addition, the name Arf4-like is misleading, as it is not more closely related to Arf4 than any other Arf. The use of the A, B, C suffix is based upon the precedent from Rab proteins and Arl8s. Extending such a system could have resulted in Arl7 and Arl9 being re-named Arl4C and D but we chose not to do so, to minimize the numbers of changes proposed. 2) One of the human proteins named Arl8 in at least one database has been renamed Arl5B (see Table 1) because it is 80% identical to human Arl5, the same length, and to prevent confusion with two other proteins, called Arl8A and B (Pasqualato et al., 2002). 3) Human Arl8A and B were so named because they are 92% identical, to avoid confusion with Arl8/Arl5B, and because they have appeared in print under these names (Pasqualato et al., 2002). 4) Human Arl10 and Arl11 were named as the next open number, after discovering these predicted proteins in database searches. They have not been named previously to our knowledge (J. Logsdon and R. A. Kahn, manuscript in preparation). 5) Several other proteins are referred to in the figures in this chapter by their orthologs, despite the fact that “accepted” names exist in public

The ARF Family Tree

17

databases. Whenever possible, both names are included; e.g., Sc Arl2/Cin4, Sp Arl2/alp41, Sc Arfrp/Arl3, and Ce Arl2/evl-20.

3.1

Evolution of the Arf family

These analyses have revealed some key evolutionary features of the Arf family. It is important to note that this family is—so far—unique to eukaryotes, not being found in either Bacteria or Archaea. In contrast, the fact that many of the Arf family members are found in Giardia, a divergent, and likely early-diverging, protist (Roger, 1999), emphasizes that the these proteins are quite ancient features of eukaryotic cells. Our phylogenetic analyses indicate that the ages of the various subfamilies are quite varied (Figure 4). The oldest families, which contain animal, fungal, and Giardia orthologs, include Arf I/II, Arl2 and Sar (and possibly Arfrp). Some apparently middle-aged families include only animal and fungal members, such as ArfI/II, ArfIII and Arl1. Young sub-families, which are restricted to animals, include Arf I, ArfII, Arl 4 and Arl5. The youngest family, Arf10/11, is restricted only to mammals. These latter groups likely represent functions unique to animals, including developmental roles (as is known for Arl4). In addition to estimating the relative ages of each of the subfamilies by their phylogenetic distribution, we can also make inferences about the likely loss of particular orthologs from particular subfamilies (assuming that the genomes in which these genes are not found are indeed complete and/or that the genes are not so divergent as to not be recognized). These losses include Arl 1 in S. pombe, Arl4 in C. elegans and Arfrp in S. pombe. The apparent absence of these orthologs in particular organisms may provide important clues as to their conserved/dispensable functions in other species. Finally, we must point out that the phylogenetic sampling used here, while useful for generally delineating the relationships among these proteins, does not encompass the breadth of eukaryotic diversity. Additional analyses including more eukaryotes, notably more protist species and plants, will be important to further refine the family tree of Arfs.

4.

SUMMARY

From the sequencing of genomes we now have a more clear idea of the size of the Arf family. Members of this family regulate a wide range of essential cellular processes and they have been performing those functions since the emergence of eukaryotes. With 23 members of the Arf family in mammals, and similar numbers of Arf effectors, Arf GEFs, and Arf GAPs (identified to

18

LOGSDON AND KAHN

date) we also have a minimal number to place on the complexity in signaling that these GTPases are involved in. Specificity in signaling is required but the mechanisms by which Arf family GTPases achieve this specificity is currently uncertain. We will continue to benefit from each of the model genetic systems used in our analysis, as well as others, to decipher the roles and mechanisms of Arf family GTPases in cell regulation.

REFERENCES Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and Lipman, D. J. (1990). Basic local alignment search tool. J. Mol. Biol., 215, 403-410. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W., and Lipman, D. J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res., 25, 3389-3402. Amor, J. C., Horton, J. R., Zhu, X., Wang, Y., Sullards, C., Ringe, D., Cheng, X., and Kahn, R. A. (2001). Structures of Yeast ARF2 and ARL1. Distinct roles for the N terminus in the structure and function of Arf family GTPases. J. Biol. Chem., 276, 42477-42484. Antoshechkin, I., and Han, M. (2002). The C. elegans evl-20 gene is a homolog of the small GTPase ARL2 and regulates cytoskeleton dynamics during cytokinesis and morphogenesis. Dev Cell, 2, 579-591. Barlowe, C., d'Enfert, C., and Schekman, R. (1993). Purification and characterization of SAR1p, a small GTP-binding protein required for transport vesicle formation from the endoplasmic reticulum. J. Biol. Chem., 268, 873-879. Bhamidipati, A., Lewis, S. A., and Cowan, N. J. (2000). ADP ribosylation factor-like protein 2 (Arl2) regulates the interaction of tubulin-folding cofactor D with native tubulin. J Cell Biol, 149, 1087-1096. Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity [see comments]. Cell, 75, 1137-1144. Cavenagh, M. M., Breiner, M., Schurmann, A., Rosenwald, A. G., Terui, T., Zhang, C., Randazzo, P. A., Adams, M., Joost, H. G., and Kahn, R. A. (1994). ADP-ribosylation factor (ARF)-like 3, a new member of the ARF family of GTP-binding proteins cloned from human and rat tissues. J. Biol. Chem., 269, 18937-18942. Clark, J., Moore, L., Krasinskas, A., Way, J., Battey, J., Tamkun, J., and Kahn, R. A. (1993). Selective amplification of additional members of the ADP-ribosylation factor (ARF) family: cloning of additional human and Drosophila ARF-like genes. Proc. Natl. Acad. Sci., USA, 90, 8952-8956. Cockcroft, S., Thomas, G. M., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of ARF in granulocytes. Science, 263, 523-526. Hanzal-Bayer, M., Renault, L., Roversi, P., Wittinghofer, A., and Hillig, R. C. (2002). The complex of Arl2-GTP and PDE{delta}: from structure to function. EMBO J., 21, 20952106. Hillig, R. C., Hanzal-Bayer, M., Linari, M., Becker, J., Wittinghofer, A., and Renault, L. (2000). Structural and biochemical properties show ARL3-GDP as a distinct GTP binding protein. Structure Fold Des., 8, 1239-1245.

The ARF Family Tree

19

Hoyt, M. A., Macke, J. P., Roberts, B. T., and Geiser, J. R. (1997). Saccharomyces cerevisiae PAC2 functions with CIN1, 2 and 4 in a pathway leading to normal microtubule stability. Genetics, 146, 849-857. Hoyt, M. A., Stearns, T., and Botstein, D. (1990). Chromosome instability mutants of Saccharomyces cerevisiae that are defective in microtubule-mediated processes. Mol. Cell. Biol., 10, 223-234. Huelsenbeck, J. P., and Ronquist, F. (2001). MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics, 8, Jacobs, S., Schilf, C., Fliegert, F., Koling, S., Weber, Y., Schurmann, A., and Joost, H. G. (1999). ADP-ribosylation factor (ARF)-like 4, 6, and 7 represent a subgroup of the ARF family characterization by rapid nucleotide exchange and a nuclear localization signal. FEBS Lett, 456, 384-388. Kahn, R. A., and Gilman, A. G. (1984). Purification of a protein cofactor required for ADPribosylation of the stimulatory regulatory component of adenylate cyclase by cholera toxin. J. Biol. Chem., 259, 6228-6234. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kahn, R. A., Kern, F. G., Clark, J., Gelmann, E. P., and Rulka, C. (1991). Human ADPribosylation factors. A functionally conserved family of GTP-binding proteins. J. Biol. Chem., 266, 2606-2614. Kimura, K., Oka, T., and Nakano, A. (1995). Purification and assay of yeast Sar1p. Methods Enzymol., 257, 41-49. Kobayashi, A., Kubota, S., Mori, N., McLaren, M. J., and Inana, G. (2003). Photoreceptor synaptic protein HRG4 (UNC119) interacts with ARL2 via a putative conserved domain. FEBS Lett., 534, 26-32. Kuge, O., Dascher, C., Orci, L., Rowe, T., Amherdt, M., Plutner, H., Ravazzola, M., Tanigawa, G., Rothman, J. E., and Balch, W. E. (1994). Sar1 promotes vesicle budding from the endoplasmic reticulum but not Golgi compartments. J. Cell Biol., 125, 51-65. Lee, F. J., Huang, C. F., Yu, W. L., Buu, L. M., Lin, C. Y., Huang, M. C., Moss, J., and Vaughan, M. (1997). Characterization of an ADP-ribosylation factor-like 1 protein in Saccharomyces cerevisiae. J. Biol. Chem., 272, 30998-31005. Lee, F. J., Moss, J., and Vaughan, M. (1992). Human and Giardia ADP-ribosylation factors (ARFs) complement ARF function in Saccharomyces cerevisiae. J. Biol. Chem., 267, 24441-24445. Lee, F. J., Stevens, L. A., Hall, L. M., Murtagh, J. J., Jr., Kao, Y. L., Moss, J., and Vaughan, M. (1994a). Characterization of class II and class III ADP-ribosylation factor genes and proteins in Drosophila melanogaster. J. Biol. Chem., 269, 21555-21560. Lee, F. J., Stevens, L. A., Kao, Y. L., Moss, J., and Vaughan, M. (1994b). Characterization of a glucose-repressible ADP-ribosylation factor 3 (ARF3) from Saccharomyces cerevisiae. J. Biol. Chem., 269, 20931-20937. Lowe, S. L., Wong, S. H., and Hong, W. (1996). The mammalian ARF-like protein 1 (Arl1) is associated with the Golgi complex. J Cell Sci, 109, 209-220. Lu, L., Horstmann, H., Ng, C., and Hong, W. (2001). Regulation of Golgi structure and function by ARF-like protein 1 (Arl1). J. Cell Sci., 114, 4543-4555. Massenburg, D., Han, J. S., Liyanage, M., Patton, W. A., Rhee, S. G., Moss, J., and Vaughan, M. (1994). Activation of rat brain phospholipase D by ADP-ribosylation factors 1,5, and 6: separation of ADP-ribosylation factor-dependent and oleate-dependent enzymes. Proc. Natl. Acad. Sci., USA, 91, 11718-11722.

20

LOGSDON AND KAHN

McElver, J., Patton, D., Rumbaugh, M., Liu, C., Yang, L. J., and Meinke, D. (2000). The TITAN5 gene of Arabidopsis encodes a protein related to the ADP ribosylation factor family of GTP binding proteins. Plant Cell, 12, 1379-1392. Mishima, K., Tsuchiya, M., Nightingale, M. S., Moss, J., and Vaughan, M. (1993). ARD 1, a 64-kDa guanine nucleotide-binding protein with a carboxyl-terminal ADP-ribosylation factor domain. J. Biol. Chem., 268, 8801-8807. Mueller, A. G., Moser, M., Kluge, R., Leder, S., Blum, M., Buttner, R., Joost, H. G., and Schurmann, A. (2002). Embryonic lethality caused by apoptosis during gastrulation in mice lacking the gene of the ADP-ribosylation factor-related protein 1. Mol. Cell Biol., 22, 1488-1494. Nakano, A., and Muramatsu, M. (1989). A novel GTP-binding protein, Sar1p, is involved in transport from the endoplasmic reticulum to the Golgi apparatus. J. Cell Biol., 109, 26772691. Nishikawa, S., and Nakano, A. (1991). The GTP-binding Sar1 protein is localized to the early compartment of the yeast secretory pathway. Biochim. Biophys. Acta, 1093, 135-143. Oka, T., Nishikawa, S., and Nakano, A. (1991). Reconstitution of GTP-binding Sar1 protein function in ER to Golgi transport. J. Cell Biol., 114, 671-679. Pasqualato, S., Renault, L., and Cherfils, J. (2002). Arf, Arl, Arp, and Sar proteins: a family of GTP-binding proteins with a structural device for front-back communication. EMBO Rep. Radcliffe, P. A., Vardy, L., and Toda, T. (2000). A conserved small GTP-binding protein Alp41 is essential for the cofactor-dependent biogenesis of microtubules in fission yeast. FEBS Lett., 468, 84-88. Randazzo, P. A., and Kahn, R. A. (1995). Myristoylation and ADP-ribosylation factor function. Methods Enzymo.l, 250, 394-405. Renault, L., Hanzal-Bayer, M., and Hillig, R. C. (2001). Coexpression, copurification, crystallization and preliminary X-ray analysis of a complex of ARL2-GTP and PDE delta. Acta Crystallogr. D.Biol. Crystallogr., 57, 1167-1170. Roger, A.J. (1999). Reconstructing Early Events in Eukaryotic Evolution. American Naturalist, 154, S146-S163. Rosenwald, A. G., Rhodes, M. A., Van Valkenburgh, H., Palanivel, V., Chapman, G., Boman, A., Zhang, C. J., and Kahn, R. A. (2002). ARL1 and membrane traffic in Saccharomyces cerevisiae. Yeast, 19, 1039-1056. Schleifer, L. S., Kahn, R. A., Hanski, E., Northup, J. K., Sternweis, P. C., and Gilman, A. G. (1982). Requirements for cholera toxin-dependent ADP-ribosylation of the purified regulatory component of adenylate cyclase. J. Biol. Chem., 257, 20-23. Schurmann, A., Massmann, S., and Joost, H. G. (1995). ARP is a plasma membraneassociated Ras-related GTPase with remote similarity to the family of ADP-ribosylation factors. J. Biol. Chem., 270, 30657-30663. Sewell, J. L., and Kahn, R. A. (1988). Sequences of the bovine and yeast ADP-ribosylation factor and comparison to other GTP-binding proteins. Proc. Natl. Acad. Sci., USA, 85, 4620-4624. Sharer, J. D., and Kahn, R. A. (1999). The ARF-like 2 (ARL2)-binding protein, BART. Purification, cloning, and initial characterization. J. Biol. Chem., 274, 27553-27561. Sharer, J. D., Shern, J. F., Van Valkenburgh, H., Wallace, D. C., and Kahn, R. A. (2001). ARL2 and BART enter mitochondria and bind the adenine nucleotide transporter (ANT). Mol. Biol. Cell, 2002, 71-83. Tamkun, J. W., Kahn, R. A., Kissinger, M., Brizuela, B. J., Rulka, C., Scott, M. P., and Kennison, J. A. (1991). The arflike gene encodes an essential GTP-binding protein in Drosophila. Proc. Natl. Acad. Sci., USA, 88, 3120-3124.

The ARF Family Tree

21

Van Valkenburgh, H., Shern, J. F., Sharer, J. D., Zhu, X., and Kahn, R. A. (2001). ADPribosylation factors (ARFs) and ARF-like 1 (ARL1) have both specific and shared effectors: characterizing ARL1-binding proteins. J. Biol. Chem., 276, 22826-22837. Van Valkenburgh, H. A., and Kahn, R. A. (2002). Coexpression of proteins with methionine aminopeptidase and/or N-myristoyltransferase in Escherichia coli to increase acylation and homogeneity of protein preparations. Methods Enzymol., 344, 186-193. Vitale, N., Moss, J., and Vaughan, M. (1998). Molecular characterization of the GTPaseactivating domain of ADP- ribosylation factor domain protein 1 (ARD1). J. Biol. Chem., 273, 2553-2560.

Chapter 2 THE GDP/GTP CYCLE OF ARF PROTEINS Structural and Biochemical Aspects

SEBASTIANO PASQUALATO, LOUIS RENAULT, AND JACQUELINE CHERFILS Laboratoire d'Enzymologie et Biochimie Structurales, CNRS, Gif sur Yvette Cedex, France

Abstract:

1.

The structural GDP/GTP cycle of Arf proteins is implemented by switch regions that change their conformations in response to the nature of their interactions. They include the classical switch regions in the nucleotidebinding site, and a unique switch region located on the opposite side of the protein. These switch regions communicate with each other and couple activation of Arfs by GTP to their recruitment to membranes. Variations in this cycle determine the specificity and differences in Arf1 and Arf6 signaling. The role of guanine nucleotide exchange factors (GEFs) as key players in the implementation of the nucleotide/membrane coupling is also described. A particular emphasis is given to the inhibition of GEF-activated nucleotide exchange by the drug Brefeldin A and by a GEF point mutation, as tools to trap intermediate states of the reaction. Taken together, structural and biochemical studies yield a comprehensive model for one of the most remarkable nucleotide cycle found in G proteins.

INTRODUCTION

Arf proteins are involved in a large and still expanding number of cellular processes, all related to their pivotal role as regulators of protein traffic and membrane dynamics (reviewed in Chavrier and Goud, 1999). These functions are operated by their ability to switch between an inactive GDPbound and an active GTP-bound form, each able to bind selectively to distinct sets of cellular regulators, while effectors bind only to their GTPbound form. In this chapter we will focus on the biochemical and structural implementation of the GDP/GTP switch, which departs from the classical switch of other small GTP-binding proteins. 23

2 24

PASQUALATO, RENAULT AND CHERFILS

Activation of Arf proteins by GTP is coupled to their recruitment to membranes. The membrane- and nucleotide-binding sites are located on opposite sides, where they feature switch regions that change their conformation in response to the nature of their interactions and communicate with each other. In the nucleotide binding site, these regions are the classical switch 1 and switch 2 regions found in all members of the G protein superfamily (reviewed in Vetter and Wittinghofer, 2001), while a unique Nterminal extension that carries a myristoyl lipid changes its conformation upon binding to membranes. Coupling of the GDP/GTP cycle to membrane interactions was first characterized by biochemical studies, which provided evidence for the role of the N-terminal helix and its lipid modification (Antonny et al., 1997; Franco et al., 1993, 1995; Kahn et al., 1992; Randazzo et al., 1995). How communication is established between the N-terminal helix and the nucleotide binding site, and the consequences for the activation mechanism, were revealed later by structural studies (Amor et al., 1994; Goldberg 1998; Greasley et al., 1995; Menetrey et al., 2000; Pasqualato et al., 2001) and recently proposed to extend to other small GTP-binding proteins homologous to Arf, including Arf-like and Sar proteins (Pasqualato et al., 2002). This chapter will first describe the structural aspects of the GDP/GTP cycle and its biochemical characteristics, and how variations in the structural cycle provide a basis to understand how the closely related Arf1 and Arf6 have specific, non-overlapping functions. It will then discuss the role of guanine nucleotide exchange factors (GEFs) as key players in the implementation of the nucleotide/membrane coupling. A particular emphasis will be given to the inhibition of GEF-activated nucleotide exchange by the drug brefeldin A (BFA) and by a GEF point mutation, as tools to trap intermediate states of the reaction. Taken together, structural and biochemical studies yield a comprehensive model for one of the most remarkable nucleotide cycle found in G proteins.

2.

THE MEMBRANE/NUCLEOTIDE CYCLE OF ARF PROTEINS

Although many biochemical studies have preceded the report of the first crystal structure of Arf in 1994 (Amor et al., 1994), we have chosen to introduce this chapter by a description of the structural aspects of the mechanism, as they provide a convenient background for subsequent discussions. The structure of human Arf1 bound to GDP (Amor et al., 1994), followed by that of rat Arf1•GDP (identical in sequence) (Greasley et al., 1995) were first reported. Structure of an activated Arf was then

THE GDP/GTP CYCLE OF ARF PROTEINS

3 25

reported for a truncated mutant of human Arf1 lacking the N-terminus extension (Arf1∆17) (Goldberg, 1998), together with the nucleotide-free complex of Arf1∆17 with the Sec7 domain of the yeast GEF, Gea2 (Goldberg, 1998). More recently, the GDP/GTP cycle of full-length human Arf6 (Menetrey et al., 2000; Pasqualato et al., 2001) and the GDP-bound form of yeast Arf2 (Amor et al., 2001) have also been solved, as well as a complex of the GTPase activating protein (GAP) domain of rat Arf GAP1 bound to Arf1∆17•GDP (that is, the product of the GTPase reaction) (Goldberg, 1999). Structures of Arf proteins display several unique features with respect to the common fold of small GTP-binding proteins (Amor et al., 1994; Amor et al., 2001; Greasley et al., 1995; Menetrey et al., 2000). They diverge most in their GDP-bound forms, with an original conformation of switch 1, a register shift of strands β2-β3 (a region we have coined the interswitch (Pasqualato et al., 2001)) and an N-terminal helix that is anchored opposite to the nucleotide-binding site (Figure 1). By contrast, active forms of Arf, either truncated (Goldberg, 1999) or not (Pasqualato et al., 2001) of the Nterminal extension, adopt conformations that are similar to those of other small GTP-binding proteins. These distinctive features are at the heart of a communication device that operates between the N-terminal helical extension and the nucleotide-binding site.

2.1

Switch 1 forms an extra β-strand in Arf•GDP

The switch 1 region is located downstream from the P-loop (a conserved signature of GTP-binding proteins that they use to bind the phosphates of guanine nucleotides). It carries an invariant threonine residue that binds the Mg2+ ion and the γ-phosphate of GTP, thereby loading a ‘spring’ that contributes to organize the active conformation (reviewed in Vetter and Wittinghofer, 2001). Unlike Ras, Rho and Rab proteins, where switch 1 interacts with the GDP nucleotide, switch 1 of Arf•GDP is remote from the nucleotide and does not interact with it (Amor et al., 1994; Amor et al., 2001; Greasley et al., 1995; Menetrey et al., 2000). Instead, it forms an additional β-strand with strand β3, thus extending the central β-sheet (Figure 1). The extra β-strand disappears in the active form as a large conformational change restores the classical interactions, including those of the invariant threonine (T48 in human Arf1, T44 in human Arf6) (Goldberg, 1998; Pasqualato et al., 2001) (Figure 1). This movement articulates on hinge glycines, located at each end of switch 1 (G40/G50 in human Arf1, G36/G46 in human Arf6), which provide the required flexibility. There are, so far, two small GTP-binding proteins whose switch 1 forms an extra βstrand remote from GDP, the other one being Ran (Scheffzek et al., 1995).

4 26

PASQUALATO, RENAULT AND CHERFILS

In the GDP-bound form of both proteins, interactions that wedge the guanine base, which are provided by an aromatic residue of the switch 1 in Ras, Rho,

Figure 1. The GDP/GTP cycle and the interswitch toggle of full-length human Arf6. Regions that change their conformation are in dark gray. Switch 1 and 2 regions are defined here based on the conformational changes between the GDP- and GTP-bound Arf structures, and therefore differ from their classical definitions in Ras. It should be noted that as the switch 1, interswitch and switch 2 regions form a continuous switch peptide, their junctions are arbitrary. In Arf6, switch 1 comprises residues 36-46, the interswitch residues 47-64 and switch 2 residues 65-80. The flexible N-terminus in Arf6•GTPγS is shown as a dashed line. D63 from the DxGGQ motif, which blocks the γ-phosphate binding-site in the GDP-bound form, and the invariant T44 from switch 1, are shown. Hinge glycine residues that allow the flexibility of switch 1 are indicated by their Cα. The DxGGQxxxRxxW signature is also shown. Protein Data Bank entries are 1E0S (Arf6•GDP) and 1HFV (Arf6•GTPγS).

and Rab families, originate from a residue from the β6-α5 loop (T161 in Arf1, K152 in Ran). However, only in Arf is this interaction maintained throughout the GDP/GTP cycle, whereas in active Ran it is replaced by an aromatic interaction from the switch 1 (Vetter et al., 1999) equivalent to that of Ras, Rho and Rab. A +1 shift of the invariant threonine within the

THE GDP/GTP CYCLE OF ARF PROTEINS

5 27

sequence of the switch 1 in Arf may explain this difference (Pasqualato et al., 2001). Edge-to-edge interaction of the switch 1 β-strand yields a symmetrical dimer in the crystal of Arf1•GDP (interface area of ū1050 Å2), which could not form with Arf1•GTP because of the entirely different conformation of switch 1. Arf6•GDP also forms a dimer that is built from a crystal symmetry. Its interface is different from that of the Arf1•GDP dimer and somewhat larger (ū1400 Å2), and it is not compatible with the GTP-bound conformation. On the other hand, yeast Arf2•GDP and the active forms of Arf do not form crystalline dimers. An interesting question is whether the observation of crystalline dimers could be relevant to the formation of equivalent dimers at the membrane in the cell. Hints for a membrane homodimerization of Arf1 have been obtained by cross-linking in the presence of Golgi membranes and GTPγS (Zhao et al., 1999). With respect to the crystal structures, an Arf1 dimer formed under these conditions must either be different from the crystalline dimer, or be an Arf1•GDP dimer despite the presence of GTPγS. Also, activation of Arf proteins by their GEFs involves the unfolding of the switch 1 β-strand (Goldberg, 1998), suggesting that, if an Arf•GDP dimer exists at the membrane, it must dissociate in order to be activated by the GEF.

2.2

The interswitch toggle

The myristoylated N-terminal extension of Arf proteins is necessary for its cellular activity, and has early been proposed to act as an additional switch (Antonny et al., 1997; Kahn et al., 1992; Randazzo et al., 1995). Crystal structures of non-myristoylated Arf proteins have provided striking insight into this issue. In Arf•GDP, the N-terminus forms a helix that lies in a hydrophobic groove comprising residues from the interswitch loop (the loop that connects strands β2 and β3) and the tip of switch 1 (Amor et al., 1994; Amor et al., 2001; Greasley et al., 1995; Menetrey et al., 2000) (Figure 1). The existence of this binding pocket is made possible by the retraction of the interswitch in the protein core, where it forms β-strand interactions with strand β1 on one edge, and with the extra β-strand of switch 1 on the other one. As a consequence, residues of switch 2 that are involved in binding the γ-phosphate of GTP are displaced out of register such that they are incompatible with the binding of GTP. Furthermore, an invariant aspartate at the interswitch-switch 2 junction (sequence DxGGQ, D67 in Arf1, D63 in Arf6) is displaced so as to occlude the binding site for the γ-phosphate (Pasqualato et al., 2002) (Figure 2). These large differences, as compared to other small GTP-binding proteins, vanish in structures of Arf proteins bound to slowly hydrolysable

6 28

PASQUALATO, RENAULT AND CHERFILS

Figure 2. Conformational changes of the nucleotide-binding site along the GDP/GTP cycle. From top to bottom: Arf1•GDP (PDB entry 1HUR), nucleotide-free Arf1∆17-Gea2-Sec7 complex (gift from J. Goldberg), Arf1∆17•GDP(NH)P (gift from J. Goldberg). Hydrogen bonds are shown as dotted lines. The carbonyls of A28 and A29 and the Cα of G69 and G70 are shown.

THE GDP/GTP CYCLE OF ARF PROTEINS

7 29

GTP analogs, as a consequence of a large amplitude movement that affects switch 1, the interswitch, switch 2, and the N-terminal helix (Figure 1). These rearrangements were first observed in the structure of Arf1∆17 bound to GDP(NH)P (Goldberg, 1998), and subsequently in full-length human Arf6 bound to GTPγS (Pasqualato et al., 2001), the latter structure ruling out an artifact of the N-terminal truncation. In these structures, the interswitch unzips its interactions with its two neighboring β-strands, and reforms the interswitch-β1 interaction with a two-residue register shift while switch 1 has moved away, reaching the GTP nucleotide. These concerted conformational changes restore a classical conformation of switch 2 that is then able to bind GTP. In particular, the aspartate from the DxGGQ motif is back in register and interacts with the P-loop (T31 in Arf1) and indirectly through a water molecule with Mg2+, both interactions found in other small GTP-binding proteins. It also interacts with a conserved asparagine (N52 in Arf1) whose mutation to arginine in Arf1 impairs activation of phospholipase D (Skippen et al., 2002). The rearrangement of the DxGGQ motif also positions the second glycine to bind the γ-phosphate of GTP, thus providing, together with the invariant threonine from switch 1, the second ‘loaded spring’ of the classical GDP/GTP switch (Figure 2). On the opposite side, the movement pushes the interswitch out of the protein core by about 7Å, thereby destroying the hydrophobic binding site for the Nterminal helix. In turn, in full-length Arf6 (the helix is truncated in the Arf1∆17 structures) this helix is displaced into the solvent and not visible because of its mobility in the crystal. Thus, the conformational changes of the N-terminus, the interswitch and the switch regions are strongly coupled: either the N-terminus interacts with the hydrophobic pocket and fastens the retracted interswitch, hence a conformation of the switch regions that is incompatible with the binding of GTP; or the N-terminus is displaced, allowing the interswitch to be pushed out of the protein core and restoring the ability of switch 1 and 2 to bind GTP. Conformational changes that occur upon replacement of GDP by GTP in Arf are the largest ever observed within the core G domain of a small GTPbinding protein. They are all the more spectacular that they are grafted onto a remarkably conserved architecture found in tens of GTP-binding proteins which undergo a GDP/GTP conformational cycle but lack this extra movement. As we will see below, these movements reflect a front-back communication device that we have termed the “interswitch toggle” (Pasqualato et al., 2002), coupling interactions of the N-terminus with membranes with the GDP/GTP nucleotide status of the protein. Comparison of the GDP/GTP cycle of Arfs to that of small GTP-binding proteins that have no interswitch toggle reveals two potential determinants for this structural mechanism (Pasqualato et al., 2002). One is the length of

8 30

PASQUALATO, RENAULT AND CHERFILS

the interswitch, which is shorter in Arf proteins than in other small GTPbinding proteins, allowing it to be eclipsed in the protein core and form the hydrophobic binding site for the N-terminal helix in the inactive form. The second is a conserved sequence signature, wDvGGQxxxRxxW (residues 6678 in Arf1, 62-74 in Arf6) that spans the β3 strand of the interswitch and switch 2. In Arf•GDP, the interswitch and switch 2 are mutually stabilized by interaction of the tryptophan (W78 in Arf1) with surrounding aromatic residues, including another tryptophan from the interswitch (W66 in Arf1). Reorganization of this peptide upon binding of GTP creates a novel network

Figure 3. The GTP-bound conformation of the DxGGQxxxRxxW signature. The carbonyls of G69 and G70 and the N of G70 are shown. The hydrogen bond network is shown as dotted lines. This close-up view is from Arf1∆17•GDP(NH)P.

of hydrogen bonds involving the glycine pair, the arginine and the tryptophan (G69, G70, R75 and W78 in Arf1) (Figure 3). This structural transition is probably made possible by the flexibility provided by the glycine pair, the first one being found only in the Arf family. The large change in environment of the tryptophan accounts for the increase in intrinsic fluorescence when GDP is exchanged for GTP (Kahn and Gilman, 1986), which provides a built-in reporter to monitor the kinetics of spontaneous as well as GEF-catalyzed nucleotide exchange (Antonny et al., 1997).

THE GDP/GTP CYCLE OF ARF PROTEINS

2.3

9 31

Arf6 vs Arf1: a structural basis for their functional specificity

We have focused so far on the features that are common to the GDP/GTP cycles of Arf1 and Arf6, and likely to other Arf proteins as well. This knowledge is essential to go a step further and identify the specific structural features beyond the ‘air de famille’ that may explain why Arf1 and Arf6 have distinct cellular functions (reviewed in Chavrier and Goud, 1999, see Chapter 14) despite having 67% sequence identity and very similar switch 1 and 2 regions. For instance, there is a single V/I change between Arf1 and Arf6 in the binding interface with the Sec7 domain as established from the nucleotide-free Arf1/Gea2-Sec7 complex (Goldberg, 1998). However, Arf6 has a family of specific exchange factors, termed EFA6, that are inactive on Arf1 in vivo and in vitro (Franco et al., 1999; Macia et al., 2001). We have

Figure 4. Arf1 and Arf6 have different GDP-bound conformations. Arf6•GDP is shown in light gray. Regions of Arf1•GDP that have significant conformational differences to Arf6•GDP are superimposed in dark gray. The flexible switch 2 region in Arf1•GDP is also shown.

10 32

PASQUALATO, RENAULT AND CHERFILS

investigated the structural basis for this specificity with the crystallographic study of the GDP/GTP cycle of human Arf6 and its comparison to that of Arf1 (Menetrey et al., 2000; Pasqualato et al., 2001). While the sequence identity between Arf1 and Arf6 predicts that they should be extremely similar, this study revealed unexpected differences affecting their GDP-bound forms (Menetrey et al., 2000) (Figure 4). First, if the N-terminus of Arf6 does form an amphipathic helix that fastens the retracted interswitch despite being shorter by four residues, the structure shows that it is the linker to the G protein core that is shortened and has a different conformation. More surprisingly, switch 1 of Arf6, although forming the extra β-sheet, is displaced by up to 3-4Å relative to Arf1, and its switch 2 is less flexible than that of Arf1. Where do these differences come from, given that Arf1 and Arf6 have almost identical switch regions? Comparison of Arf6•GDP with Arf1•GDP identifies residues located either on the remote ends of the switch regions, or in contact with the switch regions, that are different between Arf1 and Arf6 and could account for their biochemical and functional specificities. The contribution of these residues was tested by mutational studies. The crystal structure of Arf6•GDP pointed to a sequence difference in the vicinity of the nucleotide-binding site (I42 in Arf1, S38 in Arf6), which positions a conserved glutamate (E54 in Arf1, E50 in Arf6) in or out of the active site, respectively (Figure 5). Permutation of these residues was sufficient to switch the kinetics of spontaneous nucleotide dissociation between N-terminally truncated Arf1 and Arf6 (Menetrey et al., 2000), which is about 15 times faster in Arf6 than in Arf1 (Macia et al., 2001), thus supporting its proposed role in the nucleotide binding site. A functional role for this residue is further supported by the inability of the double Arf6-Q37I,S38I mutant to form the actin protrusions upon treatment with aluminum fluorides in HeLa cells that are characteristic of wild type Arf6 (Al-Awar et al., 2000). Other residues potentially involved in stabilizing Arf6•GDP in a conformation different from that of Arf1•GDP were introduced in Arf1 and tested for their activation by an unspecific GEF, ARNO, and an Arf6-specific GEF, EFA6 (Macia et al., 2001). This study showed that ARNO is very robust at activating these various Arf1 mutants, while mutations targeting separately either switch 1, the interswitch or switch 2 are not sufficient on their own to gain activation by EFA6. Thus, although specificity is demonstrated to lie within the G protein core of Arf and to be fully defined by the Sec7 domain of the GEF, its structural implementation is likely to result from the combination of subtle and localized effects. It should be noted that chimeras that paste entire sub-regions of Arf1 and Arf6 together may, from that point of view, induce structural effects that are larger than the differences between Arf1 and Arf6 that specify their interactions.

THE GDP/GTP CYCLE OF ARF PROTEINS

11 33

On the other hand, the active structures of Arf1∆17 and Arf6 are surprisingly similar (Pasqualato et al., 2001). In particular, the switch 1 and switch 2 regions do not show the discrepancies of the inactive GDP-bound forms. Altogether, the conformations of the switch regions fully identify the

Figure 5. A structural basis for the different nucleotide-binding properties of Arf1 and Arf6. Replacement of S38 in Arf6 by I42 in Arf1 modifies the interactions of a conserved glutamate residue in the nucleotide-binding site.

inactive from the active form of an individual Arf isoform, and their different conformations in Arf1•GDP and Arf6•GDP can be discriminated by partners of the inactive form, in particular their GEFs (Menetrey et al., 2000). On the contrary, as the switch regions have the same conformation and virtually identical sequences in the active forms of Arf1 and Arf6, they are expected to be poor contributors to their specific interactions with effectors (Pasqualato et al., 2001). As a consequence, specific interactions established by activated Arf must originate from contributions other than the sole recognition of the switch regions by protein partners. This can be accounted for by several, non-exclusive hypotheses. One is that regions other than the switch regions, which feature sequence differences between Arf1 and Arf6, are necessary for establishing specific interactions. Another is that specificity is initiated while Arf is still bound to GDP and subsequently propagates to Arf•GTP-effector complexes. This could be accomplished by the formation of transient Arf-GEF-effector complexes, where the identity of the effector would be specified by interactions with both Arf and the GEF (Pasqualato et al., 2001). Such a concept of GEF/effector interactions is indeed emerging for another family of small

12 34

PASQUALATO, RENAULT AND CHERFILS

GTP-binding proteins involved in intracellular traffic, the Rab family (reviewed in Zerial and McBride, 2001).

2.4

The coupled membrane-nucleotide switch

Full appreciation of the communication device in Arf comes from the combination of these structural data with the biochemical studies, many of which had been carried out before the structures were known. The first biochemical studies on Arfs purified from bovine brain showed that they are present in both the cytosol and membrane fractions (Kahn et al., 1988). Subsequent experiments showed that binding of GTP to Arf is required for its membrane association (Kahn, 1991; Walker et al., 1992) and vice versa Arf1 cannot be activated in the absence of membranes (Randazzo et al., 1995). This supported a model in which Arf activation, mediated by a membrane-associated GEF, is concomitant with its recruitment to membranes (Serafini et al., 1991). Myristoylation of Arf proteins is necessary for their GTP-dependent binding to membranes (Haun et al., 1993; Kahn et al., 1992). The relative contributions of the myristate and the N-terminal helix in binding to membranes were studied by Franco and co-workers by in vitro experiments in the presence of vesicles (Franco et al., 1993, 1995). Unmyristoylated (unmyr) Arf1•GTPγS, but not unmyr-Arf1•GDP, binds to phospholipid vesicles or phospholipid-cholate micelles, and the interaction of myristoylated (myr)-Arf1•GTPγS with membranes is stronger than that of myr-Arf•GDP (Franco et al., 1993, 1995). These observations suggested that Arf interacts with membranes through two components: the myristate, that allows a basal interaction regardless of the nucleotide state; a protein region, that becomes available for binding when Arf switches to the GTPbound form, which is the N-terminal helix that has been described in the previous section (Antonny et al., 1997; Kahn et al., 1992; Randazzo et al., 1995). This amphipathic N-terminal α-helix buries hydrophobic residues in a pocket of the G protein core, thereby fastening the inactive conformation in the absence of membrane (Amor et al., 1994; Greasley et al., 1995; Menetrey et al., 2000). Mutation of these residues to alanine, which is expected to lower the hydrophobic interactions, also decreased the affinity of mutant Arf proteins for phospholipids (Antonny et al., 1997). Thus, in the GTP-bound conformation, the N-terminal helix was predicted to insert its hydrophobic residues’ side chains at the level of phospholipid acyl chains (Antonny et al., 1997). This model is supported by NMR experiments that showed that the N-terminal Arf1 peptide, either myristoylated or not, is predominantly helical in a lipid environment, with the helix axis nearly parallel to the bilayer surface (Losonczi and Prestegard, 1998; Losonczi et

THE GDP/GTP CYCLE OF ARF PROTEINS

13 35

al., 2000). Moreover, myristoylation, while enhancing lipid binding, does not affect peptide structure or behavior to a significant degree. Altogether, biochemical and structural data are fully accounted for by a model where the N-terminal amphipathic helix buries its hydrophobic face either in a pocket of the G protein core in cytoplasmic Arf•GDP (the conformation observed in crystal structures), or in interaction with the membrane in Arf•GTP. Its conformation in Arf•GDP serves as a ‘hasp’ to fasten the retracted conformation of the interswitch, and must be released to allow nucleotide exchange. This is facilitated by interaction of the Nterminal myristate with membranes, which occurs regardless of the nature of the bound nucleotide. This interaction of the myristate with membranes possibly also favors a dynamic equilibrium between the Arf- and membranebound conformations of the N-terminus. After the N-terminal hasp is unfastened, binding of GTP pushes and/or stabilizes the interswitch in the conformation where it destroys the hydrophobic pocket, such that the Nterminal helix is now stably bound to the membrane. An important consequence is that kinetics of nucleotide exchange of Arf are correctly measured only in the presence of membranes using myr-Arf, or in solution using mutants truncated of their N-terminal ‘hasp’, while all other combinations are likely to yield artifactual exchange rates. Therefore, Arf1∆17 provides a convenient alternative to measure exchange activities in solution (Antonny et al., 1997; Paris et al., 1997).

2.5

A different GTPase mechanism in Arf proteins?

Like other small GTP-binding proteins, the active state of Arfs is terminated by hydrolysis of the bound GTP. However, spontaneous GTPase activity is not measurable in Arf proteins (Kahn and Gilman, 1986), for which the appellation of “GTPases” would therefore be illegitimate. Structural studies allow us to address this difference with other small GTP-binding proteins, which have intrinsic GTPase activity. Surprisingly, in the active forms, the nucleotide binding site and GTP have the classical arrangement found in other small GTP-binding proteins, including the localization of the catalytic glutamine (DxGGQ motif, Q71 in Arf1) in the vicinity of the γ-phosphate of GTP with which it could interact with only minor conformational changes (Figure 4). Moreover, the wDvGGQxxxRxxW signature, which is also found in the Gα subunit of heterotrimeric proteins, has the same conformation in activated Arf1 and Arf6 and in the Gαi•GDP•AlF4 complex (Coleman et al., 1994). The latter structure mimics the transition state of the GTPase reaction. One essential difference, however, lies in the replacement of G42 in the P-loop of Gα, by an aspartate (D26 in Arf1) (Figures 2 and 4). Superposition of activated Gαi•GDP•AlF4 with structures of activated Arfs

14 36

PASQUALATO, RENAULT AND CHERFILS

suggests that the side chain of this aspartate may hinder the catalytic glutamine from interacting with and stabilizing the leaving phosphate group. This aspartate is in fact equivalent to the G12 position of Ras, whose mutation to valine results in impairment in both spontaneous and GAPstimulated GTPase activity (Scheffzek et al., 1997). It is as yet unclear how the GAP releases this unfavorable contact in Arf, but this observation opens the possibility that the contribution of the catalytic glutamine may differ from other small GTP-binding proteins. The crystal structure of the complex of Arf1∆17•GDP with the catalytic domain of rat Arf GAP1 does not provide insight into this issue, as the complex features an interface that is remote from the nucleotide binding site (Goldberg, 1999). This study proposes that the actual GAP is composed of the GAP domain and the coatomer, which may provide the interactions that are missing in the GTP binding site (Goldberg, 1999). At the moment this question is the subject of an on-going debate as to the existence or not of a classical GAP mechanism involving the complementation of the active site with an ‘arginine finger’ provided by the Arf GAP (Click et al., 2002; Goldberg, 1999; Szafer et al., 2000; Szafer et al., 2001).

3.

ROLE OF GEFS IN COUPLING THE MEMBRANE AND NUCLEOTIDE SWITCHES

Activation of Arf by their guanine nucleotide exchange factors, which are characterized by a conserved Sec7 domain of §200 amino acids, is the subject of several chapters in this book (see Chapters 4-6). Here we will focus on the catalytic mechanism of Sec7-domain-stimulated nucleotide exchange and how GEF activity is grafted onto the basic membrane/nucleotide switch, emphasizing the role of membranes in this process. Further insight into this mechanism was gained from use of a GEF point mutant and BFA as tools to trap otherwise transient steps of the reaction. A comprehensive model for the GDP/GTP cycle can be formalized from this ensemble of studies.

3.1

The nucleotide-free Arf-Sec7 complex and the glutamic finger

Crystal structures of the Sec7 domain of human ARNO (ARNO-Sec7) (Cherfils et al., 1998; Mossessova et al., 1998) and the NMR structure of cytohesin1-Sec7 (Betz et al., 1998) have shown that this domain is a superhelix of helices, which form a hydrophobic groove lined with conserved amino acids, on the rim of which stands an invariant glutamate

THE GDP/GTP CYCLE OF ARF PROTEINS

15 37

(the ‘glutamic finger’, E156 in ARNO) (see Figure 6). Site-directed mutagenesis identified this groove as the binding site for Arf (Betz et al., 1998; Cherfils et al., 1998; Mossessova et al., 1998). The functional importance of the invariant glutamate was first demonstrated in a protein from A. thaliana, EMB30 (also known as GNOM and recently demonstrated to be an Arf GEF (Steinmann et al., 1999)), where its mutation to lysine resulted in various defects in embryogenesis (Shevell et al., 1994). Mutation of the glutamic finger to lysine in the Sec7 domain of ARNO (ARNO-K156) produces the most severe decrease in catalysis of nucleotide exchange as compared to other mutations within the active site (Beraud-Dufour et al., 1998; Cherfils et al., 1998). Despite its crucial catalytic defect, ARNOK156 forms an abortive complex with Arf1∆17•GDP at low Mg2+ concentration (Beraud-Dufour et al., 1998). No abortive complex is formed with ARNO-D156, suggesting that the positive charge of the lysine mutant mimics Mg2+ in the nucleotide binding site of Arf1∆17•GDP (BeraudDufour et al., 1998). From these studies it was predicted that the ‘glutamic finger’ would act in the vicinity of Mg2+ and the phosphates of the nucleotide, where its negative charges may, in a first step, destabilize the phosphates of GDP by electrostatic repulsion, before they mimic the βphosphate to stabilize the empty nucleotide-binding site (Beraud-Dufour et al., 1998). On the other hand, the surface of Arf interacting with the Sec7 groove was mapped to the switch 1 and 2 regions by mutagenesis (BeraudDufour et al., 1998) and hydroxyl-radical footprinting (Mossessova et al., 1998). These data were combined by docking simulation to predict the structure of the nucleotide-free complex before its crystal structure was solved (Beraud-Dufour et al., 1998). Unambiguous orientation was achieved by charge permutation between ARNO-Sec7-D183 and K73 in switch 2 of Arf. Individually, each mutant failed to support catalyzed nucleotide exchange, but significant activity was restored when they were assayed together, suggesting that these two residues would form a salt bridge in the complex (Beraud-Dufour et al., 1998). Taken together, these data provided sufficient information to dock switch 1 of Arf1•GDP into the hydrophobic groove of ARNO-Sec7, with the nucleotide-binding site open to the solvent. These ‘low resolution’ predictions were essentially confirmed by the crystal structure of the nucleotide-free complex of human Arf1∆17 with the Sec7 domain of yeast Gea2 (Goldberg, 1998) (Figure 6). In this complex, the glutamic finger binds to the invariant lysine of the P-loop (which is a ligand of the β- and γ- phosphates of guanine nucleotides) (Figure 2). It is interesting to note that electrostatic stabilization of the empty nucleotide binding site by a negatively charged residue has turned out to be a general feature of the GEFs of the small GTP-binding proteins (Renault et al., 2001). Unexpectedly though, the ‘glutamic finger’ is contributed by the GEF only

16 38

PASQUALATO, RENAULT AND CHERFILS

in the case of the Sec7 domain, while in other systems it is provided by the G protein itself. In addition, the complex revealed that the P-loop is distorted such that carbonyls of its main chain (A27 and A28) would conflict with binding of the α- and β-phosphates of a nucleotide

Figure 6. The nucleotide-free Arf∆17-Gea2-Sec7 complex. Coordinates of the complex were a gift from J. Goldberg.

(Figure 2). Although this may be a simple consequence of the absence of nucleotide (Goldberg, 1998), this conformation is stabilized by an elaborate network of hydrogen bonds supported by D26 in the P-loop (Figure 2), suggesting that residue may be critical for the alternative conformations of the P-loop (Pasqualato et al., 2001). Besides these aspects, the most spectacular and unpredictable finding from the nucleotide-free Arf1∆17-Gea2 complex was that nucleotide-free Arf has already undertaken the interswitch toggle that is observed in activated Arf•GTP (Goldberg, 1998). This finding has been pivotal in formalizing the model for the membrane/nucleotide switch and its regulation by exchange factors as will be discussed below.

THE GDP/GTP CYCLE OF ARF PROTEINS

3.2

17 39

Regulation of Arf GEF activity by membranes

In the nucleotide-free complex, the interswitch is pushed out of the protein core as in activated Arf, thus destroying the hydrophobic pocket that forms the binding site of the N-terminal helix in Arf•GDP (Goldberg, 1998). This is made possible in the crystal by the use of the truncated Arf1∆17 mutant, but in the cell this implies that the N-terminal hasp must be displaced by membranes to permit Sec7-catalysed nucleotide exchange. The role of membranes in assisting the mechanism of nucleotide exchange had been shown for a partially purified Arf GEF (Franco et al., 1996) and later for ARNO, which is inactive on full-length myr-Arf1 in the absence of membranes (Beraud-Dufour et al., 1999) and for its Sec7 domain under various experimental set-ups (Paris et al., 1997). In particular, the Sec7 domain of ARNO was shown to activate Arf1∆17•GDP in the absence of phospholipids (Paris et al., 1997). On the contrary, the Sec7 domain of ARNO did not readily activate the equally soluble unmyr-Arf1•GDP in solution. Thus, activation of exchange stimulated by Sec7 domains occurs either in solution with the truncated Arf mutant, where the interswitch is not fastened by the N-terminus, or in the presence of lipids if myr-Arf is used instead, in which case membranes are required to displace the N-terminal hasp. Other conditions used in exchange assays yield artifactual exchange rates. This role of membranes in Sec7-stimulated nucleotide exchange reflects the general requirement of Arfs for the unfastening of the N-terminal hasp in order to create a binding site for GTP. On the other hand, the Sec7 domain is found in multidomain proteins with different architectures, including two families, the ARNO/cytohesin and EFA6 families, that carry phospholipidinteracting pleckstrin homology (PH) domains and a novel Arf GEF, p100, with a PH domain yet to be characterized as such (Someya et al., 2001). Phospholipids stimulate the nucleotide exchange activity of GEFs of the ARNO/cytohesin family by interacting with their PH domains, hence recruiting the GEF at the membrane where it finds its substrate (see also Chapters 4-6). For instance, nucleotide exchange activity stimulated by ARNO is enhanced when increasing amounts of phosphatidylinositol 4,5bisphosphate (PI(4,5)P2) are added to phospholipid vesicles (Chardin et al., 1996). More detailed studies on the role of PI(4,5)P2 demonstrated that binding of the PH domain to PI(4,5)P2-containing vesicles does not affect the catalytic activity of ARNO, but simply promotes membrane recruitment of the exchange factor (Paris et al., 1997). Furthermore, this effect is combined in ARNO with interaction of a polybasic C-terminus with membranes (Macia et al., 2000). It should be noted that these latter experiments were done with the triglycine splice variant of ARNO (S. Paris,

18 40

PASQUALATO, RENAULT AND CHERFILS

personal communication). This variant carries three consecutive glycines in the PH domain and has a high affinity for PI(4,5)P2 and low discrimination between PI(4,5)P2 and PI(3,4,5)P3, unlike the diglycine variant (reviewed in Cullen and Chardin, 2000). It is not known yet how other families of large Arf GEFs interact with membranes, but it is likely that uncharacterized noncatalytic domains fulfill such a function, either directly or through binding to a membrane-located partner. Thus, membranes have two essentially independent functions: as a co-factor of the Sec7 domain that acts directly on Arf and unfastens its N-terminal hasp; and in recruiting the exchange factor by domains other than the catalytic Sec7 domain, thereby contributing to localizing the GEF at the compartment where Arf is to be activated.

3.3

The intermediate steps of the exchange reaction studied by inhibition by BFA and by the glutamic finger mutation

The basic GEF-catalyzed exchange reaction common to all small GTPbinding proteins involves an initial ternary complex comprising GDP and the GEF. This complex isomerizes to yield the high affinity nucleotide-free complex (reviewed in Cherfils and Chardin, 1999). The ternary complexes are usually difficult to study because their affinity is low. In Arf, the cytosol/membrane switch superimposes onto that scheme, but luckily the ternary complex can be trapped under two circumstances. One is found in nature in the form of the fungal metabolite BFA, which has been widely used over the years as a powerful inhibitor of Golgi traffic (reviewed in Klausner et al., 1992). The other one is the above-mentioned mutation of the glutamic finger to a lysine. Both interferences yield complexes with distinct partitioning to membranes, which further refine our understanding of the discrete steps of the exchange reaction. Investigation of the biochemical mechanism of action of BFA has been made possible by the identification of the amino acids that determine the sensitivity to BFA in yeast Arf GEFs using random mutagenesis, which were then introduced in the otherwise BFA-insensitive ARNO-Sec7 (Peyroche et al., 1999). Inhibition by BFA increases with the addition of Arf1•GDP, ruling out a competition between Arf and the drug for binding the Sec7 domain. Using gel filtration experiments and radiolabeled BFA and nucleotides, the target of BFA was shown to be the complex between Arf•GDP and the Sec7 domain, which it traps in an abortive complex (Peyroche et al., 1999; Robineau et al., 2000). BFA efficiently traps both Arf1∆17 and myr-Arf1•GDP, and in the latter case the trapped complex has a low affinity for membranes comparable to that of myr-Arf1•GDP (BeraudDufour et al., 1999; Robineau et al., 2000). Thus, BFA is proposed to act as

THE GDP/GTP CYCLE OF ARF PROTEINS

19 41

a double-face adhesive tape, gluing the Sec7 domain to Arf•GDP which has not yet undertaken the interswitch toggle (Beraud-Dufour et al., 1999). These studies also suggest that BFA has no direct effect on the membrane/cytosol partitioning of the trapped Sec7-Arf•GDP complex. This non-competitive mechanism was also proposed for members of the p200/BIG family of Arf GEFs, which are likely cellular targets of BFA in mammalian cells (Mansour et al., 1999). It should be noted that this noncompetitive inhibition mechanism is all the more remarkable in that it targets a low-affinity complex, yet it is very efficient in the cell (Peyroche et al., 1999; reviewed in Chardin and McCormick, 1999). It is also very specific, as it does not target Arf6 in vitro (our unpublished data) nor in vivo (Peters et al., 1995), and Arf1 is a target for BFA only with certain exchange factors. These observations make BFA a model inhibitor for GEF-catalyzed activation of small GTP-binding proteins. Differences in the structures and flexibility of Arf1•GDP and Arf6•GDP (Menetrey et al., 2000), and in the flexibility of the Sec7 domains at the residues that determine BFAsensitivity (Renault et al., 2002) may contribute to this specificity. In contrast, the complex of ARNO-K156-Sec7 with Arf•GDP forms only at low Mg2+ concentration (µM), either in solution with Arf1∆17 or in the presence of membranes with myr-Arf1, in which case its affinity for membranes is comparable to that of the nucleotide-free complex (BeraudDufour et al., 1999; Beraud-Dufour et al., 1998). Thus, the E156K mutation probably traps still another discrete step of the reaction, where the Nterminal hasp binds to membranes where it is stabilized by the interswitch toggle, while only the final nucleotide dissociation step is inhibited. Mutation of the glutamic finger does not interfere with the formation of the BFA complex (Robineau et al., 2000), reinforcing the idea that the glutamic finger is not involved until late in the course of the exchange reaction. The mechanism of the Sec7 domain can thus be separated in (at least) two steps: a docking step where the Sec7 domain binds to Arf•GDP; and a catalytic step where the glutamate has a crucial role in dissociating the GDP nucleotide. Comparison of yeast Gea2-Sec7 in complex with nucleotide-free human Arf1∆17 (Goldberg, 1998) with its unbound structure (Renault et al., 2002) shows that a subdomain motion closes the hydrophobic groove upon binding to Arf, suggesting that Arf•GDP may bind onto the side of the groove opposite to the glutamic finger at the docking step, while closure of Sec7 subdomains brings the glutamic finger into the active site at the catalytic step (Renault et al., 2002).

20 42

3.4

PASQUALATO, RENAULT AND CHERFILS

A comprehensive model for biochemical and structural front-back communication

Biochemical and structural data together can be formalized into a remarkably consistent and detailed model for the activation reaction of Arfs by their GEFs, which also provides valuable insight into the mechanisms that take place in the cell. This model is summarized in Figure 7. 1) Myr-Arf•GDP is mainly cytosolic, but it can associate weakly with membranes by its N-terminal myristate. In its cytosolic form, Arf•GDP has a retracted interswitch that is fastened by the amphipathic N-terminal helix. The small fraction of membrane-bound Arf•GDP is probably in equilibrium between a form where only the myristate is inserted in the membrane, and a form where both the myristate and the N-terminal helix interact with the membrane. In either case, the interswitch remains retracted, such that the hydrophobic pocket for the N-terminal hasp is available for reversible unbinding. The retracted interswitch prevents the binding of GTP by inserting an aspartate in the γ-phosphate binding-site. 2) The Arf GEF is recruited in the vicinity of a membrane by PH domains or by yet to be identified domains, where it encounters myrArf•GDP. It acts either on the fraction of Arf•GDP where the hasp has already been unfastened, or promotes the unfastening of the N-terminal hasp, which is facilitated by its proximity to membranes; two possibilities that are currently not resolved. In solution, interaction of myr-Arf•GDP with the Sec7 domain of soluble Arf GEFs is also likely to occur, but it is unproductive and dissociates readily. The glutamic finger is not involved at that early docking stage, which, together with the localization of residues that determine the sensitivity to BFA, suggests that the docking site of myrArf•GDP onto Sec7 domains is remote from the nucleotide-binding site, possibly centered on the switch 2 region of Arf and the C-terminal subdomain of the Sec7 domain. This step is the target of BFA in the cell. 3) The Sec7 domain then induces the interswitch toggle, which destroys the binding pocket for the N-terminal hasp and thereby anchors Arf to membranes. However, this complex is still able to bind GDP, indicating that conformational changes such as the distortion of the P-loop may not have taken place yet. Mutation of the glutamic finger traps this intermediate complex. 4) Closure of the subdomains of the Sec7 domain, as well as changes in the switch regions and possibly the orientation of Arf, eventually yield a large protein-protein interface and position the glutamic finger into the active site where it distorts the P-loop and dissociates the GDP phosphates by charge repulsion. The glutamic finger then proceeds to stabilize the

THE GDP/GTP CYCLE OF ARF PROTEINS

21 43

nucleotide-binding site by mimicking the interactions of the β-phosphate with the P-loop.

Figure 7. The model for the nucleotide/membrane switch. See text for details of steps 1 to 5. Gua=guanine base, myr=myristate, E=glutamic finger, D=aspartate from the DxGGQ motif, Sec7Nt= Sec7 N-terminal domain; Sec7Ct= Sec7 C-terminal domain.

22 44

PASQUALATO, RENAULT AND CHERFILS

5) Finally, entry of GTP, binding first by the guanine base whose binding site is readily accessible in the nucleotide-free complex, dissociates the GEF and provides alternative interactions that stabilize the active conformation of the switch/interswitch machinery. As the interswitch is now firmly stabilized, the interaction of the N-terminal hasp with membranes becomes irreversible as long as Arf is bound to GTP. The model and results discussed so far argue clearly for the need of both the guanine nucleotide exchange factor and the membrane to perform an efficient activation of Arf proteins. Notably, both membranes and the Sec7 domain induce conformational changes that drive the activation process. For these reasons, catalyzed nucleotide exchange towards Arf is composed of two factors acting in concert, but mostly inactive on their own: a classical GEF that changes the conformation of the switch regions and builds up the GTP-binding site, and a membrane co-factor unfastening the N-terminal hasp. We have recently proposed that this mechanism of front-back communication by the interswitch toggle is also found in most Arf-like proteins as well as in Sar1, suggesting that these proteins will also be activated by a “dual” exchange factor (Pasqualato et al., 2002).

4.

ACKNOWLEDGEMENTS

We thank Sonia Paris (CNRS, Valbonne) for critical reading of the manuscript. This work was supported by grants from the Association pour la Recherche contre le Cancer, and from the CNRS, INSERM and French Research Ministry.

REFERENCES Al-Awar, O., Radhakrishna, H., Powell, N. N., and Donaldson, J. G. (2000). Separation of membrane trafficking and actin remodeling functions of ARF6 with an effector domain mutant. Mol. Cell Biol., 20, 5998-6007. Amor, J. C., Harrison, D. H., Kahn, R. A., and Ringe, D. (1994). Structure of the human ADP-ribosylation factor 1 complexed with GDP. Nature, 372, 704-708. Amor, J. C., Horton, J. R., Zhu, X., Wang, Y., Sullards, C., Ringe, D., Cheng, X., and Kahn, R. A. (2001). Structures of yeast ARF2 and ARL1: distinct roles for the N terminus in the structure and function of ARF family GTPases. J. Biol. Chem., 276, 42477-42484. Antonny, B., Beraud-Dufour, S., Chardin, P., and Chabre, M. (1997). N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry, 36, 4675-4684. Beraud-Dufour, S., Paris, S., Chabre, M., and Antonny, B. (1999). Dual interaction of ADP ribosylation factor 1 with Sec7 domain and with lipid membranes during catalysis of guanine nucleotide exchange. J. Biol. Chem., 274, 37629-37636.

THE GDP/GTP CYCLE OF ARF PROTEINS

23 45

Beraud-Dufour, S., Robineau, S., Chardin, P., Paris, S., Chabre, M., Cherfils, J., and Antonny, B. (1998). A glutamic finger in the guanine nucleotide exchange factor ARNO displaces Mg2+ and the beta-phosphate to destabilize GDP on ARF1. EMBO J., 17(13), 3651-3659. Betz, S. F., Schnuchel, A., Wang, H., Olejniczak, E. T., Meadows, R. P., Lipsky, B. P., Harris, E. A., Staunton, D. E., and Fesik, S. W. (1998). Solution structure of the cytohesin1 (B2-1) Sec7 domain and its interaction with the GTPase ADP ribosylation factor 1. Proc. Natl. Acad. Sci., USA, 95, 7909-7914. Chardin, P., and McCormick, F. (1999). Brefeldin A: the advantage of being uncompetitive. Cell, 97, 153-155. Chardin, P., Paris, S., Antonny, B., Robineau, S., Beraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996). A human exchange factor for ARF contains Sec7- and pleckstrinhomology domains. Nature, 384, 481-484. Chavrier, P., and Goud, B. (1999). The role of ARF and Rab GTPases in membrane transport. Curr. Opin. Cell Biol., 11, 466-475. Cherfils, J., and Chardin, P. (1999). GEFs: structural basis for their activation of small GTPbinding proteins. Trends Biochem. Sci., 24, 306-311. Cherfils, J., Menetrey, J., Mathieu, M., Le Bras, G., Robineau, S., Beraud-Dufour, S., Antonny, B., and Chardin, P. (1998). Structure of the Sec7 domain of the Arf exchange factor ARNO. Nature, 392, 101-105. Click, E. S., Stearns, T., and Botstein, D. (2002). Systematic Structure-Function Analysis of the Small GTPase Arf1 in Yeast. Mol. Biol. Cell, 13, 1652-1664. Coleman, D. E., Berghuis, A. M., Lee, E., Linder, M. E., Gilman, A. G., and Sprang, S. R. (1994). Structures of active conformations of Gi alpha 1 and the mechanism of GTP hydrolysis. Science, 265, 1405-1412. Cullen, P. J., and Chardin, P. (2000). Membrane targeting: what a difference a G makes. Curr. Biol., 10, R876-878. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1993). Myristoylation is not required for GTP-dependent binding of ADP-ribosylation factor ARF1 to phospholipids. J. Biol. Chem., 268, 24531-24534. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1995). Myristoylation of ADPribosylation factor 1 facilitates nucleotide exchange at physiological Mg2+ levels. J. Biol. Chem., 270, 1337-1341. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1996). Myristoylation-facilitated binding of the G protein ARF1GDP to membrane phospholipids is required for its activation by a soluble nucleotide exchange factor. J. Biol. Chem., 271, 1573-1578. Franco, M., Peters, P. J., Boretto, J., van Donselaar, E., Neri, A., D'Souza-Schorey, C., and Chavrier, P. (1999). EFA6, a sec7 domain-containing exchange factor for ARF6, coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 14801491. Goldberg, J. (1998). Structural basis for activation of ARF GTPase: mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Goldberg, J. (1999). Structural and functional analysis of the ARF1-ARFGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Greasley, S. E., Jhoti, H., Teahan, C., Solari, R., Fensome, A., Thomas, G. M., Cockcroft, S., and Bax, B. (1995). The structure of rat ADP-ribosylation factor-1 (ARF-1) complexed to GDP determined from two different crystal forms. Nat. Struct. Biol., 2, 797-806. Haun, R. S., Tsai, S. C., Adamik, R., Moss, J., and Vaughan, M. (1993). Effect of myristoylation on GTP-dependent binding of ADP-ribosylation factor to Golgi. J. Biol. Chem., 268, 7064-7068.

24 46

PASQUALATO, RENAULT AND CHERFILS

Kahn, R. A. (1991). Fluoride is not an activator of the smaller (20-25 kDa) GTP-binding proteins. J. Biol. Chem., 266, 15595-15597. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kahn, R. A., Goddard, C., and Newkirk, M. (1988). Chemical and immunological characterization of the 21-kDa ADP-ribosylation factor of adenylate cyclase. J. Biol. Chem., 263, 8282-8287. Kahn, R. A., Randazzo, P., Serafini, T., Weiss, O., Rulka, C., Clark, J., Amherdt, M., Roller, P., Orci, L., and Rothman, J. E. (1992). The amino terminus of ADP-ribosylation factor (ARF) is a critical determinant of ARF activities and is a potent and specific inhibitor of protein transport. J. Biol. Chem., 267, 13039-13046. Klausner, R. D., Donaldson, J. G., and Lippincott-Schwartz, J. (1992). Brefeldin A: insights into the control of membrane traffic and organelle structure. J. Cell Biol., 116, 1071-1080. Losonczi, J. A., and Prestegard, J. H. (1998). Nuclear magnetic resonance characterization of the myristoylated, N-terminal fragment of ADP-ribosylation factor 1 in a magnetically oriented membrane array. Biochemistry, 37, 706-716. Losonczi, J. A., Tian, F., and Prestegard, J. H. (2000). Nuclear magnetic resonance studies of the N-terminal fragment of adenosine diphosphate ribosylation factor 1 in micelles and bicelles: influence of N-myristoylation. Biochemistry, 39, 3804-3816. Macia, E., Chabre, M., and Franco, M. (2001). Specificities for the small G proteins ARF1 and ARF6 of the guanine nucleotide exchange factors ARNO and EFA6. J. Biol. Chem., 276, 24925-24930. Macia, E., Paris, S., and Chabre, M. (2000). Binding of the PH and polybasic C-terminal domains of ARNO to phosphoinositides and to acidic lipids. Biochemistry, 39, 5893-5901. Mansour, S. J., Skaug, J., Zhao, X. H., Giordano, J., Scherer, S. W., and Melancon, P. (1999). p200 ARF-GEP1: a Golgi-localized guanine nucleotide exchange protein whose Sec7 domain is targeted by the drug brefeldin A. Proc. Natl. Acad. Sci., USA, 96, 7968-7973. Menetrey, J., Macia, E., Pasqualato, S., Franco, M., and Cherfils, J. (2000). Structure of Arf6GDP suggests a basis for guanine nucleotide exchange factors specificity. Nat. Struct. Biol., 7, 466-469. Mossessova, E., Gulbis, J. M., and Goldberg, J. (1998). Structure of the guanine nucleotide exchange factor Sec7 domain of human arno and analysis of the interaction with ARF GTPase. Cell, 92, 415-423. Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997). Role of protein-phospholipid interactions in the activation of ARF1 by the guanine nucleotide exchange factor Arno. J. Biol. Chem., 272, 22221-22226. Pasqualato, S., Menetrey, J., Franco, M., and Cherfils, J. (2001). The structural GDP/GTP cycle of human Arf6. EMBO Rep., 2, 234-238. Pasqualato, S., Renault, L., and Cherfils, J. (2002). Arf, Arl, Arp and Sar proteins: a family of GTP-binding proteins with a structural device for front-back communication. EMBO Rep., 3, 1035-1041. Peters, P. J., Hsu, V. W., Ooi, C. E., Finazzi, D., Teal, S. B., Oorschot, V., Donaldson, J. G., and Klausner, R. D. (1995). Overexpression of wild-type and mutant ARF1 and ARF6: distinct perturbations of nonoverlapping membrane compartments. J. Cell. Biol., 128, 1003-1017. Peyroche, A., Antonny, B., Robineau, S., Acker, J., Cherfils, J., and Jackson, C. L. (1999). Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell, 3, 275-285.

THE GDP/GTP CYCLE OF ARF PROTEINS

25 47

Randazzo, P. A., Terui, T., Sturch, S., Fales, H. M., Ferrige, A. G., and Kahn, R. A. (1995). The myristoylated amino terminus of ADP-ribosylation factor 1 is a phospholipid- and GTP-sensitive switch. J. Biol. Chem., 270, 14809-14815. Renault, L., Christova, P., Guibert, B., Pasqualato, S., and Cherfils, J. (2002). Mechanism of domain closure of Sec7 domains and role in BFA sensitivity. Biochemistry, 41, 36053612. Renault, L., Kuhlmann, J., Henkel, A., and Wittinghofer, A. (2001). Structural basis for guanine nucleotide exchange on Ran by the regulator of chromosome condensation (RCC1). Cell, 105, 245-255. Robineau, S., Chabre, M., and Antonny, B. (2000). Binding site of brefeldin A at the interface between the small G protein ADP-ribosylation factor 1 (ARF1) and the nucleotideexchange factor Sec7 domain. Proc. Natl. Acad. Sci., USA, 97, 9913-9918. Scheffzek, K., Ahmadian, M. R., Kabsch, W., Wiesmuller, L., Lautwein, A., Schmitz, F., and Wittinghofer, A. (1997). The Ras-RasGAP complex: structural basis for GTPase activation and its loss in oncogenic Ras mutants. Science, 277, 333-338. Scheffzek, K., Klebe, C., Fritz-Wolf, K., Kabsch, W., and Wittinghofer, A. (1995). Crystal structure of the nuclear Ras-related protein Ran in its GDP-bound form. Nature, 374, 378381. Serafini, T., Orci, L., Amherdt, M., Brunner, M., Kahn, R. A., and Rothman, J. E. (1991). ADP-ribosylation factor is a subunit of the coat of Golgi-derived COP-coated vesicles: a novel role for a GTP-binding protein. Cell, 67, 239-253. Shevell, D. E., Leu, W. M., Gillmor, C. S., Xia, G., Feldmann, K. A., and Chua, N. H. (1994). EMB30 is essential for normal cell division, cell expansion, and cell adhesion in Arabidopsis and encodes a protein that has similarity to Sec7. Cell, 77, 1051-1062. Skippen, A., Jones, D. H., Morgan, C. P., Li, M., and Cockcroft, S. (2002). Mechanism of ADP ribosylation factor-stimulated phosphatidylinositol 4,5-bisphosphate synthesis in HL60 cells. J. Biol. Chem., 277, 5823-5831. Someya, A., Sata, M., Takeda, K., Pacheco-Rodriguez, G., Ferrans, V. J., Moss, J., and Vaughan, M. (2001). ARF-GEP(100), a guanine nucleotide-exchange protein for ADPribosylation factor 6. Proc. Natl. Acad. Sci., USA, 98, 2413-2418. Steinmann, T., Geldner, N., Grebe, M., Mangold, S., Jackson, C. L., Paris, S., Galweiler, L., Palme, K., and Jurgens, G. (1999). Coordinated polar localization of auxin efflux carrier PIN1 by GNOM ARF GEF. Science, 286, 316-318. Szafer, E., Pick, E., Rotman, M., Zuck, S., Huber, I., and Cassel, D. (2000). Role of coatomer and phospholipids in GTPase-activating protein-dependent hydrolysis of GTP by ADPribosylation factor-1. J. Biol. Chem., 275, 23615-23619. Szafer, E., Rotman, M., and Cassel, D. (2001). Regulation of GTP hydrolysis on ADPribosylation factor-1 at the Golgi membrane. J. Biol. Chem., 276, 47834-47839. Vetter, I. R., Arndt, A., Kutay, U., Gorlich, D., and Wittinghofer, A. (1999). Structural view of the Ran-Importin beta interaction at 2.3 A resolution. Cell, 97, 635-646. Vetter, I. R., and Wittinghofer, A. (2001). The guanine nucleotide-binding switch in three dimensions. Science, 294, 1299-1304. Walker, M. W., Bobak, D. A., Tsai, S. C., Moss, J., and Vaughan, M. (1992). GTP but not GDP analogues promote association of ADP-ribosylation factors, 20-kDa protein activators of cholera toxin, with phospholipids and PC-12 cell membranes. J. Biol. Chem., 267, 3230-3235. Zerial, M., and McBride, H. (2001). Rab proteins as membrane organizers. Nat. Rev. Mol. Cell Biol., 2, 107-117.

26 48

PASQUALATO, RENAULT AND CHERFILS

Zhao, L., Helms, J. B., Brunner, J., and Wieland, F. T. (1999). GTP-dependent binding of ADP-ribosylation factor to coatomer in close proximity to the binding site for dilysine retrieval motifs and p23. J. Biol. Chem., 274, 14198-14203.

Chapter 3 ARF AND PHOSPHOLIPIDS PAUL A. RANDAZZO, ZHONGZHEN NIE, AND DIANNE S. HIRSCH Laboratory of Cellular Oncology, CCR, NCI, Bethesda, USA

Abstract:

1.

Arfs have been implicated in remodeling of both biological membranes and the actin cytoskeleton. In each case, Arf function appears to be integrally related to lipid metabolism. As regulators of membrane traffic, Arfs bind to membrane surfaces and affect membrane structure. Effects of Arf on the actin cytoskeleton may be mediated through acid phospholipids. Examination of the interactions of Arfs and lipids has revealed complex relationships. Arfs directly bind lipids. Arfs also bind and activate two enzymes that metabolize lipids, phospholipase D and phosphatidylinositol 4-phosphate 5-kinase. The products of these enzymes, phosphatidic acid and phosphatidylinositol 4,5bisphosphate, affect both positive and negative regulators of Arfs. Although precise mechanisms have yet to be defined, regulation of the concentrations of these signaling lipids appears to mediate at least some effects of Arfs.

INTRODUCTION

Membrane traffic is the transport of membrane and proteins between membrane bound organelles in eukaryotes. The actin cytoskeleton is integrally associated with biological membranes. It likely is involved in membrane transport and actin, in turn, is affected by lipid components of biological membranes. Given Arfs’ regulatory roles in both membrane traffic and actin remodeling, the importance of lipids for Arf function was anticipated. Not anticipated was the number of relationships identified between Arfs and lipids. The role of lipids in Arf function was first appreciated with the initial identification of Arf as a cofactor for cholera toxin catalyzed ADPribosylation of the heterotrimeric G-protein Gs (Kahn and Gilman, 1986; Kahn and Gilman, 1984). Mixed micelles of dimyristoyl phosphatidylcholine and cholate increased the efficiency of ADPribosylation in a reaction mixture containing GTP, NAD, cholera toxin, 49

2 50

RANDAZZO, NIE, AND HIRSCH

Gs and Arf. At least part of the effect of the micelles in this reaction is related to the properties of Arf. Arf•GTP is the cofactor for cholera toxin; however, Arf•GDP is purified from mammalian tissues. For activity in the assay, Arf has to exchange GTP for GDP. Exchange of nucleotide on native Arf is greatly enhanced by lipids. Once exchange does occur, Arf•GTP binds tightly to and is stabilized by lipid interfaces, where it can function as a cofactor for cholera toxin (see Chapter 10 for additional details). The basis of the lipid requirement for nucleotide exchange was uncovered shortly after the cloning of Arf and identification of N-terminal myristate as a modification. Arf bound to GTP was found to directly associate with hydrophobic surfaces through Arf’s N-terminal myristate and -helix. Arf also interacts with specific phospholipids, the phosphoinositides, within biological membranes. A cluster of basic amino acids mediates the binding to phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2). This specific interaction, together with the general hydrophobic association of Arf with lipids, directs Arf binding to target proteins, by both recruiting Arf to specific surfaces and by affecting the affinity of Arf for effectors. For instance, acid phospholipids have been found to influence the binding of two Arf effectors, GGA and Arfaptin, to Arf. Phosphatidic acid (PA), PI(4,5)P2, and their metabolites have also been found to influence the activity Arfdirected guanine nucleotide exchange factors (GEFs) and Arf GTPase activating proteins (GAPs). Adding complexity to the regulation of Arf by lipids is the ability of Arf to directly influence the lipid composition of the membranes to which it binds. Arf activates both phospholipase D (PLD), which hydrolyzes phosphatidylcholine (PC) to PA, and phosphatidylinositol 4-phosphate 5kinase (PIP kinase I), which transfers a phosphate from ATP to phosphatidylinositol 4-phosphate (PI(4)P) to form PI(4,5)P2 (also see Chapter 11). The physiological relevance of many of these reactions and their relationships to one another is yet to be determined. In this chapter, the literature examining the impact of phospholipids on Arf function will be discussed followed by some limited speculation about how these reactions may be related to one another and how the lipids themselves may mediate the effects of Arf on membrane traffic and the actin cytoskeleton.

2.

STRUCTURAL DETERMINANTS OF ARF INTERACTIONS WITH LIPIDS

Several characteristics of Arf proteins contribute to their unique nucleotidedependent association with membranes. The most critical are the amino

ARF AND PHOSPHOLIPIDS

3 51

terminal myristate and amphipathic -helix. Although crystal structures of the N-myristoylated Arfs (myrArf) have not been determined, crystal structures of the non-acylated proteins have provided insights into the potential role of the myristate. The crystal structure of Arf•GDP (Greasley et al., 1995; Amor et al., 1994) reveals a hydrophobic cleft in which the Nterminal myristate likely inserts. In the Arf•GTP structure, the cleft is rearranged and the myristate is predicted to be free to interact with a hydrophobic surface (Goldberg, 1999; Goldberg, 1998). Lipids would, therefore, bind to myrArf•GTP more tightly than to myrArf•GDP, stabilizing the GTP bound form of the protein. Consistent with this interpretation of the role of myristate, GTPS dissociation from myrArf was slowed at least seven-fold by mixed micelles of dimyristoyl phosphatidylcholine and cholate and GDP dissociation was accelerated 3fold. In contrast, mixed micelles had no effect on GDP dissociation from nonmyristoylated Arf1 and slowed GTPS dissociation by only 2-fold (Randazzo et al., 1995). Thus, the proposed structural differences in the myristate between the GTP and GDP structures can explain two biochemical results: (i) lipids promote GTP binding to myrArf to a much greater extent than to nonmyrArf1 (Randazzo et al., 1995; Randazzo and Kahn, 1995) and; (ii) myrArf1•GTP binds to lipid micelles much more tightly than does either myrArf1•GDP or nonmyrArf1•GTP (Randazzo et al., 1993). The amino terminal -helix of Arf1 also contributes to association with lipid surfaces. This domain of Arf1 was not present in the Arf1•GTP crystal structure as an N-terminal truncation mutant was used to obtain crystals. However, based on other differences between the GTP and GDP bound structures of Arf1, changes in the N-terminal -helix were inferred (Goldberg, 1999; Goldberg, 1998). In Arf1•GTP, the amphipathic helix is thought to be free to bind hydrophobic surfaces. Consistent with this prediction, GTPS dissociation from Arf1 is slowed by lipids. In addition, nonmyristoylated Arf1•GTPS binds to phospholipids, although with a much lower affinity than myristoylated Arf1 (Franco et al., 1993; Randazzo et al., 1993; Antonny et al., 1997a), so that detection of the interaction required the presence of 30 mM phospholipid, in contrast to 100 M to 3 mM total phospholipid used to study binding to myristoylated Arf (Franco et al., 1993; Randazzo et al., 1993; Antonny et al., 1997a; Paris et al., 1997a; Franco et al., 1995). Further supporting the role of the -helix in lipid association, deletion of either the first 13 or 17 amino acids of Arf1 resulted in proteins (Arf1- 13, Arf1- 17) that did not detectably associate with lipid vesicles. Nucleotide binding to Arf1- 13 and Arf1- 17 was not affected by phospholipids (Hong et al., 1995; Hong et al., 1994; Randazzo et al., 1995; Randazzo et al., 1994; Kam et al., 2000).

4 52

RANDAZZO, NIE, AND HIRSCH

A third structural feature of Arfs that contributes to membrane binding is a patch of solvent-exposed basic residues (Greasley et al., 1995; Amor et al., 1994). These residues interact electrostatically with acidic phospholipids that are rich in the cytoplasmic face of biological membranes. Prominent among these residues are those contributed from the N-terminal -helix and the Cterminus (K10, K15, K16, K59, R178 and K181). Consistent with a specific interaction of PI(4,5)P2 with Arf, PI(4,5)P2 stimulated GDP dissociation from Arf under a specific set of conditions (Terui et al., 1994), while other acid phospholipids were not effective. GTP dissociation was not affected by PI(4,5)P2. Although these data did not address the potential physiologic function of PI(4,5)P2 binding, they did demonstrate the specificity of the interaction between Arf and PI(4,5)P2. Subsequent studies directly examined binding (Kam et al., 2000; Randazzo, 1997a). By using nonmyristoylated protein, the role of the residues comprising this basic patch in PI(4,5)P2 binding could be examined independently of the general hydrophobic interactions with nonspecific lipids. Arf1•GTPS was found to bind PI(4,5)P2 with a Kd of approximately 40 M. Similar results were obtained whether lipids were presented in the form of large unilamellar vesicles or mixed micelles with Triton X-100. Arf1•GTPS also bound to PA and phosphatidylserine, but with lower affinity. Mutation of K15, K16, R178 or K181 to leucine reduced the affinity for PI(4,5)P2.

3.

ROLE OF LIPID BINDING IN ARF FUNCTION

3.1

General hydrophobic interaction

The importance of Arf-lipid binding in biochemical and biological functions of Arfs has been established from a number of studies. N-myristoylation of Arfs is co-translational and stable so no non-myristoylated Arf is thought to be present in cells. Although bacterially expressed nonmyristoylated Arf has been useful for dissecting the role of lipid interactions with other parts of Arf, a description of the role of the critical myristoyl group is necessary for understanding the biological function of Arf. The nonspecific association of Arf with phospholipids mediated by the N-terminal myristate has been found to have at least two distinct roles.

3.2

Nucleotide exchange

The N-terminal myristate increases the rate of nucleotide exchange on Arf catalyzed by the ARNO family of GEFs. Two mechanisms explaining the effect of the myristate have been considered. Although the membrane

ARF AND PHOSPHOLIPIDS

5 53

association is greatest for the GTP bound form of the Arf, myrArf•GDP also associates with membranes (Franco et al., 1996; Beraud-Dufour et al., 1999), while non-myristoylated Arf•GDP does not. Targeting myrArf•GDP to the membrane is thought to facilitate interaction with ARNO family exchange factors that are recruited to membranes through their PH domains (described below and in Chapters 4-6). A second effect of myristate in GEF-catalyzed nucleotide exchange is facilitation of the binding of lysines 10, 15 and 16 to acid phospholipids such as PI(4,5)P2. Chardin and colleagues (Paris et al., 1997b) found that ARNO catalyzed nucleotide exchange on both myristoylated and nonmyristoylated Arf1 in the presence of PI(4,5)P2. Under physiologic conditions of Mg2+ and ionic strength, myristoylated Arf1 was the preferred substrate. Lowering Mg2+ and ionic strength greatly enhanced the rate of exchange using nonmyristoylated Arf1. The authors noted that low Mg2+ and salt concentrations favor electrostatic interactions such as those between acid phospholipids and basic amino acid residues. Based on these observations, the authors concluded that the electrostatic binding between acid phospholipids and lysines 10, 15 and 16 of Arf1, though weak under physiologic conditions, was necessary for activity. The myristate, by providing a second binding site, would increase the affinity of Arf for the membrane, thereby driving these otherwise weak electrostatic interactions. Note that these results are consistent with the reported (Terui et al., 1994) effect of PI(4,5)P2 to increase nucleotide exchange on Arf1.

3.3

Coat proteins

The high affinity of Arf1•GTP for membranes, imparted by the myristate on Arf, is also critical for recruiting coat proteins to membranes (see Chapters 12 and 13 for details). Vesicle coat protein recruitment to Golgi membranes has been studied in reconstitution systems (Donaldson et al., 1992; Rothman, 1994; Serafini et al., 1991; Orci et al., 1993; Palmer et al., 1993; Stamnes and Rothman, 1993; Ooi et al., 1998; Dell'Angelica et al., 2000; Zhu et al., 1999; Traub et al., 1993). Coat proteins, purified away from Arfs, were incubated with recombinant Arf proteins and either isolated Golgi membranes or lipid vesicles. In these experiments, nonmyristoylated Arfs did not recruit coat proteins (including coatomer, AP-1, and GGA) to lipid vesicles or to isolated Golgi membranes, whereas myrArf1 did. The inability to recruit the coat proteins is not due to lack of protein – protein interaction. Nonmyristoylated Arf1 binds GGA as well as myristoylated Arf1 (Jacques et al., 2002). Instead, the myristate increases the affinity of Arf for the membrane surface where Arf can function as a membrane binding site for the coat proteins.

6 54

RANDAZZO, NIE, AND HIRSCH

The amino terminal -helix has a role in directing protein interaction. Although myristate is not required for interaction of Arf with GGA and GAP, the N-terminal -helix does contribute to binding. Arf1- 13 has a lower affinity for both GGA and GAP than does nonmyristoylated Arf1 (Jacques et al., 2002). Furthermore, although nonmyristoylated Arf1 can activate PLD to some extent, Arf1- 17 has no detectable effect on PLD (Jones et al., 1999). Further studies are needed to determine the basis for the decreased affinity for GGA and GAP, and decreased ability to activated PLD. Two possibilities appear most likely and are considered. First, the effects of deleting the amino terminus may be due to loss of its membrane binding and/or a specific orientation of Arf at the membrane surface. The other possibility is that, in addition to binding lipids, the N-terminal -helix of Arf directly interacts with target proteins.

4.

INTERACTION OF SIGNALING PHOSPHOLIPIDS WITH ARF AND ITS REGULATORS

As is true for nonspecific hydrophobic interactions, the binding of specific lipids to Arf also controls Arf function and regulation. In this section, the influence of phosphoinositide binding to Arf, Arf GEFs, and Arf GAPs is discussed.

4.1

Acid phospholipid binding to Arf

Acidic phospholipid binding to the cluster of basic amino acids on the surface of Arf has a role beyond membrane targeting. A role in GEFcatalyzed nucleotide exchange is discussed above. PI(4,5)P2 binding to Arf1 also has a role in the function of the GAP, ASAP1 (Kam et al., 2000; Randazzo, 1997a). PI(4,5)P2 stimulation of GAP activity correlated with PI(4,5)P2 binding to Arf1. Blocking PI(4,5)P2 binding, either by sequestering PI(4,5)P2 or by introducing mutations into Arf that reduced PI(4,5)P2 binding, also reduced PI(4,5)P2-dependent GAP activity. Similarly, PI(4,5)P2 may direct Arf interaction with PLD. Arf1 with mutations in two residues involved in PI(4,5)P2 binding, lysines 15 and 16, did not stimulate PLD activity (Jones et al., 1999). The effect of acid phospholipid binding to Arf on protein interactions is selective. PI(4,5)P2 did not have an effect on Arf binding to GGA and mutation of lysines 15 and 16 did not affect Arf binding to GGA. Instead of PI(4,5)P2, PA enhanced the interaction of GGA with Arf•GTP. Interaction of Arf with Arfaptin and Arf activity as a cofactor for cholera toxin was

ARF AND PHOSPHOLIPIDS

7 55

inhibited by acid phospholipids, including PI(4,5)P2 (K. Jacques, Z. Nie, K. Stanley and P. A. Randazzo, unpublished observations). Taken together, these results indicate that the specific phospholipid environment can affect the affinity of Arf for particular targets.

4.2

Acid phospholipid binding to Arf GEFs

The ARNO family of Arf GEFs (comprised of ARNO1, cytohesin 1/ARNO2, GRP1 and cytohesin 4 (Moss and Vaughan, 2002; Ogasawara et al., 2000; Klarlund et al., 1997; Chardin et al., 1996)) and the EFA6 family (Franco et al., 1999) have PH domains and are affected by phospholipids (see Chapters 2 and 4-6; Jackson and Casanova, 2000; Moss and Vaughan, 2002) (Figure 1A). The lipid requirements for the ARNO family proteins have been most extensively characterized. The function of the PH domain is to recruit the GEF to a lipid surface (Paris et al., 1997b; Beraud-Dufour et al., 1999; Chardin et al., 1996; Pacheco-Rodriguez et al., 1998). When using a truncated Arf, Arf1- 17, that is soluble in the GTP bound form, PI(4,5)P2 has no effect on activity. Furthermore, truncated ARNO lacking the PH domain is active when using Arf1- 17. As described above, acid phospholipid binding to Arf also contributes to catalysis of exchange by ARNO. The independent regulation of ARNO family members may be in part achieved through differential activation by specific phosphoinositide isomers. Grp1 is specifically activated by PI(3,4,5)P3 (Klarlund et al., 1998; Klarlund et al., 1997). ARNO1 and cytohesin 1 have been reported to interact with either PI(4,5)P2 or PI(3,4,5)P3 (Bourgoin et al., 2002; Venkateswarlu et al., 1998; Chardin et al., 1996; Moss and Vaughan, 2002; Venkateswarlu et al., 1999). The basis for the differences in lipid specificity was examined (Klarlund et al., 2000; Lietzke et al., 2000) and found to be due to the replacement of a triglycine motif, found in ARNO and cytohesin, with a diglycine motif in Grp1 (Figure 1B). Splice variation within the mRNAs encoding each of these proteins produce a mixture of products containing diglycine (PIP3 binding) and triglycine (promiscuous or PI(4,5)P2 binding) motifs, and therefore heterogeneity in phosphoinositide binding (Klarlund et al., 2000; Moss and Vaughan, 2002; Ogasawara et al., 2000).

8 56

RANDAZZO, NIE, AND HIRSCH

Figure 1. Phospholipid-regulated Arf GEFs and GAPs. (A) Schematic showing the domain organization of ARNO and EFA6 family Arf GEFs. (B) Amino acid sequences of regions conferring lipid specificity of ARNO1 and GRP1. (C) Schematic of the Arf GAPs. Ankyrin repeat, ANK; Arf GAP catalytic domain, Arf GAP; coiled-coil, CC; pleckstrin homology domain, PH; proline-rich domain, PR; Rho GAP domain, Rho GAP.

4.3

Arf GAPs and lipids

To date, fourteen mammalian Arf GAPs have been described (see Chapters 7-9; Cukierman et al., 1995; Premont et al., 1998; Vitale et al., 2000; Bagrodia et al., 1999; Brown et al., 1998; Premont et al., 2000; Miura et al., 2002; Krugmann et al., 2002; Nie et al., 2002). The proteins can be divided into three groups based on structure (Figure 1C). The first subfamily contains Arf GAP1 and 3. The protein has a catalytic domain, found in all proteins with GAP activity, comprised of a zinc finger motif. A second domain targets Arf GAP1 to the Golgi apparatus by binding to ERD2 (Aoe et al., 1999). A second subfamily contains the GITs (called APAP in the nomenclature introduced in this volume). Similar to Arf GAP1 in containing the Arf GAP catalytic domain at the amino terminus, the GITs also have

ARF AND PHOSPHOLIPIDS

9 57

ankyrin (ANK) repeat domains and unique domains involved in targeting the protein to the plasma membrane. The largest subgroup of Arf GAPs is the ASAPs, defined by a core of PH, Arf GAP and ANK repeat domains. All members of this subgroup contain additional domains N-terminal to this core. Some also have additional domains C-terminal of the core. The lipid dependence of Arf GAP1, GITs and ASAPs has been examined and are discussed below. 4.3.1

Arf GAP1

Arf GAP1 has unique lipid dependence among the Arf GAPs. In the first studies of Arf GAP1, PI(4,5)P2 was found to have a small but reproducible effect on activity. Those studies were performed with nonmyristoylated Arf1 (Randazzo 1997b; Makler et al., 1995). When myrArf1 was used, no effect of PI(4,5)P2 was seen (Antonny et al., 1997b). The most likely explanation for these results was that PI(4,5)P2 was recruiting the nonmyristoylated Arf1 to the hydrophobic surface with which the GAP was also associated. Myristoylated Arf1, bound to GTP, has a high affinity for the vesicles and hydrophobic surfaces, and there was no further effect of adding PI(4,5)P2. When Antonny and colleagues subsequently re-examined the lipid requirements of Arf GAP1 (Antonny et al., 1997b), they found that diacylglyerols (DAG) with monounsaturated acyl groups both increased the amount of GAP associated with vesicles and stimulated GAP activity. On examination of the effects of a number of synthetic lipids, the extent of stimulation was found to correlate with the relative volumes occupied by the polar head group and the acyl groups. The smaller the head group relative to the acyl group, the greater the activity. The monounsaturated acyl groups have a kink that results in a relatively large occupied volume. Dioleoylglycerol, with no head group and two kinked acyl groups, was the most effective of the series they examined. Saturated phosphatidylcholine, with a large head group and tightly packed acyl groups, was inhibitory. Gcs1p, a yeast homolog of Arf GAP1, was similar in lipid dependency. Based on this analysis, Cassel and colleagues proposed that Arf GAP1 binding to membranes is driven by hydrophobic contact of the protein with the acyl groups of the lipids. Large head groups and tight packing of acyl groups would tend to exclude the protein whereas small head groups and loosely packed acyl groups would favor interaction with the hydrophobic part of the protein. The importance of the activation of Arf GAP1 by DAG was thought to be related to the function of PLD. The product of PLD, PA, is rapidly hydrolyzed to ortho-phosphate and DAG. This set of reactions could

10 58

RANDAZZO, NIE, AND HIRSCH

constitute a negative feedback loop regulating Arf1 (see Figure 2, adapted from Antonny et al., 1997b).

Figure 2. Postulated feedback loop regulating the activation status of Arf1 as a result of changes in membrane lipids. The direct activation of PLD by Arf•GTP is shown, along with the product of PLD, PA, being converted to DAG, an activator of Arf GAP1.

The idea that Arf GAPs are a target of lipid signaling is consistent with in vivo studies in S. cerevisiae (Yanagisawa et al., 2002). Screens for secretory mutants in yeast had led to the identification of Sec14p, a PI transfer protein thought to regulate Golgi function by altering the lipid composition of Golgi membranes (Bankaitis et al., 1989). Deletion of Gcs1, an Arf GAP, was synthetically lethal with a temperature sensitive mutant of Sec14, sec14-1ts (Yanagisawa et al., 2002). Similarly, replacement of Gcs1 with Gcs1-K54, a mutant lacking GAP activity, was synthetically lethal with sec14-1ts. The defect imposed by loss of Sec14 function can be partially overcome by mutations (“bypass Sec14p” mutations) in genes that control phosphatidylcholine or phosphoinositide metabolism. Consistent with Gcs1 functioning downstream of lipid changes mediated by Sec14, deletion of Gcs1 re-imposed defects in growth and Golgi function in sec14-1ts strains with “bypass Sec14p” mutations. 4.3.2

GITs

Less is known about the potential regulatory role of lipids for the function of GITs. In one report, PIP3 stimulated the GAP activity of GIT, using Arf6 as a substrate (Vitale et al., 2000). However, at least 100 M PIP3 was required for a relatively modest (2 – 3 fold) stimulation (by comparison, 360 M PA with 45 M PI(4,5)P2 activate ASAP1 10,000 fold (Kam et al., 2000).

ARF AND PHOSPHOLIPIDS 4.3.3

11 59

ASAPs

Efforts to purify a phosphoinositide-dependent Arf GAP activity from bovine brain led to the first identification of ASAP family proteins (Brown et al., 1998). A core of PH, Arf GAP and ANK repeat domains defines these Arf GAPs. Domains outside the core define subfamilies within the ASAP family. ASAP1 and PAP (called AMAP1 and 2 in the nomenclature introduced in this volume) contain a unique amino terminus, proline rich domain and SH3 domain (Andreev et al., 1999; Brown et al., 1998). ACAPs have unique amino termini containing one or two coiled coil domains (Jackson et al., 2000). The AGAPs have a G-protein like domain (Nie et al., 2002). The ARAPs have five PH domains and a Rho GAP domain (Miura et al., 2002; Krugmann et al., 2002; personal observation). Two variants have sterile  motif (SAM) domains. ASAP1 has been the most extensively characterized of the ASAPs in terms of lipid dependence (Randazzo and Kahn, 1994). The GAP activity of ASAP1 purified from bovine brain or recombinant ASAP1 was synergistically and specifically activated by PA and PI(4,5)P2 (Kam et al., 2000; Brown et al., 1998; Randazzo, 1997a; Randazzo and Kahn, 1994). PA was more potent than phosphatidylserine or phosphatidylinositol. PI(4,5)P2 could not be replaced with PI(3,4,5)P3 or other D-3-phosphorylated phosphoinositides. A deletion mutant consisting of just the PH, Arf GAP and ANK repeat domains was the minimal structure in which GAP activity was detectable. The PI(4,5)P2 specificity was preserved in this protein. PI(4,5)P2 binding to both ASAP1 and Arf is required for GAP activity. As described above, Arf binding to PI(4,5)P2 facilitates its interaction with ASAP1. Stimulation of GAP activity also correlated with GAP binding to vesicles or micelles containing PI(4,5)P2 and PA. The introduction of mutations into the PH domain of ASAP1 abrogated both PI(4,5)P2 binding to ASAP1 and PI(4,5)P2 stimulation of GAP activity. The PH domain of ASAP1 functions as an allosteric binding site for PI(4,5)P2-dependent regulation of the GAP activity (Kam et al., 2000). This possibility was examined after noting that recruitment to membranes and stimulation of GAP activity could be dissociated. Probing the structure of ASAP1 by limited proteolysis led to the identification of a trypsin cleavage site within the Arf GAP domain that was exposed upon ASAP1 binding PI(4,5)P2, consistent with a PI(4,5)P2-dependent conformational change in the GAP. The possibility of a conformational change in Arf has not yet been tested. The phosphoinositide specificity within the ASAP family varies. The ACAPs are activated by either PI(4,5)P2 or PI(3,5)P2 in vitro, with equal efficacy (Z. Nie and P.A. Randazzo, unpublished observations). ARAPs are

12 60

RANDAZZO, NIE, AND HIRSCH

activated by PI(3,4)P2 and PI(3,4,5)P3 to a much greater extent than they are activated by other phosphoinositide isomers (Miura et al., 2002; Krugmann et al., 2002; personal observation). The contributions of each of the five PH domains to this activation have not been extensively examined. PH domain 1, counting from the amino terminus, likely confers at least some of the PI(3,4,5)P3 dependence. PH domain 2 is of particular interest, given the role of the PH domain of ASAP1 as an allosteric binding site for the PI(4,5)P2dependent regulation of GAP activity. The in vitro studies of the AGAPs have been the most complex to interpret (Nie et al., 2002). The PH domain is split, with an 80 amino acid insert between the predicted 4 and 5 strands of the structure. Within the first 4  strands, AGAP1 has 9 identities of the consensus sequence for PI(3,4,5)P3 binding (Lietzke et al., 2000). The one difference is conservative, lysine in place of an arginine. Consistent with this level of identity, PI(3,4,5)P3 stimulated this GAP. However, the required concentrations were relatively high (10 – 50 M) compared to that seen with ARAP1 (100 nM) and the addition of a second acid phospholipid, PA, inhibited the PI(3,4,5)P3-dependent activity. PI(4,5)P2, in the presence of PA, stimulated activity. This effect was specific for PA. Other acid phospholipids, including phosphatidylserine and PI, were much less effective in inhibiting PI(3,4,5)P3-dependent GAP activity and synergizing with PI(4,5)P2. The distinct phosphoinositide specificities of each Arf GAP and GEF could allow for the independent regulation of each. For instance, ASAP1, ACAP1 and ACAP2 have been found in the same cellular sites (Jackson et al., 2000). Changes in PI(3,5)P2 could contribute to the regulation of ACAP1 and ACAP2 whereas ASAP1 could be regulated by changes in PI(4,5)P2. Other mechanisms could provide additional regulation; e.g., ASAP1 is dependent on PA and is also a target of other signaling proteins, such as Src.

5.

SYSTEMS OF PHOSPHOINOSITIDE DEPENDENT REACTIONS

PI(4,5)P2 and PA could be elements of a system of regulatory feedforward and feedback loops. In addition to PI(4,5)P2 and PA affecting Arf, Arf affects the production of these lipids. PLD1 is activated synergistically by Arf, Rho family GTP-binding proteins, protein kinase C (PKC) and PI(4,5)P2 (see Chapter 11; Thomas et al., 1994; Liscovitch et al., 2000; Exton, 1999; Frohman et al., 1999). PIP kinase I is directly activated by Arf and PA (Anderson et al., 1999; Honda et al., 1999; Jones et al., 2000).

ARF AND PHOSPHOLIPIDS

13 61

The effect of Arf GAPs on Rho family proteins adds further complexity to this network of interactions. ARAP1 induces the activation of Cdc42 and the inactivation of RhoA in NIH 3T3 fibroblasts (Miura et al., 2002). ARAP1 has a Rho GAP domain and RhoA GAP activity in vitro and in vivo. Based on the formation of membrane ruffles and filopodia observed when the proteins are ectopically expressed, other ASAP family members also appear to be able to activate Cdc42 and/or Rac (P.A. Randazzo, unpublished observations). Furthermore, GIT family Arf GAPs bind the Rho family exchange factor called Pix/Cool (Zhao et al., 2000; Turner et al., 1999). Rho family proteins stimulate PLD and PIP kinase (Sternweis et al., 1997; Chong et al., 1994). Thus, the effects on the Rho family proteins could generate additional feedforward and feedback loops.

Figure 3. Proposed system of feedforward and feedback loops impinging on Arf and regulating phospholipid concentrations. Substrates and products of specific enzymes are indicated with a solid line and allosteric or regulatory effects with dotted lines. PI3-K, PI 3kinase, PI(4)P-5K, PI(4)P 5-kinase.

With the products of the PLD and PIP kinase I reactions synergizing with Arf and Rho to activate other enzymes, a complex network of feedforward and feedback loops can be assembled. The distinct phosphoinositide specificities of GAPs and GEFs could provide further coordination to these reactions. For instance, a PI(3,4,5)P3-dependent GEF could function with a

14 62

RANDAZZO, NIE, AND HIRSCH

PI(4,5)P2-dependent Arf GAP to drive the cycle of GTP binding and hydrolysis on Arf. A summary of these reactions is shown in Figure 3.

6.

FUNCTION OF ARF/LIPID REACTION NETWORKS IN CELL PHYSIOLOGY

Systems of reactions driving cyclical changes in Arf and phospholipids could function in several ways that need not be mutually exclusive. Here, three possible roles for the dynamic relationships between Arf and phospholipids are considered.

6.1

Rapid and precipitous switching

The simplest model for Arf action is that any particular effect (e.g., membrane traffic) is mediated by a single effector (e.g., a coat protein). Arf function in membrane traffic is dependent on the regulated cycle of GTP binding and hydrolysis. Not only is this cycle integral to Arf function, the cycle must occur on a time scale that matches the events being regulated. Therefore, one possible role of the feedforward and feedback loops is to mediate rapid and precipitous switching of Arf between on and off states. Lipid metabolites would be particularly suited for these functions as they could be generated and diffuse rapidly within the membrane at which Arf works.

6.2

PI(4,5)P2 and PA are Arf effectors

PI(4,5)P2 and PA could be Arf effectors, in which case the lipid loops could be a means of regulating the levels of either or both PI(4,5)P2 and PA. PI(4,5)P2 and PA regulate a number of events that impact membrane traffic and actin polymerization and waves of PI(4,5)P2 coincide with phagocytosis, an Arf6-dependent event (Takenawa and Itoh, 2001; Osborne et al., 2001; Gillooly et al., 2001; Simonsen et al., 2001; Payrastre et al., 2001; Cremona and De Camilli, 2001). Evidence supporting the function of PA and PI(4,5)P2 as Arf effectors has been reviewed elsewhere (Chapter 11 of this volume and Randazzo et al., 2000). Briefly, either blocking the production or sequestering the lipids has been found to block Arf-dependent cellular functions (Honda et al., 1999; Randazzo et al., 2000). Additionally, overexpressing PIP kinase mimics part of the phenotype seen with the constitutive activation of Arf6 (Honda et al., 1999; Brown et al., 2001). Mechanistic details of PA and PI(4,5)P2 function as Arf effectors are lacking. Two possibilities have been considered (Randazzo et al., 2000).

ARF AND PHOSPHOLIPIDS

15 63

First, the lipids may recruit specific proteins to membranes. For example, PA might recruit a coat protein or may activate a PIP kinase. PI(4,5)P2 may activate PLD, directly bind coat proteins or affect coat protein binding by influencing the association of actin cytoskeletal components with the membranes. In this “mesh hypothesis”, a membrane microdomain with a high content of PI(4,5)P2 recruits spectrin, an actin binding protein, which loosens the actin mesh, allowing coat binding and subsequent vesicle budding (Godi et al., 1999; Lorra and Huttner, 1999). A second possibility is that the lipid changes affect the physical properties of the membranes. The synthesis of PA may, by inducing curvature in the membrane, drive budding and release of a transport vesicle (Weigert et al., 2000; Weigert et al., 1999b; Weigert et al., 1999a; Schmidt et al., 1999).

6.3

Temporal regulation

The changes in lipids that would be driven by systems of reactions such as those schematized in Figure 3 could have a third effect. Arf, by both affecting lipid levels and having lipid-dependent protein targets, could order a series of reactions. As discussed above, PA and PI(4,5)P2 inhibit Arf binding to Arfaptin 2. PA promotes Arf binding to GGA; PA with PI(4,5)P2 promotes Arf interaction with Arf GAP. These effects create the potential for temporal regulation. For instance, with low PA and PI(4,5)P2, Arf could act through Arfaptin, perhaps mediating the recruitment of a Rho family protein to a membrane. As these lipids increase, in part driven by a feedforward loop activating Arf, which in turn activates both PIP kinase and PLD, Arf would dissociate from Arfaptin and recruit a coat protein. This event could result in the sorting of transmembrane “cargo” proteins (see Chapters 12-13). As PI(4,5)P2 and/or PA concentrations increase further, Arf GAPs could be activated, Arf•GTP hydrolyzed to Arf•GDP, and the sequence of events terminated. The sequence depicted in Figure 4 is just one of many possibilities by which Arfs may integrate lipid and protein signaling.

7.

CONCLUSION

The role of lipids in Arf function has been examined for over a decade and, rather than being simplified, more relationships have been discovered among the proteins and lipids in the Arf pathway. At the same time, the functions of phosphoinositides and phosphatidic acid have been further defined. This progress has led to an appreciation of the integral relationship between Arfs and phospholipids and a number of testable models of the mechanisms by

16 64

RANDAZZO, NIE, AND HIRSCH

which membrane traffic and the actin cytoskeleton are temporally and spatially regulated.

8.

REFERENCES

Aoe, T., Huber, I., Vasudevan, C., Watkins, S. C., Romero, G., Cassel, D., and Hsu, V. W. (1999). The KDEL receptor regulates a GTPase-activating protein for ADP-ribosylation factor1 by interacting with its non-catalytic domain. J. Biol. Chem., 274, 20545-20549. Amor, J. C., Harrison, D. H., Kahn, R. A., and Ringe, D. (1994). Structure of the human ADP-ribosylation factor-1 complexed with GDP. Nature, 372, 704-708. Anderson, R. A., Boronenkov, I. V., Doughman, S. D., Kunz, J., and Loijens, J. C. (1999). Phosphatidylinositol phosphate kinases, a multifaceted family of signaling enzymes. J. Biol. Chem., 274, 9907-9910. Andreev, J., Simon, J. P., Sabatini, D. D., Kam, J., Plowman, G., Randazzo, P. A., and Schlessinger, J. (1999). Identification of a new Pyk2 target protein with Arf-GAP activity. Mol. Cell. Biol., 19, 2338-2350. Antonny, B., Beraud-Dufour, S., Chardin, P., and Chabre, M. (1997a). N-terminal hydrophobic residues of the G-protein ADP- ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry, 36, 4675-4684. Antonny, B., Huber, I., Paris, S., Chabre, M., and Cassel, D. (1997b). Activation of ADPribosylation factor 1 GTPase-activating protein by phosphatidylcholine-derived diacylglycerols. J. Biol. Chem., 272, 30848-30851. Bagrodia, S., Bailey, D., Lenard, Z., Hart, M., Guan, J., Premont, R., Taylor, S., and Cerione, R. (1999). A tyrosine-phosphorylated protein that binds to an important regulatory region on the Cool family of p21-activated kinase-binding proteins. J. Biol. Chem., 274, 2239322400. Bankaitis, V. A., Malehorn, D. E., Emr, S. D., and Greene, R. (1989). The Saccharomycescerevisiae Sec14 gene encodes a cytosolic factor that is required for transport of secretory proteins from the yeast Golgi complex. J. Cell Biol., 108, 1271-1281. Beraud-Dufour, S., Paris, S., Chabre, M., and Antonny, B. (1999). Dual interaction of ADP ribosylation factor 1 with Sec7 domain and with lipid membranes during catalysis of guanine nucleotide exchange. J. Biol. Chem., 274, 37629-37636. Bourgoin, S. G., Houle, M. G., Singh, I. N., Harbour, D., Gagnon, S., Morris, A. J., and Brindley, D. N. (2002). ARNO but not Cytohesin-1 translocation is phosphatidylinositol 3-kinase-dependent in HL-60 cells. J. Leuk. Biol., 71, 718-728. Brown, F. D., Rozelle, A. L., Yin, H. L., Balla, T., and Donaldson, J. G. (2001). Phosphatidylinositol 4,5-bisphosphate and Arf6-regulated membrane traffic. J. Cell Biol., 154, 1007-1017. Brown, M. T., Andrade, J., Radhakrishna, H., Donaldson, J. G., Cooper, J. A., and Randazzo, P. A. (1998). ASAP1, a phospholipid-dependent Arf GTPase-activating protein that associates with and is phosphorylated by Src. Mol. Cell. Biol., 18, 7038-7051. Chardin, P., Paris, S., Antonny, B., Robineau, S., Beraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996). A human exchange factor for Arf contains Sec7- and pleckstrinhomology domains. Nature, 384, 481-484. Chong, L. D., Traynor-Kaplan, A., Bokoch, G. M., and Schwartz, M. A. (1994). The small GTP-binding protein-Rho regulates a phosphatidylinositol 4-phosphate 5-kinase in mammalian-cells. Cell, 79, 507-513.

ARF AND PHOSPHOLIPIDS

17 65

Cremona, O. and De Camilli, P. (2001). Phosphoinositides in membrane traffic at the synapse. J. Cell Sci., 114, 1041-1052. Cukierman, E., Huber, I., Rotman, M., and Cassel, D. (1995). The Arf1 GTPase-activating protein - zinc-finger motif and Golgi-complex localization. Science, 270, 1999-2002. Dell'Angelica, E. C., Puertollano, R., Mullins, C., Aguilar, R. C., Vargas, J. D., Hartnell, L. M., and Bonifacino, J. S. (2000). Ggas: a family of ADP ribosylation factor-binding proteins related to adaptors and associated with the Golgi complex. J. Cell Biol., 149, 8193. Donaldson, J. G., Cassel, D., Kahn, R. A., and Klausner, R. D. (1992). ADP-ribosylation factor, a small GTP-binding protein, is required for binding of the coatomer protein COP to Golgi membranes. Proc. Nat. Acad. Sci., USA, 89, 6408-6412. Exton, J. H. (1999). Regulation of phospholipase D. Biochim. Biophys. Acta, 1439, 121-133. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1993). Myristoylation is not required for GTP-dependent binding of ADP-ribosylation factor-Arf1 to phospholipids. J. Biol. Chem., 268, 24531-24534. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1995). Myristoylation of ADPribosylation factor-1 facilitates nucleotide exchange at physiological Mg2+ levels. J. Biol. Chem., 270, 1337-1341. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1996). Myristoylation-facilitated binding of the G protein ARF1(GDP) to membrane phospholipids is required for its activation by a soluble nucleotide exchange factor. J. Biol. Chem., 271, 1573-1578. Franco, M., Peters, P. J., Boretto, J., van Donselaar, E., Neri, A., D'Souza-Schorey, C., and Chavrier, P. (1999). EFA6, a Sec7 domain-containing exchange factor for ARF6, coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 14801491. Frohman, M. A., Sung, T. C., and Morris, A. J. (1999). Mammalian phospholipase D structure and regulation. Biochim. Biophys. Acta, 1439, 175-186. Gillooly, D. J., Simonsen, A., and Stenmark, H. (2001). Phosphoinositides and phagocytosis. J. Cell Biol., 155, 15-17. Godi, A., Santone, I., Pertile, P., Marra, P., Di Tullio, G., Luini, A., Corda, D., and De Matteis, M. A. (1999). ADP-ribosylation factor regulates spectrin skeleton assembly on the Golgi complex by stimulating phosphatidylinositol 4,5-bisphosphate synthesis. Biochem. Soc. Trans., 27, 638-642. Goldberg, J. (1998). Structural basis for activation of ARF GTPase: mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Goldberg, J. (1999). Structural and functional analysis of the ARF1-ARFGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Greasley, S. E., Jhoti, H., Teahan, C., Solari, R., Fensome, A., Thomas, G. M. H., Cockcroft, S., and Bax, B. (1995). The structure of rat ADP-ribosylation-factor-I (Arf-1) complexed to GDP determined from 2 different crystal forms. Nature Struct. Biol., 2, 797-806. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase alpha is a downstream effector of the small G protein ARF6 in membrane ruffle formation. Cell, 99, 521-532. Hong, J. X., Haun, R. S., Tsai, S. C., Moss, J., and Vaughan, M. (1994). Effect of ADPribosylation factor amino-terminal deletions on its GTP-dependent stimulation of choleratoxin activity. J. Biol. Chem., 269, 16519 -16524. Hong, J. X., Zhang, X. K., Moss, J., and Vaughan, M. (1995). Isolation of an amino-terminal deleted recombinant ADP-ribosylation factor-1 in an activated nucleotide-free state. Proc. Nat. Acad. Sci., USA, 92, 3056-3059.

18 66

RANDAZZO, NIE, AND HIRSCH

Jackson, C. L. and Casanova, J. E. (2000). Turning on ARF: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. Jackson, T. R., Brown, F. D., Nie, Z. Z., Miura, K., Foroni, L., Sun, J. L., Hsu, V. W., Donaldson, J. G., and Randazzo, P. A. (2000). ACAPs are Arf6 GTPase-activating proteins that function in the cell periphery. J. Cell Biol., 151, 627-638. Jacques, K. M, Nie, Z., Stauffer, S., Hirsch, D. S., Chen, L.-X., Stanley, K. T., and Randazzo, P. A. (2002). Arf1dissociates from the clathrin adaptor GGA prior to being inactivated by Arf GTPase-activating proteins. J. Biol. Chem., 277, 47235-47241. Jones, D. H., Bax, B., Fensome, A., and Cockcroft, S. (1999). ADP ribosylation factor 1 mutants identify a phospholipase D effector region and reveal that phospholipase D participates in lysosomal secretion but is not sufficient for recruitment of coatomer I. Biochem. J., 341, 185-192. Jones, D. H., Morris, J. B., Morgen, C. P., Kondo, H., Irvine, R. F., and Cockcroft, S. (2000). Type I phosphatidylinositol 4-phosphate 5-kinase directly interacts with ADP-ribosylation factor 1 and is responsible for phosphatidylinositol 4,5-bisphosphate synthesis in the Golgi compartment. J. Biol. Chem., 275, 13962-13966. Kahn, R. A. and Gilman, A. G. (1984). Purification of a protein cofactor required for ADPribosylation of the stimulatory regulatory component of adenylate cyclase by cholera toxin. J. Biol. Chem., 259, 6228-6234. Kahn, R. A. and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kam, J. L., Miura, K., Jackson, T. R., Gruschus, J., Roller, P., Stauffer, S., Clark, J., Aneja, R., and Randazzo, P. A. (2000). Phosphoinositide-dependent activation of the ADPribosylation factor GTPase-activating protein ASAP1 - evidence for the pleckstrin homology domain functioning as an allosteric site. J. Biol. Chem., 275, 9653-9663. Klarlund, J. K., Guilherme, A., Holik, J. J., Virbasius, J. V., Chawla, A., and Czech, M. P. (1997). Signaling by phosphoinositide-3,4,5-trisphosphate through proteins containing pleckstrin and Sec7 homology domains. Science, 275, 1927-1930. Klarlund, J. K., Rameh, L. E., Cantley, L. C., Buxton, J. M., Holik, J. J., Sakelis, C., Patki, V., Corvera, S., and Czech, M. P. (1998). Regulation of GRP1-catalyzed ADP ribosylation factor guanine nucleotide exchange by phosphatidylinositol 3,4,5-trisphosphate. J. Biol. Chem., 273, 1859-1862. Klarlund, J. K., Tsiaras, W., Holik, J. J., Chawla, A., and Czech, M. P. (2000). Distinct polyphosphoinositide binding selectivities for pleckstrin homology domains of GRP1-like proteins based on diglycine versus triglycine motifs. J. Biol. Chem., 275, 32816-32821. Krugmann, S., Anderson, K. E., Ridley, S. H., Risso, N., McGregor, A., Coadwell, J., Davidson, K., Eguinoa, A., Ellson, C. D., Lipp, P., Manifava, M., Ktistakis, N., Painter, G., Thuring, J. W., Cooper, M. A., Lim, Z. Y., Holmes, A. B., Dove, S. K., Michell, R. H., Grewal, A., Nazarian, A., Erdjument-Bromage, H., Tempst, P., Stephens, L. R., and Hawkins, P. T. (2002). Identification of ARAP3, a novel PI3K effector regulating both Arf and Rho GTPases, by selective capture on phosphoinositide affinity matrices. Mol. Cell, 9, 95-108. Lietzke, S. E., Bose, S., Cronin, T., Klarlund, J., Chawla, A., Czech, M. P., and Lambright, D. G. (2000). Structural basis of 3-phosphoinositide recognition by pleckstrin homology domains. Mol. Cell, 6, 385-394. Liscovitch, M., Czarny, M., Fiucci, G., and Tang, X. Q. (2000). Phospholipase D: molecular and cell biology of a novel gene family. Biochem. J., 345, 401-415. Lorra, C. and Huttner, W. B. (1999). The mesh hypothesis of Golgi dynamics. Nature Cell Biol., 1, E113-E115.

ARF AND PHOSPHOLIPIDS

19 67

Makler, V., Cukierman, E., Rotman, M., Admon, A., and Cassel, D. (1995). ADP-ribosylation factor-directed GTPase-activating protein - purification and partial characterization. J. Biol. Chem., 270, 5232-5237. Miura, K., Jacques, K. M., Stauffer, S., Kubosaki, A., Zhu, K. J., Hirsch, D. S., Resau, J., Zheng, Y., and Randazzo, P. A. (2002). ARAP1: A point of convergence for Arf and Rho signaling. Mol. Cell, 9, 109-119. Moss, J. and Vaughan, M. (2002). Cytohesin-1 in 2001. Arch.Biochem. Biophys., 397, 156161. Nie, Z., Stanley, K. T., Stauffer, S., Jacques, K. M., Hirsch, D. S., Takei, J., and Randazzo, P. A. (2002). AGAP1: an endosome-associated, phosphoinositide-dependent ADPribosylation factor GTPasae activating protein that affects actin cytoskeleton. J. Biol. Chem., 277, 48965-48975. Ogasawara, M., Kim, S. C., Adamik, R., Togawa, A., Ferrans, V. J., Takeda, K., Kirby, M., Moss, J., and Vaughan, M. (2000). Similarities in function and gene structure of Cytohesin-4 and Cytohesin-1, guanine nucleotide-exchange proteins for ADP- ribosylation factors. J. Biol. Chem., 275, 3221-3230. Ooi, C. E., Dell'Angelica, E. C., and Bonifacino, J. S. (1998). ADP-ribosylation factor 1 (ARF1) regulates recruitment of the AP-3 adaptor complex to membranes. J. Cell Biol., 142, 391-402. Orci, L., Palmer, D. J., Amherdt, M., and Rothman, J. E. (1993). Coated vesicle assembly in the Golgi requires only coatomer and Arf proteins from the Cytosol. Nature, 364, 732-734. Osborne, S. L., Meunier, F. A., and Schiavo, G. (2001). Phosphoinositides as key regulators of synaptic function. Neuron, 32, 9-12. Pacheco-Rodriguez, G., Meacci, E., Vitale, N., Moss, J., and Vaughan, M. (1998). Guanine nucleotide exchange on ADP-ribosylation factors catalyzed by Cytohesin-1 and its Sec7 domain. J. Biol. Chem., 273, 26543-26548. Palmer, D. J., Helms, J. B., Beckers, C. J. M., Orci, L., and Rothman, J. E. (1993). Binding of coatomer to Golgi membranes requires ADP-ribosylation factor. J. Biol. Chem., 268, 12083-12089. Paris, S., Antonny, B., Beraud-Dufour, S., Robineau, S., Chabre, M., and Chardin, P. (1997a). Arf1: interactions with phospholipids and activation by the human exchange factor ARNO. FASEB J., 11, 3425. Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997b). Role of protein-phospholipid interactions in the activation of ARF1 by the guanine nucleotide exchange factor ARNO. J. Biol. Chem., 272, 22221-22226. Payrastre, B., Missy, K., Giuriato, S., Bodin, S., Plantavid, M., and Gratacap, M. P. (2001). Phosphoinositides - key players in cell signaling, in time and space. Cell. Signal., 13, 377387. Premont, R., Claing, A., Vitale, N., Freeman, J., Pitcher, J., Patton, W., Moss, J., Vaughan, M., and Lefkowitz, R. (1998). β2-Adrenergic Receptor Regulation By GIT1, A G ProteinCoupled Receptor Kinase-Associated ADP-Ribosylation Factor GTPase-Activating Protein. Proc. Nat. Acad. Sci., USA, 95, 14082-14087. Premont, R. T., Claing, A., Vitale, N., Perry, S. J., and Lefkowitz, R. J. (2000). The GIT family of ADP-ribosylation factor GTPase-activating proteins - functional diversity of GIT2 through alternative splicing. J. Biol. Chem., 275, 22373-22380. Randazzo, P. A. (1997a). Functional interaction of ADP-ribosylation factor 1 with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 272, 7688-7692. Randazzo, P. A. (1997b) Resolution of two ADP-ribosylation factor 1 GTPase-activating proteins from rat liver. Biochem. J., 324, 413-419.

20 68

RANDAZZO, NIE, AND HIRSCH

Randazzo, P. A. and Kahn, R. A. (1994). GTP hydrolysis by ADP-ribosylation factor is dependent on both an ADP-ribosylation factor GTPase-activating protein and acid phospholipids. J. Biol. Chem., 269, 10758-10763. Randazzo, P. A. and Kahn, R. A. (1995). Myristoylation and ADP-ribosylation factor function. Meth. Enzymol., 250, 394-405. Randazzo, P. A., Nie, Z., Miura, K., and Hsu, V. W. (2000). Molecular aspects of the cellular activities of ADP-ribosylation factors. Science STKE., 2000, RE1. Randazzo, P. A., Terui, T., Sturch, S., Fales, H. M., Ferrige, A. G., and Kahn, R. A. (1995). The myristoylated amino-terminus of ADP-ribosylation factor-1 is a phospholipidsensitive and GTP-sensitive switch. J. Biol. Chem., 270, 14809-14815. Randazzo, P. A., Terui, T., Sturch, S., and Kahn, R. A. (1994). The amino-terminus of ADPribosylation factor (Arf)-1 is essential for interaction with Gs and Arf GTPase-activating protein. J. Biol. Chem., 269, 29490-29494. Randazzo, P. A., Yang, Y. C., Rulka, C., and Kahn, R. A. (1993). Activation of ADPribosylation factor by Golgi membranes - evidence for a Brefeldin A-sensitive and protease-sensitive activating factor on Golgi membranes. J. Biol. Chem., 268, 9555-9563. Rothman, J. E. (1994). Mechanism of intracellular protein-transport. Nature, 372, 55-63. Schmidt, A., Wolde, M., Thiele, C., Fest, W., Kratzin, H., Podtelejnikov, A. V., Witke, W., Huttner, W. B., and Soling, H. D. (1999). Endophilin I mediates synaptic vesicle formation by transfer of arachidonate to lysophosphatidic acid. Nature, 401, 133-141. Serafini, T., Orci, L., Amherdt, M., Brunner, M., Kahn, R. A., and Rothman, J. E. (1991). ADP-ribosylation factor is a subunit of the coat of Golgi-derived COP-coated vesicles - a novel role for a GTP-binding protein. Cell, 67, 239-253. Simonsen, A., Wurmser, A. E., Emr, S. D., and Stenmark, H. (2001). The role of phosphoinositides in membrane transport. Curr. Opin. Cell Biol., 13, 485-492. Stamnes, M. A. and Rothman, J. E. (1993). The binding of AP-1 Clathrin adapter particles to Golgi membranes requires ADP-ribosylation factor, a small GTP-binding protein. Cell, 73, 999-1005. Sternweis, P. C., Brown, H. A., Singer, W. D., Jiang, X., Faulkner, A., Wells, C., Roth, M. G., Ktistakis, N. T., and Bi, K. (1997). Regulation of phospholipase D and its role in membrane traffic. FASEB J., 11, 100. Takenawa, T. and Itoh, T. (2001). Phosphoinositides, key molecules for regulation of actin cytoskeletal organization and membrane traffic from the plasma membrane. Biochim. Biophys. Acta, 1533, 190-206. Terui, T., Kahn, R. A., and Randazzo, P. A. (1994). Effects of acid phospholipids on nucleotide exchange properties of ADP-ribosylation factor-1 - evidence for specific interaction with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 269, 28130-28135. Thomas, G., Fensome, A., Whatmore, J., Geny, B., and Cockcroft, S. (1994). The regulation of phospholipase-D by Arfs. J. Neurochem., 63, S29. Traub, L. M., Ostrom, J. A., and Kornfeld, S. (1993). Biochemical dissection of AP-1 recruitment onto Golgi membranes. J. Cell Biol., 123, 561-573. Turner, C. E., Brown, M. C., Perrotta, J. A., Riedy, M. C., Nikolopoulos, S. N., McDonald, A. R., Bagrodia, S., Thomas, S., and Leventhal, P. S. (1999). Paxillin LD4 motif binds PAK and PIX through a novel 95-Kd Ankyrin repeat, ARF-GAP protein: a role in cytoskeletal remodeling. J. Cell Biol., 145, 851-863. Venkateswarlu, K., Gunn-Moore, F., Tavare, J. M., and Cullen, P. J. (1999). EGF- and NGFstimulated translocation of Cytohesin-1 to the plasma membrane of PC12 cells requires PI 3-kinase activation and a functional cytohesin-1 PH domain. J. Cell Sci., 112, 1957-1965.

ARF AND PHOSPHOLIPIDS

21 69

Venkateswarlu, K., Oatey, P. B., Tavare, J. M., and Cullen, P. J. (1998). Insulin-dependent translocation of ARNO to the plasma membrane of adipocytes requires phosphatidylinositol 3-kinase. Curr. Biol., 8, 463-466. Vitale, N., Patton, W. A., Moss, J., Vaughan, M., Lefkowitz, R. J., and Premont, R. T. (2000). GIT proteins, a novel family of phosphatidylinositol 3,4,5- trisphosphate-stimulated GTPase-activating proteins for ARF6. J. Biol. Chem., 275, 13901-13906. Weigert, R., Silletta, M. G., Spano, S., Turacchio, G., Cericola, C., Colanzi, A., Senatore, S., Mancini, R., Polishchuk, E. V., Salmona, M., Facchiano, F., Burger, K. N. J., Mironov, A., Luini, A., and Corda, D. (1999a). Ctbp/BARS induces fission of Golgi membranes by acylating lysophosphatidic acid. Nature, 402, 429-433. Weigert, R., Silletta, M. G., Spano, S., Turacchio, G., Cericola, C., Luini, A., and Corda, D. (2000). Fission of Golgi membranes is induced by Ctbp/BARS-induced acylation of lysophosphatidic acid. Biophys. J., 78, 1878. Weigert, R., Silletta, M. G., Turacchio, G., Spano, S., Burger, K. N. J., Corda, D., and Luini, A. (1999b). BARS is an acyl-transferase involved in Golgi membrane fission. Mol. Biol. Cell, 10, 25. Yanagisawa, L., Marchena, J., Xie, Z., Poon, P., Singer, R., Johnston, G., Randazzo, P., and Bankaitis, V. (2002). Activity PF specific lipid-regulated Arf GAPs is required for Sec14p-dependent Golgi secretory function in yeast. Mol. Biol. Cell., 13, 2193-2206 Zhao, Z. S., Manser, E., Loo, T. H., and Lim, L. (2000). Coupling of PAK-interacting exchange factor PIX to GIT1 promotes focal complex disassembly. Mol. Cell. Biol., 20, 6354-6363. Zhu, Y. X., Drake, M. T., and Kornfeld, S. (1999). ADP-ribosylation factor 1 dependent clathrin-coat assembly on synthetic liposomes. Proc. Nat. Acad. Sci., USA, 96, 5013-5018.

Chapter 4 THE SEC7 FAMILY OF ARF GUANINE NUCLEOTIDE EXCHANGE FACTORS CATHERINE L. JACKSON Cell Biology and Metabolism Branch, NICHD, NIH, Bethesda, USA

Abstract:

1.

Arf GEFs are all multidomain proteins that likely integrate upstream signals with the downstream effects of GTPase activation. The presence of a highly conserved domain, the Sec7 domain, allows prediction of the number of Arf GEFs in multiple organisms and phylogenetic analyses. The profound effects of brefeldin A on membrane systems in eukaryotic cells demonstrate that its targets, the Arf GEFs, are central regulators of organelle structure and function. Ultrastructural analyses, particularly those coupled to genetic studies in S. cerevisiae, offer new insights into the roles of Arf GEFs in membrane dynamics and protein transport.

INTRODUCTION

Conversion of Arf•GDP to Arf•GTP is catalyzed by guanine nucleotide exchange factors (GEFs). Arf•GDP is largely soluble, with only a weak affinity for membranes, whereas Arf•GTP is more tightly membrane bound (Antonny et al., 1997). Activation of Arf results not only in a conformational change in the protein that allows it to interact with numerous effectors, but also in a change in localization from the cytoplasm to a membrane. It is the localization of the Arf GEFs that determine where in the cell Arf will be activated. Hence, the GEFs control not only the timing of activation of Arf GTPases, but also the spatial regulation of activation.

71

2 72

2.

JACKSON

THE SEC7 DOMAIN CATALYZES GUANINE NUCLEOTIDE EXCHANGE ON ARFS

The Arf GEFs began to be identified at the molecular level in 1996 and the presence of a conserved domain, the Sec7 domain, as that responsible for GEF activity was soon evident. A genetic approach in budding yeast, Saccharomyces cerevisiae, identified the first Arf GEF (Peyroche et al., 1996). In yeast, ARF1 and ARF2 encode functionally indistinguishable Arf proteins, which in sequence and function are most similar to mammalian Class I Arfs (see Chapter 1). ARF1 produces 90% of the total Arf protein in cells, and ARF2 only 10% (Stearns et al., 1990). In a strain deleted for ARF1 and expressing wild type Arf2p from the ARF2 gene alone, ARF2-N31 was expressed from a low-copy plasmid. This dominant negative phenotype was conditional: at 37oC, cells expressing the ARF2-N31 allele were viable but at 23oC, cells expressing the mutant protein failed to grow. A multicopy suppressor of the cold-sensitive growth defect was found and named GEA1 (for Guanine nucleotide Exchange on Arf). There is a second gene encoding a protein 50% identical to Gea1p, named GEA2. Gea1p and Gea2p are functionally redundant. Strains carrying a deletion of one or the other individually have no growth or secretion phenotype, but the double deletion strain is inviable. To date, no clear differences in function have been reported. At the time they were identified, the only sequence feature evident in the Gea1p and Gea2p proteins was the presence of a so-called “Sec7 domain”. This domain was homologous to a region found in the Sec7p protein. SEC7 was first identified in the screen for secretion mutants carried out in S. cerevisiae (Novick et al., 1980). Other proteins such as Arabidopsis thaliana GNOM/Emb30 and mammalian B-2/cytohesin had also been found to have a Sec7 domain (Sec7d), although its function was not clear from analysis of these proteins (Liu and Pohajdak, 1992; Shevell et al., 1994; Zagulski et al., 1995). In collaboration with P. Chardin, we searched for human proteins containing any sequence similarities to the Gea1p and Gea2p proteins, and found an EST clone with a Sec7d. The protein was expressed in E. coli and found to have very potent Arf GEF activity (Chardin et al., 1996). The protein was named ARNO for ADPRibosylation factor Nucleotide binding site Opener. The only sequence similarity between ARNO and the Gea proteins was in the Sec7d. This observation led to the prediction that the Sec7d was responsible for Arf GEF activity, and indeed the ARNO Sec7d alone (purified from E. coli) had Arf GEF activity (Chardin et al., 1996). Hence, the Sec7d is the catalytic domain of the Arf GEFs. Two other groups were taking different approaches to identify an Arf GEF at about the same time. A classic biochemical approach aimed at

THE SEC7 FAMILY OF ARF GEFS

3 73

identifying a protein responsible for in vitro Arf GEF activity culminated in the cloning of the bovine BIG1 cDNA, encoding a mammalian ortholog of yeast Sec7p (Morinaga et al., 1997; Morinaga et al., 1996). Shortly afterwards, the human BIG1 and BIG2 proteins were cloned and all three of these proteins were shown to have Arf GEF activity in vitro (Togawa et al., 1999) (see Chapter 6). Also, a genetic approach was used to identify the target of the drug brefeldin A (BFA; see below) in Chinese hamster ovary cells. This led to the identification of GBF1, another Sec7d-containing protein with demonstrated Arf GEF activity in vitro (Claude et al., 1999) (see Chapter 5).

3.

EVOLUTIONARY CONSERVATION OF THE ARF GEFS

The complete genome sequences for humans, D. melanogaster, C. elegans, S. pombe, S. cerevisiae and A. thaliana are now available. With the identification of the Sec7d as that responsible for Arf GEF activity, it is now possible to scan the predicted proteins from these organisms to determine all predicted Arf GEF proteins. S. cerevisiae, S. pombe, C. elegans and D. melanogaster each have 5 Sec7d proteins, A. thaliana has 7-8, and humans have ~15. Each of the Arf GEFs in C. elegans and D. melanogaster belong to a different subfamily (Figure 1). Humans have 3-4 members of four of these five subfamilies, with GBF1 being the sole human member of the fifth subfamily (Figure 2). Mice and humans have one additional putative Arf GEF, the Fbx8 protein, which is not found even in Drosophila. In addition to a Sec7d, Fbx8 contains an N-terminal F-box, normally found in E3 ubiquitin ligase complexes that target proteins for degradation via the ubiquitination pathway. The function of this protein is not known. Two subfamilies of Arf GEFs, Gea/GBF/GNOM and Sec7/BIG, are common to all five organisms. Arabidopsis has no other subfamilies (Figure 1); it has three members of the Gea/GBF/GNOM family, and five members of the Sec7/BIG family. One of the latter (BIG5-At) is quite divergent in sequence in the Sec7d, and documentation of Arf GEF activity is required to ensure its appropriate placement in the family. A member of the Sec7/BIG family was identified in Paramecium tetraurelia in a screen for ciliary proteins, and it may play a role in transport of components to cilia in this organism (Nair et al., 1999). The two yeasts S. cerevisiae and S. pombe have, in addition to the Gea/GBF/GNOM and Sec7/BIG subfamilies, two more divergent Arf GEFs. S. cerevisiae Syt1p and S. pombe ArfGEF4 share a significant level of sequence homology, as do S. cerevisiae Ybl060p and S. pombe ArfGEF5 (Figure 1). The S. cerevisiae Syt1p Sec7d, although quite

4 74

JACKSON

Figure 1. Phylogenetic tree of Arf GEFs from different species. Included are Sec7 domain proteins from five fully sequenced eukaryotic genomes, and published Sec7 domain sequences from other organisms. Names of subfamilies are given on the right. Arf GEFs from closely related organisms to those shown here are not included. See Table 1 for descriptions of each Arf GEF sequence. Full-length Sec7 domain sequences were used for the analysis, and the tree was generated using the EMBL European Bioinformatics Institute ClustalW program at http://www.ebi.ac.uk/clustalw/.

THE SEC7 FAMILY OF ARF GEFS

5 75

divergent in sequence from those of the Gea/GBF/GNOM and Sec7/BIG subfamilies, has Arf GEF activity in vitro (Jones et al., 1999). The ARNO/GRP/cytohesin subfamily is a branch of the Sec7/BIG family restricted to metazoans. In humans, there are four members of this subfamily: cytohesin-1/PSCD1, ARNO/PSCD2, GRP1/ARNO3/PSCD3 and cytohesin-4/PSCD4 and close homologues are found in mice and rats. Humans, D. melanogaster and C. elegans have two additional subfamilies of Arf GEFs. These are the EFA6 subfamily, which includes EFA6A, EFA6B/TIC, EFA6C and EFA6D (Derrien et al., 2002) (see Chapter 14), and the ArfGEP100/BRAG subfamily, including BRAG1, ArfGEP100/BRAG2 (Someya et al., 2001), and BRAG3 (see Chapter 5). S. pombe ArfGEF4 and ArfGEF5 each have a PH domain downstream of their Sec7d which shows weak homology to those of human and D. melanogaster EFA6 proteins. Hence, these S. pombe Arf GEFs (and possibly their S. cerevisiae counterparts Syt1p and Ybl060p) may be distantly related to the EFA6 proteins of higher eukaryotes (Figure 1). Included in the tree shown in Figure 1 are two Arf GEFs from human pathogens that have been demonstrated to have in vitro exchange activity towards class I Arfs. The RalF protein of Legionella pneumophila is necessary for the localization of Arf on phagosomes containing the bacterium, and plays an important role in the infection process (Nagai et al., 2002). The malarial parasite Plasmodium falciparum has a single Sec7d protein, Arf GEF-Pf, that has in vitro Arf GEF activity sensitive to BFA (Baumgartner et al., 2001).

4.

SUBSTRATE SPECIFICITY OF ARF GEFS

A major issue in the Arf GEF field is to determine which Arf substrate(s) are used by a given GEF in the cell. Arf proteins are divided into three classes: I (Arf1, 2, 3), II (Arf4, 5), and III (Arf6), based on sequence homology and function (see Chapter 1). The primary means of assessing substrate specificity for GEFs is in vitro nucleotide exchange assays. The Gea/GBF/GNOM and Sec7/BIG Arf GEFs act on class I Arf proteins, as determined by in vitro exchange assays (Jackson and Casanova, 2000; Moss and Vaughan, 1998). The full length Gea1p and GNOM proteins use mammalian class I Arfs as substrates very efficiently (Peyroche et al., 1996; Steinmann et al., 1999). The in vitro activity of yeast Gea1p on human Arf1 is equivalent to its activity on yeast Arf1p or Arf2p (A. Peyroche and C. L. Jackson, unpublished observations). However, GBF1 has not yet been reported to have GEF activity on class I Arfs in vitro, although class I Arfs

6 76

JACKSON

Table 1. Sequence information for the Arf GEFs shown in Figures 1 and 2. Name Organism Accession number GBF1-Hs Homo sapiens NM_004193 GBF1-Dm Drosophila melanogaster AE003822 GBF1-Ce Caenorhabditis elegans AL032627 Gea1p-Sc Saccharomyces cerevisiae P47102 Gea2p-Sc Saccharomyces cerevisiae P39993 Gea1p-Sp Schizosaccharomyces pombe NP_596613 GNOM-At Arabidopsis thaliana S65571 GNL1-At Arabidopsis thaliana NP_198766 GNL2-At Arabidopsis thaliana NP_197462 BIG1-Hs Homo sapiens Q9Y6D6 BIG2-Hs Homo sapiens Q9Y6D5 BIG1-Dm Drosophila melanogaster AAF53331 BIG1-Ce Caenorhabditis elegans NP_493386 Sec7p-Sc Saccharomyces cerevisiae AAB04031 Sec7p-Pp Pischia pastoris AAK40234 BIG1-Sp Schizosaccharomyces pombe NP_594555 BIG2-Sp Schizosaccharomyces pombe NP_594954 BIG1-At Arabidopsis thaliana NP_195533 BIG2-At Arabidopsis thaliana NP_191645 BIG3-At Arabidopsis thaliana NP_171698 BIG4-At Arabidopsis thaliana NP_195264 BIG5-At Arabidopsis thaliana NP_189916 Sec7-Pt Paramecium tetraurelia AAF36486 Cytohesin1-Hs Homo sapiens Q15438 ARNO-Hs Homo sapiens CAA68084 GRP1/ARNO3-Hs Homo sapiens NP_004218 Cytohesin4-Hs Homo sapiens NP_037517 ARNO-Dm Drosophila melanogaster AAF57230 ARNO-Ce Caenorhabditis elegans NP_498764 EFA6A/PSD-Hs Homo sapiens NP_002770 EFA6B/TIC-Hs Homo sapiens NP_036587 EFA6C-Hs Homo sapiens CAB66494 EFA6D/KIAA0942-Hs Homo sapiens NP_056125 EFA6-Dm Drosophila melanogaster AAF56027 EFA6-Ce Caenorhabditis elegans NP_502416 BRAG1/KIAA0522-Hs Homo sapiens T00080 ArfGEP100/BRAG2-Hs Homo sapiens NP_055684 BRAG3/KIAA 1110-Hs Homo sapiens BAA83062 BRAG-Dm Drosophila melanogaster AAF51675 BRAG-Ce Caenorhabditis elegans NP_500420 Syt1-Sc Saccharomyces cerevisiae NP_015420 ArfGEF4-Sp Schizosaccharomyces pombe NP_594894 ArfGEF5/Yb1060p-Sc Saccharomyces cerevisiae NP_009493 ArfGEF5-Sp Schizosaccharomyces pombe NP_594936 Ra1F-Lp Legionella pneumophila AAL23711 Arf GEF-Pf Plasmodium falciparum Not available

Length 1859 2040 1820 1408 1459 1462 1451 1443 1375 1849 1785 1643 1628 2009 1772 1822 1811 1698 1793 1750 1711 1669 1135 398 399 399 394 348 393 645 1056 771 534 900 818 1560 841 740 1210 614 1226 657 687 942 374 3384

THE SEC7 FAMILY OF ARF GEFS

7 77

are likely to be a substrate for this human GEF in cells (Claude et al., 1999; Kawamoto et al., 2002). Activity of GBF1 on the class II Arf, Arf5p has been shown in vitro (Claude et al., 1999). The ARNO/GRP/cytohesin subfamily of Arf GEFs can act on both class I and class III Arfs in vitro (Franco et al., 1998; Frank et al., 1998; Macia et al., 2001). An apparent exception to this rule is cytohesin-4, which has been reported to have no activity towards Arf6, and which acts preferentially on class I Arfs in vitro (Ogasawara et al., 2000). A promising newer approach to the question of substrate specificity uses pull-downs of Arf•GTP from cell extracts by the Arf effector GGA1, which binds exclusively to the GTP-bound form of all Arfs (classes I, II and III) (Boman et al., 2000; Dell'Angelica et al., 2000). This approach was used to demonstrate that over-expressed ARNO specifically activates Arf6 to modulate migratory potential and morphology in Madin-Darby canine kidney (MDCK) cells (Santy and Casanova, 2001). Nakayama and colleagues used this pull-down approach to demonstrate that BIG2 activates class I Arfs in cells (Shinotsuka et al., 2002). In the case of the EFA6 family, biochemical and cell biological approaches converge, showing in both cases that the preferred substrate is Arf6 (Franco et al., 1999; Macia et al., 2001). ArfGEP100/BRAG2 has highest in vitro activity towards Arf6, accelerating binding of GTPγS by 12-fold (Someya et al., 2001). Immunofluorescence analyses indicate that ArfGEP100/BRAG2 and Arf6 co-localize in cells in peripheral areas, strongly supporting the conclusion that Arf6 is the physiological substrate of this GEF (Someya et al., 2001).

5.

BFA TARGETS A SUBFAMILY OF ARF GEFS IN COMPLEX WITH ARF•GDP

5.1

Effects of BFA on cells

BFA has a profound effect on organelles of the secretory and endocytic pathways in a wide range of cell types (Jackson, 2000; Klausner et al., 1992). In the mid-1980’s, BFA was shown to specifically inhibit secretion of proteins in mammalian cells, with little or no immediate effect on other cellular processes (Misumi et al., 1986; Takatsuki and Tamura, 1985). Subsequent studies revealed that BFA caused the disassembly of the Golgi apparatus, and the redistribution of certain Golgi enzymes to the ER (Doms et al., 1989; Fujiwara et al., 1988; Lippincott-Schwartz et al., 1989). This

8 78

JACKSON

Figure 2 . Phylogenetic trees of the Arf GEFs from A) Homo sapiens B) Drosophila melanogaster and C) Saccharomyces cerevisiae. See Table 1 for sequence information.

effect was rapid, with complete breakdown of this organelle occurring within 10 minutes of treatment with high concentrations of BFA (LippincottSchwartz et al., 1989). These experiments were the first to show the incredibly dynamic nature of the Golgi apparatus, whose elaborate structure

THE SEC7 FAMILY OF ARF GEFS

9 79

would seem to suggest a more stable state. BFA rapidly blocks transport of secreted proteins out of the ER-Golgi (Klausner et al., 1992) and trans-Golgi network (TGN)-endosomal systems, causing fusion of the TGN and endosome (Hunziker et al., 1992; Lippincott-Schwartz et al., 1991; Reaves and Banting, 1992; Wood et al., 1991). BFA treatment of particular cell types also causes tubulation of lysosomes (Lippincott-Schwartz et al., 1991; Wood and Brown, 1992). Two cell lines, MDCK and PtK1, are resistant to the effects of BFA at the Golgi, but sensitive to the drug at the TGN/endosome (Hunziker et al., 1992). This result suggests the existence of two distinct BFA targets acting at the level of each organelle. The most rapid effect of BFA is the release of specific peripherally associated proteins from organelles, notably the Golgi and endosomes (Boehm et al., 2001; De Matteis and Morrow, 2000; Jackson, 2000). The first such protein identified was β-COP, a subunit of the COPI coat complex, which is released from Golgi membranes within 1 minute of BFA treatment (Donaldson et al., 1990; Presley et al., 2002). Class I Arfs are also rapidly released from membranes upon BFA treatment (Donaldson et al., 1991; Presley et al., 2002). GTPγS was found to antagonize the effects of BFA on the binding of COPI to membranes, suggesting that Arfs might regulate this rapidly reversible process. This prediction was elegantly verified by the demonstration that binding of Arf1 to Golgi membranes was a prerequisite for binding of β-COP, and that only the former step was inhibited by BFA (Donaldson et al., 1992). Moreover, several groups demonstrated that Golgi membranes were capable of stimulating guanine nucleotide exchange on Arf, and that BFA inhibited this GEF activity (Donaldson et al., 1992; Helms and Rothman, 1992; Randazzo et al., 1993). These results demonstrated that at least one target of BFA was an Arf GEF or an associated Golgi protein. Cells expressing the dominant negative mutant Arf1-N31 had a phenotype indistinguishable from that of BFA-treated cells (Dascher and Balch, 1994), thus providing further support for the role of Arf and Arf GEFs in BFA action and COPI binding to Golgi membranes. Arf controls binding of other coat complexes at the TGN/endosome; including AP-1/clathrin, AP-3, AP-4 and the GGAs (see Chapters 12 and 13; Boehm and Bonifacino, 2001; Drake et al., 2000; Liang and Kornfeld, 1997; Ooi et al., 1998; Robinson and Bonifacino, 2001; Zhu et al., 1998).

5.2

Target of BFA: an Arf•GDP – Arf GEF complex

5.2.1

BFA targets in live cells

The cloning of the Arf GEFs provided an opportunity to test the hypothesis that these proteins are direct targets of BFA. Surprisingly, although some

10 80

JACKSON

Arf GEFs have in vitro exchange activity sensitive to BFA (Morinaga et al., 1997; Peyroche et al., 1996), others such as ARNO, have in vitro GEF activity that is completely resistant to the drug (Chardin et al., 1996). To determine whether the Sec7d itself is responsible for specifying BFA resistance/sensitivity, chimeric proteins were constructed. The Sec7ds of the BFA-sensitive Gea1p and Sec7p GEFs were replaced with that of BFAresistant ARNO, and their sensitivity to BFA was determined in vivo. The chimeras (Gea-ARNO-Gea1p and Sec7-ARNO-Sec7p) were expressed in cells with chromosomal deletions of the wild type genes, and each functioned like the endogenous proteins in terms of both growth and secretion (Peyroche et al., 1999). Even a gea1-∆ gea2-∆ sec7-∆ strain expressing Gea-ARNO-Gea1p and Sec7-ARNO-Sec7p as the sole copies of Gea1/2p and Sec7p, respectively, grew normally. However, this strain was now almost completely resistant to the effects of BFA. Both growth and secretion in the presence of BFA were comparable to those of the untreated strain (Peyroche et al., 1999). This result demonstrates that Sec7, Gea1p and Gea2p are the major targets of BFA in the secretory pathway of yeast, that the Sec7d itself is responsible for BFA sensitivity or resistance, and that the cellular toxicity of BFA is directly attributable to its actions on these proteins in yeast. 5.2.2

Residues in the Sec7 domain determine sensitivity to BFA

To determine residues important for BFA sensitivity, a portion of the GEA1 gene was mutagenized and BFA-resistant mutants were selected. All mutants obtained had at least one amino acid substitution in a 35-amino-acid region of the Gea1p Sec7d. Remarkably, this region also contained a high frequency of residues that differed between BFA-resistant and BFAsensitive subfamilies, but were conserved between members within each subfamily. Introducing one of these mutations, M699L, into the Sec7d of Gea1p, expressed in E. coli, led to BFA-resistant exchange activity in vitro (Peyroche et al., 1999). K. Lingelbach and colleagues isolated a BFAresistant malarial parasite line, and found that the mutation responsible for conferring BFA resistance was present in a novel Plasmodium Arf GEF (Baumgartner et al., 2001). Strikingly, the mutation was a Met→Ile mutation in the residue corresponding to M699 of Gea1p. Adjacent to the M699 residue are a pair of residues that differ between BFA-resistant and BFA-sensitive classes of Arf GEFs. The double mutant ARNO-YS190-191 was sufficient to convert ARNO into a protein with BFA-sensitive Arf GEF activity in vitro (Peyroche et al., 1999). The GeaARNO-YS190-191-Gea1p chimera, expressed in yeast, was made BFAsensitive, like wild type Gea1p (Peyroche et al., 1999). Hence, these

THE SEC7 FAMILY OF ARF GEFS

11 81

residues are determinants of BFA resistance/sensitivity both in vitro and in vivo. When the F→Y or A→S mutations were introduced separately into the Gea-ARNO-Gea1p chimera, only the F→Y mutant was BFA-sensitive (A. Peyroche and C. L. Jackson, unpublished observations), indicating that the critical residue of this pair for BFA resistance/sensitivity is the tyrosine. In a parallel study (Sata et al., 1999), chimeras were constructed between the Sec7 domains of the BFA-sensitive Sec7p and BFA-resistant cytohesin-1 Sec7 domain. The different chimeras were purified and tested for in vitro Arf GEF activity and their sensitivity to BFA. These authors narrowed down a region (within the 35-aa region described above) that was responsible for BFA resistance/sensitivity of these GEFs. In particular, two residues that differed between BFA-resistant and sensitive GEFs were identified as key for determining sensitivity to BFA. These were D965 and M975 in Sec7p, which correspond to S199 and P209 in cytohesin-1 (Sata et al., 1999). Looking at an alignment of the Sec7ds of all of the human Arf GEFs, one can ask whether it is possible to predict which ones will be BFA sensitive or resistant based on these mutagenesis studies. Figure 3 shows the four residues identified as being critical for determining BFA resistance/sensitivity. Seven of the human Arf GEFs, those in the ArfGEP100/BRAG, EFA6 and Fbx8 subfamilies, have a leucine at the position equivalent to M699 of Gea1p (Figure 3), so would be predicted to be BFA resistant. Although these seven human Arf GEFs all have aspartates and methionines at the positions equivalent to cytohesin S199 and P209, respectively, corresponding to those found in BFA-sensitive GEFs, none except Fbx8 have a tyrosine at the position corresponding to Y695 in Gea1p (F190 in ARNO/F191 in cytohesin-1) (Figure 3). Indeed, ArfGEP100 and EFA6 have activity resistant to BFA in vitro (Someya et al., 2001; Franco et al., 1999). Of the remaining GEFs, the cytohesin/ARNO/GRP subfamily has been extensively studied and shown to be BFA-resistant (Jackson and Casanova, 2000). Hence, only BIG1, BIG2 and GBF1 would be predicted to be BFA-sensitive human Arf GEFs (Figure 3). Although GBF1 has been described as a BFA-resistant GEF using Arf5 as a substrate, there is evidence, at least in vivo, that GBF1 can act on class I Arfs as well. In this case, it may well be that such activity towards a class I Arf can be inhibited by BFA (Claude et al., 1999; Kawamoto et al., 2002). 5.2.3

Structure of the Sec7 domain-BFA-Arf complex

The crystal structure of a Sec7d bound to nucleotide-free Arf reveals that the residues important for BFA resistance/sensitivity lie in the heart of the

12 82

JACKSON

Arf binding region. This result might imply that BFA acts as a competitive inhibitor, binding to the Sec7d and preventing Arf from binding. However,

Figure 3. Alignment of a portion of the Sec7 domains of the human Arf GEFs. Identical residues are boxed and homologous residues shaded blue. White letters on black background indicates residues important for determining BFA resistance/sensitivity. The asterisk denotes the catalytic glutamic acid residue (Béraud-Dufour et al., 1998).

the mechanism of action of this drug turns out to be much more unusual. BFA does not bind to either Arf or the Sec7d alone, but to a normally very short-lived reaction intermediate, an Arf•GDP-Sec7d complex (Mansour et al., 1999; Peyroche et al., 1999; Robineau et al., 2000) (Figure 4). This Arf•GDP-BFA-Sec7d quaternary complex is relatively stable, with a dissociation rate (τoff) for BFA of 30 min (Robineau et al., 2000). If BFA binds to an Arf•GDP-Sec7d protein complex, residues of Arf would be expected to influence BFA binding. Indeed, B. Antonny and colleagues identified a mutation in the Switch II region, Arf1-A80, which results in a decrease in the dissociation rate of BFA from the complex (Robineau et al., 2000). Another striking feature of the mechanism of action of BFA is the type of complex stabilized by the drug. The only complex between a small G protein and its GEF that normally can be isolated is one in which nucleotide has been lost from the complex. Robineau et al. demonstrate that BFA does not bind to such a complex (Robineau et al., 2000). The unusual mechanism of action of BFA provides a paradigm for development of novel drugs. Instead of the usual search for drugs competing for a given substrate,

THE SEC7 FAMILY OF ARF GEFS

13 83

one could design screens for drugs that block reactions through stabilization of reaction intermediates (Chardin and McCormick, 1999).

Figure 4. Mechanism of action of brefeldin A (BFA).

6.

MOLECULAR ANALYSIS OF THE ARF EXCHANGE FACTORS

6.1

Protein Partners of Arf GEFs

An important step in determining the roles of Arf GEFs in cells is to identify interacting partners. Several partners of the ARNO/cytohesin/GRP subfamily have been identified. This work is less advanced for other subfamilies of Arf GEFs, but one partner for the GNOM protein has been reported (Grebe et al., 2000). An N-terminal region conserved in all members of this subfamily, the DCB domain, is involved both in dimerization of the GNOM protein, and in binding to cyclophilin (Grebe et

14 84

JACKSON

al., 2000). The physiological relevance of this interaction is not known, but it may represent an important regulatory mechanism. Indeed, interactions between cyclophilins and regions of proteins involved in their oligomerization have been previously reported (see Grebe et al., 2000) for references). Two-hybrid screens with the Sec7 domains of different Arf GEFs have identified a surprisingly large number of potential partners. These results support the idea that the catalytic domain itself is involved in regulatory interactions. We identified the yeast TGN-localized protein Drs2p as a direct binding partner of the Sec7d of Gea2p (S. Chantalat and C. L. Jackson, manuscript in preparation). Drs2p is a 12-transmembrane domain P-type ATPase that likely functions as an aminophospholipid translocase (flippase) (Chen et al., 1999). Drs2p co-localizes with Kex2p in a late Golgi compartment in yeast, and is required for the formation of clathrin-coated vesicles. There are four other proteins in yeast that are grouped into the same subfamily of P-type ATPases along with Drs2p, which have distinct localizations and which function in different transport steps (Hua et al., 2002). The region of the Gea2p Sec7d that interacts with Drs2p was mapped by a reverse two-hybrid approach. Interestingly, the residues identified as important for the interaction are clustered in the C-terminal portion of the Sec7d, adjacent to the Arf binding site (C. Jackson, unpublished observations). This region is quite divergent among different Arf GEFs, and may represent a region generally used for regulation of Arf GEF activity. Screens with different Sec7 domains identified a number of partners (E. Smirnova, S. Chantalat, and C. L. Jackson, unpublished observations) and in at least two other cases, this same C-terminal portion of the Sec7d was found to be important for interaction. This C-terminal region is also the region in which residues affecting BFA resistance/sensitivity were found (Peyroche et al., 1999). Thus, the C-terminal portion of the Sec7d is important for the binding of other proteins and likely regulation of Arf GEF activity.

6.2

Interaction with Membranes

Members of the ARNO/cytohesin/GRP and EFA6 subfamilies have PH domains that mediate membrane binding through specific interaction with phosphoinositides (Jackson and Casanova, 2000). It is not known if the PH domains are the only localization determinants, or whether other mechanisms exist for specific membrane targeting of these GEFs. Members of the ArfGEP100/BRAG subfamily also have PH-like domains, but their sequences are quite different than those for the subfamilies mentioned above, and it is not known if they mediate membrane binding (Someya et al.,

THE SEC7 FAMILY OF ARF GEFS

15 85

2001). There are no easily identifiable PH domains present in the other subfamilies of Arf GEFs, so the mechanism of membrane localization of the Sec7/BIG and Gea/GBF/GNOM proteins remains a major question. For the Gea/GBF/GNOM subfamily of Arf GEFs, is Drs2p likely to be a major membrane localization determinant? Our data for the yeast proteins indicates that there must be additional mechanisms for localization, since the Gea2p protein is still localized in cells with a complete deletion of the DRS2 gene. Subcellular fractionation of wild type and drs2∆ cells shows the same profile for Gea2p, again suggesting that another mechanism for localization exists (S. Park and C. L. Jackson, unpublished observations). A novel transmembrane domain partner of the Gea1p and Gea2p proteins, Gmh1p, has been identified, which may be involved in localization of the Gea1p and Gea2p proteins to the Golgi apparatus (Chantalat et al., 2002). Gmh1p has orthologs in all organisms examined to date, and the human homolog, hGmh1, localizes to the Golgi complex in mammalian cells (Chantalat et al., 2002).

7.

ROLE OF THE ARF GEFS IN PROTEIN AND MEMBRANE TRAFFIC

The molecular mechanism of action of BFA indicates that the effects of this drug on cells is a result not only of failure to activate Arf, but also of trapping sensitive Arf GEFs in an inactive complex. As described above, BFA exerts profound effects on both the structure and function of organelles along the secretory and endocytic pathways. Understanding precisely how BFA affects traffic has never been easy in the context of the classic models of intracellular traffic (Griffiths, 2000; Klausner et al., 1992; Sciaky et al., 1997), and suggests that new ways of understanding membrane dynamics are needed.

7.1

Membrane Transformation Events that Accompany Traffic through the Golgi Apparatus

The current unifying principle for all traffic events in eukaryotic cells is the idea that vesicles carrying cargo bud from a donor compartment membrane, then target to and fuse with an acceptor compartment. This paradigm is based on the idea that transport pathways in the cell are divided into distinct compartments defined by a set of characteristic resident proteins. Morphological studies based on electron microscopy (EM) of ultra-thin sections do not provide evidence to contradict this view of transport. However, thin section EM has significant limitations, which prompted two

16 86

JACKSON

researchers to develop alternative methods to probe the structures of the endomembrane systems of cells (Rambourg and Clermont, 1990). Beginning in 1974, Alain Rambourg, Yves Clermont and colleagues developed novel EM techniques that allowed accurate visualization of larger segments of eukaryotic cells than had traditionally been viewed at the ultrastructural level. They used selective impregnation methods that provided high levels of contrast between the membrane and the surrounding cytoplasm. Thick sections (from 200 nm to over 1µm) were observed with the aid of stereoscopic techniques. These techniques permitted the visualization of the three-dimensional structure of cellular organelles (Rambourg and Clermont, 1990). Rambourg and Clermont demonstrated that the Golgi apparatus of mammalian cells does not consist of independent, separate compartments (stacks of saccules) but rather is one continuous entity with the appearance of a continuous ribbon (Rambourg and Clermont, 1997). The cis compartment (or cis-tubular network) consists of a system of closely anastomosed tubules, which form a network of regular polygonal mesh. The medial compartment consists of stacks of saccules interconnected by intersaccular tubular zones. Each stack of saccules is formed by closely superposed flattened elements (saccules), poorly fenestrated, and strictly parallel with each other. The trans compartment comprises several superposed sacculotubular elements which appear as independent units along the main axis of the Golgi ribbon and curve away from the latter into the cytoplasm as if they were peeling off (Figure 5; Rambourg and Clermont, 1997). In secretory cells, the secretory material is first uniformly distributed in the lumen of the Golgi elements. This material then begins to accumulate in dilations located at apparently random locations of a saccular element of the compact zone situated at any level along the cis-trans axis of the Golgi ribbon (although most commonly in a saccule of the trans compartment) (Figure 5; Rambourg et al., 1987). The drainage of the secretory material towards these dilations is accompanied by a progressive perforation and tubulation of the large saccular portions of the trans element (Rambourg et al., 1987). Hence, segregation of the secretory material (which results in formation of a progranule from the original dilation) is tightly coupled to membrane transformation (from saccular to fenestrated to tubular). When the tubular elements connecting the dilations/progranules become devoid of secretory content, they fragment to liberate the pro-secretory granules at the trans face of the Golgi apparatus (Figure 5). This tubulation of the saccular portions which accompanies the segregation and maturation of the secretory material, and the subsequent rupture of the tubular systems at the trans face of the Golgi apparatus are the manifestations of an intense level of protein

THE SEC7 FAMILY OF ARF GEFS

17 87

and lipid sorting activity at this level of the organelle (Rambourg et al., 1987; Robinson and Bonifacino, 2001). The membranes of the Golgi apparatus are the object of permanent reorganization, as demonstrated by studies of Golgi structure under conditions of physiological stimulation and inhibition of secretion. In prolactin cells of the pituitary gland of lactating rats or acinar cells of the mammary gland of such animals, the Golgi rapidly breaks down when the secretion stimulus is removed, and then reforms within minutes of the pups being returned to their mother for lactation (Clermont et al., 1993; Rambourg et al., 1993). The results demonstrate that the Golgi apparatus arises as a result of continual renewal at the cis side, and consumption at the trans side. Contrary to what is observed in mammalian cells, the Golgi apparatus of S. cerevisiae comprises approximately thirty independent units, distributed seemingly randomly throughout the cytoplasm (Rambourg et al., 2001). Each unit possesses a structure that is exclusively tubular and in the form of small tubular networks of polygonal meshes (Rambourg et al., 2001). There are at least two reasons for the dispersed nature of Golgi elements in S. cerevisiae, as opposed to mammalian cells (or even other yeasts such as Pichia pastoris). First, S. cerevisiae lacks the microtubule-dependent system that carries ER-Golgi transport intermediates to the pericentriolar region (Huffaker et al., 1988). Second, there are more ER exit sites present in S. cerevisiae than in other organisms because the former lacks a system that assembles exit sites together at the level of the ER (Rossanese et al., 1999). However, despite the differences, the fundamental process of secretion is highly conserved between S. cerevisiae and higher eukaryotic cells (Duden and Schekman, 1997; Ferro-Novick and Jahn, 1994; Kaiser and Huffaker, 1992). Morphological analysis of wild type S. cerevisiae strains reveals that formation of secretory vesicles/granules occurs by a mechanism remarkably similar to that in mammalian cells. Nodular swellings resembling secretion vesicles/granules form at the intersections of polygonal meshes made up of interconnected tubules (Rambourg et al., 2001). In structures representing the next stage of development, nodules that have the appearance of fully formed secretory vesicles are connected by very fine tubes, and are adjacent to free secretory vesicles (Rambourg et al., 2001). Thus, it is likely that in yeast, as in mammalian cells (Rambourg and Clermont, 1997), the liberation of secretion vesicles/granules occurs by rupture of tubular areas and not by budding from the edges of saccular elements.

18 88

JACKSON

Figure 5. Diagram showing a small section of the Golgi ribbon in cross section (top). Selected elements are illustrated in the three dimensional depiction shown in the lower panel. ER, endoplasmic reticulum; CE, cis element; TTN, trans-tubular network; Pg, progranule; g, granule. Taken with permission from Rambourg et al., 1987.

7.2

Temperature-sensitive mutants in yeast: Evidence For Vectorial Forward Membrane Flow

In the original selection for secretory pathway mutants (sec mutants) in S. cerevisiae, a mutation in the gene encoding what would turn out to be the founding member of the Arf GEF family was isolated (Novick et al., 1981; Novick et al., 1980). At the restrictive temperature (37oC), this mutant, sec7-1, has a block in the secretory pathway at the level of exit from the Golgi apparatus (Franzusoff and Schekman, 1989; Julius et al., 1984; Nishikawa et al., 1990; Novick et al., 1981; Stevens et al., 1982). At steady

THE SEC7 FAMILY OF ARF GEFS

19 89

state, Sec7p localizes primarily to late Golgi compartments, partially colocalizing with Kex2p, a marker protein for the yeast TGN (Franzusoff et al., 1991). In the sec7-1 mutant over a time course of incubation at 37oC, the Golgi tubular networks were progressively transformed into stacks of saccules whose superimposed elements were interconnected along a cistrans axis to form a membrane structure that is anatomically continuous (Figure 6; Rambourg et al., 1993). This process can be divided into several steps. First, after only seven minutes at 37oC, one or a few of the numerous Golgi tubular networks became more extensive, but with the same type of organization as in a wild type strain. At this time, secretory vesicles/granules were rarely seen (Figure 6B). After 15 minutes at the restrictive temperature, the tubular networks within a single Golgi element had become aligned in a parallel fashion. Some cells had accumulated highly fenestrated cisternae, suggesting that the tubular elements, once aligned in parallel, had begun to undergo filling in of the tubular mesh to form fenestrated saccules (tubulation- fenestration transition) (Figure 6C). After 30-60 minutes, only a few Golgi elements per cell were seen, and all were composed of five to eight stacked, poorly fenestrated saccules. Hence the next membrane transformation step, fenestration to saccularization, had been completed (Figure 6D). Morphologically, these Golgi elements closely resemble Golgi stacks in mammalian cells (Rambourg et al., 1993). That these structures which accumulate in the sec7-1 mutant are bone fide Golgi elements is supported by the fact that Golgi-localized proteins such as Ypt1p (Segev et al., 1988), Rer1p (Sato et al., 1995), Vps1p (Rothman et al., 1990) and the TRAPP ER-Golgi vesicle-targeting complex (Sacher et al., 1998), as well as proteins in transit through the secretory pathway such as DPAP B (Roberts et al., 1989), are associated with them. The Arf GEFs, through activation of Arf, act in the process of budding of a transport vesicle from a donor membrane. However, as illustrated by Rothman and Warren, vesicle budding is topologically equivalent to fission of tubules and fenestration of cisternal membranes (see Figures 5 and 6 in (Rothman and Warren, 1994)). The sec7-1 mutant has a phenotype consistent with defects in breaking of tubules and in fenestration of cisternae. Interestingly, BFA has a very different phenotype. In both S. cerevisiae and mammalian cells, BFA causes the Golgi to disappear rather than accumulate. These morphological results are consistent with the molecular differences in the two lesions. BFA binds to the Sec7d of sensitive Arf GEFs such as yeast Sec7p, whereas mutations in the sec7-1 mutant protein lie outside of the Sec7d (Deitz et al., 2000). It would be interesting to know if the residues mutated in sec7-1 are involved specifically in interaction with proteins that mediate membrane tubule breakage and saccule fenestration.

20 90

JACKSON

Figure 6. Diagrammatic representation of the modifications of the Golgi apparatus of an S. cerevisiae sec7-1 mutant following shift to the non-permissive temperature of 37oC. A) A Golgi saccule (S) from a wild type S. cerevisiae strain or a sec7-1 mutant at the permissive temperature of 24oC. B) A Golgi element (S) in a sec7-1 mutant at 37oC for 7.5 to 15 minutes. Thin plates (P) of electron-dense material appear at this stage. C) A transverse section through a stack of saccules in the sec7-1 mutant maintained at 37oC for 15-30 minutes. Four flattened saccules (S), interconnected with each other, are stacked and separated by a space showing thin plates (P) of electron-dense material. D) A transverse section through a Golgi stack from the sec7-1 mutant incubated at 37oC for 30 min to 1 hour. Fenestrated spheres (FS) often appear at this stage. Taken with permission from Rambourg et al., 1993.

In vivo morphological analysis of ts mutants in the Gea1p-Gea2p Arf GEF pair also support a role for Arf GEFs both in Golgi structure and in

THE SEC7 FAMILY OF ARF GEFS

21 91

regulation of vectorial membrane flow. Gea1p and Gea2p have redundant functions that are largely interchangeable (Peyroche and Jackson, 2001; Peyroche et al., 1996; Spang et al., 2001). Single mutant strains have no growth or secretion phenotypes, whereas a gea1∆ gea2∆ strain is inviable (Peyroche et al., 1996). In three ts mutants, gea1-4, gea1-6 and gea1-19, proteins exit the ER, but are blocked or slowed at an earlier stage in transport than for sec7-ts mutants. All of the gea-ts mutants have severe defects in glycosylation, much more severe than in sec7-ts mutants (Peyroche et al., 2001). Interestingly, gea1-4 accumulated larger-than normal Golgi tubular networks, but for the most part their structure was similar to wild type Golgi elements (Peyroche et al., 2001). The slow-down in transport in the gea1-4 mutant occurred at a point prior to that in the sec7-1 mutant, after exit from the ER but before Golgi exit. The structures that accumulated were also upstream in the membrane transformation pathway compared to those accumulated in the sec7-1 mutant. In contrast, gea1-6 and gea1-19 had essentially complete blocks to transport at the level of the cis-Golgi. In these mutants, no Golgi tubular network structures accumulated, but rather unfenestrated membrane structures (Peyroche et al., 2001). Like mutations in the Arf GEFs, mutations in Arf itself have phenotypes consistent with an important role for this protein in regulation of vectorial membrane flow, and in membrane fenestration and tubulation in yeast (Deitz et al., 2000; Rambourg et al., 1995). The arf1∆ mutant is viable at all temperatures, but grows more slowly, especially at low temperatures. This mutant has defects in glycosylation and the rate of secretion of all proteins traversing the secretory pathway is slowed (Gaynor et al., 1998). Even more pronounced than the transport defects are marked alterations in Golgi and endosome structure (Deitz et al., 2000; Gaynor et al., 1998; Rambourg et al., 1995). Again, slowing of the rate of secretion results in an accumulation of larger-than-normal Golgi structures. Isolation of a number of randomly generated ts mutants in ARF1 showed that yeast Arf acts at multiple steps of intracellular transport, including ER-Golgi, Golgi-endosome, endosomevacuole and Golgi-PM (Yahara et al., 2001). Analysis of these mutants is consistent with Arf regulating both organelle structure and protein transport at each of these steps (Yahara et al., 2001). The studies described above show that understanding the roles of Arf GEFs even at the single cell level is currently a challenge. An even bigger challenge is understanding the roles of these proteins at the level of the organism, for example in development. The most information to date on this subject has come from studies of Arabidopsis thaliana. The gnom mutant was identified as a pattern-formation mutant, and has defects in polarized growth and asymmetric cell division (Mayer et al., 1993). GNOM is

22 92

JACKSON

involved in polarized secretion of auxin, a plant growth hormone (Geldner et al., 1999). Polar auxin transport is important for a number of developmental processes in plants, and in Arabidopsis, is dependent on the polarized localization of PIN1, an auxin efflux carrier. PIN1 localization is severely perturbed in gnom mutant embryos, and is also disrupted by BFA treatment (Geldner et al., 2001; Steinmann et al., 1999). Since GNOM is a BFAsensitive Arf GEF, both of these lines of evidence indicate that GNOMdependent membrane traffic is important for polarized localization of PIN1 and thus polarized auxin secretion. Interestingly, in contrast to Gea2p and GBF1, GNOM does not appear to be localized primarily to the Golgi apparatus at steady state. Instead, GNOM is associated with multiple membrane compartments, including the plasma membrane (Steinmann et al., 1999).

8.

FUTURE DIRECTIONS

This is an exciting time for research in the Arf GEF field. The profound effects of BFA on membrane systems in eukaryotic cells demonstrate that its targets, the Arf GEFs, are central regulators of organelle structure and function. The Arf GEFs are all multidomain proteins that, as is the case for the Rho family of GEFs, likely integrate upstream signals with the downstream effects of GTPase activation. Molecular analysis of the Arf GEFs is just now getting underway, and should soon lead to the identification of membrane localization determinants, and both positive and negative regulators of exchange activity. Analysis of membrane systems both by electron and fluorescence microscopy leads to the concept that regulated membrane flow is an essential aspect of membrane dynamics in eukaryotic cells. In vivo analysis of yeast Arf GEF mutants shows that a specific type of lesion in different Arf GEFs leads to accumulation of larger-than-normal Golgi structures. In these cases, the type of structure that accumulates depends on the point in the secretory pathway that is inhibited by the mutation. It will be interesting to determine whether these lesions abolish interaction with a particular protein or membrane partner. A given structure is a manifestation, at a given time and place, of an underlying dynamic process. The challenge for the future is to understand this process at the molecular level, and the role of the Arf GEFs in integrating upstream signals with the generation of dynamic membrane structures.

THE SEC7 FAMILY OF ARF GEFS

9.

23 93

ACKNOWLEDGEMENTS

I thank Alain Rambourg and Yves Clermont for many stimulating discussions, and for allowing me to present their views. I apologize to authors whose work could not be cited due to space limitations.

REFERENCES Antonny, B., Béraud-Dufour, S., Chardin, P., and Chabre, M. (1997). N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry, 36, 4675-4684. Baumgartner, F., Wiek, S., Paprotka, K., Zauner, S., and Lingelbach, K. (2001). A point mutation in an unusual Sec7 domain is linked to brefeldin A resistance in a Plasmodium falciparum line generated by drug selection. Mol. Microbiol., 41, 1151-1158. Béraud-Dufour, S., Robineau, S., Chardin, P., Paris, S., Chabre, M., Cherfils, J., and Antonny, B. (1998). A glutamic finger in the guanine nucleotide exchange factor ARNO displaces Mg2+ and the beta-phosphate to destabilize GDP on ARF1. EMBO J., 17, 3651-3659. Boehm, M., Aguilar, R. C., and Bonifacino, J. S. (2001). Functional and physical interactions of the adaptor protein complex AP- 4 with ADP-ribosylation factors (ARFs). EMBO J., 20, 6265-6276. Boehm, M., and Bonifacino, J. S. (2001). Adaptins: the final recount. Mol. Biol. Cell, 12, 2907-2920. Boman, A. L., Zhang, C., Zhu, X., and Kahn, R. A. (2000). A family of ADP-ribosylation factor effectors that can alter membrane transport through the trans-Golgi. Mol. Biol. Cell, 11, 1241-1255. Chantalat, S., Courbeyrette, R., Senic-Matuglia, F., Jackson, C. L., Goud, B., and Peyroche, A. (2003). A novel Golgi membrane protein is a partner of the ARF exchange factors Gea1p and Gea2p. Mol. Biol. Cell, in press, published March 7, 2003 as 10.1091/mbc.E02-10-0693. Chardin, P., and McCormick, F. (1999). Brefeldin A: the advantage of being uncompetitive. Cell, 97, 153-155. Chardin, P., Paris, S., Antonny, B., Robineau, S., Beraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996). A human exchange factor for ARF contains Sec7- and pleckstrinhomology domains. Nature, 384, 481-484. Chen, C. Y., Ingram, M. F., Rosal, P. H., and Graham, T. R. (1999). Role for Drs2p, a P-type ATPase and potential aminophospholipid translocase, in yeast late Golgi function. J. Cell Biol., 147, 1223-1236. Claude, A., Zhao, B. P., Kuziemsky, C. E., Dahan, S., Berger, S. J., Yan, J. P., Armold, A. D., Sullivan, E. M., and Melancon, P. (1999). GBF1: A novel Golgi-associated BFA-resistant guanine nucleotide exchange factor that displays specificity for ADP-ribosylation factor 5. J. Cell Biol., 146, 71-84. Clermont, Y., Xia, L., Rambourg, A., Turner, J. D., and Hermo, L. (1993). Structure of the Golgi apparatus in stimulated and non-stimulated acinar cells of mammary glands of the rat. Anat. Rec., 237, 308-317. Dascher, C., and Balch, W. E. (1994). Dominant inhibitory mutants of ARF1 block endoplasmic reticulum to Golgi transport and trigger disassembly of the Golgi apparatus. J. Biol. Chem., 269, 1437-1448.

24 94

JACKSON

De Matteis, M. A., and Morrow, J. S. (2000). Spectrin tethers and mesh in the biosynthetic pathway. J. Cell Sci., 113, 2331-2343. Deitz, S. B., Rambourg, A., Kepes, F., and Franzusoff, A. (2000). Sec7p directs the transitions required for yeast Golgi biogenesis. Traffic, 1, 172-183. Dell'Angelica, E. C., Puertollano, R., Mullins, C., Aguilar, R. C., Vargas, J. D., Hartnell, L. M., and Bonifacino, J. S. (2000). GGAs: a family of ADP ribosylation factor-binding proteins related to adaptors and associated with the Golgi complex. J. Cell Biol., 149, 8194. Derrien, V., Couillault, C., Franco, M., Martineau, S., Montcourrier, P., Houlgatte, R., and Chavrier, P. (2002). A conserved C-terminal domain of EFA6-family ARF6-guanine nucleotide exchange factors induces lengthening of microvilli-like membrane protrusions. J. Cell Sci., 115, 2867-2879. Doms, R. W., Russ, G., and Yewdell, J. W. (1989). Brefeldin A redistributes resident and itinerant Golgi proteins to the endoplasmic reticulum. J. Cell Biol., 109, 61-72. Donaldson, J. G., Cassel, D., Kahn, R. A., and Klausner, R. D. (1992). ADP-ribosylation factor, a small GTP-binding protein, is required for binding of the coatomer protein betaCOP to Golgi membranes. Proc. Natl. Acad. Sci., USA, 89, 6408-6412. Donaldson, J. G., Finazzi, D., and Klausner, R. D. (1992). Brefeldin A inhibits Golgi membrane-catalyzed exchange of guanine nucleotide onto ARF protein. Nature, 360, 350352. Donaldson, J. G., Kahn, R. A., Lippincott-Schwartz, J., and Klausner, R. D. (1991). Binding of ARF and beta-COP to Golgi membranes: possible regulation by a trimeric G protein. Science, 254, 1197-1199. Donaldson, J. G., Lippincott-Schwartz, J., Bloom, G. S., Kreis, T. E., and Klausner, R. D. (1990). Dissociation of a 110-kD peripheral membrane protein from the Golgi apparatus is an early event in brefeldin A action. J. Cell Biol., 111, 2295-2306. Drake, M. T., Zhu, Y., and Kornfeld, S. (2000). The assembly of AP-3 adaptor complexcontaining clathrin-coated vesicles on synthetic liposomes. Mol. Biol. Cell, 11, 3723-3736. Duden, R., and Schekman, R. (1997). Insights into Golgi function through mutants in yeast and animal cells. In E. G. Berger and J. Roth (Eds.), The Golgi apparatus (pp. 219-246). Basel, Switzerland: Birkhauser Verlag. Ferro-Novick, S., and Jahn, R. (1994). Vesicle fusion from yeast to man. Nature, 370, 191193. Franco, M., Boretto, J., Robineau, S., Monier, S., Goud, B., Chardin, P., and Chavrier, P. (1998). ARNO3, a Sec7-domain guanine nucleotide exchange factor for ADP ribosylation factor 1, is involved in the control of Golgi structure and function. Proc. Natl. Acad. Sci., USA, 95, 9926-9931. Franco, M., Peters, P. J., Boretto, J., van Donselaar, E., Neri, A., D'Souza-Schorey, C., and Chavrier, P. (1999). EFA6, a sec7 domain-containing exchange factor for ARF6, coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 14801491. Frank, S., Upender, S., Hansen, S. H., and Casanova, J. E. (1998). ARNO is a guanine nucleotide exchange factor for ADP-ribosylation factor 6. J. Biol. Chem., 273, 23-27. Franzusoff, A., Redding, K., Crosby, J., Fuller, R. S., and Schekman, R. (1991). Localization of components involved in protein transport and processing through the yeast Golgi apparatus. J. Cell Biol., 112, 27-37. Franzusoff, A., and Schekman, R. (1989). Functional compartments of the yeast Golgi apparatus are defined by the sec7 mutation. EMBO J., 8, 2695-2702.

THE SEC7 FAMILY OF ARF GEFS

25 95

Fujiwara, T., Oda, K., Yokota, S., Takatsuki, A., and Ikehara, Y. (1988). Brefeldin A causes disassembly of the Golgi complex and accumulation of secretory proteins in the endoplasmic reticulum. J. Biol. Chem., 263, 18545-18552. Gaynor, E. C., Chen, C. Y., Emr, S. D., and Graham, T. R. (1998). ARF is required for maintenance of yeast Golgi and endosome structure and function. Mol. Biol. Cell, 9, 653670. Geldner, N., Friml, J., Stierhof, Y. D., Jurgens, G., and Palme, K. (2001). Auxin transport inhibitors block PIN1 cycling and vesicle trafficking. Nature, 413, 425-428. Grebe, M., Gadea, J., Steinmann, T., Kientz, M., Rahfeld, J. U., Salchert, K., Koncz, C., and Jurgens, G. (2000). A conserved domain of the arabidopsis GNOM protein mediates subunit interaction and cyclophilin 5 binding. Plant Cell, 12, 343-356. Griffiths, G. (2000). Gut thoughts on the Golgi complex. Traffic, 1, 738-745. Helms, J. B., and Rothman, J. E. (1992). Inhibition by brefeldin A of a Golgi membrane enzyme that catalyses exchange of guanine nucleotide bound to ARF. Nature, 360, 352354. Hua, Z., Fatheddin, P., and Graham, T. R. (2002). An essential subfamily of Drs2p-related Ptype ATPases are required for protein trafficking between the Golgi complex and the endosomal/vacuolar system. Mol. Biol. Cell, 13, 3162-3177. Published July 16, 2002 as 2010.1091/mbc.E2002-2003-0172. Huffaker, T. C., Thomas, J. H., and Botstein, D. (1988). Diverse effects of beta-tubulin mutations on microtubule formation and function. J. Cell Biol., 106, 1997-2010. Hunziker, W., Whitney, J. A., and Mellman, I. (1992). Brefeldin A and the endocytic pathway. Possible implications for membrane traffic and sorting. FEBS Lett, 307, 93-96. Jackson, C. L. (2000). Brefeldin A: Revealing the fundamental principles governing membrane dynamics and protein transport. In H. Hilderson and S. Fuller (Eds.), Subcellular Biochemistry: Fusion of biological membranes and related problems (Vol. 34, pp. 233-272). New York: Kluwer Academic/Plenum Publishers. Jackson, C. L., and Casanova, J. E. (2000). Turning on ARF: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. Jones, S., Jedd, G., Kahn, R. A., Franzusoff, A., Bartolini, F., and Segev, N. (1999). Genetic interactions in yeast between Ypt GTPases and Arf guanine nucleotide exchangers. Genetics, 152, 1543-1556. Julius, D., Schekman, R., and Thorner, J. (1984). Glycosylation and processing of preproalpha-factor through the yeast secretory pathway. Cell, 36, 309-318. Kaiser, C., and Huffaker, T. (1992). An inside look at a small cell. Yeast Cell Biology sponsored by Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA, August 1318, 1991. New Biol., 4, 23-27. Kawamoto, K., Yoshida, Y., Tamaki, H., Torii, S., Shinotsuka, C., Yamashina, S., and Nakayama, K. (2002). GBF1, a guanine nucleotide exchange factor for ADP-ribosylation factors, is localized to the cis-Golgi and involved in membrane association of the COPI coat. Traffic, 3, 483-495. Klausner, R. D., Donaldson, J. G., and Lippincott-Schwartz, J. (1992). Brefeldin A: insights into the control of membrane traffic and organelle structure. J. Cell Biol., 116, 1071-1080. Liang, J. O., and Kornfeld, S. (1997). Comparative activity of ADP-ribosylation factor family members in the early steps of coated vesicle formation on rat liver Golgi membranes [published erratum appears in J. Biol. Chem. 1998 Jan 23;273:2488]. J. Biol. Chem., 272, 4141-4148. Lippincott-Schwartz, J., Yuan, L., Tipper, C., Amherdt, M., Orci, L., and Klausner, R. D. (1991). Brefeldin A's effects on endosomes, lysosomes, and the TGN suggest a general mechanism for regulating organelle structure and membrane traffic. Cell, 67, 601-616.

26 96

JACKSON

Lippincott-Schwartz, J., Yuan, L. C., Bonifacino, J. S., and Klausner, R. D. (1989). Rapid redistribution of Golgi proteins into the ER in cells treated with brefeldin A: evidence for membrane cycling from Golgi to ER. Cell, 56, 801-813. Liu, L., and Pohajdak, B. (1992). Cloning and sequencing of a human cDNA from cytolytic NK/T cells with homology to yeast SEC7. Biochim. Biophys. Acta, 1132, 75-78. Macia, E., Chabre, M., and Franco, M. (2001). Specificities for the small G proteins ARF1 and ARF6 of the guanine nucleotide exchange factors ARNO and EFA6. J. Biol. Chem., 276, 24925-24930. Mansour, S. J., Skaug, J., Zhao, X. H., Giordano, J., Scherer, S. W., and Melancon, P. (1999). p200 ARF-GEP1: a Golgi-localized guanine nucleotide exchange protein whose Sec7 domain is targeted by the drug brefeldin A. Proc. Natl. Acad. Sci., USA, 96, 7968-7973. Mayer, U., Büttner, G., and Jürgens, G. (1993). Apical-basal pattern formation in the Arabidopsis embryo: studies on the role of the gnom gene. Development, 117, 149-162. Misumi, Y., Miki, K., Takatsuki, A., Tamura, G., and Ikehara, Y. (1986). Novel blockade by brefeldin A of intracellular transport of secretory proteins in cultured rat hepatocytes. J. Biol. Chem., 261, 11398-11403. Morinaga, N., Moss, J., and Vaughan, M. (1997). Cloning and expression of a cDNA encoding a bovine brain brefeldin A-sensitive guanine nucleotide-exchange protein for ADP-ribosylation factor. Proc. Natl. Acad. Sci., USA, 94, 12926-12931. Morinaga, N., Tsai, S. C., Moss, J., and Vaughan, M. (1996). Isolation of a brefeldin Ainhibited guanine nucleotide-exchange protein for ADP ribosylation factor (ARF) 1 and ARF3 that contains a Sec7-like domain. Proc. Natl. Acad. Sci., USA, 93, 12856-12860. Moss, J., and Vaughan, M. (1998). Molecules in the ARF orbit. J. Biol. Chem., 273, 2143121434. Nagai, H., Kagan, J. C., Zhu, X., Kahn, R. A., and Roy, C. R. (2002). A bacterial guanine nucleotide exchange factor activates ARF on Legionella phagosomes. Science, 295, 679682. Nair, S., Guerra, C., and Satir, P. (1999). A Sec7-related protein in Paramecium. FASEB J., 13, 1249-1257. Nishikawa, S., Umemoto, N., Ohsumi, Y., Nakano, A., and Anraku, Y. (1990). Biogenesis of vacuolar membrane glycoproteins of yeast Saccharomyces cerevisiae. J. Biol. Chem., 265, 7440-7448. Novick, P., Ferro, S., and Schekman, R. (1981). Order of events in the yeast secretory pathway. Cell, 25, 461-469. Novick, P., Field, C., and Schekman, R. (1980). Identification of 23 complementation groups required for post-translational events in the yeast secretory pathway. Cell, 21, 205-215. Ogasawara, M., Kim, S. C., Adamik, R., Togawa, A., Ferrans, V. J., Takeda, K., Kirby, M., Moss, J., and Vaughan, M. (2000). Similarities in function and gene structure of cytohesin-4 and cytohesin-1, guanine nucleotide-exchange proteins for ADP-ribosylation factors. J. Biol. Chem., 275, 3221-3230. Ooi, C. E., Dell'Angelica, E. C., and Bonifacino, J. S. (1998). ADP-Ribosylation factor 1 (ARF1) regulates recruitment of the AP-3 adaptor complex to membranes. J. Cell Biol., 142, 391-402. Peyroche, A., Antonny, B., Robineau, S., Acker, J., Cherfils, J., and Jackson, C. L. (1999). Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell, 3, 275-285. Peyroche, A., Courbeyrette, R., Rambourg, A., and Jackson, C. L. (2001). The ARF exchange factors Gea1p and Gea2p regulate Golgi structure and function in yeast. J. Cell Sci., 114, 2241-2253.

THE SEC7 FAMILY OF ARF GEFS

27 97

Peyroche, A., and Jackson, C. L. (2001). Functional analysis of ADP-ribosylation factor (ARF) guanine nucleotide exchange factors Gea1p and Gea2p in yeast. Methods Enzymol., 329, 290-300. Peyroche, A., Paris, S., and Jackson, C. L. (1996). Nucleotide exchange on ARF mediated by yeast Gea1 protein. Nature, 384, 479-481. Presley, J. F., Ward, T. H., Pfeifer, A. C., Siggia, E. D., Phair, R. D., and LippincottSchwartz, J. (2002). Dissection of COPI and Arf1 dynamics in vivo and role in Golgi membrane transport. Nature, 417, 187-193. Rambourg, A., and Clermont, Y. (1990). Three-dimensional electron microscopy: structure of the Golgi apparatus. Eur. J. Cell Biol., 51, 189-200. Rambourg, A., and Clermont, Y. (1997). Three-dimensional structure of the Golgi apparatus in mammalian cells. In E. G. Berger and J. Roth (Eds.), The Golgi apparatus (pp. 37-61). Basel, Switzerland: Birkhauser Verlag. Rambourg, A., Clermont, Y., Chretien, M., and Olivier, L. (1993). Modulation of the Golgi apparatus in stimulated and non-stimulated prolactin cells of female rats. Anat. Rec., 235, 353-362. Rambourg, A., Clermont, Y., Hermo, L., and Segretain, D. (1987). Tridimensional architecture of the Golgi apparatus and its components in mucous cells of Brunner's glands of the mouse. Am. J. Anat., 179, 95-107. Rambourg, A., Clermont, Y., Jackson, C. L., and Kepes, F. (1995). Effects of brefeldin A on the three-dimensional structure of the Golgi apparatus in a sensitive strain of Saccharomyces cerevisiae. Anat. Rec., 241, 1-9. Rambourg, A., Clermont, Y., and Kepes, F. (1993). Modulation of the Golgi apparatus in Saccharomyces cerevisiae sec7 mutants as seen by three-dimensional electron microscopy. Anat. Rec., 237, 441-452. Rambourg, A., Jackson, C. L., and Clermont, Y. (2001). Three dimensional configuration of the secretory pathway and segregation of secretion granules in the yeast Saccharomyces cerevisiae. J. Cell Sci., 114, 2231-2239. Randazzo, P. A., Yang, Y. C., Rulka, C., and Kahn, R. A. (1993). Activation of ADPribosylation factor by Golgi membranes. Evidence for a brefeldin A- and proteasesensitive activating factor on Golgi membranes. J. Biol. Chem., 268, 9555-9563. Reaves, B., and Banting, G. (1992). Perturbation of the morphology of the trans-Golgi network following brefeldin A treatment: redistribution of a TGN-specific integral membrane protein, TGN38. J. Cell Biol., 116, 85-94. Roberts, C. J., Pohlig, G., Rothman, J. H., and Stevens, T. H. (1989). Structure, biosynthesis, and localization of dipeptidyl aminopeptidase B, an integral membrane glycoprotein of the yeast vacuole. J. Cell Biol., 108, 1363-1373. Robineau, S., Chabre, M., and Antonny, B. (2000). Binding site of brefeldin A at the interface between the small G protein ADP-ribosylation factor 1 (ARF1) and the nucleotideexchange factor Sec7 domain. Proc. Natl. Acad. Sci., USA, 97, 9913-9918. Robinson, M. S., and Bonifacino, J. S. (2001). Adaptor-related proteins. Curr. Opin. Cell Biol., 13, 444-453. Rossanese, O. W., Soderholm, J., Bevis, B. J., Sears, I. B., O'Connor, J., Williamson, E. K., and Glick, B. S. (1999). Golgi structure correlates with transitional endoplasmic reticulum organization in Pichia pastoris and Saccharomyces cerevisiae. J. Cell Biol., 145, 69-81. Rothman, J. E., and Warren, G. (1994). Implications of the SNARE hypothesis for intracellular membrane topology and dynamics. Curr. Biol., 4, 220-233. Rothman, J. H., Raymond, C. K., Gilbert, T., O'Hara, P. J., and Stevens, T. H. (1990). A putative GTP binding protein homologous to interferon-inducible Mx proteins performs an essential function in yeast protein sorting. Cell, 61, 1063-1074.

28 98

JACKSON

Sacher, M., Jiang, Y., Barrowman, J., Scarpa, A., Burston, J., Zhang, L., Schieltz, D., Yates, J. R., 3rd, Abeliovich, H., and Ferro-Novick, S. (1998). TRAPP, a highly conserved novel complex on the cis-Golgi that mediates vesicle docking and fusion. EMBO J., 17, 24942503. Santy, L. C., and Casanova, J. E. (2001). Activation of ARF6 by ARNO stimulates epithelial cell migration through downstream activation of both Rac1 and phospholipase D. J. Cell Biol., 154, 599-610. Sata, M., Moss, J., and Vaughan, M. (1999). Structural basis for the inhibitory effect of brefeldin A on guanine nucleotide-exchange proteins for ADP-ribosylation factors. Proc. Natl. Acad. Sci., USA, 96, 2752-2757. Sato, K., Nishikawa, S., and Nakano, A. (1995). Membrane protein retrieval from the Golgi apparatus to the endoplasmic reticulum (ER): characterization of the RER1 gene product as a component involved in ER localization of Sec12p. Mol. Biol. Cell, 6, 1459-1477. Sciaky, N., Presley, J., Smith, C., Zaal, K. J., Cole, N., Moreira, J. E., Terasaki, M., Siggia, E., and Lippincott-Schwartz, J. (1997). Golgi tubule traffic and the effects of brefeldin A visualized in living cells. J. Cell Biol., 139, 1137-1155. Segev, N., Mulholland, J., and Botstein, D. (1988). The yeast GTP-binding YPT1 protein and a mammalian counterpart are associated with the secretion machinery. Cell, 52, 915-924. Shevell, D. E., Leu, W. M., Gillmor, C. S., Xia, G., Feldmann, K. A., and Chua, N. H. (1994). EMB30 is essential for normal cell division, cell expansion, and cell adhesion in Arabidopsis and encodes a protein that has similarity to Sec7. Cell, 77, 1051-1062. Shinotsuka, C., Yoshida, Y., Kawamoto, K., Takatsu, H., and Nakayama, K. (2002). Overexpression of an ADP-ribosylation factor-guanine nucleotide exchange factor, BIG2, uncouples brefeldin A-induced adaptor protein-1 coat dissociation and membrane tubulation. J. Biol. Chem., 277, 9468-9473. Someya, A., Sata, M., Takeda, K., Pacheco-Rodriguez, G., Ferrans, V. J., Moss, J., and Vaughan, M. (2001). ARF-GEP (100), a guanine nucleotide-exchange protein for ADPribosylation factor 6. Proc. Natl. Acad. Sci., USA, 98, 2413-2418. Spang, A., Herrmann, J. M., Hamamoto, S., and Schekman, R. (2001). The ADP ribosylation factor-nucleotide exchange factors Gea1p and Gea2p have overlapping, but not redundant functions in retrograde transport from the Golgi to the endoplasmic reticulum. Mol. Biol. Cell, 12, 1035-1045. Stearns, T., Kahn, R. A., Botstein, D., and Hoyt, M. A. (1990). ADP ribosylation factor is an essential protein in Saccharomyces cerevisiae and is encoded by two genes. Mol. Cell Biol., 10, 6690-6699. Steinmann, T., Geldner, N., Grebe, M., Mangold, S., Jackson, C. L., Paris, S., Galweiler, L., Palme, K., and Jurgens, G. (1999). Coordinated polar localization of auxin efflux carrier PIN1 by GNOM ARF GEF. Science, 286, 316-318. Stevens, T., Esmon, B., and Schekman, R. (1982). Early stages in the yeast secretory pathway are required for transport of carboxypeptidase Y to the vacuole. Cell, 30, 439-448. Takatsuki, A., and Tamura, G. (1985). Brefeldin A, a specific inhibitor of intracellular translocation of vesicular stomatitis virus G protein: intracellular accumulation of highmannose type G protein and inhibition of its cell surface expression. Agric. Biol. Chem., 49, 899-902. Togawa, A., Morinaga, N., Ogasawara, M., Moss, J., and Vaughan, M. (1999). Purification and cloning of a brefeldin A-inhibited guanine nucleotide-exchange protein for ADPribosylation factors. J. Biol. Chem., 274, 12308-12315. Wood, S. A., and Brown, W. J. (1992). The morphology but not the function of endosomes and lysosomes is altered by brefeldin A. J. Cell Biol., 119, 273-285.

THE SEC7 FAMILY OF ARF GEFS

29 99

Wood, S. A., Park, J. E., and Brown, W. J. (1991). Brefeldin A causes a microtubulemediated fusion of the trans-Golgi network and early endosomes. Cell, 67, 591-600. Yahara, N., Ueda, T., Sato, K., and Nakano, A. (2001). Multiple roles of Arf1 GTPase in the yeast exocytic and endocytic pathways. Mol. Biol. Cell, 12, 221-238. Zagulski, M., Babinska, B., Gromadka, R., Migdalski, A., Rytka, J., Sulicka, J., and Herbert, C. J. (1995). The sequence of 24.3 kb from chromosome X reveals five complete open reading frames, all of which correspond to new genes, and a tandem insertion of a Ty1 transposon. Yeast, 11, 1179-1186. Zhu, Y., Traub, L. M., and Kornfeld, S. (1998). ADP-ribosylation factor 1 transiently activates high-affinity adaptor protein complex AP-1 binding sites on Golgi membranes. Mol. Biol. Cell, 9, 1323-1337.

Chapter 5 LARGE ARF GEFS OF THE GOLGI COMPLEX In search of mechanisms for the cellular effects of BFA PAUL MELANÇON, XINHUA ZHAO, AND TROY K. R. LASELL Department of Cell Biology, University of Alberta, Edmonton, Canada

Abstract:

1.

Class I and class II Arfs localize to the Golgi complex where they activate lipid-modifying enzymes and also promote the recruitment of several coat proteins implicated in the sorting of cargo into budding transport vesicles. Three families of large GEFs spatially and temporally regulate Arf activation in this organelle. GBF1, a BFA-resistant GEF with specificity for Class II Arfs in vitro, localizes to cis-compartments of the Golgi and appears to regulate assembly of the COPI coat. In contrast, BIG1 and BIG2 are BFAsensitive GEFs that concentrate on trans-compartments where they control the recruitment of clathrin adaptor proteins. BRAG1, a representative of the third family, is confined to early Golgi compartments, but unlike GBF1 displays broader Arf substrate specificity. These large GEFs are likely multidomain proteins with complex interactions responsible for regulating activity, membrane association and possibly the selection of downstream effectors. Ongoing efforts in several laboratories focus on the identification of partners for these GEFs.

INTRODUCTION

Dynamic and bi-directional transport pathways link organelles of the central vacuolar system of eukaryotic cells. The efficient and selective transport of proteins between these organelles is made possible by a complex molecular machinery that drives the budding, targeting and fusion of transport vesicles into which cargo and specific address molecules are selectively inserted (Antonny and Schekman, 2001; Spang, 2002). The initial sorting and budding processes are driven by the recruitment of specific coat proteins that are spatially and temporally regulated by small GTPases of the Arf and Sar1p families. The coatomer complex, or COPI, directs retrograde movement of escaped ER proteins from the Golgi and the vesiculo-tubular 101

2 102

MELANÇON, ZHAO, AND LASELL

clusters (VTCs) of the ER-Golgi Intermediate Compartment (ERGIC) that ferry cargo to the Golgi stack; it has also been implicated in transport of cargo across the Golgi stack (Marsh and Howell, 2002; Pelham and Rothman, 2000). In agreement with such diverse functions, COPI localizes to several compartments, with greatest abundance in VTCs and cis-elements of the Golgi stack (Griffiths et al., 1995; Oprins et al., 1993). In contrast, the clathrin coat directs transport of specific cargo to endosomes. Several types of adaptor complexes, or APs (see Chapters 12 and 13), regulate association of clathrin with membranes of the trans-Golgi network and endosomes (Hirst and Robinson, 1998; Robinson and Bonifacino, 2001). An approach to identify and characterize components of this transport machinery has involved elucidating the mechanism of action of the drug Brefeldin A (BFA). BFA is a fungal metabolite that blocks protein secretion and dramatically alters the morphology of several organelles of the secretory pathway (Jackson, 2000). Current thinking is that these effects result from blocking the activation of Arfs by BFA-sensitive GEFs, which in turn prevents assembly of several cytoplasmic coats (Donaldson et al., 1992; Helms and Rothman, 1992; Randazzo et al., 1993). Efforts to identify those cellular targets and describe how BFA disrupts so many aspects of protein traffic led to the characterization of 3 families of large Golgi-associated Arf GEFs termed Golgi-specific BFA resistance Factor 1 (GBF1), BFAInhibited GEFs (BIGs), and BFA-Resistant Arf GEFs (BRAGs). As summarized below, those studies paint a surprisingly more complex picture than was suggested from in vitro studies. Furthermore, they demonstrate that large Arf GEFs all localize to the Golgi complex but have distinct properties that may explain how Arfs can simultaneously regulate the function of so many distinct processes in the Golgi complex. As the properties of yeast Arf GEFs were reviewed in Chapter 4, we will focus on results obtained with mammalian systems.

2.

BFA-RESISTANT CELL LINES: A STARTING POINT

2.1

Resistance to BFA cytotoxicity correlates with stabilization of the COPI coat

Several years ago, we and others set out to develop a genetic approach to identify the cellular target(s) of BFA (Yan et al., 1992; Yan et al., 1994). We generated, by chemical mutagenesis, several mutant lines of Chinese hamster ovary (CHO) cells resistant to BFA in order to better understand the effects of this drug on various organelles. These studies led to the isolation

LARGE ARF GEFS OF THE GOLGI COMPLEX

1033

of 24 CHO mutant lines (termed BFY1-24) that can grow in the presence of BFA concentrations (1-2 µg/ml) 5-10 fold greater than the IC50. These BFY lines displayed organelle-specific resistance to BFA: whereas the Golgi complex maintained its appearance and function at high BFA concentrations, the endosomal system retained sensitivity (Yan et al., 1994). A similar selection procedure by Nakayama and colleagues led to isolation of several BFA resistant CHO lines termed BAR1-30 (Torii et al., 1995). In these mutant cells, COPI showed reduced sensitivity to BFA whereas γ-adaptin was released at BFA concentrations similar to those effective in the parental line. These two studies provided direct evidence for the presence of multiple, organelle-specific targets for BFA. In addition, they demonstrated that the cytotoxicity of BFA correlated with effects on the COPI coat and the Golgi complex and not on events associated with the TGN or endolysosomal systems. We concluded that characterization of factors that alter BFA cytotoxicity would lead to the discovery of proteins involved in regulation of the COPI coat, and the possible identification of those Arf GEFs or effectors impacting cell viability.

2.2

BFY cells express a BFA resistance factor

As a first step towards understanding the mechanism of resistance, we chose some of the most resistant lines for detailed biochemical analysis. BFY1 and BFY2 cells display normal Golgi morphology at a BFA concentration as high as 20 µg/ml and have been the focus of this analysis (Yan et al., 1994; Yan and Melançon, 1994). This work led to the unexpected discovery of a dominant activity that confers BFA resistance to several in vitro assays and is present in both cytosolic and membrane fractions. We then established that this “resistance factor” was clearly distinct from Arfs and coat proteins and acted either at or upstream of the nucleotide exchange reaction that activates Arfs and initiates coat recruitment. The observation that the BFA resistance factor acted in a dominant manner was particularly important since it demonstrated the feasibility of expression cloning as a method for its identification.

3.

GBF1: A GOLGI-SPECIFIC BFA RESISTANCE FACTOR

3.1

Identification of GBF1

To identify the BFA resistance factor, we constructed cDNA libraries from mutant BFY1 and BFY2 cells using an episomal vector that allowed

4 104

MELANÇON, ZHAO, AND LASELL

recovery of stable HEK-293 transformants at high frequency (Claude et al., 1999). Plasmids recovered from populations of BFA-resistant transformants were screened individually for their ability to confer BFA-resistance. This procedure yielded a single plasmid encoding the putative BFA “resistance factor”. This Golgi-specific BFA resistance factor, or GBF1, appears to be the only one that can be recovered by this approach since several subsequent selections with either the BFY1 or BFY2 libraries yielded the same cDNA. Sequence analysis revealed a novel 210 kDa protein with a central sec7 domain (Sec7d) which is now known to catalyze nucleotide exchange on Arfs and is present in all characterized Arf GEFs (see Chapter 2). As a first test of GBF1 function, we prepared Golgi-enriched membrane fractions from cells overexpressing the cDNA and confirmed that those membranes now displayed BFA-resistant Arf-specific GEF activity.

3.2

BFY and parental cells express similar levels of multiple splice forms of GBF1

Two approaches established that BFY cells express wild-type GBF1. First, we constructed a full-length cDNA for human GBF1 and characterized it (Mansour et al., 1998). To our surprise, moderate overexpression of the wild-type protein led to BFA resistance similar to that observed with the GBF1 recovered from mutant CHO cells (S. Mansour and P. Melançon, unpublished observation). In parallel, we cloned and sequenced several fulllength cDNAs from a library prepared from the parental CHO line (Zhao et al., 2002a). Analysis of these GBF1 transcripts revealed no point mutations but did identify three nucleotide positions (at bp 1010, 1864 and 4479 in the open reading frame) displaying frequent in-frame deletions or insertions. A combination of genomic DNA sequencing, RT-PCR and biological assays demonstrated that (1) the small variations were consistent with the use of alternative splicing sites, (2) the splice variants were made at similar frequencies in wild-type and BFA-resistant cells, and, (3) those variants displayed activity identical to the original GBF1 cDNA (Zhao et al., 2002a). These results led us to conclude that contrary to expectations, GBF1 was not mutated in BFY cells and that its isolation in our screen resulted from its intrinsic ability to induce BFA resistance when overexpressed. Interestingly, overexpression of the BFA-sensitive Geas similarly induces BFA resistance (Peyroche et al., 1999). This property appears unique to the GBF1/Gea subfamily of Arf GEFs: overexpression of BIGs does not reduce BFA sensitivity (see section 6.2 below) while that of ARNOs actually mimics BFA and causes dispersal of the COP1 coat (Franco et al., 1998; Monier et al., 1998).

LARGE ARF GEFS OF THE GOLGI COMPLEX

3.3

1055

GBF1 is a BFA resistant GEF with specificity for Class II Arfs in vitro

The guanine nucleotide exchange activity of GBF1 was first confirmed by using a His6-tagged form of GBF1 expressed in HEK-293 cells (Claude et al., 1999). As expected, GBF1 displayed a robust and BFA-resistant GEF activity. However, whereas GBF1 activated both Arf1 and Arf5 at µM concentrations of MgCl2, under more physiological (> 1 mM) MgCl2 concentrations GEF activity was only observed with Arf5. The Nakayama group also cloned GBF1 and recently confirmed this result using crude, recombinant GBF1 and Arf (Kawamoto et al., 2002). Interestingly, overexpression of GBF1 was found to promote recruitment of both Class I and Class II Arfs in intact cells, suggesting that the substrate specificity of GBF1 may be different in vivo.

Figure 1. Alignment of Sec7d of several GEFs showing residues of motif 2 implicated in sensitivity to BFA. Residues found to strongly correlate with BFA resistance and sensitivity are shown on first and second line, respectively. Positions marked with asterisk correspond to mutations shown to affect BFA sensitivity.

The BFA resistance observed in vitro is actually surprising considering the sequence of GBF1. Mutational analysis and comparison of several Sec7d proteins identified several residues involved in BFA binding (Peyroche et al., 1999; Sata et al., 1999). These residues are located in or near a conserved hydrophobic helix and participate in the flexibility of the C-terminal region of the Sec7d that appears critical for formation of the complex with BFA (Renault et al., 2002). The consensus motif for BFA resistant proteins is FAVIMLNTSL-X7-KP and that for BFA sensitive proteins is YSIIMLTTDL-X7-KM (see Figure 1). The hydroxyl moiety of

6 106

MELANÇON, ZHAO, AND LASELL

the tyrosine residue may be particularly crucial as it makes a key contact in the BFA-Arf•GDP-GEF complex. Not surprisingly, the FY mutation is sufficient to convert the Arf GEF activity of the Sec7d of ARNO from BFAresistant to BFA-sensitive (see Chapter 4 for details). The fact that motif 2 of GBF1 (YAVIMLNTDQ-X10-PM) contains the critical tyrosine would suggest a BFA-sensitive activity. The presence of three additional amino acid residues within motif 2, relative to the consensus sequence (see Figure 1), may account for the BFA resistance and unique Arf specificity of GBF1.

3.4

GBF1 associates with Golgi membranes

As observed for most Arf GEFs, GBF1 is a peripheral membrane protein (Claude et al., 1999). GBF1 is readily detected on membranes (Kawamoto et al., 2002), however, careful analysis of proportional amounts of cytosol and membrane fractions established that the microsome fraction corresponds to <10% of the total GBF1 in the post-nuclear supernatant (Claude et al., 1999). A proteomics-based analysis of Golgi-associated proteins identified GBF1 in the detergent phase of a TX-114 extraction of a highly stacked rat liver Golgi fraction (Bell et al., 2001). These results suggest that a fraction of membrane-associated GBF1 may be tightly bound to an integral membrane protein of the Golgi complex. This conclusion is supported by analysis of the membrane-association of GNOM, one of several orthologs of GBF1 encoded by the A. thaliana genome. GNOM is a BFA-sensitive GEF that can use Arf1 as a substrate and can complement loss of Gea function in a ¨gea2/gea1-19 temperature-sensitive strain (Steinmann et al., 1999). This study revealed that GNOM was released from microsomes by TX-100 and urea, but not by incubation with buffers of high salt concentration or alkaline pH. It may be important to note that treatment with BFA causes a significant increase in the extent of GNOM membrane association (Steinmann et al., 1999). GBF1 co-localized with well-characterized Golgi markers to perinuclear structures, as determined by indirect immunofluorescence using antiserum raised against a C-terminal peptide (Claude et al., 1999). This result was confirmed using in situ immunogold electron microscopy on rat hepatocyte cryosections. Quantitative analysis of these images established that even though GBF1 was present at the highest density on Golgi membranes, the majority of it was associated with a smooth tubular compartment adjacent to Golgi stacks that correspond to the ERGIC in this tissue (Claude et al., 1999). Subsequent analysis of rat parotid cryosections by Nakayama and colleagues confirmed this clear enrichment in vesicular and tubular structures in the vicinity of the Golgi stack (Kawamoto et al., 2002). In both studies, gold particles were also observed on ER-like elements suggesting

LARGE ARF GEFS OF THE GOLGI COMPLEX

1077

that GBF1 may cycle between the Golgi complex and the ER. Recent observations with a new antibody (9D2), raised against a fragment of GBF1, support the possibility that GBF1 functions on membranes in transit to the Golgi complex. Contrary to what is observed with most of our sera (e.g., see Figure 2A), this new antiserum reveals not only the expected juxtanuclear pattern but also shows a number of peripheral puncta (arrows in Figure 2B).

Figure 2. GBF1 localizes to the Golgi complex (A; 9D5) and peripheral puncta (B; 9D2).

3.5

GBF1: a homodimer of unknown domain structure

As is the case for large GEFs that regulate other small GTPases, such as SOS (Hall et al., 2002) and Dbl (Hoffman and Cerione, 2002), GBF1 is likely to be a multidomain protein with complex interactions responsible for regulating its membrane association and activity. Unfortunately, with the exception of the central Sec7d and some proline-rich regions near the Cterminus, the GBF1 sequence does not contain recognizable protein motifs. The only other characterized domain regulates dimerization and was identified through characterization of GNOM. The possibility that GNOM is a homodimer was first suggested by unusual allelic complementation between three classes of GNOM alleles (Busch et al., 1996). Subsequent characterization using a yeast two-hybrid assay confirmed that GNOM is a homodimer and established that the region responsible for dimerization is present at the N-terminus (residues 1 to 246), in the region encoded by the first exon (Grebe et al., 2000). This region is also able to bind cyclophilin-5, a protein with folding activity that may be important for regulation of GNOM dimerization and function. This dimerization and cyclophilinbinding, or DCB domain, is also found in Geas and GBF1 but its function in these proteins has not been confirmed. However, GBF1 and Geas likely function as dimers because other well-characterized Arf GEFs such as BIGs (Yamaji et al., 2000) and ARNOs (Chardin et al., 1996) clearly form dimers.

8 108

MELANÇON, ZHAO, AND LASELL

To identify regions implicated in membrane recruitment, a large number of tagged forms of various N- and C-terminal truncations were constructed. Contrary to what was observed with BIG1 and BIG2 (see section 4 below), this approach proved unsuccessful as no chimeras localized to the Golgi complex (S. Mansour, unpublished observations).

4.

BIGS: GOLGI-LOCALIZED TARGETS OF BFA

To further investigate the mechanism of action of BFA, we cloned and characterized the cDNA of human BFA Inhibited GEF1, or BIG1 (Mansour et al., 1998). This 205 kDa GEF was first isolated from bovine brain extracts by Joel Moss and colleagues on the basis of its ability to promote guanine nucleotide exchange on purified Arfs (Morinaga et al., 1997; Morinaga et al., 1996). It was later shown to co-purify as a heterodimer with another BFA-sensitive Arf GEF with 60% sequence identity termed BIG2 (Togawa et al., 1999; Yamaji et al., 2000).

4.1

BFA is an non-competitive inhibitor of BIGs

Using in vitro Arf GEF assays, we first characterized the activity of a 39 kDa recombinant fragment that spans the central Sec7d of BIG1 (Mansour et al., 1999). Previous work with ARNO (Chardin, 1996) had established that this domain was sufficient to catalyze nucleotide exchange (see Chapter 2). The N- and C-terminal boundaries of the BIG1 fragment were selected on the basis of our experience with numerous fragments of GBF1 and the presence of a sequence motif lying outside the Sec7d, termed the HUS box (homology upstream of Sec7d), that is shared by several large Arf GEFs. This fragment was particularly active and promoted nucleotide exchange on Arfs at enzyme:substrate ratios as low as 1:100 (Mansour et al., 1999). In contrast, full-length forms of BIGs or truncated forms of the yeast Sec7p show much lower specific activity and necessitate near equimolar amounts of enzymes and substrate (Sata et al., 1998; Togawa et al., 1999; Yamaji et al., 2000). More importantly, this high specific activity allowed confirmation that the Sec7d was the direct target of BFA. Kinetic analyses established that BFA did not compete with Arf for interaction with BIG1 but, rather, acted as a non-competitive inhibitor that bound the BIG1-Arf complex with a KI of 7 µM (Mansour et al., 1999). Peyroche et al. (1999) further confirmed that BFA promoted formation of a stable complex between Arf and a BFA-sensitive GEF that could be recovered by sizeexclusion chromatography. Interestingly, GDP remained associated with Arf1 in this complex.

LARGE ARF GEFS OF THE GOLGI COMPLEX

4.2

1099

BIGs: Golgi-localized BFA targets

To investigate the intracellular location of BIG1, we constructed tagged forms of both full-length and various fragments of the protein. HA-tagged full length BIG1 localized to tight ribbon-like juxtanuclear structures that overlapped with the Golgi marker mannosidase II (Mansour et al., 1999). Similar results were subsequently reported for BIG1 and tagged BIG2 (Shinotsuka et al., 2002b; Yamaji et al., 2000). Analysis of several truncated forms established that fragments containing the N-terminal third of BIG1 contained sufficient information to direct localization to the Golgi complex (Mansour et al., 1999). A similar analysis established that a tagged Nterminal fragment containing the first 550 residues of BIG2 likewise localized to the Golgi complex (Zhao et al., 2002b). It is not clear at this point whether this domain contains targeting information or is responsible for dimerization with a properly targeted endogenous BIG.

5.

BRAGS: A NOVEL FAMILY OF BFA-RESISTANT ARF GEFS

BLAST searches of the human genome with the GBF1 sequence identified 13 human proteins with a Sec7d, including three previously uncharacterized, large, putative Arf GEFS encoded by partial human cDNAs (KIAA0522, KIAA0763 and KIAA1110). Phylogenetic analysis of metazoan Sec7d proteins encoded by the genomes of H. sapiens (n=13), D. melanogaster (n=5) and C. elegans (n=5) revealed that these three large proteins define a fifth subclass of Arf GEFs (see Figure 3). Interestingly, whereas humans express several members of each subfamily, the D. melanogaster and C. elegans genomes encode only one member of each of the 5 groups. Surprisingly, as is also true for ARNOs and EFA6s, the genomes of S. cerevisiae, S. pombe or A. thaliana do not encode a member of this family. Furthermore, these Arf GEFs are unique among the large Arf GEFs in containing a pleckstrin homology (PH) domain, located approximately 55 residues downstream of the Sec7d.

10 110

MELANÇON, ZHAO, AND LASELL

Figure 3. Phylogenetic analysis of metazoan Sec7d proteins identifies 5 sub-families.

5.1

BRAGs are BFA-resistant Arf GEFs

The Sec7d of KIAA0522 and KIAA0763 were expressed in bacteria as GST fusion proteins. These proteins demonstrated robust Arf GEF activity (Lasell and Melançon, 2001). Of three fragments tested, the most active contained the Sec7d of KIAA0522 and also included the downstream PH domain. This fragment displayed Arf GEF activity towards myristoylated forms of all three classes of Arfs but appeared most active on Class I Arfs. The sequences of the Sec7d of all three members of this family contain 2 key residues (shaded in Figure 1) that strongly correlate with BFA resistance. We confirmed that the recombinant fusion protein was not affected by BFA, under conditions in which the activity of BIG1 measured in parallel experiments was inhibited by more than 80%. These proteins were renamed BRAGs for their BFA-resistant Arf GEF activities (personal communication). J. Moss and colleagues recently reported the characterization of the ORF encoded by KIAA0763/BRAG2 (Someya et al., 2001). A recombinant form of the His6-tagged-100 kDa protein, termed GEP-100, was expressed in insect cells and found to promote nucleotide exchange on recombinant forms

LARGE ARF GEFS OF THE GOLGI COMPLEX

11 111

of all three classes of Arfs, but with significantly greater activity towards Arf6. Furthermore, as expected, this activity was BFA resistant. Previous work established that the PH domain present in members of the ARNO and EFA6 families regulates GEF activity by facilitating recruitment to membranes containing phosphoinositides (Chardin et al., 1996; Macia et al., 2001; Paris et al., 1997). Surprisingly, varying levels of PS, PIP2 or PIP3 between 0 to 50 µM had no impact on GTP loading by GEP-100/BRAG2 on either Arf1 or Arf6 (Someya et al., 2001). The authors concluded that the PH domain does not mediate regulation of GEF activity by phosphoinositides. However, one should note that all nucleotide exchange assays were performed with Arfs lacking the essential myristate and used PS or phosphoinositides bound to BSA rather than as liposomes. Arf activation is coupled to membrane interaction and assays with full-length Arf are usually performed in the presence of vesicles (Franco et al., 1993; Paris et al., 1997). Reactions carried out with non-myristoylated ARFs in absence of liposomes may therefore not reflect the true specificity of the GEF. Furthermore, since the authors did not use liposomes doped with varying levels of phosphoinositides, their results do not exclude regulation of membrane recruitment by the PH domain.

5.2

BRAGs may localize to distinct compartments

Antibodies raised against a recombinant form of the BRAG1 Sec7d localize the endogenous protein to juxtanuclear structures (see Figure 4) that co-stain with the Golgi marker mannosidase II in NRK cells (Lasell and Melançon, 2001). Interestingly, BRAG1 is rapidly displaced from Golgi membranes upon treatment with BFA. This observation is an important distinction because GBF1 and BIGs remain membrane-associated after BFA treatment. As discussed in 7.1 below, this displacement could depend on the PH domain and result from changes in lipid composition. The displacement of BRAG1 from the Golgi could explain some of the effects of BFA on the Golgi complex of animal cells. In contrast, endogenous GEP-100/BRAG2 localizes to coarse puncta near the nucleus that coincide with the early endosomal marker EEA1 in human glioblastoma T98G cells (Someya et al., 2001). Contrary to GBF1/Geas and BIGs that are restricted to the Golgi complex, or the EFA6s/ARNOs that function in the endosomal system, BRAGs may localize to both of these compartments.

12 112

MELANÇON, ZHAO, AND LASELL

Figure 4. Endogenous BRAG1 localizes in NRK cells to cis-elements of the Golgi complex (A), well resolved from trans-elements positive for the clathrin adaptor protein GGA1 (B).

6.

ARF GEFS CONTROL DISTINCT REACTIONS IN THE GOLGI COMPLEX

6.1

GBF1 and BIGs localize to cis- and transcompartments of the Golgi complex, respectively

The results summarized in the previous sections established that three families of Arf GEFs localize to the Golgi complex. To test the possibility that these GEFs have specialized function that may be reflected in distinct localization, we examined their relative distribution using quantitative confocal laser scanning microscopy (Zhao et al., 2002b). Co-staining for endogenous GBF1 and either HA-tagged or endogenous BIG1 revealed clear differences in the staining patterns of these two Arf GEFs. Such separation between GBF1 and BIGs was observed in at least two cell types and with various combinations of endogenous or HA-tagged-forms of BIGs. Further work showed that GBF1 co-localizes with the cis-marker p115 (86%) while BIGs overlap extensively with the trans-Golgi marker TGN38 (83%). This localization of GBF1 is consistent with its relative abundance on vesiculotubular structures detected by immunogold labeling (Claude et al., 1999; Kawamoto et al., 2002). These observations suggest that GBF1 and BIGs activate Arfs in specific locations to regulate different coating reactions. In agreement with this possibility, the COPI coat overlaps to a greater extent with GBF1 (64%) than BIG1 (31%), while clathrin shows limited overlap with BIG1, and virtually none with GBF1. A similar analysis of BRAG1 (see Figure 4) suggests that it localizes preferentially to cis-compartments where it may participate with GBF1 in regulation of the COPI coat. The localization of GBF1 and BIGs to early and late compartments of the secretory pathway, respectively, is supported by their response to various treatments. As expected of a cis-Golgi protein, GBF1 partially redistributes

LARGE ARF GEFS OF THE GOLGI COMPLEX

13 113

from the Golgi complex to peripheral sites following incubation at 15˚C, while it remains membrane-associated and appears in a diffuse, reticular, pattern (consistent with an ER location) in the presence of BFA (Kawamoto et al., 2002; Zhao et al., 2002b). In contrast, the perinuclear localization of BIGs is not altered at 15ºC (Zhao et al., 2002b). Furthermore, BIGs are not released from membranes upon BFA treatment but, instead, redistribute (like TGN38) to a dense collection of fine punctate structures in the perinuclear region. The apparent discrepancy with the conclusion of Vaughan and colleagues, that BFA causes little change in BIGs’ distribution (Yamaji et al., 2000), most likely reflects the slow kinetics of BFA effects on the TGN and the short (10 min) treatment used by those authors (X. Zhao, unpublished data).

6.2

GBF1 and BIGs regulate the recruitment of distinct coats

GBF1 was originally identified by its ability to allow moderately overexpressing cells (2-6 fold) to grow at BFA concentrations that blunt the growth of control cells (Claude et al., 1999). This effect is unique to GBF1 since dramatic overexpression of BIG1 (10-20-fold) has minimal impact on the BFA-sensitivity of transformants (Claude et al., 2003). GBF1 mitigates the effects of BFA on cell growth by stabilizing the COPI coat (Claude et al., 2003; Kawamoto et al., 2002). As expected from its intra-cellular localization, GBF1 overexpression does not protect BFA-induced redistribution of the trans-Golgi marker furin (Claude et al., 2003). In contrast, overexpression of BIG1 does not protect either Arf or COPI from BFA-induced dissociation. The Golgi disperses with normal kinetics in cells overexpressing BIG1 (Claude et al., 2003). Similarly, overexpression of BIG2 has no protective effects on COPI or Golgi organization (Shinotsuka et al., 2002b). However, overexpression of BIG2 does antagonize some of the effects of BFA, but only those related to the trans-Golgi. For example, overexpression of BIG2 prevents the BFAinduced release of the clathrin adaptor complex, AP-1, to cytosol. Furthermore, a dominant-negative mutant of BIG2 lacking a critical Glu residue in its Sec7d does not interfere with COPI recruitment but specifically affects membrane traffic from the TGN through inhibition of clathrin adaptors such as AP-1 and GGA (Shinotsuka et al., 2002a). Interestingly, overexpression of BIG2 does not prevent tubulation of the TGN; the resulting tubules appear as queues of spherical swellings labeled with BIG2, AP-1 and low levels of Arf1 (Shinotsuka et al., 2002a). Preliminary results show that endogenous BIG1 does not appear in BFAinduced TGN tubules (Figure 5). The presence of HA-BIG2 in TGN-

14 114

MELANÇON, ZHAO, AND LASELL

derived tubules could result from its overexpression. Alternatively, BIG2 homodimers may localize differently and have functions distinct from BIG1homo and heterodimers within the TGN.

Figure 5. BFA-induced tubules from the TGN contain TGN38 (A) but lack BIG1 (B).

6.3

GBF1 and BIG1 have distinct but interrelated functions in the Golgi complex

Analysis of Arf GEF expression levels in single and double GBF1/BIG1 transformants revealed unexpected relationships between these two GEF families (Claude et al., 2003). Expression of any of several GBF1 splice variants substantially decreased endogenous BIG1 protein levels. This reduction in BIG1 levels was even observed when BIG1 synthesis was driven from a plasmid’s viral promoter suggesting effects on translation or stability rather than transcription. In contrast, overexpression of BIG1 had no impact on GBF1 levels. However, BIG1 overexpression clearly interfered with GBF1 function: in double transformants, GBF1 no longer protected COPI from BFA-induced dissociation, and no longer caused an increase of the LD50 in cytotoxicity assays. These results underscore the poorly understood and probably complex functional relations between Arf GEF families at the Golgi complex.

7.

EXPLANATION FOR BFA EFFECTS ON THE GOLGI COMPLEX REMAINS ELUSIVE

7.1

BARS-50 and BFA effects on early compartments of the Golgi complex

It is generally assumed that the effects of BFA result from inhibition of an Arf GEF. However, as summarized above, BIGs are the only mammalian

LARGE ARF GEFS OF THE GOLGI COMPLEX

15 115

Arf GEF clearly inhibited by BFA and they localize and function within trans-elements of the Golgi complex. How then can one explain the effects of BFA on early steps in protein traffic? Although the precise mechanism remains unclear, several possibilities are discussed here. For example, BARS-50 is a lysophosphatidate acyl transferase (thus, like phospholipase D produces phosphatidic acid) that becomes ADP-ribosylated in response to BFA treatment of cells (Weigert et al., 1999). BARS-50 also promotes scission of COPI vesicles in in vitro assays (Spano et al., 1999; Weigert et al., 1999). Thus, BFA-induced ADP-ribosylation of BARS-50 may alter its enzymatic activity and result in alterations in the lipid composition of the Golgi stacks and thereby interfere, directly or indirectly, with COPI recruitment or prevent the release of VTCs from ER exit sites. Changes in lipid composition could also be responsible for the rapid loss of BRAG1 from Golgi membranes observed after BFA treatment. Some of the effects of BFA on the Golgi could therefore result not from blocking a cis-acting GEF but from interfering with its recruitment. We also cannot exclude the possibility that GBF1 is BFA sensitive, under physiological conditions.

7.2

The mechanism of BFA resistance in BFY cells remains unknown

Several lines of evidence established that GBF1 was not the “resistance factor” originally detected in extracts of BFY1 and BFY2 cells. Comparison of GBF1 levels in wild-type and mutant cells first confirmed that BFA resistance did not result from GBF1 overexpression (Claude et al., 1999). As described in section 3.2 above, subsequent analysis eliminated the possibility that mutant lines expressed unusual splice forms of GBF1 with differential effects on BFA sensitivity (Zhao et al., 2002a). These observations prompted us to examine if a mutation in the Sec7d of BIGs might be responsible for the reduced BFA-sensitivity of mutant lines. Sequencing of several RT-PCR products encoding the Sec7d of BIG1 and BIG2 established that mutant cells expressed wild-type forms of both of these BFA-sensitive Arf GEFs (personal communication). The nature of the mutation in our BFY lines therefore remains unknown and open to investigation. The mechanism responsible for the “natural” BFA resistance of cell lines such as Ptk1 (Ktistakis et al., 1991) and MDCK (Hunziker et al., 1991) remains similarly unknown. Clearly the cellular effects of BFA are more complex than originally anticipated, and further investigation of the BFY lines may shed new light on regulation of Arf activation in early Golgi compartments.

16 116

8.

MELANÇON, ZHAO, AND LASELL

CONCLUDING REMARKS: ARF GEFS AS MOLECULAR SCAFFOLDS

The work summarized above provides convincing evidence that the Arf GEFs that function at the Golgi complex display both substrate preference and distinct localization. These properties may help define which Arf is activated where. What remains largely unexplored, however, is the mechanism(s) that controls the choice of effector targeted by the activated Arf to elicit a specific cellular response. As discussed in several other chapters, Arf•GTP promotes not only membrane recruitment of COPI and the several classes of clathrin adaptor proteins; it also regulates several noncoat effectors, including the membrane remodeling enzymes phospholipase D and PI(4)P 5-kinase (Brown et al., 1993; Godi et al., 1999; Roth, 1999), as well as proteins of poorly defined function such as Arfaptins and Arfophilin (Kanoh et al., 1997; Shin et al., 1999; Van Valkenburgh et al., 2001; Williger et al., 1999). By acting as transient molecular scaffolds, the large Arf GEFs could provide some of the required specificity to direct specific interactions between GEFs and downstream effectors in a spatially and temporally regulated manner. Ongoing efforts in several laboratories to identify partners for Arfs and their regulators will in time test this hypothesis.

REFERENCES Antonny, B., and Schekman, R. (2001). ER export: public transportation by the COPII coach. Curr. Opin. Cell Biol., 13, 438-443. Bell, A. W., Ward, M. A., Blackstock, W. P., Freeman, H. N., Choudhary, J. S., Lewis, A. P., Chotai, D., Fazel, A., Gushue, J. N., Paiement, J., Palcy, S., Chevet, E., Lafreniere-Roula, M., Solari, R., Thomas, D. Y., Rowley, A., and Bergeron, J. J. (2001). Proteomics characterization of abundant Golgi membrane proteins. J. Biol. Chem., 276, 5152-5165. Brown, H. A., Gutowsk, S., Moomaw, R. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity. Cell, 75, 1137-1144. Busch, M., Mayer, U., and Jürgens, G. (1996). Molecular analysis of the Arabidopsis patternformation gene GNOM: gene structure and intragenic complementation. Molec. Gen. Genet., 250, 681-691. Chardin, P., Paris, S., Antonny, B., Robineau, S., Beraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996). A human exchange factor for ARF contains Sec7- and pleckstrinhomology domains. Nature, 384, 481-484. Claude, A., Zhao, B. P., Kuziemsky, C. E., Dahan, S., Berger, S. J., Yan, J. P., Armold, A. D., Sullivan, E. M., and Melançon, P. (1999). GBF1: A novel Golgi-associated BFA-resistant guanine nucleotide exchange factor that displays specificity for ADP-ribosylation factor 5. J. Cell Biol., 146, 71-84.

LARGE ARF GEFS OF THE GOLGI COMPLEX

17 117

Claude A, Zhao BP, Melancon P. (2003) Characterization of alternatively spliced and truncated forms of the Arf guanine nucleotide exchange factor GBF1 defines regions important for activity. Biochem Biophys Res Commun 303(1): 160-9. Donaldson, J. G., Finazzi, D., and Klausner, R. D. (1992). Brefeldin A inhibits Golgi membrane catalyzed exchange of guanine nucleotide onto ARF protein. Nature, 360, 350352. Franco, M., Boretto, J., Robineau, S., Monier, S., Goud, B., Chardin, P., and Chavrier, P. (1998). ARNO3, a Sec7-domain guanine nucleotide exchange factor for ADP ribosylation factor 1, is involved in the control of Golgi structure and function. Proc. Natl. Acad. Sci., USA, 95, 9926-9931. Franco, M., Chardin, P., Chabre, M., and Paris, S. (1993). Myristoylation is not required for GTP-dependent binding of ADP-ribosylation factor ARF1 to phospholipids. J. Biol. Chem., 268, 24531-24534. Godi, A., Pertile, P., Meyers, R., Marra, P., Di Tullio, G., Iurisci, C., Luini, A., Corda, D., and De Matteis, M. A. (1999). ARF mediates recruitment of PtdIns-4-OH kinase-beta and stimulates synthesis of PtdIns(4,5)P2 on the Golgi complex [see comments]. Nat. Cell Biol., 1, 280-287. Grebe, M., Gadea, J., Steinmann, T., Kientz, M., Rahfeld, J. U., Salchert, K., Koncz, C., and Jurgens, G. (2000). A conserved domain of the arabidopsis GNOM protein mediates subunit interaction and cyclophilin 5 binding. Plant Cell, 12, 343-356. Griffiths, G., Pepperkok, R., Locker, J. K., and Krei, T. E. (1995). Immunocytochemical localization of beta-COP to the ER-Golgi boundary and the TGN. J. Cell Sci., 108, 28392856. Hall, B. E., Yang, S. S., and Bar-Sagi, D. (2002). Autoinhibition of Sos by intramolecular interactions. Front. Biosci., 7, d288-d294. Helms, J. B., and Rothman, J. E. (1992). Inhibition by brefeldin A of a Golgi membrane enzyme that catalyzes exchange of guanine nucleotide bound to ARF. Nature, 360, 352354. Hirst, J., and Robinson, M. S. (1998). Clathrin and adaptors. Biochim. Biophys. Acta, 1404, 173-193. Hoffman, G. R., and Cerione, R. A. (2002). Signaling to the Rho GTPases: networking with the DH domain. FEBS Lett., 513, 85-91. Hunziker, W., Whitney, J. A., and Mellman, I. (1991). Selective inhibition of transcytosis by brefeldin A in MDCK cells. Cell, 67, 617-627. Jackson, C. L. (2000). Brefeldin A: Revealing the fundamental principles governing membrane dynamics and protein transport. In H. A. Fuller (Ed.), Fusion of biological membranes and related problems, Vol. 34 (pp. 233-272). New York, Kluwer Academic. Kanoh, H., Williger, B. T., and Exton, J. H. (1997). Arfaptin 1, a putative cytosolic target protein of ADP-ribosylation factor, is recruited to Golgi membranes. J. Biol. Chem., 272, 5421-5429. Kawamoto, K., Yoshida, Y., Tamaki, H., Torii, S., Shinotsuka, C., Yamashina, S., and Nakayama, K. (2002). GBF1, a guanine nucleotide exchange factor for ADP-ribosylation factors, is localized to the cis-Golgi and involved in membrane association of the COPI coat. Traffic 3, 483-495. Ktistakis, N. T., Roth, M. G., and Bloom, G. S. (1991). PtK1 cells contain a nondiffusible, dominant factor that makes the Golgi apparatus resistant to brefeldin A. J. Cell Biol., 113, 1009-1023. Lasell, T. K., and Melançon, P. (2001). Characterization of a novel family of large ADPribosylation factor specific guanine nucleotide exchange factors. FASEB J., 15, A1176.

18 118

MELANÇON, ZHAO, AND LASELL

Macia, E., Chabre, M., and Franco, M. (2001). Specificities for the small g proteins arf1 and arf6 of the guanine nucleotide exchange factors arno and efa6. J. Biol. Chem., 276, 2492524930. Mansour, S. J., Herbrick, J. A., Scherer, S. W., and Melançon, P. (1998). Human GBF1 is a ubiquitously expressed gene of the sec7 domain family mapping to 10q24 [In Process Citation]. Genomics, 54, 323-327. Mansour, S. J., Skaug, J., Zhao, X. H., Giordano, J., Scherer, S. W., and Melançon, P. (1999). p200 ARF-GEP1: a Golgi-localized guanine nucleotide exchange protein whose Sec7 domain is targeted by the drug brefeldin A. Proc. Natl. Acad. Sci., USA, 96, 7968-7973. Marsh, B. J., and Howell, K. E. (2002). The mammalian Golgi - complex debates. Nature Rev. Cell Biol., 3, 789-795. Monier, S., Chardin, P., Robineau, S., and Goud, B. (1998). Overexpression of the ARF1 exchange factor ARNO inhibits the early secretory pathway and causes the disassembly of the Golgi complex [In Process Citation]. J. Cell Sci., 111, 3427-3436. Morinaga, N., Moss, J., and Vaughan, M. (1997). Cloning and expression of a cDNA encoding a bovine brain brefeldin A-sensitive guanine nucleotide-exchange protein for ADP-ribosylation factor. Proc. Natl. Acad. Sci., USA, 94, 12926-12931. Morinaga, N., Tsai, S. C., Moss, J., and Vaughan, M. (1996). Isolation of a brefeldin Ainhibited guanine nucleotide-exchange protein for ADP ribosylation factor (ARF) 1 and ARF3 that contains a Sec7-like domain. Proc. Natl. Acad. Sci., USA, 93, 12856-12860. Oprins, A., Duden, R., Kreis, T. E., Geuze, H. J., and Slot, J. W. (1993). Beta-COP localizes mainly to the cis-Golgi side in exocrine pancreas. J. Cell Biol., 121, 49-59. Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997). Role of protein-phospholipid interactions in the activation of ARF1 by the guanine nucleotide exchange factor Arno. J. Biol. Chem., 272, 22221-22226. Pelham, H. R., and Rothman, J. E. (2000). The debate about transport in the Golgi – two sides of the same coin? Cell, 102, 713-719. Peyroche, A., Antonny, B., Robineau, S., Acker, J., Cherfils, J., and Jackson, C. L. (1999). Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell, 3, 275-285. Randazzo, P. A., Yang, Y. C., Rulka, C., and Kahn, R. A. (1993). Activation of ADPribosylation factor by Golgi membranes. J. Biol. Chem., 268, 9555-9563. Renault, L., Christova, P., Guibert, B., Pasqualato, S., and Cherfils, J. (2002). Mechanism of domain closure of Sec7 domains and role in BFA sensitivity. Biochemistry, 41, 36053612. Robinson, M. S., and Bonifacino, J. S. (2001). Adaptor-related proteins. Curr. Opin. Cell Biol., 13, 444-453. Roth, M. G., (1999). Lipid regulators of membrane traffic through the Golgi complex. Trends Cell Biol., 9, 174-179. Sata, M., Donaldson, J. G., Moss, J., and Vaughan, M. (1998). Brefeldin A-inhibited guanine nucleotide-exchange activity of Sec7 domain from yeast Sec7 with yeast and mammalian ADP ribosylation factors. Proc. Natl. Acad. Sci., USA, 95, 4204-4208. Sata, M., Moss, J., and Vaughan, M. (1999). Structural basis for the inhibitory effect of brefeldin A on guanine nucleotide-exchange proteins for ADP-ribosylation factors. Proc. Natl. Acad. Sci., USA, 96, 2752-2757. Shin, O. H., Ross, A. H., Mihai, I., and Exton, J. H. (1999). Identification of arfophilin, a target protein for GTP-bound class II ADP-ribosylation factors. J. Biol. Chem., 274, 36609-36615. Shinotsuka, C., Waguri, S., Wakasugi, M., Uchiyama, Y., and Nakayama, K. (2002a). Dominant-negative mutant of BIG2, an ARF-guanine nucleotide exchange factor,

LARGE ARF GEFS OF THE GOLGI COMPLEX

19 119

specifically affects membrane trafficking from the trans-Golgi network through inhibiting membrane association of AP-1 and GGA coat proteins. Biochem. Biophys. Res. Comm., 294, 254-260. Shinotsuka, C., Yoshida, Y., Kawamoto, K., Takatsu, H., and Nakayama, K. (2002b). Overexpression of an ADP-ribosylation factor-guanine nucleotide exchange factor, BIG2, uncouples brefeldin A-induced adaptor protein-1 coat dissociation and membrane tubulation. J. Biol. Chem., 277, 9468-9473. Someya, A., Sata, M., Takeda, K., Pacheco-Rodriguez, G., Ferrans, V. J., Moss, J., and Vaughan, M. (2001). ARF-GEP100, a guanine nucleotide-exchange protein for ADPribosylation factor 6. Proc. Natl. Acad. Sci., USA, 98, 2413-2418. Spang, A. (2002). ARF1 regulatory factors and COPI vesicle formation. Curr. Op. Cell Biol., 14, 423-27. Spano, S., Silletta, M. G., Colanzi, A., Alberti, S., Fiucci, G., Valente, C., Fusella, A., Salmona, M., Mironov, A., Luini, A., Corda, D., and Spanfo, S. (1999). Molecular cloning and functional characterization of brefeldin A-ADP- ribosylated substrate. A novel protein involved in the maintenance of the Golgi structure. [Published erratum appears in J. Biol. Chem., 274, 25188 (1999).] J. Biol. Chem., 274, 17705-17710. Steinmann, T., Geldner, N., Grebe, M., Mangold, S., Jackson, C. L., Paris, S., Galweiler. L., Palme, K., and Jurgens, G. (1999). Coordinated polar localization of auxin efflux carrier PIN1 by GNOM ARF GEF. Science, 286, 316-318. Togawa, A., Morinaga, N., Ogasawara, M., Moss, J., and Vaughan, M. (1999). Purification and cloning of a brefeldin A-inhibited guanine nucleotide-exchange protein for ADPribosylation factors. J. Biol. Chem., 274, 12308-12315. Torii, S., Banno, T., Watanabe, T., Ikehara, Y., Murakami, K., and Nakayama, K. (1995). Cytotoxicity of brefeldin A correlates with its inhibitory effect on membrane binding of COP coat proteins. J. Biol. Chem., 270, 11574-11580. Van Valkenburgh, H., Shern, J. F., Sharer, J. D., Zhu, X., and Kahn, R. A. (2001). ADPribosylation factors (ARFs) and ARF-like 1 (ARL1) have both specific and shared effectors: characterizing ARL1-binding proteins. J. Biol. Chem., 276, 22826-22837. Weigert, R., Silletta, M. G., Spano, S., Turacchio, G., Cericola, C., Colanzi, A., Senatore, S., Mancini, R., Polishchuk, E. V., Salmona, M., Facchiano, F., Burger, K. N., Mironov, A., Luini, A., and Corda, D. (1999). CtBP/BARS induces fission of Golgi membranes by acylating lysophosphatidic acid. Nature, 402, 429-433. Williger, B. T., Provost, J. J., Ho, W. T., Milstine, J., and Exton, J. H. (1999). Arfaptin 1 forms a complex with ADP-ribosylation factor and inhibits phospholipase D. FEBS Lett. 454, 85-89. Yamaji, R., Adamik, R., Takeda, K., Togawa, A., Pacheco-Rodriguez, G., Ferrans, V. J. Moss, J., and Vaughan, M. (2000). Identification and localization of two brefeldin Ainhibited guanine nucleotide-exchange proteins for ADP-ribosylation factors in a macromolecular complex. Proc. Natl. Acad. Sci., USA, 97, 2567-2572. Yan, J. P., Beebe, L., Melançon, P. (1992). Isolation and characterization of Chinese hamster ovary cell lines resistant to brefeldin A. Mol. Biol. Cell, 3, 118a. Yan, J. P., Colon, M. E., Beebe, L. A., and Melançon, P. (1994). Isolation and characterization of mutant CHO cell lines with compartment-specific resistance to brefeldin A. J. Cell Biol., 126, 65-75. Yan, J. P., and Melançon, P. (1994). Characterization of a soluble BFA-resistance factor which regulates association of coatomer to Golgi membranes. FASEB J., 8, 1458. Zhao, X., Lasell, T. K., and Melançon, P. (2002b). Localization of large ADP-ribosylation factor-guanine nucleotide exchange factors to different Golgi compartments: evidence for distinct functions in protein traffic. Mol. Biol. Cell, 13, 119-133.

Chapter 6 BIG1 AND BIG2: BREFELDIN A-INHIBITED EXCHANGE FACTORS FOR ARFS Actions of brefeldin A G. PACHECO-RODRIGUEZ, J. MOSS, AND M. VAUGHAN Pulmonary-Critical Care Medicine Branch, NHLBI, NIH, Bethesda, USA

Abstract:

1.

Brefeldin A-inhibited guanine nucleotide exchange factor 1 (BIG1) and BIG2 were purified as components of a ~670 kDa protein complex. BIG1 and BIG2 partially colocalize with the Golgi 58 kDa protein and Ȗ-adaptin, a component of the AP-1 complex. The distribution of BIG1 in the TGN is distinct from that of GBF-1, a brefeldin A (BFA)-resistant GEF that was confined to the cisGolgi. Recently reported studies implicate BIG2 in the transport of lysosomal enzymes between the TGN and late endosomes. Recognition that BFA inhibition of BIG1 and Gea1 is non-competitive was the key to understanding its mechanism of action. [3H]BFA binding was used to define structural requirements for the formation of a BFA-Arf-Sec7 domain complex. A recently published synthesis of structure-function information includes a valuable model of the Arf-Sec7 domain interaction and the mechanism of BFA inhibition.

INTRODUCTION

Brefeldin A (BFA) is a fungal, fatty acid metabolite described over 30 years ago as an anti-viral agent (Tamura et al., 1968). Later, investigation of its effect on protein synthesis in cultured rat hepatocytes revealed that protein secretion was completely inhibited by BFA at concentrations that were without effect on protein synthesis (Misumi et al., 1986). This and subsequent work demonstrated that BFA reversibly blocked transport of protein from the ER to Golgi and caused apparent disintegration of Golgi structure (Fujiwara et al., 1988). Only three years later, studies of BFA action by several groups had resulted in significant clarification of the processes of ER-Golgi transport, anterograde and retrograde (Orci et al., 1991). 121

2 122

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

Since Palade (1975) described a vesicular transport process for protein secretion, the use of cell-free systems to characterize biochemically and microscopically individual steps in membrane traffic pathways has accelerated the recognition that analogous molecular mechanisms and machinery are employed at multiple sites in all eukaryotic cells. Present understanding of these processes is based on work of many investigators including major contributions, experimental and conceptual, from the Rothman laboratory. In the Rothman model (Rothman, 1994), inactive, cytosolic Arf•GDP interacts with a guanine nucleotide-exchange protein (GEF), which catalyzes the replacement of bound GDP with GTP generating Arf•GTP that moves to a Golgi membrane, where it recruits coatomer (a cytosolic complex of seven coat proteins) to initiate vesicle formation. Sequential accumulation of Arf•GTP and coatomer results in deformation of the membrane to form a bud. By fusion of the donor membrane at the base of the bud (fission), a transport vesicle is released. Demonstration of bud and vesicle formation from lipid model membranes (liposomes) later confirmed that the only additions required were Arf•GTP and coatomer (Spang et al., 1998). After the arrival of a coated vesicle at its target membrane and before membrane fusion can complete the transport process, release of coatomer must occur. This was believed to be triggered by the hydrolysis of Arfbound GTP, followed sequentially by dissociation of Arf•GDP and coatomer. Earlier this year, a publication by Lippincott-Schwartz and coworkers provided valuable new insight into the dynamics of Arf and coatomer interactions in vesicle formation (Presley et al., 2002). Based on detailed quantification and comparison of movements of fluorescently labeled Arf1 and ß-COP in living cells, they established clearly that inactivation of Arf and release of Arf•GDP is required for release of coatomer. In the cell, however, dissociation of Arf from Golgi membranes was much more rapid than that of coatomer, i.e., release of Arf did not “trigger” release of coatomer. As was pointed out, the continuing Arfdependent recruitment and independently regulated release of coatomer during vesicle formation is consistent with the critical functions of coatomer in the modification of membrane morphology (composition) and the assembly/accumulation of appropriate cargo through direct and indirect interactions with phospholipids and other proteins (Presley et al., 2002). BFA was, in fact, employed to advantage in these studies to inhibit vesicle formation from Golgi membranes. In 1992, two groups had reported simultaneously that Golgi membranes accelerated in vitro binding of GTP to Arf, and this GEF activity was inhibited by BFA (Donaldson et al., 1992; Helms and Rothman, 1992). Purification of a BFA-inhibited GEF was finally reported in 1996 (Morinaga

BIGS: BFA-SENSITIVE ARF GEFS

1233

et al., 1996). That protein, p200 (later named BIG1), and a second BFAinhibited GEF (p190 or BIG2) were present together in a ~670 kDa complex (Morinaga et al., 1996; Togawa et al., 1999). Although many effects of BFA on cells are due to its interference with Arf activation via GEF inhibition, BFA also alters Golgi structures by stimulating the ADP-ribosylation of a 50 kDa protein termed BARS, for BFA-ADP-ribosylated substrate (Spanfó et al., 1999). A 38 kDa glyceraldehyde-3-phosphate dehydrogenase is also modified by apparently the same BFA-stimulated ADP-ribosyltransferase. Comparisons of effects of a number of inhibitors of bacterial ADP-ribosyltransferases revealed parallel effects of coumarin and quinone class compounds on the ability of BFA to influence Golgi morphology and to inhibit ADP-ribosylation in vitro. This result is consistent with the involvement of an ADP-ribosylated protein in the disruption of Golgi architecture (Weigert et al., 1997). BFAinduced dissociation of coatomer from Golgi membranes, which results from its inhibition of Arf activation, was unaffected by prevention of the protein ADP-ribosylation and the perturbation of Golgi structure (Mironov et al., 1997). These observations implicated the ADP-ribosylated BARS in Golgi disintegration, and were consonant with the notion that this effect of BFA does not result from its inhibition of Arf activation. With the demonstration that BARS is a lysophosphatidic acid:acyl-CoA transferase, it was suggested that this activity (or the phosphatidic acid that it produces) is required for fission of the Golgi tubular compartments/elements, which in its absence become massively enlarged/elongated as vesicle formation ceases (Weigert et al., 1999). It is notable that endophilin, a protein that interacts with dynamin and is essential for endocytosis, is also a lysophosphatidic acid:acyl-CoA transferase (Schmidt et al., 1999). Both groups suggested similar mechanisms for the effects of phosphatidic acid on membrane structure at the site of vesicle fission. It is of interest that BARS is very similar in structure to CtBP1 and CtBP2, proteins involved in transcription, which interact with the Cterminus of the adenovirus transforming protein E1A. As those proteins were also substrates for BFA-stimulated ADP-ribosylation, it was suggested that BARS be designated CtBP3, and that dual roles for these proteins in the nucleus and in Golgi be considered (Spanfó et al., 1999).

2.

STRUCTURE OF BIG1 AND BIG2

The human BIG1 gene was mapped to chromosome 8q13 (Mansour et al., 1999) and BIG2 to 20q13.13. cDNA clones for bovine BIG1 (Morinaga et al., 1997), human BIG1 (Togawa et al., 1999; Mansour et al., 1999), and

4 124

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

human BIG2 (Togawa et al., 1999; Shinotsuka et al., 2002a) have been reported, BIG1 and BIG2 proteins and mRNAs are present in many bovine and human tissues (Morinaga et al., 1997; Togawa et al., 1999; Mansour et al., 1999). BIG1 and BIG2 were initially purified together in a ~670 kDa protein complex (Morinaga et al., 1996). Amino acid sequences of peptides from BIG1 resembled portions of the yeast Sec7 protein, an Arf GEF involved in protein secretion (Franzusoff and Schekman, 1989). BIG1 and BIG2 were later cloned (Morinaga et al., 1997; Togawa et al., 1999; Mansour et al., 1999; Shinotsuka et al., 2002a), and found to contain Sec7 domains of ~200 amino acids that are 50% identical to the yeast Sec7 domain and 90% identical to each other. The Sec7 domain of yeast Sec7 synthesized as a recombinant protein exhibited GEF activity that was inhibited by BFA (Sata et al., 1998), as did the Sec7 domains of BIG1 (Morinaga et al., 1999), BIG2 (Togawa et al., 1999), and Gea1 (Peyroche et al., 1999). BFA analogues that had not inhibited the GEF activity of Golgi membranes (Donaldson et al., 1992) also failed to inhibit the Sec7 domain of yeast Sec7 (Sata et al., 1998). Until very recently, the only functional region identified in these proteins, was the Sec7 domain, initially recognized in the Saccharomyces cerevisiae Sec7 protein (Achstetter et al., 1988). Sec7 domain proteins, like Arfs, are ubiquitous in eukaryotes. Structures of the BIG1 and BIG2 Sec7 domains have not been reported, but those for other Sec7 domains are known, including those for ARNO or cytohesin-2 (Mossesova et al., 1998; Cherfils et al., 1998), cytohesin-1 (Betz et al., 1998), and Gea2, a BFAinhibited GEF from yeast (Goldberg, 1998; Renault et al., 2002). The Sec7 domain is formed by two elements, each composed of five Į-helices, which contain motifs 1 and 2. These short amino acid sequences are critical for Arf binding and catalysis of nucleotide exchange. A diagram of the domain organization of BIG1 is shown in Figure 1. Specific amino acids in a site formed by apposition of the two motifs appear to interact with complementary side chains in switch 1 (residues 41-55) and switch 2 (residues 70-80) of Arf (Mossesova et al., 1998; Goldberg, 1998; Betz et al., 1998; Cherfils et al., 1998; Renault et al., 2002).

BIGS: BFA-SENSITIVE ARF GEFS

1255

Figure 1. Functional domains of BIG1. Mansour et al. (1999) reported that the N-terminal fragment (residues 1-560) was required for Golgi localization and identified the HUS (homology upstream of the Sec7 domain) box as important for protein stability. The central Sec7 domain contains Motifs 1 and 2. Asterisk indicates region that increases GEF activity of the Sec7 domain 100-fold (Morinaga et al., 1999).

Functions of regions outside the Sec7 domains of BIG1 and BIG2 are, for the most part, obscure. Arf GEF activity of BIG1(1-885), which lacks the 964 residues C-terminal to the Sec7 domain, was equivalent to that of the full-length protein, when recombinant proteins synthesized in Sf9 cells were compared. Additional studies of BIG1(1-885), truncated at the N-terminus, indicated that removal of amino acids 1-594 (BIG1 (595-885) had no effect, but an additional truncation of 35 residues (BIG1 (630-885) resulted in the loss of 99% of the Arf GEF activity (Morinaga et al., 1999).This sequence, with a predicted pI of 6.30, appears to form a coil, integrated by three helices that are connected by segments without defined structure, which probably represent loops. The effect of this segment is consistent with earlier observations that regions outside the Sec7 domain altered cytohesin Arf GEF activity (Chardin et al., 1996; Pacheco-Rodriguez et al., 1998). Mansour et al. (1999) identified a sequence in BIG1 (amino acids 562VVDIYVNYDCDLNAANIFERLVNDLSKI-590), which they referred to as the region of “homology upstream of the Sec7 domain (HUS)” that is also present in other large GEFs, such as Sec7, GBF1, and EMB30. They suggested that this sequence, which is identical in BIG1 and BIG2 (Togawa et al., 1999), could be important for protein folding and stability. The region N-terminal to it appears to contain a Golgi localization signal. Among five epitope-tagged BIG1 mutants that were overexpressed in BHK cells, only the one containing residues 1-560 coincided with the Golgi marker mannosidase II, as did the full-length BIG1 (Mansour et al., 1999). Subcellular localization signals have not been recognized in other Arf GEFs leaving in question the determinants of their subcellular distributions, which

6 126

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

undoubtedly depends on specific molecular interactions and very possibly reversible covalent modification. FBS (F-box/Sec7), a human protein of 319 amino acids, was identified in a data base search for new E3 ubiquitin-protein ligases that contain F-box motifs. Its deduced sequence contains also a Sec7-like domain (residues 139-203), which includes the glutamate that is critical for GEF activity in other Sec7 domains (Ilyin et al., 2000). The F-box was recognized as a structural element required for the binding of Skp1p, a cyclin F-binding protein functionally implicated in cell cycle regulation (Bai et al., 1996). Regulated proteolytic degradation of these proteins is critical to the control of their function. Many proteins in the F-box family contain additional conserved structural motifs (e.g., leucine-rich repeats, WD40 domains) that are involved in specific protein-protein interactions (Ilyin et al., 2000). More information regarding FBS and its potential GEF activity is clearly needed. Sec7 domains contain at least some of the determinants for specificity of Arf GEF activity and/or molecular interaction (see also section 3, Mechanism of BFA inhibition). The cytohesin-1 Sec7 domain, however, had lower substrate specificity than the intact molecule (Pacheco-Rodriguez et al., 1998). Cytohesin-1 accelerated GTPȖS binding to recombinant human Arf1, yeast Arf3, and ARD1, a 64 kDa protein that contains a C-terminal Arf domain (Mishima et al., 1993). The cytohesin-1 Sec7 domain, on the other hand, was active with human Arfs 1, 5, and 6, yeast Arfs 1, 2, and 3, ARD1, two ARD1 mutants that contain the Arf domains, and ǻ13Arf1, which lacks the N-terminal Į-helix and terminal glycine. Neither cytohesin-1 nor its Sec7 domain increased GTPȖS binding to human ARL 1, 2, or 3. It was concluded that the N-terminus of Arf is important for functional interaction with cytohesin-1 and that critical determinants of substrate specificity reside outside the Sec7 domain of cytohesin-1 (Pacheco-Rodriguez et al., 1998). This seems likely to be true also for BIG1 and BIG2, but remains to be investigated. Knowledge of the functional specificity of Arf-Arf GEF interactions in cells is, at best, limited, albeit improving in the last few years. The GEF activity of a protein assayed in vitro by its acceleration of GTPȖS binding to a particular Arf substrate has no necessary or predictable relationship to the physiological actions of the molecules studied. Inside the cell, Arfs and Arf GEFs operate in an environment of numerous proteins and lipids, which dynamically influence multiple simultaneous and sequential interactions. An in vitro system with GTPȖS includes none of these and is lacking also the cyclic interconversion of Arf•GDP and Arf•GTP that is required for its biological function. Application of the procedure described by Santy and Casanova (Santy and Casanova, 2001) to “pull down” Arf•GTP from cell

BIGS: BFA-SENSITIVE ARF GEFS

1277

lysates using an immobilized fragment of GGA3 could prove very useful for identifying and quantifying specific intracellular Arf-Arf GEF interactions. A recent publication compared the crystal structure of the free (without ligand) Sec7 domain of Gea2 (Renault et al., 2002) with that of Gea2 bound to nucleotide-free Arf1 that lacks the N-terminal myristate and amino acids 1-17 (Goldberg, 1998). This information dramatically increases our understanding of the molecular mechanisms of Arf GEF action and its inhibition by BFA (reviewed in the next section). Renault et al. (Renault et al., 2002) suggested also that differences among Arf GEFs in the width of the empty catalytic groove that accommodates the Arf molecule could serve to discriminate among different structures of the individual Arfs (Amor et al., 1994; Ménétrey et al., 2000). Additional detailed data of this kind will surely facilitate identification of the specific roles and interactions of these molecules.

3.

INTRACELLULAR LOCALIZATION OF BIG1 AND BIG2

BIG1 and BIG2 had been purified from bovine brain cytosol as components of a ~670 kDa complex (Morinaga et al., 1996), and were, to a large extent, immunoprecipitated together from HepG2 cell cytosol by antibodies specific for BIG1 or BIG2 (Yamaji et al., 1999). BIG1 and BIG2 were also partially co-immunoprecipitated from CHAPS-solubilized, microsomal fractions. After centrifugation of microsomal preparations on a sucrose density gradient, BIG, BIG2, and ß-COP (one of seven proteins that compose the COPI coatomer) were recovered only in the 1.15 M/0.85 M sucrose (heavy Golgi) fraction (Yamaji et al., 1999). By fluorescence microscopy, both endogenous BIG1 and BIG2 in HepG2 cells were partially colocalized with the Golgi-specific 58 kDa protein and partially with Ȗ-adaptin, a component of the AP-1 complex. The distribution of over-expressed BIG2 and endogenous BIG1, BIG2, and also appeared quite similar in HepG2 and HeLa cells. After exposure of HepG2 cells to BFA for 10 min with complete release of ȕ-COP from Golgi structures, both BIG1 and BIG2 were still Golgi-associated (Yamaji et al., 1999). Mansour, et al. (Mansour et al., 1999), however, found complete dispersal of over-expressed BIG1 from Golgi in BHK2 cells incubated for 10 min with BFA, which likely reflects simply a difference in the time required for BFA inhibition in different cells. Zhao et al. (2002) reported that endogenous BIG1 in NRK cells was localized in the TGN. Its concentration in the Golgi compartment was contrasted with that of GBF1, a 208 kDa, BFA-resistant Arf GEF that preferentially activates Arf5 (Claude et al., 1999), which was confined to a

8 128

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

cis-Golgi compartment. Using quantitative confocal microscopy, Zhao et al. (Zhao et al., 2002) demonstrated less than 25% overlap of the two proteins. This methodology was also employed to quantify the extent of overlap of the two Arf GEFs with Golgi and coat protein markers. The distribution of over-expressed BIG2(2-552) carrying an N-terminal hemagglutinin (HA) tag was very similar to that of endogenous BIG1 (Zhao et al., 2002). Whereas BIG1 and BIG2 most likely function with class I Arfs in BFA-sensitive anterograde traffic from the TGN, GBF1 may operate with Arf5 in retrograde transport from Golgi to ER. It seems possible that the BFAinsensitive binding of Arf5 to Golgi membranes from rat brain reported by Tsai et al. (Tsai et al., 1993) was a reflection of GBF1 activity. An association of BIG1 and BIG2 with the Ȗ-adaptin component of AP-1 is consistent with the reported BFA inhibition of formation of clathrin-coated vesicles (Robinson and Bonifacino, 2001). Earlier this year, Shinotsuka et al. (Shinotsuka et al., 2002a) reported that HA- or Myc-tagged BIG2 over-expressed in human HeLa cells or Clone 9 rat hepatocytes was localized in the perinuclear region. Staining overlapped with that for ȕ-COP, mannosidase II, Ȗ-adaptin, and mannose-6-phosphate receptor, but not with LAMP1 or EEA1. They used a glutathione Stransferase (GST)-tag attached to the domain of GGA1 that specifically binds Arf•GTP to show that the amount of activated Arf in cell lysates was increased by overexpression of BIG2. BFA caused dispersal of overexpressed Arf1-HA from the perinuclear area to the cytosol. When BIG2 was also overexpressed it appeared localized with Arf1-HA. On exposure of cells to BFA, both Arf1-HA and Myc-BIG2 were seen on tubulovesicular structures radiating from the perinuclear region. Comparison of the effects of BFA on the behavior of ȕ-COP and Ȗ-adaptin revealed that whereas overexpression of BIG2 did not interfere with BFA-induced release of ȕCOP, it did cause retention of Ȗ-adaptin which was seen along the tubular structures with which BIG2 was associated. The authors concluded that BIG2 was acting primarily at the TGN in the process of AP-1 recruitment to form clathrin-coated vesicles. Over-expression of BIG2 did not prevent the rapid BFA-induced release of ȕ-COP or the slower disruption of Golgi structure evidenced by dispersal of mannosidase, consistent with the notion that its function differs from that of BIG1, which presumably activates class I Arfs for anterograde transport from the cis-Golgi (Shinotsuka et al., 2002a). Following their demonstration that BIG2 is responsible for the initiation of clathrin-coated vesicles by activating Arf to recruit Ȗ-adaptin, Shinotsuka et al. (Shinotsuka et al., 2002b), reported that over-expression in HEK 293 cells of dominant negative BIG2-K738, which lacks Arf GEF activity, caused redistribution of Ȗ-adaptin from the perinuclear region, without

BIGS: BFA-SENSITIVE ARF GEFS

1299

affecting ȕ-COP. They showed that the adaptor protein GGA1, like Ȗadaptin and BIG2-K738, was dispersed along tubular structures resembling those seen in BFA-treated cells. These and other observations implicated BIG2 in transport of lysosomal enzymes between the TGN and late endosomes (Shinotsuka et al., 2002b). Earlier conclusions regarding the apparently similar subcellular distributions of BIG1 and BIG2, both of which appeared to colocalize partially with both Ȗ-adaptin and coatomer, were undoubtedly due to insufficiently precise techniques for immunofluorescence microscopy (Yamaji et al., 1999). They were perhaps accepted with less question than they might otherwise have been because of the knowledge that BIG1 and BIG2 had remained associated throughout six steps of purification and were immunoprecipitated together, both phenomena that demand, now more than ever, detailed investigation and explanation.

4.

MECHANISMS OF BFA INHIBITION

The recognition that BFA inhibition of Gea1 (Peyroche et al., 1999) and BIG1 (Mansour et al., 1999) is non-competitive was the key to understanding its mechanism of action. Peyroche et al. (Peyroche et al., 1999) demonstrated directly that BFA stabilized and thus increased the amount of a stoichiometric complex containing the Gea1 Sec7 domain and Arf1-∆17•GDP (lacking the N-terminal myristate and 17 amino acids). They identified two amino acids in the Sec7 domain that were critical for BFA inhibition of secretion in the mutant yeast used to test activity of the recombinant proteins. A chimeric Gea1 containing an ARNO Sec7 domain was used to demonstrate that replacement of F190 and A191, respectively, with the residues found in the corresponding position in Gea1 (tyrosine and serine), resulted in BFA sensitivity and the capacity to form a stable BFAArf•GDP-Sec7 domain complex, each of which was lacking in the parent molecule (Peyroche et al., 1999). Sata et al. (Sata et al., 1999) showed that replacement of S198 and P208 in the Sec7 domain of cytohesin-1 (an Arf GEF very similar to ARNO) with aspartate and methionine, respectively, as found in the BFA-inhibited BIG1 (p200), made it sensitive to BFA. Sequences in these regions of BFA-inhibited and BFA-insensitive GEFs are aligned in Figure 2. In the crystal structure of the ARNO Sec7 domain, these four amino acid side chains are exposed on the surface, in or near the hydrophobic groove in which the Arf molecule binds through its switch 1 and switch 2 regions (Mossesova et al., 1998). Specific interactions of individual amino acids in the two switch regions of Arf with complementary residues in motifs 1 and 2 of the Sec7 domain were identified in crystal

10 130

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

structures of the Gea2 Sec7 domain complex with a mutant Arf1, either nucleotide-free or bound to Gpp(NH)p, a poorly hydrolysable GTP analogue (Goldberg, 1998).

Figure 2. Amino acids in the Sec7 domain that influence sensitivity to BFA. Sequences shown contain Motif 2 (Mossesova et al., 1998), which forms a side of the Arf-binding groove and includes one pair of residues (**) that modify BFA sensitivity (Peyroche et al., 1999). The adjacent sequence contains another pair of amino acids that affects BFA sensitivity, (*) and the boxed position, in which interchanging Q (as in BIG1) with N, (as in cytohesins), did not affect BFA sensitivity (Sata et al., 1999). Five BFA-sensitive GEFs, from human (h), yeast (y), and Arabadopsis (a), are shown, along with four BFA-insensitive cytohesins.

Robineau et al. (Robineau et al., 2000) used [3H]BFA-binding to define structural requirements for the formation of a BFA-Arf•GDP-Sec7 domain complex. BFA binding was similar whether the Sec7 domain contained Y190 and S191 (Peyroche et al., 1999) or D198 and M208 (Sata et al., 1999), but was essentially doubled when both pairs of amino acids were present. Differences in the kinetics of [3H]BFA binding and release from complexes with Sec7 domains containing either only two or all four of the

BIGS: BFA-SENSITIVE ARF GEFS

11 131

critical amino acids provided further insight into molecular interactions (Robineau et al., 2000). All Sec7 domains contain a specific glutamic acid that is responsible for accelerating the release of bound guanine nucleotide from Arf, i.e., for GEF activity. Replacement of this E156 with lysine in the BFA-sensitive quadruple mutant Sec7 domain markedly accelerated its [3H]BFA binding, indicative of complex formation. Without Arf GEF activity the concentration of Arf•GDP remains high, whereas on interaction with an active Sec7 domain, Arf•GDP is rapidly converted to nucleotidefree Arf, which forms a much more stable interaction, so that the concentration of the short-lived Arf•GDP-Sec7 domain that can bind BFA is always very low. Effects of mutations in switch 2 of Arf1 that had little effect on Arf GEF activity were also evaluated using the BFA-binding assay. When H80 in Arf 1 was replaced with alanine, release of BFA from the complex formed with the quadruple mutant Sec7 domain was slowed. Comparison of Sec7 domains containing only one of the two pairs of mutant amino acids, demonstrated it was D198 and M208 (found in BIG1) that were necessary for this effect of the A80 mutation, one of the examples of contributions of elements in both of the molecules to interaction specificity (Robineau et al., 2000). These observations, together with the three-dimensional structures, provide important insights into mechanisms of Arf GEF action, as well as BFA inhibition. Wittinghofer and colleagues have made major contributions to the understanding of mechanisms by which guanine nucleotide-binding proteins function as molecular switches in the regulation of cell signaling and metabolism. A recent review presents a valuable analysis of the differences, as well as similarities, among the diverse families of GTPases, including the Arfs (Vetter and Wittinghofer, 2001). In considering the mechanism of Sec7 domain actions in Arf activation and BFA sensitivity, Renault et al. (Renault et al., 2002) began with evidence of the structural flexibility of Arf, relative to that of the Sec7 domain. To evaluate flexibility, they compared the crystal structure of the Gea2 Sec7 domain alone with that of the domain bound to nucleotide-free Arf (Goldberg, 1998). They found that the N- and C-terminal halves of the Sec7 domain, which form sides of the Arf-binding groove, were significantly closer together when Arf was bound. This was termed “Sec7 domain closure” and was accompanied by changes in the hinge region that links the two halves, involving the conformation of helix Į7 (Renault et al., 2002). A similar effect of Arf binding on the structure of cytohesin-1 had been reported (Betz et al., 1998). Renault et al. (Renault et al., 2002) considered the behavior of Gea2 in light of the published data on the structure of the ARNO Sec7 domain. They noted that three of the four positions that had been implicated in determination of BFA sensitivity, form more bonds in the

12 132

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

empty Sec7 domain when they are occupied by tyrosine, aspartate, and methionine (as in BFA-sensitive Arf GEFs) than when occupied by phenylalanine, serine, and proline (BFA-insensitive). They concluded that these three positions are important in the flexibility of the Sec7 domain that is necessary for alternation between the open (empty) and closed (Arfbound) conformations of the Arf-binding groove. Both flexibility and the capacity to form hydrogen bonds may contribute to BFA sensitivity. Renault et al. (Renault et al., 2002) observed that the specificity of an ArfArf GEF interaction should probably be established by the initial reaction of the two molecules. They suggested that differences among GEFs in the width of the empty catalytic groove, which is significantly wider in ARNO than Gea2, could serve to discriminate among structures of the GDP-bound forms of individual Arfs (Amor et al., 1994; Ménétrey et al., 2000) in the switch 1 and switch 2 regions with which they interact.

5.

CONCLUSIONS

BFA continues to be a valuable tool in the dissection of molecular processes responsible for vesicular transport, very recently in the direct demonstration that during vesicle formation in the Golgi, release of coatomer does not immediately follow the inactivation and dissociation of Arf. Active Arf is required for the recruitment of coatomer, but not for its continued function in the assembly of cargo and modification of membrane morphology (Presley et al., 2002). BIG1 and BIG2 were isolated together from bovine brain cytosol through six steps of purification (Morinaga et al., 1996), and were to a large extent co-immunoprecipitated from HepG2 cell cytosol by specific antibodies against one or the other (Yamaji et al., 1999). Growing evidence that they can regulate different intracellular transport pathways (Shinotsuka et al., 2002a,, 2002b), however, indicates the necessity of further investigation and explanation of those phenomena. Extensive analysis of the interactions of individual amino acids in the Sec7 domain of Gea2, which is identical to BIG1 and BIG2 in the residues that determine BFA inhibition, with different nucleotide-bound forms of Arf1, which is also a preferred substrate for BIG1 and BIG2, has provided detailed new insight into mechanisms of GEF action and BFA inhibition (Renault et al., 2002).

BIGS: BFA-SENSITIVE ARF GEFS

13 133

REFERENCES Achstetter, T., Franzusoff, A., Field, C., and Schekman, R. (1988). SEC7 encodes an unusual, high molecular weight protein required for membrane traffic from the yeast Golgi apparatus. J. Biol. Chem., 263, 11711-11717. Amor, J. C., Harrison, D. H., Kahn, R. A., and Ringe, D. (1994). Structure of human ADPribosylation factor 1 complexed with GDP. Nature, 372, 704-708. Bai, C., Sen, P., Hofmann, K., Ma, L., Goebl, M., Harper, J. W., and Elledge, S J. (1996). SKP1 connects cell cycle regulators to the ubiquitin proteolysis machinery through a novel motif, the f-box. Cell, 86, 263-274. Betz, S. F., Schnuchel, A., Wang, H., Olejniczak, E. T., Meadows, R. P., Lipsky, B. P., Harris, E. A. S., Staunton, D. E., and Fesik, S. W. (1998). Solution structure of the cytohesin-1 (B2-1) Sec7 domain and its interaction with the GTPase ADP-ribosylation factor 1. Proc. Natl. Acad. Sci., USA, 95, 7909-7914. Chardin, P., Paris, S., Antonny, B., Robineau, S., Béraud-Dufour, S., Jackson, C. L., and Chabre, M. (1996). A human exchange factor for ARF contains Sec7- and pleckstrinhomology domains. Nature, 384, 481-484. Cherfils, J., Ménétrey, J., Mathieu, M., LeBrass, G., Robineau, S., Béraud-Dufour, S., Antonny, B., and Chardin, P. (1998). Structure of the Sec7 domain of the ARF exchange factor ARNO. Nature, 392, 101-105. Claude, A., Zhao, B.-P., Kuziemsky, C. E., Dahan, S., Berger, S. J., Yan, J. P., Arnold, A. D., Sullivan, E. M., and Melançon, P. (1999). GBF1: a novel Golgi-associated BFA-resistant guanine nucleotide exchange factor that displays specificity for ADP-ribosylation factor 5. J. Cell Biol., 146, 71-84. Donaldson, J. G., Finazzi, D., and Klausner, R. D. (1992). Brefeldin A inhibits Golgi membrane-catalyzed exchange of guanine nucleotide into ARF protein. Nature, 360, 350352. Franzusoff, A., and Schekman, R. (1989). Functional compartments of the yeast Golgi apparatus are defined by the Sec7 mutation. EMBO J., 8, 2695-2702. Fujiwara, T., Oda, K., Yokota, S., Takatsuki, A., and Ikehara, Y. (1988). Brefeldin A causes disassembly of the Golgi complex and accumulation of secretory proteins in the endoplasmic reticulum. J. Biol. Chem., 263, 18545-18552. Goldberg, J. (1998). Structural basis for activation of ARF GTPase: Mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Helms, J. B., and Rothman, J. E. (1992). Inhibition by brefeldin A of a Golgi membrane enzyme that catalyzes exchange of guanine nucleotide bound to ARF. Nature, 360, 352354. Ilyin, G. P., Rialland, M., Pigeon, C., and Guguen-Guillouzo, C. (2000). CDNA cloning and expression analysis of new members of the mammalian f-box protein family. Genomics, 67, 40-47. Mansour, S. J., Skaug, J., Zhao, X.-H., Giordano, J., Scherer, S. W., and Melancon, P. (1999). p200 ARF-GEP: a Golgi-localized guanine nucleotide exchange protein whose Sec7 domain is targeted by the drug brefeldin A. Proc. Natl. Acad. Sci., USA, 96, 7968-7973. Ménétrey, J., Macia, E., Pasqualato, S., Franco, H., and Cherfils, J. (2000). Structure of ARF6-GDP suggests a basis for guanine nucleotide exchange factors specificity. J. Nat. Struct. Biol., 7, 466-469. Mironov, A., Colanzi, A., Silletta, M. G., Fiucci, S., Fusella, A., Polishchuk, R., Mironov, A., Jr., DiTullio, G., Weigert, R., Malhotra, V., Corda, D., deMatteis, M. A., and Luini, A. (1997). Role of NAD+ and ADP-ribosylation in the maintenance of the Golgi structure. J. Cell Biol., 139, 1109-1118.

14 134

PACHECO-RODRIGUEZ, MOSS, AND VAUGHAN

Mishima, K., Tsuchiya, M., Nightingale, M. S., Moss, J., and Vaughan, M. (1993). ARD1, a 64 kDa guanine nucleotide-binding protein with a carboxyl-terminal ADP-ribosylation factor domain. J. Biol. Chem., 268, 8801-8807. Misumi, Y., Misumi, Y., Miki, K., Takatsuki, A., Tamura, G., and Ikehara, Y. (1986). Novel blockade by brefeldin A of intracellular transport of secretory proteins in cultured rat hepatocytes. J. Biol. Chem., 261, 11398-11403. Morinaga, N., Adamik, R., Moss, J., and Vaughan, M. (1999). Brefeldin A inhibited activity of the Sec7 domain of p200, a mammalian guanine nucleotide-exchange protein for ADPribosylation factors. J. Biol. Chem., 274, 17417-17423. Morinaga, N., Moss, J., and Vaughan, M. (1997). Cloning and expression of a cDNA encoding a bovine brain brefeldin A-sensitive guanine nucleotide-exchange protein for ADP-ribosylation. Proc. Natl. Acad. Sci., USA, 94, 1226-1231. Morinaga, N., Tsai, S.-C., Moss, J., and Vaughan, M. (1996). Isolation of a brefeldin Ainhibited guanine nucleotide-exchange protein for ADP-ribosylation factor (ARF) 1 and ARF 3 that contains a Sec7-like domain. Proc. Natl. Acad. Sci., USA, 93, 12856-12860. Mossesova, E., Gulbis, J. M., and Goldberg, J. (1998). Structure of the guanine nucleotide exchange factor Sec7 domain of human ARNO and analysis of the interaction with ARF GTPase. Cell, 92, 415-423. Orci, L., Tagaya, M., Amherdt, M., Perrelet, A., Donaldson, J. G., Lippincott-Schwartz, J., Klausner, R. D., and Rothman, J. E. (1991). Brefeldin A, a drug that blocks secretion, prevents the assembly of non-clathrin-coated buds on Golgi cisternae. Cell, 64, 11831195. Pacheco-Rodriguez, G., Meacci, E., Vitale, N., Moss, J., and Vaughan, M. (1998). Guanine nucleotide exchange on ADP-ribosylation factors catalyzed by cytohesin-1 and its Sec7 domain. J. Biol. Chem., 273, 26543-26548. Palade, G. E. (1975). Intracellular aspects of the process of protein synthesis. Science, 189, 347-358. Peyroche, A., Antonny, B., Robineau, S., Acker, J., Cherfils, J., and Jackson, C. L. (1999). Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell., 3, 275-285. Presley, J. F., Ward, T. H., Pfeifer, A. C., Siggla, E. D., Phair, R. D., Lippincott-Schwartz, J. (2002). Dissection of COPI and ARF1 dynamics in vivo and role in Golgi membrane transport. Nature, 417, 187-193. Renault, L., Christova, P., Guibert, B., Pasqualato, S., and Cherfils, J. (2002). Mechanism of domain closure of Sec7 domains and role in BFA sensitivity. Biochemistry, 41, 36053612. Robineau, S., Chabre, M., and Antonny, B. (2000). Binding site of brefeldin A at the interface between the small G protein ADP-ribosylation factor 1 (ARF1) and the nucleotideexchange factor Sec7 domain. Proc. Natl. Acad. Sci., USA, 97, 9913-9918. Robinson, M. S., and Bonifacino, J. S. (2001). Adaptor-related proteins. Curr. Opin. Cell Biol., 13, 444-453. Rothman, J. E. (1994). Mechanisms of intracellular protein transport. Nature, 372, 55-63. Santy, L. C., and Casanova, J. E. (2001). Activation of ARF6 by ARNO stimulates epithelial cell migration through downstream activation of both Rac1 and phospholipase D. J. Cell. Biol., 154, 599-610. Sata, M., Donaldson, J. G., Moss, J., and Vaughan, M. (1998). Brefeldin A-inhibited guanine nucleotide-exchange activity of Sec7 domain from yeast Sec7 with yeast and mammalian ADP ribosylation factors. Proc. Natl. Acad. Sci., USA, 95, 4204-4208.

BIGS: BFA-SENSITIVE ARF GEFS

15 135

Sata, M., Moss, J., and Vaughan, M. (1999). Structural basis for the inhibitory effect of brefeldin A on guanine nucleotide-exchange proteins for ADP-ribosylation factors. Proc. Natl. Acad. Sci., USA, 96, 2752-2757. Schmidt, T. A., Wolde, M., Thiele, C., Fest, W., Kratz, H., Podtelejnikov, A., Witke, W., Hultner, W. B., and Söling, H.-D. (1999). Endophilin 1 mediates synaptic vesicle formation by transfer of arachidonate to lysophosphatidic acid. Nature, 401, 133-141. Shinotsuka, C., Waguri, S., Wakasugi, M., Uchiyama, Y., and Nakayama, K., (2002b). Dominant-negative mutant of BIG2, an ARF-guanine nucleotide exchange factor, specifically affects membrane trafficking from the trans-Golgi network through inhibiting membrane association of AP-1 and GGA coat proteins. Biochem. Biophys. Res. Comm., 294, 254-260. Shinotsuka, C., Yoshida, Y., Kawamoto, K., Takatsu, H., and Nakayama, K. (2002a). Overexpression of an ADP-ribosylation factor-guanine nucleotide exchange factor, BIG2, uncouples brefeldin A-induced adaptor protein-1 coat dissociation and membrane tubulation. J. Biol. Chem., 277, 9468-9472. Spanfó, S., Silletta, M. G., Colanzi, A., Alberti, S., Fiucci, G., Valente, C., Fusella, A., Salmona, M., Mironov, A., Luini, A., and Corda, D. (1999). Molecular cloning and functional characterization of brefeldin A-ADP-ribosylated substrate. J. Biol. Chem., 274, 17705-17710. Spang, A., Matsuoka, K., Hamamoto, S., Schekman, R., and Orci, L. (1998). Coatomer, ARF1p, and nucleotide are required to bud coat protein complex I-coated vesicles from large synthetic liposomes. Proc. Natl. Acad. Sci., USA, 95, 11199-11204. Tamura, G., Ando, K., Suzuki, S., Takatsuki, A., and Arima, K. (1968). Antiviral activity of brefeldin A and verrucarin A. J. Antibiot., 21, 160-161. Togawa A., Morinaga, N., Ogasawara, M., Moss, J., and Vaughan, M. (1999). Purification and cloning of a brefeldin A-inhibited guanine nucleotide-exchange protein for ADPribosylation factors. J. Biol. Chem., 274, 12308-12315. Tsai, S.-C., Adamik, R., Haun, R. S., Moss, J., and Vaughan, M. (1993). Effects of brefeldin A and accessory proteins on association of ADP-ribosylation factors 1, 3, and 5 with Golgi. J. Biol. Chem., 268, 10820-10825. Vetter, I. R., and Wittinghofer, A. (2001) The guanine nucleotide binding switch in three dimensions. Science, 294, 1299-1304. Weigert, A., Colanzi, A., Mironov, R., Buccione, C., Cericola, M. G., Santini, G., Flati, S., Fusella, A., Donaldson, J. G., DiGirolamo, M., Corda, D., DeMatteis, M. A., and Luini, A. (1997). Characterization of chemical inhibitors of brefeldin A-activated mono-ADPribosylation. J. Biol. Chem., 272, 14200-14207. Weigert, R., Silletta, M. G., Spano, S., Turacchio, G., Cericola, C., Colanzi, A., Sematore, S., Mancini, R., Polishchuk, E. V., Salmona, M., Facchiano, F., Burger, K. N. J., Mironov, A., Luini, A., and Corda, D. (1999). CtBP/BARS induces fission of Golgi membranes by acylating lysophosphatidic acid. Nature, 402, 429-433. Yamaji, R., Adamik, R., Takeda, K., Togawa, A., Pacheco-Rodriguez, G., Ferrans, V. J., Moss, J., and Vaughan, M. (1999). Identification and localization of two brefeldin Ainhibited guanine nucleotide-exchange proteins for ADP-ribosylation factors in a macromolecular complex. Proc. Natl. Acad. Sci., USA, 97, 2567-2572. Zhao, X., Lasell, T. K. R., and Melançon, P. (2002). Localization of large ADP-ribosylation factor-guanine nucleotide exchange factors to different Golgi compartments: evidence for distinct functions in protein traffic. Mol. Biol. Cell, 13, 119-133.

Chapter 7 ARF GTPASE-ACTIVATING PROTEIN 1 DAN CASSEL Department of Biology, Technion-Israeli Institute of Technology, Haifa, Israel

Abstract:

1.

Regulators of Arf activity include a family of proteins with a shared domain, the cysteine-rich Arf GAP domain, that is responsible for activating the latent GTPase activity of Arfs. The first of these to be discovered, Arf GAP1 is the focus of this chapter. It’s role in the cellular actions of Arfs, particularly vesicular traffic, and the regulation of Arf GAP1 by other factors, e.g., lipids, is discussed.

PERSPECTIVES AND SCOPE

Like all regulatory GTPases, Arfs function as molecular switches, oscillating between the active, GTP-bound form and the inactive, GDP-bound form. These oscillations are tightly regulated by guanine nucleotide exchange factors (GEFs) that “switch on” the Arf proteins, and GTPase-activating proteins (GAPs) that act as off switches by facilitating the GTP hydrolysis (GTPase) activity of Arfs. Accompanying the on/off switching of Arfs is a cycle of membrane association/dissociation. In the GTP-bound state Arfs associate with cellular membranes to which they recruit their various effector molecules, thereby facilitating membrane traffic. GTP hydrolysis, brought about by an Arf GAP, results in the rapid dissociation of Arfs from membranes. Although Arf proteins were discovered in the 1980s (Kahn and Gilman, 1986), their regulators began to be identified only in the next decade. To date, multiple regulators of each part of the GTPase cycle of Arfs have been identified. These regulators possess domains that function in the binding to Arf and subsequent effects on GTP binding or hydrolysis. Thus, all Arf GEFs contain a Sec7 domain, and all Arf GAPs contain an “Arf GAP domain”. This chapter begins with the introduction of the family of mammalian Arf GAPs. A detailed discussion of Arf GAP1, the first Arf GAP to be 137

2 138

CASSEL

purified and cloned, is then presented. This is the smallest of the many Arf GAPs found in mammals, is localized to the cytosolic surface of Golgi membranes, and acts to regulate traffic through this organelle. Where appropriate, discussions will be extended to Arf GAPs from the yeast S. cerevisiae that are structurally and/or functionally related to Arf GAP1.

2.

ARF GAP PROTEINS

2.1

Conserved features and the catalytic domain

All Arf GAPs known to date contain a highly conserved “signature” sequence of approximately 50 amino acids that comprises the Arf GAP domain. The most notable feature of this sequence is the presence of a zinc finger motif, consisting of 4 cysteine residues arranged in the sequence CX2CX16CX2C, where X is any amino acid (see Figure 1). Residues between pairs of cysteines frequently include a charged amino acid, whereas

Figure 1. Expression of Arf GAP1 in rat tissues. Tissues were homogenized in isotonic sucrose in the presence of a cocktail of protease inhibitors, separated by centrifugation (100,000xg, 1 h) into cytosolic fraction and total membrane fractions, and analyzed by Western blotting with polyclonal, rabbit anti-Arf GAP1 antibodies.

the X16 “loop” is enriched in neutral and hydrophobic amino acids, reflecting the fact that this loop is localized in the core of the protein and does not face the solvent (Goldberg, 1999; Mandiyan et al., 1999). Particularly high sequence conservation is observed in a 20-25 amino acid stretch that follows the last cysteine residue of the zinc coordination site. This sequence is thought to be part of the GAP “catalytic domain” that mediates the stimulation of GTP hydrolysis on ARFs. In all, the GAP domain includes approximately 90 residues starting from the last cysteine of the zinc coordination site (see section 3.3, below). While some short sequence

ARF GTPASE-ACTIVATING PROTEIN 1

1393

elements are conserved between certain subgroups of Arf GAPs in the part of the catalytic domain that extends beyond the above-mentioned conserved sequence, sequences in this region are in general quite variable, and do not show any consensus when all Arf GAPs are compared.

2.2

Arf GAP subfamilies

The completed sequence of the human genome has revealed more than 20 genes encoding proteins with an Arf GAP signature. Many of these proteins have been shown to possess Arf GAP activity. A notable exception is centaurin-α (Hammonds-Odie et al., 1996), for which no GAP activity could be demonstrated. Arf GAPs can be divided into five subtypes or subfamilies, according to similarities in sequence and domain structure (see Table 1). One type is Arf GAP1, which does not contain any domains found Table 1: Mammalian Arf GAPs. Notes: 1Size of the longest splice variant (when multiple variants exist). 2Paxillin does not interact with Git2. 3Absent in ACAPs. 4Defective in ARAP2. 5Absent in ARAP1. 6Present only in ARAP2. Abbreviations: PH, plekstrin homology domain; ANK, ankyrin repeats; PrR, proline-rich, SH3-binding domain; SH3, Srchomology-3 domain; RA, Ras associating/Ral GDS domain; SAM, sterile α motif; RHD, Rab13 homology domain; GRK, G protein receptor kinase; FA, focal adhesions; ND, not determined. Family Size Other Substrate Location Binding Pathways kD1 Domains Preference Proteins Effected Arf GAP1 45 Arf1, 3, 5>6 Golgi KDEL ER-Golgi receptor traffic Arf GAP3 57 ND PeriND ND nuclear Git1/Cat1 85 ANK Arf1, 3, 5, 6 PM/ GRK, Surface Git2/Cat2 85 Arf1, 3, 5, 6 Endopaxilin2, receptor somes PIX internalization, integrin signaling ASAP1 Arf1, 5>6 130 CC, PH, FA Src, Crk Focal adheANK, Arf1, 5>6 Pyk2, sions, Papα/ PrR3, SH33 Arf6>1, 5 paxillin Integrin PAG3 signaling ACAP1, 2 ARAP1 PH (x5), Arf1,5>6 Golgi/PM ND CoordiARAP2 RhoGAP, ND ND nation of ARAP3 ANK, RA4, Arf6 PM Arf and SAM5, Rho family RHD6 GTPases

in other proteins besides the Arf GAP domain (Cukierman et al., 1995). This is also the case for Arf GAP3 (Zhang et al., 2000), a more recently discovered Arf GAP. Arf GAP1 and Arf GAP3 have orthologs in yeast,

4 140

CASSEL

named Gcs1 and Glo3, respectively (Poon et al., 1999; Poon et al., 1996). The second subfamily is the Git proteins, including three highly related members that contain ankyrin repeats, also present in most other Arf GAPs (Premont et al., 1998; Premont et al., 2000; Turner et al., 1999). Two additional Arf GAP families, the ASAPs/ACAPs (Andreev et al., 1999; Brown et al., 1998; Jackson et al., 2000; Kondo et al., 2000) and ARAPs (Krugmann et al., 2002; Miura et al., 2002) contain PH domain(s) and show phosphoinositide-dependent Arf GAP activity. Alternative splicing is common in most Arf GAP genes, adding further to the diversity of the Arf GAP family of proteins. The domain structure, substrate specificity, cellular localization and the general function of Arf GAPs are summarized in Table 1. For detail on some of these proteins, see subsequent chapters (e.g., Chapters 8 and 9).

3.

ARF GAP1

3.1

Arf GAP role in the COPI system

Because Arf GAP1 is thought to function in the COPI system, we will first briefly introduce this system. For additional details of the role of Arfs in vesicle coat protein recruitment at Golgi and other membranes, see Chapters 12 and 13. The COPI system functions primarily in the retrograde traffic of proteins from the Golgi to the endoplasmic reticulum (ER), although there is evidence for its function in anterograde transport from ER to Golgi as well (reviewed in (Lippincott-Schwartz et al., 1998; Rothman and Wieland, 1996; Schekman and Orci, 1996)). The COPI coat (termed coatomer) is a sevensubunit complex that is recruited en block from cytosol onto Golgi membrane. Coatomer recruitment is initiated when Arf1 is converted to the GTP-bound form by the action of a GEF at the Golgi. Subsequently, Golgibound Arf1•GTP acts to recruit coatomer, apparently by direct binding to coatomer subunits (Eugster et al., 2000; Zhao et al., 1997; Zhao et al., 1999). Coatomer also interacts with cytoplasmic domains of membrane proteins bearing the KKXX retrograde transport (sorting) signal, thereby incorporating this cargo into COPI vesicles (Cosson and Letourneur, 1994; Letourneur et al., 1994). p23/p24 transmembrane proteins, that are abundant in the ER-to Golgi shuttle, also interact with coatomer through signals on their cytoplasmic domain (Dominguez et al., 1998; Fiedler et al., 1996; Nickel et al., 1997; Sohn et al., 1996). The interaction with coatomer is likely to function in the recruitment of p23/24 proteins into COPI vesicles, but the precise function of these proteins remains uncertain. Following its binding to the Golgi membrane, coatomer is thought to polymerize and

ARF GTPASE-ACTIVATING PROTEIN 1

1415

facilitate the deformation of the membrane, that is required to initiate vesicle budding. Before fusion with the target membrane, the vesicle must shed its protein coat. This process of vesicle uncoating depends on the hydrolysis of Arf1bound GTP (Tanigawa et al., 1993; Teal et al., 1994). When COPI vesicles are formed in the presence of the hydrolysis-resistant GTP derivative, GTPγS, both uncoating and fusion are prevented. Because Arfs are devoid of any intrinsic GTPase activity, GTP hydrolysis is completely dependent on the action of an Arf GAP. Whether Arf GAP acts as an “uncoating factor” at a time after completion of vesicle biogenesis or whether GTP hydrolysis is initiated prior to the completion of the budding process, is still a matter of some debate. It is noteworthy that Arf1 also regulates the interaction of coat proteins with the trans-Golgi network. Because evidence so far has not indicated a role for Arf GAP1 in this organelle, we will limit our discussion to the COPI system, but one should bear in mind that a role of Arf GAP1 at the TGN cannot be ruled out at present.

3.2

Discovery of Arf GAP1 and its basic features

Arf GAP1 was identified following its purification from rat liver cytosol (Cukierman et al., 1995; Makler et al., 1995). The purification was monitored with a GAP assay in which the substrate was myristoylated Arf1 that had been loaded with [γ-32P]GTP in the presence of dimyristoylphosphatidylcholine and cholate. The absence of phosphoinositides in this assay probably precluded the detection of the phosphoinositide-dependent Arf GAPs. Purification was greatly facilitated by the finding that GAP activity could be readily recovered following denaturing conditions, allowing for a critical purification step in the presence of 8 M urea, and reverse-phase HPLC in organic solvents. The ability to renature Arf GAP1 by simple dilution with buffer is shared by yeast Arf GAPs Gcs1 and Glo3 (Poon et al., 1999; Poon et al., 1996), and is probably attributable to the zinc coordination domain, as this domain retains the bound zinc even after treatment in the presence of 6 M guanidine and 1 mM EDTA (I. Huber and D. Cassel, unpublished observation). However, this resistance to denaturants is not a property of all Arf GAPs, possibly because additional domains that are required for GAP activity (such as the PH domain in phosphoinositide-dependent GAPs) may not refold under these conditions. Molecular cloning of the Arf GAP1 cDNA from a rat liver cDNA library (Cukierman et al., 1995) revealed a protein containing 415 amino acids (accession NP_659558) (Figure 1). Database entries indicate that the mouse ortholog of Arf GAP1 (accession XP_130740) contains 414 amino acids,

6 142

CASSEL

whereas two human splice variants encode proteins of 406 (NP_060679) and 414 (AAH28233) amino acids. Sequencing of different rat liver cDNA library-derived clones has revealed the presence of multiple splice variants (Cukierman et al., 1995). One variant had an insert containing an early stop codon, and thus encoded a truncated protein that nevertheless had full Arf GAP activity when expressed in E. coli, while another variant lacked the entire zinc finger domain. In this variant the initiating methionine became out of frame, but translation could be initiated from the second methionine in the sequence (Met64), resulting in an inactive protein. Additional splice variants were identified but have not been characterized in detail. Western blot analysis of different rat tissues and cell lines showed that the long form (containing 415 amino acids) that we consider the “normal” protein is the most abundant form expressed. In some cases the expression of smaller proteins was observed (see Figure 2), but whether these forms are splice variants or degradation products has not been definitively established. Attempts to express the full-length Arf GAP1 protein in E. coli were unsuccessful, whereas fragments containing up to approximately 80% of the amino terminal sequence could be readily expressed in bacteria (Cukierman et al., 1995). Amino terminal fragments with full catalytic activity were employed in numerous studies of the catalytic mechanism (see below). More recently, full-length Arf GAP1 was successfully expressed in insect cells using the baculovirus system (Huber et al., 2001; Vitale et al., 2000). An amino terminal fragment containing the first 257 amino acids of Arf GAP1 was employed for immunization of rabbits, generating antibodies that recognized the protein with high affinity and specificity. Western blot analysis (Figure 2) showed that the protein is ubiquitously expressed in rat tissues, with the highest levels observed in brain and less in liver and lung. In all tissues examined, the protein was distributed between cytosol and total cellular membranes. Comparison to purified, recombinant protein revealed that Arf GAP1 comprises ~0.004% of liver cytosolic proteins, and is yet less abundant in most cultured cells. Use of the antibodies in indirect immunofluorescence showed that Arf GAP1 distributes between cytosol and the Golgi complex, where it colocalizes with coatomer. This distribution pattern was observed in normal rat kidney (NRK) cells (Cukierman et al., 1995) as well as in many other cell types, including baby hamster kidney (BHK), HeLa and COS-7 cells (E. Cukierman and D. Cassel, unpublished observations). When examined in GAP assays in vitro, Arf GAP1 was reported to prefer Arf1, as compared with Arf6 ((Brown et al., 1998; Vitale et al., 2000) and B. Antonny and D. Cassel, unpublished observation). The preference for Arf1 over Arf6 may be further increased in vivo, due to the localization of Arf

ARF GTPASE-ACTIVATING PROTEIN 1

1437

GAP1 to Golgi, and the differential distribution of Arf1 and Arf6 to Golgi and plasma membrane/endosomes, respectively.

Figure 2. (A) The sequence of rat Arf GAP1 is shown with the four cysteines that comprise the zinc finger (ZF) motif and the essential arginine residue (R50) highlighted. (B) The domain organization of Arf GAP1 is shown. The N-terminal GAP catalytic domain is conserved in all known Arf GAPs, whereas the non-catalytic domain is unique to Arf GAP1 and does not contain any motif present in other proteins.

When overexpressed in mammalian cells, Arf GAP1 caused the disappearance of the Golgi complex and its fusion with ER, whereas inactive mutants bearing substitutions in the scaffold cysteines of the zinc finger (Cukierman et al., 1995) or in an invariant arginine residue (Arg50, Pick and Cassel, unpublished observation) had no effect. A simple model, whereby an excess of Arf GAP activity causes the conversion of Arf1 to the GDP form with subsequent dissociation of coat protein from the Golgi, may explain these findings. This would result in the fusion of the Golgi with the ER, an effect similar to that of the drug brefeldin A (BFA), which inactivates Arf by inhibiting a subset of Arf GEFs (Donaldson, 1992b; Donaldson and Jackson, 2000; Helms and Rothman, 1992; Peyroche et al., 1999). Structure-function studies showed that the catalytic domain of Arf GAP1 by itself was ineffective in causing Golgi dissociation, and that this effect requires determinants positioned in the carboxyl-terminal, non-catalytic part of Arf GAP1 (Huber et al., 1998). Apparently, these determinants increase the efficiency of GAP-induced Golgi disassembly by mediating GAP binding to the Golgi, thereby increasing its proximity to its substrate,

8 144

CASSEL

membrane-bound Arf1-GTP. This prediction has recently received strong support by a study employing Arf GAP1 fragments fused to the green fluorescence protein (GFP) (Yu and Roth, 2002), which demonstrated that carboxyl terminal residues are both necessary and sufficient for targeting Arf GAP1 to the Golgi. Additionally, this study has suggested a role of phosphorylation by casein kinase-I in Arf GAP1 binding to the Golgi.

3.3

Catalytic core

When various amino terminal fragments of Arf GAP1 were tested for in vitro GAP activity, high activity was observed in proteins containing the first 146 amino acids or more. Substantial, although much lower, activity was still observed in a mutant protein containing only the first 121 amino acids, whereas shorter fragments were inactive (Cukierman et al., 1995). These findings suggested that the minimal GAP catalytic core resides within the first ~120 amino acids. An essential role of the zinc finger motif was initially suggested by the finding that mutations in its scaffold cysteines abolished GAP activity (Cukierman et al., 1995). However, more recent crystallographic data (Goldberg, 1999) indicated that the 16 amino acid “loop” of the zinc finger domain is buried within the protein core, and is thus unlikely to directly participate in catalysis. The loss of activity upon mutation of the scaffold cysteines is probably the result of unfolding of the protein upon the destruction of the zinc finger structure that results from the loss of the chelated zinc atom. That this is indeed the case is suggested by the decreased solubility of the mutant proteins and their aberrant chromatographic behavior (D. Cassel, unpublished observations). Limited proteolysis of the catalytically active Arf GAP1(1-146) resulted in a fragment consisting of the first 136 amino acids, presumably representing the complete catalytic core. The crystal structure of the catalytic core of the Arf GAP domain was solved by Goldberg (Goldberg, 1999) and demonstrated a unique fold, not found in GAPs for other families of GTPases. In this structure the zinc finger, composed of β strands, is nested within an array of one β strand and six α helices. Some mechanistic implications of this structure are discussed below (see also Chapter 2).

3.4

Catalytic mechanism

Small GTPases catalyze the hydrolysis of GTP by a conserved mechanism, whereby a glutamine residue presents the water molecule at the active site (Scheffzek et al., 1998). This mechanism has a built-in low catalytic efficiency due to poor positioning of the glutamine residue for catalysis. Structural studies of Ras and Rho GAPs complexed with their respective

ARF GTPASE-ACTIVATING PROTEIN 1

1459

GTPases (Nassar et al., 1998; Rittinger et al., 1997; Scheffzek et al., 1997) showed that the GAPs stimulate GTP hydrolysis in two ways. First, the GAPs bind to the switch 2 regions of the GTPases, acting to stabilize this region and to orient the glutamine residue in a position that is optimal for catalysis. Second, the GAPs possess an “arginine finger” mechanism, whereby an arginine residue is directly introduced into the active site of the GTPase. This arginine residue is thought to facilitate GTP hydrolysis by interacting with the glutamine residue and further orienting it for catalysis, and by interacting with the phosphate groups of GTP, leading to a decrease in their effective charge and thereby assisting in the nucleophilic attack of the water molecule. However, recent studies have indicated that the arginine finger mechanism does not take place in all GTPases, and is absent in Ran and Rap GAPs (Brinkmann et al., 2002; Seewald et al., 2002) The mode of interaction between Arf GAP1 and its substrate has been studied by Goldberg (Goldberg, 1999), who co-crystallized Arf1•GDP and the catalytic core of Arf GAP1. The structure revealed multiple interactions of helix α3 of Arf1 with the GAP catalytic core. A number of interactions were also observed between the GAP catalytic core and switch 2 of Arf1, but the significance of these interactions remained unclear as switch 2 undergoes a significant conformational change upon the GDP-to GTP switch, and presumably the transition state of the GTPase reaction resembles the GTP state of Arf1 rather than the GDP state which was used in this study. The mode of interaction between Arf1 and GAP as observed in the above study has subsequently been challenged (Click et al., 2002; Mandiyan et al., 1999), although an alternative model has yet to be presented. The mechanism that emerged from the structural study by Goldberg suggested that Arf GAP1 does not contribute an arginine residue to facilitate catalysis. The single arginine residue in the GAP catalytic core that is invariant in all known Arf GAPs (Arg50 in Arf GAP1) pointed away from the protein-protein interface. Additionally, the same study demonstrated that coatomer greatly stimulated the activity of the catalytic core of GAP. These findings raised the possibility that coatomer is required to complement the catalytic activity of GAP, possibly by supplying the arginine residue to the active site. The effect of coatomer on GAP activity will be further discussed below. However, site-directed mutagenesis studies by several groups have shown that the invariant arginine residue in Arf GAPs (always positioned 5 residues from the carboxyl-terminal cysteine of the zinc finger (see Figure 1)) is essential for GAP activity. Even conservative replacements with a lysine residue completely abolished catalytic activity of Arf GAP1 (Szafer et al., 2000), and the invariant arginine was also found essential for activity in other Arf GAPs (Jackson et al., 2000; Mandiyan et al., 1999; Miura et al., 2002; Randazzo et al., 2000). These findings suggested that this residue

10 146

CASSEL

might in fact act in an arginine-finger mechanism. Once again, structural studies of the complex between Arf GAPs and their substrate in the GTPbound state will be required in order to resolve this issue, as well as other mechanistic detail of Arf GAP action.

3.5

Regulation of Arf GAP1 activity

3.5.1

Effect of membrane lipids

The substrate of GAP, Arf•GTP, is tightly associated with membrane lipids. This is due to the myristoyl group attached to the amino terminus, as well as the amphipathic amino-terminal helix, that interacts with the protein core in the GDP-bound state and is thus shielded from solvent, but become exposed in the GTP-bound state (Antonny et al., 1997a; Cherfils et al., 1998; Goldberg, 1998). The loading of Arf1 with GTP (in preparation for a GAP assay) can be achieved using a large variety of phospholipids and detergents. By systematically varying the composition of liposomes used during Arf1 loading, Antonny (Antonny et al., 1997b) found that the activity of Arf GAP1(1-257) (a fragment of Arf GAP1 that can be readily expressed in bacteria) is increased in the presence of phospholipids with small head groups such as phosphatidylethanolamine, and is dramatically stimulated upon addition of a small percentage of diacylglycerols bearing at least one unsaturated fatty acid chain. The effects of the different lipid compositions on Arf GAP1 activity closely correlated with the binding of Arf GAP1 to the liposomes. From these results, it was concluded that a hydrophobic domain in GAP can bind to membrane lipids, that this binding is facilitated by the presence of open spaces in the lipid bilayer created by cone-shaped lipids, and is sterically inhibited by the presence of large phospholipid head groups. The good correlation between GAP binding to liposomes and GAP activity suggested a mechanism whereby binding of Arf GAP1 to membrane lipids results in an increase in GAP activity simply due to an increase in the effective GAP concentration in the proximity of its Arf•GTP substrate, the latter being tightly bound to liposomes regardless of their composition. Phosphoinositides, such as PI(4,5)P2, were reported to increase Arf GAP1 activity in some studies (Makler et al., 1995; Vitale et al., 2000), but not in others (Antonny et al., 1997b). How this effect is brought about is uncertain because unlike ASAP- and ARAP-like Arf GAPs, Arf GAP1 does not have a PH domain. It should be borne in mind that the interactions of proteins with lipid vesicles may differ markedly from their interaction with biological membranes. Whereas reported effects of lipids on Arf and Arf GAP activities are suggestive of regulatory events in vivo, the latter are difficult to demonstrate and thus remain uncertain.

ARF GTPASE-ACTIVATING PROTEIN 1 3.5.2

11 147

Role of coatomer

Goldberg has demonstrated that coatomer, which in itself has no effect on GTP hydrolysis on Arf1, dramatically stimulates Arf GAP1-dependent GTP hydrolysis (Goldberg, 1999). In this study GAP activity was assayed in solution, by employing a truncated Arf1 mutant that lacks the first 17 amino acids (∆17-Arf1). Unlike wild-type Arf1 that depends on a lipid environment (Antonny et al., 1997a; Paris et al., 1997) or effectors (Zhu et al., 2000) for GTP binding, ∆17-Arf1 can bind GTP to high stoichiometry in solution (Paris et al., 1997). This is apparently due to an alleviation of a restriction brought about by the amphipathic amino terminus of the wild type Arf1 protein, which becomes exposed in the GTP state and must be shielded by binding to a hydrophobic surface (Amor et al., 1994; Goldberg, 1998). The stimulation of Arf GAP1 activity was demonstrated using the catalytic core of Arf GAP1 (residues 1-136). One obvious conclusion from these findings was that Arf GAP1 and coatomer do not compete for the same binding site on Arf1. Interestingly, the opposite situation was recently reported for the coat adaptors termed GGAs, which compete with GAP for Arf1 binding and thus inhibit GAP activity, thereby stabilizing their binding to membranes (Puertollano et al., 2001). Our group has extended the studies of the role of coatomer to more naturally-occurring conditions, using wild type, myristoylated Arf1 in place of ∆17-Arf1, and wild type, full length Arf GAP1 in place of the catalytic core. When using myr-Arf1 loaded with GTP in the presence of liposomes as substrate, Arf GAP1 activity was higher by two orders of magnitude than that observed with ∆17-Arf1, and this activity was not further increased in the presence of coatomer (Szafer et al., 2000). This suggested that while coatomer strongly stimulates Arf GAP1 activity in a soluble system, its effect vanishes when Arf1 and Arf GAP1 are brought to proximity by their mutual binding to phospholipids (see 3.5.1 above). Additionally, with a different type of Arf GAP, the phosphoinositide-dependent ASAP1 (Brown et al., 1998), coatomer did not increase GAP activity even when using ∆17Arf1 as substrate (Szafer et al., 2000). We concluded from these studies that coatomer might not be an essential component of the mechanism of GTP hydrolysis on Arf1, and that Arf GAPs may possess the entire catalytic machinery of this mechanism. In a subsequent study (Szafer et al., 2001) we analyzed the effect of coatomer using Golgi membranes, the biological site of coatomer assembly. We designed an assay in which myr-Arf1 is loaded with GTP by Golgi membrane-bound GEFs and thus becomes bound to the membrane. In rat liver Golgi membrane preparations endogenous GAP activity is relatively low (particularly so after salt wash), and thus the “spontaneous” rate of

12 148

CASSEL

decay of the GTP state of Arf1 is slow. The effect of external GAPs on GTP hydrolysis can then be readily measured. Arf GAP1 exhibited high activity in this assay, but unlike the liposome assay, addition of coatomer approximately doubled this activity. A similar two-fold effect of coatomer was observed when using the whole cellular repertoire of Arf GAPs, represented by coatomer-depleted liver cytosol. Immunodepletion experiments suggested that although Arf GAP1 contributes significantly to cytosolic GAP activity, most of the activity in cytosol is mediated by additional Arf GAPs under these conditions. The identity of these soluble Arf GAPs is currently unknown. A different pattern was observed with an Arf GAP from S. cerevisiae. This organism contains at least 3 Arf GAPs, termed Gcs1, Glo3 and Age2 (Poon et al., 1999; Poon et al., 2001; Poon et al., 1996). When Glo3 was tested in the Golgi assay, its GAP activity showed large dependency on coatomer, while in the liposome assay Glo3 activity was low both in the absence and presence of coatomer (Szafer et al., 2001). These quantitatively different effects of coatomer on GAP1 and on Glo3 can be rationalized as follows. The above mentioned studies in Golgi membranes suggest that coatomer can increase the efficiency of Arf GAP1 stimulation of GTP hydrolysis in a biological setting. Moreover, evidence for a similar role of a coat in GTP hydrolysis was recently presented for the COPII system, whose mechanisms appear to resemble those of the COPI system (Antonny and Schekman, 2001; Springer et al., 1999). However, unlike in the soluble system (Goldberg, 1999), a recruitment of GAP to its site of action, either by binding to membrane lipids (Antonny et al., 1997b) or by interaction with membrane proteins (Aoe et al., 1997), should provide significant GAP activity also in the absence of a coat complex, as suggested by our experiments with liposomes and Golgi membranes. The role of coatomer would then be to fine-tune the rate of GTP hydrolysis at COPI assembly sites, a process that could play an important role in cargo sorting (see below). In the case of Glo3, the large effect of coatomer on GAP activity might be related to the fact that this protein forms a complex with coatomer (Eugster et al., 2000), which should bring this GAP into proximity with Arf1•GTP. Glo3 is not an ortholog of Arf GAP1, and perhaps unlike Arf GAP1, Glo3 does not interact with membrane lipids, thus explaining its low activity in the presence of liposomes. The reason for the low activity of Glo3 with Golgi membranes is less clear; one possible reason might be the use of a heterologous system consisting of a yeast Arf GAP and a mammalian Golgi. Future studies should reveal whether the mammalian ortholog of Glo3 (Arf GAP3) shows a coatomer-dependent GAP activity on Golgi membranes.

ARF GTPASE-ACTIVATING PROTEIN 1 3.5.3

13 149

Regulation by cytoplasmic domains of p23/24 proteins

p23/24 proteins are abundant membrane proteins that shuttle between the ER, the ER-Golgi intermediate compartment (ERGIC) and cis-Golgi. Multiple members of this family are found in hetero-oligomeric complexes (Emery et al., 2000; Fullekrug et al., 1999; Gommel et al., 1999). They present short cytoplasmic tails, many of which bind coatomer through a diphenylalanine (FF) and/or a dibasic motif (Dominguez et al., 1998; Fiedler et al., 1996; Nickel et al., 1997; Sohn et al., 1996). Recent studies suggest that some of these tails may act as inhibitors of Arf GAP1 activity. The first demonstration of this effect came from Goldberg’s study (Goldberg, 2000), where he showed that a peptide representing the cytoplasmic tail of a p24 protein (with the sequence YLKRFFEVRRVV, termed hp24a) selectively inhibited coatomer-stimulated activity of the catalytic core of Arf GAP1 in solution. The peptide had no effect on GAP activity observed in the absence of coatomer. Several peptides representing cytoplasmic tails of other p24 proteins, as well as p24a variant peptides containing replacements in residues that are critical for coatomer binding, had no effect of coatomerdependent Arf GAP1 activity. Based on these findings, Goldberg suggested a model for cargo sorting on the basis of GTP hydrolysis on Arf1. In this model, the interaction of coatomer with cargo (such as the p24 tails) prevents coatomer stimulation of GAP, resulting in stabilization of Arf1•GTP. In this way the correct match of cargo, coat complex, and Arf is “proofread” and specificity in vesicle assembly can be assured. In the absence of the cargo component, coatomer-dependent Arf GAP activity rapidly hydrolyzes GTP on Arf1, thereby preventing the formation of COPI vesicles that lack cargo. Although the effect on GTP hydrolysis was restricted to p24a, the existence of p24 proteins in hetero-oligomeric complexes could explain how at least some other members of this family may be sorted into COPI vesicles. Recently, Nilsson’s group has reported the results of their studies of the role of GTP hydrolysis in cargo sorting, using an in vitro assay for vesicle budding from liver Golgi membranes (Lanoix et al., 1999; Lanoix et al., 2001). In this system, Golgi enzymes as well as the p23/24 proteins are sorted into vesicles by a COPI-dependent mechanism, which is in accordance with the model of Golgi maturation (Bonfanti et al., 1998; Mironov et al., 2001; Pelham, 2001). Significantly, sorting was observed in the presence of GTP, but not in the presence of the hydrolysis-resistant GTP derivative, GTPγS. Several lines of evidence suggested that p23/24 proteins might be involved in a GTP hydrolysis-dependent sorting event. Addition of low concentrations of certain p23/24 peptides prevented the sorting of all Golgi markers examined. The most active peptide was similar to the one

14 150

CASSEL

found by Goldberg to inhibit coatomer-dependent GAP activity (CQIYYLKRFFEVRRVV, derived from the same p24 protein but named in this study p24β1). When examined for their effect on Arf GAP1, some p24 peptides were found to bind the GAP and to inhibit its activity. Comparison of different p24 peptides, as well as variant p24β1 peptides, revealed a good correlation between Arf GAP1 binding capacity, effect on Arf GAP1 activity, and the ability of the peptides to inhibit cargo sorting. Unlike the study by Goldberg (Goldberg, 2000), the peptides were found in this study to inhibit GAP activity in the absence of coatomer. Possible reasons for these different results are difference in the lengths of the p24 peptides employed, as well as types of assays and conditions employed. The p24 tail-derived peptides used in both studies contain a number of hydrophobic residues as well as multiple positive charges, and could therefore exhibit non-specific effects, particularly so when membranes are concerned. Thus, additional approaches are required in order to establish a role for cytoplasmic tails of p24 proteins in sorting cargo into Golgi-derived vesicles and the role of GAP activity.

3.6

Role of Arf GAP1 in KDEL receptor signaling

The KDEL receptor (KDEL-R) serves as a specialized mechanism for the retrieval of escaped soluble ER-resident proteins, many of which bear the Lys-Asp-Glu-Lys (KDEL) signal (Andres et al., 1991). The receptor is a seven-pass transmembrane protein with a short cytoplasmic tail. It shuttles between the Golgi and the ER, based on a cycle of ligand-induced Golgi to ER transport followed by the release of the ligand in the ER and the recycling of the empty receptor back to the Golgi. The directionality of these cycles is apparently determined by an increased affinity of the receptor for KDEL ligands at the lower pH that prevails at the Golgi lumen as compared with the ER (Wilson et al., 1993). The KDEL-R is transported from the Golgi through COPI vesicles (Majoul et al., 1998; Orci et al., 1997). Studies by Hsu and colleagues have suggested a role of Arf GAP1 in this process. Co-immunoprecipitation experiments demonstrated that the KDEL-R interacts with the GAP, apparently following receptor oligomerization (Aoe et al., 1997). Moreover, using cells harboring a plasmid that allows the regulated expression of a KDEL ligand, Aoe (Aoe et al., 1998) demonstrated that interaction to the KDEL-R with Arf GAP1 depends on ligand binding to the receptor. A noncatalytic domain at the carboxyl-terminal part of Arf GAP1 that is required for the ability of overexpressed GAP to disassemble the Golgi in transfected cells (Aoe et al., 1999) was also required for GAP-KDEL-R interaction. More recently, an interaction between Arf GAP1 and the KDEL-R, as well

ARF GTPASE-ACTIVATING PROTEIN 1

15 151

as a KDEL ligand-induced recruitment of Arf GAP1 to the Golgi, were demonstrated in living cells using fluorescence resonance energy transfer (FRET) (Majoul et al., 2001). What could be the function of Arf GAP1 recruitment to the Golgi membrane following ligand binding to KDEL-R? One possibility is that GAP interaction with the receptor inhibits its catalytic activity. Such (putative) inhibition would promote coatomer assembly onto membrane patches containing the KDEL-R, as suggested for p24 proteins (section 3.5.3). In this way the KDEL-R may act analogously to the p24 proteins to help sort cargo into Golgi-derived, COPI coated vesicles. Another way to envision the role of GAP in cargo sorting is based on lessons from the COPII system (Antonny and Schekman, 2001; Springer et al., 1999). Like the COPI system, coat binding in the COPII system is triggered by the activation of the GTPase, Sar1, at the ER membrane. However, the GAP in the COPII system (the Sec23/24 complex) plays a direct and essential role in the assembly of the coat protein complex, by binding directly to both Sar1 and the other coat components, Sec13/31. Additionally, interactions of Sec23/24 with SNAREs are thought to ensure the inclusion of the fusion machinery in COPII vesicles. There is growing evidence for parallel mechanisms in the two systems. Thus, the yeast Arf GAP Glo3 was shown to interact with coatomer (Eugster et al., 2000), although this has not so far been demonstrated for either Arf GAP1 or its yeast ortholog Gcs1. As in the COPI system, the Sec13/31 coat was found to stimulate GTP hydrolysis mediated by Sec23/24 (Antonny et al., 2001). Yeast Arf GAPs show genetic interactions with SNAREs (Poon et al., 1999), and recent evidence suggests that a transient interaction of yeast Arf GAPs with v-SNAREs may activate the v-SNAREs for Arf1 binding to membranes (Rein et al., 2002). Together, these findings suggest that Arf GAPs could function as an intimate component of the COPI complex. According to this view, cargo such as the KDEL-R could be recruited to COPI vesicles through its interaction with GAP, rather than vice versa. However, there is one obvious distinction between the COPI and COPII systems. Unlike the COPII system, coatomer can bind directly to Arf, without any obvious requirement for GAP (Zhao et al., 1997), and can induce vesicle budding from liposomes in a minimal system containing only Arf1 and the guanine nucleotide (Donaldson et al., 1992a). Thus, further studies will be required to find out how far the analogy between the two systems can be taken.

16 152

4.

CASSEL

CONCLUSION

GTPase-activating proteins acting in the early secretory system appear to function not merely as signal terminators but also as components in key processes of the system such as cargo sorting. While accumulating evidence suggests such a role for Arf GAP1, there are still many questions that need to be answered more definitely. One should also bear in mind that Arf GAP1 may not be the only player at the Golgi, and other Arf GAPs that might fulfill this function have been described (Andreev et al., 1999; Liu et al., 2001; Mazaki et al., 2001; Miura et al., 2002). Likewise, at least two Arf GAPs were reported to function in the ER-Golgi shuttle in S. cerevisiae (Poon et al., 1999; Poon et al., 1996). Thus the specific role of each of these GAPs in COPI-mediated trafficking will have to be further analyzed. The recent advances in the area of membrane traffic, as well as the introduction of modern techniques that allow studies in living cells, seem to promise rapid progress in the understanding of the role of GAPs in this system.

REFERENCES Amor, J. C., Harrison, D. H., Kahn, R. A., and Ringe, D. (1994). Structure of the human ADP-ribosylation factor 1 complexed with GDP. Nature, 372, 704-708. Andreev, J., Simon, J. P., Sabatini, D. D., Kam, J., Plowman, G., Randazzo, P. A., and Schlessinger, J. (1999). Identification of a new Pyk2 target protein with Arf-GAP activity. Mol. Cell. Biol., 19, 2338-2350. Andres, D. A., Rhodes, J. D., Meisel, R. L., and Dixon, J. E. (1991). Characterization of the carboxyl-terminal sequences responsible for protein retention in the endoplasmic reticulum. J. Biol. Chem., 266, 14277-14282. Antonny, B., Beraud-Dufour, S., Chardin, P., and Chabre, M. (1997a). N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry, 36, 4675-4684. Antonny, B., Huber, I., Paris, S., Chabre, M., and Cassel, D. (1997b). Activation of ADPribosylation factor 1 GTPase-activating protein by phosphatidylcholine-derived diacylglycerols. J. Biol. Chem., 272, 30848-30851. Antonny, B., Madden, D., Hamamoto, S., Orci, L., and Schekman, R. (2001). Dynamics of the COPII coat with GTP and stable analogues. Nat. Cell. Biol., 3, 531-537. Antonny, B., and Schekman, R. (2001). ER export: public transportation by the COPII coach. Curr. Opin. Cell. Biol., 13, 438-443. Aoe, T., Cukierman, E., Lee, A., Cassel, D., Peters, P. J., and Hsu, V. W. (1997). The KDEL receptor, ERD2, regulates intracellular traffic by recruiting a GTPase-activating protein for ARF1. EMBO J., 16, 7305-7316. Aoe, T., Huber, I., Vasudevan, C., Watkins, S. C., Romero, G., Cassel, D., and Hsu, V. W. (1999). The KDEL receptor regulates a GTPase-activating protein for ADP- ribosylation factor 1 by interacting with its non-catalytic domain. J. Biol. Chem., 274, 20545-20549.

ARF GTPASE-ACTIVATING PROTEIN 1

17 153

Aoe, T., Lee, A. J., van Donselaar, E., Peters, P. J., and Hsu, V. W. (1998). Modulation of intracellular transport by transported proteins: insight from regulation of COPI-mediated transport. Proc. Natl. Acad. Sci., USA, 95, 1624-1629. Bonfanti, L., Mironov, A. A., Jr., Martinez-Menarguez, J. A., Martella, O., Fusella, A., Baldassarre, M., Buccione, R., Geuze, H. J., Mironov, A. A., and Luini, A. (1998). Procollagen traverses the Golgi stack without leaving the lumen of cisternae: evidence for cisternal maturation. Cell, 95, 993-1003. Brinkmann, T., Daumke, O., Herbrand, U., Kuhlmann, D., Stege, P., Ahmadian, M. R., and Wittinghofer, A. (2002). Rap-specific GTPase activating protein follows an alternative mechanism. J. Biol. Chem., 277, 12525-12531. Brown, M. T., Andrade, J., Radhakrishna, H., Donaldson, J. G., Cooper, J. A., and Randazzo, P. A. (1998). ASAP1, a phospholipid-dependent Arf GTPase-activating protein that associates with and is phosphorylated by Src. Mol. Cell. Biol., 18, 7038-7051. Cherfils, J., Menetrey, J., Mathieu, M., Le Bras, G., Robineau, S., Beraud-Dufour, S., Antonny, B., and Chardin, P. (1998). Structure of the Sec7 domain of the Arf exchange factor ARNO. Nature, 392, 101-105. Click, E. S., Stearns, T., and Botstein, D. (2002). Systematic Structure-Function Analysis of the Small GTPase Arf1 in Yeast. Mol. Biol. Cell., 13, 1652-1664. Cosson, P., and Letourneur, F. (1994). Coatomer interaction with di-lysine endoplasmic reticulum retention motifs. Science, 263, 1629-1631. Cukierman, E., Huber, I., Rotman, M., and Cassel, D. (1995). The ARF1 GTPase-activating protein: zinc finger motif and Golgi complex localization. Science, 270, 1999-2002. Dominguez, M., Dejgaard, K., Fullekrug, J., Dahan, S., Fazel, A., Paccaud, J. P., Thomas, D. Y., Bergeron, J. J., and Nilsson, T. (1998). gp25L/emp24/p24 protein family members of the cis-Golgi network bind both COP I and II coatomer. J. Cell. Biol., 140, 751-765. Donaldson, J. G., Cassel, D., Kahn, R. A., and Klausner, R. D. (1992a). ADP-ribosylation factor, a small GTP-binding protein, is required for binding of the coatomer protein betaCOP to Golgi membranes. Proc. Natl. Acad. Sci., USA, 89, 6408-6412. Donaldson, J. G., Finazzi, D., and Klausner, R. D. (1992b). Brefeldin A inhibits Golgi membrane-catalyzed exchange of guanine nucleotide onto ARF protein. Nature, 360, 350352. Donaldson, J. G., and Jackson, C. L. (2000). Regulators and effectors of the ARF GTPases. Curr. Opin. Cell Biol., 12, 475-482. Emery, G., Rojo, M., and Gruenberg, J. (2000). Coupled transport of p24 family members. J. Cell. Sci., 113 ( Pt 13), 2507-2516. Eugster, A., Frigerio, G., Dale, M., and Duden, R. (2000). COP I domains required for coatomer integrity, and novel interactions with ARF and ARF-GAP. EMBO J., 19, 39053917. Fiedler, K., Veit, M., Stamnes, M. A., and Rothman, J. E. (1996). Bimodal interaction of coatomer with the p24 family of putative cargo receptors. Science, 273, 1396-1399. Fullekrug, J., Suganuma, T., Tang, B. L., Hong, W., Storrie, B., and Nilsson, T. (1999). Localization and recycling of gp27 (hp24gamma3): complex formation with other p24 family members. Mol. Biol. Cell, 10, 1939-1955. Goldberg, J. (1998). Structural basis for activation of ARF GTPase: mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Goldberg, J. (1999). Structural and functional analysis of the ARF1-ARFGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Goldberg, J. (2000). Decoding of Sorting Signals by Coatomer through a GTPase Switch in the COPI Coat complex. Cell, 100, 671-679.

18 154

CASSEL

Gommel, D., Orci, L., Emig, E. M., Hannah, M. J., Ravazzola, M., Nickel, W., Helms, J. B., Wieland, F. T., and Sohn, K. (1999). p24 and p23, the major transmembrane proteins of COPI-coated transport vesicles, form hetero-oligomeric complexes and cycle between the organelles of the early secretory pathway. FEBS Lett., 447, 179-185. Hammonds-Odie, L. P., Jackson, T. R., Profit, A. A., Blader, I. J., Turck, C. W., Prestwich, G. D., and Theibert, A. B. (1996). Identification and cloning of centaurin-alpha. A novel phosphatidylinositol 3,4,5-trisphosphate-binding protein from rat brain. J. Biol. Chem., 271, 18859-18868. Helms, J. B., and Rothman, J. E. (1992). Inhibition by brefeldin A of a Golgi membrane enzyme that catalyses exchange of guanine nucleotide bound to ARF. Nature, 360, 352354. Huber, I., Cukierman, E., Rotman, M., Aoe, T., Hsu, V. W., and Cassel, D. (1998). Requirement for both the amino-terminal catalytic domain and a non-catalytic domain for in vivo activity of ADP-ribosylation factor GTPase-activating protein. J. Biol. Chem., 273, 24786-24791. Huber, I., Rotman, M., Pick, E., Makler, V., Rothem, L., Cukierman, E., and Cassel, D. (2001). Expression, purification, and properties of ADP-ribosylation factor (ARF) GTPase activating protein-1. Methods Enzymol., 329, 307-316. Jackson, T. R., Brown, F. D., Nie, Z., Miura, K., Foroni, L., Sun, J., Hsu, V. W., Donaldson, J. G., and Randazzo, P. A. (2000). ACAPs are arf6 GTPase-activating proteins that function in the cell periphery. J. Cell Biol., 151, 627-638. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kondo, A., Hashimoto, S., Yano, H., Nagayama, K., Mazaki, Y., and Sabe, H. (2000). A New Paxillin-binding Protein, PAG3/Papalpha/KIAA0400, Bearing an ADP-Ribosylation Factor GTPase-activating Protein Activity, Is Involved in Paxillin Recruitment to Focal Adhesions and Cell Migration. Mol. Biol. Cell, 11, 1315-1327. Krugmann, S., Anderson, K. E., Ridley, S. H., Risso, N., McGregor, A., Coadwell, J., Davidson, K., Eguinoa, A., Ellson, C. D., Lipp, P., Manifava, M., Ktistakis, N., Painter, G., Thuring, J. W., Cooper, M. A., Lim, Z. Y., Holmes, A. B., Dove, S. K., Michell, R. H., Grewal, A., Nazarian, A., Erdjument-Bromage, H., Tempst, P., Stephens, L. R., and Hawkins, P. T. (2002). Identification of ARAP3, a novel PI3K effector regulating both Arf and Rho GTPases, by selective capture on phosphoinositide affinity matrices. Mol. Cell, 9, 95-108. Lanoix, J., Ouwendijk, J., Lin, C. C., Stark, A., Love, H. D., Ostermann, J., and Nilsson, T. (1999). GTP hydrolysis by arf-1 mediates sorting and concentration of Golgi resident enzymes into functional COP I vesicles. EMBO J., 18, 4935-4948. Lanoix, J., Ouwendijk, J., Stark, A., Szafer, E., Cassel, D., Dejgaard, K., Weiss, M., and Nilsson, T. (2001). Sorting of Golgi resident proteins into different subpopulations of COPI vesicles: a role for ArfGAP1. J. Cell Biol., 155, 1199-1212. Letourneur, F., Gaynor, E. C., Hennecke, S., Demolliere, C., Duden, R., Emr, S. D., Riezman, H., and Cosson, P. (1994). Coatomer is essential for retrieval of dilysine-tagged proteins to the endoplasmic reticulum. Cell, 79, 1199-1207. Lippincott-Schwartz, J., Cole, N. B., and Donaldson, J. G. (1998). Building a secretory apparatus: role of ARF1/COPI in Golgi biogenesis and maintenance. Histochem. Cell Biol., 109, 449-462. Liu, X., Zhang, C., Xing, G., Chen, Q., and He, F. (2001). Functional characterization of novel human ARFGAP3. FEBS Lett., 490, 79-83.

ARF GTPASE-ACTIVATING PROTEIN 1

19 155

Majoul, I., Sohn, K., Wieland, F. T., Pepperkok, R., Pizza, M., Hillemann, J., and Soling, H. D. (1998). KDEL receptor (Erd2p)-mediated retrograde transport of the cholera toxin A subunit from the Golgi involves COPI, p23, and the COOH terminus of Erd2p. J. Cell Biol., 143, 601-612. Majoul, I., Straub, M., Hell, S. W., Duden, R., and Soling, H. D. (2001). KDEL-cargo regulates interactions between proteins involved in COPI vesicle traffic: measurements in living cells using FRET. Dev. Cell, 1, 139-153. Makler, V., Cukierman, E., Rotman, M., Admon, A., and Cassel, D. (1995). ADP-ribosylation factor-directed GTPase-activating protein. Purification and partial characterization. J. Biol. Chem., 270, 5232-5237. Mandiyan, V., Andreev, J., Schlessinger, J., and Hubbard, S. R. (1999). Crystal structure of the ARF-GAP domain and ankyrin repeats of PYK2-associated protein beta. Embo J., 18, 6890-6898. Mazaki, Y., Hashimoto, S., Okawa, K., Tsubouchi, A., Nakamura, K., Yagi, R., Yano, H., Kondo, A., Iwamatsu, A., Mizoguchi, A., and Sabe, H. (2001). An ADP-ribosylation factor GTPase-activating protein Git2-short/KIAA0148 is involved in subcellular localization of paxillin and actin cytoskeletal organization. Mol. Biol. Cell, 12, 645-662. Mironov, A. A., Beznoussenko, G. V., Nicoziani, P., Martella, O., Trucco, A., Kweon, H. S., Di Giandomenico, D., Polishchuk, R. S., Fusella, A., Lupetti, P., Berger, E. G., Geerts, W. J., Koster, A. J., Burger, K. N., and Luini, A. (2001). Small cargo proteins and large aggregates can traverse the Golgi by a common mechanism without leaving the lumen of cisternae. J. Cell Biol., 155, 1225-1238. Miura, K., Jacques, K. M., Stauffer, S., Kubosaki, A., Zhu, K., Hirsch, D. S., Resau, J., Zheng, Y., and Randazzo, P. A. (2002). ARAP1: a point of convergence for Arf and Rho signaling. Mol. Cell, 9, 109-119. Nassar, N., Hoffman, G. R., Manor, D., Clardy, J. C., and Cerione, R. A. (1998). Structures of Cdc42 bound to the active and catalytically compromised forms of Cdc42GAP. Nat. Struct. Biol., 5, 1047-1052. Nickel, W., Sohn, K., Bunning, C., and Wieland, F. T. (1997). p23, a major COPI-vesicle membrane protein, constitutively cycles through the early secretory pathway. Proc. Natl. Acad. Sci., USA, 94, 11393-11398. Orci, L., Stamnes, M., Ravazzola, M., Amherdt, M., Perrelet, A., Sollner, T. H., and Rothman, J. E. (1997). Bidirectional transport by distinct populations of COPI-coated vesicles. Cell, 90, 335-349. Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997). Role of protein-phospholipid interactions in the activation of ARF1 by the guanine nucleotide exchange factor Arno. J. Biol. Chem., 272, 22221-22226. Pelham, H. R. (2001). Traffic through the Golgi apparatus. J. Cell Biol., 155, 1099-1101. Peyroche, A., Antonny, B., Robineau, S., Acker, J., Cherfils, J., and Jackson, C. L. (1999). Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell, 3, 275-285. Poon, P. P., Cassel, D., Spang, A., Rotman, M., Pick, E., Singer, R. A., and Johnston, G. C. (1999). Retrograde transport from the yeast Golgi is mediated by two ARF GAP proteins with overlapping function. EMBO J., 18, 555-564. Poon, P. P., Nothwehr, S. F., Singer, R. A., and Johnston, G. C. (2001). The Gcs1 and Age2 ArfGAP proteins provide overlapping essential function for transport from the yeast transGolgi network. J. Cell Biol., 155, 1239-1250.

20 156

CASSEL

Poon, P. P., Wang, X., Rotman, M., Huber, I., Cukierman, E., Cassel, D., Singer, R. A., and Johnston, G. C. (1996). Saccharomyces cerevisiae Gcs1 is an ADP-ribosylation factor GTPase- activating protein. Proc. Natl. Acad. Sci., USA, 93, 10074-10077. Premont, R. T., Claing, A., Vitale, N., Freeman, J. L., Pitcher, J. A., Patton, W. A., Moss, J., Vaughan, M., and Lefkowitz, R. J. (1998). beta2-Adrenergic receptor regulation by GIT1, a G protein-coupled receptor kinase-associated ADP ribosylation factor GTPase-activating protein. Proc. Natl. Acad. Sci., USA, 95, 14082-14087. Premont, R. T., Claing, A., Vitale, N., Perry, S. J., and Lefkowitz, R. J. (2000). The GIT family of ADP-ribosylation factor GTPase-activating proteins. Functional diversity of GIT2 through alternative splicing. J. Biol. Chem., 275, 22373-22380. Puertollano, R., Randazzo, P. A., Presley, J. F., Hartnell, L. M., and Bonifacino, J. S. (2001). The GGAs promote ARF-dependent recruitment of clathrin to the TGN. Cell, 105, 93-102. Randazzo, P. A., Andrade, J., Miura, K., Brown, M. T., Long, Y. Q., Stauffer, S., Roller, P., and Cooper, J. A. (2000). The Arf GTPase-activating protein ASAP1 regulates the actin cytoskeleton. Proc. Natl. Acad. Sci., USA, 97, 4011-4016. Rein, U., Andag, U., Duden, R., Schmitt, H. D., and Spang, A. (2002). ARF-GAP-mediated interaction between the ER-Golgi v-SNAREs and the COPI coat. J. Cell Biol., 157, 395404. Rittinger, K., Walker, P. A., Eccleston, J. F., Smerdon, S. J., and Gamblin, S. J. (1997). Structure at 1.65 A of RhoA and its GTPase-activating protein in complex with a transition-state analogue. Nature, 389, 758-762. Rothman, J. E., and Wieland, F. T. (1996). Protein sorting by transport vesicles. Science, 272, 227-234. Scheffzek, K., Ahmadian, M. R., Kabsch, W., Wiesmuller, L., Lautwein, A., Schmitz, F., and Wittinghofer, A. (1997). The Ras-RasGAP complex: structural basis for GTPase activation and its loss in oncogenic Ras mutants. Science, 277, 333-338. Scheffzek, K., Ahmadian, M. R., and Wittinghofer, A. (1998). GTPase-activating proteins: helping hands to complement an active site. Trends Biochem. Sci., 23, 257-262. Schekman, R., and Orci, L. (1996). Coat proteins and vesicle budding. Science, 271, 15261533. Seewald, M. J., Korner, C., Wittinghofer, A., and Vetter, I. R. (2002). RanGAP mediates GTP hydrolysis without an arginine finger. Nature, 415, 662-666. Sohn, K., Orci, L., Ravazzola, M., Amherdt, M., Bremser, M., Lottspeich, F., Fiedler, K., Helms, J. B., and Wieland, F. T. (1996). A major transmembrane protein of Golgi-derived COPI-coated vesicles involved in coatomer binding. J. Cell Biol., 135, 1239-1248. Springer, S., Spang, A., and Schekman, R. (1999). A primer on vesicle budding. Cell, 97, 145-148. Szafer, E., Pick, E., Rotman, M., Zuck, S., Huber, I., and Cassel, D. (2000). Role of coatomer and phospholipids in GTPase-activating protein- dependent hydrolysis of GTP by ADPribosylation factor-1. J. Biol. Chem., 275, 23615-23619. Szafer, E., Rotman, M., and Cassel, D. (2001). Regulation of GTP hydrolysis on ADPribosylation factor-1 at the Golgi membrane. J. Biol. Chem., 276, 47834-47839. Tanigawa, G., Orci, L., Amherdt, M., Ravazzola, M., Helms, J. B., and Rothman, J. E. (1993). Hydrolysis of bound GTP by ARF protein triggers uncoating of Golgi- derived COPcoated vesicles. J. Cell Biol., 123, 1365-1371. Teal, S. B., Hsu, V. W., Peters, P. J., Klausner, R. D., and Donaldson, J. G. (1994). An activating mutation in ARF1 stabilizes coatomer binding to Golgi membranes. J. Biol. Chem., 269, 3135-3138.

ARF GTPASE-ACTIVATING PROTEIN 1

21 157

Turner, C. E., Brown, M. C., Perrotta, J. A., Riedy, M. C., Nikolopoulos, S. N., McDonald, A. R., Bagrodia, S., Thomas, S., and Leventhal, P. S. (1999). Paxillin LD4 motif binds PAK and PIX through a novel 95-kD ankyrin repeat, ARF-GAP protein: A role in cytoskeletal remodeling. J. Cell Biol., 145, 851-863. Vitale, N., Patton, W. A., Moss, J., Vaughan, M., Lefkowitz, R. J., and Premont, R. T. (2000). GIT proteins, A novel family of phosphatidylinositol 3,4, 5-trisphosphate-stimulated GTPase-activating proteins for ARF6. J. Biol. Chem., 275, 13901-13906. Wilson, D. W., Lewis, M. J., and Pelham, H. R. (1993). pH-dependent binding of KDEL to its receptor in vitro. J. Biol. Chem., 268, 7465-7468. Yu, S., and M. G. Roth. (2002). Casein kinase-I regulates membrane-binding by ARF GAP1. Mol. Biol. Cell, 13, 2559-2570. Zhang, C., Yu, Y., Zhang, S., Liu, M., Xing, G., Wei, H., Bi, J., Liu, X., Zhou, G., Dong, C., Hu, Z., Zhang, Y., Luo, L., Wu, C., Zhao, S., and He, F. (2000). Characterization, chromosomal assignment, and tissue expression of a novel human gene belonging to the ARF GAP family. Genomics, 63, 400-408. Zhao, L., Helms, J. B., Brugger, B., Harter, C., Martoglio, B., Graf, R., Brunner, J., and Wieland, F. T. (1997). Direct and GTP-dependent interaction of ADP ribosylation factor 1 with coatomer subunit beta. Proc. Natl. Acad. Sci., USA, 94, 4418-4423. Zhao, L., Helms, J. B., Brunner, J., and Wieland, F. T. (1999). GTP-dependent binding of ADP-ribosylation factor to coatomer in close proximity to the binding site for dilysine retrieval motifs and p23. J. Biol. Chem., 274, 14198-14203. Zhu, X., Boman, A. L., Kuai, J., Cieplak, W., and Kahn, R. A. (2000). Effectors increase the affinity of ADP-ribosylation factor for GTP to increase binding. J. Biol. Chem., 275, 13465-13475.

Chapter 8 GIT PROTEINS: ARF GAPS AND SIGNALING SCAFFOLDS ROBERT SCHMALZIGAUG AND RICHARD PREMONT Department of Medicine (Gastroenterology), Duke University Medical Center, Durham, USA

Abstract:

1.

GIT proteins are GTPase-activating proteins (GAPs) for the ADP-ribosylation factor (Arf) family of GTPases. The two GIT family members, GIT1 and GIT2, stimulate GTP hydrolysis on all the known Arf subtypes about equally well, and this activity is stimulated by PI(3,4,5)P3. As such, GITs are negative regulators of the cellular functions of Arfs, such as intracellular vesicle traffic and cytoskeletal rearrangement. In addition, GITs form the core of a large protein complex involved in integrating signals through multiple pathways, involving G protein-coupled receptors, receptor tyrosine kinases, integrins, and at least two classes of small GTP-binding proteins, the Arfs and Rac/Cdc42, Rho families.

INTRODUCTION

GIT proteins are members of the fast-growing family of Arf GAP domaincontaining proteins, which presently has 22 known human members (Figure 1). Based on overall sequence similarity and domain organization, the members of this family can be divided into two major subgroups that contain subfamilies with multiple members. We classify these into the smaller GAPs (Arf GAP1, Arf GAP3/Zfp289, RIP/Hrb, SMAP1/2, centaurinα1/α2), and the large, multidomain GAPs (GITs, ASAP1/PAP, ACAPs, ARAPs, AGAPs). These proteins all share the ~100 amino acid CX2CX16CX2C zinc finger-containing Arf GAP domain, although several members have yet to be directly tested for GAP activity. In another classification scheme, the centaurin subfamily of Arf GAPs has been proposed for those Arf GAPs also containing ankyrin repeats and PH domains, and includes two centaurin-α proteins, ASAP1/PAP (centaurinβ’s), ACAPs (also centaurin-β’s), AGAPs (centaurin-γ’s) and ARAPs 159

2 160

SCHMALZIGAUG AND PREMONT

(centaurin-δ’s) (Jackson et al., 2000a). In all these GAPs (except the centaurin-α subgroup), the GAP domain is found centrally in the protein. In contrast to this centaurin group, GITs do not have identifiable PH domains, and share the amino terminal GAP domain localization with the small GAP proteins and centaurin-α’s. GITs have a multidomain structure and participate in multiple intracellular signaling events (as detailed below). This suggests that other, less well-characterized multidomain GAP family members will also be found to participate in multiple cellular signaling pathways.

Figure 1. Mammalian Arf GAP Family Domain Structure. PBS, paxillin binding sequence; PH, pleckstrin homology domain; Rho GAP, GTPase-activating protein domain for Rho family small GTPases; SH3 Src homology type 3 domain; Spa2, region of homology with yeast Spa2p and Sph1p

2.

GITS: DISCOVERY AND NOMENCLATURE

Over the past few years, GITs have been identified in several distinct contexts by many laboratories. This has led to a wealth of information about GITs on the one hand, and to multiple names and a confusing literature on the other hand. In short, there are only two GIT genes in humans or mouse, and two identified GIT sequences from chicken. These sequences have been called GIT1/Cat-1/p95-APP1 and GIT2/KIAA0148/Cat-2/p95-PKL/p95-APP2. These are summarized in Table 1.

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

1613

Table 1. GIT protein nomenclature. °Mouse Cat-1 was partially peptide sequenced, but not cloned or fully sequenced. *PKL was originally reported as being a human cDNA sequence, but has since been shown to be from chicken and identical to p95-APP2. GIT1 GIT2 Species Reference GIT1 Rat Premont et al., 1998 GIT1 GIT2 Human Premont et al., 2000 (Cat-1)º Cat-2 Mouse Bagrodia et al., 1999 PKL Chicken* Turner et al., 1999 p95APP1 Chicken Di Cesare et al., 2000 p95APP2 Chicken Paris et al., 2002 KIAA0148 Human Nagase et al., 1996

The initial discovery of the GIT proteins came from a search for G protein-coupled receptor kinase 2 (GRK2) interacting proteins. In a yeast two-hybrid screen, GRK2 was employed as the bait and a rat cDNA library served as the prey. One positive clone was found to encode the C-terminus of a previously unknown protein that was named GRK interactor 1, or GIT1 (Premont et al., 1998). The full-length rat GIT1 sequence was found to contain an N-terminal zinc finger-like motif with similarity to the then recently identified rat Arf GAP1 GAP domain (Cukierman et al., 1995), suggesting that it too was a GAP for Arf proteins. Searching DNA databases for rat GIT1-related sequences revealed two classes of GIT1-related sequences: the human KIAA0148 sequence encoding a putative protein of unknown function with high similarity to the first two-thirds of GIT1 (Nagase et al., 1995), and single human and mouse ESTs with similarity to the extreme C-terminus of GIT1. Reverse transcriptase PCR revealed that these two sequences were in fact derived from the same alternatively spliced gene, and then named GIT2 (Premont et al., 2000). GITs were next found in connection with two small GTPases, Rac1 and Cdc42, and their effectors, the p21-activated protein kinases (PAKs). Bagrodia et al. (1999) immunoprecipitated PAK3 from mouse fibroblasts and co-purified a tyrosine phosphorylated protein of 95 kDa. Analysis of the protein revealed an indirect binding to PAK3 via the PIX/Cool protein, a guanine nucleotide exchange factor (GEF) for Rac and Cdc42. Because the protein directly bound to Cool and was found to be tyrosine phosphorylated, it was called Cool-associated tyrosine phosphorylated protein 1 or Cat-1. From the limited peptide sequences obtained from this purified protein, it is likely mouse GIT1. cDNA cloning based on Cat-1 peptide sequences yielded the related mouse Cat-2 cDNA, the mouse ortholog of human GIT2 (Bagrodia et al., 1999). Zhao et al. (2000b) used a similar strategy to identify GIT1 as a binding partner of PIX/Cool and PAK complexes. Both GIT1 and GIT2 bind to PIX proteins, and indirectly to PAKs (Bagrodia et al., 1999; Zhao et al., 2000b; Premont et al., 2000). Another study investigated the link between membrane transport and Rac-mediated actin

4 162

SCHMALZIGAUG AND PREMONT

reorganization by identifying proteins that specifically bound to Rac•GTP. Two proteins, p95-APP1 and p95-APP2 (95 kDa, Arf GAP-putative, PIXinteracting, Paxillin-interacting proteins), were isolated (Di Cesare et al., 2000; Paris et al., 2002). These are the chicken orthologs of mammalian GIT1 and GIT2. Yet another group applied a GST-paxillin pull-down assay to isolate paxillin-interacting proteins, and identified a protein that linked paxillin to the PIX/PAK complex, which they named p95-PKL (paxillin-kinase linker) (Turner et al., 1999). PKL binds to a very small region of paxillin, the LD4 motif, and this motif directs the localization of PKL (and the PIX/PAK complex) to focal adhesions. The PKL clone was reported to come from a human HeLa cell cDNA library, and originally was thought to encode a third mammalian GIT family member. Despite much effort, no authentic PKL sequences could be amplified from human cDNA, and no human gene for PKL could be mapped or found in database searches (R. T. Premont, unpublished observations). The PKL cDNA is identical to the chicken p95APP2 cDNA, and has since been acknowledged to be derived from a chicken cDNA library (Turner et al., 2001). Thus, PKL/p95-APP2 is the chicken ortholog of mammalian GIT2. Another study sought to find proteins that interact with paxillin specifically in the perinuclear region (Mazaki et al., 2001). From a GST-paxillin pulldown experiment, the shortest splice variant of GIT2 (GIT2-short, KIAA0148) was isolated. GIT2-short co-localized with paxillin in the perinuclear region but not in focal adhesions (Mazaki et al., 2001). An elegant screen for substrates of the receptor-type protein tyrosine phosphatase ζ (PTPζ) used a yeast substrate-trapping screen, where proteins were phosphorylated by over-expressed Src and trapped by an inactive mutant of PTPζ (Kawachi et al., 2001). GIT1 is a known Src substrate (Bagrodia et al., 1999), and was found to interact with the mutant PTPζ and to be a substrate for dephosphorylation by active PTPζ. As GIT1 phosphorylation by protein kinases is associated with focal adhesions and cell attachment, this finding potentially completes the regulatory cycle of GIT1 phosphorylation by extracellular stimuli. In summary, GIT proteins have been identified multiple times in very different biological contexts. The current challenge is to organize these diverse functional roles into a better understanding of the physiological function of these proteins.

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

3.

GIT GENES, SPLICING AND TISSUE DISTRIBUTION

3.1

Genes and Genomic location

1635

Human GIT1 and GIT2 proteins are encoded by two genes, located at chromosome bands 17p11.2 and 12q24.1, respectively (Premont et al., 2000). The two mouse GIT genes are located on chromosomes 1 and 5 (R. T. Premont, unpublished observation). No additional GIT-like sequences were detected in the human or mouse genomes. Chickens also appear to have two GIT genes (Di Cesare et al., 2000; Paris et al., 2002). To date, neither GIT gene locus has been associated with any disease. The human GIT1 gene consists of 20 coding exons distributed over 15 kb. The mouse GIT1 gene has 19 exons, as there is no intron separating the sequences equivalent to human exons 13 and 14. The human and mouse GIT2 genes have an identical structure of 21 coding exons spread over 50 kb. The location of introns is highly conserved between the two GIT genes, with 18 introns inserted in equivalent positions in the two human and mouse sequences (R. T. Premont, unpublished observations). In GIT1, an extra exon is present in a non-conserved part of the C-terminus, and in GIT2, an additional intron is present within the region encoding the second ankyrin repeat, and the 21st exon encodes the C-terminal nine amino acids of the GIT2-short splice variant (see below).

3.2

Alternative splicing

GIT1 appears to be expressed only as the full-length protein, as splice variants have not been found in human or rat tissues using RT-PCR (Premont et al., 2000). However, the rat GIT1 sequence contains a nine amino acid insertion after residue 253, as compared to the human, mouse and chicken GIT1 sequences, due to a short extension at the junction of exons 7 and 8. Thus, rat GIT1 is 770 residues long, while human and mouse GIT1 have 761 residues. In contrast, human GIT2 mRNA is expressed as multiple splice variants (Premont et al., 2000), and chicken GIT2/PKL was suggested to have multiple splice variants as well (Turner et al., 1999). Analysis of mRNA from multiple human tissues suggests that the GIT2 transcript undergoes tissue-specific, alternative splicing. The longest form of the protein, called GIT2-long, is co-linear with GIT1 and uses 20 exons, but alternative splicing leads to two classes of shorter GIT2 variants, in which either internal sequences appear to be deleted or the C-terminus is truncated (see Figure 2). First, direct sequencing of eight distinct human GIT2 cDNAs identified five

6 164

SCHMALZIGAUG AND PREMONT

contiguous and independently spliced blocks of sequence between residues 333 and 578 (originally termed sequences A through E) that are deleted in individual clones. Sequence A corresponds to exons 12 and 13, B plus C correspond to exon 14 (which appears to have an internal splice acceptor site in the codon for residue 449), and sequences D and E match exons 15 and 16, respectively. Totally independent splicing of these five blocks would potentially lead to 32 internally spliced variants. In the second class of alternative splicing, exon 14 is spliced to a distinct exon (i.e., exon 21, shown as D’ in Figure 2) that leads to early translation termination, producing the shortest known variant of GIT2 (called GIT2-short or KIAA0148) that lacks just over 300 amino acids at the C-terminus (including the major paxillin binding site). However, it remains unknown whether all GIT2 mRNA splice variants are translated into distinct variants of the GIT2 protein. Other groups have identified GIT2-long from chicken (i.e., p95-APP2/PKL) (Turner et al., 1999; Di Cesare et al., 2000), and one shorter GIT2 version (i.e., mouse Cat-2) that lacks sequences B and C (exon 14) (Bagrodia et al., 1999). The GIT2-short/KIAA0148 variant has been reported as well (Nagase et al., 1995; Mazaki et al., 2001).

Figure 2. Domains identified in GITs and alternative splicing of GIT2.

3.3

GIT tissue distribution

The mRNAs encoding GIT proteins appear to be ubiquitously expressed in rat, human and chicken (Premont et al., 1998; Premont et al., 2000; Paris et al., 2002). Low levels of GIT1 mRNA are found in liver and spleen, medium levels in lung, skeletal muscle, heart and kidney and highest levels in testes and brain (Premont et al., 1998; Premont et al., 2000). Direct RTPCR amplification of GIT2 splice variants from a variety of human tissues suggests that this message undergoes extensive tissue-specific alternative splicing, as detailed above. GIT2 mRNA is expressed at moderate to high levels in many tissues including the heart, brain, kidney and spleen (Nagase et al., 1995; Premont et al., 2000). The GIT2-short variant in particular was found at high levels in spleen, testes and leukocytes (Premont et al., 2000).

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

1657

The presence of native GIT1 protein has only been reported for brain and testes, as well as the HeLa, COS7 and CHOK1 cell lines (Zhao et al., 2000b; Manabe et al., 2002). The GIT2-short protein has been detected in several cell lines (U937, HeLa and NIH3T3 cells), with the highest levels in the immunoderived Jurkat cell line (Mazaki et al., 2001). Neither a comprehensive study of GIT1 and GIT2 protein expression in various tissues, nor the identification of most of the multiple predicted GIT2 protein variants has been reported so far. The main limitation to such a study has been the lack of specific antibodies able to distinguish between GIT1 and GIT2. Specific antibodies are crucial because the full-length forms have the same molecular weight. Antibodies are commercially available from Santa Cruz Biotechnology and Transduction Laboratories. The GIT1 and GIT2 antibodies from Santa Cruz do not appear to be very avid and give only faint signals in immunoblots, although they do distinguish recombinant GIT1 and GIT2. The GIT1 monoclonal antibody from Transduction Labs has good specificity for GIT1 versus GIT2, but only weakly recognizes endogenous levels of protein. The PKL monoclonal antibody from Transduction Labs (raised against chicken GIT2) is fairly avid but not at all specific for PKL/GIT2 versus GIT1 (R. T. Premont, unpublished observations). Thus, for example, it remains unclear which GIT protein(s) is present in the Jurkat T cell line, as detected with the PKL monoclonal antibody (Ku et al., 2001).

4.

GIT STRUCTURE AND DOMAINS

4.1

GAP domain

In GIT1, the GAP domain was the first functional domain characterized. It was found by similarity to the two known Arf GAPs at the time, rat Arf GAP1 and yeast Gcs1. All proteins that have been shown to have Arf GAP activity share a conserved domain containing a CX2CX16CX2C zinc finger (Figure 1). A second feature essential for GAP activity is an arginine residue that has been suggested to serve as the ‘arginine finger’ that enhances the rate of GTP hydrolysis by several orders of magnitude (Scheffzyk et al., 1998). Mutation of either of the first two cysteine residues of the zinc finger abolished GAP activity of Arf GAP1 (Cukierman, et al., 1995). Mutation of this arginine residue in other Arf GAPs (Arf GAP1, ASAP1/PAPα, ACAPs, ARAP1) results in ablation of GAP activity (Mandiyan et al., 1999; Randazzo et al., 2000; Jackson et al., 2000a; Szafer et al., 2000; Miura et al., 2002). The GAP domain of GIT1 is located at the very N-terminus of the protein and contains both the cysteine residues of the zinc finger followed by the

8 166

SCHMALZIGAUG AND PREMONT

essential arginine residue five amino acids downstream of the zinc finger (Figure 2). The GAP domain of GIT2 has the conserved cysteines and arginine at the exact same positions as in GIT1 (Premont et al., 2000). Purified recombinant GIT1 and GIT2 contain stoichiometric amounts of zinc (Vitale et al., 2000), and the X-ray crystal structure of the GAP domains of rat Arf GAP1 and human β-PAP show a single zinc atom bound by the four cysteine residues (Goldberg, 1999; Mandiyan et al., 1999). Arf GAP1, GIT1 and GIT2 have very similar GAP activities, with GIT1 being slightly more potent than Arf GAP1 and GIT2, using Arf1 as a substrate (Vitale et al., 2000). GITs promote GTP hydrolysis in all three subtypes of Arfs with similar efficiency and thus do not show any obvious enzymatic specificity. In contrast, Arf GAP1 stimulated the GTP hydrolysis on Arf1-5, but had little effect on Arf6 (Vitale et al., 2000). The distinct stimulation of Arfs by the different Arf GAP subtypes introduces a level of specificity within the large GAP family. The GAP activity of GIT1 is regulated by the phosphatidylinositiol 3,4,5trisphosphate (PI(3,4,5)P3) in vitro. The presence of PI(3,4,5)P3 increases GAP activity of GIT1 3.5-fold and of GIT2 2-fold (Vitale et al., 2000). The components of the in vitro assay were limited to the Arf1 and GIT proteins and vesicles carrying PI(3,4,5)P3. Thus, it is likely that the stimulatory effect of PI(3,4,5)P3 comes from lipid-dependent localization of Arf and GIT to the vesicles. Stimulation of Arf GAP activity through a direct effect on the Arf itself by PI(3,4,5)P3 is unlikely, as the GAP activity of Arf GAP1 is not promoted by PI(3,4,5)P3 under identical assay conditions (Vitale et al., 2000). However, GIT proteins do not have an obvious PH domain to bind to PI(3,4,5)P3 (or PIP2) like many other Arf GAPs (Figure 1) (Jackson et al., 2000b). The mechanism of regulation of the GIT GAP activity by PI(3,4,5)P3 remains elusive. Perhaps GITs can interact with PI(3,4,5)P3 through sequences distinct from the PH domain, as reported for neurogranin (Lu et al., 1997). It is interesting to note that various Arf GAPs are stimulated by distinct sets of phosphoinositides and their metabolites; e.g. Arf GAP1 activity is stimulated by PIP2 and diacylglycerol, ASAP1, β-PAP and ACAPs by PIP2, and ARAPs by PI(3,4,5)P3 (Antonny et al., 1997; Randazzo, 1997, Vitale et al., 2000; Jackson et al., 2000a; Krugmann et al., 2002, Miura et al., 2002). The regulation of GAPs by distinct phosphoinositides adds another level of specificity to the regulation of the Arf GTPases by their GAPs. The first GAP-dead mutant of GIT1 was made by deleting the 45 Nterminal residues, including the cysteine residues of the zinc finger and the conserved arginine residue. GIT1-¨45 failed to activate GTPase activity of Arf1 (Premont et al., 1998). The single mutation of the conserved arginine 39 (to alanine) abolished the Arf GAP activity of GIT1 and GIT2 as well (R.

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

1679

T. Premont, unpublished observations). In addition, the first cysteine residue of the zinc finger in GIT2-short has been mutated to alanine. This mutation abolished the Arf GAP activity of GIT2-short (Mazaki et al., 2001). Therefore, both the arginine and the zinc finger are required for the GAP activity of GIT proteins.

4.2

Ankyrin repeats

GITs contain three ankyrin repeats immediately adjacent to the GAP domain. In the Arf GAP1 and β-PAP structures, the zinc finger and the immediately downstream region (which is the ankyrin repeats in PAP) form a single structural domain, with the downstream region serving as a foundation for the zinc finger (Goldberg, 1999; Mandiyan et al., 1999). This suggests a structural role in GIT proteins as well. To date, no precise function has been described for these motifs in GITs, although they may also function as protein - protein interaction sites. A predicted coiled-coil motif partially overlapping this region has been suggested to be a secondary paxillin binding site, particularly in the GIT2-short splice variant (Turner et al., 1999; Mazaki et al. 2001), but this has not been verified through mutagenesis.

4.3

Spa2 homology domain

In the yeast Spa2 and Sph1 proteins, two short, near repeat units that also are predicted to form a coiled-coil appear to mediate binding to the yeast MEK proteins Mkk1, Mkk2 and Ste7 (Sheu et al., 1998; Roemer et al., 1998). GIT1 and GIT2 contain Spa2 homology domains (SHD), consisting of two 30 amino acid near-repeat units separated by 30 residues. This region appears to be the site by which GITs bind tightly to α- and β-PIX (Zhao et al., 2000b), guanine nucleotide exchange factors (GEFs) for Rac1/Cdc42 as well as major binding partners for the Rac1/Cdc42 p21-activated protein kinases (PAKs) (Manser et al., 1998). GIT1 fragments containing the SHD bind to β-PIX in protein overlay assays (Zhao et al., 2000b). The GIT binding site was identified as the Cterminal region of β-PIX using a variant of β-PIX (p50-Cool1) lacking the C-terminal 200 residues, which failed to bind (Bagrodia et al., 1999), and in two-hybrid assays (Premont et al., 2000). The GIT binding site was later mapped to a 60 residue sequence within the C-terminus of β-PIX by direct binding in overlay assays (Zhao et al., 2000b). Recently, a valine to alanine point mutation within this region of β-PIX was reported to decrease GIT1 binding dramatically (Feng et al., 2002). In addition to PIX binding, the SHD of GIT1 has been reported to interact directly with focal adhesion

10 168

SCHMALZIGAUG AND PREMONT

kinase (FAK) (Zhao et al., 2000b). The direct demonstration that these SHDs mediate specific protein-protein interactions is currently lacking and will make an important contribution to our understanding of signaling by GIT1 and other SHD-containing proteins.

4.4

Coiled-coil dimerization domain

A predicted coiled-coil region is found following the Spa2 homology domain in GIT1, and is conserved in GIT2. Recent experiments have shown that GITs dimerize (R. T. Premont, unpublished observation), although an earlier report suggested that they do not (Zhao et al., 2000b). Flag-tagged GIT1 is able to efficiently co-immunoprecipitate 6xHis-tagged GIT1 or GIT2, and GIT1 interacts directly with GIT1 or GIT2 in yeast two-hybrid assays. The smallest GIT1 fragments able to co-immunoprecipitate 6xHistagged GIT1 or endogenous GIT2 contained this coiled-coil domain. GIT1 fragments containing the coiled-coil region also interacted with endogenous PIX proteins, independently of the Spa2 homology region, suggesting that GIT proteins contain two distinct PIX binding sites (R. T. Premont, unpublished observations). β-PIX has been reported to dimerize through the interaction of a coiledcoil region (Kim et al., 2001). Indeed, α-PIX and β-PIX can form homoand heterodimers (R. T. Premont, unpublished observation). GITs also homo- and heterodimerize, and bind tightly to PIX proteins through two distinct sites, the SHD and coiled-coil region (R. T. Premont, unpublished observation). Therefore, there is the potential for GIT-PIX complex oligomerization rather than simple heterodimerization. Knowing that all binding sites for GIT1 are contained in the final 186 amino acids of β-PIX, we prepared a β-PIX C-terminal (CT) fragment as a GIT–PIX interaction inhibitor. Although Flag-tagged β-PIX CT co-immunoprecipitated GIT1 as expected, HA-tagged β-PIX CT failed to inhibit the interaction of Flagtagged GIT1 with β-PIX, but simply added into the immunoprecipitated GIT – PIX complex (R. T. Premont, unpublished observation). Multiple recent studies have suggested models for regulated GIT – PIX interaction where these two proteins are assumed to dissociate (Turner et al., 1999; Zhao et al., 2000b; Di Cesare et al., 1999; Nishiya et al., 2002). Our model for GIT – PIX interaction, in contrast, suggests that GIT and PIX proteins are tightly associated oligomers, and form the non-dissociable core of a complex of signaling proteins. Indeed, separation of cellular proteins on a gel filtration column reveals that endogenous GIT1 and β-PIX proteins co-migrate at an extremely high molecular size, far larger than even a GIT/PIX heterotetramer. We hypothesize that GIT and PIX proteins are always associated in this large complex, which is critical to both GIT and PIX

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

11 169

functions, and suggest that other multidomain Arf GAPs might also behave similarly with their own specific partners.

4.5

Paxillin binding site

Another prominent feature of GIT proteins is binding to paxillin. Paxillin binding in vitro and in vivo was initially discovered with chicken PKL/GIT2 (Turner et al., 1999). Comparisons to the paxillin binding sites of FAK and vinculin revealed two putative paxillin binding sites in GIT2. One putative site (PBS1) was predicted between the GAP domain and the first ankyrin repeat, while the second site (PBS2) was predicted in the C-terminal half of GIT2, within another coiled-coil region. Crude mapping suggested that the C-terminal half of PKL contained the binding site, and deletion of each putative site identified the C-terminal PBS2 as the major paxillin-binding domain (West et al., 2001). The deleted C-terminal site included 35 amino acids and represents the smallest paxillin-binding site reported in GIT proteins so far. Surprisingly, human GIT2-short, which lacks this major paxillin-binding region, was also found to bind paxillin (Mazaki et al., 2001). Deletion mutants of GIT2-short identified a region around the Cterminal end of the GAP domain and the first ankyrin repeat, which correlates with the PBS1 site described above (Mazaki et al., 2001). In rat GIT1, the paxillin-binding site was mapped to the 120 C-terminal residues, and deletion of this region abolished paxillin binding to GIT1 (Zhao et al., 2000b). However, the paxillin binding site in GIT1 has not been further refined. A direct comparison of mammalian GIT1 and GIT2 suggested that GIT1 binds more tightly to paxillin than does GIT2 (Premont et al., 2000). It has been reported that GIT1 binding to paxillin can be enhanced by coexpression of β-PIX, leading to the suggestion that GIT-PIX interaction modulates the interaction of GIT proteins with paxillin (and thus with focal adhesions) (Zhao et al., 2000b). However, in our hands, co-expression of either PIX protein with GIT1 or GIT2 has no apparent effect on the amount of co-immunoprecipitated paxillin, and we believe that GIT and PIX proteins are constitutively associated (R. T. Premont, unpublished). The GIT binding site on paxillin is the LD4 motif (LDXLLXXL), which is also a binding site for FAK, vinculin and clathrin heavy chain (Tumbarello et al., 2002). Deletion of the paxillin LD4 motif prevented the paxillin - GIT interaction and retarded cell migration (Turner et al., 1999). GIT1 (Nishiya et al., 2002) and GIT2 (Turner et al., 1999) also bind to the paxillin family member Hic-5 through its cognate LD3 motif. The regulation of GIT binding to paxillin, and localization to focal adhesions, is an active area of interest, because of the apparent role of this interaction in cell attachment to matrix and cell motility (see below).

12 170

4.6

SCHMALZIGAUG AND PREMONT

Phosphorylation

GIT1 phosphorylation was first observed when tyrosine-phosphorylated Cat1 was pulled down with PAK3 and activated Cdc42 from fibroblast lysates (Bagrodia et al., 1999). Tyrosine phosphorylation of GIT1 is regulated in a cell cycle-dependent manner. GIT1 is highly phosphorylated in a random cycling population of cells, but is not phosphorylated in cells blocked in mitosis. When cells are released into G1 phase, GIT1 rapidly becomes phosphorylated (Bagrodia et al., 1999). Cell adhesion through integrins also regulates GIT1 phosphorylation. Detachment of cells from the substratum abolishes GIT1 phosphorylation, while plating cells on fibronectin induces GIT1 phosphorylation (Bagrodia et al., 1999). FAK and Src can each enhance GIT1 phosphorylation in cells, and have been suggested as the primary GIT1 kinases (Bagrodia et al., 1999; Zhao et al., 2000b; Kawachi et al., 2001). In contrast, little is known about GIT2 phosphorylation. As described above, GIT1 is a substrate of PTPζ (Kawachi et al., 2001). PTPζ activity is regulated by its ligand, pleiotrophin. Binding of pleiotrophin to PTPζ inactivates the phosphatase and increases the tyrosine phosphorylation of its substrates, i.e., β-catenin (Meng et al., 2000). Neuroblastoma cells expressing GIT1 were treated with pleiotrophin, leading to increased GIT1 tyrosine phosphorylation and the conclusion that PTPζ acts on GIT1 in vivo. GIT1 co-localized with PTPζ in cells, e.g., both GIT1 and PTPζ are expressed in processes of hippocampal and cortical pyramidal neurons (Kawachi et al., 2001).

5.

GIT PROTEIN CELLULAR LOCALIZATION

Subcellular fractionation of cultured cells expressing endogenous GIT2 or recombinant GIT1 or GIT2 suggests that the GIT proteins are primarily soluble, with only a small fraction of the protein associated with the cytoskeleton or cellular membranes (R. T. Premont, unpublished observations). Endogenous CHO-K1 cell ‘PKL’ (likely GIT1 based on the use of a GIT1-specific antiserum in these cells by Manabe et al. (2002)) and 3Y1 cell GIT2-short have also been reported to be primarily soluble (Mazaki et al., 2001; Brown et al., 2002). Expression of green fluorescent proteintagged GIT proteins in various cell types leads to a non-specific cytosolic expression pattern in most of the cells. However, in a subset of cells, GIT proteins display striking patterns of localization (Figure 3). Thus, under circumstances that are generally only poorly understood, GITs can be found associated with distinct cellular structures. This has led to suggestions that GITs are highly regulated and involved in transiently localizing associated

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

13 171

proteins to certain cellular structures, and in transport between distinct cellular compartments (Turner et al., 1999; Di Cesare et al., 1999; Manabe et al., 2002). These regulated localizations are discussed below.

Figure 3. Cellular localization of GIT1/EGFP in an unusually flattened, migrating HeLa cell. Three distinct patterns are evident in this cell. GIT is found at the membrane periphery in membrane ruffles at the leading edge of the cell, in small finger-like focal adhesions just behind the spreading leading edge, and in a diffuse cytoplasmic distribution in the thicker part of the cell near the nucleus.

5.1

Focal adhesions

Focal adhesions are integrin-dependent sites linking the extracellular matrix to the actin-rich cytoskeleton. Their formation is influenced particularly by the RhoA, Cdc42 and Rac1 GTPases. Integrins form the transmembrane link to several focal complex proteins, thereby directly or indirectly recruiting vinculin, talin, α-actinin, focal adhesion kinase and paxillin. GITs co-localize with paxillin in focal adhesions at the ends of actin stress fibers (Turner et al., 1999; Zhao et al., 2000b; Manabe et al., 2002) and provide the link allowing PIX and PAK proteins to transiently associate with paxillin-containing focal adhesions. PKL/GIT2 interacts directly with the LD4 motif of paxillin (Turner et al., 1999). In cells expressing a paxillin mutant lacking the LD4 motif, GIT2 could no longer localize to focal adhesions (Brown et al., 2002). The GIT2-short isoform, which lacks the Cterminal paxillin binding site, does not localize to focal adhesions (Mazaki et al., 2001). GIT1 also binds paxillin and co-localizes with paxillin in focal adhesions (Premont et al., 2000; Zhao et al., 2000b; Manabe et al., 2002).

14 172

SCHMALZIGAUG AND PREMONT

Deletion of the C-terminal 120 amino acids (including the paxillin binding site) of GIT1 caused the loss of paxillin binding and localization to focal adhesions, while C-terminal fragments containing the paxillin-binding site did localize to focal adhesions (Zhao et al., 2000b; Manabe et al., 2002; Brown et al., 2002). The interaction between paxillin and GIT1 is weak compared to the GITPIX interaction, as overexpressed GIT1 can only pull-down a small fraction of endogenous paxillin (Premont et al., 2000; Zhao et al., 2000b). Paxillin binds GIT2 weaker than GIT1 (Premont et al., 2000), and the GIT2-short variant even weaker (Mazaki et al., 2001). The weakness of these interactions likely results from assays being performed on a mixed population of cells, in which most cells were not spreading or migrating and thus did not have GIT/PIX/PAK complexes stably bound to focal adhesions. Co-expression of GIT1 and β-PIX was reported to enhance paxillin binding to GIT1 (Zhao et al., 2000b). However, studies in my lab cannot confirm this enhancing effect of PIX on the GIT-paxillin interaction in the same cell system (R. T. Premont, unpublished observation). This putative interactionenhancing effect of PIX on the GIT-paxillin interaction has not been shown to lead to enhanced localization to focal adhesions in cells. Thus, the effect of PIX on GIT localization remains unclear. Enhanced GIT1 localization to focal adhesions has been shown in three cases. In the first case, expression of the PAK auto-inhibitory domain peptide, PAK83-149, led to increased localization of PAK to focal adhesions (Zhao et al., 2000a). The autoinhibitory domain eliminates PAK autophosphorylation and prevents full activation (Zhao et al., 1998). The autoinhibitory domain also enhanced the localization of GIT1 to focal adhesions (Zhao et al., 2000b). The same effect of the PAK auto-inhibitory domain was observed with GIT2 in CHO cells (Brown et al., 2002). In contrast, a constitutively active form of PAK prevented GIT2 localization to focal adhesions (Brown et al., 2002). These experiments provide evidence that PAK kinase activity is an important regulator of the localization of GIT proteins to focal adhesions. In the other two cases, GIT localization to focal adhesions was stimulated by constitutively active forms of the GTPases Rac1 (Manabe et al., 2002; Brown et al., 2002) and Cdc42 (Brown et al., 2002). Constitutively active Rac1 enhanced GIT1 and GIT2 localization to focal adhesions, while dominant negative Rac1 had no effect on GIT localization (Manabe et al., 2002; Brown et al., 2002). In addition, constitutively active Cdc42, but not RhoA, enhanced GIT2 localization (Brown et al., 2002). Overall, GIT proteins (in complex with PIX and presumably PAK) are recruited to paxillin in focal adhesions both as the adhesions are forming and as they are dissociating. Little GIT (or PIX or PAK) appears associated with

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

15 173

mature, stable adhesions. Cell spreading (after mitosis or replating) or cell migration, conditions that activate Rac1 and Cdc42, stimulate GIT/PIX/PAK localization to focal adhesions. These conditions are also associated with phosphorylation of GIT1 on tyrosine residues, although the relationship between GIT phosphorylation, GIT binding to paxillin, and GIT localization to focal adhesions remains unclear. In contrast, activity of PAK appears associated with disassembly of focal adhesions and loss of GIT localization.

5.2

Membranes and vesicles

GITs have been found associated with distinct plasma membrane domains. In chicken embryonic fibroblasts, overexpressed GIT1/p95-APP1 localized to the plasma membrane at sites of ruffling, along with actin, β1-integrin and Arf6 (Di Cesare et al., 1999). GIT1 also localized to membrane ruffles in neuroblastoma cells (Kawachi et al., 2001). Treatments stimulating ruffle formation localized a C-terminal fragment of GIT1 to the plasma membrane, where it co-localized with actin (Matafora et al., 2001). In COS7 cells, GIT1 co-localized with paxillin to the plasma membrane (Zhao et al., 2000b). GIT1 and a fragment consisting of the paxillin-binding site localized to membrane ruffles at the leading edge of CHO cells (Manabe et al., 2002). GIT2-short was found with actin and paxillin at the plasma membrane (Mazaki et al., 2001). GIT1 was also found in mobile, punctate structures in the cytosol (Manabe et al., 2002). Localization to these structures required the presence of the Spa2 homology domain of GIT1. Using specific markers of intracellular compartments, these structures were shown to be of neither endosomal nor Golgi nature. In addition to GIT1, these structures also contained PIX, PAK and paxillin. In migrating cells, the structures moved from the cytoplasm towards a pre-existing focal adhesion at the leading edge of the cells. The reverse movement was found at the retracting side of the cells, where the GIT1-containing structures moved away from adhesions towards the cell center (Manabe et al., 2002). The authors conclude that this represents a transport vesicle carrying GIT/PIX/PAK complexes to and from sites of focal adhesion assembly and disassembly. Overexpressed GIT fragments have been found associated with large cytoplasmic vacuoles that are not seen in untransfected cells. Both Nterminal and C-terminal fragments of GIT1 containing the SHD can be found in such structures (Di Cesare et al., 1999). The large vacuoles were later identified as large Rab11-positive vesicles (Matafora et al., 2001). Rab11 is a marker for pericentriolar recycling endosomes. Deletion or inactivation of the GAP domain also induced accumulation of GIT1 to Rab11-positive structures (Matafora et al., 2001). These findings led to a

16 174

SCHMALZIGAUG AND PREMONT

model for GIT1 function in membrane recycling from the endocytic compartment through recycling endosomes (Matafora et al., 2001). Much like GIT1, C-terminal fragments of GIT2 lacking GAP activity localized to large vesicles (Paris et al., 2002). In contrast to co-localization of GIT1 with Rab11 in the large vesicles, GAP-deficient fragments of GIT2 co-localized with EEA1, a marker for the early endosomal compartment (Paris et al., 2002). However, it remains unclear whether endogenous GIT proteins actually traverse or function in these intracellular compartments, and whether the apparently distinct localizations of GIT1 and GIT2 reflect distinct roles in these two compartments.

6.

GIT FUNCTIONS

6.1

Receptor endocytosis and GRKs

Agonist activation of G protein-coupled receptors (GPCRs) leads first to the activation of coupled heterotrimeric G proteins, which activate a multitude of intracellular messenger-mediated signaling pathways (Lefkowitz, 2000; Pierce et al., 2002). The second result of agonist activation of receptor is the activation and membrane recruitment of G protein-coupled receptor kinases (GRKs), which phosphorylate the activated receptors. Phosphorylated receptors then bind to β-arrestins, leading to uncoupling of the receptor from G protein signaling. This results in desensitization of the receptor to further stimulation. GRK phosphorylation of activated receptors and subsequent arrestin binding also promote internalization of the inactivated receptor proteins from the cell surface, facilitating their resensitization by dephosphorylation in an intracellular acidic vesicle compartment, prior to return to the cell surface (see Chapter 15 for additional details; Lefkowitz, 1998; Pitcher et al., 1998). As GIT1 was originally isolated as a GRK-interacting protein, its role in GRK-mediated GPCR regulation was studied first. GIT1 interacts with the four widely expressed GRKs: GRK2, GRK3, GRK5 and GRK6. GIT1 is not phosphorylated by GRK2 and the kinase activity of GRK2 is not regulated by GIT1 binding in vitro. However, agonist-dependent phosphorylation of the β2-adrenergic receptor (β2AR) was increased in cells overexpressing GIT1. GIT1 also affected β2AR signaling, as detected by reduced receptorstimulated cAMP production, and inhibited the agonist-stimulated internalization of the β2AR in living cells (Premont et al., 1998). GIT2 overexpression also has the ability to reduce β2AR internalization (Premont et al., 2000). As GIT1 was unable to increase β2AR phosphorylation by GRK2 in vitro, the increased phosphorylation in cells is likely due to the

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

17 175

inhibition of receptor internalization and subsequent dephosphorylation. Similarly, the inhibitory effect on adenylyl cyclase stimulation might arise from the desensitized receptors remaining at the plasma membrane and failing to be resensitized by dephosphorylation (Premont et al., 1998). Although GIT1 and GRK2 were found to interact in a yeast two-hybrid screen, GIT1 binding to the β2AR by GRK2 has not been confirmed in vivo. With the two proteins not affecting each other biochemically, localization is still the best potential function of the interaction. GIT1 was tested for its specificity to regulate internalization of various GPCRs (Claing et al., 2000). GIT1 expression does inhibit the sequestration of many receptors, but not all. The specificity of the GIT1 effect does not correlate with the G protein to which a receptor couples, since GIT1 inhibited internalization of some receptors coupled to Gs, Gi or Gq. GPCRs are sequestered from the cell surface by several distinct pathways. The predominant pathway is the clathrin-coated pit pathway. Activated receptors are phosphorylated by GRKs and bind to β-arrestins. In addition to uncoupling receptors from G proteins, β-arrestins also serve as adaptors to link receptors to be internalized with clathrin-coated pits, through β-arrestin interaction with clathrin itself and with the AP-2 adaptor protein complex (Lefkowitz, 1998, 2000). This pathway regulates β2AR endocytosis, among others, and is sensitive to mutants of β-arrestin, clathrin and dynamin II, as well as several chemical inhibitors of endocytosis. Receptors displaying GIT1-dependent regulation of internalization also turned out to be sensitive to mutations in β-arrestin and dynamin. Furthermore, a drug known to stabilize clathrin cage assembly and hence inhibit internalization affected only the class of receptors sensitive to GIT1. Thus, GIT1 affects only the class of G protein-coupled receptors that use clathrin-coated pits as their route to endocytic internalization (Claing et al., 2000). The function of GIT1 on receptor internalization is not restricted to GPCRs, as EGFR sequestration is also regulated by GIT1 (Claing et al., 2000). In contrast to GPCRs and the EGFR, which require ligand binding for sequestration, the transferrin receptor internalizes continuously and independently of ligand binding (Claing et al., 2000). Because GIT1 overexpression did not affect the uptake of the transferrin receptor, this suggested that the GIT1-promoted defect in receptor internalization is specific for recruiting agonist activated receptors to clathrin-coated pits, rather than inhibiting pit formation or endocytosis per se. The functional effects of GIT1 on β2AR (and other receptor) sequestration and function appear to be a result of its ability to stimulate the GTPase activity of Arfs or interact with Arfs, as deletion of the Arf GAP domain ablated the inhibitory effect of GIT1 on β2AR-stimulated cAMP synthesis and receptor sequestration (Premont et al., 1998). Several other

18 176

SCHMALZIGAUG AND PREMONT

Arf GAPs have been tested and do not alter β2AR sequestration (Claing et al., 2001), suggesting a unique role for GIT proteins (perhaps through GRK interaction) in regulating receptor trafficking. Arfs function primarily as GTP-dependent regulators of vesicular traffic through vesicle coat recruitment. Arfs are involved in regulating plasma membrane endocytosis, from experiments in which transferrin receptor traffic is inhibited by Arf6 mutants unable to cycle between the GTP and GDP-bound states (D’SouzaSchorey et al., 1995). The results with GITs suggested that Arfs, and most likely Arf6, may be involved in regulating clathrin-coated pit-mediated internalization of many GPCRs. Direct testing of these Arf6 mutants revealed that they also inhibited the agonist-dependent sequestration of the β2AR (Claing et al., 2001). Further, overexpression of the Arf GEF, ARNO, which activates Arf6 in the cell, led to increased sequestration of the β2AR (Claing et al., 2001). Thus, it appears that GITs and Arf6 are key regulators of the recruitment of activated receptors (cargo) to clathrin-coated pits on the cell surface for their subsequent internalization (see Chapter 15).

6.2

Cell motility and focal adhesions

Cell migration requires the continuous formation and disassembly of cellular adhesion sites at opposing poles of the cell. The basic migration cycle includes extension of a protrusion, formation of stable attachments near the leading edge of the protrusion, translocation of the cell body forward and release of adhesions and retraction at the cell rear. Focal adhesions are relatively stable structures and tend to inhibit cell migration (Webb et al., 2002). In addition to their structural role in linking the extracellular matrix with the actin cytoskeleton, focal adhesions also serve as centers for cellular signaling. Many important signaling molecules, particularly non-receptor tyrosine kinases like Src, FAK and Pyk2, are highly localized to focal adhesions, and participate in both “outside-in” and “inside-out” signaling through integrins. Through paxillin, GITs link a large complex of proteins (PIX/PAK/POPX/Rac1/Cdc42) to focal adhesions. GITs have been reported to affect cell migration. Overexpressed GIT1 promoted cell migration towards a fibronectin and collagen gradient (Zhao et al., 2000b) and increased both the number of migrating cells and their migration speed (Manabe et al., 2002). In contrast, the GIT-associated proteins PIX and PAK were less effective or had no effect on cell migration, respectively. Myristoylated fragments of GIT1 that lack the GAP and Spa2 homology domains stimulated cell migration to a similar degree as full GIT1, while a GIT1 mutant lacking the paxillin binding site did not stimulate migration (Zhao et al., 2000b). Thus, the paxillin binding but not the Arf GAP activity of GITs appears to drive cell migration.

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

19 177

Overexpressed GIT1 can also stimulate paxillin dissociation from focal adhesions but not adhesion disassembly, since cells with decreased levels of localized paxillin still contain normally localized vinculin, another marker of focal adhesions (Zhao et al., 2000b). The ability of GIT1 to stimulate cell migration appears to correlate with its ability to promote the loss of paxillin from adhesions. Because paxillin dissociation is thought to be a marker for focal adhesion remodeling, GIT1 has been suggested to promote this process (Zhao et al., 2000b). GIT2-short was also able to increase the number of cells migrating and their speed of migration, even though it lacks the C-terminal, major paxillin binding site (Mazaki et al., 2001). In contrast to GIT1, expression of GIT2short caused the loss of paxillin only from central (perinuclear) but not from peripheral adhesions (Mazaki et al., 2001). In this case, vinculin staining was limited to peripheral adhesions as well. GIT2-short expression abolished the perinuclear accumulation of paxillin and reduced the number of actin stress fibers (Mazaki et al., 2001). The effects of GIT2-short on paxillin, vinculin and actin could be reversed by mutation to ablate its Arf GAP activity. Furthermore, constitutively active Arf1 partially restored the localization of paxillin to focal adhesions and actin stress fibers (Mazaki et al., 2001). The precise mechanism by which GITs affect cell migration remains unknown, but presumably involves its role in localizing PIX and PAK to paxillin-containing focal adhesions during new adhesion formation at the leading edge of the cell and old focal adhesion disassembly at the rear of the cell, as well as coordinating vesicle trafficking required for leading edge protrusion.

6.3

Cell spreading and protrusions

Changes in cell shape occur during cell spreading and the formation of protrusions, and require changes in the cytoskeleton, the formation or disassembly of focal adhesions and membrane expansion driven by intracellular vesicle fusion with the leading edge of the cell. Cell spreading appeared unaffected by GIT1 or a C-terminal fragment consisting of the coiled-coil domain and the paxillin-binding site, but these same proteins were found to promote the formation of cell protrusions (Di Cesare et al., 1999; Manabe et al., 2002). Similarly, GIT2 lacking the PBS2 paxillin binding region increased cell protrusions (West et al., 2001). However, the paxillin- and Hic-5-binding fragment GIT1-CT was able to block Hic-5induced inhibition of cell spreading (Nishiya et al., 2002). The formation of protrusions could be prevented by blockade of Rac cycling by dominant active or negative mutants of Rac (Di Cesare et al., 2000). Co-expression of

20 178

SCHMALZIGAUG AND PREMONT

full-length GIT1 with Arf6 also inhibited cell spreading and caused cells to retract (Di Cesare et al., 1999). Myristoylated GIT1, and particularly in combination with β-PIX, induced a severe retraction leading to grossly elongated cells resembling neurons (Zhao et al., 2000b). Overall, these studies suggest that GITs are important in regulating a Rac- and PAKdependent pathway that requires paxillin-dependent localization, to initiate cell protrusions. Although cell spreading appears to use the same mechanisms, it is much less grossly affected by GIT proteins.

6.4

Signaling

PIX and PAK interact with each other but bind paxillin indirectly (Manser et al., 1998). PKL/GIT2 was then discovered to bind directly to paxillin, allowing PAK and PIX to indirectly link to paxillin through GITs (Turner et al., 1999). This GIT/PIX/PAK complex has been found in several situations and cell lines, and can contain GIT1 or GIT2, α-PIX or β-PIX, and PAK1, PAK2 or PAK3 (Bagrodia et al., 1999; Premont et al., 2000, Di Cesare et al., 1999; Ku et al., 2001; Turner et al., 1999). This GIT/PIX/PAK complex can bind to active Rac1•GTP but not Rac1•GDP (Di Cesare et al., 1999, Paris et al., 2002). One report, however, suggests that PAK binds directly to paxillin (Hashimoto et al., 2001), although a mutant of PAK that cannot interact with PIX also fails to co-immunoprecipitate paxillin (R. T. Premont, unpublished observation). The GIT/PIX/PAK complex provides a versatile signaling module to cells. This complex contains a GAP for Arfs, a GEF for the Rac1/Cdc42 GTPases, and a serine/threonine kinase regulated by Rac1/Cdc42. Activated Rac1/Cdc42 can also associate with the complex, and the complex can activate these small GTP-binding proteins. This suggests that recruitment of this complex to sites within the cell will lead to a local shift from Arf- to Rac1/Cdc42-dependent pathways, as Arf6 would be deactivated while Rac1/Cdc42 would be activated. As mentioned earlier, our recent experiments indicate that GIT and PIX proteins form a tightly associated oligomeric complex containing multiple GIT and PIX monomers (R. T. Premont, unpublished observation). This oligomer appears to be at least a heterotetramer, and may well be far larger. One GIT monomer does not bind to only one PIX monomer, and the tight association of this oligomeric GIT–PIX complex suggests that these proteins likely do not dissociate in the cell but form a stable multiprotein complex. In any case, one clear result of this oligomeric model is that GIT–PIX complexes should function by spatially augmenting the activity of the GIT (Arf GAP) and PIX (Rac1/Cdc42 GEF) proteins through the colocalization of multiple enzyme

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

21 179

units, as well as function to facilitate the co-localization of multiple copies of the various GIT and PIX binding partners. Activated PAKs can bind the SH2-SH3 adaptor protein Nck, allowing PAK association with other signaling and adaptor proteins (McCarty, 1998). α-PIX binds the p85 regulatory subunit of phosphatidylinositol 3-kinase, itself an SH2-SH3 adaptor protein, in response to PDGF activation, and associates with the activated PDGF receptor itself (Yoshii et al., 1999). Since PIX also binds to the POPX protein phosphatases, that dephosphorylate and inactivate PAKs (Koh et al., 2002), these phosphatases may also be found in a GIT-containing PIX/PAK/POPX complex, although this has not yet been formally demonstrated. Other GIT and PIX binding partners currently being identified will only add to the complexity. This suggests that the confluence of these multiple signaling pathways, and multiple copies of each molecule, play an important role in regulating crosstalk among these ‘distinct’ pathways.

7.

FUTURE PROSPECTS

GITs have been identified in several distinct cellular contexts. The major challenge is in understanding how these apparently distinct functions are integrated. The realization that GIT and PIX proteins are tightly associated in oligomeric complexes raises many questions about the role of this association in both what had been thought to be the distinct functions of the GIT and PIX proteins, and in how this association regulates and integrates the separate activities of these two GTPase regulators. Further experiments are necessary to understand the many facets of this complex, including: defining the structure of the complex, identifying the full set of associated proteins, understanding the regulation of the localization of the GIT–PIX complex and its associated proteins, and defining the crosstalk among the many signaling pathways that intersect within or around this complex. With regard to regulation of Arf proteins, it seems clear that GITs mainly regulate Arf6 in the cell. This stems primarily from co-localization, rather than any enzymatic specificity of the GAP domain for any particular Arf subtype. The GAP activity is activated in vitro by PI(3,4,5)P3, which is an intracellular messenger activated by many signaling pathways. The cellular regulation of GITs by PI(3,4,5)P3 and its role in GIT function in vivo remains to be shown. It seems clear that GIT and Arf6 are important in regulating the formation of vesicles containing activated receptors from cell surface clathrin-coated pits. Whether GITs have other roles in transport processes is unknown, although recent video microscopy of GIT complexes shuttling about in cells (Manabe et al., 2002) suggests that this is true.

22 180

SCHMALZIGAUG AND PREMONT

A major open question is the role of Arf regulation in other activities of the GITs. It appears that the GAP activity is required neither for cell motility effects, nor for binding to paxillin at focal adhesions. On the other hand, one study suggests that Arfs directly promote paxillin recruitment to focal adhesions through an unknown mechanism (Norman et al., 1998). Whether the Arf GAP activity of GITs is important for cellular localization and signaling roles of GITs, or whether GITs play a primarily scaffolding function, remains to be determined. It is also interesting to consider whether GITs function in part as Arf effectors, by transiently localizing associated proteins to active Arf6 at the plasma membrane or elsewhere. In a broader view, it will be interesting to determine whether other members of the Arf GAP family also share functional and structural similarity with the GITs. In this regard, we have been able to demonstrate that several other Arf GAP proteins also appear to dimerize, suggesting that dimerization is an important aspect for the proper function of the multidomain Arf GAPs. The tight association of GIT and PIX proteins also hints that other multidomain GAPs might also have tightly-associated protein partners (preliminary evidence suggests that PIX does not play such a role for several other GAP proteins), and form the core of distinct complexes of signaling and scaffolding proteins. Much further work will be required to fully elucidate the answers to these questions.

REFERENCES Antonny, B., Huber, I., Paris, S., Chabre, M., and Cassel, D. (1997). Activation of ADPribosylation factor 1 GTPase-activating protein by phosphatidylcholine-derived diacylglycerols. J. Biol. Chem., 272, 30848-30851. Bagrodia, S., Bailey, D., Lenard, Z., Hart, M., Guan, J. L., Premont, R. T., Taylor, S. J., and Cerione, R. A. (1999). A tyrosine-phosphorylated protein that binds to an important regulatory region on the Cool family of p21-activated kinase-binding proteins. J. Biol. Chem., 274, 22393-22400. Brown, M. C., West, K. A., and Turner, C. E. (2002). Paxillin-dependent paxillin kinase linker and p21-activated kinase localization to focal adhesions involves a multi-step activation pathway. Mol. Biol. Cell, 13, 1550-65. Claing, A., Perry, S. J., Archilio, M., Walker, J. K. L., Albanesi, J. P., Lefkowitz, R. J., and Premont, R. T. (2000). Multiple endocytic pathways of G protein-coupled receptors delineated by GIT1 sensitivity. Proc. Natl. Acad. Sci., USA, 97, 1119-24. Claing, A., Chen, W., Miller, W. E., Vitale, N., Moss, J., Premont, R. T., and Lefkowitz, R. J. (2001). β-Arrestin-mediated ADP-ribosylation factor 6 activation and β2-adrenergic receptor endocytosis. J. Biol. Chem., 276, 42509-13. Cukierman, E., Huber, I., Rotman, M., and Cassel, D. (1995). The ARF1 GTPase-activating protein: zinc finger motif and Golgi complex localization. Science, 270: 1999-2002. de Curtis, I. (2001). Cell migration: GAPs between membrane traffic and the cytoskeleton. EMBO Reports, 2, 277-281

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

23 181

Di Cesare, A., Paris, S., Albertinazzi, C., Dariozzi, S., Andersen, J., Mann, M., Longhi, R., and de Curtis, I. (2000). p95-APP1 links membrane transport to Rac-mediated reorganization of actin. Nature Cell Biol., 2, 521-530. D’Souza-Schorey, C., Li, G., Colombo, M. I., and Stahl, P. D. (1995). A regulatory role for ARF6 in receptor-mediated endocytosis. Science, 267, 1175-1178. Feng, Q., Albeck, J. G., Cerione, R. A., and Yang, W. (2002). Regulation of the Cool/Pix proteins. J. Biol. Chem., 277, 5644-5650. Goldberg, J. (1999). Structural and functional analysis of the ARF1-ARFGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Hashimoto, S., Tsubouchi, A., Mazaki, Y., and Sabe, H. (2001). Interaction of paxillin with p21-activated kinase (PAK). J. Biol. Chem., 276, 6037-6045. Jackson, T. R., Brown, F. D., Nie, Z., Miura, K., Foroni, L., Sun, J., Hsu, V. W., Donaldson, J. G., and Randazzo, P. A. (2000a). ACAPs are Arf6 GTPase-activating proteins that function in the cell periphery. J. Cell. Biol., 151, 627-638. Jackson, T. R., Kearns, B. G., and Theibert, A. B. (2000b). Cytohesins and centaurins: mediators of PI 3-kinase-regulated Arf signaling. Trends Biochem. Sci., 25, 489-495. Kawachi, H., Fujikawa, A., Maeda, N., and Noda, M. (2001). Identification of GIT1/Cat-1 as a substrate molecule of protein tyrosine phosphatase ζ/β by the yeast substrate-trapping system. Proc. Natl. Acad. Sci., USA, 98, 6593-6598. Kim, S., Lee, S.-H., and Park, D. (2001). Leucine zipper-mediated homodimerization of the p21-activated kinase-interacting factor, βPIX. J. Biol. Chem., 276, 10581-10584. Koh, C.-G., Tan, E.-J., Manser, E., and Lim, L. (2002). The p21-activated kinase PAK is negatively regulated by POPX1 and POPX2, a pair of serine/threonine phosphatases of the PP2C family. Current Biol., 12, 317-321. Krugmann, S., Anderson, K. E., Ridley, S. H., Risso, N., McGregor, A., Coadwell, J., Davidson, K., Eguinoa, A., Ellson, C. D., Lipp, P., Manifava, M., Ktistakis, N., Painter, G., Thuring, J. W., Cooper, M. A., Lim, Z.-Y., Holmes, A. B., Dove, S. K., Michell, R. H., Grewal, A., Nazarian, A., Erdjument-Bromage, H., Tempst, P., Stephens, L. R., and Hawkins, P. T. (2002). Identification of ARAP3, a novel PI3K effector regulating both Arf and Rho GTPases, by selective capture on phosphoinositide affinity matrices. Mol. Cell, 9, 95-108. Ku, G. M., Yablonski, D., Manser, E., Lim, L., and Weiss, A. (2001). A PAK1-PIX-PKL complex is activated by the T-cell receptor independent of Nck, Slp-76 and LAT. EMBO J., 20, 457-465. Lefkowitz, R.J. (1998). G protein-coupled receptors. J. Biol. Chem., 273, 18677-18680. Lefkowitz, R. J. (2000). The superfamily of heptahelical receptors. Nature Cell Biol., 2, E133-E136. Lu, P.J., and Chen, C.S. (1997). Selective recognition of phosphatidylinositol 3,4,5trisphosphate by a synthetic peptide. J. Biol. Chem., 272, 466-472. Manabe, R., Kovalenko, M., Webb, D., and Horwitz, A. R. (2002). GIT1 functions in a motile, multi-molecular signaling complex that regulates protrusive activity and cell migration. J. Cell Sci., 115, 1497-1510. Mandiyan, V., Andreev, J., Schlessinger, J., and Hubbard, S. R. (1999). Crystal structure of the ARF-GAP domain and ankyrin repeats of PYK2-associated protein beta. EMBO J., 18, 6890-6898. Manser, E., Loo, T.-H., Koh, C.-G., Zhao, Z.-S., Chen, X.-Q., Tan, L., Tan, I., Leung, T., and Lim, L. (1998). PAK kinases are directly coupled to the PIX family of nucleotide exchange factors. Mol. Cell, 1, 183-192. Matafora, V., Paris, S., Dariozzi, S., and de Curtis, I. (2001). Molecular mechanisms regulating the subcellular localization of p95-APP1 between the endosomal recycling

24 182

SCHMALZIGAUG AND PREMONT

compartment and sites of actin organization at the cell surface. J. Cell Sci., 114, 45094520. Mazaki, Y., Hashimoto, S., Okawa, K., Tsubouchi, A., Nakamura, K., Yagi, R., Yano, H., Kondo, A., Iwamatsu, A., Mizoguchi, A., and Sabe, H. (2001). An ADP-ribosylation factor GTPase-activating protein GIT2-short/KIAA0148 is involved in subcellular localization of paxillin and actin cytoskeletal organization. Mol. Biol. Cell, 12, 645-662. McCarty, J. H. (1998). The Nck SH2/SH3 adaptor protein: a regulator of multiple intracellular signal transduction events. BioEssays, 20, 913-921. Meng, K., Rodriguez-Pena, A., Dimitrov, T., Chen, W., Yamin, M., Noda, M., and Deuel, T.F. (2000). Pleiotrophin signals increased tyrosine phosphorylation of β-catenin through inactivation of the intrinsic catalytic activity of the receptor-type protein phosphatase β/ζ. Proc. Natl. Acad. Sci., USA, 97, 2603-2608. Miura, K., Jacques, K. M., Stauffer, S., Kubosaki, A., Zhu, K., Hirsch, D. S., Resau, J., Zheng, Y., and Randazzo, P. A. (2002). ARAP1: a point of convergence for Arf and Rho signaling. Mol. Cell, 9, 109-119. Nagase, T., Seki, N., Tanaka, A., Ishikawa, K., and Nomura, N. (1995). Prediction of the coding sequences of unidentified human genes. DNA Res., 2, 167-174. Nishiya, N., Shirai, T., Suzuki, W. and Nose, K. (2002). Hic-5 interacts with GIT1 with a different binding mode from paxillin. J. Biochem., 132, 279-289. Norman, J. C., Jones, D., Barry, S. T., Holt, M. R., Cockcroft, S., and Critchley, D. R. (1998). ARF1 mediates paxillin recruitment to focal adhesions and potentiates Rho-stimulated stress fiber formation in intact and permeabilized Swiss 3T3 fibroblasts. J. Cell Biol., 143, 1981-1995. Paris, S., Za, L., Sporchia, B., and de Curtis, I. (2002). Analysis of the subcellular distribution of avian p95-APP2, an ARF-GAP orthologous to mammalian paxillin kinase linker. Int. J. Biochem. Cell Biol., 34, 826-837. Pierce, K. L., Premont, R. T., and Lefkowitz, R. J. (2002). Seven-transmembrane receptors. Nature Rev. Mol. Cell Biol., 3, 639-650. Pitcher, J. A., Friedman, N. J. and Lefkowitz, R. J. (1998). G protein-coupled receptor kinases. Ann. Rev. Biochem., 67, 653-692. Premont, R. T., Claing, A., Vitale, N., Freeman, J. L. R., Pitcher, J. A., Patton, W. A., Moss, J., Vaughan, M., and Lefkowitz, R. J. (1998). β2-Adrenergic receptor regulation by GIT1, a G protein-coupled receptor kinase-associated ADP-ribosylation factor GTPase-activating protein. Proc. Natl. Acad. Sci., USA, 95, 14082-14087. Premont, R. T., Claing, A., Vitale, N., Perry, S. J., and Lefkowitz, R. J. (2000). The GIT family of ADP-ribosylation factor GTPase-activating proteins. J. Biol. Chem., 275, 2237322380. Randazzo, P. A. (1997). Resolution of two ADP-ribosylation factor 1 GTPase-activating proteins from rat liver. Biochem. J., 324, 413-419. Randazzo, P. A., Andrade, J., Miura, K., Brown, M. T., Long, Y.-Q., Stauffer, S., Roller, P., and Cooper, J. A. (2000). The Arf GTPase-activating protein ASAP1 regulates the actin cytoskeleton. Proc. Natl. Acad. Sci., USA, 97, 4011-4016. Roemer, T., Vallier, L., Sheu Y.J., and Snyder, M. (1998). The Spa2-related protein, Sph1p, is important for polarized growth in yeast. J. Cell. Sci., 111, 479-494. Scheffzek, K., Ahmadian, M. R., and Wittinghofer, A. (1998). GTPase-activating proteins: helping hands to complement an active site. Trends Biochem. Sci., 23, 257-262. Sheu, Y. J., Santos, B., Fortin, N., Costigan, C., and Snyder, M. (1998). Spa2p interacts with cell polarity proteins and signaling components involved in yeast morphogenesis. Mol. Cell Biol., 18, 4053-4069.

GITS: ARF GAPS AND SIGNALING SCAFFOLDS

25 183

Szafer, E., Pick, E., Rotman, M., Zuck, S., Huber, I., and Cassel, D. (2000). Role of coatomer and phospholipids in GTPase-activating protein-dependent hydrolysis of GTP by ADPribosylation factor-1. J. Biol. Chem., 275, 23615-23619. Tumbarello, D. A., Brown, M. C., and Turner, C. E. (2002). The paxillin LD motifs. FEBS Lett., 513, 114-118. Turner, C. E., Brown, M. C., Perrotta, J. A., Riedy, M. C., Nikolopoulos, S. N., McDonald, A. R., Bagrodia, S., Thomas, S., and Leventhal, P. S. (1999). Paxillin LD4 motif binds PAK and PIX through a novel 95-kD ankyrin repeat, ARF-GAP protein. J. Cell Biol., 145, 851863. Turner, C. E., West, K. A., and Brown, M. C. (2001). Paxillin-ARF GAP signaling and the cytoskeleton. Current Opin. Cell Biol., 13, 593-599. Vitale, N., Patton, W. A., Moss, J., Vaughan, M., Lefkowitz, R. J., and Premont, R. T. (2000). GIT proteins, a novel family of phosphatidylinositol 3,4,5-trisphosphate-stimulated GTPase-activating proteins for ARF6. J. Biol. Chem., 275, 13901-13906. Webb, D. J., Parsons, J. T., and Horwitz, A. F. (2002). Adhesion assembly, disassembly and turnover in migrating cells - over and over and over again. Nature Cell Biol., 4, E97-100. West, K. A., Zhang, H., Brown, M. C., Nikolopoulos, S. N., Riedy, M. C., Horwitz, A. R., and Turner, C. E. (2001). The LD4 motif of paxillin regulates cell spreading and motility through interaction with paxillin kinase linker. J. Cell Biol., 154, 161-176. Yoshii, S., Tanaka, M., Otsuki, Y., Wang, D.-Y., Guo, R.-J., Zhu, Y., Takeda, R., Hanai, H., Kaneko, E., and Sugimura, H. (1999). αPIX nucleotide exchange factor is activated by interaction with phosphatidylinositol 3-kinase. Oncogene, 18, 5680-5690. Zhao, Z.-S., Manser, E., Chen, X.-Q., Chong, C., Leung, T., and Lim, L. (1998) A conserved negative regulatory region in αPAK. Mol. Cell. Biol., 18, 2153-2163. Zhao, Z.-S., Manser, E., and Lim, L. (2000a) Interaction between PAK and Nck. Mol. Cell. Biol., 20, 3906-3917. Zhao, Z.-S., Manser, E., Loo, T.-H., and Lim, L. (2000b). Coupling of PAK-interacting exchange factor PIX to GIT1 promotes focal complex disassembly. Mol. Cell. Biol., 20, 6354-6363.

Chapter 9 PAXILLIN-ASSOCIATED ARF GAPS Their Isoform Specificities and Roles in Coordination HISATAKA SABE Department of Molecular Biology, Osaka Bioscience Institute, Osaka, Japan

Abstract:

1.

Cell migration is a multifactorial process in which a number of distinct events occur simultaneously. Paxillin is an integrin-assembly adaptor protein. Here, properties of Arf GAPs bearing paxillin-binding capacity are described with their isoform specificity. Roles in linking Arf function with other intracellular events all need to be coordinately regulated during integrin-mediated cell adhesion and migration.

INTRODUCTION

Integrin-mediated cell movement is a multifactorial process in which a number of distinct intracellular events must occur simultaneously and coordinately. These include signal transduction, cytoskeletal remodeling, and membrane traffic and remodeling. These are required for proper adhesion, protrusion, traction and retraction, physical force generation and polarity formation. Cell migration is initiated by a variety of extracellular cues, such as by growth factors, cytokines, chemoattractants and extracellular matrix components. The intracellular biochemical events that are evoked by these stimuli (such as ion fluxes, protein activation, protein phosphorylation, and gene expression) have been extensively studied during the last two decades, and each has been well documented. The existence of cross talk between these events is also well known. However, several fundamental questions still remain largely unsolved. Prominent among these are the molecular mechanisms involved in the temporal and spatial coordination of distinct processes as well as the orchestration of distinct events occurring at different parts within a single cell, all of which result in a unified, efficient, directional migration. What is the principal or primordial 185

2 186

SABE

mechanism employed in such coordination and orchestration in integrinmediated migration? Arf GTPases are important factors regulating intracellular vesicle traffic and membrane remodeling. They are also important for actin-cytoskeletal remodeling. Recently, several Arf GTPase activating proteins (GAPs) have been found to bind paxillin, an integrin-assembly scaffolding/adaptor protein. In this chapter, the roles of a subset of cellular Arf GAPs in integrin-mediated cell adhesion and migration are discussed. We have focused on aspects of the mechanisms by which Arf activities coordinate the remodeling of membrane and actin-cytoskeleton in motile cells, together with the isoform specificity of the Arf GAPs involved. We will describe how paxillin-associated Arf GAPs differ structurally and functionally, and discuss the model in which different Arfs work with different subsets of Arf GAPs to effect changes in the actin cytoskeleton and membrane remodeling. The specificity of Arf GAPs for different Arfs, assessed both biochemically and in cell biological assays, will also be addressed. A hypothesis that explains the dual regulation of substrate specificities toward the different Arfs in vivo is proposed.

2.

INTEGRINS, PAXILLIN AND CELL MIGRATION

Integrins are bi-directional, cell surface α/β heterodimeric signaling proteins. Integrins require assembly of several different types of proteins at their cytoplasmic tails to transmit their signals (Hynes, 2002). These integrinassembly proteins also play a role in transmitting intracellular information into changes in the ability of integrins to bind extracellular ligands. Integrins exhibit dynamic properties during cell adhesion and migration. Newly synthesized integrins are transported to the plasma membrane via intracellular membrane/vesicle traffic. In actively migrating cells, subsets of integrins (including, e.g., fibronectin receptors) are recycled actively and can be brought to the leading edges of the cell periphery, perhaps coupled with their endocytosis at the trailing edge of the cell (Bretscher, 1989; Bretscher, 1992; Lawson and Maxfield, 1995). However, it appears that not all types of integrins can enter recycling endosomes (Bretscher, 1992; Roberts et al., 2001). In contrast, in cultured fibroblasts most of the β1 integrin molecules adhering to the extracellular matrix (ECM) at the rear of the cell are left behind as the cell body moves forward (Regan and Horwitz, 1992). Most integrin-assembly proteins are recruited to integrins by the binding of integrin to extracellular environments. A hierarchical and step-wise regulation exists in the recruitment of integrin-assembly proteins (Miyamoto et al., 1995a; Miyamoto et al., 1995b). How then do such proteins assemble

PAXILLIN-ASSOCIATED ARF-GAPS

1873

at the cytoplasmic tails of integrins at the cell surface, while dynamically entering cycling endosomes in motile cells? Paxillin is an integrin-assembly adaptor protein (Turner, 2000). Paxillin is not constitutively associated with the cytoplasmic tails of integrins, but is recruited to these sites upon integrin engagement (Miyamoto et al., 1995b). Paxillin was one of the first integrin assembly proteins to be successfully tagged with EGFP (Mazaki et al., 1998). Viewed by video recording, EGFP-paxillin is swiftly recruited to the leading edges of lamellipodial structures in migrating NMuMG epithelial cells (Nakamura et al., 2000) and in Chinese hamster ovary cells (Laukaitis et al., 2001). Macroaggregates of paxillin-containing focal adhesions, connected to stress fibers, formed as a cell body extended forward, apparently as a consequence of the accumulation of paxillin molecules formerly localized laterally along lamellipodia (Nakamura et al., 2000). Further, paxillin was not diffuse in the cytoplasm but accumulated at the perinuclear regions, including the Golgi apparatus and ER structures (Mazaki et al., 1998; Mazaki et al., 2001). Perhaps closely related to this, Norman, et al. (1998) have shown that Arf1 is involved in the recruitment of paxillin to focal adhesions. It was later shown that all three classes of Arfs affect the subcellular localization and focal accumulation of paxillin (Kondo et al., 2000; Mazaki et al., 2001).

3.

PAXILLIN-ASSOCIATED ARF GAPS

3.1

Structures and Nomenclatures

Protein pull-down assays have demonstrated that a large number of different proteins can associate with paxillin (Mazaki et al., 1997; Turner et al., 1999). These proteins include those bearing the Arf GAP domain; e.g., p95PKL (Turner et al., 1999), PAG3/KIAA0400 (Kondo et al., 2000), PAG2/KIAA1249 (clone 43 and 81; Kondo et al., 2000), p95APP1 (Di Cesare, 2000), Git2 and its alternatively spliced variants (Mazaki et al., 2001). The current nomenclature of these proteins is confusing, primarily due to their simultaneous but independent identification by different groups as binding proteins for different proteins. The Arf GAPs discussed here, and alternative names, are summarized in Table 1. p95PKL was affinity-purified using GST-paxillin as a probe, and turned out to be orthologous to human Git2. PAG1 and PAG1-short were also identified by affinity-purification from HeLa lysates, and turned out to be the same molecules as Git2 and Git2-short, respectively. The mouse ortholog of PAG1/Git2, Cat2, was originally identified as a Cool/Pix binding protein (Bagrodia and Cerione, 1999). PAG1-short/Git2-short is a C-terminal

4 188

SABE

truncated isoform of PAG1/Git2. Git2 and p95PKL have two paxillin binding sequences (PBS; Turner et al., 1999), only one of which binds strongly to paxillin (Mazaki et al., 2001). Git2-short only contains the weaker PBS, suggesting the possibility of a cellular role independent of paxillin binding. p95APP1 was originally isolated from chicken embryonic brain lysates as a binding partner of activated Rac1A and Rac1B (Di Cesare et al., 2000), and is an ortholog of human Git1 (Premont et al., 1998), a close relative of Git2. Table 1. Names of paxillin-associated ArfGAPs Original Names

Names as Proposed Standard PAGs Names GIT1 (human), CAT-1 (mouse), p95APP1 (chick) ⎯ APAP1 GIT2 (human), CAT-2 (mouse), p95APP2 / p95PKL (chick) PAG1 APAP2 ASAP1 (mouse), DEF-1 (bovine) PAG2 AMAP1 Pap α (human) PAG3 AMAP2

Integrins are activated upon maturation of monocytes into adherent macrophage-like cells, at which time paxillin expression is augmented (Mazaki et al., 1997). PAG2 and PAG3 were identified from human monocyte U937 cells, by far-western screening of a λgt11 cDNA library with the N-terminal half of paxillin as a probe. PAG2 is the human ortholog of mouse ASAP1 (Brown et al., 1998) and bovine DEF1 (King et al., 1999). PAG3 is identical to Papα, which was first identified as a protein tyrosine kinase, Pyk2, binding protein (Andreev et al., 1999). ASAP1 and PAPα each contain a src homology 3 (SH3) domain at their C-terminus, to which paxillin binds (Kondo et al., 2000). More than a dozen Arf GAP domain-containing proteins exist in humans (see Figure 1), though not all have been demonstrated to exhibit Arf GAP activity. Based on structural similarity, they can be classified into several distinct classes, with the paxillin-binding Arf GAPs falling into two groups, Git and ASAP1/PAPα proteins (Figure 1). The Git group all have the Arf GAP domain at the N-terminus, followed by ankyrin repeats. The ASAP1/PAPα group bear a single PH domain for putative lipid binding, followed by the Arf GAP domain, ankyrin repeats, the proline-rich domain (PRD), and the SH3 domain at the C-terminus. Predicted proteins that contain Arf GAP domains were found in the Kazusa database (http://www.kazusa.or.jp/en/database.html); including KIAA0050/ACAP1, γ1, δ1 KIAA0167/centaurin KIAA0580/ARAP2/centaurin and KIAA0782/ARAP1/centaurin δ2. These were examined for paxillin binding, but none exhibited such an activity (H. Uchida and H. Sabe, unpublished observations). Thus, binding to paxillin is an activity that is specific to only two groups of Arf GAPs.

PAXILLIN-ASSOCIATED ARF-GAPS

1895

In this chapter, PAG1 is called APAP2, and PAG2 and PAG3 are called AMAP1 and AMAP2, respectively, following the proposed unified naming of ArfGAPs. Other ArfGAPs are also called accordingly.

Figure 1. Arf GAP proteins in human genome. The class I and class II paxillin-associated Arf GAPs are indicated on the right, with the paxillin-binding regions underlined.

3.2

Isoform Specificities of Paxillin-associated Arf GAPs

3.2.1

Biochemical Arf GAP Assays

Arf GAP activity can be measured in vitro using GTP-loaded Arfs as substrates. Both APAP1 and its short isoform exhibit comparable activities towards Arf1, 2, 3, 5 and 6 in vitro (Vitale et al., 2000). Their activities towards Arf1 were augmented several fold by the presence of phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3), but not by PIP, PI(4,5)P2, or diacylglycerol. On the other hand, both AMAP1 (Brown et al., 1998) and AMAP2 (Andreev et al., 1999) exhibit PI(4,5)P2-dependent GAP activity for Arf1 and Arf5, with much less activity for Arf6 in vitro. Neither AMAP1 and AMAP2 exhibit much GAP activity in the absence of PI(4,5)P2, and this effect is specific to PI(4,5)P2 as other lipids (i.e., phosphatidic acid or phosphatidylinositol) were almost ineffective in the augmentation of their GAP activities. Phospholipids added in these assays

6 190

SABE

may interact with the N-terminus of Arfs or with the PH domains of the Arf GAPs, when present (see Chapter 3 for details). Biochemical assays were also performed in vivo. Mammalian cells, overexpressing epitope-tagged Arfs and Arf GAPs were labeled with radioactive orthophosphate. Tagged Arfs were subsequently pulled-down and bound radiolabeled nucleotides analysed by thin layer chromatography. With this assay, AMAP1 again appeared to exhibit GAP activity for Arf1, but not for Arf6 (Furman et al., 2002). 3.2.2

Biological Arf GAP Assays

Biochemical assays have verified the presence of Arf GAP activity in every paxillin-associated Arf GAP. However, in vitro GAP assays often show relatively broad specificity that can be contradictory to results obtained from in vivo studies (e.g., see Settleman et al., 1992 for p190RhoGAP; Ridley et al., 1993 for rhoGAP-C). Moreover, given the known abilities of other proteins (e.g., COPI, see below) or lipids (see Chapter 3) to dramatically alter Arf GAP activities and the potential loss of specificity inherent in the use of purified proteins, it seems to be important to define Arf GAP activities and specificities in a cellular context whenever possible. Of course, the value of such data from intact cells must be weighed against the dangers of using modified and overexpressed proteins. In this section I describe several cell biological approaches to assay Arf GAP activities for the paxillin binding Arf GAPs. The in vivo specificity of these Arf GAPs was verified using mutants within the GAP domains (Cukierman et al., 1995) as negative controls in each experiment. A cell based approach to assaying the GAP activity for Arf1 involved analyzing the subcellular localization of β-COP, a subunit of the COPI coat that is involved in shuttling of vesicles between the Golgi and ER compartments. Activated Arf1 specifically recruits COPI to the appropriate intracellular membrane. Thus, β-COP can be an excellent marker for the intracellular activity of Arf1; it is diffusely redistributed in the cytosol when Arf1 is inactivated (Peters et al., 1995; Ooi et al., 1998). This redistribution is specific to Arf1, and is not seen by inactivation of Arf5 or Arf6. Overexpression of APAP2-short caused the redistribution of β-COP in fibroblasts (Mazaki et al., 2001). This redistribution was suppressed by coexpression of the GTP hydrolysis-deficient form of Arf1, Arf1-L67. On the other hand, overexpression of AMAP2 did not exert such an effect on βCOP (Andreev et al., 1999; Kondo et al., 2000). A second cell biological assay involved analysis of the GAP activity for Arf6. The class III Arf, Arf6, but not class I or II Arfs, is involved in Fcγ receptor-mediated phagocytosis in macrophage cells (Zhang et al., 1998).

PAXILLIN-ASSOCIATED ARF-GAPS

1917

Dominant Arf6 mutants, one defective in GTP hydrolysis (Arf6-L67) and the other defective in GTP binding (Arf6-N27), each inhibit Fcγ receptormediated phagocytosis, while mutants at homologous sites of the other Arfs do not exert such inhibitory effects. Moreover, brief treatment of macrophages with brefeldin A had no effect on Fcγ receptor-mediated phagocytosis, despite the disruption in the organization and function of the Golgi apparatus (Zhang et al., 1998). Brefeldin A inhibits the activity of several guanine nucleotide exchange factors (GEFs) for Arfs and thereby inhibits the function of Arfs, with the exception of Arf6 (Peters et al., 1995; Radhakrishna et al., 1996; Cavenagh et al., 1996; Chardin and McCormick, 1999). Thus, this experiment also indicates that brefeldin A-sensitive Arf GEFs and their target Arfs appear to be nonessential for phagocytosis. We originally identified human AMAP1 and AMAP2 from mature monocytes of macrophage-like cells (Kondo et al., 2000). Paxillin has also been shown to accumulate at phagocytic cups, though its function in phagocytosis is still unclear (Greenberg et al., 1994; Allen and Adrem, 1996). Both AMAP1 and AMAP2 accumulate at phagocytic cups in P388D1 mouse macrophage cells, while APAP2 and its short isoform do not (Uchida et al., 2001; H. Uchida and H. Sabe, unpublished observations). Such accumulation was also seen with EGFP-tagged AMAP2 that was exogenously expressed at moderate levels (2-3 fold higher than endogenous AMAP2). Overexpression of AMAP1 or AMAP2 inhibited phagocytic activity, while overexpression of APAP2 or APAP2-short did not (Uchida et al., 2001). In these assays, the levels of overexpressed Arf GAPs were more than twenty fold higher than the endogenous levels, assessed optically with each plasmid-transfected cell. The phenotypes induced by AMAP2 overexpression were essentially the same as those seen with expression of the Arf6 mutants (Zhang et al., 1998). Moreover, co-overexpression of Arf6, but not the other Arfs, with AMAP2 restored phagocytic activity (Uchida et al., 2001). These results are all consistent with roles for AMAP1 and AMAP2 in Arf6-mediated Fcγ receptor phagocytosis, and further suggest that these two Arf GAPs have specificity for Arf6 in this context. A third assay involved the addition of aluminum fluoride to COS-7 cells that over express different Arf proteins. AlF (a mixture of 30 mM NaF and 50 µM AlCl3; (Radhakrishna et al., 1996)), is an activator of heterotrimeric G proteins (Sternweis and Gilman, 1982; Al-Awar et al., 2000; Boshans et al., 2000). AlF treatment of Arf cDNA transfected cells gave rise to distinct cell phenotypes depending on the cDNAs transfected; it increased the number and size of Arf1-associated punctuate structures in the Arf1transfected cells (Ooi et al., 1998) and induced membrane protrusion in the Arf6-transfected cells (Radhakrishna et al., 1996). Overexpression of APAP2 or APAP2-short (Mazaki et al., 2001) suppressed the Arf1-mediated

8 192

SABE

phenotype in AlF treated cells, while it was ineffective in suppressing the Arf6-mediated membrane protrusions. In contrast, overexpression of AMAP2 suppressed the Arf6-induced membrane protrusions, but not the Arf1-mediated phenotype (Kondo et al., 2000). The three different in vivo assays described above revealed that the APAP group of paxillin-associated Arf GAPs act as GAPs for Arf1, and those of AMAP group act as GAPs for Arf6. However, it is also possible that a single Arf GAP acts on different isoforms of Arf, as suggested by in vitro biochemical assays. Therefore, although these assays indicate specificity for Arfs exists in certain cellular contexts for each Arf GAP in vivo, they do not exclude the possibility that these Arf GAPs also act on other Arfs in different cellular contexts. For example, chicken APAP1 colocalizes with Arf6 at the cell periphery in cultured neuron cells, and was hence proposed to act as a GAP for Arf6 in vivo (Di Cesare et al., 2000). In Arf6-transfected COS-7 cells treated with AlF, a significant fraction of APAP2-short was also recruited by and colocalized with Arf6 at the plasma membrane, without significant suppression of Arf6 activity even when overexpressed (Mazaki et al., 2001).

3.3

Selective and Stable Binding of the Arf GAP Domain to Arf•GTP

The results described above from biochemical and cell biological assays are not consistent with each other. One of the critical questions is why AMAP2 does not exhibit GAP activity for Arf6 when measured in vitro, although it acts as an efficient GAP for Arf6 in vivo. This issue is discussed further below. Protein pull-down assays have demonstrated that the GEF domains of several GEF proteins, such as intersectin, can selectively and stably bind to their substrate GTPases in their GDP bound forms (Hart et al., 1994; Hussain et al., 2001). In the case of the GAP proteins, on the other hand, assessment of their substrates by a similar method has not been applicable, because the binding sites for the GAPs and effectors on Ras family GTPases have been shown to largely overlap (Goldberg, 1999), which may result in large areas of their GAP binding domains being occluded by effectors, and therefore unavailable for being pulled-down efficiently by exogenously expressed GAP domains. Moreover, GTP bound to GTPases may quickly be hydrolysed upon binding to their GAPs in vivo so that GAPs cannot bind stably to the GTPases. However, it has been proposed that the binding sites for the GAPs and coatomers or other effectors do not overlap in Arf1 (Goldberg, 1999). These observations may explain the success in the use of pull-down assays with at least one Arf GAP and Arfs. AMAP2 was found to

PAXILLIN-ASSOCIATED ARF-GAPS

1939

bind stably to wild type Arf6 in vivo, but not to other Arfs, in cells overexpressing all of these proteins (S. Hashimoto and H. Sabe, unpublished observation). The presence of divalent cations did not affect the binding. Moreover, the Arf GAP domain of AMAP2 selectively bound Arf6-L67, but not Arf6-N27 in vitro. Consistent with this, AMAP2 was efficiently recruited to the plasma membrane by activated Arf6, but not by other Arfs, and both proteins colocalize stably at the plasma membrane of these transfected cells (S. Hashimoto and H. Sabe, unpublished observation). Therefore, it is conceivable that this 'non-catalytic' and stable binding of AMAP2 to Arf6 prevents AMAP2 from exhibiting efficient catalytic activity against Arf6, when measured using the above mentioned biochemical method.

3.4

Dual Regulation of the Arf GAP activities towards Arfs in vivo-a Hypothesis

The fact that AMAP2 can bind stably to Arf6•GTP without hydrolyzing the GTP appears to be consistent and analogous to a current model for the interaction of Arf GAP1 with Arf1. Arf1 is the best studied of the Arf isoforms, and may provide some insight into the mechanism of regulation of the other Arf isoforms. Arf1 mediates transport between the ER and Golgi by employing coat protein complexes, such as COPI (Roth, 1999). GAP molecules for Arf1 are found as components of a priming complex induced by Arf1-GTP in the COPI system (Schekman and Orci, 1996). The GTPase activity of other small GTPases, such as Ras and Rho, is accelerated dramatically upon binding to their cognate GAPs. With these protein complexes, the ‘arginine finger’ model has been proposed, in which the corresponding GAP supplies a critical arginine, which hydrogen bonds to the γ-phosphate of GTP and stabilizes the transition state (Scheffzek et al., 1998). In contrast to this model, Goldberg (1999) has proposed that Arf1/ArfGAP1 constitutes only part of the usual protein-binding interface, and that the key residue(s) required for maximal rates of GTP hydrolysis, like the 'arginine finger' in the case of other GAPs, may not be provided by ArfGAP1. Rather, further association with the coatomer and/or accumulation of the coatomer with the Arf1/ArfGAP1 complex may be required for the maximal catalytic activity of the Arf1/ArfGAP1 complex. This model may explain why ArfGAP1 by itself does not function as an efficient catalytic GAP upon binding to Arf1•GTP (Roth, 1999; Springer et al., 1999). It should be noted, however, that there are several reports apparently contradictory to this model, which suggest that the coatomer may simply aid the productive association of Arf and ArfGAP (Mandiyan et al., 1999; Szafer et al., 2000; Eugster, et al., 2000). Nevertheless, a recent

10 194

SABE

extensive study of the 'interface array' of protein interactions of the small GTPases also indicates that the protein-protein contacts in the Arf-ArfGAP complex are significantly different from those in other GAP complex, and supports the finding that the Arf/ArfGAP structure may represent a complex with only low catalytic activity (Corbett and Alber, 2001). Therefore, it is also possible that coatomer proteins may act as allosteric activators of the catalytic activity of the Arf1/ArfGAP1 complex. Further scrutiny is needed to clarify the picture of ArfGAP action, and roles of the coatomer or other proteins regarding their possible involvement in the temporal-spatial determination of the onset of the GTP-hydrolysis activity of Arf1/ArfGAP1 complexes. On the other hand, structural analysis of AMAP1 (Papβ) suggests that the critical residues for catalysis are directly provided by the Arf GAP (Mandiyan et al., 1999), consistent with in vitro results showing that AMAP1 exhibits higher basal GAP activity for Arf1 than does Arf GAP1. Considering that AMAP2 exhibits efficient GAP activity for Arf1 and Arf5 in the presence of PI(4,5)P2 in vitro (Andreev et al., 1999), I propose that AMAP2 may function as a GAP for different Arfs in at least two different ways: one mechanism requires other protein(s) or modification(s) for AMAP2 to be active for Arf6, in the other mechanism AMAP2 is 'constitutively' active as a GAP for Arf1 and Arf5, as long as phospholipids such as PI(4,5)P2 exist at the required levels. This putative dual regulation of the GAP activities of AMAP2 may be important for the intracellular regulation of Arfs. The former mechanism may be necessary for priming complex formation of Arf6 on the donor membrane, similar to that proposed for the interaction of Arf GAP1 with Arf1 on Golgi or ER membranes (Roth, 1999; Springer et al., 1999). Subsequent association with additional proteins, such as with coat or cargo proteins, as proposed in the case of Arf GAP1 (Goldberg, 1999; Springer et al., 1999), or some modification of AMAP2, may be necessary for the full activation of the GAP activity of the Arf6/AMAP2 complex. On the other hand, unlike other members of Rassuperfamily GTPases, Arfs do not have intrinsic GTPase activity (Kahn and Gilman, 1986). Hence, it is possible that at least some Arf GAPs retain a relatively “non-specific” GAP activity that may be used to silence any Arf that was inappropriately activated while the same Arf GAP may retain a highly specific Arf GAP function that is context and possibly co-factor dependent. This putative “non-specific GAP” function of AMAP2 may be important, for example, after the mitosis of cells where proper intracellular compartmentalization of membrane components could be lost, at least transiently. Such function of AMAP2 may also contribute excluding inappropriate activities of the class I and class II Arfs from areas where Arf6 has to function.

PAXILLIN-ASSOCIATED ARF-GAPS

11 195

A previous structural study was performed of the Arf1•GDP/Arf GAP complex (Goldberg, 1999). However, since we have shown that the Arf GAP domain of AMAP2 binds stably to Arf6•GTP (S. Hashimoto and H. Sabe, unpublished observation), structural analysis of the complex of Arf6•GTP and the Arf GAP domain of AMAP2 might be informative for further analyses of the regulation of the catalytic activity of Arf GAPs. Moreover, this analysis will also help to assess our hypothesis regarding the possible dual regulation of the GAP activity of AMAP2, described above.

3.5

BIOCHEMICAL CIRCUITRY OF PAXILLINASSOCIATED Arf GAPS IN MOTILE CELLS

3.5.1

Links between Arf1-mediated membrane traffic and actincytoskeletal remodeling -roles for paxillin-associated Arf1GAPs

The paxillin-associated Arf GAPs of the APAP group, Git1, Git2, and Git2short, are orthologs of murine Cat-1 and Cat-2 and their splice variants, identified as Cool/Pix binding proteins (Bagrodia and Cerione, 1999). Cool/Pix proteins were identified as binding to Pak (p21-activated serine/threonine kinase; Manser et al., 1998; Bagrodia et al., 1998); a kinase activated by GTP-bound Cdc42 or Rac1 (Manser et al., 1994). Binding of Cool/Pix to Pak also activates its kinase activity, either by cooperating with the binding of activated Cdc42 or Rac or by altering Pak conformation directly. There are several isoforms of Cool/Pix proteins, p50Cool-1, p85Cool-1/βPix and Cool-2/αPix, and all contain the Dbl-homology (DH) domain that can activate Rho family members by stimulating guanine nucleotide exchange (Bagrodia and Cerione, 1999). p85Cool-1/βPix increases GTP binding to Rac but not Cdc42 (Manser et al., 1998) and Cool2/αPix may increase the levels of GTP-bound Cdc42 (Daniels et al., 1999). On the other hand, in vitro measurements of Cool/Pix-catalyzed GDP release from small GTPases have failed to detect any exchange activity (Bagrodia et al., 1998), suggesting that additional factors may be required for this exchange activity (Bagrodia and Cerione, 1999). Chicken APAP2, p95PKL, was identified as a paxillin-binding protein at the same time that Cat proteins and Cool/Pix proteins were identified, and was originally proposed to be a protein kinase-linker (PKL). p95PKL plays a role in connecting paxillin to Pak through Pix: i.e., paxillin binds to p95PKL, p95PKL to Pix, and Pix to Pak in a linear configuration (Turner et al., 1999). APAP2 and its short isoform in other species also bind to both Cool-2/αPix and p85Cool-1/βPix (Premont et al., 2000; Hashimoto et al., 2001; Mazaki et al., 2001). Given that the binding of Pix to Pak activates its

12 196

SABE

kinase activity, the above model implies that the connection of Pak with paxillin may be activated together with activation of the kinase activity of Pak. Paxillin can associate with both the kinase-inactive and the Cdc42activated forms of Pak3 in vivo, in which the kinase activities were assayed by phosphorylation of myelin basic protein in vitro (Hashimoto et al., 2001). Nck, an adaptor protein, binds directly to Pak1 to link Pak1 to growth factor receptors (Galisteo et al., 1996; Bockoch et al., 1996; Lu et al., 1997; Zhao et al., 2000). Nck binding to Pak1 had no stimulatory effect on Pak1 kinase activity, and it did not alter the ability of Pak1 to be stimulated by activated Rac or Cdc42. Pak1 can phosphorylate Nck. Hashimoto, et al. (2001) have shown that the biochemical properties of the binding of paxillin to Pak are similar to those ascribed to Nck: paxillin can bind directly to Pak without affecting its kinase activity, and can be phosphorylated by Pak3. Moreover, paxillin and Nck exhibit comparable affinities and bind competitively to Pak3. Therefore, paxillin appears to function as a template linking integrin to Pak, as Nck links growth factors to Pak (Hashimoto et al., 2001). Therefore, there seem to be at least two different ways in which paxillin and Pak can associate: one is through the aid of APAP2 , in which Paks may be activated by binding to Pix, and the other is direct, in which Paks can either remain inactive or be activated by Cdc42 or Rac1. Chicken APAP2 and paxillin colocalize to focal adhesions, and are connected to stress fibers, at the edges of CHO.K1 fibroblast cells (Turner et al., 1999; Brown et al., 2002). On the other hand, human APAP2-short primarily localize to perinuclear areas in 3Y1 and NIH3T3 fibroblasts, largely overlapping with paxillin as well as β-COP (Mazaki et al., 2001). However, AMAP2-short does not accumulate at focal adhesions in 3Y1 and NIH3T3 fibroblasts, while a fraction colocalizes with paxillin at focal complexes of lamellipodia (Mazaki et al., 2001). The cell morphology induced by overexpression of Git2-short is similar to that induced by active Rac1, exhibiting active membrane ruffles with concomitant substantial loss of actin stress fibers and focal adhesions (Mazaki et al., 2001). Consistently, rat p78Pix, a homolog of human p85Cool-1/βPix, and paxillin colocalize to Cdc42 and Rac1-induced focal complexes in HeLa cells (Manser et al., 1998). In CHO.K1 cells, on the other hand, APAP2 was recruited to focal adhesions connected to stress fibers, by the expression of activated Cdc42 and Rac1, but not RhoA (Brown et al., 2002). These data suggest that focal adhesions may be regulated differently in CHO.K1 cells compared to 3Y1 or NIH3T3 cells. Accumulation of APAP1 to adhesion-like structures, which apparently resemble focal adhesions, was also observed in REF52 fibroblast cells (Manabe et al., 2002). It has also been shown that paxillin plays a role in recruiting APAP2 and Pak to focal adhesions in CHO.K1 cells (Brown et al., 2002). On the other hand, it is yet to be clarified as to whether APAP1

PAXILLIN-ASSOCIATED ARF-GAPS

13 197

and APAP2 localize to the Golgi or ER to regulate their structure and function, as reported with the APAP2-short isoform. Arf1 is involved in the regulation of ER-Golgi membrane/vesicle traffic. Besides the perinuclear areas, Arf1 also functions at the cell periphery and is involved in the recruitment of paxillin to focal adhesions (Norman et al., 1998). Given that the Git group of paxillin-associated Arf GAPs act as GAPs for Arf1 (as discussed above), a molecular complex between paxillin, these Arf GAPs, βPIX, and Pak may be important for the temporal and spatial coordination of the Arf1-mediated processes with several Cdc42/Rac1-mediated processes, resulting in efficient transport and secretion of newly synthesized proteins into areas where focal complexes and focal adhesions are formed in motile cells (see later). 3.5.2

3.5.2.1

Links between Arf6-mediated processes and membrane remodeling at leading edges - roles for paxillin-associated Arf6 GAPs

Link to membrane protrusion: a lesson from Fcγ receptormediated phagocytosis in macrophages One of the interesting features of Arf GAPs of the AMAP group, PAG2 and PAG3, is that their overexpression in epithelial or monocyte cells blocks haptotactic migratory activities towards different extracellular matrices, measured using modified Boyden chambers in the absence of serum (Kondo et al., 2000; H. Uchida and H. Sabe, unpublished). In contrast, overexpression of APAP2 and its short isoform did not exhibit such inhibitory effects. Rather, each has a slightly stimulatory effect on migratory activities (Zhao et al., 2000; Mazaki et al., 2001). Cell migration proceeds in several discrete stages, in which membrane protrusion is one of the earliest. There are striking similarities between the processes involved in the remodeling of the cytoskeleton and membrane, and the formation of leading edges during cell migration and extension of pseudopods during Fcγ +receptor-mediated phagocytosis in macrophages (Aggeler and Werb, 1982; Aderem and Underhill, 1999). Moreover, engulfment of even large numbers of IgG-opsonized particles by macrophages is not accompanied by a reduction in surface area, indicating that membrane mass is supplied and maintained during phagocytosis, possibly by exocytosis of intracellular vesicles or endosomal membranes (Greenberg, 1999; Cox et al., 1999; Bajno et al., 2000). To address whether AMAP2 is involved in the maintenance of membrane mass and/or formation of membrane protrusion, Fcγ receptor-mediated phagocytosis in macrophages was used as a model system. Overexpression of AMAP2 was found to inhibit phagocytosis at very early stages, and the

14 198

SABE

formation of membrane protrusions was almost completely blocked (Uchida et al., 2001). The focal accumulation of F-actin beneath the test particles was also blocked. This result is essentially the same as that observed by inhibition of Arf6 (Zhang et al., 1998), as also mentioned earlier. A GAPinactive mutant of AMAP2, as well as APAP2 and its short isoform, did not exert such inhibitory effects on membrane protrusion. Therefore, the inhibitory effects on membrane protrusion were specific to the AMAP group, PAG2 and PAG3, and did not extend to the APAP group of Arf GAPs. The inhibition of membrane protrusion may be one of the mechanisms by which overexpression of AMAP1 and AMAP2 block cell migration. As Arf6 is the only Arf that participates in Fcγ receptor-mediated phagocytosis, it is conceivable that Arf6 is involved in this membrane mass supply. Effects of AMAP overexpression in this system could result either from its activity as a GAP for Arf6 or as a result of its stable binding to Arf6, without GTP hydrolysis, as described above. 3.5.2.2

Link between Arf6-mediated membrane recycling and Arf GAPs By what mechanism do the paxillin-associated Arf GAPs participate in the processes of membrane protrusion? Arf6 primarily functions in membrane recycling at the cell periphery, probably through its GTPase cycle (D'SouzaSchorey et al., 1995; Peters et al., 1995; Radhakrishna et al., 1996; Radhakrishna and Donaldson, 1997; D'Souza-Schorey et al., 1998; Al-Awar et al., 2000; Brown et al., 2001). The majority of the GTP-hydrolysis defective mutant Arf6-Q67L localizes to plasma membrane invaginated pits or accumulates in intracellular vacuoles, where it acts to block endocytosis of several cell surface molecules. On the other hand, the majority of the GTP-binding defective mutant Arf6-T27N localizes to intracellular tubulovesicular structures (thought to represent the pericentriolar recycling compartment) and blocks cell surface molecules from recycling back to the cell surface. It has therefore been proposed that activated Arf6, may regulate the outward flow of the recycling endosomes. Hydrolysis of the GTP is then thought to signal the return of Arf6 to the tubular endosome. There are several distinct processes for endocytosis of different types of cell surface molecules, including clathrin-dependent endocytosis and a variety of clathrin-independent processes. In which types of endocytosis Arf6 is involved has not been well established and appears to be rather cell typespecific. In HeLa cells, for example, it has been reported that Arf6 might not act in clathrin-dependent endocytosis, but is involved in recycling of Tac and MHC class I molecules, which apparently do not contain cytoplasmic tails conferring clathrin/AP-2 localization (Radhakrishna et al., 1996; Radhakrishna and Donaldson, 1997; Brown et al., 2001). β1 integrin also

PAXILLIN-ASSOCIATED ARF-GAPS

15 199

colocalized with Arf6-L67-positive intracellular vacuoles (Brown et al., 2001), although it is not yet established whether β1 integrin is under the control of Arf6 activity. Arf GAPs of the AMAP group contain protein interaction domains (i.e., the PH, PRD and SH3 domains) that are not present in those of the APAP group. As already mentioned, the SH3 domain of human AMAP2 binds to Pyk2 (Andreev et al., 1999) and paxillin (Kondo et al., 2000). Further identification of the interaction partners of these Arf GAPs may help reveal the molecular mechanism by which these Arf GAPs act in vivo. The PRDs of AMAP1 and AMAP2 bind to SH3 domains of several proteins, most of which are components of the endocytic machinery, such as amphiphysin-II and intersectin (S. Hashimoto and H. Sabe, unpublished observation; Y. Onodera and H. Sabe, unpublished observation). The PRDs of AMAP1 and AMAP2 contain several distinct proline-rich sequences, and the binding to endocytic proteins largely overlap in specificity, but are not identical. The SH3 domains of AMAP1 and AMAP2 also bind POB1 (Oshiro et al., 2002), which is a binding partner for RalBP1, a Ral effector that is also involved in receptor-mediated endocytosis (Nakashima et al., 1999). These results are consistent with the conclusion that the AMAP group of paxillin-associated Arf GAPs act to regulate Arf6-mediated endocytosis. However, it still remains established how and when these Arf GAPs attack Arf6 to down regulate its activity, during the Arf6-mediated membrane recycling back to the cell surface or the endocytosis. It also remains to be established which cell surface molecules are involved in the membrane recycling regulated by AMAP. 3.5.2.3 Links to Cell Migratory Activity The intracellular circuitry of paxillin-associated Arf GAPs has been described above and shown to be related to the machinery that regulates cell motility. Overexpression of APAPs enhances cell migration, while overexpression of the AMAPs inhibits it. Cell adhesion was not significantly affected by the overexpression of any of these Arf GAPs. These results do not necessarily imply that each of the Arf GAPs have a role either in enhancing or in inhibiting cell migratory activities under physiological conditions. AMAP1 and AMAP2, for example, are induced during monocyte maturation and accumulate at phagocytic cups, implying that these Arf GAPs might be necessary for, rather than inhibitory to, monocyte maturation and phagocytic activity. Because cell migration does not operate by a simple linear cascade of signaling events but is a multifactorial process in which a number of distinct, linked events take place, the results of the overexpression of Arf GAPs might give rise to either enhancing or blocking effects depending on the cellular context,

16 200

SABE

experimental conditions, and the presence/expression of other proteins. It is also possible that Arf GAP overexpression gives rise to indirect effects, depending on the degree of their overexpression. Therefore, effects caused by the overexpression of Arf GAPs, regardless of whether the wild type or mutant proteins are used, should be interpreted as evidence for their involvement in the processes examined. Indeed, unlike our results of AMAP2 overexpression, moderate levels of its exogenous expression have recently been shown to slightly enhance haptotaxis of NIH3T3 cells towards a gradient of fibronectin plus platelet derived growth factor (Furman et al., 2002).

4.

PROSPECT

In this chapter, we have focused on the Arf GAPs that bind paxillin, and described their isoform specificity and roles in cytoskeletal remodeling and membrane remodeling. The latter two are coordinately regulated during integrin-mediated cell adhesion, migration and intracellular signaling. An important distinction that emerges from the results described is that the paxillin binding Arf GAPs fall into two groups. One, the APAPs, has specificity for class I Arfs (e.g., Arf1) and cellular roles in membrane traffic and ruffling at the leading edge of motile cells. The other, the AMAPs, is active on Arf6 in cell-based Arf GAP assays and coordinates effects on endocytosis and recycling at the plasma membrane and on membrane protrusions in motile cells (see Figure 2). Migrating cells secrete proteins, such as cytokines and membrane metalloproteases, at the leading edges of cells. The APAPs, exhibiting a GAP activity for Arf1, localize both at the cell periphery and perinuclear areas, and bind Pix proteins, that contain Rac/Cdc42 GEF activity and bind Paks. Arf1 plays important roles in the secretion of newly synthesized proteins, and Cdc42 and Rac1 are important in polarity and membrane transport. I speculate that these Arf GAPs may be able to influence the determination of the cell polarity or movements, as mentioned earlier. Moreover, Cdc42 and Rac1 appear to be activated at the perinuclear areas, in addition to the cell periphery (Stowers et al., 1995; Erickson et al., 1996; Kroschewski et al., 1999; Wu et al., 2000; Kraynov et al., 2000; Del Pozo et al., 2001; Ridley, 2001). It will thus be interesting to explore whether the APAPs, as well as their interaction with Pix or Pak, are involved in the perinuclear activation of Cdc42 and Rac1. Possible relationships between the perinuclear activation of Cdc42 and Rac1, and activity of Arf1 may also provide interesting results.

PAXILLIN-ASSOCIATED ARF-GAPS

17 201

Figure 2. Biochemical circuitry of paxillin-associated Arf GAP proteins. See text for details.

Given that Arf6 mediates recycling of some integrins via endosomes, it is possible that overexpressed AMAPs inhibit integrin recycling by shutting off Arf6 activities, resulting in the inhibition of cell migration. However, it remains to be determined whether Arf6 regulates the recycling of integrins through endosomes. Moreover, not all isoforms of integrins recycle (Bretscher, 1992). It is not yet established whether recycling of subsets of integrins is necessary for and contribute to cell migratory activity, rather than merely associate with it. Membrane protrusion is also inhibited by the overexpression of these Arf GAPs. Membrane protrusion must be coordinated with cytoskeletal remodeling, and there are a number of reports indicating that Arf6 is involved in actin-cytoskeletal remodeling at the cell periphery. At least in the early stages of protrusion, membranes may be pushed forward from the inside by polymerizing F-actins (Mitchison and Cramer, 1996; Borisy and Svitkina, 2000). On the other hand, the supply of membrane mass from the inside is clearly demonstrated in the case of phagosome protrusions, in which Arf6 activity is likely to be required, as already mentioned. The source of membrane components supplied for the protrusions involved in cell migration and the precise mechanism how Arf6 activity is involved in the protrusive process remain uncertain (Bretscher, 1996; de Curtis, 2001). It will also be important to clarify the molecular mechanism by which membrane protrusions are tightly coupled with actin cytoskeletal remodeling, occurring underneath the protrusions. Furthermore, it remains to be determined how the polarity of integrin traffic to the cell surface and endosomal recycling are coupled to the direction of cell

18 202

SABE

movement. The AMAP Arf GAPs remain excellent tools to explore these issues. Paxillin is found at all subcellular areas where the APAPs and AMAPs exist. Paxillin is tyrosine phosphorylated in most of these areas, and linked to the regulation of Rho-family GTPases (Nakamura et al., 2000; Tsubouchi et al., 2002). Binding of paxillin to these Arf GAPs may be necessary for the recruitment of paxillin to certain subcellular areas where paxillin is subsequently phosphorylated and acts as an adaptor linking the upstream signals to the activities of Rho GTPases. This model, however, also needs further experimental scrutiny. Lastly, paxillin-binding Arf GAPs are all tyrosine phosphorylated (Andreev et al., 1999; Bagrodia et al., 1999), though the biological significance of tyrosine phosphorylation awaits description. Although the discussion in this chapter has been restricted to those Arf GAP that were identified by their binding to paxillin, most of the other Arf GAPs also have multiple protein interaction modules and are thus likely to act to coordinate multiple signaling events. Arfs are found in all eukaryotes. Hence, Arf GAPs are great targets for the analysis of the principles of intracellular multifactorial processes, where membranes have been employed as primordial superintendents of cell function, perhaps from the very beginning of the existence of eukaryotes on this planet (http://smart.emblheidelberg.de/smart/do_annotation.pl?DOMAIN=Arf&BLAST=DUMMY& EVOLUTION=Show#Evolution).

REFERENCES Aderem, A., and Underhill, D. M. (1999). Mechanisms of phagocytosis in macrophages. Annu. Rev. Immunol., 17, 593-623. Aggeler, J., and Werb, Z. (1982). Initial events during phagocytosis by macrophages viewed from outside and inside the cell: membrane-particle interactions and clathrin. J. Cell Biol., 94, 613-623. Al-Awar, O., Radhakrishna, H., Powell, N. N., and Donaldson, J. G. (2000). Separation of membrane trafficking and actin remodeling functions of ARF6 with an effector domain mutant. Mol. Cell. Biol., 20, 5998-6007. Allen, L. H., and Aderem, A. (1996). Molecular definition of distinct cytoskeletal structures involved in complement- and Fc receptor-mediated phagocytosis in macrophages. J. Exp. Med., 184, 627-637. Andreev, J., Simon, J., Sabatini, D. D., Kam, J., Plowman, G., Randazzo, P.A., and Schlessinger, J. (1999). Identification of a new Pyk2 target protein with Arf-GAP activity. Mol. Cell. Biol., 19, 2338-2350. Bajno, L., Peng, X. R., Schreiber, A. D., Moore, H. P., Trimble, W. S., and Grinstein, S. (2000). Focal exocytosis of VAMP3-containing vesicles at sites of phagosome formation. J. Cell Biol., 149, 697-706. Bagrodia, S., Taylor, S. J., Jordon, K. A., Van Aelst, L.,and Cerione, R. A. (1998). A novel regulator of p21-activated kinases. J. Biol. Chem., 273, 23633-23636.

PAXILLIN-ASSOCIATED ARF-GAPS

19 203

Bagrodia, S., Bailey, D., Lenard, Z., Hart, M., Guan, J. L., Premont, R. T., Taylor, S. J., and Cerione, R. A. (1999). A tyrosine-phosphorylated protein that binds to an important regulatory region on the cool family of p21-activated kinase-binding proteins. J. Biol. Chem., 274, 22393-22400. Bagrodia, S., and Cerione, R. A. (1999). Pak to the future. Trends Cell Biol., 9, 350-355. Bockoch, G. M., Wang, Y., Bohl, B. P., Sells, M. A., Quilliam, L. A., and Knaus, U. G. (1996). Interaction of the Nck adapter protein with p21-activated kinase (PAK1). J. Biol. Chem., 271, 25746-25749. Borisy, G. G., and Svitkina, T. M. (2000). Actin machinery: pushing the envelope. Curr. Opin. Cell Biol., 12, 104-112. Boshans, R. L., Szanto, S., van Aelst, L., and D'Souza-Schorey, C. (2000). ADP-ribosylation factor 6 regulates actin cytoskeleton remodeling in coordination with Rac1 and RhoA. Mol. Cell Biol., 20, 3685-3694 Bretscher, M. S. (1996). Getting membrane flow and the cytoskeleton to cooperate in moving cells. Cell, 87, 601-606. Bretscher, M. S. (1989). Endocytosis and recycling of the fibronectin receptor in CHO cells. EMBO J., 8, 1341-1348. Bretscher, M. S. (1992). Circulating integrins: α5β1, α6β4 and Mac-1, but not α3β1, α4β1 or LFA-1. EMBO J., 11, 405-410. Brown, M. T., Andrade, J., Radhakrishna, H., Donaldson, J. G., Cooper, J. A., and Randazzo, P. A. (1998). ASAP1, a phospholipid-dependent arf GTPase-activating protein that associates with and is phosphorylated by Src. Mol. Cell Biol., 18, 7038-7051. Brown, F. D., Rozelle, A. L., Yin, H. L., Bella, T., and Donaldson, J. G. (2001). Phosphatidylinositol 4,5-bisphosphate and Arf6-regulated membrane traffic. J. Cell Biol., 154, 1007-1017. Brown, M. C., West, K. A., and Turner, C. E. (2002). Paxillin-dependent Paxillin Kinase Linker and p21-Activated Kinase Localization to Focal Adhesions Involves a Multistep Activation Pathway. Mol. Biol. Cell, 13, 1550-1565. Cavenagh, M. M., Whitney, J. A., Carrol, K., Zhang, C. -J., Boman, A. L., Rosenwald, A.G., Mellman, I., and Kahn, R. A. (1996). Intracellular distribution of Arf proteins in mammalian cells. Arf6 is uniquely localized to the plasma membrane. J. Biol. Chem., 271, 21767-21774. Chardin, P., and McCormick, F. (1999). Brefeldin A: the advantage of being uncompetitive. Cell, 97, 153-155. Corbett, K. D., and Alber, T. (2001). The many faces of Ras: recognition of small GTPbinding proteins. Trends Biochem. Sci., 26, 710-716. Cox, D., Tseng, C. C., Bjekic, G., and Greenberg, S. (1999). A requirement for phosphatidylinositol 3-kinase in pseudopod extension. J. Biol. Chem., 274, 1240-1247. Cukierman, E., Huber, I., Rotman, M., and Cassel. D. (1995). The ARF1 GTPase-activating protein: zinc finger motif and Golgi complex localization. Science, 270, 1999-2002. Daniels, R. H., Zenke, F. T. and Bokoch, G. M. (1999). αPix stimulates p21-activated kinase activity through exchange factor-dependent and -independent mechanisms. J. Biol. Chem., 274, 6047-6050. de Curtis, I. (2001) Cell migration : GAPs between membrane traffic and the cytoskeleton. EMBO R., 2, 277-281. Del Pozo, M. A., Kiosses, W. B., Alderson, N. B., Meller, N., Hahn, K. M., and Schwartz, M. A. (2002). Integrins regulate GTP-Rac localized effector interactions through dissociation of Rho-GDI. Nature Cell Biol., 4, 232-239.

20 204

SABE

Di Cesare, A., Paris, S., Albertinazzi, C., Dariozzi, S., Andersen, J., Mann, M., Longhi, R., and de Curtis, I. (2000). p95-APP1 links membrane transport to Rac-mediated reorganization of actin. Nat. Cell Biol., 2, 521-530. D'Souza-Schorey, C., Li, G., Colombo, M. I, and Stahl, P. D. (1995). A reglatory role for ARF6 in receptor-mediated endocytosis. Science, 267, 1175-1178. D'Souza-Schorey, C., van Donselaar, E., Hsu, V. W., Yang, C., Stahl, P. D., and Peters, P. J. (1998). ARF6 targets recycling vesicles to the plasma membrane: insights from an ultrastructural investigation. J. Cell Biol., 140, 603-616. Erickson, J. W., Zhang, C., Kahn, R. A., Evans, T., and Cerione, R. A. (1996). Mammalian Cdc42 is a brefeldin A-sensitive component of the Golgi apparatus. J. Biol. Chem., 271, 26850-26854. Eugster, A., Frigerio, G., Dale, M., and Duden, R. (2000). COP I domains required for coatomer integrity, and novel interactions with ARF and ARF-GAP. EMBO J., 19, 39053917. Furman, C., Short, S. M., Subramanian, R. R., Zetter, B. R., and Roberts, T. M. (2002). DEF1/ASAP1 is a GTPase-activating protein (GAP) for ARF1 that enhances cell motility through a GAP-dependent mechanism. J. Biol. Chem., 277, 7962-7969. Galisteo,M. L., Chernoff, J., Su, Y.C., Skolnik, E. Y., and Schlessinger, J. (1996). The adaptor protein Nck links receptor tyrosine kinases with the serine-threonine kinase Pak1. J. Biol. Chem., 271, 20997-21000. Goldberg, J. (1999). Structural and functional analysis of the ARF1-ARFGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Greenberg, S., Chang, P., and Silverstein, S. C. (1994). Tyrosine phosphorylation of the γ subunit of Fcγ receptors, p72syk, and paxillin during Fc receptor-mediated phagocytosis in macrophages. J. Biol. Chem., 269, 3897-3902. Greenberg, S. (1999). Molecular components of phagocytosis. J. Leukoc. Biol., 66, 712-717. Hart, M. J., Eva, A., Zangrilli, D., Aaronson, S. A., Evans, T., Cerione, R. A., and Zheng, Y. (1994). Cellular transformation and guanine nucleotide exchange activity are catalyzed by a common domain on the dbl oncogene product. J. Biol. Chem., 269, 62-65. Hashimoto, S., Tsubouchi, A., Mazaki, Y., and Sabe, H. (2001). Interaction of paxillin with p21-activated Kinase (PAK). Association of paxillin α with the kinase-inactive and the Cdc42-activated forms of PAK3. J. Biol. Chem., 278, 6037-6045. Hussain, N. K., Jenna, S., Glogauer, M., Quinn, C. C., Wasiak, S., Guipponi, M., Antonarakis, S. E., Kay, B. K., Stossel, T. P., Lamarche-Vane, N., and McPherson, P. S. (2001). Endocytic protein intersectin-l regulates actin assembly via Cdc42 and N-WASP. Nat. Cell Biol., 3, 927-932. Hynes, R. O. (2002). Integrins: bidirectional, allosteric signaling machines. Cell, 110, 673687. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. King, F. J., Hu, E., Harris, D.F., Sarraf, P., Spiegelman, B. M., and Roberts, T. M. (1999). DEF-1, a novel Src SH3 binding protein that promotes adipogenesis in fibroblastic cell lines. Mol. Cell. Biol., 19, 2330-2337. Kondo, A., Hashimoto, S., Yano, H., Nagayama, K., Mazaki, Y., and Sabe, H. (2000). A new paxillin-binding protein, PAG3/Papalpha/KIAA0400, bearing an ADP-ribosylation factor GTPase-activating protein activity, is involved in paxillin recruitment to focal adhesions and cell migration. Mol. Biol. Cell, 11,1315-1327. Kraynov, V. S., Chamberlain, C., Bokoch, G. M., Schwartz, M. A., Slabaugh, S., and Hahn, K. M. (2000). Localized Rac activation dynamics visualized in living cells. Science, 290, 333-337.

PAXILLIN-ASSOCIATED ARF-GAPS

21 205

Kroschewski, R., Hall, A., and Mellman, I. (1999). Cdc42 controls secretory and endocytic transport to the basolateral plasma membrane of MDCK cells. Nature Cell Biol., 1, 8-13. Lawson, M. A., and Maxfield, F. R. (1995). Ca(2+)- and calcineurin-dependent recycling of an integrin to the front of migrating neutrophils. Nature, 377, 75-79. Laukaitis, C. M., Webb, D. J., Donais, K., and Horwitz, A. F. (2001). Differential dynamics of α5 integrin, paxillin, and α-actinin during formation and disassembly of adhesions in migrating cells. J. Cell Biol., 153, 1427-1440. Lu, W., Katz, S., Gupta, R., and Mayer, B. J. (1997). Activation of Pak by membrane localization mediated by an SH3 domain from the adaptor protein Nck. Curr. Biol., 7, 8594. Manabe, R., Kovalenko, M., Webb, D. J.,and Horwitz, A. R. (2002). GIT1 functions in a motile, multi-molecular signaling complex that regulates protrusive activity and cell migration. J. Cell Sci., 115, 1497-1510. Mandiyan, V., Andreev, J., Schlessinger, J., and Hubbard, S. R. (1999). Crystal structure of the ARF-GAP domain and ankyrin repeats of PYK2-associated protein beta. EMBO J., 18, 6890-6898. Manser, E., Leung, T., Salihuddin, H., Zhao, Z. S., and Lim, L. (1994). A brain serine/ threonine ptotein kinase activated by Cdc42 and Rac1. Nature, 367, 40-46. Manser, E., Loo, T. H., Koh, C. G., Zhao, Z. S., Chen, X. Q., Tan, L., Tan, I., Leung, T., and Lim, L. (1998). PAK kinases are directly coupled to the PIX family of nucleotide exchange factors. Mol. Cell, 1, 183-192. Mazaki, Y., Hashimoto, S., and Sabe, H. (1997). Monocyte cells and cancer cells express novel paxillin isoforms with different binding properties to focal adhesion proteins. J. Biol. Chem., 272, 7437-7444. Mazaki, Y., Uchida, H., Hino, O., Hashimoto, S., and Sabe, H. (1998). Paxillin isoforms in mouse. Lack of the γ isoform and developmentally specific beta isoform expression. J. Biol. Chem., 273, 22435-22441. Mazaki, Y., Hashimoto, S., Okawa, K., Nakamura, K., Yagi, R., Yano, H., Kondo, A., Iwamatsu, A., Mizoguchi, A., and Sabe, H. (2001). An ADP-ribosylation factor GTPaseactivating protein Git2-short/KIAA0148 is involved in subcellular localization of paxillin and actin cytoskeletal organization. Mol. Biol. Cell. 12, 645-662. Mitchison, T. J., and Cramer, L. P. (1996). Actin-Based Cell Motility and Cell Locomotion. Cell, 84, 371-379. Miyamoto, S., Akiyama, S. K., and Yamada, K. M. (1995a). Synergistic roles for receptor occupancy and aggregation in integrin transmembrane function. Science, 267, 883-885. Miyamoto, S., Teramoto, H., Coso, O. A., Gutkind, J. S., Burbelo, P. D., Akiyama, S. K., and Yamada, K. M. (1995b). Integrin function: molecular hierarchies of cytoskeletal and signaling molecules. J. Cell Biol., 131, 791-805. Nakamura, K., Yano, H., Uchida, H., Hashimoto, S., and Sabe, H. (2000). Tyrosine phosphorylation of paxillin α is involved in temporospatial regulation of paxillincontaining focal adhesion formation and F-actin organization in motile cells. J. Biol. Chem., 275, 27155-27164. Nakashima, S., Morinaka, K., Koyama, S., Ikeda, M., Kishida, M., Okawa, K., Iwamatsu, A., Kishida, S., and Kikuchi, A. (1999). Small G protein Ral and its downstream molecules regulate endocytosis of EGF and insulin receptors. EMBO J., 18, 3629-4362. Norman, J. C., Jones, D., Barry, S. T., Holt, M. R., Cockcroft, S., and Critchley, D.R. (1998). ARF1 mediates paxillin recruitment to focal adhesions and potentiates Rho-stimulated stress fiber formation in intact and permeabilized Swiss 3T3 fibroblasts. J. Cell Biol., 143, 1981-1995.

22 206

SABE

Ooi, C. E., Dell'Angelica, E. C., and Bonifacino, J. S. (1998). ADP-Ribosylation factor 1 (ARF1) regulates recruitment of the AP-3 adaptor complex to membranes. J. Cell Biol., 142, 391-402. Oshiro,T., Koyama, S., Sugiyama, S., Kondo, A., Onodera, Y., Asahara, T., Sabe, H., and Kikuchi A.(2002). Interaction of POB1, a downstream molecule of small G protein Ral, with PAG2, a paxillin binding protein,is involved in cell migration. J. Biol. Chem., 277, 38618-38626. Peters, P. J., Hsu, V. W., Ooi, C. E., Finazzi, D., Teal, S. B., Oorschot, V., Donaldson, J. G., and Klausner, R. D. (1995). Overexpression of wild-type and mutant ARF1 and ARF6: distinct perturbations of nonoverlapping membrane compartments. J. Cell Biol., 128, 1003-1017. Premont, R. T., Claing, A., Vitale, N., Freeman, J. L., Pitcher, J. A., Patton, W. A., Moss, J., Vaughan, M., and Lefkowitz, R. J. (1998). β2-Adrenergic receptor regulation by GIT1, a G protein-coupled receptor kinase-associated ADP ribosylation factor GTPase-activating protein. Proc. Natl. Acad. Sci., USA, 95, 14082-14087. Premont, R. T., Claing, A., Vitale, N., Perry, S. J., and Lefkowitz, R. J. (2000). The GIT family of ADP-ribosylation factor GTPase-activating proteins. Functional diversity of GIT2 through alternative splicing. J. Biol. Chem., 275, 22373-22380. Radhakrishna, H., Klausner, R. D., and Donaldson, J. G. (1996). Aluminum fluoride stimulates surface protrusions in cells overexpressing the ARF6 GTPase. J. Cell Biol., 134, 935-947. Radhakrishna, H., and Donaldson, J. G. (1997). ADP-ribosylation factor 6 regulates a novel plasma membrane recycling pathway. J. Cell Biol., 139, 49-61. Regen, C. M., and Horwitz A. F. (1992). Dynamics of β1 integrin-mediated adhesive contacts in motile fibroblasts. J Cell Biol., 119, 1347-1359. Ridley, A. J., Self, A. J., Kasmi, F., Paterson, H. F., Hall, A., Marshall, C. J., and Ellis, C. (1993). rho family GTPase activating proteins p190, bcr and rhoGAP show distinct specificities in vitro and in vivo. EMBO J., 12, 5151-5160. Ridley, A. J. (2001). Rho proteins: linking signaling with membrane trafficking. Traffic, 2, 303-310. Roberts, M., Barry, S., Woods, A., van der Sluijs, P., and Norman, J. (2001). PDGF-regulated rab4-dependent recycling of αvβ3 integrin from early endosomes is necessary for cell adhesion and spreading. Curr. Biol., 11, 1392-1402. Roth, M. G. (1999). Snapshots of ARF1: Implications for Mechanisms of Activation and Inactivation. Cell, 97, 149-152. Schekman, R., and Orci, L. (1996). Coat proteins and vesicle budding. Science, 271, 15261533. Scheffzek, K., Ahmadian, M. R., Wiesmuller, L., Kabsch, W., Stege, P., Schmitz, F., and Wittinghofer, A. (1998). Structural analysis of the GAP-related domain from neurofibromin and its implications. EMBO J., 17, 4313-4327. Settleman, J., Albright, C. F., Foster, L. C., and Weinberg, R. A. (1992). Association between GTPase activators for Rho and Ras families. Nature, 359, 153-155. Springer, S., Spang, A., and Schekman, R. (1999). A primer on vesicle budding. Cell, 97, 145-148. Sternweis, P. C., and Gilman, A. G. (1982). Aluminum: a requirement for activation of the regulatory component of adenylate cyclase by fluoride. Proc. Natl. Acad. Sci., USA, 79, 4888-4891. Stowers, L., Yelon, D., Berg, L. J., and Chant, J. (1995) Regulation of the polarization of T cells toward antigen-presenting cells by Ras-related GTPase CDC42. Proc. Natl. Acad. Sci., USA, 92, 5027-5031.

PAXILLIN-ASSOCIATED ARF-GAPS

23 207

Szafer, E., Pick, E., Rotman, M., Zuck, S., Huber, I., and Cassel, D. (2000). Role of coatomer and phospholipids in GTPase-activating protein-dependent hydrolysis of GTP by ADPribosylation factor-1. J. Biol. Chem., 275, 23615-23619. Tsubouchi, A., Sakakura, J., Yagi, R., Mazaki, Y., Schaefer, E., Yano, H., and Sabe, H. (2002). Localized suppression of RhoA activity by Tyr31/118-phosphorylated paxillin in cell adhesion nd migration. J. Cell Biol., 159, 673-683. Turner, C. E., Brown, M. C., Perrotta, J. A., Riedy, M. C., Nikolopoulos, S. N., McDonald, A. R., Bagrodia, S., Thomas, S., and Leventhal, P. S. (1999). Paxillin LD4 motif binds PAK and PIX through a novel 95-kD ankyrin repeat, ARF-GAP protein: A role in cytoskeletal remodeling. J. Cell Biol., 145, 851-863. Turner, C. E. (2000). Paxillin and focal adhesion signalling. Nature Cell Biol., 2, E231-E236. Uchida, H., Kondo, A., Yoshimura, Y., Mazaki, Y., and Sabe, H. (2001). PAG3/Papα/KIAA0400, a GTPase-activating protein for ADP-ribosylation factor (ARF), regulates ARF6 in Fcγ receptor-mediated phagocytosis of macrophages. J. Exp. Med., 193, 955-966. Vitale, N., Patton, W. A., Moss, J., Vaughan, M., Lefkowitz, R. J., and Premont, R. T. (2000). GIT proteins, A novel family of phosphatidylinositol 3,4, 5-trisphosphate-stimulated GTPase-activating proteins for ARF6. J. Biol. Chem., 275, 13901-13906. Wu, W. J., Erickson, J. W., Lin, R., and Cerione, R. A. (2000). The gamma-subunit of the coatomer complex binds Cdc42 to mediate transformation. Nature, 405, 800-804. Zhang, Q., Cox, D., Tseng, C.C., Donaldson, J.G., and Greenberg, S.(1998). A requirement for ARF6 in Fcγ receptor-mediated phagocytosis in macrophages. J. Biol. Chem., 273, 19977-19981. Zhao, Z. S., Manser, E., Loo, T. H., and Lim, L. (2000). Coupling of PAK-interacting exchange factor PIX to GIT1 promotes focal complex disassembly. Mol. Cell. Biol., 20, 6354-6363.

Chapter 10 ACTIVATION OF CHOLERA TOXIN AND E. COLI HEAT-LABILE ENTEROTOXIN (LT) BY ARF G. PACHECO-RODRIGUEZ1, NAOKO MORINAGA2, MASATOSHI NODA2, J. MOSS1, AND M. VAUGHAN1 1

Pulmonary-Critical Care Medicine Branch, NHLBI, NIH, Bethesda, USA;22nd Department of Microbiology, School of Medicine, Chiba University, Chiba, Japan.

Abstract:

1.

Cholera (CT) and heat labile E. coli (LT) enterotoxins are members of the AB5 family of bacterial toxins that contain one catalytically active A subunit associated with a complex of five B subunits. The B subunits of CT and LT bind the receptor, ganglioside GM1, at the cell surface as a required first step in entry of the holotoxin into target cells. The A subunit is processed into the catalytic A1 and linker A2 segments. The holotoxin-receptor complex is internalized and transported to the endoplasmic reticulum before release of the active A1 protein into the cytosol. The ADP-ribosyltransferase activity of A1 is a critical determinant of toxin potency. When bound to GTP, all Arfs are allosteric activators of CT in vitro. The importance of such interactions in the intoxication of cells is, however, less clear. Recent reports suggest possibilities for clinically effective prevention of CT action by interference at the critical initial step of binding to the GM1 receptor or by inhibition of CTA1 ADPribosyltransferase activity in intestinal mucosa cells.

INTRODUCTION

Purification of a protein, termed ADP-ribosylation factor or Arf, which was required for the demonstration of cholera toxin (CT)-catalyzed ADPribosylation of purified GĮs, the stimulatory guanine nucleotide-binding regulatory protein of adenylyl cyclase, was reported by Kahn and Gilman in 1984. For several years thereafter, the physiological function(s) of Arfs remained obscure. Possible roles in eukaryotic ADP-ribosylation reactions were considered, as was the possibility that by defining the mechanisms of Arf activation of cholera toxin something might be learned about Arf action 209

2 210

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

in animal cells. Identification of a critical role for Arf in protein secretion in Saccharomyces cerevisiae (Botstein et al., 1988) initiated a remarkable and continuing expansion of understanding diverse aspects of vesicular traffic at a level of complexity, molecular and regulatory, not anticipated at the time that Palade first established the mechanisms of transport of newly synthesized secretory proteins from the site of synthesis in the endoplasmic reticulum to the plasma membrane (Palade, 1975). This chapter deals mainly with the possible involvement of Arf in the action of CT or E. coli heat-labile enterotoxin (LT) on intact cells. After summarizing information regarding the structure and function of the toxins, we consider briefly: 1) the role of Arf in toxin internalization and processing by cells, and, 2) the mechanism of Arf activation of LT and CT in vivo and in vitro.

2.

STRUCTURE AND FUNCTION OF CT AND LT

CT and LT are members of the AB5 family of bacterial toxins that contain one catalytically active A subunit associated with a complex of five B subunits. The latter are responsible for toxin binding to specific receptors on the cell surface and the subsequent entry of toxin into the cell. In a given toxin, the five B subunits may or may not be identical, but all bind to glycoproteins or ganglioside oligosaccharides. In particular, CT and LT each have a single type of B subunit and the pentamer associates with the monosialoganglioside GM1 (galactosyl-N-acetylgalactosaminyl-[Nacetylneuraminyl]-galactosylglucosylceramide), the oligosaccharide of which is responsible for toxin binding. A subunits of the AB5 toxins can have different enzymatic activities, although several of them are mono-ADPribosyltransferases (reviewed by Merritt and Hol, 1995). CT and LT are very similar in structure (ca. 80% amino acid sequence identity in both A and B subunits) and function, including enzymatic properties, ganglioside specificity, and immunoreactivity. However, they differ significantly in toxicity for human cells (Rodighiero et al., 1999). Systematic comparison of the effects of chimeric CT/LT molecules on secretion of Cl¯ ion by human T84 epithelial cells grown in culture demonstrated that differences in the receptor binding B subunits, or in the catalytic A1 fragment were not responsible for the differences in toxicity. However, when the holotoxins were bound to GM1 the ability of CT to withstand denaturation by acid or urea was greater than that of LT. This was accounted for by sequence differences in a ten-residue segment of the A2 fragments. Stability of the holotoxin-receptor complex as it is internalized and transported to the

ACTIVATION OF CT AND LT BY ARF

2113

intracellular site where release of A1 will have the greatest toxicity is critical to toxin potency (Rodighiero et al., 1999).

Figure 1. Structure of LT (reproduced by permission from Merritt and Hol, 1995). In LT, the A subunit is positioned above a ring of five identical B subunits with the A2 domain extending through the center of the B pentamer. In this image, the catalytic A1 subunit is at the top and connected to the pentameric B subunit ring (bottom) by the long alpha helix that comprises most of the A2 subunit.

The crystal structure of LT was first reported in 1991 (Sixma et al., 1991) and that of the CTB complex, with GM1 oligosaccharide bound, in 1994 (Merritt et al., 1994). Information from these and other structural studies is all consistent with a model in which the B pentamer that binds GM1 on the cell surface forms a circular base on which the single A subunit rests, with its C-terminal 5 kDa A2 segment extending into the B subunit ring (see

4 212

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

Figure 1). The KDEL (RDEL in LT) sequence at the C-terminus marks the toxin for transport to the endoplasmic reticulum (Lencer et al., 1995). Additional critical sequence in A2 is that required for appropriate release of active A1 from the A2 fragment. Proteolytic cleavage (nicking) of the holotoxin at R192 can occur even before toxin entry into the cell, but the A1 and A2 fragments remain covalently linked until reduction of the single disulfide bond between them releases A1. This is probably catalyzed by a protein disulfide isomerase in the endoplasmic reticulum (Moss et al., 1980; Orlandi, 1997; Tsai et al., 2001). In the mono-ADP-ribosyltransferase reaction catalyzed by CT or LT, the ADP-ribose moiety of NAD is transferred to a specific arginine in an acceptor protein (Moss and Vaughan, 1977; Moss and Richardson, 1978). These toxins also catalyze a reaction (NAD glycohydrolase) in which water serves as the ADP-ribose acceptor (Moss et al., 1976). Both activities are intrinsic to the A1 fragment of CT (Moss et al., 1979a). In addition, CTA can ADP-ribosylate itself (auto-ADP-ribosylation) and Arf (Tsai et al., 1988). In the stereospecific reaction catalyzed by CT, the Į-anomer of ADP-ribosylarginine is produced from ȕNAD+ (Moss et al., 1979b). The major ADP-ribose acceptor in CT- and LT-catalyzed reactions in eukaryotic cells is GĮs, but in vitro, CT and LT can use several guanidinium compounds as ADP-ribose acceptors. Agmatine is one of those used in a simple assay (Moss and Vaughan, 1977) for quantification of CT ADP-ribosyltransferase activity (and that of other transferases that use arginine as an ADP-ribose acceptor).

3.

CELL BINDING AND PROCESSING OF CT AND LT

Ganglioside GM1 is the cell surface binding site/receptor for the B subunit pentamer through which CT or LT is internalized (Fishman et al., 1978). CT bound also to GM1 oligosaccharide that had been covalently attached to proteins on the cell surface (neoganglioproteins), but was not internalized, indicating that the toxin-GM1 interaction contributes to intoxication beyond the initial cell-surface binding (Pacuszka and Fishman, 1992). It also suggests that toxin internalization should be evaluated along with cell binding to assess the biological relevance of the interaction (Pacuszka and Fishman, 1992). The CT receptor, GM1, appears to be concentrated in caveolae or lipid rafts (Tran et al., 1987). The toxin was localized at nonclathrin-coated invaginations of fibroblast membranes by electron microscopy (Tran et al., 1987). In the intestine, CT binds to and is internalized via ganglioside GM1 on the apical membrane of the mucosal

ACTIVATION OF CT AND LT BY ARF

2135

cells. Cell polarity may be, in fact, an essential property of a CT (or LT) target cell in vivo, although many studies of toxin action are performed in non-polarized cells, e.g., CHO cells, that are also susceptible to CT (Morinaga et al., 2001). In intestinal mucosa cells, it is the GĮs subunit located on the basolateral membrane that is the main CT substrate. Covalent modification of GĮs and the resulting essentially irreversible activation of adenylyl cyclase leads to sustained elevation of cellular cAMP content, efflux of Na+ and water into the intestinal lumen, and eventual dehydration of the host. The complex of CT with GM1 ganglioside is internalized by receptormediated endocytosis, which is selective, regulated, and carried out via either clathrin-coated or non-clathrin-coated invaginations of the plasma membrane (e.g., caveolae and lipid rafts). Clathrin-coated pits are lined on the cytoplasmic surface with light and heavy clathrin chains plus the adaptor proteins required for assembly of the clathrin coat (for review see Higgins and McMahon, 2002). The process includes: 1) binding and localization of the ligand (toxin)-receptor complex in a nascent clathrin-coated vesicle, and, 2) accumulation of other specific molecules to complete vesicle formation. Caveolae are flask-shaped invaginations of the plasma membrane characterized by their size (50-100 nm), morphology, and biochemical constituents, which include cholesterol, sphingomyelin, and ceramide among other lipids (Anderson, 1998). Protein components include caveolin, glycosylphosphatidylinositol-anchored proteins, and receptors (Anderson, 1998). Lipid rafts or caveolae appear to be the main site of holotoxin entry, although access to some cells can occur via clathrin-coated pits or nonclathrin-coated invaginations. The site of entry can depend on the cell type, as was shown by studies of the Caco-2, HeLa, and BHK cell lines (Torgersen et al., 2001). CT-induced morphological changes of CHO cells were not affected by over-expression of dynamin, but were suppressed significantly by over-expression of the mutant dynamin in which K44 is replaced by alanine, resulting in an impaired capacity to hydrolyze GTP (personal communication, N. Morinaga). Dynamin functions in internalization via caveolae (Henley et al., 1998; Oh et al., 1998). As chlorpromazine, an inhibitor of clathrin-dependent endocytosis (Sofer and Futerman, 1995; Orlandi and Fishman, 1998), did not affect CT-induced cAMP accumulation in CHO cells, the entry of CT into those cells may be via caveolae. The mutant of Eps15, a component of clathrin-coated pits, should clarify which is involved. Independent of the membrane site of receptor interaction or mechanism of internalization, specific binding of CTB is the critical initial step in CT action. It represents, therefore, an important potential site for intervention with drugs that could prevent the CT-GM1 interaction (Merritt et al., 2002).

6 214

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

Fishman reported more than 20 years ago that CT binding and internalization had occurred well before the increase in cell cAMP was detected. The delay was considered to be the time required for intracellular transport and processing of the holotoxin (Fishman, 1982). Detailed studies by several groups, plus the greatly improved understanding of intracellular vesicular traffic have clarified the processes involved in intoxication of different types of cells by CT and LT. In the human T84 intestinal cell line, the A and B subunits separate as a result of cleavage by a serine protease in the endoplasmic reticulum (Lencer et al., 1997). The B subunit complex was returned to the plasma membrane via a secretory pathway, while the A subunit was further processed to generate A1 and A2 peptides by a proteindisulfide isomerase in the endoplasmic reticulum (Moss et al., 1980; Orlandi, 1997; Tsai et al., 2001). Finally, the A1 subunit was released into the cytosol, potentially via the Sec61p complex (Schmitz et al., 2000), from which it moved to its site of action. In polarized cells, that site is on the basolateral membrane where CTA1 ADP-ribosylates GĮs. In cultured fibroblasts incubated with CT, detection of the enzymatically active A1 peptide paralleled the activation of adenylate cyclase, from which it was inferred that the final transit of A1 to the GĮs substrate is very rapid (Kassis et al., 1982). Although the molecular interactions that result in toxin transit from plasma membrane to the Golgi are not completely understood, the transport appears to be vesicle-mediated. BFA, which disrupts Golgi formation and structure, prevented the action of CT, suggesting the requirement for an intact Golgi (Orlandi et al., 1993). Known targets of BFA in mammalian cells are BFA-inhibited guanine nucleotide-exchange proteins BIG1 and BIG2 (see Chapter 6), which activate Arf, and the BFAstimulated ADP-ribosyltransferase, which is a lysophosphatidic acid: acylCoA transferase (Weigert et al., 1999). All of these proteins are important for Golgi function. Whether or not all are important for CT and LT processing remains to be determined.

4.

ROLE OF ARF IN CT AND LT ACTION

The interactions of CT and LT with Arfs and resulting effects on ADPribosyltransferase activity have been extensively investigated. The importance of such interactions in the intoxication of cells is, however, less clear. The six mammalian Arfs are grouped in three classes: class I (Arfs 13), class II (Arfs 4, 5), and class III (Arf6) (reviewed in Moss and Vaughan, 1995 and 1998). When bound to GTP, each of the Arfs can activate CT in vitro, and the allosteric nature of this interaction was demonstrated in kinetic studies (Noda et al., 1990). Questions remain, however, regarding the role

ACTIVATION OF CT AND LT BY ARF

2157

of Arf in CT effects on cells. Interaction of Arf with CT or LT in the Golgi could precede the toxin effects that result from elevation of cellular cAMP concentration. Arf1 function was shown to be necessary for the morphological effects of CT on CHO cells (Morinaga et al., 2001). These cultured cells assume an elongated or bi-polar shape after incubation with CT. Over-expression of mutant Arf1 proteins indicated that the cyclic activation of Arf, i.e., the continuing inter-conversion of Arf•GTP and Arf•GDP, was necessary for CT action. The results were also consistent with the conclusion that Arf1 was critical for the induction of cAMP formation, and probably participated early in the process of CT intoxication of the cell. The morphological changes induced by CT were mimicked by treatment of cells with dibutyrylcAMP, 8-Br-cAMP, or forskolin, each of which increases cAMP without activating GĮs. The changes in morphology appeared about one hour later with CT than with 8-Br-cAMP, corresponding to the difference between the time required for CT to activate cyclase and increase cAMP and that needed for the cAMP analogue to reach its site of action. In cells over-expressing Arf1 mutants that blocked CT actions, 8-Br-cAMP had effects similar to those seen in cells transfected with wild-type Arf1 or empty vector, presumably because its action does not depend on Arf function, which was required for the morphological effects of CT (Morinaga et al., 2001). To evaluate the specificity of the effect of Arf1, similar over-expression experiments were carried out with the analogous mutants of Arf5 and Arf6. Unlike the Arf1 mutants, the over-expressed Arf6 and Arf5 mutants did not interfere with CT induction of morphological changes, consistent with the different functions of individual Arfs in vesicular transport. In CHO cells, COPI-coated vesicles may be involved in CT movement from Golgi to endoplasmic reticulum. Therefore, the distribution of Arfs and ȕ-COP (one of the seven protein components of COPI) was investigated. Arf1-L71, a mutant that is essentially entirely GTP-bound, was largely colocalized with ȕ-COP, and Arf5-L71 less so. Arf6-L67 was partially concentrated at the plasma membrane and was not associated with ȕ-COP. The observation that both GTP- and GDP-bound forms of Arf1 were required for CT action in CHO cells is consistent with the conclusion that Arf1 influences CT action in those cells, not via the direct interaction by which all GTP-bound Arfs can activate CT in vitro, but by its specific function in the vesicular transport pathway that CT must travel from plasma membrane to Golgi to endoplasmic reticulum before it can ADP-ribosylate GĮs and alter CHO cell morphology (Morinaga et al., 2001). A different kind of study in CHO cells, however, showed that Arf interaction with toxin in the Golgi, resulting in stimulation of toxin ADPribosyltransferase activity, could be important for the activation of adenylyl

8 216

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

cyclase and effects on cell morphology. To assess the role of Arf in localization of the toxin to Golgi and its subsequent action, Zhu and Kahn (2001) used CHO cells for transient over-expression of LTA1 (the A1 fragment of LT) with myc and his6 sequences at the C-terminus (Zhu and Kahn, 2001). These experiments eliminated factors related to toxin internalization, transport from the plasma membrane, and protein processing. By 16 hr after transfection of cells with myc-his6-LTA1, it had accumulated throughout the Golgi compartments where an increased amount of endogenous Arf was also seen. With greater intracellular accumulation of recombinant LTA1, it became dispersed throughout the cell, along with endogenous Arf and several Golgi marker proteins. The authors suggested that this effect of large amounts of LTA1 on Golgi morphology, which resembled that of BFA, was due to toxin binding of Arf•GTP, leaving insufficient active Arf to maintain Golgi function, just as results from BFA inhibition of the guanine nucleotide-exchange proteins required for formation of Arf•GTP. It was unclear whether the toxin-induced elevation of cellular cAMP content might also have affected Golgi structure (Zhu and Kahn, 2001). Co-expression of Arf3 with LTA1 was used to investigate its role in Golgi localization and cAMP production. The effects of a variety of Arf mutants with impaired intrinsic activity and/or capacity for toxin interaction were evaluated. It was concluded that Arf function was necessary for the concentration of LTA1 on Golgi membranes and its associated effects on Golgi structure, as well as for toxin-induced accumulation of cAMP and its effects on cell morphology (Zhu and Kahn, 2001). Over-expression of LTA1 in the endoplasmic reticulum and transport of holotoxin from the extracellular medium for processing in the region of the Golgi/endoplasmic reticulum obviously result in very different numbers of molecules of active A1 subunit at that site, as pointed out by Zhu and Kahn (2001). In the study of Morinaga et al. (2001), it seemed clear that the action of Arf1, and not that Arf5 or Arf6, was specifically important for the internalization and delivery of active CTA1 from the CHO cell surface to the site of ADPribosylation of GĮs, although the possibility of Arf enhancement of CTA1 ADP-ribosyltransferase activity or participation in effects of CTA1 on Golgi structure was not evaluated. It is difficult to know how to relate the effects of relatively massive amounts of LTA1 synthesized in the CHO cell (Zhu and Kahn, 2001) to the action of much smaller amounts of CTA1 accumulated in the Golgi when it is generated, presumably much more slowly, from holotoxin internalized after binding to its GM1 receptor on the cell surface. In studies with intact cells, only a limited amount of the total bound toxin may actually enter the cell and be delivered to the Golgi. If CTA1 generated in the Golgi moves rapidly to its site of action, as was

ACTIVATION OF CT AND LT BY ARF

2179

concluded by Kassis et al. (1982), CTA1 (or LTA1) may not, in fact, accumulate in the Golgi of intoxicated cells. As the multiple actions of BFA become better understood, it may be useful in answering some of these kinds of questions, just as it continues to be useful in elucidating mechanisms of intracellular vesicular transport (Presley et al., 2002).

5.

MECHANISMS OF ACTIVATION OF LT AND CT BY ARF

In their early experiments, Kahn and Gilman (1986) had demonstrated that Arf activation of CT-catalyzed ADP-ribosylation of GĮs was dependent on GTP, and quantification of GTP binding to Arf can often be a more convenient measure of Arf activation than is measurement of the activated Arf by its effects on an enzyme activity. Low stoichiometry of GTPȖS binding to Arf in vitro has been found in many laboratories, and it is known that the assay conditions, notably Mg2+ concentration and the presence of specific phospholipids or detergents are of major importance (reviewed in Moss and Vaughan, 1998). A dramatic effect of recombinant LTA1 on the stoichiometry of GTPȖS binding to recombinant Arf3 was reported by Zhu et al. (2000). They observed that LTA1, LTA, and CTA had similar effects; those of POR1∆N, GGA1, and MKLP1, were analogous but smaller. The authors interpreted their findings as evidence that these proteins acted as Arf effectors, which interacted with inactive Arf, to increase its affinity for GTP, resulting in the formation of stoichiometric effector-Arf•GTP complexes. For several years, it had seemed plausible that the mechanism by which Arf activates CT (or LT) might also be the mechanism employed for Arf activation of its endogenous effectors. It is difficult, however, to compare quantitatively Arf activity in vitro with its physiological actions in a cell. Perhaps the most obvious reason is that Arf function in the cell requires the continuing alternation between GTP- and GDP-bound forms that must be constantly regulated (temporally and spatially) by the proteins that modulate GDP release and GTP binding and hydrolysis in response to diverse signals and molecular interactions. This Arf activity (or its absence) can be observed or inferred (e.g., in effects of BFA on Golgi morphology). Unlike a difference in catalytic rate, however, it is not readily quantified. Following reports that Arf1 and Arf3 activate phospholipase D (PLD) in animal cells (Brown et al., 1993; Cockcroft et al., 1994), an Arf-activated PLD in rat brain extract was separated from an oleate-dependent PLD and shown to be activated by recombinant Arf5 (class II) and Arf6 (class III), as well as by class I Arfs (Massenburg et al., 1994). It appeared that a quantitative comparison of Arf effects on the activities of the partially

10 218

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

purified PLD and the pure CTA1 protein might permit identification of elements of the Arf structure that are responsible for activation of the two enzymes. Experiments were facilitated by the production of recombinant chimeric proteins in which segments of the human Arf1 and Arl1 proteins were exchanged (Zhang et al., 1995). These two proteins are, overall 53% identical in amino acid sequence (181 residues), but Arl1 is deficient in Arf activity, as defined by the usual criteria of 1) ability to stimulate cholera toxin ADP-ribosyltransferase activity, and, 2) to rescue yeast bearing the lethal double mutation of both arf1 and arf2 genes (Moss and Vaughan, 1995). On assay of the chimeric molecules, the protein containing the first 73 amino acids of Arf1 (and the remainder of Arl1) activated PLD to the same extent as did Arf1, but failed to activate the toxin. The latter was activated by the reciprocal chimera containing Arf sequence in positions 74181 and the first 73 amino acids of Arl1. Positions 37-55 in Arf correspond to the “effector loop” in Ras and Ras-related proteins, through which the 20 kDa GTPases produce many of their effects (Vetter and Wittinghofer, 2001). Thus, the mechanism of Arf activation of the ADP-ribosyltransferase activity of the toxin seems unlike that of its phospholipase D activation and at least some of its actions in cells (Zhang et al., 1995). These observations do not, of course, rule out the possibility that Arf stimulation of CTA1 and LTA1 activity proceeds by a mechanism that is employed by the GTPase for other, as yet unrecognized, biological purposes.

6.

CONCLUSIONS

Studies of the mechanism of action of CT continue to extend the understanding of diverse aspects of biology and medicine. Unfortunately, however, infection and disease caused by Vibrio cholerae remain serious worldwide problems that must be addressed. Cholera is especially prevalent and damaging in those parts of the world in which its prevention, through improvement of sanitation and other public health measures, is particularly difficult. When a person is infected, oral rehydration, with or without drugs that reduce diarrhea, can be effective therapy, only however, when accomplished in the relatively short time before effects of dehydration are irreversible. Too often, such seemingly simple measures are unavailable in the environment of epidemic cholera. Based on present knowledge of the mechanism of CT action, at least two events may be potentially useful targets for therapeutic intervention. Binding of CT to the intestinal mucosal cell might be prevented by the use of ganglioside GM1 analogues, as suggested by Merritt et al. (2002). This group synthesized a molecule composed of five m-nitrophenyl-Į-D-

ACTIVATION OF CT AND LT BY ARF

11 219

galactoside moieties, which interacts in a 1:1 molar ratio with the AB5 holotoxin. Inhibition of ADP-ribosyltransferase activity is another potentially efficacious approach. On investigation of herbal medications from several countries, it was determined that kampo formulations, or the gallate compounds that they contain, can effectively prevent or reduce the diarrheal effects of CT. The gallate derivatives inhibited ADP-ribosyltransferase activity of CT in vitro, when agmatine was used as a model substrate. They also inhibited the diarrheogenic activity of CT in an ileal loop model of in vivo action on the intestine (Oi et al., 2002), consistent with a significant role for this activity in disease. Thus, study of the mechanisms of action of CT and LT, their cell surface binding and entry into cells, as well as activation of GĮs by ADP-ribosylation can continue to provide clinically useful information.

REFERENCES Anderson, R. G. W. (1998). The caveolae membrane system. Annu. Rev. Biochem., 67, 199225. Botstein, D., Segev, N., Stearns, T., Hoyt, M. A., Holden, J., and Kahn, R. A. (1988). Diverse biological functions of small GTP-binding proteins in yeast. Cold Spring Harbor Symp. Quant. Biol., 53, 629-636. Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D. Cell, 75, 1137-1144. Cockcroft, S., Thomas, G. M. H., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of Arf in granulocytes. Science, 263, 523-526. Fishman, P. H. (1982). Internalization and degradation of cholera toxin by cultured cells: relationship to toxin action. J. Cell. Biol., 93, 860-865. Fishman, P. H., Moss, J., and Osborne, J. C. Jr. (1978). Interaction of choleragen with the oligosaccharide of ganglioside GM1: evidence for multiple oligosaccharide binding sites. Biochemistry, 17, 711-716. Henley, J. R., Krueger, E. W. A., Oswald, B. J., and McNiven, M. A. (1998). Dynaminmediated internalization of caveolae. J. Cell Biol., 141, 85-99. Higgins, M. K., and McMahon, H. T. (2002). Snap-shots of clathrin-mediated endocytosis. Trends Biochem. Sci., 27, 257-263. Kahn, R. A., and Gilman, A. G. (1984). Purification of a protein cofactor required for ADPribosylation of the stimulatory regulatory component of adenylate cyclase by cholera toxin. J. Biol. Chem., 259, 6228-6234. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP-binding protein. J. Biol. Chem., 261, 7906-7911. Kassis, S., Hagmann, J., Fishman, P. H., Chang, P. P., and Moss, J. (1982). Mechanism of action of cholera toxin on intact cells. Generation of A1 peptide and activation of adenylate cyclase. J. Biol. Chem., 257, 12148-12152.

12 220

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

Lencer, W. I., Constable C., Moe, S., Jobling, M. G., Webb, H. M., Ruston, S., Madara, J. L., Hirst, T. R., and Holmes, R. K. (1995). Targeting of cholera toxin and Escherichia coli heat-labile toxin in polarized epithelia: Role of COOH-terminal KDEL. J. Cell. Biol., 131, 951-962. Lencer, W. I., Constable, C., Moe, S., Rufo, P. A., Wolf, A., Jobling, M. G., Ruston, S. P., Madara, J. L., Holmes, R. K., and Hirst, T. R. (1997). Proteolytic activation of cholera toxin and Escherichia coli heat-labile toxin by entry into host epithelial cells. Signal transduction by a protease-resistant toxin variant. J. Biol. Chem., 272, 15562-15568. Massenburg, D., Han, J.-S., Liyanage, M., Patton, W. A., Rhee, S. G., Moss, J., and Vaughan, M. (1994). Activation of rat brain phospholipase D by ADP-ribosylation factors, 1,5, and 6: separation of ADP-ribosylation factor-dependent and oleate-dependent enzymes. Proc. Natl. Acad. Sci., USA, 91, 11718-11722. Merritt, E. A., and Hol, W. G. J. (1995). AB5 toxins. Curr. Opin. Struct. Biol., 5, 165-171. Merritt, E. A., Zhang, Z., Pickens, J. C., Ahn, M., Hol, W. G. J., and Fan, E. (2002). Characterization and crystal structure of a high-affinity pentavalent receptor-binding inhibitor for cholera toxin and E. coli heat-labile enterotoxin. J. Am. Chem. Soc., 124, 8818-8824. Merritt, E. A., Sarfaty, S., Van den Akker, F., L’hoir, C., Martial, J. A., and Hol, W. G. J. (1994). Crystal structure of cholera toxin B-pentamer bound to receptor GM1 pentasaccharide. Prot. Sci., 3, 166-175. Morinaga, N., Kaihou, Y., Vitale, N., Moss, J., and Noda, M. (2001). Involvement of ADPribosylation factor 1 in cholera toxin-induced morphological changes of Chinese hamster ovary cells. J. Biol. Chem., 276, 22838-22843. Moss, J., Manganiello, V. C., and Vaughan, M. (1976). Hydrolysis of nicotinamide adenine dinucleotide by choleragen and it’s a protomer: possible role in the activation of adenylate cyclase. Proc. Natl. Acad. Sci., USA, 73, 4424-4427. Moss, J., and Richardson, S. H. (1978). Activation of adenylate cyclase by heat-labile Escherichia coli enterotoxin. Evidence for ADP-ribosyltransferase activity similar to that of choleragen. J. Clin. Invest., 62, 281-285. Moss, J., Stanley, S. J., and Lin, M. C. (1979b). NAD glycohydrolase and ADPribosyltransferase activities are intrinsic to the A1 peptide to choleragen. J. Biol. Chem., 254, 11993-11996. Moss, J., Stanley, S. J., Morin, J. E., and Dixon, J. E. (1980). Activation of choleragen by thiol: protein disulfide oxidoreductase. J. Biol. Chem., 255, 11085-11087. Moss, J., Stanley, S. J., and Oppenheimer, N. J. (1979a). Substrate specificity and partial purification of a stereospecific NAD- and guanidine-dependent ADP-ribosyltransferase from avian erythrocytes. J. Biol. Chem., 254, 8891-8894. Moss, J., and Vaughan, M. (1977). Mechanism of action of choleragen. Evidence for ADPribosyltransferase activity with arginine as an acceptor. J. Biol. Chem., 252, 2455-2457. Moss, J., and Vaughan, M. (1995). Structure and function of Arf proteins: Activators of cholera toxin and critical components of intracellular vesicular transport processes. J. Biol. Chem., 270, 21431-21434. Moss, J., and Vaughan, M. (1998). Molecules in the Arf orbit. J. Biol. Chem., 273, 1232712330. Noda, M., Tsai, S.-C., Adamik, R., Moss, J., and Vaughan, M. (1990). Mechanism of cholera toxin activation by a guanine nucleotide-dependent 19 kDa protein. Biochim. Biophys. Acta, 1034, 195-199.

ACTIVATION OF CT AND LT BY ARF

13 221

Oh, P., McIntosh, D. P., and Schnitzer, J. E. (1998). Dynamin at the neck of caveolae mediates their budding to form transport vesicles by GTP-driven fission from the plasma membrane of endothelium. J. Cell Biol., 141, 101-114. Oi, H., Matsuura, D., Miyake, M., Ueno, M., Takai, I., Yamamoto, T., Kubo, M., Moss, J., and Noda, M. (2002). Identification in traditional herbal medications and confirmation by synthesis of factors that inhibit cholera toxin-induced fluid accumulation. Proc. Natl. Acad. Sci., USA, 99, 3042-3046. Orlandi, P. A. (1997). Protein-disulfide isomerase-mediated reduction of the A subunit of cholera toxin in a human intestinal cell line. J. Biol. Chem., 272, 4591-4599. Orlandi, P. A., Curran, P. K., and Fishman, P. H. (1993). Brefeldin A blocks the response of cultured cells to cholera toxin. J. Biol. Chem., 268, 12010-12016. Orlandi, P. A., and Fishman, P. H. (1998). Filipin-dependent inhibition of cholera toxin: evidence for toxin internalization and activation through caveolae-like domains. J. Cell Biol. 141, 905-915. Pacuszka, T., and Fishman, P. H. (1992). Intoxication of cultured cells by cholera toxin: evidence for different pathways when bound to ganglioside GM1 or neoganglioproteins. Biochemistry, 31, 4773-4778. Palade, G. E. (1975). Intracellular aspects of the process of protein synthesis. Science, 189, 347-358. Peterson, J. W., and Ochoa, L. G. (1989). Role of prostaglandins and cAMP in the secretory effects of cholera toxin. Science, 245, 857-859. Presley, J. F., Ward, T. H., Pfeifer, A. C., Siggla, E. D., Phair, R. D., and LippincottSchwartz, J. (2002). Dissection of COPI and Arf1 dynamics in vivo and role in Golgi membrane transport. Nature, 417, 187-193. Rodighiero, C., Aman, A. T., Kenny, M. J., Moss, J., Lencer, W. I., and Hirst, T. R. (1999). Structural basis for the differential toxicity of cholera toxin and Escherichia coli heatlabile enterotoxin. J. Biol. Chem., 274, 3962-3969. Schmitz, A., Herrgen, H., Winkeler, A., and Herzog, V. (2000). Cholera toxin is exported from microsomes by the Sec61p complex. J. Cell Biol., 140, 1203-1212. Sixma, T. K., Pronk, S. E., Kalk, K., Wartna, E. S., Van Zanten, B. A. M., Withold, B., and Hol, W. G. J. (1991). Crystal structure of a cholera toxin-related heat-labile enterotoxin from E. coli. Nature, 351, 371-377. Sofer, A., and Futerman, A. H. (1995). Cationic amphiphilic drugs inhibit the internalization of cholera toxin to the Golgi apparatus and the subsequent elevation of cyclic AMP. J. Biol. Chem., 270, 12117-12122. Torgersen, M. L., Skretting, G., Van Deurs, B., Sandvig, K. (2001). Internalization of cholera toxin by different endocytic mechanism. J. Cell Sci., 114, 3737-3747. Tran, D., Carpenter, J. L., Sawano, F., Gorden, P., and Orci, L. (1987). Ligands internalized through coated or non-coated invaginations follow a common intracellular pathway. Proc. Natl. Acad. Sci., USA, 84, 7957-7961. Tsai, S.-C., Noda, M., Adamik, R., Chang, P., Chen, H.-G., Moss, J., and Vaughan, M. (1988). Stimulation of choleragen enzymatic activities by GTP and two soluble proteins purified from bovine brain. J. Biol. Chem., 263, 1768-1772. Tsai, B., Rodighiero, C., Lencer, W. I., and Rapoport, T. A. (2001). Protein disulfide isomerase acts as a redox-dependent chaperone to unfold cholera toxin. Cell, 104, 937948. Vetter, J. R., and Wittinghofer, A. (2001). The guanine nucleotide-binding switch in three dimensions. Science, 294, 1299-1304.

14 222

PACHECO-RODRIGUEZ, MORINAGA, NODA, MOSS, AND VAUGHAN

Weigert, R., Silletta, M. G., Spano, S., Turrachio, G., Cericola, C., Conzi, A., Senatore, S., Mancini, R., Polishchuk, E. V., Salmona, M., Facchiang, F., Burger, K. N. J., Miranov, A., Luini, A., and Corda, D. (1999). CtBP/BARS induces fission of Golgi membranes by acylating lysophosphatidic acid. Nature, 402, 429-433. Zhang, G.-F., Patton, W. A., Lee, F.-J. S., Liyanage, M., Han, J.-S., Rhee, S. G., Moss, J., and Vaughan, M. (1995). Different Arf domains are required for the activation of cholera toxin and phospholipase D. J. Biol. Chem., 270, 21-24. Zhu, X., Boman, A. L., Kuai, J., Cieplak, W., and Kahn, R. A. (2000). Effectors increase the affinity of ADP-ribosylation factor for GTP to increase binding. J. Biol. Chem., 275, 13465-13475. Zhu, X., and Kahn, R. A. (2001). The Escherichia coli heat-labile toxin binds to Golgi membranes and alters Golgi and cell morphologies using ADP-ribosylation factordependent processes. J. Biol. Chem., 276, 25014-25021.

Chapter 11 PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN JOHN H. EXTON Department of Molecular Physiology, Biophysics and Pharmacology, Howard Hughes Medical Institute, Vanderbilt University, Nashville, USA n

Abstract:

The effects of Arfs on three potential effectors (phospholipase D, Arfaptins and Arfophilin) are described. Arfs activate mammalian PLD1 in vitro, but their roles in the regulation of PLD in vivo are not well defined. Direct binding of Arfs to PLD1 has not been demonstrated in vitro and activation may involve PI(4,5)P2, which binds to the enzyme. Evidence for a role of Arf6 in the regulation of PLD in vivo is accumulating and may partly involve a PIP kinase that is activated by this Arf. The role of Arf-stimulated PLD in Golgi traffic remains controversial and evidence for and against this is presented. The properties of the Arf-binding proteins Arfaptins 1 and 2 are described, and interactions of POR1, a truncated form of Arfaptin 2, with Arfs and Rac1 are discussed. A hypothesis for Arfaptin 2 as a communicator between these two small GTPases in vivo is presented. A possible role of Arfaptin 2 in Huntington’s disease is discussed. Finally, the properties of Arfophilin, which selectively binds ARFs 4, 5 and 6 are described.

1.

PHOSPHOLIPASE D

1.1

Regulation of the PLD1 Isozyme of Phospholipase D by Arfs in vitro

Arf was recognized as an activator of phospholipase D (PLD) in early experiments with partially purified preparations of the enzyme (Brown et al., 1993) or with neutrophils depleted of cytosol (Cockcroft et al., 1994). Studies with individual members of the Arf family indicated that all mammalian Arfs were capable of activating PLD in the presence of GTPγS (Massenburg et al., 1994; Brown et al., 1995; Tsai et al., 1998). The myristoylated forms of the Arfs were much more effective than the 223

2 224

EXTON

unmodified forms (Massenburg et al., 1994; Brown et al., 1995; Tsai et al., 1998). No large differences were observed in the potency or efficacy of different Arf members to activate PLD in vitro (Massenburg et al., 1994; Brown, 1995; Tsai et al., 1998). As noted for all other stimulators of PLD, phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) was required for the effect of Arf on the enzyme. Cloning studies have revealed the existence of two mammalian isozymes of PLD (PLD1 and PLD2), each of which exist as splice variants. These isozymes have four conserved sequences (I-IV) and sequences I and IV each contain the highly conserved HKD motif (Exton, 2002). Searches of the human genome database have not revealed other PLD isozymes. PLD1 is strongly stimulated by Arfs 1 and 3 in vitro (Hammond et al., 1995, 1997; Min et al., 1998; Sung et al., 1999b) whereas PLD2 shows little or no response (Colley et al., 1997; Kodaki and Yamashita, 1997; Lopez et al., 1998; Sung et al., 1999a). Marked synergism is observed between Arf and RhoA or protein kinase C in the activation of PLD1 or partially purified brain PLD (Hammond et al., 1997; Singer et al., 1996). The site at which Arf interacts with PLD1 has not been defined, but deletional mutagenesis of the first 325 amino acids of the N-terminus indicates that the site is not contained within this sequence (Sung et al., 1999b; Park et al., 1998). Surprisingly, deletion of the first 308 amino acids of PLD2 decreases the basal activity of the enzyme, but renders it highly responsive to Arfs 1 and 5 in vitro (Sung et al., 1999a). This suggests that the N-terminus may block binding of Arf or otherwise interfere with its ability to stimulate catalysis. This observation, plus the finding that PLD2 does show a weak response to Arf1 in vitro (Kodaki and Yamashita, 1997; Lopez et al., 1998; Sung et al., 1999a), suggests that this PLD isozyme could mediate part of the action of Arfs on PLD. The domain in Arf1 that interacts with PLD has been defined (Zhang et al., 1995; Jones et al., 1999). Initial work with chimeric proteins constructed between Arf1 and the structurally related, but inactive, protein ARL1 indicated that the first 73 amino acids at the N-terminus of Arf1 were required for activation of PLD (Zhang et al., 1995). The region comprising residues 35-94 was shown in another study to be required for PLD1 activation and AP-1 recruitment, but not for Arf1 activation by a Golgiassociated GEF (Liang et al., 1997). A later study utilized substitution and deletion mutagenesis and knowledge of the crystal structure of Arf1 to map the PLD interaction site to the α2 helix, part of the β2 strand and the Nterminal helix and its ensuing loop (Jones et al., 1999). Mutants that showed increased or decreased ability to activate PLD produced similar effects on secretion in HL60 cells (Jones et al., 1999). However, another study utilizing selective mutations showed that amino acid residues (I49T, F51Y)

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

3 225

in the switch I region of Arf3 were required for PLD activation, but only one of these (F51Y) was needed to promote vesiculation of Golgi (Kuai, et al., 2000). The mechanism by which Arf activates PLD is unclear. In in vitro experiments, recombinant Arf1 can activate recombinant PLD1 in the presence of GTPγS and PI(4,5)P2, which is included in the PLD assay mixture (Hammond et al., 1995; Min et al., 1998). Because PI(4,5)P2 is required for PLD activity (Brown et al., 1993; Hammond et al., 1995; Min et al., 1998; Sciorra et al., 1999), it is difficult to explore the role of this lipid in the action of Arf on the enzyme. As stated above, no interaction site for Arf on PLD has been defined, although binding of PI(4,5)P2 to the enzyme has been demonstrated (Sciorra et al., 1999; Hodgkin, et al., 2000). The binding of PI(4,5)P2 to Arf1 has been demonstrated and shown to increase nucleotide exchange (Randazzo, 1997; Terui et al., 1994). The binding promotes the interaction of Arf1 with an Arf GAP (Randazzo and Kahn, 1994) and with the Arf GEF ARNO (Paris et al., 1997), and this may be true for PLD. As discussed below, indirect effects of Arf family members on PLD isozymes in vivo are also likely.

1.2

Regulation of PLD by Arf in vivo – Roles of Arf6 and PIP Kinase

A role for Arf in the regulation of PLD in vivo is not as well defined as the roles of protein kinase C and Rho family members. Early experiments showed that depletion of the cytosol of HL-60 cells through detergent permeabilization caused loss of GTPγS-stimulated PLD activity, which could be restored by addition of brain cytosol (Geny et al., 1993). The restorative agent was subsequently shown to be Arf (Cockcroft et al., 1994). A similar role for Arf in the regulation of PLD in permeabilized neutrophils, mast cells and HEK cells has been reported (Rumenapp et al., 1995, 1997; Fensome et al., 1998; Way, et al., 2000). In the studies with neutrophils and mast cells, Arf also restored the response to FMLP (Fensome et al., 1998) or antigen (Way, et al., 2000). In other studies, addition of Arf1 or Arf6 to permeabilized A10 vascular smooth muscle cells or Rat 1 fibroblasts did not restore the activation of PLD by GTPγS, but restored the ability of angiotensin II, PDGF or PMA to activate the enzyme (Shome et al., 1998;, 2000). However, these results have not been observed in all cells (Glenn et al., 1998). Transfection of wild type or constitutively active forms of Arf3 does not activate PLD1 in COS-7 cells (Park et al., 1997), but this may be due to mislocalization of the expressed proteins or disruption of the Golgi. Isolated Golgi fractions possess Arf-stimulated PLD activity (Provost et al., 1996; Ktistakis et al., 1995, 1996), and PLD1 has been localized to the

4 226

EXTON

perinuclear region in many studies of intact cells. However, efforts to localize PLD1 specifically to Golgi in vivo have not given uniformly positive results (Colley et al., 1997; Chen et al., 1997; Freyberg, et al., 2001; Brown et al., 1998). Several studies of the regulation of PLD by Arf in vivo have utilized brefeldin A and have given divergent results. This is probably because this compound does not inhibit all Arf GEFs (Kam and Exton, 2001). Thus, a subclass of Arf GEFs containing certain residues in their Sec7 domains are targets of brefeldin A, whereas others e.g.. ARNO and cytohesin-1 are not (Moss and Vaughan, 1998; see Chapters 4-6 for additional details). Brefeldin A has been reported to inhibit carbachol stimulation of PLD in HEK cells expressing M3 muscarinic receptors (Rumenapp et al., 1995) and to impair activation of PLD by PMA and PDGF in Rat 1 fibroblasts (Shome et al., 1998). Other studies have shown inhibitory activity of this fungal metabolite on PLD activation by angiotensin II, endothelin 1, carbachol and PDGF in A10 vascular smooth muscle cells and 1321N1 astrocytoma cells (Shome, et al., 2000; Andresen, et al., 2001; Mitchell et al., 1998). In contrast, brefeldin A has no effects on the stimulation of PLD by ATP and FMLP in HL60 cells (Guillemain and Exton, 1997; Bourgoin et al., 1995) or by PMA, sphingosine 1-P or bradykinin in A549 adenocarcinoma cells (Meacci et al., 1999). It is not known whether these differences reflect different complements of Arf GEFs in these cells or different roles for Arfs in agonist regulation of PLD. Arf6 differs from the other mammalian Arfs in that it is associated with the plasma membrane (Meacci et al., 1999; Cavenagh et al., 1996; Gaschet and Hsu, 1999; Radhakrishna and Donaldson, 1997; Caumont et al., 1998, 2000; D’Souza-Schorey et al., 1995; Yang et al., 1998; Al-Awar, et al., 2000) and cycles between an intracellular compartment and the plasma membrane as a function of its GTP/GDP binding or agonist stimulation (Gaschet and Hsu, 1999; Radhakrishna and Donaldson, 1997; Caumont et al., 1998, 2000; Al-Awar, et al., 2000). Stimulation of chromaffin cells with nicotine causes activation of PLD, and there is a translocation of Arf6 to the plasma membranes that correlates with an enhancement of GTPγSstimulated PLD activity in this fraction (Caumont et al., 1998). In addition, a myristoylated peptide corresponding to residues 2-13 in Arf6 inhibited Ca2+-stimulated PLD activity in permeabilized chromaffin cells (Caumont et al., 1998). In contrast, the corresponding peptide from Arf1 was ineffective. The introduction of antibodies to ARNO, a brefeldin A-insensitive GEF for Arfs 1 and 6, into permeabilized chromaffin cells also inhibited Ca2+stimulated PLD activity. These findings indicate a role for Arf6 in the regulation of PLD in chromaffin cells, and this has been related to the regulation of exocytosis in these cells (Caumont et al., 1998, 2000).

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

5 227

Translocation of Arf1 to a membrane fraction in response to GTPγS, FMLP or PMA has been reported in HL-60 cells (Fensome et al., 1998; Houle et al., 1995; Martin et al., 1996), although the nature of the membrane fraction was not defined. Arf1-stimulated PLD activity has been localized to both the plasma membrane and endomembranes in subcellular fractionation studies in HL-60 cells (Whatmore et al., 1996). Thus Arf1 may also be involved in the regulation of PLD in these cells. As alluded to earlier, Arf may indirectly regulate PLD activity in vivo. This is through its effects on the synthesis of PI(4,5)P2, which is a major regulator of this phospholipase. Phosphatidylinositol phosphate kinase, the enzyme that synthesizes PI(4,5)P2 from PIP, exists in multiple forms in mammalian cells (Anderson et al., 1999). Thus Type I comprises several 5kinase isoforms that convert PI 4-P to PI(4,5)P2, and Type II comprises several 4-kinase isoforms that convert PI 5-P to PI(4,5)P2. Members of the Rho family of GTPases have been demonstrated to activate Type I PIP kinase (Exton, 2002), and recent work has shown that Arfs 1 and 6 can also activate this kinase (Martin et al., 1996; Honda et al., 1999; Godi et al., 1999; Jones et al., 2000; Skippen et al., 2002). As PI(4,5)P2 is absolutely required for PLD activity, this could represent a cellular mechanism for activation of PLD by Arf. This idea is supported by the co-localization of Type I PIP kinase with PLD2 in pulmonary artery endothelial cells coexpressing both enzymes (Jones et al., 2000) and the demonstration that both PLD1 and PLD2 interact with the Type I kinase in vivo, as shown by immunoprecipitation studies (Jones et al., 2000). Arf6 has also been shown to co-localize with Type I PIP kinase and PI(4,5)P2 in HeLa cells (Honda et al., 1999; Brown et al., 2001).

1.3

Roles of Arf-stimulated PLD in Golgi Traffic and Exocytosis

A critical issue is the role of Arf-regulated PLD in membrane traffic. This remains a controversial area. Early work showing that Arf could activate PLD and that Golgi preparations possessed Arf-stimulable PLD activity suggested that PLD was involved in traffic at the Golgi (Ktistakis et al., 1995; Kahn et al., 1993). This was supported by the observation that, in a cell line with high constitutive PLD activity, Arf was not required for the formation of COPI-coated vesicles (Ktistakis et al., 1996). Furthermore, addition of ethanol, which reduces phosphatidic acid (PA) formation by PLD, inhibited the formation of coated vesicles from Golgi, whereas addition of PLD to Golgi membranes induced COPI binding and vesicle formation (Ktistakis et al., 1996). Ethanol was also found to inhibit

6 228

EXTON

transport between the endoplasmic reticulum and Golgi, and the block could be reversed by addition of PA (Bi et al., 1997). In permeabilized GH3 pituitary cells, purified PLD was observed to stimulate secretory vesicle budding from the trans Golgi, whereas this was inhibited by primary butanol (Chen et al., 1997). Generation of PA through the combined action of phospholipase C and diacylglycerol kinase also stimulated vesicle budding (Siddhanta and Shields, 1998). Treatment of GH3 cells with primary butanol reduced transport from the endoplasmic reticulum and the Golgi, and also inhibited hormone secretion (Siddhanta et al., 2000). There was an associated decrease in PI(4,5)P2 synthesis by Golgi membranes and disassembly and fragmentation of the Golgi (Siddhanta et al., 2000). Exogenous PLD has also been shown to recruit AP-2 adaptors onto plasma membranes and thus mimic the effects of GTPγS plus Arf (West et al., 1997). This suggests a role for PLD in vesicle traffic at other cellular sites. Although it has been assumed that the effects of PLD on vesicle traffic are due to the effects of PA per se, they may be due to changes in PI(4,5)P2. PA has been reported to stimulate Type I PIP kinase (Skippen et al., 2002; Jenkins et al., 1994; Moritz et al., 1992; Ren et al., 1996) and there is much evidence that PI(4,5)P2 is required for traffic at several cellular sites (Skippen et al., 2002; Martin, 1998,, 2001; Simonsen et al., 2001). There is other evidence implicating a role for PLD in exocytosis. This has mainly relied on the ability of primary alcohols to inhibit secretion from many cell types (for references, see Cockcroft, 1996, and, Jones et al., 1999). However, it must be recognized that these reagents may not be specific for PLD when used in intact cells. However, exogenous PLD stimulates secretion in many cell types (Cockcroft, 1996) and permeabilization of cells to release Arf decreases GTPγS-stimulated PLD activity and exocytosis, both of which can be restored by adding Arf (Jones et al., 1999). Additional evidence for a role of PLD in exocytosis comes from studies of the effects of overexpressed wild type and catalytically inactive PLD1 in PC12 cells. These active and inactive enzymes stimulated and inhibited secretion, respectively (Vitale et al., 2001). On the other hand, wild type and inactive PLD2 were without effect. Ceramide is known to inhibit the stimulation of PLD by PMA and agonists in several cell types (Exton, 2002) and Arf may possibly be involved. Addition of C2 ceramide to HL-60 cells inhibits membrane translocation of Arf and PLD activation induced by FMLP (Abousalham et al., 1997). In chromaffin cells, C2 ceramide inhibits both PLD activity and catecholamine release (Vitale et al., 2001). In contrast to the preceding evidence supporting a role of Arf-stimulated PLD in membrane traffic at the Golgi and in exocytosis, there are findings that do not support this hypothesis. For example, measurements of PA

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

7 229

levels in Golgi during budding of COPI coated vesicles indicated that this lipid declined rather than increased (Abousalham et al., 1997). Another study examined the effects of mutations in Arf3 on PLD activity and the recruitment of COPI to Golgi membranes (Kuai et al., 2000) and observed a poor correlation. Similarly, an N-terminally deleted form of Arf1 was found to be without effect on Arf-stimulated PLD activity, but competed with Arf1 to prevent COPI binding to Golgi membranes (Jones et al., 1999). In another study, exogenous PLD was observed to recruit AP-2 adaptors onto the plasma membrane, but was unable to recruit AP-1 adaptors onto the trans Golgi network (West et al., 1997). In summary, it seems very unlikely that PLD alone mediates the effects of Arf on vesicle traffic at various sites. However, it probably plays a supporting role through the generation of PI(4,5)P2 and perhaps other lipids.

2.

ARFAPTINS

2.1

Properties of Arfaptins 1 and 2

Arfaptins 1 and 2 were identified as Arf-binding proteins by the yeast twohybrid system using constitutively active Arf3 as bait and an HL-60 cDNA library (Kanoh et al., 1997). The interactions were not seen with wild type Arf3 and were independent of the myristoylation state of the Arf. Sequencing revealed that the arfaptins were 60% identical at the amino acid level and 81% homologous, with the greatest homology being in the Cterminal half (Kanoh et al., 1997). Northern blot analysis revealed that the mRNAs for both Arfaptins were ubiquitously expressed, with high levels in liver, pancreas and placenta (Kanoh et al., 1997). The proteins bound predominantly to Class I Arfs compared with Class II and III Arfs, and only when the Arfs were in the GTP-bound form. Arfaptin 1 was recruited to Golgi membranes in vitro by GTPγS and this was dependent on Arf (Kanoh et al., 1997). Arfaptin 1 also localized to Golgi when transfected into COS 7cells. Arfaptin 2 was not examined in these experiments. The function of Arfaptin 1 was subsequently tested by examining its effects on the activation of PLD and cholera toxin ADP-ribosyltransferase by Arfs (Tsai et al., 1998). Arfaptin 1 inhibited the activation of both enzymes by all classes of Arf in a dose-dependent manner. The effects on Class II and III Arfs (Arfs 5 and 6) were less than those on Class I Arfs (Tsai et al., 1998). In agreement with previous results (Kanoh et al., 1997), myristoylation of Arf did not alter the inhibitory actions of Arfaptin 1. When tested against Arf1 in which the N-terminal 13 residues were replaced by the corresponding sequence for Arl1, Arfaptin 1 was unable to inhibit the

8 230

EXTON

action of this Arf mutant on the cholera toxin activity (Tsai et al., 1998). The corresponding experiments could not be conducted with PLD, because the Arf mutant was unable to activate this enzyme. These data imply the existence of an Arfaptin 1 binding site at the N-terminus of Arf1. In studies of guanine nucleotide binding and release from Arf3, Arfaptin 1 was found to increase the binding of GTPγS and GDP, but the effects were not large and may have been due to stabilization of Arf (Tsai et al., 1998). The release of GTPγS from Arf3 was also only slightly altered by Arfaptin 1. When tested in the presence of a guanine nucleotide exchange factor, only slight effects of Arfaptin 1 were observed on GTPγS or GDP binding (Tsai et al., 1998). These results indicate that Arfaptin 1 has little intrinsic GEF activity and does not modify the action of GEFs on Arf. The possible effects of Arfaptin 1 on the inactivation of Arf by GTPase-activating proteins were not examined. In binding studies of the interaction of Arfaptin 1 with Arf3, deletional mutagenesis of Arfaptin 1 indicated that sequences in both the N-and Ctermini were involved (Williger et al., 1999). Both were required for the inhibitory effect of Arfaptin 1 on Arf3 stimulation of PLD1 (Williger et al., 1999). Arfaptin 1 was also shown to associate with a 'high speed' fraction containing Golgi membranes in a GTPγS- and Arf3-dependent manner (Williger et al., 1999). The functional consequences of the effects of Arfaptin 1 on Arf interactions with PLD in Golgi were explored in NIH 3T3 cells overexpressing Arfaptin 1 (Williger et al., 1999). In these cells, the activation of PLD by PMA was suppressed and there was a slowing of vesicular traffic through the Golgi, as measured by the glycosylation of vesicular stomatitis virus (Williger et al., 1999). Consistent with an earlier report (Tsai et al., 1998), Arfaptin inhibited the effect of Arf3 on PLD activity in Golgi membranes.

2.2

Functions of POR1, a Truncated Form of Arfaptin2

An intriguing development in knowledge of the function of Arfaptin 2 came when an N-terminally truncated form (POR1) was identified in a yeast twohybrid screening of a Jurkat T cell cDNA library using active Rac1 as bait (Van Aelst et al., 1996). Its sequence was identical to that of Arfaptin 2, except that it lacked the N-terminal 38 amino acids (Van Aelst et al., 1996). In studies of the interactions between POR1 and members of the Rho family, the Arfaptin 2 truncation mutant was found to be selective for Rac1, but surprisingly, the interaction with wild type Rac1 was as strong as with constitutively active Rac1-V12. POR1 was localized to membrane ruffles in cells expressing Rac1-V12, and it was able to synergize with Ras-V12 (but not Rac1-V12) in the induction of membrane ruffles (Van Aelst et al., 1996).

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

9 231

Furthermore, N-terminally deleted POR1 was able to inhibit the effect of Rac1 to cause ruffling. In an extension of this work, mutants of Rac1 were found to interact differentially with POR1 and the protein kinase PAK1 (Joneson et al., 1996). PAK1 is implicated in the activation of the c-Jun Nterminal kinase (JNK) pathway by Rho family proteins, but not in the actin cytoskeleton changes (Lamarche et al., 1996). These results implicate POR1 in the actin skeleton rearrangements induced by Rac1. Future experiments directed toward defining the impact of the N-terminal truncation of Arfaptin2 on localization and function, as well as comparisons between the localization and cellular roles for Arfaptin 1 and 2 will help define the functions of this family of Arf effectors.

2.3

Interactions between Arfaptin 2/POR1 and Arfs and Rac1

In a later study of the actin cytoskeleton rearrangements induced by constitutively active Arf6-L67, it was found that the changes were different from those induced by Rac1-V12 and were not observed with Arf1-L71 (D’Souza-Schorey et al., 1997). The actin rearrangements caused by Arf6L67 were inhibited by N-terminally and C-terminally deleted POR1. Coexpression of wild type Arf6 and POR1 synergistically induced actin polymerization at the cell periphery (D’Souza-Schorey et al., 1997). The effects of Arf6-L67 were not blocked by dominant negative Rac1-N17, and dominant negative Arf6-N27 did not modify Rac1-V12-induced lamellipodia formation (D’Souza-Schorey et al., 1997). These data indicated that Arf6 and Rac1 were not in the same pathway to regulate cytoskeletal architecture, but operated in parallel. Further work showed that POR1 bound directly to the GTP-bound form of Arf6, and also to Arfs 1 and 3 (D’Souza-Schorey et al., 1997). In summary, these results provided evidence that Arfaptin 2/POR1 mediated the effects of Arf6 on the actin cytoskeleton, but its role in the actions of Rac1 was unclear. The role of Arfaptin 1 as a possible mediator of both Arf6 and Rac1 actions was explored further in two studies. The first involved a crystallographic study of the interaction of Arfaptin 2 and Rac1 (Tarricone et al., 2001). An N-terminally truncated form of the Arfaptin crystallized as a dimer encompassing the entire predicted coiled-coil region and dimerization involved packing of 2 or 3 helices from each monomer (Tarricone et al., 2001). Binding measurements utilizing isothermal titration calorimetry indicated that Rac1•GMPPNP and Rac1•GDP bound to Arfaptin 2 with similar KDs, and with 1:1 stoichiometry per dimer. Efforts to crystallize Arfaptin 2 in complex with Rac1•GMPPNP were not successful since the GMPPNP was hydrolyzed during the crystallization leading to a crystal with

10 232

EXTON

Rac1•GDP, i.e., Rac1 bound to GDP with a β-amidophosphate group. The structure showed that Rac1 sat at the midpoint of the Arfaptin 2 dimer and that the interaction involved both the switch I and II regions of the GTPase and helix A of the Arfaptin (Tarricone et al., 2001). It also revealed why Arfaptin 2 is selective for Rac1 rather than Rho or Cdc42. Crystallization of Rac1-L61•GMPPNP in complex with Arfaptin 2 was also undertaken. This GTPase-deficient form of Rac1 did not hydrolyze GMPPNP. However, the structure of the complex was very similar to that of Rac1•GDP-Arfaptin 2 i.e. there was little difference in the two switch regions (Tarricone et al., 2001). In particular, Thr35 of Rac1 did not interact with nucleotide-associated Mg2+, as in other Rac1•GTP-containing complexes, but interacted with Arfaptin 2. This is characteristic of GDPbound GTPases (Tarricone et al., 2001). It seems that, in contrast to other GTPases of the Ras superfamily, the switch regions of Rac and other Rho proteins are able to adopt similar conformations in the GDP- and GTPbound forms in complex with certain binding proteins (Tarricone et al., 2001). The binding of Arfaptin 2 to Arfs 1 and 6 was also studied (Tarricone et al., 2001). In contrast to the findings with Rac, and in agreement with earlier findings (Kanoh et al., 1997) Arfaptin 2 bound only to the GTP-liganded forms of these GTPases, and binding to Arf6 was weaker than to Arf1. Binding of Rac1 and Arf to Arfaptin 2 was mutually exclusive, indicating that they share part of the same binding site. This raises the possibility that activated Arf could displace Rac1 from its binding to Arfaptin 2. Subsequent activation of Rac1 could lead to signaling through this GTPase. In support of this idea (Tarricone et al., 2001), overexpression of EFA6, a GEF for Arf6, was found to lead to Rac activation and membrane ruffling (Franco et al., 1999). In another study of the interaction of Arfaptin 2 with Arfs and Rac1, it was found that constitutively active Arf1-L71, Arf5-L71 and Arf6-L67 all interacted with Arfaptin 2 in the yeast two-hybrid system (Shin and Exton, 2001), whereas the dominant negative forms of these GTPases did not. On the other hand, active Rac1-L61 did not interact with Arfaptin 2, whereas wild type and dominant negative Rac-N17 did (Shin and Exton, 2001). Pulldown experiments with GST-Arfaptin 2 also indicated that active Rac1-L61 was not bound, whereas the active forms of Arfs 1, 5 and 6 were. Finally, there was significant binding of Rac1 to the Arfaptin in the presence of GDPβS, but little in the presence of GTPγS, and the reverse was true for the Arfs (Shin and Exton, 2001). The differential interactions of Arfaptin 2 with the GTP-bound versus GDP-bound forms of Rac1 and Arf6 would provide for novel forms of communication between these two GTPases, as indicated above.

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

11 233

In a structural study of the interaction of Arfaptin 2 with Rac1 (Cherfils, 2001), it was found that the dimeric helical domain of Arfaptin 2 could be superimposed on a monomeric helical domain of Tiam1, which is a GEF for Rac. This structural mimicry was utilized to explore the mechanism of the GEF reaction, in which there is low affinity binding of the GDP-bound GTPase to the GEF, which converts to a high affinity complex with nucleotide-free GTPase. Entry of GTP then induces a low affinity complex leading to its dissociation. The Rac1•GDP complex with Arfaptin 2 was thus considered to mimic the first step of a GEF reaction.

2.4

PICK1 Proteins are Homologous to Arfaptins

Another interesting finding indicating similarity between Arfaptins and other proteins potentially involved in signaling is the observation that the PICK1 proteins have sequence homology with Arfaptins 1 and 2 (Takeya et al., 2000). The PICK1 proteins are PDZ domain-containing proteins that were cloned in a yeast two-hybrid screen using PKCα as bait (Staudinger et al., 1995). However, PICK1 does not interact with Arfs. On the other hand, its PDZ domain binds to GTP-bound Arfs 1 and 3, but not Arfs 5 and 6, and not when Arfs 1 and 3 are GDP-bound. The association is lost if the PDZ domain is mutated and if the C-terminus of Arf1 is deleted (Staudinger et al., 1995). It was suggested that PICK1 might participate in Arf1- and 3mediated cellular processes.

2.5

Possible Role of Arfaptin2 in Huntington’s Disease

A possible role of Arfaptin 2 in Huntington’s disease is suggested by a recent report (Peters et al., 2002). This inherited neurodegenerative disease is associated with nuclear and cytosolic aggregates of the protein huntingtin. Expression of Arfaptin 2 in PC12 cells resulted in deposition of dense aggregates of this protein in circular nuclear structures and also around the centriole (Peters et al., 2002). Because the pattern resembled that seen with mutant huntingtin, the distribution of endogenous huntingtin was examined, and it was found to co-localize with expressed Arfaptin 2 in the nuclear region. Introduction of GFP-tagged mutant huntingtin also showed colocalization with Arfaptin 2 (Peters et al., 2002). Deletion mutants of Arfaptin 2 were then tested. A C-terminally truncated form promoted aggregate function as well as the wild-type protein, but an N-terminally truncated form had greatly impaired ability, and huntingtin was diffusely present in the cytoplasm. Immunoprecipitation experiments also revealed that huntingtin and Arfaptin 2 were able to interact in vitro (Peters et al., 2002). Involvement of the ubiquitin-proteasome system in the effects of

12 234

EXTON

Arfaptin 2 was suggested by the effects of lactacystin, an inhibitor of this system. In wild-type cells, this inhibitor enhanced huntingtin-induced aggregation, whereas in cells transfected with Arfaptin 2, aggregation was minimal (Peters et al., 2002). It was suggested that Arfaptin 2 promotes huntingtin aggregation by inhibiting ubiquitin-proteasome function. In transgenic mice expressing huntingtin, increased levels of Arfaptin 2 were found in striatum, cortex and cerebellum, and overlapped with huntingtin aggregates (Peters et al., 2002). These data suggest that Arfaptin 2 can regulate huntingtin aggregates, possibly by inhibiting proteasome activity.

3.

ARFOPHILIN

Arfophilin is an Arf-binding protein discovered in a yeast two-hybrid screen of a kidney cDNA library using Arf5 as bait (Shin et al., 1999). This 82 kDa protein was initially shown to be specific for Class II Arfs (Arfs 4 and 5) and only when these were in the GTP-liganded form. Northern analysis showed Arfophilin to be expressed at high levels in heart and skeletal muscle (Shin et al., 1999). The Arf binding site was determined by deletional analysis to be located in the C-terminus (amino acids 612-756). This contains a coiled coil structure, but this alone is not enough for Arf5 binding (Shin et al., 1999). Through the use of Arf5/Arf3 chimeras the binding site for Arfophilin in Arf5 was localized to the N-terminal 17 amino acids. Because this sequence is different in Arf6 it was assumed incorrectly that Class III Arfs would not bind to Arfophilin (see below). Arfophilin could be translocated to the membrane fraction in CHO-K1 cell lysates in response to GTPγS or Arf5-L71, suggesting that active Arf5 and arfophilin could associate in vivo. More detailed studies of the interactions of Arfophilin with different classes of Arf confirmed that it did not bind to Arf1, but surprisingly bound to Arf6 as well as Arf5 (Shin et al., 2001). Using chimeric constructs, the Arf6 binding site was localized to a sequence towards the N-terminus (amino acids 37-80). Surprisingly, this is different from the C-terminal sequence (amino acids 612-756) involved in binding Arf5 (Shin et al., 1999). Overexpression of Arfophilin in CHO-K1 cells led to its localization to the perinuclear region, including Golgi, presumably in association with Arf5. There was also some localization to the periphery of the cell and this may involve association with Arf6. The function of Arfophilin remains unknown. However, there is evidence that it is overexpressed in human pancreatic cancer (Shin et al., 2001).

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

13 235

REFERENCES Abousalham, A., Loissis, C., O’Brien, L., and Brindley, D. N. (1997). Cell-permeable ceramides prevent the activation of phospholipase D by ADP-ribosylation factor and RhoA. J. Biol. Chem., 272, 1069-1075. Al-Awar, O., Radhakrishna, H., Powell, N. N., and Donaldson, J. G. (2000). Separation of membrane trafficking and actin remodeling functions of Arf6 with an effector domain mutant. Mol. Cell. Biol., 20, 5998-6007. Anderson, R. A., Boronenkov, I. V., Doughman, S. D., Kunz, J., and Loijens, J. C. (1999). Phosphatidylinositol phosphate kinases, a multifaceted family of signaling enzymes. J. Biol. Chem., 274, 9907-9910. Andresen, B. T., Jackson, E. K., and Romero, G. G. (2001). Angiotensin II signaling to phospholipase D in renal microvascular smooth muscle cells in SHR. Hepatology, 37, 635-639. Bi, K., Roth, M. G., and Ktistakis, N. T. (1997). Phosphatidic acid formation by phospholipase D is required for transport from the endoplasmic reticulum to the Golgi complex. Curr. Biol., 7, 301-307. Bourgoin, S., Harbour, D., Desmarais, Y., Takai, Y., and Beaulieu, A. (1995). Low molecular weight GTP-binding proteins in HL-60 granulocytes. J. Biol. Chem., 270, 3172-3178. Brown, F. D., Thomas, N., Saqib, K. M., Clark, J. M., Powner, D., Thompson, N. T., Solari, R., and Wakelam, M. J. O. (1998). Phospholipase D1 localizes to secretory granules and lysosomes and is plasma membrane translocated on cellular stimulation. Curr. Biol., 8, 835-838. Brown, H. A., Gutowski, S., Kahn, R. A., and Sternweis, P. C. (1995). Partial purification and characterization of Arf-sensitive phospholipase D from porcine brain. J. Biol. Chem., 270, 14935-14943. Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity. Cell, 75, 1137-1144. Caumont, A.-S., Galas, M.-C., Vitale, N., Aunis, D., and Bader, M.-F. (1998). Regulated exocytosis in chromaffin cells. Translocation of Arf6 stimulates a plasma membraneassociated phospholipase D. J. Biol. Chem., 273, 1373-1379. Caumont, A.-S., Vitale, N., Gensse, M., Galas, M-C, Casanova, J. E., and Bader, M.-F. (2000). Identification of a plasma membrane-associated guanine nucleotide exchange factor for Arf6 in chromaffin cells. Possible role in the regulated exocytotic pathway. J. Biol. Chem., 275, 15637-15644. Cavenagh, M. M., Whitney, J. A., Carroll, K., Zhang, C.-J., Bowman, A.L., Rosenwald, A. G., Mellman, I., and Kahn, R. A. (1996). Intracellular distribution of Arf proteins in mammalian cells. Arf6 is uniquely localized to the plasma membrane. J. Biol. Chem., 271, 21767-21774. Chen, Y.-G., Siddhanta, A., Austin, C. D., Hammond, S. M., Sung, T.-C., Frohman, M. A., Morris, A. J., and Shields, D. (1997). Phospholipase D stimulates release of nascent secretory vesicles from the trans-Golgi network. J. Cell Biol. 138, 495-504 Cherfils, J. (2001). Structural mimicry of DH domains by Arfaptin suggests a model for the recognition of Rac-GDP by its guanine nucleotide exchange factor. FEBS Lett., 507, 280284. Cockcroft, S. (1996). Arf-regulated phospholipase D: a potential role in membrane traffic. Chem. Physics Lipids, 80, 59-80.

14 236

EXTON

Cockcroft, S., Thomas, G. M. H., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of Arf in granulocytes. Science, 263, 523-526. Colley, W. C., Sung, T.-C., Roll, R., Jenco, J., Hammond, S. M., Altshuller, Y., Bar-Sagi, D., Morris, A. J., and Frohman, M. A. (1997). Phospholipase D2, a distinct phospholipase D isoform with novel regulatory properties that provokes cytoskeletal reorganization. Curr. Biol., 7, 191-20 D’Souza-Schorey, C., Li, G., Colombo, M. I., and Stahl, P. D. (1995). A regulatory role for Arf6 in receptor-mediated endocytosis. Science, 267, 1175-1178. D’Souza-Schorey, C., Boshans, R. L., McDonought, M., Stahl, P. D., and Van Aelst, L. (1997). A role for POR1, a Rac1-interacting protein, in Arf6-mediated cytoskeletal rearrangements. EMBO J., 16, 5445-5454. Exton, J. H. (2002). Phospholipase D – structure, regulation and function. In S. G. Amara, E. Bamberg, M. P. Blaustein, H. Grunick, R. J. Gottingen, W. J. Lederer, A. Miyajima, H. Murer, N. Pfanner, G. Schultz, M. Schweiger (Eds.), Reviews of physiology, biochemistry and pharmacology (pp. 1-94). Berlin, Heidelberg: Springer-Verlag. Fensome, A., Whatmore, J., Morgan, C., Jones, D., and Cockcroft, S. (1998). ADPribosylation factor and Rho proteins mediate fMLP-dependent activation of phospholipase D in human neutrophils. J. Biol. Chem., 273, 13157-13164. Franco, M., Peters, P. J., Boretton, J., van Donselaar, E., Neri, A., D’Souza-Schorey, C., and Chavrier, P. (1999). EFA6 a Sec7 domain-containing exchange factor for Arf6 coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 1480-1491. Freyberg, Z., Sweeney, D., Siddhanta, A., Bourgoin, S., Frohman, M., and Shields, D. (2001). Intracellular localization of phospholipase D1 in mammalian cells. Mol. Biol. Cell, 12, 943-955. Gaschet, J., and Hsu, V. W. (1999). Distribution of Arf6 between membrane and cytosol is regulated by its GTPase cycle. J. Biol. Chem., 274, 20040-20045. Geny, B., Fensome, A., and Cockcroft, S. (1993). Rat brain cytosol contains a factor which reconstitutes guanine-nucleotide-binding-protein-regulated phospholipase–D activation in HL 60 cells previously permeabilized with streptolysin O. Eur. J. Biochem., 215, 389-396. Glenn, D. E., Thomas, G. M. H., O’Sullivan, A. J., and Burgoyne, R. D. (1998). Examination of the role of ADP-ribosylation factor and phospholipase D activation in regulated exocytosis in Chromaffin and PC12 cells. J. Neurochem., 71, 2023-2033. Godi, A., Pertile, P., Meyers, R., Marra, P., DiTullio, G., Iurisci, C., Luini, A., Corda, D., and DeMatteis, M. A. (1999). Arf mediates recruitment of PtdIns-4-OH kinase-β and stimulates synthesis of PtdIns(4,5)P2 on the Golgi complex. Nat. Cell Biol., 1, 280-287. Gress, T. M., Muller-Pillasch, F., Geng, M., Zimmerhackl, F., Zehetner, G., Friess, H., Buchler, M., Adler, G., and Lehrach, H. (1996). A pancreatic cancer-specific expression profile. Oncogene, 13, 1819-1830. Guillemain, I., and Exton, J. H. (1997). Effects of brefeldin A on phosphatidylcholine phospholipase D and inositol phospholipid metabolism in HL-60 cells. Eur. J. Biochem., 249, 812-819. Hammond, S. M., Altshuller, Y. M., Sung, T.-C., Rudge, S. A., Rose, K., Engebrecht, J. A., Morris, A. J., and Frohman, M. A. (1995). Human ADP-ribosylation factor-activated phosphatidylcholine-specific phospholipase D defines a new and highly conserved gene family. J. Biol. Chem., 270, 29640-29643. Hammond, S. M., Jenco, J. M., Nakashima, S., Cadwallader, K., Gu, Q.-m., Cook, S., Nozawa, Y., Prestwich, G. D., Frohman, M. A., and Morris, A. J. (1997). Characterization of two alternately spliced forms of phospholipase D1. Activation of the purified enzymes

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

15 237

by phosphatidylinositol 4,5-bisphosphate, ADP-ribosylation factor, and Rho family monomeric GTP-binding proteins and protein kinase C-α. J. Biol. Chem., 272, 3860-3868. Hodgkin, M. N., Masson, M. R., Powner, D., Saqib, K. M., Ponting, C. P., and Wakelam, M. J. O. (2000). Phospholipase D regulation and localization is dependent upon a phosphatidylinositol 4,5-bisphosphate-specific PH domain. Curr. Biol., 10, 43-46. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase α is a downstream effector of the small G protein Arf6 in membrane ruffle formation. Cell, 99, 521-532. Houle, M. G., Kahn, R. A., Naccache, P. H., and Bourgoin, S. (1995). ADP-ribosylation factor translocation correlates with potentiation of GTPγS-stimulated phospholipase D activity in membrane fractions of HL-60 cells. J. Biol. Chem., 270, 22795-22800. Jackson, C. L., and Casanova, J. E. (2000). Turning on Arf: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. Jenkins, G. H., Fisette, P. L., and Anderson, R. A. (1994). Type I phosphatidylinositol 4phosphate 5-kinase isoforms are specifically stimulated by phosphatidic acid. J. Biol. Chem., 269, 11547-11554. Jones, D., Morgan, C., and Cockcroft, S. (1999). Phospholipase D and membrane traffic. Potential roles in regulated exocytosis, membrane delivery and vesicle budding. Biochim. Biophys. Acta, 1439, 229-244. Jones, D. H., Bax, B., Fensome, A., and Cockcroft, S. (1999). ADP ribosylation factor 1 mutants identify a phospholipase D effector region and reveal that phospholipase D participates in lysosomal secretion but is not sufficient for recruitment of coatomer 1. Biochem. J., 341, 185-192. Jones, D. H., Morris, J. B., Morgan, C. P., Kondo, H., Irvine, R. F., and Cockcroft, S. (2000). Type I phosphatidylinositol 4-phosphate 5-kinase directly interacts with ADP-ribosylation factor 1 and is responsible for phosphatidylinositol 4,5-bisphosphate synthesis in the Golgi compartment. J. Biol. Chem,. 275, 13962-13966. Joneson, T., McDonough, M., Bar-Sagi, D., and Van Aelst, L. (1996). RAC regulation of actin polymerization and proliferation by a pathway distinct from Jun kinase. Science, 274, 1374-1376. Kahn, R. A., Yucel, J. K., and Malhotra, V. (1993). Arf signaling: A potential role for phospholipase D in membrane traffic. Cell, 75, 1045-1048. Kam, Y., and Exton, J. H. (2001). Phospholipase D activity is required for actin stress fiber formation in fibroblasts. Mol. Cell. Biol., 21, 4055-4066. Kanoh, H., Williger, B.-T., and Exton, J. H. (1997). Arfaptin 1, a putative cytosolic target protein of ADP-ribosylation factor, is recruited to Golgi membranes. J. Biol. Chem., 272, 5421-5429. Kodaki, T., and Yamashita, S. (1997). Cloning, expression, and characterization of a novel phospholipase D complementary DNA from rat brain. J. Biol. Chem., 272, 11408-11413. Ktistakis, N. T., Brown, H. A., Sternweis, P. C., and Roth, M. G. (1995). Phospholipase D is present in Golgi-enriched membranes and its activation by ADP-ribosylation factor is sensitive to brefeldin A. Proc. Natl. Acad. Sci., USA, 92, 4952-4956. Ktistakis, N. T., Brown, H. A., Waters, M. G., Sternweis, P. C., and Roth, M. G. (1996). Evidence that phospholipase D mediates ADP ribosylation factor-dependent formation of Golgi coated vesicles. J. Cell Biol., 134, 295-306. Kuai, J., Boman, A. L., Arnold, R. S., Zhu, X., and Kahn, R. A. (2000). Effects of activated ADP-ribosylation factors on Golgi morphology require neither activation of phospholipase D1 nor recruitment of coatomer. J. Biol. Chem., 275, 4022-4032.

16 238

EXTON

Lamarche, N., Tapon, N., Stowers, L., Burbelo, P. D., Aspenstrom, P., Bridge, T., Chant, J., and Hall, A. (1996). Rac and Cdc42 induce actin polymerization and G1 cell cycle progression independently of p65pAK and the JNK/SAPK MAP kinase cascade. Cell, 87, 519-529. Liang, J. O., Sung, T.-C., Morris, A. J., Frohman, M.A., and Kornfeld, S. (1997). Different domains of mammalian ADP-ribosylation factor 1 mediate interaction with selected target proteins. J. Biol. Chem., 272, 33001-33008. Lopez, I., Arnold, R. S., and Lambeth, J. D. (1998). Cloning and initial characterization of a human phospholipase D (hPLD2). J. Biol. Chem., 273, 12846-12852. Martin, A., Brown, F. D., Hodgkin, M. N., Bradwell, A. J., Cook, S. J., and Hart, M. (1996). Activation of phospholipase D and phosphatidylinositol 4-phosphate 5-kinase in HL60 membranes is mediated by endogenous Arf but not Rho. J. Biol. Chem., 271, 1739717403. Martin, T. F. J. (1998). Phosphoinositide lipids as signaling molecules: common themes for signal transduction, cytoskeletal regulation, and membrane trafficking. Annu. Rev. Cell Dev. Biol., 14, 231-264. Martin, T. F. J. (2001). PI(4,5)P2 regulation of surface membrane traffic. Curr. Opin. Cell Biol., 13, 493-499. Massenburg, D., Han, J.-S., Liyanage, M., Patton, W. A., Rhee, S. G., Moss, J., and Vaughan, M. (1994). Activation of rat brain phospholipase D by ADP-ribosylation factors 1, 5, and 6: separation of ADP-ribosylation factor-dependent and oleate-dependent enzymes. Proc. Natl. Acad. Sci., USA, 91, 11718-11722. Meacci, E., Vasta, V., Moorman, J. P., Bobak, D. A., Bruni, P., Moss, J., and Vaughan, M. (1999). Effect of Rho and ADP-ribosylation factor GTPases on phospholipase D activity in intact human adenocarcinoma A549 cells. J. Biol. Chem., 274, 18605-18612. Min, D. S., Park, S.-K, and Exton, J. H. (1998). Characterization of a rat brain phospholipase D isozyme. J. Biol. Chem., 273, 7044-7051. Mitchell, R., McCulloch, D., Lutz, E., Johnson, M., MacKenzie, C., Fennell, M., Fink, G., Zhou, W., and Sealfon, S. C. (1998). Rhodopsin-family receptors associate with small G proteins to activate phospholipase D. Nature, 392, 411-414. Moritz, A., DeGraan, P. N. E., Gispen, W. H., and Wirz, K. W. A. (1992). Phosphatidic acid is a specific activator of phosphatidylinositol-4-phosphate kinase. J. Biol. Chem., 267, 7207-7210. Moss, J., and Vaughan, M. (1998). Molecules in the Arf orbit. J. Biol. Chem., 273, 2143121434. Paris, S., Beraud-Dufour, S., Robineau, S., Bigay, J., Antonny, B., Chabre, M., and Chardin, P. (1997). Role of protein-phospholipid interactions in the activation of Arf1 by the guanine nucleotide exchange factor Arno. J. Biol. Chem., 272, 22221-22226. Park, S.-K., Min, D. S., and Exton, J. H. (1998). Definition of the protein kinase C interaction site of phospholipase D. Biochem. Biophys. Res. Comm., 244, 364-367. Park, S.-K., Provost, J. J., Bae, C. D., Ho, W.-T., and Exton, J. H. (1997). Cloning and characterization of phospholipase D from rat brain. J. Biol. Chem., 272, 29263-29271. Peters, P. J., Ning, K., Palacios, F., Boshans, R. L., Kazanstev, A., Thompson, L. M., Woodman, B., Bates, G. P., and D’Souza-Schorey, C. (2002). Arfaptin 2 regulates the aggregation of mutant huntingtin protein. Nature Cell Biol., 4, 240-245. Provost, J. J., Fudge, J., Israelit, S., Siddiqi, A. R., and Exton, J. H. (1996). Tissue-specific distribution and subcellular distribution of phospholipase D in rat: evidence for distinct RhoA- and ADP-ribosylation factor (Arf)-regulated isoenzymes. Biochem. J., 319, 285291.

PHOSPHOLIPASE D, ARFAPTINS AND ARFOPHILIN

17 239

Radhakrishna, H., and Donaldson, J. G. (1997). ADP-ribosylation factor 6 regulates a novel plasma membrane recycling pathway. J. Cell Biol., 139, 49-61. Randazzo, P. A. (1997). Functional interaction of ADP-ribosylation factor 1 with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 272, 7688-7692. Randazzo, P. A., and Kahn, R. A. (1994). GTP hydrolysis by ADP-ribosylation factor is dependent on both an ADP-ribosylation factor GTPase-activating protein and acid phospholipids. J. Biol. Chem., 269, 10758-10763. Ren, X.-D., Bokoch, G. M., Traynor-Kaplan, A., Jenins, G. H., Anderson, R. A., and Schwartz, M. A. (1996). Physical association of the small GTPase Rho with a 68-kDa phosphatidylinositol 4-phosphate 5-kinase in Swiss 3T3 cells. Mol. Biol. Cell, 7, 435-442. Rümenapp, U., Geiszt, M., Wahn, F., Schmidt, M., and Jakobs, K. H. (1995). Evidence for ADP-ribosylation-factor-mediated activation of phospholipase D by m3 muscarinic acetylcholine receptor. Eur. J. Biochem., 234, 240-244. Rümenapp, U., Schmidt, M., Wahn, F., Tapp, E., Grannass, A., and Jakobs, K. H. (1997). Characteristics of protein-kinase-Cand ADP-ribosylation-factor-stimulated phospholipase D activities in human embryonic kidney cells. Eur. J. Biochem., 248, 407414. Sciorra, V. A., Rudge, S. A., Prestwich, G. D., Frohman, M. A., Engebrecht, J. A., and Morris, A. J. (1999). Identification of a phosphoinositide-binding motif that mediates activation of mammalian and yeast phospholipase D isoenzymes. EMBO J., 20, 59115921. Shin, O.-H., Couvillon, A. D., and Exton, J. H. (2001). Arfophilin is a common target of both Class II and Class III ADP-ribosylation factors. Biochem., 36, 10846-10852. Shin, O.-H., and Exton, J. H. (2001). Differential binding of Arfaptin 2/POR1 to ADPribosylation factors and Rac1. Biochem. Biophys. Res. Comm., 285, 1267-1273. Shin, O.-H., Ross, A. H., Mihai, I., and Exton, J. H. (1999). Identification of arfophilin, a target protein for GTP-bound Class II and ADP-ribosylation factors. J. Biol. Chem., 274, 36609-36615. Shome, K., Nie, Y., and Romero, G. (1998) ADP-ribosylation factor proteins mediate agonist-induced activation of phospholipase D. J. Biol. Chem., 273, 30836-30841. Shome, K., Rizzo, M. A., Vasudevan, C., Andresen, B., and Romero, G. (2000). The activation of phospholipase D by endothelin-1, angiotensin II, and platelet-derived growth factor in vascular smooth muscle A10 cells is mediated by small G proteins of the ADPribosylation factor family. Endocrinol., 141, 2200-2208. Siddhanta, A., Backer, J. M., and Shields, D. (2000). Inhibition of phosphatidic acid synthesis alters the structure of the Golgi apparatus and inhibits secretion in endocrine cells. J. Biol. Chem, 275, 12023-12031. Simonsen, A., Wurmster, A. E., Emr, S. D., and Stenmark, H. (2001). The role of phosphoinositides in membrane transport. Curr. Opin. Cell Biol., 13, 485-492. Singer, W. D., Brown, H. A., Jiang, X., and Sternweis, P. C. (1996). Regulation of phospholipase D by protein kinase C is synergistic with ADP-ribosylation factor and independent of protein kinase activity. J. Biol. Chem., 271, 4504-4510. Skippen, A., Jones, D. H., Morgan, C. P., Li, M., and Cockcroft, S. (2002). Mechanism of ADP ribosylation factor-stimulated phosphatidylinositol 4,5-bisphosphate synthesis in HL60 cells. J. Biol. Chem., 277, 5823-5831. Staudinger, J., Zhou, J., Burgess, R., Elledge, S. J., and Olsen, E. N. (1995). PICK1: a perinuclear binding protein and substrate for protein kinase C isolated by the yeast twohybrid system. J. Cell Biol., 128, 263-271. Sung, T.-C., Altshuller, Y. M., Morris, A. J., and Frohman M. A. (1999a). Molecular analysis of mammalian phospholipase D2. J. Biol. Chem., 274, 494-502.

18 240

EXTON

Sung, T.-C., Zhang, Y., Morris, A. J., and Frohman, M. A. (1999b). Structural analysis of human phospholipase D1. J. Biol. Chem., 274, 3659-3666. Takeya, R., Takeshige, K., and Sumimoto, H. (2000). Interaction of the PDZ domain of human PICK1 with class 1 ADP-ribosylation factors. Biochem. Biophys. Res. Comm., 267, 149-155. Tarricone, C., Xiao, B., Justin, N., Walker, P. A., Rittinger, K., Gamblin, S. J., and Smnerdon, S. J. (2001). The structural basis of Arfaptin-mediated cross talk between Rac and Arf signaling pathways. Nature, 411, 215-219. Terui, T., Kahn, R. A., and Randazzo, P. A. (1994) Effects of acid phospholipids on nucleotide exchange properties of ADP-ribosylation factor 1. Evidence for specific interaction with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 269, 28130-28135. Tsai, S.-C., Adamik, R., Hong, J.-X., Moss, J., Vaughan, M., Kanoh, H., and Exton, J. H. (1998). Effects of Arfaptin 1 on guanine nucleotide-dependent activation of phospholipase D and cholera toxin by ADP-ribosylation factor. J. Biol. Chem., 273, 20697-20701. Van Aelst, L., Joneson, T., and Bar-Sagi, D. (1996). Identification of a novel Rac1-interacting protein involved in membrane ruffling. EMBO J., 15, 3778-3786. Vitale, N., Caumont, A-S, Chasserot-Golaz, S., Du, G., Wu, S., Sciorra, V. A., Morris, A. J., Frohman, M. A., and Bader, M-F. (2001) Phospholipase D1: a key factor for the exocytotic machinery in neuroendocrine cells. EMBO J., 20, 2424-2434. Way, G., O’Luanaigh, N., and Cockcroft, S. (2000). Activation of ecovytosis by cross-linking of the IgE receptor is dependent on ADP-ribosylation factor 1-regulated phopholipase D in RBL-2H3 mast cells: evidence that the mechanism of activation is via regulation of phosphatidylinositol 4,5-bisphosphate synthesis. Biochem. J., 345, 63-70. West, M. A., Bright, N. A., and Robinson, M. S. (1997). The role of ADP-ribosylation factor and phospholipase D in adaptor recruitment. J. Cell Biol., 138, 1239-1254, Whatmore, J., Morgan, C. P., Cunningham, E., Collison, K. S., Willison, K. R., and Cockcroft, S. (1996) ADP-ribosylation factor 1-regulated phospholipase D activity is localized at the plasma membrane and intracellular organelles in HL60 cell. Biochem. J., 320, 785-794. Williger, B.-T., Osterman, J., and Exton, J. H. (1999). Arfaptin 1, an Arf-binding protein, inhibits phospholipase D and endoplasmic reticulum/Golgi protein transport. FEBS Lett., 443, 197-200. Williger, B.-T., Provost, J. J., Ho, W.-T., Milstine, J., and Exton, J. H. (1999). Arfaptin 1 forms a complex with ADP-ribosylation factor and inhibits phospholipase D. FEBS Lett., 454, 85-89. Yang, C. Z., Heimberg, H., D’Souza-Schorey, C., Mueckler, M. M., and Stahl, P. D. (1998). Subcellular distribution and differential expression of endogenous ADP-ribosylation factor 6 in mammalian cells. J. Biol. Chem., 273, 4006-4011. Zhang, G.-F., Patton, W. A., Lee, F.-J. L., Liyanage, M., Han, J.-S., Rhee, S.G., Moss, J., and Vaughan, M. (1995). Different Arf domains are required for the activation of cholera toxin and phospholipase D. J. Biol. Chem., 270, 21-24.

Chapter 12 COAT PROTEINS ANNETTE BOMAN1 AND TOMMY NILSSON2* 1

Department of Biochemistry and Molecular Biology, University of Minnesota School of Medicine Duluth, Duluth, USA; 2European Molecular Biology Laboratory Heidelberg, Heidelberg, Germany; * contact person for this chapter

Abstract:

1.

A major cellular function of Arfs is the regulation of membrane traffic. Many studies, using several independent approaches, have shown that activated Arfs directly recruit coat complexes to membranes. These coats sort cargo and facilitate vesicle formation. Three classes of Arf-dependent coat proteins or complexes have been described so far, including COPs, adaptins, and GGAs. This chapter focuses on aspects of Arf-dependent coat proteins and complexes. We will discuss the binding between Arfs and coats, with an emphasis on COPI and GGAs, as well as the localization and function of each coat.

ARF AND MEMBRANE TRAFFIC: BACKGROUND

A long-standing question in the membrane traffic field is how Arf proteins regulate vesicle formation. Coat recruitment is clearly downstream of Arf activation. The cycle of coat recruitment and vesicle formation is intimately tied to the GTP cycle of Arf. Upon activation, Arf•GTP recruits coat proteins to a membrane surface. Blocking nucleotide exchange causes rapid dissociation of Arf and coats from membranes. Conversely, preventing GTP hydrolysis stabilizes Arf and coats on membranes and/or vesicles. Blocking the GTP cycle of Arf with drugs (e.g., brefeldin A (BFA)), slowly hydrolyzable GTP analogs (e.g., GTPγS), Arf mutants, or an Arf peptide affects many transport pathways: ER-Golgi transport, intra-Golgi transport, nuclear envelope assembly, endosome fusion and endocytosis, and synaptic vesicle fusion (Balch et al., 1992; Boman et al., 1992; Faundez et al., 1997; Lenhard et al., 1992; Taylor et al., 1992). These experiments showed that Arf activity is critical at many steps of the biosynthetic and 241

2 242

BOMAN AND NILSSON

endocytic pathways. Interestingly, a distinct coat protein or protein complex is associated with each distinct transport pathway. A major question is whether Arf recruits coats directly or indirectly, for example via lipid modification. The direct interaction hypothesis has gained support in recent years following reports of direct interaction between Arf and many of these coats (Austin et al., 2002; Austin et al., 2000; Boman et al., 2000; Dell'Angelica et al., 2000; Zhao et al., 1999). The major Arf-dependent coat complexes are described here.

2.

ARF-DEPENDENT COAT COMPLEXES

There are currently three different classes of Arf-dependent adaptors that act in the biogenesis of vesicles and sorting of their cargo: the tetrameric adaptins (APs), heptameric coatomers (COPs), and monomeric GGAs (Golgi-localized, γ-adaptin ear homology domain, Arf-binding). Each class has more than one member in mammals with at least four APs, two COP complexes, and three GGAs.

2.1

Adaptor complexes and clathrin

Clathrin-coated vesicles were first discovered by electron microscopy in the mid 1960s. An electron-dense, regular lattice was observed to coat buds and vesicles formed from the plasma membrane (Roth and Porter, 1964) and the trans-Golgi network (Friend and Farquhar, 1967). Assembly proteins (APs) or adaptins were found to recruit clathrin onto sites of vesicle budding and facilitate clathrin assembly into a triskelion. To date, four adaptin complexes have been described in mammals (AP-1 through AP-4) and three are present in S. cerevisiae (AP-1 to AP-3). These adaptin complexes are each composed of 4 subunits with conserved structures: 2 large subunits (100 kDa; γ and β1 in AP-1, α and β2 in AP-2, δ and β3 in AP-3, ε and β4 in AP-4), one medium (50 kDa; µ1-µ4), and one small (20 kDa; σ1-σ4). The heterotetramers assemble into body, hinge, and ear domains. AP-1, AP2, and AP-4 bind clathrin; it is yet unclear if AP-3 binds clathrin (Simpson et al., 1996; Dell'Angelica et al., 1998). The AP complexes each bind the cytoplasmic tails of transmembrane cargo, but function in distinct pathways, as described below. Adaptin complexes AP-1, AP-3, and AP-4 were found to be downstream of Arf activation due to their BFA sensitivity and GTPγSstimulated recruitment (Boehm et al., 2001; Ooi et al., 1998; Stamnes and Rothman, 1993; Traub et al., 1993). Recent studies show cross-linking of AP-1 and AP-3 to Arf (see below).

COAT PROTEINS

2433

AP-2 appears to bind membranes independently of Arf. First, membrane association of AP-2 is insensitive to BFA (Robinson and Kreis, 1992). Second, expression of Arf6 mutants, which dramatically perturbs the plasma membrane, has no effect on AP-2 localization (Radhakrishna et al., 1996). Third, AP-2 was not cross-linked to Arf under conditions where AP-1 and AP-3 were cross-linked (Austin et al., 2002). However, AP-2 recruitment to plasma membranes is dependent upon PLD activation (West et al., 1997), suggesting that an Arf, most likely Arf6, may play a role in the regulation of the budding of AP-2 coated vesicles. Such a role would be indirect and distinct from that played by Arfs in recruitment of other coat proteins.

2.2

Coatomer complexes

Like AP/clathrin-coated vesicles, COPI coated vesicles can be visualized as electron dense coats by electron microscopy (Orci et al., 1986). Assays designed to reconstitute ER-Golgi and intra-Golgi transport each required GTP and were blocked by GTPγS (Beckers and Balch, 1989). Coated vesicles accumulated in the presence of GTPγS (Orci et al., 1989), allowing their visualization by electron microscopy and later purification. These Golgi-derived, non-clathrin-coated vesicles (Waters et al., 1991) were highly enriched for seven proteins, defined as coatomer or COPI, that was later also found as a stable, cytosolic complex. Arf was also abundant on these vesicles when purified from cell-free, intra-Golgi transport assays containing GTPγS (Serafini et al., 1991). The first subunit to be cloned (Duden et al., 1991), β-COP, has been used extensively as a marker for COPI, due to the availability of high quality antisera. Binding of β-COP to membranes was sensitive to BFA; within 1 minute after BFA addition, β−COP dissociated from the Golgi to the cytosol (Donaldson et al., 1991). This was similar to the time course of Arf dissociation from membranes, suggesting that β−COP was downstream of Arf activation. Though genetic studies in yeast (Stearns et al., 1990) and biochemical studies involving reconstitution of vesicle budding revealed functional ties between Arf and subunits of COPI, it wasn’t until 1999 that direct interactions between Arf and COPI were demonstrated, using cross linking techniques (see below). A related coat complex, named COPII, functions in ER exit (Barlowe et al., 1994). COPII comprises five subunits: Sec23 and Sec24 (which form a stable dimer), Sec13 and Sec31 (which form a stable dimer), and Sar1, a GTPase related to Arf (Barlowe et al., 1993). Membrane recruitment of the COPII subunits requires Sar1 rather than Arf.

4 244

2.3

BOMAN AND NILSSON

GGAs

The GGAs are a recently identified family of Arf effectors that are monomeric, clathrin-binding adaptors (for review, see Boman, 2001). Three GGAs are present in mammalian cells: GGA1, GGA2, and GGA3. Two are present in S. cerevisiae (Gga1p and Gga2p), and D. melanogaster, C. elegans, and S. pombe genomes each have at least one GGA gene. GGA proteins have four domains (Figure 1) that possess many or all the functions N

C VHS

GAT

hinge

ear

Figure 1. Schematic of GGA domain structure.

of AP complexes. These domains were named based on homology with other proteins. The VHS domain binds to cargo proteins, the GAT (GGA and Tom1) domain binds Arf, the hinge domain binds clathrin, and the ear domain binds accessory proteins. The monomeric nature of GGA proteins has allowed rapid testing of the model that Arf directly regulates vesicle coating. GGAs were originally identified through their direct interaction with Arf proteins (Boman et al., 2000). This interaction has been studied both in vitro and in vivo, and all data indicate that GGAs are direct effectors of Arf. All three human GGA proteins and both yeast Gga proteins interact with activated Arf in two-hybrid assays. Purified GST-GGA proteins (mammalian and yeast) interact specifically in vitro with Arf•GTP, but not Arf•GDP, which confirms the two-hybrid results (Boman et al., 2000; Zhdankina et al., 2001). These interactions do not require membranes or cargo. However, binding to Arf may be enhanced by cargo or other protein interactions; each domain of yeast Ggas contributes to Golgi localization and function (Boman et al., 2002). Like most Arf effectors, purified GST-GGA1 stabilizes and increases the amount of the GTP-bound form of Arf in in vitro nucleotide binding assays (Zhu et al., 2000). The rate of GTP-binding to Arf is not affected; hence GGA proteins are not exchange factors. GGAs have no GTPase-activating protein (GAP) activity (Boman et al., 2000; Puertollano et al., 2001); nor do they enhance the GAP activity of Arf GAP1 (Puertollano et al., 2001). Rather, GGAs can compete with Arf-GAPs for binding to Arf (Puertollano et al., 2001). In the presence of GGAs, therefore, GTP hydrolysis on Arf is slowed. Arf has no intrinsic GTPase activity. The dual effects of GGAs to

COAT PROTEINS

2455

stabilize Arf•GTP and to reduce GAP-dependent GTP hydrolysis will likely act synergistically in vivo. Several in vivo studies have shown that human GGAs are Arf effectors. First, stable expression of a dominant activating mutant of Arf, Arf1-L71, in NRK cells caused dramatic expansion of the Golgi lumen through an unknown effector (Zhang et al., 1994). Overexpression of GGA1 in these cells prevented the Arf1-L71-induced expansion of the Golgi apparatus (Boman et al., 2000), indicating a direct interaction between GGA1 and Arf in vivo and a function distinct from Golgi expansion. Second, an increased amount of Arf is associated with Golgi membranes in cells overexpressing GGA1, as visualized by immunofluorescence (Zhu et al., 2000). Similarly, overexpression of GGA3 slows the BFA-induced dissociation of Arf1 from Golgi membranes (Puertollano et al., 2001). These results show that the membrane-associated, GTP-bound form of Arf is stabilized in vivo, as expected from the in vitro experiments described above. Finally, a number of experiments reveal that the localization of GGA proteins at the Golgi is due to interaction with Arf. First, treatment with BFA causes rapid translocation of GGA proteins to the cytosol in a time frame indistinguishable from that of Arf itself (Boman et al., 2000; Dell'Angelica et al., 2000; Hirst et al., 2000; Poussu et al., 2000). Second, the interaction between the GAT domain of GGA proteins and Arf is strong enough to drive a reporter construct (GFP) onto the Golgi, even in the absence of the VHS, hinge and ear domains (Dell'Angelica et al., 2000). Third, and most convincingly, point mutations within the Arf-binding domain that cause loss of Arf interaction also cause loss of Golgi localization (Boman et al., 2002; Puertollano et al., 2001; Takatsu et al., 2002). Together, these data suggest that Arf•GTP recruits GGAs from the cytosol onto late Golgi membranes by interacting with the GAT domain. Arfs localize to many organelles and can selectively recruit one or more coat complexes onto a target membrane. Yet each Arf-dependent coat has a distinct localization and some organelles are the targets of multiple coats (Figure 2). It is not known how this selectivity is achieved. In two-hybrid and in vitro binding experiments, each of the human Arfs bind equally well to GGA proteins (Boman et al., 2000). Similarly, each Arf can recruit COPI and crosslink to AP-1 in vitro (Austin et al., 2002). These data suggest that Arf isoform specificity does not account for the specific localization of coats. Perhaps other proteins, such as cargo, stabilize the interaction between Arf and coat proteins at the target membrane, whereas interaction at other locations might be transient. A role for Arf GAP(s) in timing and proofreading of the Arf-coat-cargo complex has been proposed and is discussed in more detail in Chapters 7-9. This section summarizes the

6 246

BOMAN AND NILSSON

current understanding of determinants of localization and traffic of each coat, as diagrammed in Figure 2.

Figure 2. Sites of action of each coat protein. COPs function in the early secretory pathway, mediating transport between the ER, cis-Golgi network (CGN), and Golgi. APs and GGAs function in later pathways, primarily between TGN, endosomes, and plasma membrane. However, the specific pathways mediated by each coat are still being debated.

3.

LOCALIZATION AND TRANSPORT PATHWAYS OF EACH COAT

3.1

Adaptin complexes

Each of the adaptin complexes aids in the recruitment of clathrin to nascent vesicles, though binding of clathrin to AP-3 is still contentious (see Chapter

COAT PROTEINS

2477

13). Each adaptin functions at a distinct location and in distinct pathways. AP-2, which appears to be either Arf-independent or indirectly subject to Arf regulation, is the most abundant in many cell types and functions at the plasma membrane (for review see Hirst and Robinson, 1998). AP-1 is localized predominantly to the trans-Golgi network (TGN), endosomes, and small vesicles (Seaman et al., 1996). It is still unclear precisely which step (or steps) are mediated by AP-1. AP-1 binds to the cytoplasmic tail of both the cation-dependent and cation-independent mannose 6-phosphate receptors (M6PR). Two leucine residues in the M6PR tail and other cargo proteins define a dileucine motif and are essential for AP binding. In yeast, deletion of AP-1 has no known phenotype on its own, but has synthetic defects when combined with a clathrin temperature sensitive mutant or with GGA deletions (Costaguta et al., 2001; Stepp et al., 1995; Yeung et al., 1999). These results, and many others from studies using mammalian cells, suggest that AP-1 functions from TGN to endosomes (Hirst and Robinson, 1998). However, mice lacking the medium subunit of AP-1 accumulate M6PR at endosomes, suggesting that AP-1 functions in retrograde transport from endosomes to TGN (Meyer et al., 2000). AP-3 is also localized predominantly to the TGN, but the AP-3 pathway is distinct from that of AP1. In both yeast and mammalian cells, AP-3 functions in a direct route from the TGN to the lysosome (Cowles et al., 1997; Dell'Angelica et al., 1997; Simpson et al., 1997; Stepp et al., 1997). In neurons, AP-3 also forms synaptic vesicles from endosomes (Blumstein et al., 2001). AP-4 localizes to the endosomal/lysosomal system, but neither the function nor the pathway is known (Dell'Angelica et al., 1999; Hirst et al., 1999).

3.2

COPI and COPII

COPI localizes to the Golgi apparatus, and can be seen distributing mainly to the cis side, including the cis-Golgi network (CGN) (Oprins et al., 1993). Though COPI has been extensively studied, the role of this coat complex is still being debated. The debate can be divided into two categories: one debating the very existence of COPI (and COPII) as a coat for vesicles, the other what type of cargo goes into COPI vesicles. Though each issue is far from settled, there is good biochemical, morphological, and functional evidence in support of COPI vesicles. However, COPI is also implicated in other processes such as tubulation of membranes (e.g., such as that seen upon removal of the coat) and regulation (with Arf) of lipid modifying enzymes, and the sorting process itself. The issue of cargo sorting is also far from settled. For a long time, it was assumed that anterograde cargo was the main cargo of COPI vesicles. There is now growing evidence that COPI vesicles recycle Golgi resident proteins. Whether recycling is the sole

8 248

BOMAN AND NILSSON

function of COPI vesicles remains to be seen (for a recent review, see Storrie and Nilsson, 2002). COPII localizes to the endoplasmic reticulum (ER) and functions in ER exit (Barlowe et al., 1994). As with COPI, COPII components have been implicated in sorting through direct association to cytoplasmic domains of both newly synthesized proteins as well as to resident proteins that shuttle between the ER and the Golgi apparatus (e.g., p24 proteins). One of the COPII components, Sec13, has also been localized close to the cis side of the Golgi apparatus in mammalian cells (Tang et al., 1997), raising the possibility of a functional link between COPII and COPI. There is less controversy regarding the role of COPII vesicles though important questions remain. For example, large supra-molecular structures (e.g., collagen and some viruses) too large to fit in the typical 60nm COPII vesicle can still leave the ER. Does this occur with the help of COPII and if so, how?

3.3

GGA proteins

Several laboratories have examined the localization of GGA proteins in mammalian cells, using indirect immunofluorescence and immunoelectron microscopy. Endogenous GGA1, GGA2 and GGA3 localize predominantly to the trans-Golgi region in NRK, HeLa, Cos7 and human embryonic skin cells (Boman et al., 2000; Dell'Angelica et al., 2000; Hirst et al., 2000; Poussu et al., 2000; Takatsu et al., 2000). Mammalian GGAs isolated following cell lysis are soluble (Boman et al., 2000; Hirst et al., 2000), which suggests that the membrane-associated proteins rapidly exchange with a cytosolic pool of GGAs. The three GGAs show overlapping but subtle differences in staining patterns. In addition to the shared TGN staining, GGA1 shows a highly punctate pattern within the TGN and late Golgi region, GGA2 shows diffuse and cytosolic staining, and GGA3 stains larger puncta in the cytosol. To date, no clear functional distinctions have been found for the GGA isoforms. Hence, it is not clear if each mediates a distinct pathway, if each sorts different cargo into the same vesicles, or if they are functionally redundant. In yeast, Gga proteins also localize to the late Golgi (Boman et al., 2002), with apparent differences between Gga1p and Gga2p. GFP-tagged Gga1p shows almost exclusively punctate, Golgi staining, whereas Gga2p has a higher level of cytosolic staining. Yeast Ggas also fractionate predominantly with membranes rather than cytosol, even in the absence of Arf1p (Boman et al., 2002). These data suggest that membrane association in yeast is stabilized by non-Arf interactions to a greater extent than in mammalian cells. Interestingly, at least five of the coats described here localize to and function at the TGN of mammalian cells: GGA1, GGA2, GGA3, AP-1, and

COAT PROTEINS

2499

AP-3. Overexpression of GGA proteins can compete for AP-1 binding at the TGN, suggesting competition for Arf or other binding sites (Puertollano et al., 2001). In contrast, GGA overexpression does not alter COPI localization (Boman et al., 2000). It is not known how Arf can recruit each of these proteins to the TGN and facilitate proper sorting and vesicle formation. Regulation of coat binding must occur, but the type and mechanism of such regulation are not yet known.

4.

INTRAMOLECULAR INTERACTIONS BETWEEN ARFS AND COATS

Several experiments have been performed to identify the region of Arf that binds to coats and vice versa. The similarities and differences reveal interesting mechanistic insights into coat recruitment.

4.1

Binding site on Arf

The crystal structure of Arf contains two flexible switch regions, comparable to switch I and II of Ras (see Chapter 2). These regions have different conformations in the GDP-bound versus GTP-bound states, suggesting that they are effector binding regions. Using directed and random mutagenesis approaches, these switch regions were found to be important for coat binding. In a directed approach, three residues in switch I of Arf1 were mutated to allow photo-activated cross-linking to neighboring residues. In the presence of membranes and GTP, residues Ile46 and Ile49 were cross-linked to COPI (Zhao et al., 1999). Under the same conditions, residue Val43 was not specifically cross-linked (Zhao et al., 1999). Similarly, Ile46 was crosslinked to AP-1 and AP-3 (Austin et al., 2002; Austin et al., 2000). These data suggest that switch I is in close proximity to COPI, AP-1, and AP-3 at membrane surfaces. Notably, the Ile46 cross-linking experiments did not identify GGAs or AP-2 (Austin et al., 2002). In a random approach to define effector-binding sites on human Arf3, reverse two-hybrid screening showed that switch I and switch II are each important for binding to different effectors (Kuai et al., 2000). Several mutations were identified that eliminated binding to one effector yet retained binding to other effectors. Two mutations in switch I were found to eliminate binding to GGA1: I49T and F51Y (Kuai et al., 2000). These mutants were later shown to fail to interact with the GAT domain of GGA3, as expected (Puertollano et al., 2001). In contrast, several mutations in

10 250

BOMAN AND NILSSON

switch II had no effect on GGA binding. These data suggest that switch I is essential for binding to GGA proteins. In the absence of direct binding tests with COPI, these mutants were also tested for recruitment of COPI to Golgi membranes. Interestingly, the switch II mutations were much more impaired in COPI recruitment to Golgi membranes than were the switch I mutations (Kuai et al., 2000), suggesting that switch II is essential for COPI recruitment. This suggested that the binding sites for GGAs and COPI/APs are distinct, though perhaps overlapping. The binding site for Arf GAP includes switch II, as determined by x-ray crystallography. The activity of Arf GAP1 is enhanced by a factor of at least a thousand by COPI. However, this effect has only been observed in solution using truncated Arf GAP1 and Arf1 devoid of its N-terminal domain (Goldberg, 1999; Szafer et al., 2000; Szafer et al., 2001). Thus, the possibility of a stimulatory role of coatomer is open and additional experiments, preferably including some in vivo, are required. In contrast, binding and activity of Arf GAP1 is reduced by GGAs (Puertollano et al., 2001). These data suggest that COPI binds predominantly to switch I and GGAs bind to switch II. But such a conclusion is inconsistent with the mutational analyses. The potential differences between the cross-linking, GAP, and mutant analyses have not yet been resolved, and will likely require structure determinations of the protein complexes and additional data.

4.2

Binding site on coats

The binding site for Arf on each coat has been mapped to different extents, with the most detailed analysis performed on GGA proteins. There is no apparent homology between coats in the Arf-binding regions. Perhaps the structures are conserved, but the linear sequence is not. The binding site on GGA proteins was mapped by truncation analysis to a highly conserved domain corresponding to residues 170-280 in GGA1 (Boman et al., 2000; Zhdankina et al., 2001). This region lies in a domain that has been named the Arf-Binding Domain (ABD), GGA and TOM1 (GAT) domain, or GGA Homology (GGAH) domain (Boman et al., 2002; Dell'Angelica et al., 2000; Takatsu et al., 2002). There appears to be consensus to use the term GAT in the future. A cluster of residues within this domain (180-200 of GGA1) is extremely conserved between all yeast and human GGAs. Point mutations within this region of human and yeast GGA proteins eliminate binding to Arf (Boman et al., 2000; Puertollano et al., 2001; Takatsu et al., 2002). In human GGAs, these mutations abolish membrane association, showing that Arf recruits GGAs to the late Golgi. Interestingly, in yeast these mutations have minor effects on function and no effect on localization, suggesting that interactions with proteins other than

COAT PROTEINS

11 251

Arfs are more important for recruitment or stabilization at the membrane surface (Boman et al., 2002). Cross-linking experiments show that the binding site for Arf on COPI lies within the β and/or γ subunits (Zhao et al., 1999). Binding sites on AP-1 and AP-3 have been mapped to the ‘body’ (Austin et al., 2002) but have not been narrowed further.

4.3

Cargo binding and vesicle formation

The mechanisms of coat recruitment, cargo selection, vesicle formation, timing, and coat dissociation are still unclear, though a model for Arf’s involvement in cargo selection and vesicle formation is emerging. Some cargoes have been defined for each coat protein. Analyses of coordinated binding between Arf, coats, and cargoes show that Arf may play an important role in cargo selection. This was most clearly demonstrated in the Golgi with COPI. When GTP was used in vesicle formation assays, accumulated vesicles contained cargo proteins; either newly synthesized proteins (Ostermann et al., 1993) or resident proteins of the Golgi (Lanoix et al., 1999). Addition of GTPγS or activating Arf mutants caused accumulation of empty vesicles (Austin et al., 2002; Pepperkok et al., 2000). This suggests that the GTP cycle of Arf is critical for sorting and packaging of cargo into vesicles. As COPI can also interact directly with the cytoplasmic domains of resident proteins (Cosson and Letourneur, 1994) and perhaps stimulate Arf GAP activity, this may provide a level of regulation and cross talk between coats and the GTP cycle on Arf. It is not yet known if similar cargo selection mechanisms will occur at other sites of Arf action. Also, the cargo itself has been suggested to influence Arf GAP activity, effectively down-regulating its activity as more cargo accumulates (Goldberg, 2000; Lanoix et al., 2001; Springer et al., 1999). This couples coat recruitment and vesicle formation to cargo sorting ensuring that vesicles can only form in the presence of cargo. Though an attractive scenario, more work is needed to fully understand the details of both COPI and COPII coat proteins and how they recruit their cargo and form vesicles. The known cargoes for mammalian GGA proteins are transmembrane receptors that contain an acidic-cluster dileucine motif in their cytoplasmic tails. To date, these include M6PR, sortilin, sor-LA, LRP3, and yeast Vps10p. Other proteins contain this motif and will likely join the ranks of GGA-sorted cargo. The VHS domain of GGA proteins binds these motifs, regulated by phosphorylation of the cargo tail (Doray et al., 2002; Kato et al., 2002) and phosphorylation of GGAs at an internal regulatory site in the hinge domain (Doray et al., 2002). The crystal structure of the VHS domain bound to the M6PR tail has been solved by two groups (Misra et al., 2002;

12 252

BOMAN AND NILSSON

Shiba et al., 2002). These models show that 12 residues in two α helices (α6 and α8) are involved in the interaction (Shiba et al., 2002). Only three of these residues are conserved in yeast Gga1 and Gga2, none of which make contact with the dileucine motif. Consistent with this, human GGAs bind to yeast Vps10p tail (Dennes et al., 2002), but yeast Ggas do not (A. Boman, unpublished observation). These data suggest that yeast Gga proteins bind cargo with a different motif than reported for human GGA proteins. However, the VHS domain of yeast Gga proteins is required for localization and function, strongly suggesting that this domain binds to cargo or other important regulatory proteins (Boman et al., 2002). The VHS and GAT domains of yeast Ggas work synergistically for Golgi localization, suggesting that Arf activation will lead to efficient cargo packaging by Gga proteins. We speculate that Arf binding to yeast or human GGAs may allow kinases or phosphatases to act on GGA proteins, thus regulating cargo binding. It is still unclear how long Arf and coats remain associated with vesicles in vivo. Based on early experiments with GTPγS, it was proposed that hydrolysis of GTP on Arf triggered uncoating, just prior to vesicle fusion. More recent data show that vesicles formed with GTP do not retain their coats, suggesting that GTP hydrolysis occurs during or shortly after coat recruitment (Lanoix et al., 1999). Unlike COP and AP complexes, GGA proteins are not stabilized on membranes by GTPγS (A. Boman, unpublished observations), nor are they cross-linked to Arf (Austin et al., 2002). These data suggest that distinct mechanisms are involved in vesicle formation by each of the different coat proteins.

5.

FUTURE QUESTIONS

Many intriguing questions remain, including: What specifies coat localization? What regulates coat binding and cargo recruitment? What are the recognition sites on coats, and how does each bind to Arf? Do Arf GEFs and Arf GAPs have functions in membrane traffic distinct from those of activating or inhibiting Arf activity? The models for Arf and coat function presented in this chapter will provide a basis for addressing each of these questions in the future.

COAT PROTEINS

13 253

REFERENCES Austin, C., Hinners, I., and Tooze, S. A. (2000). Direct and GTP-dependent interaction of ADP-ribosylation factor 1 with clathrin adaptor protein AP-1 on immature secretory granules. J. Biol. Chem., 275, 21862-21869. Austin, C., Boehm, M., and Tooze, S. A. (2002). Site-specific cross-linking reveals a differential direct interaction of class 1, 2, and 3 ADP-ribosylation factors with adaptor protein complexes 1 and 3. Biochemistry, 41, 4669-4677. Balch, W. E., Kahn, R. A., and Schwaninger, R. (1992). ADP-ribosylation factor is required for vesicular trafficking between the endoplasmic reticulum and the cis-Golgi compartment. J. Biol. Chem., 267, 13053-13061. Barlowe, C., d'Enfert, C., and Schekman, R. (1993). Purification and characterization of SAR1p, a small GTP-binding protein required for transport vesicle formation from the endoplasmic reticulum. J. Biol. Chem., 268, 873-879. Barlowe, C., Orci, L., Yeung, T., Hosobuchi, M., Hamamoto, S., Salama, N., Rexach, M. F., Ravazzola, M., Amherdt, M., and Schekman, R. (1994). COPII: a membrane coat formed by Sec proteins that drive vesicle budding from the endoplasmic reticulum. Cell, 77, 895907. Beckers, C. J., and Balch, W. E. (1989). Calcium GTP: essential components in vesicular trafficking between the endoplasmic reticulum and Golgi apparatus. J. Cell Biol., 108, 1245-1256. Blumstein, J., Faundez, V., Nakatsu, F., Saito, T., Ohno, H., and Kelly, R. B. (2001). The neuronal form of adaptor protein-3 is required for synaptic vesicle formation from endosomes. J. Neurosci., 21, 8034-8042. Boehm, M., Aguilar, R. C., and Bonifacino, J. S. (2001). Functional and physical interactions of the adaptor protein complex AP- 4 with ADP-ribosylation factors (Arfs). EMBO J., 20, 6265-6276. Boman, A. L., Taylor, T. C., Melancon, P., and Wilson, K. L. (1992). A role for ADPribosylation factor in nuclear vesicle dynamics. Nature, 358, 512-514. Boman, A. L., Zhang, C., Zhu, X., and Kahn, R. A. (2000). A family of ADP-ribosylation factor effectors that can alter membrane transport through the trans-Golgi. Mol. Biol. Cell, 11, 1241-1255. Boman, A. L. (2001). GGA proteins: new players in the sorting game. J. Cell Science, 114, 3413-3418. Boman, A. L., Salo, P. D., Hauglund, M. J., Strand, N. L., Rensink, S. J., and Zhdankina, O. (2002). Arf interaction is not sufficient for yeast Gga function or localization. Mol. Biol. Cell, 13, 3078-3095. Cosson, P., and Letourneur, F. (1994). Coatomer interaction with di-lysine endoplasmic reticulum retention motifs. Science, 263, 1629-1631. Cosson, P., and Letourneur, F. (1997). Coatomer (COPI)-coated vesicles: role in intracellular transport and protein sorting. Curr. Opin. Cell Biol., 9, 484-487. Costaguta, G., Stefan, C. J., Bensen, E. S., Emr, S. D., and Payne, G. S. (2001). Yeast gga coat proteins function with clathrin in golgi to endosome transport. Mol. Biol. Cell, 12, 1885-1896. Cowles, C. R., Odorizzi, G., Payne, G. S., and Emr, S. D. (1997). The AP-3 adaptor complex is essential for cargo-selective transport to the yeast vacuole. Cell, 91, 109-118. Dell'Angelica, E. C., Ohno, H., Ooi, C. E., Rabinovich, E., Roche, K. W., and Bonifacino, J. S. (1997). AP-3: an adaptor-like protein complex with ubiquitous expression. EMBO J., 16, 917-928.

14 254

BOMAN AND NILSSON

Dell'Angelica, E. C., Klumperman, J., Stoorvogel, W., and Bonifacino, J. S. (1998). Association of the AP-3 adaptor complex with clathrin. Science, 280, 431-434. Dell'Angelica, E. C., Mullins, C., and Bonifacino, J. S. (1999). AP-4, a novel protein complex related to clathrin adaptors. J. Biol. Chem., 274, 7278-7285. Dell'Angelica, E. C., Puertollano, R., Mullins, C., Aguilar, R. C., Vargas, J. D., Hartnell, L. M., and Bonifacino, J. S. (2000). GGAs: A family of ADP ribosylation factor-binding proteins related to adaptors and associated with the Golgi complex. J. Cell Biol., 149, 8194. Dennes, A., Madsen, P., Nielsen, M. S., Petersen, C. M., and Pohlmann, R. (2002). The yeast Vps10p cytoplasmic tail mediates lysosomal sorting in mammalian cells and interacts with human GGAs. J. Biol. Chem., 277, 12288-12293. Donaldson, J. G., Kahn, R. A., Lippincott-Schwartz, J., and Klausner, R. D. (1991). Binding of Arf and beta-COP to Golgi membranes: possible regulation by a trimeric G protein. Science, 254, 1197-1199. Doray, B., Bruns, K., Ghosh, P., and Kornfeld, S. A. (2002). Autoinhibition of the ligandbinding site of GGA1/3 VHS domains by an internal acidic cluster-dileucine motif. Proc. Natl. Acad. Sci,. USA, 99, 8072-8077. Faundez, V., Horng, J. T., and Kelly, R. B. (1997). ADP ribosylation factor 1 is required for synaptic vesicle budding in PC12 cells. J. Cell Biol., 138, 505-515. Friend, D. S., and Farquhar, M. G. (1967). Functions of coated vesicles during protein absorption in the rat vas deferens. J. Cell Biol., 35, 357-376. Goldberg, J. (1999). Structural and functional analysis of the Arf1-ArfGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Goldberg, J. (2000). Decoding of sorting signals by coatomer through a GTPase switch in the COPI coat complex. Cell, 100, 671-679. Hirst, J., and Robinson, M. S. (1998). Clathrin and adaptors. Biochim. Biophys. Acta, 1404, 173-193. Hirst, J., Bright, N. A., Rous, B., and Robinson, M. S. (1999). Characterization of a fourth adaptor-related protein complex. Mol. Biol. Cell, 10, 2787-2802. Hirst, J., Lui, W. W., Bright, N. A., Totty, N., Seaman, M. N., and Robinson, M. S. (2000). A family of proteins with gamma-adaptin and VHS domains that facilitate trafficking between the trans-Golgi network and the vacuole/lysosome. J. Cell Biol., 149, 67-80. Kato, Y., Misra, S., Puertollano, R., Hurley, J. H., and Bonifacino, J. S. (2002). Phosphoregulation of sorting signal VHS domain interactions by a direct electrostatic mechanism. Nat. Struct. Biol., 9, 532-536. Kuai, J., Boman, A. L., Arnold, R. S., Zhu, X., and Kahn, R. A. (2000). Effects of activated ADP-ribosylation factors on Golgi morphology require neither activation of phospholipase D1 nor recruitment of coatomer. J. Biol. Chem., 275, 4022-4032. Lanoix, J., Ouwendijk, J., Lin, C. C., Stark A., Love, H. D., Ostermann, J., and Nilsson, T. (1999). GTP hydrolysis by arf-1 mediates sorting and concentration of Golgi resident enzymes into functional COP I vesicles. EMBO J., 18, 4935-4948. Lenhard, J. M., Kahn, R. A., and Stahl, P. D. (1992). Evidence for ADP-ribosylation factor (Arf) as a regulator of in vitro endosome-endosome fusion. J. Biol. Chem., 267, 1304713052. Meyer, C., Zizioli, D., Lausmann, S., Eskelinen, E. L., Hamann, J., Saftig, P., von Figura, K., and Schu, P. (2000). mu1A-adaptin-deficient mice: lethality, loss of AP-1 binding and rerouting of mannose 6-phosphate receptors. EMBO J., 19, 2193-2203.

COAT PROTEINS

15 255

Misra, S., Puertollano, R., Kato, Y., Bonifacino, J. S., and Hurley, J. H. (2002). Structural basis for acidic-cluster-dileucine sorting-signal recognition by VHS domains. Nature, 415, 933-937. Ooi, C. E., Dell'Angelica, E. C., and Bonifacino, J. S. (1998). ADP-Ribosylation factor 1 (Arf1) regulates recruitment of the AP-3 adaptor complex to membranes. J. Cell Biol., 142, 391-402. Oprins, A., Duden, R., Kreis, T. E., Geuze, H. J., and Slot, J. W. (1993). Beta-COP localizes mainly to the cis-Golgi side in exocrine pancreas. J. Cell Biol., 121, 45-59. Orci, L., Glick, B. S., and Rothman, J. E. (1986). A new type of coated vesicular carrier that appears not to contain clathrin: its possible role in protein transport within the Golgi stack. Cell, 46, 171-184. Orci, L., Malhotra, V., Amherdt, M., Serafini, T., and Rothman, J. E. (1989). Dissection of a single round of vesicular transport: sequential intermediates for intercisternal movement in the Golgi stack. Cell, 56, 357-368. Osterman, J., Orci, L., Tani, K., Amherdt, M., Ravazzola, M., Elazar, Z., and Rothman, J. E. (1993). Stepwise assembly of functionally active transport vesicles. Cell, 75, 1015-1025. Pepperkok, R., Whitney, J. A., Gomez, M., and Kreis, T. E. (2000). COPI vesicles accumulating in the presence of a GTP restricted arf1 mutant are depleted of anterograde and retrograde cargo. J. Cell Sci., 113, 135-144. Poussu, A., Lohi, O., and Lehto, V. P. (2000). Vear, a novel Golgi-associated protein with VHS and gamma-adaptin "Ear" domains. J. Biol. Chem., 275, 7176-7183. Puertollano, R., Randazzo, P. A., Presley, J. F., Hartnell, L. M., and Bonifacino, J. S. (2001). The ggas promote arf-dependent recruitment of clathrin to the tgn. Cell, 105, 93-102. Radhakrishna, H., Klausner, R. D., and Donaldson, J. G. (1996). Aluminum fluoride stimulates surface protrusions in cells overexpressing the Arf6 GTPase. J. Cell Biol., 134, 935-947. Robinson, M. S., and Kreis, T. E. (1992). Recruitment of coat proteins onto Golgi membranes in intact and permeabilized cells: effects of brefeldin A and G protein activators. Cell, 69, 129-138. Roth, T. F., and Porter, K. R. (1964). Yolk protein uptake in the oocyte of the mosquito Aedes aegypti. J. Cell Biol., 20, 313-332. Seaman, M. N., Sowerby, P. J., and Robinson, M. S. (1996). Cytosolic and membraneassociated proteins involved in the recruitment of AP-1 adaptors onto the trans-Golgi network. J. Biol. Chem., 271, 25446-25451. Serafini, T., Orci, L., Amherdt, M., Brunner, M., Kahn, R. A., and Rothman, J. E. (1991). ADP-ribosylation factor is a subunit of the coat of Golgi-derived COP-coated vesicles: a novel role for a GTP-binding protein. Cell, 67, 239-253. Shiba, T., Takatsu, H., Nogi, T., Matsugaki, N., Kawasaki, M., Igarashi, N., Suzuki, M., Kato, R., Earnest, T., Nakayama, K., and Wakatsuki, S. (2002). Structural basis for recognition of acidic-cluster dileucine sequence by GGA1. Nature, 415, 937-941. Simpson, F., Bright, N. A., West, M. A., Newman, L. S., Darnell, R. B., and Robinson, M. S. (1996). A novel adaptor-related protein complex. J. Cell. Biol., 133, 749-760. Simpson, F., Peden, A. A., Christopoulou, L., and Robinson, M. S. (1997). Characterization of the adaptor-related protein complex, AP-3. J. Cell Biol., 137, 835-845. Springer, S., Spang, A., and Schekman, R. (1991). A primer on vesicle budding. Cell, 97, 145-148. Stamnes, M. A., and Rothman, J. E. (1993). The binding of AP-1 clathrin adaptor particles to Golgi membranes requires ADP-ribosylation factor, a small GTP-binding protein. Cell, 73, 999-1005.

16 256

BOMAN AND NILSSON

Stearns, T., Kahn, R. A., Botstein, D., and Hoyt, M. A. (1990). ADP ribosylation factor is an essential protein in Saccharomyces cerevisiae and is encoded by two genes. Mol. Cell. Biol., 10, 6690-6699. Stearns, T., Willingham, M. C., Botstein, D., and Kahn, R. A. (1990). ADP-ribosylation factor is functionally and physically associated with the Golgi complex. Proc. Natl. Acad. Sci., USA, 87, 1238-1242. Stepp, J. D., Pellicena-Palle, A., Hamilton, S., Kirchhausen, T., and Lemmon, S. K. (1995). A late Golgi sorting function for Saccharomyces cerevisiae Apm1p, but not for Apm2p, a second yeast clathrin AP medium chain-related protein. Mol. Biol. Cell, 6, 41-58. Stepp, J. D., Huang, K., and Lemmon, S. K. (1997). The yeast adaptor protein complex, AP3, is essential for the efficient delivery of alkaline phosphatase by the alternate pathway to the vacuole. J. Cell Biol., 139, 1761-1774. Storrie, B., and Nilsson, T. (2002). The Golgi apparatus: balancing new with old. Traffic, 3 521-529. Szafer, E., Pick, E., Rotman, M., Zuck, S., Huber, I., and Cassel, D. (2000). Role of coatomer and phospholipids in GTPase-activating protein-dependent hydrolysis of GTP by ADPribosylation factor-1. J. Biol. Chem., 275, 23615-23619. Szafer, E., Rotman, M., and Cassel, D. (2001). Regulation of GTP hydrolysis on ADPribosylation factor-1 at the Golgi membrane. J. Biol. Chem., 276, 47834-47839. Takatsu, H., Yoshino, K., and Nakayama, K. (2000). Adaptor gamma ear homology domain conserved in gamma-adaptin and GGA proteins that interact with gamma-synergin. Biochem. Biophys. Res. Comm., 271, 719-725. Takatsu, H., Yoshino, K., Toda, K., and Nakayama, K. (2002). GGAproteins associate with Golgi membranes through interaction between their GGAH domains and ADPribosylation factors. Biochem. J., 365, 369-78. Tang, B. L., Peter, F., Krijnse-Locker, J., Low, S. H., Griffiths, G., and Hong, W. (1997). The mammalian homolog of yeast Sec13p is enriched in the intermediate compartment and is essential for protein transport from the endoplasmic reticulum to the Golgi apparatus. Mol. Cell. Biol., 17, 256-266. Taylor, T. C., Kahn, R. A., and Melancon, P. (1992). Two distinct members of the ADPribosylation factor family of GTP-binding proteins regulate cell-free intra-Golgi transport. Cell, 70, 69-79. Traub, L. M., Ostrom, J. A., and Kornfeld, S. (1993). Biochemical dissection of AP-1 recruitment onto Golgi membranes. J. Cell Biol., 123, 561-573. Waters, M. G., Serafini, T., and Rothman, J. E. (1991). 'Coatomer': a cytosolic protein complex containing subunits of non-clathrin-coated Golgi transport vesicles. Nature, 349, 248-251. West, M. A., Bright, N. A., and Robinson, M. S. (1997). The role of ADP-ribosylation factor and phospholipase D in adaptor recruitment. J. Cell Biol., 138, 1239-1254. Yeung, B. G., Phan, H. L., and Payne, G. S. (1999). Adaptor complex-independent clathrin function in yeast. Mol. Biol. Cell, 10, 3643-3659. Zhang, C. J., Rosenwald, A. G., Willingham, M. C., Skuntz, S., Clark, J., and Kahn, R. A. (1994). Expression of a dominant allele of human Arf1 inhibits membrane traffic in vivo. J. Cell Biol., 124, 289-300. Zhao, L., Helms, J. B., Brunner, J., and Wieland, F. T. (1999). GTP-dependent binding of ADP-ribosylation factor to coatomer in close proximity to the binding site for dilysine retrieval motifs and p23. J. Biol. Chem., 274, 14198-14203. Zhdankina, O., Strand, N. L., Redmond, J. M., and Boman, A. L. (2001). Yeast GGA proteins interact with GTP-bound Arf and facilitate transport through the Golgi. Yeast, 18, 1-18.

COAT PROTEINS

17 257

Zhu, X., Boman, A. L., Kuai, J., Cieplak, W., and Kahn, R. A. (2000). Effectors increase the affinity of GTP to Arf and lead to increased binding. J. Biol. Chem., 275, 13465-13475.

Chapter 13 HETEROTETRAMERIC COAT PROTEIN-ARF INTERACTIONS M. L. STYERS AND V. FAUNDEZ Department of Cell Biology and Center for Neurodegenerative Diseases, Emory University School of Medicine, Atlanta, USA

Abstract:

1.

The mechanisms by which COPI and the heterotetrameric adaptor complexes interact with Arf are analyzed from a comparative evolutionary perspective. Phylogenetic, biochemical, and structural analyses support a common ancestry of the heterotetrameric adaptors and COPI. Consistent with a structural conservation between COPI and adaptors, the functional interactions between Arf and adaptors possess similarities with the way COPI and Arf associate. However, these interactions have become increasingly complex and appear to have reached a summit with the adaptor AP-2. A common theme across different carrier-forming machines is the requirement of interacting networks of functionally diverse Arf effectors to recruit adaptor complexes to membranes.

INTRODUCTION

The exocytic and endocytic routes are organized as a series of interdependent membrane-enclosed compartments, each one possessing its own morphological and functional identity (Palade, 1975; LippincottSchwartz et al., 2000). These compartments communicate via a continuous flow of membrane carriers of a tubulovesicular nature. This raises the question of how the specific protein and lipid compositions of compartments are maintained. A steady state composition is maintained in each organelle by controlling both the molecules that exit from it as well as the fusion of incoming membrane carriers (Lippincott-Schwartz et al., 2000; Wickner and Haas, 2000; Chen and Scheller, 2001). Tight control over the formation and fusion of these carriers defines the identity of an exocytic or endocytic organelle. 259

2 260

STYERS AND FAUNDEZ

To analyze the complex process of membrane carrier formation we will divide the process into several steps: 1) the formation of membrane microdomains, or ‘hot spots’, 2) cargo recognition and concentration in discrete regions of the donor membrane, or sorting, 3) imposition of membrane curvature and bud formation, 4) bud severing, or pinching off, and, 5) dissociation of the budding machinery from the vesicle, or uncoating (Springer et al., 1999). In this sequence of events, Arf GTPases and coat/adaptor molecules play a central role in the formation of membrane microdomains and sorting of membrane components into the nascent bud. The roles of GTPases and coats/adaptors in sorting and vesiculation have been explored to different extents, depending upon: 1) the coat and GTPase involved, 2) the particular pathway analyzed, and, 3) the cell model system. The complexity of these mechanisms is daunting as five different Arfs functionally interact with a diverse set of effectors, among them monomeric and multimeric coats/adaptors. An even greater complexity emerges when coat/adaptor isoforms and splicing variants are considered, as well as their corresponding interacting molecules. We will attempt to formulate unitary principles that govern multiple vesiculation processes that require Arf and a coat. We will approach this question from a comparative evolutionary perspective in order to define the mechanisms by which Arf and coat/adaptor regulate the formation of membrane carriers. Our hypothesis is that the primordial coat-Arf interaction defines a basic mechanism upon which selective pressures have acted to diversify the membrane carrier formation machinery. We propose that Arf and its effectors involved in membrane traffic can be fitted to a “molecular mutualism model”. We have borrowed and adapted the mutualism concept as used in ecology. Mutualism is defined here as a network of interactions of two or more elements that have a positive net impact on the fitness of those elements to perform their functions. Mutualistic interactions between organisms control growth and/or density of the network components. Moreover, there is no need for the elements of the network to physically interact or for the interactions to be permanent or obligatory. In our particular case, the lipids and proteins that take part in the formation of membrane carriers constitute the network. We will concentrate on the relationship between Arf GTPases and their coat and non-coat effectors. As we will discuss, and similar to the consequences of mutualism between organisms, the interactions between Arf and facultative effectors control the density or growth rate of membrane-associated components like coats. We will first describe the evolutionary relationships among the coatomer (COPI) and the adaptor complexes as a way to define conserved features

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

2613

with regard to the coats and their interaction with Arf that could be obligatory properties of the network.

2.

PHYLOGENETIC ANALYSIS OF COPI AND OTHER ADAPTOR COMPLEX STRUCTURES

If mechanisms common to all coats and adaptors have been maintained throughout evolution, it is reasonable to expect structural similarities between coats and adaptor subunits. The first indication of the common evolutionary origins of the COPI complex and other adaptors came from the sequence comparisons between subunits that form the COPI complex and the heterotetrameric adaptors AP-1, AP-2, AP-3 and AP-4 (Lewin and Mellman, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001) (See Figure 1).

Figure 1. Proposed evolutionary relationships of the heterotetrameric coat subunits. Asterisks denote rounds of gene duplication. Diagram has been adapted from Lowe and Kreis, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001.

Phylogenetically, COPI is the oldest coat complex (Lewin and Mellman, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001). COPI is organized as a heptamer made up of two biochemically and functionally distinguishable sub-components, the heterotetrameric F-subcomplex,

4 262

STYERS AND FAUNDEZ

containing β-COP, γ-COP, δ-COP and ζ-COP, and a trimer consisting of αCOP, β’-COP and ε-COP (Lowe and Kreis, 1998; Kirchhausen, 2000). The subunits of the F-subcomplex possess sequence similarity to the different adaptor subunits (Schledzewski et al., 1999; Eugster et al., 2000). Like the F-subcomplex, the tetrameric adaptors contain two large subunits β1-4 and γ, α, δ, ε; a medium chain (µ1-4) and a small chain (σ1-4) (Boehm and Bonifacino, 2001) (see Figure 2).

Figure 2. The pair of dimers model for the assembly of the heterotetrameric coat complexes. Diagram represents a putative evolutionary sequence for the formation of heterotetramers from a set of dimers (see Figure 1). The putative proto-dimer pair has been inferred from the hierarchy of subunit interactions obtained for COPI as well as the adaptor complexes (see text). Diagram of the canonical heterotetramer structure was derived from the crystal structure of the AP-2 core complex (Collins et al., 2002).

Heterotetrameric adaptors have likely evolved from the COPI Fsubcomplex through successive rounds of gene duplication (Lewin and Mellman, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001; Chow et al., 2001). However, whether this tetrameric organization emerged as an assembly of independently evolved monomers or as coevolved dimers is an open question. If a pair of dimers generated the primordial COPI Fsubcomplex it could be possible to observe preferential interaction between some subunits while others could serve as bridges between subunits that do not interact extensively. Two-hybrid analysis has revealed that the β-COP interaction with the medium chain δ-COP is privileged over all the other possible molecular links. Alternatively, γ-COP forms a stable complex with

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

2635

the small ζ-COP subunit (Eugster et al., 2000; Faulstich et al., 1996; Takatsu et al., 2001). Triple-hybrid analysis has further shown that ζ-COP acts as a bridge between the β and γ subunits because negligible interaction occurs between β and γ-COP in the absence of ζ-COP (Takatsu et al., 2001). This hierarchy of contacts detected by yeast hybrid analysis can also be biochemically defined, as it is possible to disassemble purified coatomer into stable β−δ and γ−ζ sub-complexes (Pavel et al., 1998). In particular, the isolated β−δ sub-complex is functional, as this isolated sub-complex can be recruited to membranes in an Arf1•GTP-dependent manner (Pavel et al., 1998). As we will see, the β subunit of the COPI F-subcomplex is involved in Arf1 recognition. These results suggest that the most primitive interaction between Arf1 and a coat occurred between the GTPase and a β−δ dimer. This molecular analysis is supportive of the phylogenetic model proposed by Schledzewski, et al. (Schledzewski et al., 1999), who suggested that the primordial COPI precursor emerged by the fusion of two proto-dimers (see Figure 2). Does adaptor subunit organization fit the proto-dimer model? The crystal structure of the AP-2 adaptor core complex (Collins et al., 2002), which phylogenetic studies indicate is the most recent heterotetramer addition to the tetrameric adaptor family (Lewin and Mellman, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001), has provided insight into this question. The core of AP-2 is organized as a rectangular structure with its outside flanked by curved α and β2 chains. Nestled in the elbow of α and β2 are the σ2 and the N-terminus of the µ2 subunits, respectively. Each pair, α−σ2 and β2-µ2, forms a tight heterodimer, which creates a flat platform where the C-terminal end of µ2 sits. Remarkably, the crystal structure of σ2 and the N-terminal end of µ2 are almost identical (Collins et al., 2002) (see Figure 2). Also, the large subunits α and β2 share substantial structural homology (Collins et al., 2002). These results are in agreement with the phylogenetic prediction that the large subunits as well as the σ and µ-type chain originated by gene duplication from common ancestor genes that encoded a primitive dimer (Lewin and Mellman, 1998; Schledzewski et al., 1999; Boehm and Bonifacino, 2001). Genetic and biochemical evidence regarding other tetrameric adaptors also support a “proto-dimer” model”. Two-hybrid analysis has revealed the existence of preferential dimer associations within AP-1 (Takatsu et al., 2001; Page and Robinson, 1995; Aguilar et al., 1997), AP-3 (Peden et al., 2002) and AP-4 (Takatsu et al., 2001; Hirst et al., 1999), in addition to AP-2 (Page and Robinson, 1995). Stable AP-3 dimers are also found in vivo. Naturally occurring AP-3 mutations in mice, either in δ or β3A subunits, lead to degradation of the non-mutated subunits to different extents (Kantheti et al., 1998; Feng et al., 1999; Dell’Angelica et al., 1999a; Zhen et

6 264

STYERS AND FAUNDEZ

al., 1999). A small amount of undegraded β3-µ3 dimers can be immunoprecipitated from δ-deficient cytosol despite the complete absence of σ3. Conversely, a δ−σ3 dimer can be immunoprecipitated from β3deficient cytosol in which all µ3 is degraded (Peden et al., 2002; Zhen et al., 1999). These results suggest a preferential association of undegraded subunits following a pair of dimers rule. Each subunit of these dimers cannot exist in the absence of the other, suggesting that the dimer is the minimal stable assembly of monomers. These results imply that the dimer organization is an obligatory mutualistic interaction. Phylogenetic, biochemical, and structural analyses support a common ancestry of the heterotetrameric adaptors. The pair of dimers feature of the most primitive coat, the COPI F-subcomplex, is conserved in one of the most recent evolutionary additions to the heterotetrameric adaptor family, AP-2. This core structural organization likely represents the minimal scaffold where other molecules, like the Arf GTPases, assist the adaptor complexes in their sorting, coating, and vesiculation functions.

3.

COMPARATIVE ANALYSIS OF THE MOLECULAR INTERACTIONS BETWEEN ARFS AND HETEROTETRAMERIC ADAPTORS.

With the exception of Arf6 (Cavenagh et al., 1996), all Arfs cycle between the cytosol and membranes, depending upon whether GDP or GTP is bound, respectively (Moss and Vaughan, 1998). Cycling of Arf between cytosol and membranes brings its effectors in close proximity to membranes. Among those effectors are COPI and other adaptors. These coat complexes cycle between membrane-bound and cytosolic states in a mechanism that is functionally linked to the Arf GTPase cycle. Three non-exclusive models have been proposed that differ in the nature of the Arf effector. These models include the hypotheses that: a) Arfs interact directly with COPI and other adaptor complexes (Zhao et al., 1997). b) Arfs interact with or modify a putative docking site, inducing a change that would allow the docking site to bind the coat molecule (Zhu et al., 1998). The docking site could be either a lipid (Arneson et al., 1999) and/or a protein (Zhu et al., 1998). c) Arfs interact with lipid modifying effectors that create membrane microdomains favorable for adaptor binding (Ktistakis et al., 1996). It is likely that in a single cell, different membranes organize networks of Arf effectors that build vesiculation machines that differ in the extent that

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

2657

these three mechanisms are used. Variations in the relative amounts of different effectors present on membranes, differences in lipid composition, the presence of different adaptors, and the use of different model systems might explain the emergence of these different models.

3.1

Arf interactions with coat molecules: the direct link

The first indication that Arf was involved in the recruitment of adaptors to membranes came from the use of the fungal metabolite brefeldin A (BFA), an inhibitor of some Arf GEFs (Klausner et al., 1992; Jackson and Casanova, 2000). Shortly after drug treatment COPI (Klausner et al., 1992; Lippincott-Schwartz et al., 1991; Orci et al., 1991); AP-1 (Seaman et al., 1993; Robinson and Kreis, 1992), AP-3 (Simpson et al., 1996; Dell’Angelica et al., 1997; Ooi, et al., 1998) and AP-4 (Hirst et al., 1999; Dell’Angelica, et al., 1999b) are redistributed to the cytosol. Similar effects can be triggered using Arf mutants that mimic the GDP-bound state of the GTPase (Hirst et al., 1999; Ooi et al., 1998). Demonstration of direct binding between coat molecules and Arf1 was first obtained in an in vitro reconstituted Golgi membrane system (Zhao et al., 1997; Zhao, et al., 1999). Purified COPI can be recruited to Golgi membranes by in vitro transcribed-translated-Arf1. Using tRNA loaded with photo-reactive amino acids it is possible to label selected positions on in vitro transcribed-translated Arf1. Loading the GTPase with a photo-reactive phenylalanine in position 82 leads to a crosslinked product with β-COP (Zhao et al., 1997). In contrast, Arf1 loaded with photo-reactive residues either at position 46 or 49 forms cross-linked products with each large coatomer subunit, β-COP and γ-COP (Zhao et al., 1999). The GTP•Arf1-COPI interaction has been confirmed for the β-COP subunit using two-hybrid analysis (Eugster et al., 2000). The exact regions in the β-COP and γ-COP subunits where Arf1 binds are presently not known. However, if we draw a parallel between the AP-2 core crystal structure and COPI, these results suggest that Arf1 is wedged in the interface of the two large COPI subunits in close proximity to the membrane (see Figure 3). In light of the structural conservation observed between COPI and the tetrameric adaptors, a likely prediction is that Arf1 labeled on positions 46, 49 or 82 should be cross-linked to the large subunits of the adaptor complexes known to be recruited by Arf1. Immature secretory granules purified from adrenomedullary cells recruit AP-1 and AP-3 in a GTP•Arfdependent fashion (Dittie et al., 1996). In this in vitro adaptor recruitment system, photoreactive Arf1•GTP or Arf5•GTP, labeled at position 46, can be cross-linked to the AP-1 complex (Austin et al., 2000; Austin et al., 2002). Immunoprecipitation with β1 and γ antibodies has confirmed that the

8 266

STYERS AND FAUNDEZ

large subunits of the AP-1 complex interact with Arf1. Moreover, removing the ear domains of both large subunits by controlled proteolysis does not abrogate the formation of β1/γ-Arf1 cross-linked products (Austin et al., 2002). These results suggest that the membrane-facing region of AP-1, composed of the amino-terminal part of β1 and γ, interacts with Arf1. Similar results have been obtained with AP-3, although the identity of the high molecular weight cross-linked products could not be defined by immunoprecipitation with β3 or δ antibodies. However, native immunoprecipitation with antibodies against the small σ3 subunit brings down Arf1 cross-linked to polypeptides that correspond to the molecular weight of the large AP-3 subunits (Austin et al., 2002).

Figure 3. Proposed molecular interactions of the ARF GTPases with heterotetrameric coats. See text for details.

Like other adaptor complexes, AP-4 is recruited to membranes by a mechanism linked to the GDP-GTP cycle of Arfs. AP-4 remains cytosolic when cells are either briefly exposed to brefeldin A or transfected with the mutants of Arf1 or Arf3 that bind guanine nucleotides poorly. Similar effects are obtained by introducing Arf GAP or the GGA1 VHS-GAT domain into cells (Hirst et al., 1999; Dell'Angelica, 1999b). The interaction of Arf1•GTP and AP-4 has been mapped, using two-hybrid analysis, to the first 260 amino acids of ε-adaptin, a putative membrane–facing region of the large subunit, and the switch I region of Arf1 (Boehm et al., 2001). These results are in agreement with the hypothesis that the structure and localization of the Arf-binding region in the COPI F-subcomplex and tetrameric adaptors is evolutionarily conserved. However, no interactions were detected with the other large AP-4 subunit, β4-adaptin (Boehm et al., 2001). This might reflect an intrinsic limitation of the two-hybrid technique,

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

2679

which cannot detect interactions dependent on binding to membranes and assembly of a heterotetramer. Is a simple model of an Arf molecule wedged between the two large subunits of a multimeric adaptor enough (see Figure 3)? It is possible that this model represents the primordial mechanism to which all Arf-coat interactions conform, though perhaps to different extents. If these interactions have remained constrained to the large coat subunits with the conserved switch I and II regions of Arf; we can predict that the interactions between adaptors and Arfs should be indistinguishable among Arf isotypes known to cycle between cytosol and membranes (Arf1-5). Conversely, modifications of the putative core mechanism by demands imposed by new coat-dependent traffic pathways might be reflected through differential interactions between adaptors and Arf isotypes. Detailed analysis of the AP3 and AP-4 interactions with Arfs supports the latter. Expression of an Arf1 mutant that mimics the inactive state induces detachment of coats from membranes, including AP-3 (Ooi et al., 1998). However, the expression in cells of the homologous mutant of Arf5 does not release AP-3 from membranes (Ooi et al., 1998). Notably, in vitro photoreactive Arf5 does not generate detectable cross-linked species with AP-3; yet photoreactive Arf1 can be cross-linked to AP-3 (Austin et al., 2002). An even more extensively studied example is AP-4. Expression of the dominant negative mutants of Arf1 and Arf3 leads to a cytosolic redistribution of AP-4 (Aguilar et al., 2001). However, neither a similar mutation in Arf4 nor Arf5 modifies the membrane distribution of AP-4 (Aguilar et al., 2001). Overall, these results indicate that selectivity determinants further control the mechanism by which Arfs and adaptor effectors interact. Switch I and II regions of Arfs are conserved, suggesting that other contacts between Arf and adaptors might define specificity. In support of this idea, the C-terminal region of µ4 is capable of interacting with Arf1, in GTP or GDP-bound forms, through regions other than switches I and II of Arf (Aguilar et al., 2001). Similar analysis performed with the COPI subunits has not led to a detectable interaction of Arf1 with the µ-adaptin homolog, δ-COP (Eugster et al., 2000). In summary, these results suggest that although the primary mechanism of adaptor-Arf associations is highly conserved and involves the large subunits, the medium and small subunits may have diverged to add further specificity to the adaptor-Arf interactions.

10 268

STYERS AND FAUNDEZ

3.2

Arf interactions with coat molecules: the indirect link

3.2.1

Is a direct Arf-coat interaction sufficient to control coat membrane recruitment?

A major indication that other factors are required to define the binding of an adaptor to membranes has come from the reconstitution of Arf-dependent coat recruitment onto protein-free synthetic membranes. AP-1 and clathrin can be recruited to liposomes with a similar efficiency to that achieved on natural membranes, such as Golgi-enriched fractions (Zhu et al., 2001a; Zhu et al., 1999a). However, AP-1 recruitment to Golgi membranes requires cytosolic factors other than Arf1 fractions (Zhu et al., 2001a; Zhu et al., 1999a; Seaman et al., 1996). Moreover, the AP-1-liposome interaction is significantly influenced by the liposome lipid composition. Acidic phospholipids, such as phosphatidic acid or phosphatidylinositides (PIs), notably increase the adaptor recruitment to liposomes (Zhu et al., 1999a). Phosphatidic acid is generated from phosphatidylcholine by phospholipase D (PLD). This enzyme is an Arf effector (Brown et al., 1993; Cockcroft et al., 1994) that is found in association with the Golgi complex, to which it is recruited by a GTP•Arf-dependent mechanism (Ktistakis et al., 1995; Freyberg et al., 2001). In addition, PI-4-kinase β is recruited to the Golgi complex by an Arf-1 dependent mechanism and once on the membrane it stimulates the production of PI(4,5)P2 (Godi et al., 1999). Thus, it is likely that lipid-modifying enzymes, recruited to membranes by Arf-dependent mechanisms, control the dynamics of the Golgi complex (Sweeney et al., 2002) by creating membrane microdomains favorable for adaptor and coat recruitment. AP-1 and AP-3 recruitment to liposomes differ in the specificity of lipids required. AP-3 recruitment to synthetic liposomes occurs irrespective of the presence of acidic phospholipid (Drake et al., 2000). Moreover, AP-3, but not AP-1 or AP-2, can be recruited to synaptic vesicles in an Arf1-dependent manner (Faundez et al., 1998). Interestingly, synaptic vesicles do not possess detectable levels of PIs (Micheva et al., 2001). Additionally, the ATP-requirement to recruit AP-3 to membranes can be completely replaced by ATPγS (Faundez and Kelly, 2000), a poor substrate for lipid modifying enzymes (Hay and Martin, 1992). In this regard, AP-3 is closely related to COPI, which can also be recruited to synthetic liposomes irrespective of the presence of acidic phospholipids (Spang et al., 1998; Bremser et al., 1999). These observations agree with the close evolutionary relationship between AP-3 and COPI (Scheldzewski et al., 1999; Boehm and Bonifacino, 2001). In addition, they suggest that the interaction of adaptors with acidic

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

11 269

phospholipids is a recent evolutionary addition superimposed onto the general mechanism constituted by a direct interaction of Arf and the large subunits of the tetrameric adaptors. 3.2.2

What controls AP-2 membrane recruitment?

Initially, adaptor recruitment to membranes was considered to be primarily governed by interactions with membrane protein cargo. This model was supported by observations that adaptors could bind the cytosolic tails of membrane cargo proteins (Glickman et al., 1989; Pearse, 1989; Pearse, 1988; Ohno et al., 1995), that AP-1 recruitment to Golgi membranes was controlled by the abundance of mannose 6-phosphate receptors (Le Borgne and Hoflack, 1997; Pearse, 1985), and by the identification of synaptotagmin I and other isoforms as high affinity, membrane receptors for AP-2 (Haucke et al., 2000; Zhang et al., 1994; Ullrich et al., 1994; von Poser et al., 2000). However, the role of M6PRs as a receptor for AP-1 has been recently challenged (Zhu et al., 2001b; Zhu et al., 1999b; Puertollano et al., 2001). Moreover, increasing or decreasing the cargo concentration available for AP-2 at plasma membrane domains, either by ligand induced receptor clustering or by overexpression, did not affect the recruitment of AP-2 to plasma membrane (Santini et al., 1998; Santini and Keen, 1996). How then can we understand AP-2 recruitment to membranes? The synapse has provided a useful model system to understand the mechanisms controlling adaptor recruitment. Synaptojanin is a 5’ PIphosphate phosphatase; known to control the levels of PI(4,5)P2 in plasma membrane (McPherson et al., 1996) and to play a central role in uncoating AP-2/clathrin coated vesicles. Neurons from synaptojanin I -/- mice possess a substantial increase in the number of AP-2/clathrin coated vesicles, suggesting that the levels of PIP2 phospholipids could control AP-2 recruitment to membranes (Cremona et al., 1999; Harris et al., 2000). In fact, AP-2 possesses two PI-binding surfaces containing conserved lysine residues (Collins et al., 2002). These regions lie on the µ2 and α chains of AP-2. The crystal structure of the AP-2 core reveals that these lysines generate positively charged surfaces on the N-terminal region of α-adaptin facing the membrane and the C-terminal half of µ2. Mutating these residues in either adaptin completely abrogates the binding of AP-2 to membranes (Gaidarov and Keen, 1999; Rohde et al., 2002). Replacing the N-terminus of α-adaptin with the equivalent domain of γ-adaptin also abrogates plasma membrane AP-2 recruitment (Page and Robinson, 1995). Consistently, neomycin (West et al., 1997) and the pleckstrin homology domain of PLD-δ (Jost et al., 1998), molecules known to bind to PI(4,5)P2, hinder AP-2

12 270

STYERS AND FAUNDEZ

binding to membranes in cell free assays that recapitulate adaptor binding to membranes or AP-2-dependent plasma membrane vesiculation. The critical residues for phosphatidylinositols binding in α-adaptin, lysines 55-57 (Gaidarov and Keen, 1999), and µ2-adaptin, lysines 344-346 (Rohde et al., 2002) are absolutely conserved among AP-2 chains from invertebrates to mammals. These residues are not present in similar regions of the subunits of other adaptors, suggesting that quantitative and qualitative differences distinguish the lipid requirements to recruit different adaptors to membranes. The PI(4,5)P2-AP-2 interaction and functionality has been nicely demonstrated using reconstituted liposomes. In this experimental system the presence of negatively charged lipids, like PI(4,5)P2, in the donor membrane is sufficient to trigger adaptor recruitment and vesicle formation even in the absence of GTP (Takei et al., 1998). Although the most recently diverged adaptor, AP-2, maintains the pair of dimers structure, the mechanism of its recruitment may have diverged from being dependent upon a direct interaction with Arf to a mechanism controlled by a network of Arf-effectors. These Arf effectors control the lipid composition of the membrane and consequently, AP-2 recruitment to the membrane. 3.2.3

Does Arf control AP-2 membrane recruitment?

Early indications that GTPases could modulate AP-2 came from studies using perforated cells and cell-free assays (Seaman et al., 1993; West et al., 1997). AP-2 can be recruited to a variety of membranes by mechanisms that differ in their nucleotide requirement (Arneson et al., 1999; Zhang et al., 1994; Traub et al., 1996). The best characterized example is the AP-2 recruitment to synaptic membrane preparations (Takei et al., 1996). In this system, AP-2 binding to membranes requires ATP and is stimulated by the addition of GTPγS. GTPγS-dependent AP-2 recruitment to membranes has been initially regarded as dependent on the GTPase dynamin. However, as we will discuss, it could reflect the involvement of other GTPases. If PIP2 lipids are involved in the recruitment of AP-2, then a critical regulatory step is likely to reside in the localized synthesis of PIP2. Among those proteins responsible to control PIP2 levels, for this discussion we will concentrate on the enzymes that regulate AP-2/clathrin recruitment. PIP-5kinase Iγ (Wenk et al., 2001) and the class II PI 3-kinase C2α (Gaidarov et al., 2001) control the production of different PIP2’s at the plasma membrane. Both enzymes are necessary to recruit AP-2 to either liposomes or membranes. As expected, the effects of the PIP-5-phosphatase (synaptojanin I) and the PIP 5-kinase (PIP kinase Iγ) are antagonistic with regard to AP-2 recruitment to membranes (Wenk et al., 2001).

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

13 271

PIP-5-kinase Iγ and the class II PI 3-kinase C2α are both regulated by proteins involved in vesicle formation. The class II PI 3-kinase C2α activity is regulated by clathrin binding to the enzyme (Gaidarov et al., 2001). Importantly, PIP 5-kinase activity is regulated by Arf6 (Honda et al., 1999; Jones et al., 2000) and phosphatidic acid (Arneson et al., 1999; Honda et al., 1999), a product of the Arf6 effector, PLD. Arf6, the PI(4)P 5kinase Iγ, and PLD are present in plasma membrane, suggesting a coordinated role in regulating endocytosis. Indeed, dominant negative mutants of PLD have been described recently with the ability to inhibit EGF receptor endocytosis (Shen et al., 2001). If Arf6 controls two effectors at the plasma membrane that are known to regulate AP-2 recruitment, then a logical prediction would be that Arf6 should modulate the formation of AP2-containing vesicles. Dominant Arf6 mutants perturb the AP-2-dependent endocytosis of transferrin (D’Souza-Schorey et al., 1995) and β2-adrenergic receptors Claing et al., 2001). In polarized epithelial MDCK cells, Arf6 is selectively targeted to the apical plasma membrane where it regulates pIgA receptormediated endocytosis and the levels of plasma membrane-bound clathrin (Altschuler et al., 1999). These effects are restricted to the apical pole in polarized cells, without affecting pIgA receptor-mediated endocytosis at the basolateral domain. Differential sorting of transferrin receptors and β2adrenergic receptors in distinct clathrin-coated plasma membrane domains support the proposal that specialized types of clathrin-dependent endocytosis exist in a single cell type (Cao et al., 1998; Prasad et al., 2002). This argues in favor of evolutionary divergence of the mechanisms that control the recruitment of coats and adaptors to membranes. The synapse is a highly specialized region of the neuron where endocytosis contributes to the retrieval of synaptic vesicle components after regulated secretion (Hannah et al., 1999; Brodin et al., 2000; Gad et al., 2000; Shupliakov et al., 1997). Deficiencies in the synaptic endocytic machinery lead to modifications in synaptic transmission (Brodin et al., 2000; Gad et al., 2000; Shupliakov et al., 1997; Stimson and Ramaswami, 1999). Microinjection of the Arf6-GEF msec7-1, a rat homolog of human cytohesin I, to the neurons of embryonic Xenopus neuromuscular preparations increases excitatory post-synaptic potentials, consistent with a putative role of Arf6 in pre-synaptic membrane traffic (Ashery et al., 1999). Similarly, expression of a dominant activating Arf6 mutant increases the targeting of selected membrane protein markers to synaptic-like microvesicles in PC12 cells, a model neuroendocrine cell line (Powelka and Buckley, 2001). So far, there is no evidence of a direct interaction between Arf6 and AP2. However, β-arrestin, a clathrin adaptor involved in G protein coupled-

14 272

STYERS AND FAUNDEZ

receptor traffic, can bind Arf6•GDP and ARNO, forming a complex that catalyzes the exchange of GDP for GTP (Claing et al., 2001). Because βarrestin possesses binding sites for AP-2 (Laporte et al., 2000; Miller and Lefkowitz, 2001) as well as PIP2 (Gaidarov and Keen, 1999) it is possible that β-arrestin brings together Arf6 and AP-2 in a PIP2 membrane patch. Alternatively, AP-2 may interact with the by-product of an Arf-effector interaction, such as PIP2, which is generated by activating a PIP 5-kinase. In either case, the nature of the interaction is indirect. As previously discussed for COPI, AP-1, AP-3 and AP-4, it is likely that Arf is wedged between the membrane facing regions of the two large subunits of the heterotetrameric coat complexes. Instead, in AP-2, α-adaptin and the µ2 subunit, by virtue of tight binding to PI(4,5)P2, dock AP-2 to membrane lipids, likely obliterating an Arf-binding pocket. PI(4,5)P2 binds to Arfs with affinities on the order of -5 -7 10 M (Randazzo, 1997) whereas it binds α-adaptin with a KD of 10 M (Gaidarov and Keen, 1999). Thus, it is unlikely that an Arf molecule could co-exist with AP-2 on membrane patches rich in AP-2 and PI(4,5)P2. The -7 simultaneous binding of arrestin to AP-2 and PI(4,5)P2 (KD=10 M) could provide a selective means to generate clathrin-coated vesicles preferentially enriched in β-adrenergic receptors.

4.

MUTUALISM IN ARF-EFFECTOR INTERACTIONS INVOLVED IN MEMBRANE CARRIER BIOGENESIS

The concept of mutualism, interactions with a positive net impact on the fitness of the elements to perform their functions, is not restricted to either the nature of the interaction (direct or indirect) or to the number of elements involved. The spectrum of mutualistic interactions extends from that in lichens, algae living inside of the body of a fungus that are not facultative and require permanent direct contact between the organisms, to interactions where each organism lives independently. We propose that a similar spectrum of interactions operates in the mechanisms that control the formation of membrane carriers. First, at a simple organizational level, it appears that the pair of dimers model is an obligatory feature in the organization of adaptor heterotetramers. As we discussed, those subunits that are not dimers in vivo cannot exist independently. This core interaction provides a basis for the interactions with Arf GTPases. The binding of COPI, AP-1, AP-3 and AP-4 with Arf, though transient, is at some point direct. Finally, as in the case of AP-2 and AP-1, intermediaries of the network (such as lipid-modifying enzymes), that are themselves Arf

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

15 273

effectors, either completely (AP-2) or partially (AP-1), modulate the recruitment of the vesiculation machine to the membrane. Depending upon the adaptor, single or multiple Arf effector interactions appear to dominate the coating reactions. The simplest vesiculation systems, those generated by liposomes, purified coat, and GTPase, suggest that simple interactions dominate primordial coats like COPI (Bremser et al., 1999) while multiple and indirect mechanisms prevail in the case of AP-2 (Takei et al., 1998), an evolutionarily younger coat. Thus, in the absence of any other Arf1 effector but COPI, it is possible to bud COPI vesicles from liposomes by an Arf1•GTP-dependent mechanism, irrespective of the lipid composition of the artificial membranes (Bremser et al., 1999). On the other extreme, AP-2/clathrin coats and buds can be generated in brain lipidderived liposomes with clathrin coated vesicle enriched mixtures in the absence of any nucleotide; suggesting that the nature of the membrane changes the requirement for GTP and therefore the GTPase (Takei et al., 1998). However, it does appear that the GTPase controls the generation of lipid microdomains to which adaptors can be recruited. How can we understand the interactions between a GTPase and adaptors? Does AP-2 follow a completely different mechanism from other adaptors? As we have previously discussed, Arf1-5 are recruited to membranes where they interact with effectors only when they bind GTP. In contrast, Arf6 remains constitutively associated with membranes, irrespective of the nucleotide bound. In addition, at steady state, a great proportion of AP-2 is bound to membranes (Pearse and Robinson, 1990; Keen, 1990). Thus, a strategy based on Arf-adaptor interaction after membrane recruitment seems unlikely. Instead, an interaction between Arf6 and possibly AP-2 could be mediated by arrestin and/or lipid modifying Arf effectors; such as PLD or PI(4)P 5-kinase. This mechanism opens the possibility to control AP-2 recruitment by extracellular signals capable of controlling arrestins, PLD, PIP-kinase and lipid-phosphatases. These considerations are likely not only restricted to AP-2, because the recruitment of AP-1 to membranes can also be controlled by acidic phospholipids or lipid modifying enzymes. These modifications on an ancient mechanism add new levels of regulation that likely improve the system in response to new selective pressures. The “mutualistic network” vision is compatible with an effector molecule, such as an adaptor, having a life separate from its interactions with Arf. Rather than Arf-effector interactions being seen as a linear sequence of cause-effect relationships, we prefer to describe their organization as a collection of Arf-effector intermediaries that branch and can feed back into the network. Each effector could bring different inputs to generate a similar outcome without necessarily invoking the same sequence and/or extent of events. The mutualistic network idea can account for observations such as

16 274

STYERS AND FAUNDEZ

the recent description that β-arrestin can bind Arf6•GDP and induce the GDP-GTP exchange (Claing et al., 2001). These results suggest the counterintuitive idea that an adaptor (arrestin) could arrive at the membrane first and then activate the GTPase (Zhu et al., 2000). If Arf activation is the primary input to a branched network with feedback regulation, it is not unreasonable to expect that even after shutdown of Arf function, other network components (Arf-activated effectors or membrane cargo) could maintain the network functional status (such as an adaptor bound to membranes). We predict that the kinetic parameters that govern the activation or membrane recruitment of Arf as well as distinct effectors should become independent, as the initial Arf input is disseminated in the network. Indeed, in vivo Arf1 and COPI membrane association at the Golgi complex obey different kinetic and thermodynamic parameters; indicating that Arf and coat recruitment to membranes can be dissociated (Presley et al., 2002). Simultaneous in vivo analysis of the dynamics of Arfs and their effectors on membranes will provide the evidence required to test the mutualism model. The interactions between Arf and adaptors have become increasingly disparate over time and appear to have reached a summit with the adaptor AP-2. This divergence may have been selected as the cell-cell communication and intracellular traffic demands have progressively increased from a free-living unicellular organism to a multi-cellular organism comprised of highly specialized cells and tissues. However, the common theme across different levels of complexity in carrier forming machines is the formation of networks that interact in a mutualistic manner.

5.

ACKNOWLEDGEMENT

This work was supported by grants from the Basil O’Connor March of Dimes Foundation, The Melanoma Research Foundation, and the Emory University URC. M. L. Styers is a member of the Biochemistry, Cell and Developmental Biology Graduate program from the Graduate Division of Biomedical Sciences, Emory University School of Medicine.

REFERENCES Aguilar, R. C., Boehm, M., Gorshkova, I., Crouch, R. J., Tomita, K., Saito, T., Ohno, H., and Bonifacino, J. S. (2001). Signal-binding specificity of the mu4 subunit of the adaptor protein complex AP-4. J. Biol. Chem., 276, 13145-13152.

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

17 275

Aguilar, R. C., Ohno, H., Roche, K. W., and Bonifacino, J. S. (1997). Functional domain mapping of the clathrin-associated adaptor medium chains mu1 and mu2. J. Biol. Chem., 272, 27160-27166. Altschuler, Y., Liu, S., Katz, L., Tang, K., Hardy, S., Brodsky, F., Apodaca, G., and Mostov, K. (1999). ADP-ribosylation factor 6 and endocytosis at the apical surface of MadinDarby canine kidney cells. J. Cell Biol., 147, 7-12. Arneson, L. S., Kunz, J., Anderson, R. A., and Traub, L. M. (1999). Coupled inositide phosphorylation and phospholipase D activation initiates clathrin-coat assembly on lysosomes. J. Biol. Chem., 274, 17794-17805. Ashery, U., Koch, H., Scheuss, V., Brose, N., and Rettig, J. (1999). A presynaptic role for the ADP ribosylation factor (ARF)-specific GDP/GTP exchange factor msec7-1. Proc. Natl. Acad. Sci., USA, 96, 1094-1099. Austin, C., Boehm, M., and Tooze, S. A. (2002). Site-specific cross-linking reveals a differential direct interaction of class 1, 2, and 3 ADP-ribosylation factors with adaptor protein complexes 1 and 3. Biochemistry, 41, 4669-4677. Austin, C., Hinners, I., and Tooze, S. A. (2000). Direct and GTP-dependent interaction of ADP-ribosylation factor 1 with clathrin adaptor protein AP-1 on immature secretory granules. J. Biol. Chem., 275, 21862-21869. Boehm, M., Aguilar, R. C., and Bonifacino, J. S. (2001). Functional and physical interactions of the adaptor protein complex AP- 4 with ADP-ribosylation factors (ARFs). EMBO J., 20, 6265-6276. Boehm, M., and Bonifacino, J. S. (2001). Adaptins: the final recount. Mol. Biol. Cell, 12, 2907-2920. Bremser, M., Nickel, W., Schweikert, M., Ravazzola, M., Amherdt, M., Hughes, C. A., Sollner, T. H., Rothman, J. E., and Wieland, F. T. (1999). Coupling of coat assembly and vesicle budding to packaging of putative cargo receptors. Cell, 96, 495-506. Brodin, L., Low, P., and Shupliakov, O. (2000). Sequential steps in clathrin-mediated synaptic vesicle endocytosis. Curr. Opin. Neurobiol., 10, 312-320. Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity. Cell, 75, 1137-1144. Cao, T. T., Mays, R. W., and von Zastrow, M. (1998). Regulated endocytosis of G-proteincoupled receptors by a biochemically and functionally distinct subpopulation of clathrincoated pits. J. Biol. Chem., 273, 24592-24602. Cavenagh, M. M., Whitney, J. A., Carroll, K., Zhang, C., Boman, A. L., Rosenwald, A. G., Mellman, I., and Kahn, R. A. (1996). Intracellular distribution of Arf proteins in mammalian cells. Arf6 is uniquely localized to the plasma membrane. J. Biol. Chem., 271, 21767-21774. Chen, Y. A., and Scheller, R. H. (2001). SNARE-mediated membrane fusion. Nat. Rev. Mol. Cell Biol., 2, 98-106. Chow, V. T., Sakharkar, M. K., Lim, D. P., and Yeo, W. M. (2001). Phylogenetic relationships of the seven coat protein subunits of the coatomer complex, and comparative sequence analysis of murine xenin and proxenin. Biochem. Genet., 39, 201-211. Claing, A., Chen, W., Miller, W. E., Vitale, N., Moss, J., Premont, R. T., and Lefkowitz, R. J. (2001). beta-Arrestin-mediated ADP-ribosylation factor 6 activation and beta 2adrenergic receptor endocytosis. J. Biol. Chem., 276, 42509-42513. Cockcroft, S., Thomas, G. M., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of ARF in granulocytes. Science, 263, 523-526.

18 276

STYERS AND FAUNDEZ

Collins, B. M., McCoy, A. J., Kent, H. M., Evans, P. R., and Owen, D. J. (2002). Molecular architecture and functional model of the endocytic AP2 complex. Cell, 109, 523-535. Cremona, O., Di Paolo, G., Wenk, M. R., Luthi, A., Kim, W. T., Takei, K., Daniell, L., Nemoto, Y., Shears, S. B., Flavell, R. A., McCormick, D. A., and De Camilli, P. (1999). Essential role of phosphoinositide metabolism in synaptic vesicle recycling. Cell, 99, 179188. Dell'Angelica, E. C., Mullins, C., and Bonifacino, J. S. (1999b). AP-4, a novel protein complex related to clathrin adaptors. J. Biol. Chem., 274, 7278-7285. Dell'Angelica, E. C., Ohno, H., Ooi, C. E., Rabinovich, E., Roche, K. W., and Bonifacino, J. S. (1997). AP-3: an adaptor-like protein complex with ubiquitous expression. EMBO J., 16, 917-928. Dell'Angelica, E. C., Shotelersuk, V., Aguilar, R. C., Gahl, W. A., and Bonifacino, J. S. (1999a). Altered trafficking of lysosomal proteins in Hermansky-Pudlak syndrome due to mutations in the beta 3A subunit of the AP-3 adaptor. Mol Cell, 3, 11-21. Dittie, A. S., Hajibagheri, N., and Tooze, S. A. (1996). The AP-1 adaptor complex binds to immature secretory granules from PC12 cells, and is regulated by ADP-ribosylation factor. J. Cell Biol., 132, 523-536. Drake, M. T., Zhu, Y., and Kornfeld, S. (2000). The assembly of AP-3 adaptor complexcontaining clathrin-coated vesicles on synthetic liposomes. Mol. Biol. Cell, 11, 3723-3736. D'Souza-Schorey, C., Li, G., Colombo, M. I., and Stahl, P. D. (1995). A regulatory role for ARF6 in receptor-mediated endocytosis. Science, 267, 1175-1178. Eugster, A., Frigerio, G., Dale, M., and Duden, R. (2000). COP I domains required for coatomer integrity, and novel interactions with ARF and ARF-GAP. EMBO J., 19, 39053917. Faulstich, D., Auerbach, S., Orci, L., Ravazzola, M., Wegchingel, S., Lottspeich, F., Stenbeck, G., Harter, C., Wieland, F. T., and Tschochner, H. (1996). Architecture of coatomer: molecular characterization of delta-COP and protein interactions within the complex. J. Cell Biol., 135, 53-61. Faundez, V., Horng, J. T., and Kelly, R. B. (1998). A function for the AP3 coat complex in synaptic vesicle formation from endosomes. Cell, 93, 423-432. Faundez, V. V., and Kelly, R. B. (2000). The AP-3 complex required for endosomal synaptic vesicle biogenesis is associated with a casein kinase Ialpha-like isoform. Mol. Biol. Cell, 11, 2591-2604. Feng, L., Seymour, A. B., Jiang, S., To, A., Peden, A. A., Novak, E. K., Zhen, L., Rusiniak, M. E., Eicher, E. M., Robinson, M. S., Gorin, M. B., and Swank, R. T. (1999). The beta3A subunit gene (Ap3b1) of the AP-3 adaptor complex is altered in the mouse hypopigmentation mutant pearl, a model for Hermansky- Pudlak syndrome and night blindness. Hum. Mol. Genet., 8, 323-330. Freyberg, Z., Sweeney, D., Siddhanta, A., Bourgoin, S., Frohman, M., and Shields, D. (2001). Intracellular localization of phospholipase D1 in mammalian cells. Mol. Biol. Cell, 12, 943-955. Gad, H., Ringstad, N., Low, P., Kjaerulff, O., Gustafsson, J., Wenk, M., Di Paolo, G., Nemoto, Y., Crun, J., Ellisman, M. H., De Camilli, P., Shupliakov, O., and Brodin, L. (2000). Fission and uncoating of synaptic clathrin-coated vesicles are perturbed by disruption of interactions with the SH3 domain of endophilin. Neuron, 27, 301-312. Gaidarov, I., and Keen, J. H. (1999). Phosphoinositide-AP-2 interactions required for targeting to plasma membrane clathrin-coated pits. J. Cell Biol., 146, 755-764.

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

19 277

Gaidarov, I., Smith, M. E., Domin, J., and Keen, J. H. (2001). The class II phosphoinositide 3-kinase C2alpha is activated by clathrin and regulates clathrin-mediated membrane trafficking. Mol. Cell, 7, 443-449. Glickman, J. N., Conibear, E., and Pearse, B. M. (1989). Specificity of binding of clathrin adaptors to signals on the mannose-6- phosphate/insulin-like growth factor II receptor. EMBO J., 8, 1041-1047. Godi, A., Pertile, P., Meyers, R., Marra, P., Di Tullio, G., Iurisci, C., Luini, A., Corda, D., and De Matteis, M. A. (1999). ARF mediates recruitment of PtdIns-4-OH kinase-beta and stimulates synthesis of PtdIns(4,5)P2 on the Golgi complex. Nat. Cell Biol., 1, 280-287. Hannah, M. J., Schmidt, A. A., and Huttner, W. B. (1999). Synaptic vesicle biogenesis. Annu. Rev. Cell Dev. Biol., 15, 733-798. Harris, T. W., Hartwieg, E., Horvitz, H. R., and Jorgensen, E. M. (2000). Mutations in synaptojanin disrupt synaptic vesicle recycling. J. Cell Biol., 150, 589-600. Haucke, V., Wenk, M. R., Chapman, E. R., Farsad, K., and De Camilli, P. (2000). Dual interaction of synaptotagmin with mu2- and alpha-adaptin facilitates clathrin-coated pit nucleation. EMBO J., 19, 6011-6019. Hay, J. C., and Martin, T. F. (1992). Resolution of regulated secretion into sequential MgATP-dependent and calcium-dependent stages mediated by distinct cytosolic proteins. J. Cell Biol., 119, 139-151. Hirst, J., Bright, N. A., Rous, B., and Robinson, M. S. (1999). Characterization of a fourth adaptor-related protein complex. Mol. Biol. Cell, 10, 2787-2802. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase alpha is a downstream effector of the small G protein ARF6 in membrane ruffle formation. Cell, 99, 521-532. Jackson, C. L., and Casanova, J. E. (2000). Turning on ARF: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. _00001699 _00001699. Jones, D. H., Morris, J. B., Morgan, C. P., Kondo, H., Irvine, R. F., and Cockcroft, S. (2000). Type I phosphatidylinositol 4-phosphate 5-kinase directly interacts with ADP-ribosylation factor 1 and is responsible for phosphatidylinositol 4,5-bisphosphate synthesis in the golgi compartment. J. Biol. Chem., 275, 13962-13966. Jost, M., Simpson, F., Kavran, J. M., Lemmon, M. A., and Schmid, S. L. (1998). Phosphatidylinositol-4,5-bisphosphate is required for endocytic coated vesicle formation. Curr. Biol., 8, 1399-1402. Kantheti, P., Qiao, X., Diaz, M. E., Peden, A. A., Meyer, G. E., Carskadon, S. L., Kapfhamer, D., Sufalko, D., Robinson, M. S., Noebels, J. L., and Burmeister, M. (1998). Mutation in AP-3 delta in the mocha mouse links endosomal transport to storage deficiency in platelets, melanosomes, and synaptic vesicles. Neuron, 21, 111-122. Keen, J. H. (1990). Clathrin and associated assembly and disassembly proteins. Annu. Rev. Biochem., 59, 415-438. Kirchhausen, T. (2000). Three ways to make a vesicle. Nat Rev Mol Cell Biol, 1, 187-198. Klausner, R. D., Donaldson, J. G., and Lippincott-Schwartz, J. (1992). Brefeldin A: insights into the control of membrane traffic and organelle structure. J. Cell Biol., 116, 1071-1080. Ktistakis, N. T., Brown, H. A., Sternweis, P. C., and Roth, M. G. (1995). Phospholipase D is present on Golgi-enriched membranes and its activation by ADP ribosylation factor is sensitive to brefeldin A. Proc. Natl. Acad. Sci., USA, 92, 4952-4956. Ktistakis, N. T., Brown, H. A., Waters, M. G., Sternweis, P. C., and Roth, M. G. (1996). Evidence that phospholipase D mediates ADP ribosylation factor- dependent formation of Golgi coated vesicles. J. Cell Biol., 134, 295-306.

20 278

STYERS AND FAUNDEZ

Laporte, S. A., Oakley, R. H., Holt, J. A., Barak, L. S., and Caron, M. G. (2000). The interaction of beta-arrestin with the AP-2 adaptor is required for the clustering of beta 2adrenergic receptor into clathrin-coated pits. J. Biol. Chem., 275, 23120-23126. Le Borgne, R., and Hoflack, B. (1997). Mannose 6-phosphate receptors regulate the formation of clathrin-coated vesicles in the TGN. J. Cell Biol., 137, 335-345. Lewin, D. A., and Mellman, I. (1998). Sorting out adaptors. Biochim. Biophys. Acta, 1401, 129-145. Lippincott-Schwartz, J., Roberts, T. H., and Hirschberg, K. (2000). Secretory protein trafficking and organelle dynamics in living cells. Annu. Rev. Cell Dev. Biol., 16, 557-589. Lippincott-Schwartz, J., Yuan, L., Tipper, C., Amherdt, M., Orci, L., and Klausner, R. D. (1991). Brefeldin A's effects on endosomes, lysosomes, and the TGN suggest a general mechanism for regulating organelle structure and membrane traffic. Cell, 67, 601-616. Lowe, M., and Kreis, T. E. (1998). Regulation of membrane traffic in animal cells by COPI. Biochim. Biophys. Acta, 1404, 53-66. McPherson, P. S., Garcia, E. P., Slepnev, V. I., David, C., Zhang, X., Grabs, D., Sossin, W. S., Bauerfeind, R., Nemoto, Y., and De Camilli, P. (1996). A presynaptic inositol-5phosphatase. Nature, 379, 353-357. Micheva, K. D., Holz, R. W., and Smith, S. J. (2001). Regulation of presynaptic phosphatidylinositol 4,5-biphosphate by neuronal activity. J. Cell Biol., 154, 355-368. Miller, W. E., and Lefkowitz, R. J. (2001). Expanding roles for beta-arrestins as scaffolds and adapters in GPCR signaling and trafficking. Curr. Opin. Cell Biol., 13, 139-145. Moss, J., and Vaughan, M. (1998). Molecules in the ARF orbit. J. Biol. Chem., 273, 2143121434. Ohno, H., Stewart, J., Fournier, M. C., Bosshart, H., Rhee, I., Miyatake, S., Saito, T., Gallusser, A., Kirchhausen, T., and Bonifacino, J. S. (1995). Interaction of tyrosine-based sorting signals with clathrin-associated proteins. Science, 269, 1872-1875. Ooi, C. E., Dell'Angelica, E. C., and Bonifacino, J. S. (1998). ADP-Ribosylation factor 1 (ARF1) regulates recruitment of the AP-3 adaptor complex to membranes. J. Cell Biol., 142, 391-402. Orci, L., Tagaya, M., Amherdt, M., Perrelet, A., Donaldson, J. G., Lippincott-Schwartz, J., Klausner, R. D., and Rothman, J. E. (1991). Brefeldin A, a drug that blocks secretion, prevents the assembly of non- clathrin-coated buds on Golgi cisternae. Cell, 64, 11831195. Page, L. J., and Robinson, M. S. (1995). Targeting signals and subunit interactions in coated vesicle adaptor complexes. J. Cell Biol., 131, 619-630. Palade, G. (1975). Intracellular aspects of the process of protein synthesis. Science, 189, 347358. Pavel, J., Harter, C., and Wieland, F. T. (1998). Reversible dissociation of coatomer: functional characterization of a beta/delta-coat protein subcomplex. Proc. Natl. Acad. Sci., USA, 95, 2140-2145. Pearse, B. M. (1985). Assembly of the mannose-6-phosphate receptor into reconstituted clathrin coats. EMBO J., 4, 2457-2460. Pearse, B. M. (1988). Receptors compete for adaptors found in plasma membrane coated pits. EMBO J., 7, 3331-3336. Pearse, B. M. (1989). Characterization of coated-vesicle adaptors: their reassembly with clathrin and with recycling receptors. Methods Cell Biol., 31, 229-246. Pearse, B. M., and Robinson, M. S. (1990). Clathrin, adaptors, and sorting. Annu. Rev. Cell Biol., 6, 151-171.

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

21 279

Peden, A. A., Rudge, R. E., Lui, W. W., and Robinson, M. S. (2002). Assembly and function of AP-3 complexes in cells expressing mutant subunits. J. Cell Biol., 156, 327-336. Powelka, A. M., and Buckley, K. M. (2001). Expression of ARF6 mutants in neuroendocrine cells suggests a role for ARF6 in synaptic vesicle biogenesis. FEBS Lett., 501, 47-50. Prasad, S. V., Laporte, S. A., Chamberlain, D., Caron, M. G., Barak, L., and Rockman, H. A. (2002). Phosphoinositide 3-kinase regulates beta2-adrenergic receptor endocytosis by AP2 recruitment to the receptor/beta-arrestin complex. J. Cell Biol., 158, 563-575. Presley, J. F., Ward, T. H., Pfeifer, A. C., Siggia, E. D., Phair, R. D., and LippincottSchwartz, J. (2002). Dissection of COPI and Arf1 dynamics in vivo and role in Golgi membrane transport. Nature, 417, 187-193. Puertollano, R., Aguilar, R. C., Gorshkova, I., Crouch, R. J., and Bonifacino, J. S. (2001). Sorting of mannose 6-phosphate receptors mediated by the GGAs. Science, 292, 17121716. Randazzo, P. A. (1997). Functional interaction of ADP-ribosylation factor 1 with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 272, 7688-7692. Robinson, M. S., and Kreis, T. E. (1992). Recruitment of coat proteins onto Golgi membranes in intact and permeabilized cells: effects of brefeldin A and G protein activators. Cell, 69, 129-138. Rohde, G., Wenzel, D., and Haucke, V. (2002). A phosphatidylinositol (4,5)-bisphosphate binding site within mu2- adaptin regulates clathrin-mediated endocytosis. J. Cell Biol., 158, 209-214. Santini, F., and Keen, J. H. (1996). Endocytosis of activated receptors and clathrin-coated pit formation: deciphering the chicken or egg relationship. J. Cell Biol., 132, 1025-1036. Santini, F., Marks, M. S., and Keen, J. H. (1998). Endocytic clathrin-coated pit formation is independent of receptor internalization signal levels. Mol. Biol. Cell, 9, 1177-1194. Schledzewski, K., Brinkmann, H., and Mendel, R. R. (1999). Phylogenetic analysis of components of the eukaryotic vesicle transport system reveals a common origin of adaptor protein complexes 1, 2, and 3 and the F subcomplex of the coatomer COPI. J. Mol. Evol., 48, 770-778. Seaman, M. N., Ball, C. L., and Robinson, M. S. (1993). Targeting and mistargeting of plasma membrane adaptors in vitro. J. Cell Biol., 123, 1093-1105. Seaman, M. N., Sowerby, P. J., and Robinson, M. S. (1996). Cytosolic and membraneassociated proteins involved in the recruitment of AP-1 adaptors onto the trans-Golgi network. J. Biol. Chem., 271, 25446-25451. Shen, Y., Xu, L., and Foster, D. A. (2001). Role for phospholipase D in receptor-mediated endocytosis. Mol. Cell Biol., 21, 595-602. Shupliakov, O., Low, P., Grabs, D., Gad, H., Chen, H., David, C., Takei, K., De Camilli, P., and Brodin, L. (1997). Synaptic vesicle endocytosis impaired by disruption of dynaminSH3 domain interactions. Science, 276, 259-263. Simpson, F., Bright, N. A., West, M. A., Newman, L. S., Darnell, R. B., and Robinson, M. S. (1996). A novel adaptor-related protein complex. J. Cell Biol., 133, 749-760. Spang, A., Matsuoka, K., Hamamoto, S., Schekman, R., and Orci, L. (1998). Coatomer, Arf1p, and nucleotide are required to bud coat protein complex I-coated vesicles from large synthetic liposomes. Proc. Natl. Acad. Sci., USA, 95, 11199-11204. Springer, S., Spang, A., and Schekman, R. (1999). A primer on vesicle budding. Cell, 97, 145-148. Stimson, D. T., and Ramaswami, M. (1999). Vesicle recycling at the Drosophila neuromuscular junction. Int. Rev. Neurobiol., 43, 163-189.

22 280

STYERS AND FAUNDEZ

Sweeney, D. A., Siddhanta, A., and Shields, D. (2002). Fragmentation and re-assembly of the Golgi apparatus in vitro. A requirement for phosphatidic acid and phosphatidylinositol 4,5- bisphosphate synthesis. J. Biol. Chem., 277, 3030-3039. Takatsu, H., Futatsumori, M., Yoshino, K., Yoshida, Y., Shin, H. W., and Nakayama, K. (2001). Similar subunit interactions contribute to assembly of clathrin adaptor complexes and COPI complex: analysis using yeast three-hybrid system. Biochem. Biophys. Res. Commun., 284, 1083-1089. Takei, K., Haucke, V., Slepnev, V., Farsad, K., Salazar, M., Chen, H., and De Camilli, P. (1998). Generation of coated intermediates of clathrin-mediated endocytosis on proteinfree liposomes. Cell, 94, 131-141. Takei, K., Mundigl, O., Daniell, L., and De Camilli, P. (1996). The synaptic vesicle cycle: a single vesicle budding step involving clathrin and dynamin. J. Cell Biol., 133, 1237-1250. Traub, L. M., Bannykh, S. I., Rodel, J. E., Aridor, M., Balch, W. E., and Kornfeld, S. (1996). AP-2-containing clathrin coats assemble on mature lysosomes. J. Cell Biol., 135, 18011814. Ullrich, B., Li, C., Zhang, J. Z., McMahon, H., Anderson, R. G., Geppert, M., and Sudhof, T. C. (1994). Functional properties of multiple synaptotagmins in brain. Neuron, 13, 12811291. von Poser, C., Zhang, J. Z., Mineo, C., Ding, W., Ying, Y., Sudhof, T. C., and Anderson, R. G. (2000). Synaptotagmin regulation of coated pit assembly. J. Biol. Chem., 275, 3091630924. Wenk, M. R., Pellegrini, L., Klenchin, V. A., Di Paolo, G., Chang, S., Daniell, L., Arioka, M., Martin, T. F., and De Camilli, P. (2001). PIP kinase Igamma is the major PI(4,5)P(2) synthesizing enzyme at the synapse. Neuron, 32, 79-88. West, M. A., Bright, N. A., and Robinson, M. S. (1997). The role of ADP-ribosylation factor and phospholipase D in adaptor recruitment. J. Cell Biol., 138, 1239-1254. Wickner, W., and Haas, A. (2000). Yeast homotypic vacuole fusion: a window on organelle trafficking mechanisms. Annu. Rev. Biochem., 69, 247-275. Zhang, J. Z., Davletov, B. A., Sudhof, T. C., and Anderson, R. G. (1994). Synaptotagmin I is a high affinity receptor for clathrin AP-2: implications for membrane recycling. Cell, 78, 751-760. Zhao, L., Helms, J. B., Brugger, B., Harter, C., Martoglio, B., Graf, R., Brunner, J., and Wieland, F. T. (1997). Direct and GTP-dependent interaction of ADP ribosylation factor 1 with coatomer subunit beta. Proc. Natl. Acad. Sci., USA, 94, 4418-4423. Zhao, L., Helms, J. B., Brunner, J., and Wieland, F. T. (1999). GTP-dependent binding of ADP-ribosylation factor to coatomer in close proximity to the binding site for dilysine retrieval motifs and p23. J. Biol. Chem., 274, 14198-14203. Zhen, L., Jiang, S., Feng, L., Bright, N. A., Peden, A. A., Seymour, A. B., Novak, E. K., Elliott, R., Gorin, M. B., Robinson, M. S., and Swank, R. T. (1999). Abnormal expression and subcellular distribution of subunit proteins of the AP-3 adaptor complex lead to platelet storage pool deficiency in the pearl mouse. Blood, 94, 146-155. Zhu, X., Boman, A. L., Kuai, J., Cieplak, W., and Kahn, R. A. (2000). Effectors increase the affinity of ADP-ribosylation factor for GTP to increase binding. J. Biol. Chem., 275, 13465-13475. Zhu, Y., Doray, B., Poussu, A., Lehto, V. P., and Kornfeld, S. (2001b). Binding of GGA2 to the lysosomal enzyme sorting motif of the mannose 6- phosphate receptor. Science, 292, 1716-1718.

TETRAMERIC COAT PROTEIN-ARF INTERACTIONS

23 281

Zhu, Y., Drake, M. T., and Kornfeld, S. (1999a). ADP-ribosylation factor 1 dependent clathrin-coat assembly on synthetic liposomes. Proc. Natl. Acad. Sci., USA, 96, 50135018. Zhu, Y., Drake, M. T., and Kornfeld, S. (2001a). Adaptor protein 1-dependent clathrin coat assembly on synthetic liposomes and Golgi membranes. Methods Enzymol., 329, 379-387. Zhu, Y., Traub, L. M., and Kornfeld, S. (1998). ADP-ribosylation factor 1 transiently activates high-affinity adaptor protein complex AP-1 binding sites on Golgi membranes. Mol. Biol. Cell, 9, 1323-1337. Zhu, Y., Traub, L. M., and Kornfeld, S. (1999b). High-affinity binding of the AP-1 adaptor complex to trans-golgi network membranes devoid of mannose 6-phosphate receptors. Mol. Biol. Cell, 10, 537-549.

Chapter 14 ARF6 JAMES E. CASANOVA Department of Cell Biology, University of Virginia Health Sciences Center, Charlottesville, USA

Abstract:

1.

Arf6 is the sole representative of the class III Arfs and is the most divergent (both structurally and functionally) of the Arfs. Its localization to the plasma membrane and endosomal system also distinguishes it from the other five Arfs. Both the localization of Arf6 and its functional roles in endocytosis, secretion, and desensitization of G-protein coupled receptors exhibit cell typespecificity, making comparative studies difficult and often confusing. Finally, models for Arf6 in actin organization and phospholipid signaling are discussed.

INTRODUCTION

Arf6 was first identified in 1991 by low-stringency hybridization with a bovine Arf2 probe (Tsuchiya et al., 1991). It is the sole representative of the Class III Arfs and is the most divergent in terms of both structure and function. In this case, “divergent” is a relative term; human Arf6 is 68% identical to human Arf1 at the amino acid level, and the sequences of their effector binding “switch regions” are virtually identical. Like the other Arfs, Arf6 functions in the regulation of membrane traffic. However, in contrast to the class I and II Arfs, which operate primarily at the Golgi apparatus, Arf6 localizes to the plasma membrane and a subset of endosomes, where it regulates the recycling of internalized membrane proteins. In addition to this function, and perhaps integrated with it, Arf6 also modulates assembly of the cortical actin cytoskeleton, a role not ascribed to the other Arfs. In this chapter, we will critically examine the (often contradictory) Arf6 literature and attempt to distill some unifying principles that can be used to define the role of this intriguing protein in eukaryotic cells. 283

2 284

2.

CASANOVA

BIOCHEMICAL PROPERTIES

As noted above, human Arf6 is 68% identical to human Arf1, yet the two proteins differ in important ways. First, the pI of Arf6 is significantly higher than that of Arf1 (8.5 vs 5.9; Maranda et al., 2001) and modeling of surface electrostatic potential indicates that it has an overall cationic appearance (Amor et al., 2001). In particular, the region of the protein surrounding and including the N-terminal helix is strongly basic, while in Arf1 the primarily basic helix is surrounded by an acidic ring or “cushion” (Amor et al., 2001). The strong cationic character of Arf6 may promote electrostatic interactions with acidic head groups of membrane lipids, contributing to its observed higher affinity for membranes even in the GDP-bound state. A second key difference is the presence of a four-residue deletion near the N terminus of Arf6 relative to Arf1. Crystal structures indicate that this deletion shortens the loop between the N-terminal helix and the first beta strand, resulting in subtle changes in the nature of the nucleotide-binding pocket (Menetrey et al., 2000). In Arf1, E54 helps coordinate the bound Mg++ ion, while in Arf6, the corresponding residue, E50, instead forms a hydrogen bond with S38, thus weakening the interaction with Mg++ (Menetrey et al., 2000). The overall effect of these changes is that Arf6 exhibits a higher rate of spontaneous nucleotide exchange, particularly in the presence of acidic lipids (Terui et al., 1994). As mentioned above, the effector-interacting domains of Arf6 are virtually identical with those of the other Arfs, and it is therefore not surprising that, so far, all of the effectors that have been described for Arf6 also interact with at least one other Arf. Because effectors are discussed in detail in the preceding chapters, we will describe them here only in the specific context of Arf6 function.

3.

LOCALIZATION

Unlike Arf1, which localizes to the Golgi complex, the subcellular distribution of Arf6 is complex and can vary from one cell type to another. Using epitope-tagged constructs, Peters and coworkers were the first to demonstrate the presence of Arf6 in the plasma membrane/endosomal system (Peters et al., 1995). In cells expressing relatively high levels of exogenous Arf6, immunostaining was observed at the plasma membrane and a population of perinuclear endosomes. In cells expressing low levels of the tagged protein, plasma membrane staining was reduced and the predominant staining was endosomal. This distribution has been subsequently confirmed for the endogenous protein using isoform-specific antibodies (Song et al.,

ARF6

2853

1998). Colocalization studies in CHO cells (D'Souza-Schorey et al., 1998) revealed that the Arf6-positive endosomes contained transferrin receptors and the v-SNARE cellubrevin, but were poorly accessible to fluid-phase markers, such as horseradish peroxidase. These data suggest that they represent recycling endosomes, a compartment through which many internalized membrane constituents pass before returning to the plasma membrane. Unlike Arf1, whose association with Golgi membranes is readily perturbed by brefeldin A (BFA), the binding of Arf6 to endosomal membranes was BFA-insensitive (D'Souza-Schorey et al., 1998). This is in agreement with subsequent work demonstrating that most of the Arf GEFs found in the cell periphery are insensitive to BFA (Jackson and Casanova, 2000). Further evidence of a unique localization for Arf6 came from the expression of constitutively active (Arf6-L67) or dominant inhibitory (Arf6N27) mutants. While an activated Arf1 mutant (Arf-L71) localized to Golgi membranes and stabilized the membrane association of coat proteins (Zhang et al., 1994) (Peters et al., 1995), the corresponding Arf6 mutant was typically found at the plasma membrane (Peters et al., 1995). In HEK293 cells, expression of Arf6-L67 induced the formation of numerous, deep plasma membrane invaginations and sheet-like extensions (Peters et al., 1995). Importantly, electron microscopy revealed no concentration of the mutant protein in clathrin-coated pits or vesicles, and the number and appearance of these structures was unchanged (Peters et al., 1995). These observations suggested that Arf6 does not become incorporated into clathrin coated vesicles at the cell surface and thus is less likely to play a direct role in the recruitment of AP-2. The distribution of a mutant predicted to be in the GDP-bound conformation, Arf6-N27, also differed from that of the corresponding Arf1 mutant and from the activating mutant, Arf6-L67. While Arf1-N31 is predominantly cytosolic, Arf6-N27 was found primarily on endosomal membranes (Peters et al., 1995). This observation revealed two important characteristics of Arf6: it is capable of interacting with membranes while in the GDP conformation, and its localization within the plasma membrane/endosomal system is regulated by its GTPase cycle. Interestingly, in some cells (HEK293) but not others (e.g., CHO, HeLa; D'Souza-Schorey et al., 1998; Peters et al., 1995; Peters et al., 2001), expression of the Arf6-N27 mutant induced the formation of coated endosomal structures. Recent biochemical characterization of the protein coats revealed that they do not contain components of other wellcharacterized coats involved in vesicular transport, including clathrin, adaptor complexes AP-1, AP-2, AP-3 or AP-4, or the coatomer complexes COP-I or COP-II (Peters et al., 2001). However, it should be noted that the

4 286

CASANOVA

ability of Arf6•GDP to stimulate coat formation runs counter to the paradigm for class I Arfs, in which it is the GTP-bound form that nucleates coat assembly. Clearly it will be important to determine the nature of this novel coat, and whether its assembly is a general feature of Arf6 function in all cells. The nature of the endosomal compartment(s) containing Arf6 appears to vary from one cell type to another. For example, in Chinese hamster ovary (CHO) cells Arf6 is present on endosomes that contain transferrin receptor and cellubrevin (D'Souza-Schorey et al., 1998). However, in HeLa cells Arf6 and transferrin appear in distinct endosomal structures and Arf6 mutants do not perturb transferrin traffic in these cells (Radhakrishna and Donaldson, 1997) (see below). Moreover, in renal tissue, endogenous Arf6 is concentrated at the apical pole of epithelia, on the apical plasma membrane, in apical endosomes, or both (Londono et al., 1999; Maranda et al., 2001). Finally, in adrenal chromaffin cells, the majority of endogenous Arf6 was found on the membranes of secretory granules rather than endosomes, and translocated to the plasma membrane in response to secretagogues (Caumont et al., 1998). Such diversity in the subcellular localization of Arf6 suggests an equally diverse set of functions, however, it is likely that Arf6 utilizes a small number of effectors to perform similar functions at multiple cellular sites (see below).

4.

FUNCTIONS

4.1

Endocytosis

The localization of Arf6 at the plasma membrane and in endocytic compartments suggested a potential role in endocytosis. Considering that Arf1 regulates the binding of the clathrin-associated AP-1 adaptor complex to membranes of the trans-Golgi network (Stamnes and Rothman, 1993; Traub et al., 1993; Seaman et al., 1996), it seemed plausible that Arf6 may perform the same function for the AP-2 adaptor complex at the plasma membrane. However, as noted above, early morphological studies determined that wild-type Arf6 did not concentrate in clathrin-coated pits or vesicles, nor did expression of Arf6 mutants perturb the number or distribution of these structures (Peters et al., 1995), implying that any effect of Arf6 on clathrin-mediated endocytosis is likely to be indirect. Moreover, several studies have determined that expression of either wild-type or constitutively active Arf6, or the Arf6 exchange factor EFA6 in CHO cells actually inhibits the clathrin-dependent endocytosis of transferrin receptor (D'Souza-Schorey et al., 1995; Franco et al., 1999). This ability of Arf6 to

ARF6

2875

affect clathrin-mediated endocytosis without actually localizing to clathrincoated pits suggested that it might regulate endocytosis through an indirect mechanism. An alternative hypothesis for Arf6 function in the endocytic pathway is that Arf6 controls the recycling of internalized membrane components from endosomes back to the plasma membrane. A prediction of this model is that expression of the dominant inhibitory mutant, Arf6-N27, should induce the accumulation of normally recycling proteins in an endosomal compartment and reduce their levels at the plasma membrane. In general, available experimental data support this model, although the specific proteins affected vary with cell type. For example, the endosomal distribution of transferrin receptor (which is often used as a model for recycling receptors) overlaps significantly with that of Arf6 in CHO cells (D'Souza-Schorey et al., 1998), and expression of Arf6-N27 inhibited recycling of internalized transferrin in these cells (D'Souza-Schorey et al., 1995). In contrast, Arf6 exhibited little overlap with transferrin receptors in HeLa cells, and instead appeared to localize to a distinct endosomal compartment, enriched in proteins internalized by clathrin-independent mechanisms (e.g., MHC class I or IL-2 receptor α chain [also known as Tac] (Radhakrishna and Donaldson, 1997)). Perhaps not surprisingly, Arf6-N27 has little effect on transferrin recycling in these cells, and instead dramatically inhibited the recycling of Tac (Radhakrishna and Donaldson, 1997). While the observations in CHO and HeLa cells were superficially disparate, they share important similarities. In both cases, Arf6 regulates the flow of membrane from an endosomal compartment to the cell surface. In both cases, the compartment on which Arf6 resides is perinuclear, exhibits tubular extensions and is relatively inaccessible to fluid-phase markers such as HRP. All of these are defining characteristics of “recycling endosomes”. Maxfield and colleagues have shown in CHO cells that the default transport pathway for internalized “bulk” membrane components (i.e., lipids or membrane proteins lacking cytoplasmic sorting signals that would divert them from this route) passes through the recycling endosome (Mayor et al., 1993). The enrichment of MHC class I and Tac in the Arf6-labeled compartment in HeLa cells suggests that it is functionally similar to the recycling endosome in CHO. The major difference between the two cell lines therefore appears to be that Arf6 regulates transferrin recycling in CHO but not HeLa cells. This may simply reflect a difference in the site at which transferrin receptor is sorted following endocytosis. In CHO cells, a large fraction of the receptor passes through the recycling endosome, whereas in HeLa it is possible that most of it is returned to the plasma membrane directly from more peripheral sorting endosomes. A direct comparison between the two cell lines, using the same parameters, will be required to definitively resolve this issue.

6 288

4.2

CASANOVA

Recycling of the GLUT4 Glucose Transporter

Insulin stimulates glucose uptake into adipose and muscle tissues by triggering translocation of the glucose transporter GLUT4 from an intracellular storage pool to the plasma membrane. This shift in distribution is accomplished by an insulin-induced increase in the rate of GLUT4 recycling, coupled with a decrease in endocytic rate. However, the transporters continue to cycle through the endocytic and recycling pathways even in the presence of insulin, indicating that GLUT4 transport is a form of regulated recycling (Pessin et al., 1999). The effects of Arf6 mutants on the recycling of other membrane proteins suggested that Arf6 may regulate GLUT4 traffic in insulin-responsive cells. However, initial tests of this hypothesis produced somewhat contradictory results. In one report, addition of myristoylated peptides, corresponding to the Arf6 N-terminal helix, to permeabilized adipocytes significantly impaired insulin-induced GLUT4 translocation, while corresponding peptides from Arf1 or Arf5 had little effect (Millar et al., 1999). However, in a second report, expression of a nucleotide-free, dominant inhibitory Arf6 mutant (Arf6-N125) in intact adipocytes had no effect on either GLUT4 translocation or glucose uptake (Yang and Mueckler, 1999). More recent data from two other laboratories suggest that the latter report is correct; expression of the more widely used dominant inhibitory form, Arf6-N27, failed to inhibit insulin-induced GLUT4 translocation (Lawrence and Birnbaum, 2001; Bose et al., 2001). However, glucose uptake can also be stimulated by a different agonist, endothelin-1, and it was discovered that Arf6-N27 did inhibit GLUT4 translocation in adipocytes treated with endothelin (Lawrence and Birnbaum, 2001; Bose et al., 2001). In contrast to insulin, which binds to and activates a receptor tyrosine kinase, endothelin binds to a G-protein coupled receptor, linked to the heterotrimeric GTPases Gαq or Gα11. In fact, the effects of endothelin on GLUT4 traffic could be mimicked by expression of activated Gαq or Gα11 mutants. Several other important differences between the two signaling pathways were also noted. First, insulin-induced glucose transport was sensitive to inhibition of PI 3kinase by wortmannin, whereas endothelin-induced transport was not (Bose et al., 2001). Second, while insulin induced translocation of both GLUT4 and a second glucose transporter, GLUT1, endothelin stimulated GLUT4, but not GLUT1 transport (Lawrence and Birnbaum, 2001). Third, while both insulin and endothelin induced reorganization of cortical actin, only endothelin-induced actin assembly was inhibited by Arf6-N27 (Bose et al., 2001). These results, while intriguing, raise several important questions. First, how can the same transport process be differentially sensitive to Arf6? One

ARF6

2897

possibility, for which there is experimental evidence, is that GLUT4 exists in two distinct endosomal pools (Millar et al., 1999). One of these pools is thought to reside in a specialized storage compartment, analogous to regulated secretory granules, which would presumably be sensitive to insulin but not endothelin. The other pool may represent transporters residing in the constitutive recycling pathway, and would respond to endothelin signaling via Arf6 (Figure 1).

Figure 1. Alternative models for the role of Arf6 in GLUT4 transport. In model A, ligation of endothelin receptors activates Arf6 through a PI3-kinase independent mechanism (possibly EFA6) and this stimulates the recycling of GLUT4 transporter via a constitutive recycling pathway. Insulin mediated signaling stimulates translocation of both GLUT4 and GLUT1 via a distinct, insulin sensitive pathway that does not require Arf6 function. The same results could also be explained by model B, in which expression of Arf6-N27 leads to selective sequestration of endothelin receptors within the recycling endosomal compartment, reducing the number of receptors at the cell surface.

8 290

CASANOVA

A more puzzling question is the role of PI 3-kinase. Direct evidence exists in adipocytes that insulin stimulates the PI 3-kinase-dependent recruitment of two different exchange factors for Arf6, ARNO and GRP1, to the plasma membrane via their PH domains (Venkateswarlu et al., 1998; Langille et al., 1999). Presumably Arf6 is activated under these conditions, yet GLUT4 translocation is not inhibited by Arf6-N27. Conversely, endothelin triggers translocation independently of PI 3-kinase, yet Arf6 activation is necessary for this event. It is possible that endothelin and insulin recruit distinct Arf GEFs (e.g., EFA6 vs. ARNO or GRP-1), however this remains to be tested. Another possibility is that Arf6-N27 sequesters endothelin, but not insulin receptors, within an endosomal compartment, so that they are not available for activation at the cell surface, a prospect that was not examined in either study. Although it is clear that more work is necessary to clarify its precise role in glucose transporter traffic, these data suggest that Arf6 regulates some, but not all forms of endosomal recycling.

4.3

Regulation of endocytosis in epithelial cells

An endocytic function for Arf6 has also been described in polarized epithelial cells. As noted above, endogenous Arf6 has been localized by electron microscopy to the apical brush border membrane and to apical endocytic structures in kidney proximal tubule epithelia (Londono et al., 1999; Maranda et al., 2001). Altschuler et al. (Altschuler et al., 1999) have demonstrated a similar distribution for ectopically expressed Arf6 in two renal tubule cell lines, MDCK and LLC-PK1. However, in contrast to its inhibitory effects in non-polarized cells, expression of Arf6-L67 in MDCK cells actually stimulated clathrin-mediated endocytosis (Altschuler et al., 1999). Moreover, the observed increase in endocytic rate was restricted to the apical pole, consistent with the predominantly apical distribution of both the wild-type and mutant Arf6. The enhanced endocytic rate correlated with an increased density of clathrin-coated pits at the apical membrane (Altschuler et al., 1999), again in contrast with observations in non-polarized cells (Peters et al., 1995). The ability of Arf6 to influence apical endocytosis may be related to its capacity to modulate actin cytoskeleton assembly (see below). Although the precise relationship between actin and clathrin assembly at the apical membrane is unclear, apical endocytosis is acutely sensitive to actin depolymerizing drugs such as cytochalasin D, whereas basolateral endocytosis is not (Gottlieb et al., 1993). A more puzzling observation, however, was the finding that the dominant inhibitory mutant, Arf6-N27, also stimulated apical endocytosis (Altschuler et al., 1999). The reasons for this are not readily apparent, and will clearly require further study to resolve.

ARF6

2919

Palacios et al. (Palacios et al., 2001) have reported a very different distribution of Arf6 in MDCK cells. In their hands, the same epitope-tagged Arf6 construct localized to lateral rather than apical membranes. Although apical endocytosis was not measured, these authors found that expression of Arf6-L67 led to disassembly of adherens junctions, accompanied by internalization of the laterally localized cell adhesion molecule, E-cadherin. Endocytosis of E-cadherin is not thought to be clathrin-mediated, and was not assayed by Altshuler, et al. In contrast, the dominant negative mutant (Arf6-N27) inhibited junction breakdown and E-cadherin internalization in response to hepatocyte growth factor (HGF), which normally induces separation and migration of epithelial cells. Again, this observation disagrees with the results in non-polarized cells, where Arf6-N27 inhibits recycling, not endocytosis. However, in this context the effect of Arf6 on Ecadherin internalization may be indirect. Cell surface cadherin is anchored to the actin cytoskeleton, which limits its diffusion in the plane of the membrane, but may also prevent its constitutive internalization. If Arf6 is activated by HGF, this could lead to reorganization of cortical actin, dissociation of E-cadherin from the cytoskeleton, and its subsequent endocytosis. The ability of Arf6-N27 to inhibit this process would be consistent with such a hypothesis, but this remains to be tested directly.

4.4

Regulated Secretion

A role for Arf in regulated secretion has been described in a number of experimental systems including HL-60 cells, neutrophils, RBL basophilic leukemia cells (a mast cell model), adrenal chromaffin cells, and their tumor derivative, PC12 cells. A consensus among workers in this field seems to be that Arf6 promotes regulated secretion by stimulating the activity of the well-characterized Arf effector phospholipase D (PLD) at the plasma membrane. For instance, Bader and coworkers (Caumont et al., 1998) have shown that in resting adrenal chromaffin cells Arf6 is localized to the membranes of secretory granules, while PLD resides on the plasma membrane. Stimulation of these cells resulted in translocation of Arf6 to the plasma membrane, coincident with an increase in PLD activity and granule content release. PLD activity in permeabilized cells could be stimulated by recombinant, myristoylated Arf6, and inhibited by isoform-specific Arf6 antibodies (Caumont et al., 1998), demonstrating that Arf6 is not simply carried with the granule membrane to the cell surface but plays an important role in the exocytic event. In contrast, Wakelam and coworkers (Brown et al., 1998) have shown that PLD resides on secretory granules in RBL cells (rather than the plasma membrane), and that it translocates to the plasma membrane in response to

10 292

CASANOVA

antigen stimulation. Little colocalization of Arf6 with PLD was observed prior to stimulation, but significant overlap was demonstrated in stimulated cells. Thus it appears that both Arf6 and PLD are required for regulated secretion but the mechanisms for bringing the two proteins together at the plasma membrane differ among cell types. Notably, it is PLD1b that appears to be the target of Arf6 in most cells. In transfected PC12 cells, a catalytically inactive PLD1b mutant inhibited the K+-induced release of stored growth hormone, while a similar PLD2 mutant did not (Vitale et al., 2001). Powner et al. demonstrated directly, using surface plasmon resonance, that PLD1b can bind Arf6 (Powner et al., 2002). Furthermore, PLD1b also binds Rac1 and PKCα simultaneously with Arf6, and maximal PLD activity is achieved only when all three regulatory proteins are bound (Powner et al., 2002). Fractionation of membranes from resting chromaffin cells revealed that ARNO colocalized with plasma membrane markers, suggesting that Arf6 encounters its GEF upon delivery to the plasma membrane. At least in chromaffin cells, the GEF responsible for the agonist-induced activation of Arf6 appears to be ARNO (Caumont et al., 2000). Incubation of permeabilized cells with anti-ARNO antibodies inhibited both calciumevoked PLD activity and catecholamine secretion in a dose-dependent manner. Similarly, expression of a catalytically inactive ARNO mutant in PC12 cells also inhibited secretion induced by membrane depolarization. Together, these data provide strong support for the notion that Arf6 plays an important role in regulated secretion, primarily through its modulation of PLD activity.

4.5

Regulation of G-protein-coupled receptor desensitization

Accumulating evidence indicates that Arf6 can regulate the internalization and/or desensitization of a subset of G-protein coupled receptors, particularly those that are linked to arrestins (Mukherjee et al., 2000; Perry and Lefkowitz, 2002). Because this topic is covered in other chapters (see Chapters 8 and 15) it will not be described in detail here.

4.6

A role for Arf6 in actin cytoskeleton dynamics

A large body of evidence indicates that the actin cytoskeleton can be dynamically assembled and disassembled in response to extracellular cues, through signaling pathways that involve small GTPases of the Rho family (Van Aelst and D'Souza-Schorey, 1997). In most cell types, activation of

ARF6

11 293

RhoA leads to formation of stress fibers, while activation of Rac or Cdc42 results in the formation of lamellipodia or filopodia, respectively. Members of the Arf family, particularly Arf6, can also modulate cytoskeleton assembly, and may do so coordinately with Rho GTPases. Endogenous Arf6 has been localized to areas of the cell undergoing active cytoskeletal remodeling, including the leading edge of motile cells (Santy and Casanova, 2001) and the periphery of spreading cells (Song et al., 1998). Donaldson and colleagues first provided evidence that Arf6 plays an active role in actin remodeling by showing that treatment of Arf6expressing HeLa cells with aluminum fluoride (AlFn) induced the formation of actin rich membrane protrusions (Radhakrishna et al., 1996). AlFn stabilizes the interaction between many GTPases and their associated GAPs, thereby inhibiting GAP function and shifting the GTPase cycle toward the GTP-bound state. (It should be noted that AlFn also activates heterotrimeric G proteins directly by mimicking the gamma phosphate. An additional complication of this reagent is that it can inhibit a wide variety of phosphotransfer reactions in cells. Therefore, when added to intact cells, AlFn presumably activates many GTPases.) However protrusions only appeared in cells expressing exogenous Arf6 and similar structures were observed in cells transiently expressing Arf6-L67, suggesting that their formation in this context was Arf6-dependent (Radhakrishna et al., 1996). Similar observations were made in CHO cells, where expression of Arf6L67 led to formation of structures more closely resembling classical lamellipodia (D'Souza-Schorey et al., 1997). Several lines of evidence suggest that Arf6 modulates actin assembly through crosstalk with GTPases of the Rho family, probably by acting upstream of the Rho protein. First, expression of Arf6-L67 (Boshans et al., 2000) or of the Arf exchange factors ARNO (Frank et al., 1998) or EFA6 (Franco et al., 1999) interferes with assembly of actin stress fibers, a process mediated by RhoA. Direct measurement of RhoA activation revealed a robust increase in RhoA•GTP levels in control cells treated with lysophosphatidic acid (a stimulator of Rho activation), and this was paralleled by an increase in stress fiber formation (Boshans et al., 2000). However, in cells expressing Arf6-L67, no such increase in RhoA activation or stress fiber assembly was observed. In contrast, abundant stress fibers were formed in cells co-expressing Arf6-L67 and an activated RhoA mutant (RhoA-V14), indicating that Arf did not affect signaling downstream of Rho leading to stress fiber assembly (Boshans et al., 2000). Presumably Arf6 inhibits Rho activation by decreasing Rho-GEF activity, increasing RhoGAP activity, or both. The recent discovery of several Arf GAPs that also possess Rho GAP domains (Miura et al., 2002; Krugmann et al., 2002)

12 294

CASANOVA

suggests that recruitment of such Arf GAPs to sites of Arf6 activation could lead directly to the down regulation of Rho. As mentioned above, in many cell types activation of Arf6 also leads to the formation of flat, actin-rich structures resembling lamellipodia. Motile cells extend lamellipodia in the direction of movement, and inhibition of this process dramatically slows cell motility (Ridley et al., 1999). Membrane ruffling and lamellipodial extensions can be induced by a variety of extracellular signals including growth factors (PDGF, EGF, HGF, CSF-1, insulin), integrin ligation, and a number of G-protein coupled receptor agonists (e.g., bombesin) through signaling pathways coupled to the activation of Rac. Initial evidence of crosstalk between Arf and Rac came from the discovery that the two GTPases share a common effector, POR1/Arfaptin2 (D'Souza-Schorey et al., 1997; see Chapter 11). Although the function of this protein is not well understood, D’Souza-Schorey et al. showed that POR1 mutants interfered with actin remodeling induced by either activated Rac1 or Arf6-L67, suggesting that these two proteins operate on parallel pathways to modulate actin dynamics. However, POR1/Arfaptin2 interacts with Arf1 as well as Arf6, and exhibits a striking localization to the Golgi rather than the plasma membrane (Kanoh et al., 1997). While it remains possible that a fraction of POR1/Arfaptin2 functions in the cell periphery to modulate cytoskeleton assembly, it seems more likely that its primary function is at the Golgi, and the inhibitory effects of mutant expression on lamellipodial extension are due to mislocalization of the overexpressed constructs. Further evidence of crosstalk between Rac and Arf6 is provided by the observation that Arf6-N27 inhibits membrane ruffling in response to EGF (Honda et al., 1999), bombesin (Boshans et al., 2000) or CSF-1 (Zhang et al., 1999). In macrophages, Arf6-N27 dramatically inhibits Fcγ-mediated phagocytosis (Zhang et al., 1998), a process also known to require activation of both Rac and Cdc42. Similarly, activation of Arf6 by expression of the exchange factor EFA6 induces membrane ruffling that is inhibited by coexpression of dominant negative Rac1-N17 (Franco et al., 1999). Furthermore, expression of the Arf GEF ARNO in MDCK cells is sufficient to induce a dramatic morphological transition from a cuboidal, stationary phenotype to a flattened, motile phenotype with pronounced lamellipodia (Santy and Casanova, 2001). This transition was accompanied by the activation of both endogenous Arf6 and Rac1. As in the case of EFA6induced ruffling (Franco et al., 1999), lamellipodia did not form in the presence of dominant inhibitory Rac1-N17, suggesting that Rac lies downstream of Arf6 in a GTPase signaling pathway. How might Arf6 activation lead to activation of Rac? One model that has gained favor in recent years is that Arf6 regulates the translocation of an

ARF6

13 295

endosomal pool of Rac to the plasma membrane (Radhakrishna et al., 1999). Several groups have reported the localization of either endogenous (Radhakrishna et al., 1999) or exogenous (Boshans et al., 2000) Rac1 to endosomal membranes and the colocalization of this endosomal pool with Arf6. Moreover, stimulation of cells with either AlFn (Radhakrishna et al., 1999) or bombesin (Boshans et al., 2000) resulted in an apparent shift of both Rac and Arf6 from the endosomal pool to the plasma membrane. Importantly, expression of Arf6-N27 both prevented this translocation and inhibited membrane ruffling. Together, these data form the basis of a model in which Arf6 regulates the translocation of endosomal Rac to the plasma membrane where Racspecific GEFs would presumably activate it. This model is appealing because it incorporates the role of Arf6 in endosomal membrane recycling with its observed effects on the organization of the actin cytoskeleton. However, there are several problems with this model that bear further investigation. First, in most cells, a substantial pool of cytosolic Rac exists bound to Rho-GDI (Del Pozo et al., 2002), which could be recruited to the plasma membrane without invoking a vesicular transport mechanism. Second, Arf6-N27 inhibits membrane ruffling, induced by an activated Rac construct anchored to the plasma membrane by fusion to a heterologous transmembrane domain (Zhang et al., 1999) (suggesting that translocation to the plasma membrane is not sufficient to induce ruffling). Third, inhibition of Arf6 function impairs some, but not all aspects of Rac signaling; Boshans et al. have shown that while Arf6-N27 dramatically inhibits bombesininduced membrane ruffling, it has no effect on bombesin-induced MAP kinase activation, which is also Rac-dependent (Boshans et al., 2000). This last observation indicates that physiological agonists can activate Rac even if Arf6 function is inhibited, and supports previous evidence that the effects of Rac on membrane ruffling and on signaling to the nucleus can be uncoupled. Direct measurement of Rac•GTP levels in the presence and absence of Arf6N27 should resolve this issue. An alternative mechanism for Arf-induced Rac activation has been proposed, in which Arf6 activation is coupled to the recruitment of a Rac GEF. A subset of the Arf GAPs, the GITs (see Chapter 8) interact directly with the Rac GEF, β-PIX, which is in turn coupled to a Rac effector, the p21-activated kinase, PAK (Manser et al., 1998). The intrinsic nucleotide exchange activity of β-PIX is quite low, however it is significantly enhanced by its interaction with PAK. In this model, the GIT/PIX/PAK complex is recruited to sites of local Arf6 activation through the interaction of the GIT Arf GAP domain with Arf6•GTP. This would lead to the localized activation of Rac, the subsequent activation of the bound PAK [if recruited as a complex], phosphorylation of PAK substrates and consequent

14 296

CASANOVA

remodeling of the actin cytoskeleton (Figure 2). A variation on this theme is that the GIT/PIX/PAK complex is constitutively localized to the plasma membrane in an inactive conformation, and is then activated by its interaction with Arf6•GTP. Testing of each model is currently underway.

Figure 2. Assembly of a GIT/PIX/PAK complex. Binding of GIT to Arf could stimulate either membrane recruitment or enzymatic activity of PIX and PAK.

4.7

Arf6, phospholipid signaling and actin remodeling

Another mechanism by which Arf6 may regulate actin dynamics is through its effects on membrane phospholipid composition. All Arfs are allosteric activators of two lipid-modifying enzymes, phospholipase D (PLD, see Chapter 11), and an isoform of phosphatidylinositol 4-phosphate, 5-kinase (PI(4P) 5-kinase α or PI(4P)5K). These two enzymes are functionally linked, in that PI(4P)5K is dependent on both phosphatidic acid, the product of PLD, and interaction with Arf•GTP for its activity (Honda, et al., 1999). The result of this linkage is lipid signaling in which the activation of PLD by Arf leads, through PI(4P)5K, to enhanced local production of PI(4,5)P2 (see Figure 3). Phosphatidic acid (PA) itself has a variety of biologically important properties; its charged head group can attract proteins to the membrane through electrostatic interactions, it can increase the fluidity of membrane lipids, and it can be further metabolized to generate diacylglycerol (an activator of PKC) as well as lysophosphatidic acid (a Gprotein coupled receptor agonist) (for review see Liscovitch et al., 2000). PI(4,5)P2 also has a number of biological properties, including functions in signaling, endocytosis, exocytosis, regulation of ion channels, and a wellcharacterized ability to modulate the activity of a variety of actin-associated proteins including vinculin, ezrin, profilin, α-actinin and WASp. However, a confounding aspect of the PI(4,5)P2/PLD relationship is that PI(4,5)P2 is also required for PLD activity. Both PLD1 and PLD2 possess PH domains whose binding to PI(4,5)P2 mediate the interaction of these proteins with cellular membranes (Liscovitch et al., 2000). Thus, in this

ARF6

15 297

context, a basal level of PI(4,5)P2 is necessary to bring PLD together with its substrate. Therefore, determining whether PI(4,5)P2 is serving solely to recruit PLD or is truly required as a downstream effector necessitates careful experimental design. Recently, Cockroft and coworkers used an Arf1 mutant that can activate PI(4P)5K but not PLD (Arf1-R52), to demonstrate that 45% of the Arf-stimulated PI(4,5)P2 synthesis in HL-60 cells was dependent on PLD activity, whereas the remaining 55% was PLDindependent (Skippen et al., 2002). Thus, a significant fraction of PI(4,5)P2 can be generated by a pathway that does not require new PA synthesis.

Figure 3. Phospholipid signaling downstream of Arf6.

It has become possible in recent years to localize phosphoinositides in cells using GFP fusions to PH domains of varying phosphoinositide specificity. Using such a GFP construct fused to the PH domain of PLC-δ, Brown et al. have shown that PI(4,5)P2 colocalizes extensively with Arf6, particularly at the plasma membrane and in the distal portions of tubular endosomes (Brown et al., 2001). Moreover, these authors found that activation of Arf6 by expression of the exchange factor EFA6 induced the formation of numerous large pinosomes whose membranes were enriched in both EFA6 and PI(4,5)P2. While these pinosomes were dynamic, similar large pinocytic structures were observed in cells expressing Arf6-L67, except that these accumulated intracellularly (Brown et al., 2001). Schafer et al. have also demonstrated increased numbers of pinosomes in cells expressing activated Arf6, as well as actin-rich foci on the ventral surface of the cells (Schafer et al., 2000). The pinosome-like structures often had actin-rich “tails” associated with them, and were highly motile. This motility was actin-dependent, required PLD activity and D-3 phosphoinositides, but was apparently not dependent on PI(4,5)P2 as microinjection of anti-PI(4,5)P2 antibodies did not impair movement. Activation of Arf6 in MDCK cells resulted in increased activation of Rac, formation of broad lamellipodia and the onset of motility (Santy and

16 298

CASANOVA

Casanova, 2001). In testing the role of PLD in these phenomena, we found, to our surprise, that induction of lamellipodia was dependent on PA production while the activation of Rac was not. In cells treated with low concentrations of 1-butanol (which substitutes for water in the hydrolysis of PC, leading to the production of biologically inert phosphatidylbutanol rather than PA), levels of active Rac remained high, yet lamellipodia did not form (Santy and Casanova, 2001). These findings suggest that PA is required for the dynamic reorganization of the actin cytoskeleton that is orchestrated by Rac; however, it is unclear whether it is PA itself, PI(4,5)P2, or one of their metabolites that is the necessary lipid component. Given the caveats described above, a combined approach using pharmacological inhibitors, dominant negative proteins and/or anti-sense inhibition of PLD and PIP5K isoforms will be necessary to resolve this question. Taken together, these results indicate that Arf6 plays an important role in the regulation of actin cytoskeleton dynamics. The mechanisms by which Arf6 modulates actin assembly are complex, and probably include effects on endosomal membrane traffic, crosstalk with Rho family GTPases and changes in local membrane lipid composition. Deciphering the contribution of each will be a difficult, but achievable goal.

5.

SUMMARY

In attempting to summarize the data presented above, one thing is abundantly clear: the function of Arf6 in eukaryotic cells is complex, highly variable between cell types, and incompletely understood. However, it is possible to extract some unifying principles. First, the ability of Arf6 to regulate post-endocytic recycling appears to be a common feature in all cells where it has been examined. The mechanisms by which this is accomplished remain mysterious, but may involve the assembly of a novel protein coat on endosomal membranes. Alternatively, or additionally, recycling may utilize components of the actin cytoskeleton whose function is modulated by Arf6. The ability of actin-depolymerizing drugs to inhibit recycling in HeLa cells (Radhakrishna and Donaldson, 1997) and adipocytes (Bose et al., 2001) suggests that this may be the case. The capacity of Arf6 to affect actin cytoskeleton dynamics is related in part to its interaction with GTPases of the Rho family. In the case of Rho itself, this interaction is likely to occur through a subfamily of Arf GAPs that also possess Rho GAP domains (Miura et al., 2002; Krugmann et al., 2002; Santy and Casanova, 2002). Crosstalk between Arf6 and Rac also occurs in all cells where it has been examined. Although the precise nature of the

ARF6

17 299

interaction remains unclear, it may also involve the participation of Arf GAPs or shared effectors, e.g., PLD or Arfaptin 2/POR1. Finally, it is likely that the functions of Arf6 in both membrane traffic and cytoskeleton dynamics can be attributed, at least in part, to its modulation of membrane lipid composition. In most cases where it has been examined, inhibition of PLD activity abrogates the downstream effects of Arf6 activation. Unfortunately, due to the lack of specific pharmacological inhibitors, the tool of choice for this purpose is often treatment of cells or membrane preparations with primary alcohols. While effective, this treatment does not distinguish between PLD isoforms and may have other nonspecific effects. The emerging use of catalytically inactive PLD mutants as dominant inhibitors should help to clarify the specific roles of PLD isoforms in Arf-dependent processes.

REFERENCES Altschuler, Y., Liu, S., Katz, L., Tang, K., Hardy, S., Brodsky, F., Apodaca, G., and Mostov, K. (1999). ADP-ribosylation factor 6 and endocytosis at the apical surface of MadinDarby canine kidney cells. J. Cell Biol., 147, 7-12. Amor, J. C., Horton, J. R., Zhu, X., Wang, Y., Sullards, C., Ringe, D., Cheng, X., and Kahn, R. A. (2001). Structures of yeast Arf2 and ARL1: distinct roles for the N terminus in the structure and function of Arf family GTPases. J. Biol. Chem., 276, 42477-42484. Bose, A., Cherniack, A. D., Langille, S. E., Nicoloro, S. M., Buxton, J. M., Park, J. G., Chawla, A., and Czech, M. P. (2001). G(alpha)11 signaling through Arf6 regulates F-actin mobilization and GLUT4 glucose transporter translocation to the plasma membrane. Mol. Cell Biol., 21, 5262-5275. Boshans, R. L., Szanto, S., van Aelst, L., and D'Souza-Schorey, C. (2000). ADP-ribosylation factor 6 regulates actin cytoskeleton remodeling in coordination with Rac1 and RhoA. Mol. Cell Biol., 20, 3685-3694. Brown, F. D., Rozelle, A. L., Yin, H. L., Balla, T., and Donaldson, J. G. (2001). Phosphatidylinositol 4,5-bisphosphate and Arf6-regulated membrane traffic. J. Cell Biol., 154, 1007-1018. Brown, F. D., Thompson, N., Saqib, K. M., Clark, J. M., Powner, D., Thompson, N. T., Solari, R., and Wakelam, M. J. (1998). Phospholipase D1 localizes to secretory granules and lysosomes and is plasma-membrane translocated on cellular stimulation. Curr. Biol., 8, 835-838. Caumont, A. S., Galas, M. C., Vitale, N., Aunis, D., and Bader, M. F. (1998). Regulated exocytosis in chromaffin cells. Translocation of Arf6 stimulates a plasma membraneassociated phospholipase D. J. Biol. Chem., 273, 1373-1379. Caumont, A. S., Vitale, N., Gensse, M., Galas, M. C., Casanova, J. E., and Bader, M. F. (2000). Identification of a plasma membrane-associated guanine nucleotide exchange factor for Arf6 in chromaffin cells. Possible role in the regulated exocytotic pathway. J. Biol. Chem., 275, 15637-15644.

18 300

CASANOVA

D'Souza-Schorey, C., Boshans, R. L., McDonough, M., Stahl, P. D., and Van Aelst, L. (1997). A role for POR1, a Rac1-interacting protein, in Arf6-mediated cytoskeletal rearrangements. EMBO J., 16, 5445-5454. D'Souza-Schorey, C., Li, G., Colombo, M. I., and Stahl, P. D. (1995). A regulatory role for Arf6 in receptor-mediated endocytosis. Science, 267, 1175-1178. D'Souza-Schorey, C., van Donselaar, E., Hsu, V. W., Yang, C., Stahl, P. D., and Peters, P. J. (1998). Arf6 targets recycling vesicles to the plasma membrane: insights from an ultrastructural investigation. J. Cell Biol., 140, 603-616. Del Pozo, M. A., Kiosses, W. B., Alderson, N. B., Meller, N., Hahn, K. M., and Schwartz, M. A. (2002). Integrins regulate GTP-Rac localized effector interactions through dissociation of Rho-GDI. Nat. Cell Biol., 4, 232-239. Franco, M., Peters, P. J., Boretto, J., van Donselaar, E., Neri, A., D'Souza-Schorey, C., and Chavrier, P. (1999). EFA6, a sec7 domain-containing exchange factor for Arf6, coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 14801491. Frank, S. R., Hatfield, J. C., and Casanova, J. E. (1998). Remodeling of the actin cytoskeleton is coordinately regulated by protein kinase C and the ADP-ribosylation factor nucleotide exchange factor ARNO. Mol. Biol. Cell, 9, 3133-3146. Gottlieb, T. A., Ivanov, I. E., Adesnik, M., and Sabatini, D. D. (1993). Actin microfilaments play a critical role in endocytosis at the apical but not the basolateral surface of polarized epithelial cells. J. Cell Biol., 120, 695-710. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase alpha is a downstream effector of the small G protein Arf6 in membrane ruffle formation. Cell, 99, 521-532. Jackson, C. L., and Casanova, J. E. (2000). Turning on Arf: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. Kanoh, H., Williger, B. T., and Exton, J. H. (1997). Arfaptin 1, a putative cytosolic target protein of ADP-ribosylation factor, is recruited to Golgi membranes. J. Biol. Chem., 272, 5421-5429. Krugmann, S., Anderson, K. E., Ridley, S. H., Risso, N., McGregor, A., Coadwell, J., Davidson, K., Eguinoa, A., Ellson, C. D., Lipp, P., Manifava, M., Ktistakis, N., Painter, G., Thuring, J. W., Cooper, M. A., Lim, Z. Y., Holmes, A. B., Dove, S. K., Michell, R. H., Grewal, A., Nazarian, A., Erdjument-Bromage, H., Tempst, P., Stephens, L. R., and Hawkins, P. T. (2002). Identification of ARAP3, a novel PI3K effector regulating both Arf and Rho GTPases, by selective capture on phosphoinositide affinity matrices. Mol. Cell, 9, 95-108. Langille, S. E., Patki, V., Klarlund, J. K., Buxton, J. M., Holik, J. J., Chawla, A., Corvera, S., Czech, M. P. (1999). ADP-ribosylation factor 6 as a target of guanine nucleotide exchange factor GRP1. J. Biol. Chem., 274, 27099-27104. Lawrence, J. T., and Birnbaum, M. J. (2001). ADP-ribosylation factor 6 delineates separate pathways used by endothelin 1 and insulin for stimulating glucose uptake in 3T3-L1 adipocytes. Mol. Cell Biol., 21, 5276-5285. Liscovitch, M., Czarny, M., Fiucci, G., and Tang, X. (2000). Phospholipase D: molecular and cell biology of a novel gene family. Biochem. J., 345, 401-415. Londono, I., Marshansky, V., Bourgoin, S., Vinay, P., and Bendayan, M. (1999). Expression and distribution of adenosine diphosphate-ribosylation factors in the rat kidney. Kidney Int., 55, 1407-1416.

ARF6

19 301

Manser, E., Loo, T. H., Koh, C. G., Zhao, Z. S., Chen, X. Q., Tan, L., Tan, I., Leung, T., and Lim, L. (1998). PAK kinases are directly coupled to the PIX family of nucleotide exchange factors. Mol. Cell, 1, 183-192. Maranda, B., Brown, D., Bourgoin, S., Casanova, J. E., Vinay, P., Ausiello, D. A., and Marshansky, V. (2001). Intra-endosomal pH-sensitive recruitment of the Arf-nucleotide exchange factor ARNO and Arf6 from cytoplasm to proximal tubule endosomes. J. Biol. Chem., 276, 18540-18550. Mayor, S., Presley, J. F., and Maxfield, F. R. (1993). Sorting of membrane components from endosomes and subsequent recycling to the cell surface occurs by a bulk flow process. J. Cell Biol., 121, 1257-1269. Menetrey, J., Macia, E., Pasqualato, S., Franco, M., and Cherfils, J. (2000). Structure of Arf6GDP suggests a basis for guanine nucleotide exchange factors specificity. Nat. Struct. Biol., 7, 466-469. Millar, C. A., Powell, K. A., Hickson, G. R., Bader, M. F., and Gould, G. W. (1999). Evidence for a role for ADP-ribosylation factor 6 in insulin-stimulated glucose transporter-4 (GLUT4) trafficking in 3T3-L1 adipocytes. J. Biol. Chem., 274, 1761917625. Miura, K., Jacques, K. M., Stauffer, S., Kubosaki, A., Zhu, K., Hirsch, D. S., Resau, J., Zheng, Y., and Randazzo, P. A. (2002). ARAP1: a point of convergence for Arf and Rho signaling. Mol. Cell, 9, 109-119. Mukherjee, S., Gurevich, V. V., Jones, J. C., Casanova, J. E., Frank, S. R., Maizels, E. T., Bader, M. F., Kahn, R. A., Palczewski, K., Aktories, K., and Hunzicker-Dunn, M. (2000). The ADP ribosylation factor nucleotide exchange factor ARNO promotes beta-arrestin release necessary for luteinizing hormone/choriogonadotropin receptor desensitization. Proc. Natl. Acad. Sci., USA, 97, 5901-5906. Palacios, F., Price, L., Schweitzer, J., Collard, J. G., and D'Souza-Schorey, C. (2001). An essential role for Arf6-regulated membrane traffic in adherens junction turnover and epithelial cell migration. EMBO J., 20, 4973-4986. Perry, S. J., and Lefkowitz, R. J. (2002). Arresting developments in heptahelical receptor signaling and regulation. Trends Cell Biol., 12, 130-138. Pessin, J. E., Thurmond, D. C., Elmendorf, J. S., Coker, K. J., and Okada, S. (1999). Molecular basis of insulin-stimulated GLUT4 vesicle trafficking. Location! Location! Location! J. Biol. Chem., 274, 2593-2596. Peters, P. J., Gao, M., Gaschet, J., Ambach, A., van Donselaar, E., Traverse, J. F., Bos, E., Wolffe, E. J., and Hsu, V. W. (2001). Characterization of coated vesicles that participate in endocytic recycling. Traffic, 2, 885-895. Peters, P. J., Hsu, V. W., Ooi, C. E., Finazzi, D., Teal, S. B., Oorschot, V., Donaldson, J. G., and Klausner, R. D. (1995). Overexpression of wild-type and mutant Arf1 and Arf6: distinct perturbations of non-overlapping membrane compartments. J. Cell Biol., 128, 1003-1017. Powner, D. J., Hodgkin, M. N., and Wakelam, M. J. (2002). Antigen-stimulated activation of phospholipase D1b by Rac1, Arf6, and PKC-alpha in RBL-2H3 Cells. Mol. Biol. Cell, 13, 1252-12562. Radhakrishna, H., Al-Awar, O., Khachikian, Z., and Donaldson, J. G. (1999). Arf6 requirement for Rac ruffling suggests a role for membrane trafficking in cortical actin rearrangements. J. Cell Sci., 112, 855-866. Radhakrishna, H., and Donaldson, J. G. (1997). ADP-ribosylation factor 6 regulates a novel plasma membrane recycling pathway. J. Cell Biol., 139, 49-61.

20 302

CASANOVA

Radhakrishna, H., Klausner, R. D., and Donaldson, J. G. (1996). Aluminum fluoride stimulates surface protrusions in cells overexpressing the Arf6 GTPase. J. Cell Biol., 134, 935-947. Ridley, A. J., Allen, W. E., Peppelenbosch, M., and Jones, G. E. (1999). Rho family proteins and cell migration. Biochem. Soc. Symp., 65, 111-q23. Santy, L. C., and Casanova, J. E. (2001). Activation of Arf6 by ARNO stimulates epithelial cell migration through downstream activation of both Rac1 and phospholipase D. J. Cell Biol., 154, 599-610. Santy, L. C., and Casanova, J. E. (2002). GTPase signaling: bridging the GAP between Arf and Rho. Curr. Biol., 12, R360-2. Schafer, D. A., D'Souza-Schorey, C., and Cooper, J. A. (2000). Actin assembly at Membranes controlled by Arf6. Traffic, 1, 896-907. Seaman, M. N., Sowerby, P. J., and Robinson, M. S. (1996). Cytosolic and membraneassociated proteins involved in the recruitment of AP-1 adaptors onto the trans-Golgi network. J. Biol. Chem., 271, 25446-25451. Skippen, A., Jones, D. H., Morgan, C. P., Li, M., and Cockcroft, S. (2002). Mechanism of ADP ribosylation factor-stimulated phosphatidylinositol 4,5-bisphosphate synthesis in HL60 cells. J. Biol. Chem., 277, 5823-5831. Song, J., Khachikian, Z., Radhakrishna, H., and Donaldson, J. G. (1998). Localization of endogenous Arf6 to sites of cortical actin rearrangement and involvement of Arf6 in cell spreading. J. Cell Sci., 111, 2257-67. Stamnes M. A, and Rothman J. E. (1993). The binding of AP-1 clathrin adaptor particles to Golgi membranes requires ADP-ribosylation factor, a small GTP-binding protein. Cell, 73, 999-1005. Terui, T., Kahn, R. A., and Randazzo, P. A. (1994). Effects of acid phospholipids on nucleotide exchange properties of ADP-ribosylation factor 1. Evidence for specific interaction with phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem., 269, 28130-28135. Traub, L. M., Ostrom, J. A., and Kornfeld, S. (1993). Biochemical dissection of AP-1 recruitment onto Golgi membranes. J. Cell Biol., 123, 561-573. Tsuchiya, M., Price, S. R., Tsai, S. C., Moss, J., and Vaughan, M. (1991). Molecular identification of ADP-ribosylation factor mRNAs and their expression in mammalian cells. J. Biol. Chem., 266, 2772-2777. Van Aelst, L., and D'Souza-Schorey, C. (1997). Rho GTPases and signaling networks. Genes Dev., 11, 2295-2322. Venkateswarlu, K., Oatey, P. B., Tavare, J. M., and Cullen, P. J. (1998). Insulin-dependent translocation of ARNO to the plasma membrane of adipocytes requires phosphatidylinositol 3-kinase. Curr. Biol., 8, 463-466. Vitale, N., Caumont, A. S., Chasserot-Golaz, S., Du, G., Wu, S., Sciorra, V. A., Morris, A. J., Frohman, M. A., and Bader, M. F. (2001). Phospholipase D1: a key factor for the exocytotic machinery in neuroendocrine cells. EMBO J., 20, 2424-2434. Yang, C. Z., and Mueckler, M. (1999). ADP-ribosylation factor 6 (Arf6) defines two insulinregulated secretory pathways in adipocytes. J. Biol. Chem., 274, 25297-25300. Zhang, C. J., Rosenwald, A. G., Willingham, M. C., Skuntz, S., Clark, J., and Kahn, R. A. (1994). Expression of a dominant allele of human Arf1 inhibits membrane traffic in vivo. J. Cell Biol., 124, 289-300. Zhang, Q., Calafat, J., Janssen, H., and Greenberg, S. (1999). Arf6 is required for growth factor- and rac-mediated membrane ruffling in macrophages at a stage distal to rac membrane targeting. Mol. Cell Biol., 19, 8158-8168.

ARF6

21 303

Zhang, Q., Cox, D., Tseng, C. C., Donaldson, J. G., and Greenberg, S. (1998). A requirement for Arf6 in Fc-gamma receptor-mediated phagocytosis in macrophages. J. Biol. Chem., 273, 19977-19981.

Chapter 15 A ROLE FOR ARF6 IN G PROTEIN-COUPLED RECEPTOR DESENSITIZATION MARY HUNZICKER-DUNN Department of Cell and Molecular Biology, Northwestern University Medical Schoo1, Chicago, USA

Abstract:

1.

In addition to its roles in changes in the actin cytoskeleton and lipid metabolism at the plasma membrane, recent results suggest that Arf6 also participates in desensitization and internalization of G-protein-coupled receptors. The evidence to support this novel function for Arf6 is reviewed in this chapter.

INTRODUCTION

The superfamily of seven membrane-spanning receptors that can couple to heterotrimeric guanine nucleotide binding (G) proteins responds to a multitude of extracellular signals. These G-protein-coupled receptors (GPCRs) in turn regulate a variety of cellular responses that reflect their diverse group of specific agonists, including muscle contractility, ovulation, hormone secretion, gene expression, vision, taste, and neuronal excitation. All GPCRs exhibit conserved amino acid residues that comprise seven transmembrane (TM) alpha helices connected by variable extracellular and intracellular loops. In the absence of agonist, GPCRs exist in an inactive conformation, based on the interactions of key amino acid residues contained in highly conserved domains. Upon binding agonist, these interactions are altered and the receptors assume an activated conformation in which they catalyze the rate-limiting release of GDP from heterotrimeric G proteins. G protein activation then drives increased effector activity, resulting, for example, in increased formation of second messengers such as cAMP, inositol trisphosphate and diacylglycerol, and opening or closing of ion channels. 305

2 306

HUNZICKER-DUNN

Coupling of GPCRs to G protein(s) is regulated both by the rates of association and dissociation of receptor ligands as well as by an apparently common receptor desensitization mechanism. Desensitization functions to uncouple agonist-bound receptor from its G protein and is mediated by a family of proteins known as Arrestins. The two models of GPCR desensitization that have been most intensely investigated are light regulation of rhodopsin and isoproterenol regulation of the β2-adrenergic receptor (β2-AR). It is well established that GPCR desensitization is triggered by agonist-dependent receptor phosphorylation catalyzed by a family of G protein receptor kinases (GRKs). Arrestin then binds with high affinity to the active, phosphoreceptor. Arrestin binding to receptor interrupts the ability of the receptor to activate G proteins. Recent results reveal an unexpected function for Arf6 in this sequence of events leading to homologous (agonist-specific) receptor desensitization. Utilizing the luteinizing hormone receptor (LHR) as a prototypical GPCR in a model in which only endogenous proteins are present in their native environment, we have shown that Arf6 functions to sequester a pool of Arrestin that is obligatory for LHR desensitization. Agonist activated receptor not only activates its heterotrimeric G protein but also activates Arf6, resulting in the undocking of Arrestin so that it is now available to bind to this agonist-bound receptor. In this chapter we will review the results that led to these conclusions and attempt to integrate our data obtained using a plasma membrane model with the bulk of the literature in which the Arrestins and GPCRs are overexpressed in intact cellular models.

2.

CLASSIC MODEL FOR GPCR DESENSITIZATION

With more than 1,000 GPCRs and less than 20 G proteins, there is a clear need for regulation of the temporal aspects of G protein mediated signaling or simply alternative mechanisms of signal termination. Homologous desensitization is the phenomenon in which exposure to agonist produces a state in which reduced responses are observed upon subsequent challenge by agonist. This type of desensitization can result from covalent modifications to the receptor (e.g., phosphorylation), changes in the binding partners of the receptor (e.g., Arrestins), or changes in the location of the receptor (internalization). Most agonist-bound GPCRs require GRK-catalyzed phosphorylation of the receptor in order to achieve high-affinity Arrestin binding (Krupnick and Benovic, 1998). The family of Arrestins consists of the visual Arrestins (Arrestin1 in rod cells and Arrestin4 in cone cells) and the ubiquitous

ROLE OF ARF6 IN GPCR DESENSITIZATION

3073

Arrestin2 (β-Arrestin1) and Arrestin3 (β-Arrestin2) (Palczewski, 1994). Arrestin is converted into an active conformation that binds GPCRs with high affinity (in the pM range) upon engagement of two binding sites: one site that recognizes the active conformation of the receptor and a second site that interacts directly with GRK-phosphorylated residues on the receptor (Gurevich and Benovic, 1993; Gurevich et al., 1995; Kovoor, 1999; Celver et al., 2002). The basic requirements for GRK-catalyzed receptor phosphorylation and the obligatory binding of Arrestin molecules in GPCR desensitization were established in studies using a reconstituted model consisting of phospholipid vesicles, receptor agonist, and recombinant β2AR, GRK and Arrestin (Lohse et al., 1990; Lohse et al., 1992). The validity of this paradigm for GPCR desensitization has now been established for a large number of GPCRs using intact cell models in which receptors, GRKs, and (often tagged) Arrestins are co-transfected and expressed at artificially high levels (Krupnick and Benovic, 1998; Ferguson, 2001; Claing et al., 2002). Because Arrestins also target the receptor for translocation to an intracellular compartment (internalization), based on their ability to bind both clathrin and β2-adaptin of the AP-2 protein complex (Goodman et al., 1996; Laporte et al., 1999; Laporte et al., 2000), receptor internalization is often assayed as an indicator of desensitization. However, desensitization and internalization of GPCRs are not necessarily linked events. For example, there is evidence for some GPCRs that desensitization and internalization are mediated by the binding of distinct Arrestins to distinct receptor sites (Cheng et al., 1998; Kohout et al., 2001; Paing et al., 2002; Mukherjee et al., 2002). In such cases desensitization and internalization cannot be sequential events in a linear pathway. There is also evidence that internalization of some GPCRs is Arrestin-independent (Pals-Rylaarsdam et al., 1995; Bhatnagar et al., 2001) while, to date, GPCR desensitization is Arrestin-dependent. However, despite these caveats in data interpretation, using either receptor desensitization or internalization as the endpoint, overexpression of GRKs and/or Arrestins drives desensitization or internalization of nearly all GPCRs that have been investigated (Krupnick and Benovic, 1998). Indeed, the ability of GRKs and Arrestins to promote desensitization or internalization has often been used as evidence that a specific GPCR requires receptor phosphorylation and Arrestin binding to achieve desensitization or internalization (Carmen and Benovic, 1999). Upon receptor activation, overexpressed, tagged Arrestins are uniformly recruited from a cytosolic compartment to the plasma membrane by an unidentified mechanism (Ferguson et al., 1996). The recruitment of Arrestins to the membrane has even been developed as an assay to detect receptor activation (Barak et al., 1997) as well as to estimate the affinity of

4 308

HUNZICKER-DUNN

Arrestin’s interaction with agonist-bound receptor (Oakley et al., 2000). Consistent with the fundamental role of Arrestin binding for GPCR desensitization, Arrestin2 knockout mice show impaired β2-AR desensitization (Conner et al., 1997) and Arrestin3 knockout mice show prolonged responsiveness to analgesics like morphine (Bohn et al., 2000). Mouse embryonic fibroblasts from Arrestin2/Arrestin3 knockout mice also show impaired desensitization of the angiotensin II type 1A receptor (Kohout et al., 2001). Thus, the levels of Arrestins and their regulated interactions with GPCRs clearly play a role in GPCR signaling.

3.

DEVELOPMENT OF A NEW MODEL FOR LHR DESENSITIZATION

3.1

Arrestin2 is stably bound to plasma membranes

Desensitization of some GPCRs, including the LHR, has been demonstrated under cell-free conditions utilizing a receptor/effector-enriched plasma membrane model (Bockaert et al., 1976; Roy et al., 1976; Anderson and Jaworski, 1979; Ezra and Salomon, 1980; Ezra and Salomon, 1981). This assay involves a two-stage incubation: in stage 1, membranes are preincubated without or with agonist to promote receptor desensitization; in stage 2, the ability of receptor in the absence or presence of agonist to couple to the stimulatory G protein (Gs) is evaluated using adenylyl cyclase activity. Diminished adenylyl cyclase activity upon restimulation with agonist in stage 2 results from receptor desensitization. Under the appropriate in vitro incubation conditions, 80-100% desensitization of LHR is achieved with a time-course equivalent to that seen in the intact animal (Marsh et al., 1973; Hunzicker-Dunn, 1981; Ekstrom and Hunzicker-Dunn, 1990). These results challenge the conclusion that Arrestins must be recruited from a cytosolic compartment to effect receptor desensitization in all cases. For a GPCR to be desensitized in a phosphorylation- and Arrestindependent manner in this cell-free model, both the Arrestin and the GRK must be present in the plasma membrane-enriched fraction. As the high affinity binding of Arrestin to a GPCR is the crucial step that uncouples receptor from its cognate G protein, this model leads to the prediction of a pool of Arrestin, localized to the membrane fraction used in the two step assay. Arrestin2 is readily detected by western blotting in the membrane fraction isolated from ovarian follicles and available to the LHR to mediate desensitization (Mukherjee et al., 1999a). Similarly, subcellular

ROLE OF ARF6 IN GPCR DESENSITIZATION

3095

fractionation of other cells has revealed detectable levels of immunoreactive, endogenous Arrestins in membrane fractions (Bennett et al., 2000; Santini et al., 2000). The inclusion of Arrestin antibodies that block the ability of Arrestin to interact with GPCRs (Knospe et al., 1988; Palczewski, 1994) completely abrogated agonist-dependent LHR desensitization in follicular membranes (Mukherjee et al., 1999a). This result shows that the endogenous Arrestin2 present in this LHR-enriched membrane fraction is necessary and sufficient for LHR desensitization. Addition to the membrane fraction of either visual Arrestin1 or Arrestin2 promoted desensitization of active LHR in a dose-dependent manner, while Arrestin3 did not (Mukherjee et al., 1999a; Mukherjee et al., 2002). Moreover, the concentration of Arrestin2 that promoted 50% desensitization of LHR activity was in the subnM range, consistent with the probable high affinity binding of Arrestin2 to the LHR. In order to identify the Arrestin2 binding sites on the receptor that promoted LHR desensitization, we first determined whether synthetic peptides made to various regions of the receptor could compete with receptor for Arrestin2 and block LHR desensitization. Only peptides corresponding to the third intracellular (3i) loop, and not those corresponding to a scrambled 3i peptide, the 2i loop, or an N-terminal portion of the cytoplasmic tail, abrogated LHR desensitization (Mukherjee et al., 1999b). The blockade by the 3i loop peptide was rescued by addition of recombinant Arrestin2, indicating that the 3i peptide blocked LHR desensitization by competing with receptor for binding of Arrestin2. Arrestin2 bound directly to the 3i peptide with pM affinity while Arrestin3 bound with only mM affinity (Mukherjee et al., 2002). Asp-564 in the 3i loop, which interacts with Arg-464 at the base of TM 3 to maintain the receptor in an inactive conformation (Ballesteros et al., 2001; Shenker, 2002), is required for Arrestin2 to bind to the agonist-bound receptor (Mukherjee et al., 2002). Even substitution with a negatively charged amino acid (Glu) did not support Arrestin2 binding to the receptor, suggesting that the specific geometry of Asp-564 at this site in the 3i loop is required for Arrestin2 binding to the LHR. In contrast to many GPCRs, Arrestin2 binds to the LHR without the requirement for receptor phosphorylation (Ekstrom and Hunzicker-Dunn, 1989; Ekstrom et al., 1992; Zhu et al., 1993; Lamm and Hunzicker-Dunn, 1994; Lamm et al., 1994; Wang et al., 1997). This high affinity binding appears to be mediated, in part, by Asp-564 in the 3i loop of the receptor, although the negative charge of this residue is not simply mimicking the negatively charged phosphate group of phosphorylated GPCRs (Mukherjee et al., 2002). The homologous Asp residue is conserved in the glycoprotein hormone and cannabinoid receptors but is replaced with a negatively

6 310

HUNZICKER-DUNN

charged Glu in most other receptors (Schulz et al., 2000). Although the other glycoprotein hormone receptors are believed to require GRK-catalyzed phosphorylation to bind Arrestins for desensitization (Iacovelli et al., 1996; Troispoux et al., 2000), it is possible that the presence of this Asp residue in their 3i loops, as with the LHR, may allow them to bind Arrestin for desensitization in a phosphorylation-independent manner. For Arrestin2 to promote LHR desensitization it should bind only to active and not to inactive receptor. Moreover, the Arrestin2 binding site at Asp-564 in the 3i loop is not expected to be accessible in the inactive receptor conformation. Under conditions in which neither receptor nor Arrestin is overexpressed, Arrestin2 preferentially co-immunoprecipitated with agonist-bound LHR (Mukherjee et al., 1999a; Mukherjee et al., 2000). Because the presence of Arrestin2 in LHR-enriched membranes is independent of agonist and interacts only with agonist-bound LHR, Arrestin2 must be bound at a site distinct from the LHR, when the receptor is not active (as shown in Figure 1, panel A). Moreover, there must be a mechanism that stimulates the “undocking” of Arrestin2 so it is available to bind to agonist-bound receptor. A clue to the regulation of the Arrestin2 undocking mechanism was revealed by the demonstration that the pool of Arrestin2 present in the plasma membrane-enriched fraction requires GTP to be released and effect the desensitization of LHRs.

3.2

LHR desensitization is GTP-dependent

That GPCR desensitization might be more complex and require proteins in addition to the GRKs and Arrestins was suggested by studies in the early 1980’s by Ezra and Salomon. They used a cell-free model to demonstrate a GTP requirement for desensitization of the LHR (Ezra and Salomon, 1980; Ezra and Salomon, 1981). Similar reports soon followed for the β2-AR and vasopressin receptor in kidney cells (Roy et al., 1976; Anderson and Jaworski, 1979). Due to the relatively high concentrations of GTP in intact cells, the GTP requirement for GPCR desensitization could not be investigated in an intact cell model. This GTP requirement for GPCR desensitization could theoretically be fulfilled by the activation of a heterotrimeric G protein, possibly to recruit GRKs via released βγ subunits (Krupnick and Benovic, 1998). However, as LHR desensitization is independent of receptor phosphorylation (as reviewed above), βγ-dependent GRK recruitment cannot be a prerequisite for LHR desensitization. A requirement for heterotrimeric G protein activation in LHR desensitization

ROLE OF ARF6 IN GPCR DESENSITIZATION

3117

Figure 1. Proposed model depicting LHR desensitization: the LHR promotes the activation of ARF6, leading to cessation of LHR coupling to Gs and adenylyl cyclase. (A) The inactive receptor and key proteins detected by immunoblot or activity assays in ovarian follicle plasma membrane are shown. (B) Agonist binding promotes a conformational change in the LHR, resulting in activation of Gs and adenylyl cyclase (AC). Dashed arrows represent the directed movements of ARNO to bind the NPXXY motif in TM7 of the LHR with consequent activation of ARF6. This, in turn, releases Arrestin2, allowing its binding to the 3i loop of the LHR and effecting desensitization. The large dot in the 3i peptide represents Asp-564. (C) The desensitized LHR is shown with Arrestin2 bound to the 3i loop, ARNO bound to TM 7, and Gs uncoupled and returned to its basal state.

8 312

HUNZICKER-DUNN

was ruled out by the inability of C-terminal Gα peptides, Gα antibodies, or βγ internalization to affect LHR desensitization (Rajagopalan-Gupta et al., 1999). Interestingly, LHR desensitization could be completely reversed in the membrane model by incubating desensitized LHR with GDPβS, resulting in re-coupling of the receptor to Gs and adenylyl cyclase (Ekstrom et al., 1992). More recent studies have led to the identification of the site of action of GTP that is required for desensitization in in vitro systems.

3.3

Arf6 sequesters Arrestin2

Recombinant Arrestin2 promoted desensitization of the active LHR both in the presence and absence of GTP (Mukherjee et al., 2000), so the GTPdependent step of LHR desensitization could not be the binding of Arrestin2 to the receptor. To determine whether the GTP-dependent step consisted of the undocking of Arrestin2 from its membrane binding site, membranes were preincubated with water or 100 µM GTP, diluted ~25-fold, then repelleted and analyzed. Preincubation of membranes with 100 µM GTP (in the absence of receptor agonist) resulted in release from the membranes of more than 80% of the bound Arrestin2 (Mukherjee et al., 2000), which was detectable in the wash buffer (S. Mukherjee, unpublished observation). Insufficient Arrestin2 was retained in the membranes to promote LHR desensitization (Mukherjee et al., 2000). These results demonstrate that the GTP-dependent component of LHR desensitization is the release of Arrestin2 from its docking site in the membrane. The identity of the GTP binding protein responsible for release of Arrestin2 from membranes was sought. We turned to the large family of monomeric GTPases as candidates for the regulator of the availability of Arrestin2 for LHR desensitization. Members of the Rho, Ras, Rac and Rap families of GTPases, were discounted based on evidence that the active fragments of the large Clostridial toxins, C. difficile toxin B, and C. sordelli lethal toxin, did not affect LHR desensitization (Mukherjee et al., 2000). Arfs 1-5 were also ruled out based on the inability of brefeldin A, which inhibits many Arf guanine nucleotide exchange factors (GEFs), to block LHR desensitization (Mukherjee et al., 2000). However, the Arf GEFs which have been identified as Arf6 activators are resistant to brefeldin A (Jackson and Casanova, 2000), and Arf6 is often localized to plasma membrane-enriched cellular fractions (Cavenagh et al., 1996). To investigate a possible role for Arf6 in LHR desensitization, we utilized a synthetic peptide corresponding to the Nterminus of Arf6, which has been shown to block Arf6 activation (Kahn et al., 1992; Lenhard et al., 1992; Galas et al., 1997; Caumont et al., 1998). The corresponding N-terminal Arf1 peptide was used as a control. The N-

ROLE OF ARF6 IN GPCR DESENSITIZATION

3139

terminal Arf6 peptide abrogated LHR desensitization while the corresponding Arf1 peptide was ineffective (Mukherjee et al., 2000). Moreover, the Arf6 peptide blocked the GTP-dependent release of Arrestin2 (Mukherjee et al., 2000). These results suggest that Arf6•GDP sequesters Arrestin2, either directly or indirectly, and that upon activation, Arrestin2 is released from its membrane docking site, as shown in Figure 1, panel B.

3.4

ARNO promotes Arrestin2 undocking and LHR desensitization

Consistent with an obligatory role for Arf6 activation in liberating Arrestin2 for receptor desensitization, addition of the Arf1/6 activator ARNO (see Chapters 2, 4-6) to plasma membranes decreased the concentration of GTP required for release of Arrestin 2 or for receptor desensitization from 100 µM to 1 µM. Neither catalytically inactive ARNO-K156 (Frank et al., 1998) nor a catalytically active ARNO containing a mutation in its pleckstrin homology domain (which interferes with its binding of phosphoinositides (Nagel et al., 1998)) promoted LHR desensitization (Mukherjee et al., 2000). Moreover, LHR agonist-dependent Arrestin2 undocking and consequent LHR desensitization were blocked when membranes were incubated with catalytically inactive ARNO-K156 (Mukherjee et al., 2001). These results suggest that the exogenous, catalytically inactive ARNO mutant competes with endogenous ARNO, or a similar Arf6 activator, to block Arf6 activation Arrestin2 release from its membrane docking site. In support of a primary role for Arf6 in regulating the availability of Arrestin2 for LHR desensitization, Arf6 is readily detected in a Triton X100-insoluble fraction in plasma membrane preparations from ovarian follicles by western blotting (Mukherjee et al., 2000; Salvador et al., 2001) and is activated upon LHR activation (Salvador et al., 2001). Moreover, Arf6 activation by LHR is significantly reduced in the presence of ARNOK156 (Salvador et al., 2001). Additional evidence in support of a linear pathway from the LHR to regulate the activation of Arf6 is provided by recent evidence that ARNO binds directly to the conserved NPXXY motif located at the base of TM 7 of the LHR (personal observation). This conclusion is based on results showing that ARNO binds to a synthetic peptide containing this conserved amino acid sequence and not to one in which the amino acid residues are scrambled. ARNO also preferentially co-immunoprecipitates with the agonist-activated form of the receptor. Moreover, this TM 7 peptide promotes Arrestin2 release from its membrane docking site in the absence of LHR agonist, and supports LHR desensitization in the presence of agonist. LHR desensitization stimulated by this TM 7 peptide is blocked by

10 314

HUNZICKER-DUNN

catalytically inactive ARNO-K156 and by the N-terminal Arf6 peptide, indicating that the TM 7 peptide directs activation of ARNO, or a similar GEF, and consequently Arf6.

3.5

Model for LHR desensitization

Taken together, these results indicate that Arf6•GDP, present in the receptor/effector-enriched follicular membrane fraction, sequesters Arrestin2 either directly or indirectly (see Figure 1, panel A). Agonist-dependent LHR activation results in movement of TM 3 and TM 6 and the consequent shifting of the 3i loop into a more exposed position (Ballesteros et al., 2001), resulting in activation of the heterotrimeric Gs and consequent activation of adenylyl cyclase (see Figure 1, panel B). Gαs appears to interact with the 3i loop of the LHR (Abell and Segaloff, 1997). LHR activation may also be associated with the movement of the cytoplasmic end of TM 7 (Abdulaev and Ridge, 1998; Altenbach et al., 1999; Shenker, 2002), resulting in exposure of the highly conserved NPXXY motif. ARNO, also present in the membrane preparation (Mukherjee et al., 2001), binds to the region of the LHR at the base of TM 7 containing the NPXXY motif and promotes activation of Arf6, resulting in the undocking of Arrestin2 (panel B). Arrestin2 then binds to the agonist-bound LHR at the 3i loop, resulting in uncoupling of the receptor from Gs and reduced production of cAMP (panel C). As indicated in panel B, there is limited time period (~8-10 min in our membrane model) after agonist addition when the LHR is actively signaling to adenylyl cyclase (i.e., prior to desensitization). Addition of exogenous, recombinant Arrestin2 to active receptor (see panel B, when endogenous Arrestin2 is still docked with Arf6) likely abbreviates this time course and promotes desensitization. Similarly, addition of exogenous recombinant ARNO to active receptor likely stimulates desensitization by accelerating Arf6 activation, resulting in the undocking of endogenous Arrestin2. The apparent rate-limiting step in LHR desensitization is the relatively slow binding of ARNO to the LHR. Neither the events involved in ARNO binding to the LHR nor the mechanism by which activated Arf6 promotes the release of Arrestin2 is understood.

4.

ARRESTIN SEQUESTRATION: A WIDESPREAD FUNCTION OF ARF6?

The principal unique feature of LHR desensitization that we have uncovered is the finding that Arrestin2 is sequestered by Arf6•GDP and liberated with

ROLE OF ARF6 IN GPCR DESENSITIZATION

11 315

Arf6 activation. Arrestin2 is released from its membrane docking site only in response to (a) Arf6 activation by ARNO, in the presence of 1 µM GTP, (b) high concentrations of GTP that override the GEF requirement and activate all GTPases, or (c) LHR activation by agonist. Thus, the release of Arrestin2 is regulated and specific for reagents that activate Arf6. Moreover, sufficient Arrestin2 is retained in a membrane fraction to achieve complete LHR desensitization under cell-free conditions. These results suggest that the Arrestin available for receptor desensitization in cells may be limiting under some circumstances. Indeed, there is evidence to support this hypothesis in intact cells. In COS-7 cells in which a constitutively active parathyroid hormone receptor was expressed and coupled to Gs, the overexpression of Arrestin3 was required to reduce cAMP production by this receptor (Ferrari and Bisello, 2001). Even more compelling is the report that internalization of the β2-AR in stably transfected HEK293 cells was reduced >65% by activation of the V2 vasopressin receptor coexpressed in the same cells, an effect that was reversed by overexpression of Arrestin2 and Arrestin3 (Klein et al., 2001). Does the Arrestin2 that is not localized at the plasma membrane by Arf6 participate in LHR desensitization in intact cells? We do not know. Subcellular fractionation of ovarian granulosa cells revealed that the majority of immunoreactive Arrestin2, in contrast to Arf6 (Mukherjee et al., 2000), was detected in a soluble fraction (S. Mukherjee, unpublished observation). Perhaps this soluble Arrestin has additional functions. For example, Arrestin2 binds to cytosolic Disheveled proteins, which lie downstream of the Frizzled receptor in the Wnt pathway (Chen et al., 2001). It is also not known if the Arrestin present in the soluble fraction of untransfected cells is docked at specific cellular locations, perhaps bound (directly or indirectly) to other Arfs, or if it is diffusely distributed throughout the cell. In one of the few studies in which the distribution of endogenous Arrestins was evaluated by immunofluorescence, both nonvisual Arrestins were found distributed in a punctate pattern throughout the interior of RBL-2H3 cells (a rat mast cell line) (Santini et al., 2000), consistent with the existence of specific Arrestin binding sites. Does Arf6 sequester Arrestin in cells other than ovarian follicular cells? And does GPCR activation universally promote the undocking of Arrestin from an Arf6-dependent binding site, or is this pathway unique to the ovary and the LHR? We do not yet know the complete answer to these questions. However, the findings that LHR artificially expressed in kidney cells can activate a pathway that leads to Arf6 activation and Arrestin2-dependent LHR desensitization, shows that the GPCR-dependent Arf6 activation pathway to undock Arrestin is not restricted to ovarian cells (Mukherjee et al., 2002). Additional studies are required to ascertain whether or not other

12 316

HUNZICKER-DUNN

GPCRs use this GTP-dependent pathway to undock Arrestins and if so, whether Arf6 universally fulfills this role. However, in cells expressing recombinant receptors, GRKs, and Arrestins, the hallmark of GPCR action is the recruitment of Arrestin from a cytosolic fraction to the receptor (Ferguson et al., 1996; Barak et al., 1999; Evans et al., 2001). Does this result simply reflect the exposure of high affinity binding sites for Arrestins on activated receptors and resulting movement of Arrestin to these localized high-affinity binding sites, and does it represent the normal cellular response to receptor activation? Additional studies are required to resolve these important questions. Perhaps in the native cellular environment, the two non-visual Arrestins can be recruited to receptors by distinct mechanisms. While Arrestin2 is recruited from its plasma membrane-localized docking site in response to LHR-stimulated Arf6 activation for desensitization, Arrestin3 is not detected in follicular membranes (Mukherjee et al., 1999a) and therefore is not sequestered by Arf6 at this location. Rather, Arrestin3 appears to promote LHR internalization (Min et al., 2001) and must be recruited from another cellular location to perform this function. Using a cell line that expresses relatively high and equivalent levels of endogenous Arrestin2 and Arrestin3 and in which β2-AR was stably expressed (at ~80,000 receptors/cell), Arrestin3 was selectively translocated from a soluble compartment to clathrin-coated pits in response to agonist stimulation of the β2-ARs (Santini et al., 2000). This result indicates that under conditions in which Arrestins are not overexpressed, Arrestin3 is still recruited from the cytosol to the receptor. In a number of cell types, Arf6 is localized not only to the plasma membrane but also to a tubulovesicular endosomal membrane compartment (Peters et al., 1995; Radhakrishna and Donaldson, 1997; D'Souza-Schorey et al., 1998; Frank et al., 1998; Peters et al., 2001; Brown et al., 2001) which could co-fractionate with the plasma membrane even on sucrose gradient centrifugation (Radhakrishna and Donaldson, 1997). We therefore cannot rule out the possibility that in intact granulosa cells, Arrestin2 along with Arf6 and possibly ARNO, are recruited to the plasma membrane from this tubulovesicular compartment. However, immunofluorescence results with rat granulosa cells indicate that the majority of Arf6 is localized at the plasma membrane both in the absence and presence of receptor agonist (Mukherjee et al., 2000). Finally, there is recent evidence for the β2-AR and possibly other GPCRs that ARNO and Arf6 are involved in receptor internalization. This conclusion was based on studies by Premont and collaborators, as discussed in Chapter 8, in which they demonstrated that overexpression of the GRK interacting protein GIT1, which possesses Arf GTPase activating protein (GAP) activity, reduced β2-AR internalization (Premont et al., 1998). This effect of GIT1 was dependent on its GAP activity (Premont et al., 1998).

ROLE OF ARF6 IN GPCR DESENSITIZATION

13 317

GIT1 overexpression reduced the internalization only of those GPCRs that were internalized in an Arrestin-dependent manner and not those internalized by an Arrestin-independent pathway (Bhatnagar et al., 2001). That overexpression of GIT1 inhibits Arrestin-dependent GPCR internalization suggests that Arf•GTP is required for GPCR internalization. Subsequent studies by this group showed that transient overexpression of ARNO, in cells overexpressing β2-AR, accelerated receptor internalization; although inexplicably, overexpression of either constitutively active Arf6L67 or dominant negative Arf6-N27 reduced β2-AR internalization (Claing et al., 2001). Interestingly, these authors also showed that Arrestin2 and Arrestin3 interacted directly with ARNO and Arf6•GDP but not with Arf6•GTP in GST pull-down assays. When Arrestins, Arf6, ARNO, and β2AR were overexpressed in HEK293 cells and proteins were cross-linked, Arf6 and Arrestins co-immunoprecipitated in an agonist-dependent manner while Arrestins and ARNO were constitutively associated. These authors concluded that with Arrestin/ARNO recruitment to the agonist-activated and phosphorylated β2-AR, Arrestin functions both to recruit Arf6•GDP and as a cofactor to enhance the activity of ARNO to activate Arf6. Arf6•GTP was predicted to dissociate from the Arrestin/receptor complex and to participate in recruitment of vesicle coat proteins leading to β2-AR endocytosis (Claing et al., 2001) in a manner analogous to that of Arf1 and Arf5 in Golgi membranes where they stimulate coat assembly by binding AP-1 (Zhu et al., 1999; Drake et al., 2000). However, Arf6 is not generally associated with clathrin-coated vesicles (Peters et al., 1995; Cavenagh et al., 1996; Radhakrishna et al., 1996; Radhakrishna and Donaldson, 1997; Brown et al., 2001) and recent studies do not support an association of Arf6 and AP-2 (Austin et al., 2002). Arf6 could stimulate clathrin coated vesicle endocytosis at the plasma membrane indirectly, e.g., as a result of changes in lipid metabolism (Honda et al., 1999; Godi et al., 1999). Alternatively, the stimulatory effects of ARNO and inhibitory effects of GIT1 on β2-AR internalization could be interpreted to support our model (see Figure 1) in which ARNO stimulates Arf6 activation to liberate Arrestin to promote receptor desensitization and, for some GPCRs, receptor internalization, while GIT1-driven Arf6•GDP sequesters Arrestin, thereby inhibiting receptor internalization.

5.

SUMMARY

Taken together these data support a novel function for Arf6 in Arrestindependent GPCR desensitization and internalization. The key features of the role of Arf6 in receptor desensitization are as follows. In the resting state

14 318

HUNZICKER-DUNN

there exists a pool of Arrestin that is bound to Arf6•GDP. Upon ligand binding, receptors assume an activated conformation, couple to G proteins, and recruit ARNO. The ARNO then activates Arf6, resulting in release of Arrestins, which in turn bind the receptor and promote uncoupling and internalization. This model is based largely on results from follicular cells and LRH but Arf6•GDP appears to fulfill a similar function in human embryonic kidney cells (Mukherjee et al., 2002). As a result, signaling of the LHR to Gs is halted, and perhaps signaling to other pathways is then initiated (Pierce et al., 2001). However, because the domain to which ARNO binds on the LHR is highly conserved among the superfamily of GPCRs, it is tempting to speculate that ARNO binds to the agonist-bound conformation of other GPCRs to activate Arf6 and regulate the availability of Arrestins to GPCRs. Roles for lipid signaling (e.g., ARNO-activated Arf activating phospholipase D) in these processes cannot be ruled out as playing a role in receptor desensitization or internalization. The β2-AR may use a similar mechanism to localize and regulate the release of Arrestin from an Arf6-bound site (Claing et al., 2001). For Arrestin to bind to the β2-AR, both agonist binding and GRK-dependent receptor phosphorylation are required to convert Arrestin to a conformation that binds with high affinity (Pierce et al., 2001). While ARNO binds to the LHR and promotes Arf6 activation and LHR desensitization, the participation of another Arf6 GEF, like EFA6 (Franco et al., 1999; Dulin et al., 2001), in these events has not been ruled out. These studies thus reveal a previously unappreciated level of complexity in the signaling pathway regulating LHR coupling to heterotrimeric G proteins, which places Arf6 at a pivotal position to direct the availability of Arrestin2. As our understanding of how the Arrestins are delivered to GPCRs unfolds, new avenues for basic research and drug development for the treatment of a variety of diseases will likely emerge.

6.

ACKNOWLEDGMENTS

I would like to thank the members of my laboratory, especially Sutapa Mukherjee, Evelyn Maizels, Regina Horvat, and Lisa Salvador, as well as Andrew Shenker, Vsevolod Gurevich, and James Casanova for helpful discussions. I would also like to thank Andrew Shenker for providing me with the receptor bundles used in the figure.

ROLE OF ARF6 IN GPCR DESENSITIZATION

15 319

REFERENCES Abdulaev, N. G., and Ridge, K. D. (1998). Light-induced exposure of the cytoplasmic end of transmembrane helix seven in rhodopsin. Proc. Natl. Acad. Sci., USA, 95, 12854-12859. Abell, A. N. and Segaloff, D. L. (1997). Evidence for the direct involvement of transmembrane region 6 and the lutropin/choriogonadotropin receptor in activating Gs. J. Biol. Chem., 272, 14586-14591. Altenbach, C., Cai, K., Khorana, H. G., and Hubbell, W. L. (1999). Structural features and light-dependent changes in the sequence 306-322 extending from helix VII to the palmitoylation sites in rhodopsin: a site-directed spin-labeling study. Biochemistry, 38, 7931-7937. Anderson, W. B. and Jaworski, C. J. (1979). Isoproterenol-induced desensitization of adenylate cyclase responsiveness in a cell-free system. J. Biol. Chem., 254, 4596-4601. Austin, C., Boehm, M., and Tooze, S. A. (2002). Site-specific cross-linking reveals a differential direct interaction of class 1, 2, and 3 ADP-ribosylation factors with adaptor protein complexes 1 and 3. Biochemistry, 41, 4669-4677. Ballesteros, J. A., Jensen, A. D., Liapakis, G., Rasmussen, S. G., Shi, L., Gether, U., and Javitch, J. A. (2001). Activation of the beta 2-adrenergic receptor involves disruption of an ionic lock between the cytoplasmic ends of transmembrane segments 3 and 6. J. Biol. Chem., 276, 29171-29177. Barak, L. S., Ferguson, S. S., Zhang, J., and Caron, M. G. (1997). A beta-arrestin/green fluorescent protein biosensor for detecting a G protein-coupled receptor activation. J. Biol. Chem., 272, 27497-27500. Barak, L. S., Warabi, K., Feng, X., Caron, M. G., and Kwatra, M. M. (1999). Real-time visualization of the cellular redistribution of G protein-coupled receptor kinase 2 and betaarrestin 2 during homologous desensitization of the substance P receptor. J. Biol. Chem., 274, 7565-7569. Bennett, T. A., Maestas, D. C., and Prossnitz, E. R. (2000). Arrestin binding to the G proteincoupled N-formyl peptode receptor is regulated by the conserved "DRY" sequence. J. Biol. Chem., 275, 24590-24594. Bhatnagar, A., Willins, D. L., Gray, J. A., Woods, J., Benovic, J. L., and Roth, B. L. (2001). The dynamin-dependent, arrestin-independent Internalization of 5-hydroxytryptamine 2A (5-HT2A) serotonin receptors reveals differential sorting of arrestins and 5-HT2A receptors during endocytosis. J. Biol. Chem., 276, 8269-8277. Bockaert, J., Hunzicker-Dunn, M., and Birnbaumer, L. (1976). Hormone-stimulated desensitization of hormone-dependent adenylyl cyclase. J. Biol. Chem., 251, 2653-2663. Bohn, L. M., Gainetdinov, R. R., Lin, F. T., Lefkowitz, R. J., and Caron, M. G. (2000). Muopioid receptor desensitization by beta-arrestin-2 determines morphine tolerance but not dependence. Nature, 408, 720-723. Brown, F. D., Rozelle, A. L., Yin, H. L., Balla, T., and Donaldson, J. G. (2001). Phosphatidylinositol 4,5-bisphosphate and Arf6-regulated membrane traffic. J. Cell. Biol., 154, 1007-1017. Carmen, C. V. and Benovic, J. L. (1999). G-protein-coupled receptors: turn-ons and turn-offs. Curr. Opin. Neurobiol., 8, 335-344. Caumont, A.-S., Galas, M.-C., Vitale, N., Aunis, D., and Bader, M.-F. (1998). Regulated exocytosis in chromaffin cells.Translocation of ARF6 stimulates a plasma membraneassociated phosphlipase D. J. Biol. Chem., 273, 1373-1379.

16 320

HUNZICKER-DUNN

Cavenagh, M. M., Whitney, J. A., Carroll, K., Zhang, C., Boman, A. L., Rosenwald, A. G., Mellman, I., and Kahn, R. A. (1996). Intracellular distribution of Arf proteins in mammalian cells. J. Biol. Chem., 271, 21767-21774. Celver, J., Vishnivetskiy, S. A., Chavkin, C., and Gurevich, V. V. (2002). Conservation of the phosphate-sensitive elements in the arrestin family of proteins. J. Biol. Chem., 277, 90439048. Chen, W., Hu, L. A., Semenov, M. V., Yanagawa, S., Kikuchi, A., Lefkowitz, R. J., and Miller, W. E. (2001). β-Arrestin1 modulates lymphoid enhancer factor transcriptional activity through interaction with phosphorylated dishevelled proteins. Proc. Natl. Acad. Sci., USA, 98, 14889-14894. Cheng, Z.-J., Yu, Q.-M., Wu, Y.-L., Ma, L., and Pei, G. (1998). Selective interference of βarrestin 1 with κ and δ but not µ opiod receptor/G protein coupling. J. Biol. Chem., 273, 24328-24333. Claing, A., Chen, W., Miller, W. E., Vitale, N., Moss, J., Premont, R. T., and Lefkowitz, R. J. (2001). β arrestin-mediated ARF6 activation and β2-adrenergic receptor endocytosis. J. Biol. Chem. 276, 42509-42513. Claing, A., Laporte, S. A., Caron, M. G., and Lefkowitz, R. J. (2002). Endocytosis of G protein-coupled receptors: roles of G protein-coupled receptor kinases and beta-arrestin proteins. Prog. Neurobiol., 66, 61-79. Conner, D. A., Mathier, M. A., Mortensen, R. M., Christe, M., Vatner, S. F., Seidman, C. E., and Seidman, J. G. (1997). β-Arrestin1 knockout mice appear normal but demonstrate altered cardiac responses to β-adrenergic stimulation. Circ. Res., 81, 1021-1026. D'Souza-Schorey, C., van Donselaar, E., Hsu, V. W., Yang, C., Stahl, P. D., and Peters, P. J. (1998). ARF6 targets recycling vesicles to the plasma membrane: insights from an ultrastructural investigation. J. Cell Biol., 140, 603-616. Drake, M. T., Zhu, Y., and Kornfeld, S. (2000). The assembly of AP-3 adaptor complexcontaining clathrin-coated vesicles on synthetic liposomes. Mol. Biol. Cell, 11, 3723-3736. Dulin, N. O., Niu, J., Browning, D. D., Ye, R. D., and Voyno-Yasenetskaya, T. (2001). Cyclic AMP-independent Activation of Protein Kinase A by Vasoactive Peptides. J. Biol. Chem., 276, 20827-20830. Ekstrom, R. C., Carney, E. M., Lamm, M. L. G., and Hunzicker-Dunn, M. (1992). Reversal of the desensitized state of pig ovarian follicular human choriogonadotropin-sensitive adenylylcyclase by guanosine 5'-0-(2-thiodiphosphate). J. Biol. Chem., 267, 22183-22189. Ekstrom, R. C. and Hunzicker-Dunn, M. (1989). Homologous desensitization of ovarian luteinizing hormone/human chorionic gonadotropin-responsive adenylyl cyclase is dependent upon GTP. Endocrinology, 124, 956-963. Ekstrom, R. C. and Hunzicker-Dunn, M. (1990). Opposing effects of ethanol on pig ovarian adenylyl cyclase desensitized by human choriogonadotropin or isoproterenol. Endocrinology. 127, 2578-2586. Evans, N. A., Groarke, D. A., Warrack, J., Greenwood, C. J., Dodgson, K., Milligan, G., and Wilson, S. (2001). Visualizing differences in ligand-induced beta-arrestin-GFP interactions and trafficking between three recently characterized G protein-coupled receptors. J. Neurochem., 77, 476-485. Ezra, E. and Salomon, Y. (1980). Mechanism of desensitization of adenylate cyclase by lutropin. J. Biol. Chem., 255, 653-658. Ezra, E. and Salomon, Y. (1981). Mechanism of desensitization of adenylate cyclase by lutropin. J. Biol. Chem., 256, 5377-5382. Ferguson, S. S. G. (2001). Evolving concepts in G protein-coupled receptor endocytosis: the role in receptor desensitization and signaling. Pharm. Rev., 53, 1-24.

ROLE OF ARF6 IN GPCR DESENSITIZATION

17 321

Ferguson, S. S. G., Barak, L. S., Zhang, J., and Caron, M. G. (1996). G-protein-coupled receptor regulation: role of G-protein-coupled receptor kinases and arrestins. Can. J. Physiol. Pharmacol., 74, 1094-1110. Ferrari, S. L. and Bisello, A. (2001). Cellular distribution of constitutively active mutant parathyroid hormone (PTH)/PTH-related protein receptors and regulation of cyclic adenosine 3',5'-monophosphate signaling by β-arrestin2. Mol. Endo., 15, 149-163. Franco, M., Peters, P. J., Boretto, J., van Donselaar, E., Neri, A., D'Souza-Schorey, C., and Chavrier, P. (1999). EFA6, a sec7 domain-containing exchange factor for ARF6, coordinates membrane recycling and actin cytoskeleton organization. EMBO J., 18, 14801491. Frank, S. R., Hatfield, J. C., and Casanova, J. E. (1998). Remodeling of the actin cytoskeleton is coordinately regulated by protein kinase C and the ADP-ribosylation factor nucleotide exchange factor ARNO. Mol. Biol. Cell, 9, 3133-3146. Galas, M.-C., Helms, J. B., Vitale, N., Thierse, D., Aunis, D., and Bader, M.-F. (1997). Regulated exocytosis in chromaffin cells. A potential role for a secretory granuleassociated ARF6 protein. J. Biol. Chem., 272, 2788-2793. Godi, A., Pertile, P., Meyers, R., Marra, P., Di Tullio, G., Iurisci, C., Luini, A., Corda, D., and De Matteis, M. A. (1999). ARF mediates recruitment of Ptdins-4-OH kinase-β and stimulates synthesis of Ptdins(4,5)P2 on the golgi complex. Nature Cell Biol., 1, 280-287. Goodman, O. B., Krupnick, J. G., Santini, F., Gurevich, V. V., Penn, R. B., Gagnon, A. W., Keen, J. H., and Benovic, J. L. (1996). β-arrestin acts as a clathrin adaptor in endocytosis of the β2-adrenergic receptor. Nature, 383, 447-450. Gurevich, V. V. and Benovic, J. L. (1993). Visual arrestin interaction with rhodopsin. J. Biol. Chem., 268, 11628-11638. Gurevich, V. V., Dion, S. B., Onorato, J. J., Ptasienski, J., Kim, C. M., Sterne-Marr, M., Hosey, M. M., and Benovic, J. L. (1995). Arrestin interactions with G protein-coupled receptors. J. Biol. Chem., 270, 720-731. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 4-kinase alpha is a downstream effector of the small G protein ARF6 in membrane ruffle formation. Cell, 99, 521-532. Hunzicker-Dunn, M. (1981). Rabbit follicular adenylyl cyclase activity. II. Gonadotropininduced desensitization in granulosa cells and follicle shells. Biol. Repro., 24, 279-286. Iacovelli, L., Franchetti, R., Masini, N. M., and DeBlasi, A. (1996). GRK2 and β−arrestin1 as negative regulators of thyrotropin receptor-stimulated response. Mol. Endo., 10, 11381146. Jackson, C. L. and Casanova, J. E. (2000). Turning on ARF: the Sec7 family of guaninenucleotide-exchange factors. Trends Cell Biol., 10, 60-67. Kahn, R. A., Randazzo, P., Serafini, T., Weiss, O., Rulka, C., Clark, J., Amherdt, M., Roller, P., Orci, L., and Rothman, J. E. (1992). The amino terminus of ADP-ribosylation factor (ARF) is a critical determinant of ARF activities and is a potent and specific inhibitor of protein transport. J. Biol. Chem., 267, 13039-13046. Klein, U., Muller, C., Chu, P., Birnbaumer, M., and von Zastrow, M. (2001). Heterologous inhibition of G protein-coupled receptor endocytosis mediated by receptor-specific trafficking of beta-arrestins. J. Biol. Chem., 276, 17442-17447. Knospe, V., Donoso, L. A., Banga, J. P., Yue, S., Kasp, E., and Gregerson, D. S. (1988). Epitope mapping of bovine retinal S-antigen with monoclonal antibodies. Eye Res., 7, 1137-1147.

18 322

HUNZICKER-DUNN

Kohout, T. A., Lin, F. T., Perry, S. J., Conner, D. A., and Lefkowitz, R. J. (2001). β-Arrestin 1 and 2 differentially regulate heptahelical receptor signaling and trafficking. Proc. Natl. Acad. Sci., 98, 1601-1606. Kovoor, A., Celvert, J., Abdryashitov, R. I., Chavkin, C., and Gurevich, V. V. (1999). Targeted construction of phosphorylation-independent beta-arrestin mutants with constitutive activity in cells. J. Biol. Chem., 274, 6831-6834. Krupnick, J. G. and Benovic, J. L. (1998). The role of receptor kinases and arrestins in G protein-coupled receptor regulation. Annu. Rev. Pharmacol. Toxicol., 38, 289-319. Lamm, M. L. G., Ekstrom, R. C., Maizels, E. T., Rajagopalan, R. M., and Hunzicker-Dunn, M. (1994). The effect of protein kinases on desensitization of the porcine follicular membrane luteinizing hormone/chorionic gonadotropin-sensitive adenylyl cyclase. Endocrinology, 134, 1745-1754. Lamm, M. L. G. and Hunzicker-Dunn, M. (1994). Phosphorylation-independent desensitization of the luteinizing hormone/chorionic gonadotropin receptor in porcine follicular membranes. Mol. Endo., 8, 1537-1546. Laporte, S. A., Oakley, R. H., Holt, J. A., Barak, L. S., and Caron, M. G. (2000). The interaction of β-arrestin with the AP-2 adaptor is required for clustring of β2-adrenergic receptor into clathrin-coated pits. J. Biol. Chem., 275, 23120-23126. Laporte, S. A., Oakley, R., Zhang, J., Holt, J. A., Ferguson, S. S. G., Caron, M. G., and Barak, L. S. (1999). The β2-adrenergic receptor/β arrestin complex recruits the clathrin adaptor AP-2 during endocytosis. Proc. Natl. Acad. Sci., USA, 96, 3712-3717. Lenhard, J. M., Kahn, R. A., and Stahl, P. D. (1992). Evidence for ADP-ribosylation factor (ARF) as a regulator of in vitro endosome-endosome fusion. J. Biol. Chem., 267, 1304713052. Lohse, M. J., Andexinger, S., Pitcher, J., Trukawinski, S., Codina, J., Faure, J.-P., Caron, M. G., and Lefkowitz, R. J. (1992). Receptor-specific desensitization with purified proteins. J. Biol. Chem., 267, 8558-8564. Lohse, M. J., Benovic, J. L., Codina, J., Caron, M. G., and Lefkowitz, R. J. (1990). β-arrestin: a protein that regulates β-adrenergic receptor function. Science, 248, 1547-1550. Marsh, J. M., Mills, T. M., and Lemaire, W. J. (1973). Preovulatory changes in the synthesis of cyclic AMP by rabbit Graafian follicles. Biochem. Biophys. Acta, 304, 197-202. Min, L., Galet, C., and Ascoli, M. (2001). The association of arrestin-3 with the human lutropin/choriogonadotropin receptor depends mostly on receptor activation rather than on receptor phosphorylation. J. Biol. Chem., 277, 702-710. Mukherjee, S., Casanova, J. E., and Hunzicker-Dunn, M. (2001). Desensitization of the luteinizing hormone/choriogonadotropin receptor in ovarian follicular membranes is inhibited by catalytically inactive ARNO. J. Biol. Chem., 276, 6524-6528. Mukherjee, S., Gurevich, V. V., Jones, J. C. R., Casanova, J. E., Frank, S. R., Maizels, E. T., Bader, M.-F., Kahn, R. A., Palczewski, K., Aktories, K., and Hunzicker-Dunn, M. (2000). The ADP ribosylation factor nucleotide exchange factor ARNO promotes β-arrestin release necessary for luteinizing hormone/choriogonadotropin receptor desensitization. Proc. Natl. Acad. Sci., USA, 97, 5901-5906. Mukherjee, S., Gurevich, V. V., Preninger, A., Hamm, H. E., Bader, M. F., Fazleabas, A. T., Birnbaumer, L., and Hunzicker-Dunn, M. (2002). Aspartic acid 564 in the third cytoplasmic loop of luteinizing hormone/choriogonadotropin receptor is crucial for phosphorylation-independent interaction with arrestin2. J. Biol. Chem., 277, 17916-17927. Mukherjee, S., Palczewski, K., Gurevich, V. V., Benovic, J. L., Banga, J. P., and HunzickerDunn, M. (1999a). A direct role for arrestins in desensitization of the luteinizing

ROLE OF ARF6 IN GPCR DESENSITIZATION

19 323

hormone/choriogonadotropin receptor in porcine ovarian follicular membranes. Proc. Natl. Acad. Sci., USA , 96, 493-498. Mukherjee, S., Palczewski, K., Gurevich, V. V., and Hunzicker-Dunn, M. (1999b). βarrestindependent desensitization of luteinizing hormone/choriogonadotropin receptor is prevented by a synthetic peptide corresponding to the third intracellular loop of the receptor. J. Biol. Chem., 274, 12984-12989. Nagel, W., Zeitlmann, L., Schilcher, P., Geiger, C., Kolanus, J., and Kolanus, W. (1998). Phosphoinositide 3-OH kinase activates the beta2 integrin adhesion pathway and induces membrane recruitment of Cytohesin-1. J. Biol. Chem., 273, 14853-14861. Oakley, R. H., Laporte, S. A., Holt, J. A., Caron, M. G., and Barak, L. S. (2000). Differential affinities of visual arrestin, β arrestin1, and β arrestin2 for G protein-coupled receptors delineate two major classes of receptors. J. Biol. Chem., 275, 17201-17210. Paing, M. M., Stutts, A. B., Kohout, T. A., Lefkowitz, R. J., and Trejo, J. (2002). β-Arrestins regulate protease-activated receptor-1 desensitization but not internalization or Downregulation. J. Biol. Chem., 277, 1292-1300. Palczewski, K. (1994). Structure and functions of arrestins. Prot. Sci., 3, 1355-1361. Pals-Rylaarsdam, R., Zu, Y., Witt-Enderby, P., and Benovic, J. L. (1995). Desensitization and internalization of the m2 muscarinic acetylcholine receptor are directed by independent mechanisms. J. Biol. Chem., 270, 29004-29011. Peters, P. J., Gao, M., Gaschet, J., Ambach, A., van Donselaar, E., Traverse, J. F., Bos, E., Wolffe, E. J., and Hsu, V. W. (2001). Characterization of coated vesicles that participate in endocytic recycling. Traffic, 2, 885-895. Peters, P. J., Hsu, V. W., Ooi, C. E., Finazzi, D., Teal, S. B., Oorschot, V., Donaldson, J. G., and Klausner, R. D. (1995). Overexpression of wild-type and mutant ARF1 and ARF6: distinct perturbations of nonoverlapping membrane compartments. J. Cell. Biol., 128, 1003-1017. Pierce, K. L., Lefkowitz, R. J., and Lefkowitz, R. J. (2001). Classical and new roles of βarrestins in the regulation of G-protein coupled receptors. Nat. Rev. Neurosci., 2, 727-733. Premont, R. T., Claing, A., Vitale, N., Freeman, J. L. R., Pitcher, J. A., Patton, W. A., Moss, J., Vaughan, M., and Lefkowitz, R. J. (1998). β2-adrenergic receptor regulation of GIT1, a G protein-coupled receptor kinase-associated ADP ribosylation factor GTPase-activating protein. Proc. Natl. Acad. Sci., USA, 95, 14082-14097. Radhakrishna, H. and Donaldson, J. G. (1997). ADP-ribosylation factor 6 regulates a novel plasma membrane recycling pathway. J. Cell Biol., 139, 49-61. Radhakrishna, H., Klausner, R. D., and Donaldson, J. G. (1996). Aluminum fluoride stimulates surface protrusions in cells overexpressing the ARF6 GTPase. J. Cell Biol., 134, 935-947. Rajagopalan-Gupta, R. M., Mukherjee, S., Zhu, X., Ho, Y.-K., Hamm, H., Birnbaumer, M., Birnbaumer, L., and Hunzicker-Dunn, M. (1999). Roles of Gi and Gq/11 in mediating desensitization of the luteinizing hormone/choriogonadotropin receptor in porcine ovarian follicular membranes. Endocrinology, 140, 1612-1621. Roy, C., Guillon, G., and Jard, S. (1976). Hormone-dependent desensitization of vasopressinsensitive adenylate cyclase. Biochem. Biophys. Res. Comm., 72, 1265-1270. Salvador, L. M., Mukherjee, S., Kahn, R. A., Lamm, M. L. G., Fazleabas, A. T., Maizels, E. T., Bader, M.-F., Hamm, H., Rasenick, M. M., Casanova, J. E., and Hunzicker-Dunn, M. (2001). Activation of the luteinizing hormone/choriogonadotropin hormone receptor promotes ADP ribosylation factor (ARF) 6 activation in porcine ovarian follicular membranes. J. Biol. Chem., 276, 33773-33781.

20 324

HUNZICKER-DUNN

Santini, F., Penn, R. B., Gagnon, A. W., Benovic, J. L., and Keen, J. H. (2000). Selective recruitment of arrestin-3 to clathrin coated pits upon stimulation of G protein-coupled receptors. J. Cell. Sci., 113, 2463-2470. Schulz, A., Bruns, K., Henklein, P., Krause, G., Schubert, M., Gudermann, T., Wray, V., Schultz, G., and Schoneberg, T. (2000). Requirement of specific intrahelical interactions for stabilizing the inactive conformation of glycoprotein hormone receptors. J. Biol. Chem., 275, 37860-37869. Shenker, A. (2002). Activating mutations of the lutropin/choriogonadotropin receptor in precocious puberty. Receptors Channels, 8, 3-18. Troispoux, C., Guillou, F. E. J.-M., Firsov, D., Iacovelli, L., DeBlasi, A., Combarnous, Y., and Reiter, E. (2000). Involvement of G protein-coupled receptor kinases and arrrestins in desensitization to follicle sitmulating hromone action. Mol. Endo., 13, 1599-1614. Wang, Z., Liu, X., and Ascoli, M. (1997). Phosphorylation of the lutropin/choriogonadotropin receptor facilitates uncoupling of the receptor from adenylyl cyclase and endocytosis of the bound hormone. Mol. Endo., 11, 183-192. Zhu, X., Gudermann, T., Birnbaumer, M., and Birnbaumer, L. (1993). A luteinizing hormone receptor with a severely truncated cytoplasmic tail (LHR-ct628) desensitizes to the same degree as the full-length receptor. J. Biol. Chem., 268, 1723-1728. Zhu, Y., Drake, M. T., and Kornfeld, S. (1999). ADP-ribosylation factor 1 dependent clathrincoat assembly on synthetic liposomes. Proc. Natl. Acad. Sci., USA, 96, 5013-5018.

Chapter 16 ARF-LIKE PROTEINS ANNETTE SCHÜRMANN AND HANS-GEORG JOOST German Institute of Human Nutrition, Potsdam-Rehbrücke, Germany

Abstract:

1.

At least 10 different proteins have homology to members of the Arf family but each lacks activities that define “true” Arfs. In addition, the percent identity between these 10 proteins, when aligned with Arfs or to each other, is lower (typically 40-60%) than that between Arfs (greater than 65% identity). A summary of current information on each is presented below. Despite extensive conservation of structure between Arfs and Arls, the Arls represent a much more functionally divergent collection of GTPases, with apparent roles in vesicle traffic (Arl1, Sar1), microtubule organization (Arl2, Arl3), spermatogenesis (Arl4), and differentiation (Arfrp1). However, the overlap in binding partners between functionally distinct Arls suggests that Arfs and Arls are likely to have a complex web of overlapping and distinct functions.

INTRODUCTION

ADP-ribosylation factor-like (Arl) proteins are GTPases that were identified on the basis of their sequence similarity with Arfs (33-60% identical amino acids). At present, 10 mammalian isotopes (Arl1-7, with two different proteins designated Arl6, Arfrp1 and Sar1) have been described (see Figure 1). GTP-binding has been shown for most Arl proteins, but all Arls are essentially devoid of GTPase activity and activities described for Arf isotypes, i.e., activity as cofactor in the ADP-ribosylation of GαS by bacterial toxins (Kahn and Gilman, 1984; Kahn and Gilman, 1986), activation of phospholipase D (Brown et al., 1993; Cockcroft et al., 1994), or activation of phosphatidylinositol 4-phosphate 5-kinase (Honda et al., 1999). No Arl has ever been purified or cloned on the basis of an activity. Furthermore, no Arl can rescue the synthetic lethality of arf1-/arf2- mutants in Saccharomyces cerevisiae, an effect that is characteristic for mammalian (Kahn et al., 1991; Lee et al., 1992) and other Arfs (Drosophila melanogaster, Murtagh et al., 1993; Giardia lamblia, Lee et al., 1992). 325

2 326

SCHÜRMANN AND JOOST

These data lead to the conclusion that Arfs and Arls lack a functional overlap, except for their role as GTPase-operated molecular switches. In addition, there appears to be little functional overlap between Arls, as reflected by a considerable heterogeneity of their expression in tissues and of their subcellular distribution. Some Arls appear to be involved in the regulation of protein and/or vesicle transport between cell organelles (Arl1, Arl4, Sar1) or in the regulation of enzymatic activities controlling these processes (Arfrp1). Furthermore, the physiological functions of the different family members, in as far as they are known, appear to be quite diverse (e.g., Arl4, spermatogenesis; Arfrp1, differentiation of mesoderm during embryogenesis).

2.

ARL1

Arl1 is the closest relative to the Arfs, appears to have some functional similarities to the Arfs, and should rather be considered an atypical Arl protein. Although the subcellular localization and several interaction partners of Arl1 were identified in the past, the exact role of Arl1 is still unknown.

2.1

Similarities and differences of Arl1 to the Golgirelated function of Arf

Arl1 was originally cloned from Drosophila melanogaster and is essential for normal development because arl1- mutants exhibited a recessive lethal phenotype (Tamkun et al., 1991). The mammalian isoform that was subsequently identified (Schürmann et al., 1994) exhibits 79% amino acid identity with Drosophila Arl1, and 57% with human Arf1. The identity with other Arl isotypes is lower (40% with Arl2, 45% with Arl3, and 41% with Arl4; Figure 1). Because mammalian Arls are so highly conserved (97-99% identity between human, mouse, rat and bovine) the human proteins are used in alignments and sequence analyses (see Figure 1) but results will hold true for Arls from at least these other mammals. Northern blots revealed that Arl1 is ubiquitously expressed in rat tissues (brain, heart, spleen, lung, liver, skeletal muscle, kidney, testis and fat (Schürmann et al., 1994; Lowe et al., 1996). Binding of GTPγS to recombinant Arl1 appears to be partially dependent on the presence of lipids and detergent (Tamkun et al., 1991), a feature prominent for the Arfs.

ARF-LIKE PROTEINS

3273

Figure 1. Dendrogram of the Arf family. The human proteins were aligned with the CLUSTAL program. Bars denote the classes I-III of Arfs and the Arl4 subgroup. All mammalian Arl proteins are highly conserved (97% to 99% identity between human, mouse, rat and bovine).

2.2

Subcellular localization and conditions of redistribution between Golgi and cytosol

Like most Arf proteins, Arl1 is associated with the Golgi complex. Lowe et al. (1996) detected endogenous Arl1 in the perinuclear region of normal rat kidney (NRK) cells by indirect immunofluorescence, co-localized with the Golgi markers p28 (a cis-Golgi protein) and mannosidase II. The same subcellular localization was found in other cell lines. Furthermore, treatment of cells with nocodazole, a microtubule depolymerizing drug that causes fragmentation of the Golgi complex, dispersed Arl1 labeling into many

4 328

SCHÜRMANN AND JOOST

punctate structures throughout the cell. A similar pattern was observed for both Golgi markers p28 and mannosidase II (Lowe et al., 1996). A different pattern of cellular redistribution was observed for Arl1 and mannosidase II when brefeldin A was used to disrupt the Golgi structures. After short treatment (2 min) with brefeldin A, the labeling pattern of Arl1 and mannosidase II was unaffected. After 5 min of brefeldin A-treatment, mannosidase II was detected in tubulovesicular extensions. Arl1 labeling did not colocalize with these structures, but was detected in the cytosol. After 30 min of brefeldin A treatment, Arl1 was completely redistributed to the cytosol, whereas mannosidase II was transferred into an extensive reticular network. The observed response of Arl1 to brefeldin A differs from those of Arf and COPI (β-COP), components of non-clathrin-coated transport vesicles. Arf and COPI redistributed into the cytosol faster (within 2 min) than Arl1. This suggests that the effect of brefeldin A on Arl1 may be indirect or secondary, as compared to Arfs. As is the case for Arfs, brefeldin A could block guanine nucleotide exchange via interaction with a brefeldin Asensitive, Arl1-specific exchange factor (GEF). If the normal turnover rate of Arl1 cycling on and off Golgi membranes is slower than that of Arfs, then the inhibition of this GEF could cause the delayed release of Arl1 from the Golgi (Lowe et al., 1996). Alternatively, brefeldin A may affect Arl1 indirectly. For example, inhibition of an Arf-specific GEF by brefeldin A leads to the fast redistribution of Arf1 into the cytosol. Subsequently, the lack of Arf1 within Golgi membranes could result (via an unknown mechanism) in the release of Arl1 into the cytosol, suggesting that Arf1 is acting upstream of Arl1. Somewhat different results were obtained by van Valkenburgh et al. (2001). They found that epitope tagged Arl1 (Arl1-myc) was released from the Golgi with similar kinetics to endogenous Arfs. This discrepancy may result from the use of epitope tagging or overexpression of the GTPase. There is now extensive evidence that either approach can produce data that differ from the location or behavior of endogenous GTPases, particularly for members of the Arf family. Unfortunately, the levels of expression of Arl1 are typically low and make studies of endogenous protein more difficult than for Arfs. We should probably consider tentative any conclusions based solely on results obtained with overexpressed and or epitope tagged proteins. Additional evidence that the subcellular distribution of Arf1 and Arl1 is regulated independently was obtained with aluminum fluoride. AlF activates heterotrimeric G-proteins and reverses the effects of brefeldin A on COPI, mannosidase II, and Rab6, but not that on TGN38. When NRK cells were preincubated with AlF for 20 min prior to addition of brefeldin A, both Arl1 and COPI were more tightly associated with the Golgi complex (Lowe

ARF-LIKE PROTEINS

3295

et al., 1996). In contrast, AlF failed to reverse the effect of brefeldin A on the redistribution of Arfs into the cytosol (Donaldson et al., 1991). Association of Arl1 with the Golgi appears saturable (Lu et al., 2001). Moderate overexpression of Arl1 gave rise to enrichment of the protein at the trans side of the Golgi apparatus. In contrast, when expression was high, the majority of Arl1 was detected in the cytosol. This saturable Golgi association was explained by a rate-limiting activation step that depends on nucleotide-exchange factors. As expected, the Golgi association depends on N-terminal myristoylation. A mutant of Arl1 lacking the myristoylation motif (Arl1-A2) was found in the cytosol and the nucleus but was absent from the Golgi. Arl1 seems to be necessary for maintaining normal Golgi structure, as overexpression of Arl1-N31, a nucleotide exchange defective mutant, causes a disintegration of the Golgi apparatus (Lu et al., 2001). This phenotype was similar to that observed in cells treated with brefeldin A (Klausner et al., 1992) or cells overexpressing Arf1-N31 (Dascher and Balch, 1994). In addition, the perinuclear, Golgi-associated localization of AP-1 (γ-adaptin) was also disrupted in Arl1-N31 cells, suggesting that Arl1 is involved in Golgi recruitment of AP-1 (Lu et al., 2001). Overexpression of Arl1-L71, the activated, GTPase-defective mutant, resulted in an expanded perinuclear Golgi-like structure (Lu et al., 2001). Van Valkenburgh et al. (2001) also described a more diffuse Golgi staining in normal rat kidney (NRK) cells stably expressing Arl1-L71-myc. The Golgi of cells expressing Arl1-L71 exhibited an engorged lumen, particularly at the ends of the stacks. This phenotype was comparable to early time points in the expression of Arf1-L71, suggesting similarities in function of these proteins at the Golgi. However, the change in Golgi morphology induced by Arl1-L71 was subtle and was not accompanied by the progressive development of large perinuclear vesicles which often surrounded the nucleus, as seen for Arf1-L71 (van Valkenburgh et al., 2001). The expression of Arl1-L71 caused a massive recruitment of both COPI (β-COP) and AP-1 (γ-adaptin) into the expanded Golgi (Lu et al., 2001). Treatment with brefeldin A normally inhibits Golgi-association of COPI and AP-1, and the formation of their respective vesicles. In cells expressing Arl1-L71, in contrast, COPI and AP-1 could not be dissociated by brefeldin A. Moreover, extensive association of Arfs with the expanded Golgi-like structure was observed in cells overexpressing Arl1-L71, suggesting that active Arl1 could stimulate Golgi recruitment of Arfs. Arfs remain Golgi-associated even after brefeldin A treatment when Arl1-L71 was expressed in these cells (Lu et al., 2001). These results indicate that Arl1 has the potential to regulate COPI, AP-1, and Arf-mediated vesicle formation in the Golgi.

6 330

SCHÜRMANN AND JOOST

The effects of Arl1 on membrane traffic in the Golgi apparatus appear to be specific (Lu et al., 2001). Overexpression of Arl1-L71 selectively stimulated recruitment of COPI and AP-1, which mediate export from the Golgi. In contrast, the localization of COPII (which mediates ER export), and of α-adaptin (which mediates endocytosis from the plasma membrane), was not affected by the Arl1 mutant.

2.3

Arl1 binding proteins

In order to test whether Arl1 and Arf1 have overlapping functions, van Valkenburgh et al. (2001) analyzed the interaction of known Arf-binding proteins with Arl1. Of these, Arfaptin2 (also termed POR1, see Chapter 11 for details) and mitotic kinesin-like protein-1 (MKLP1) exhibited GTPdependent binding with Arl1. Furthermore, Arfaptin2 increased the stoichiometry of GTP binding to Arl1 (van Valkenburgh et al., 2001). Arfaptin2 was previously identified as an effector of Rac (D’Souza-Schorey et al., 1997; Kanoh et al., 1997), and has been proposed to be a mediator of cross talk between Arf proteins and Rac. Arfaptin2 binds specifically to GTP-bound Arf1 and Arf6, and binds to Rac•GTP and Rac•GDP with similar affinities (Tarricone et al., 2001). The C-terminal domain of MKLP1 binds to all activated Arf isoforms (Boman et al., 1999) and also to Arl1, but not to Arl2 or Arl3 (van Valkenburgh et al., 2001). MKLP1 is present in central spindles and midbodies of mammalian cells, and is thought to be involved in the re-organization of membranes during the terminal phase of cytokinesis (Kuriyama et al., 2002). In a two-hybrid screen performed with the GTPase-defective mutants of Arl1-3, van Valkenburgh et al. (2001) identified PDEδ and HRG4 as GTPdependent binding proteins. In addition, the authors identified Golgin-245 to bind to Arl1-L71 and Arl3-L71 (see Figure 2). Golgins are Golgi-associated proteins with antigenic potency in autoimmune diseases. The 245 kDa protein Golgin-245 contains a coiled-coil domain and is peripherally associated with the cis-Golgi. Golgin-245 comprises a C-terminal GRIPdomain that is believed to direct associated proteins to the Golgi apparatus. Arl1 and the GRIP domain of Golgin-245 are co-localized when expressed in NRK cells. In contrast to Arl1, binding of the GRIP-domain to the Golgi was not disrupted by brefeldin A, suggesting that this domain can bind the Golgi independently of Arl1. SCOCO (Short Coiled-Coil; see Figure 2) is another binding protein of Arl1-L71 that colocalizes with Arl1 in the Golgi. The function of SCOCO is still unknown (van Valkenburgh et al., 2001). Its 82 amino acid residues are predicted to fold almost completely into a coiled-coil structure. SCOCO message was widely distributed in human tissues with the highest levels

ARF-LIKE PROTEINS

3317

detected in brain, heart, and skeletal muscle. The protein was found associated with the Golgi apparatus and the plasma membrane. Like Arl1, the binding of SCOCO to the Golgi was rapidly reversed by brefeldin A. Purified SCOCO had no detectable GTPase activating protein (GAP) activity for Arl1.

Figure 2. Specific and shared binding partners of Arfs, Arls and Arfrp1 are shown. Arf family members are shown in boxes (center), with binding partners that are shared by multiple members shown above them and those specific to the Arfs or an Arl, shown below.

3.

ARL2

Arl2 is widely expressed in mammals, Drosophila melanogaster, Caenorhabditis elegans and yeast. Arl2 is a regulator of microtubule dynamics and therefore plays an essential role in cytokinesis. A subpopulation of Arl2 is located in the mitochondria and may be involved in adenine nucleotide transport. Arl2 and Arl3 are closely related, exhibit a similar structure, and share some features. Both proteins interact with PDEδ and are most likely involved in PDEδ-mediated transport of prenylated proteins.

8 332

SCHÜRMANN AND JOOST

3.1

Potential involvement of Arl2 in the biogenesis of microtubules and interaction with the mitochondrial adenine nucleotide transporter (ANT)

Arl2 was identified in a PCR-based cloning approach on the basis of its sequence similarity with Arf (Clark et al., 1993). It is expressed in all mammalian tissues with highest levels in neural cells. Arl2 binds GDP or GTP much faster than Arf, even in the absence of lipids or detergents (Clark et al., 1993). The presumed budding yeast ortholog of Arl2, CIN4, was identified in a genetic screen for mutants defective in microtubule function (Hoyt et al., 1990; Stearns et al., 1990). In addition, Radcliffe et al. (2000) cloned the Arl2/CIN4 ortholog, alp41, from fission yeast, and demonstrated that it is essential for the cofactor-dependent tubulin pathway. Disruption of the alp41 gene resulted in microtubule dysfunction and growth polarity defects. The phenotype observed for the alp41- mutant was very similar to that of cells lacking the gene encoding cofactor B or E (alp11 and alp21, respectively), suggesting that Alp41 is involved in the microtubule biogenesis pathway. Interestingly, expression of Alp1 (ortholog of cofactor D) or expression of Alp21 (cofactor E) rescued the phenotype of alp41mutants. These genetic analyses indicate that Alp41/Arl2 plays an essential role in the cofactor-dependent tubulin pathway (Radcliffe et al., 2000).

Figure 3. Model depicting the diversity in localization and actions of Arl proteins in (1) the biogenesis of microtubules (Arl2; Bhamidipati et al., 2000; R. A. Kahn, unpublished observations), (2) mitochondrial adenine nucleotide transport (Arl2; Sharer et al., 2002), (3) nuclear transport (Arl4; Jacobs et al., 1999), (4) vesicle budding from the endoplasmic reticulum (Sar1; Springer et al., 1999), membrane traffic through the Golgi (Arl1), and nuclear localization of Arl4, Arl6/Arf4L and Arl7. (C, D, E = cofactors C, cofactor D and cofactor E).

ARF-LIKE PROTEINS

3339

Microtubules are polarized polymers of α/β tubulin heterodimers that participate in several cellular functions. The dynamic behavior of microtubules is controlled by polymerization-dependent GTP hydrolysis catalyzed by the β-subunit and associated proteins. Tubulin heterodimers are generated in a multi-step process that involves several chaperone proteins designated cofactor A-E. They assemble the α/β-tubulin heterodimer, interact with native tubulin, and act as a β-tubulin GTPase activating protein (GAP; see Figure 3; Levis et al., 1997; Tian et al., 1997). Regulation of microtubule dynamics by Arl2 was also shown in mammalian cells (Bhamidipati et al., 2000). The authors demonstrated that Arl2 interacts with cofactor D and modulates the interaction of this tubulinfolding protein with native tubulin (Figure 3). In addition, Arl2 down regulates the tubulin GAP activity of cofactors C, D and E, and inhibits the binding of cofactor D to native tubulin (Figure 3). Overexpression of cofactor D or E in cultured cells resulted in the destruction of the tubulin heterodimer and of microtubules. Arl2 specifically prevented the destruction of tubulin and microtubules by cofactor D, but not by cofactor E. Using different mutants of Arl2, Bhamidipati et al. (2000) demonstrated that Arl2 preferentially interacts with cofactor D in its GDP-bound form. As binding of the effector loop mutant Arl2-A47 to cofactor D was indistinguishable from that of wild-type Arl2, the authors concluded that the switch from the GDP-bound to the GTP-bound conformation is not essential for the interaction between Arl2 and cofactor D. However, a second effector loop mutant, Arl2-A50, failed to bind cofactor D, suggesting that the cofactor is an effector of Arl2•GDP. While Bhamidipati et al. (2000) suggested that Arl2 is involved in a tubulin destruction pathway others (Y. Li, W. Kelly, and R. A. Kahn, unpublished observations) described an essential role for Arl2 in tubulin biosynthesis during embryogenesis and in the germline. The authors used RNA-mediated interference to target Arl2 in C. elegans. Injection of Arl2 double-stranded RNA resulted in embryonic lethality or sterility in early offspring. Germ cells often exhibited defects in cellularization, forming regions with numerous nuclei sharing a complete cytoplasm. As embryos lacking Arl2 exhibited decreased tubulin levels, Li, et al. (unpublished observations) concluded that Arl2 is required for the biosynthesis of tubulin. Interestingly, the phenotype observed after injection of double stranded Arl2 RNA was similar to that seen with injection of the

10 334

SCHÜRMANN AND JOOST

C. elegans ortholog of cofactor D RNA, leading to the conclusion that Arl2 and cofactor D co-participate in cell-cycle related events in C. elegans. A second important function of Arl2 has been established by Sharer and Kahn (1999). Arl2•GTP interacts with a 19 kDa protein, designated Bart (Binder of Arl Two), with an apparent KD of, 20 nM. Like Arl2, Bart is expressed ubiquitously and is most abundant in brain. Binding of Arl2 to Bart was detected in solution (Sharer and Kahn, 1999), and no detectable membrane association was found for either protein. This observation is of particular interest, because binding of all other interacting proteins to Arf isotypes requires the presence of membranes or lipids. Thus, binding of Arfs to membranes had been considered a critical part of Arf action. In contrast to every other member of the Arf family, no incorporation of [³H]myristate into Arl2 could be detected either when coexpressed in bacteria with N-myristoyltransferases, or in cultured mammalian cells (Sharer et al., 2002). Lack of covalent N-myristoylation preserves the Nterminal amphipathic α-helix as a potential mitochondrial import sequence. Recently, Sharer et al. (2002) indeed demonstrated that the Arl2-Bart complex is located in the mitochondria within the intermembrane space (Figure 3). By overlay assays with mitochondria isolated from five different tissues, the mitochondrial adenine nucleotide transporter (Ant) was identified as a binding partner of the Arl2-Bart complex (Sharer et al., 2002). Ant1 appeared to be a specific binding partner of the Arl2•GTP-Bart complex, as the Ant2 isotype did not bind the complex. Ant proteins are abundant, integral proteins of the mitochondrial inner membrane, where they homodimerize to form channels used for exchange of ATP and ADP, thereby controlling the levels of ATP in the cytoplasm (Fiore et al., 1998). In rodents, Ant1 is the major isotype present in skeletal muscle, cardiac muscle, and brain. In contrast, expression of Ant2 is ubiquitous with lower levels than those of Ant1 in heart and skeletal muscle (Stepien et al., 1992; Graham et al., 1997; Levy et al., 2000).

4.

ARL3: A POTENTIAL REGULATOR OF THE PDEδ-MEDIATED PROTEIN TRANSPORT

Arl3 (Cavenagh et al., 1994) exhibits 43% identity with Arf1. The GTPase is expressed in most tissues, with elevated mRNA levels detected in several human tumor cell lines. Recombinant Arl3 binds guanine nucleotides but lacks intrinsic GTPase activity (Cavenagh et al., 1994). However, Arl3 exhibits an unusually low affinity for guanine nucleotides, with a KD of 24 nM for mant-GDP and 48 µM for mant-GTP (Linari et al., 1999a). The Arl3•GTP, but not the Arl3•GDP binds the δ subunit of the cGMP

ARF-LIKE PROTEINS

11 335

phosphodiesterase (PDEδ) (Figure 2). The binding decreased the dissociation rate of GTP, thereby stabilizing Arl3•GTP (Linari et al., 1999a). PDEδ is also an effector of Arl2 (Hanzal-Bayer et al., 2002). PDEδ was isolated as the fourth subunit of bovine rod PDE (PDE6; Gillespie et al., 1989). PDE6 is the effector for the trimeric G protein transducin after its activation by light-sensitive rhodopsin. PDEδ binds to the prenylated C-terminus of PDEα,β, resulting in a translocation of the PDE6 holoenzyme from the membranes to the cytoplasmic fraction (Florio et al., 1996). PDEδ also transfers membrane-bound Rab13 to the cytoplasm (Marzesco et al., 1998), suggesting that the function of PDEδ is similar that of GDP-dissociation inhibitors (GDIs), which extract prenylated proteins from cellular membranes. This conclusion is strongly supported by structural similarities between PDEį and Rho GDI (Hanzal-Bayer et al., 2002). PDEδ also interacts with the retinitis pigmentosa guanine receptor (RPGR; Linari et al., 1999b) and a number of prenylated GTPases (HanzalBayer et al., 2002). RPGR contains a domain homologous to the GEF for the GTPase Ran (RCC1, regulator of chromosome condensation; Meindl et al., 1996; Renault et al., 1998). The finding that PDEδ binds to RPGR (Linari et al., 1999b) and Arl3 (Linari et al., 1999a) in the absence of any post-translational modifications indicates that PDEδ binding is not prenylation-dependent, despite the presence of a hydrophobic pocket similar to the one in Rho GDI that binds prenyl groups. Because PDEδ not only binds the subunits of PDE6 but also several other proteins (Arl2, Arl3, Rab13, RPGH, H-Ras, Rheb, Rho6 and Gαi1; HanzalBayer et al., 2002), it was suggested that PDEδ is not a genuine subunit of the PDE holoenzyme, but merely an associated protein involved in protein transport. In addition, the structure of PDEδ in a complex with Arl2•GTP (see below) led to the hypothesis that PDEδ is a transport factor for prenylated proteins, and that Arl2 and Arl3 regulate the PDEδ-mediated transport (Hanzal-Bayer et al., 2002).

5.

ARL4

Arl4, Arl6/Arf4-like and Arl7 represent a subgroup of the Arl proteins that share sequence similarities and the presence of C-terminal nuclear localization signals (NLSs). Arl4 is highly expressed in testis and is involved in the differentiation of spermatocytes. It has been suggested that Arl4 is involved in the regulation of nuclear protein transport during cell division or differentiation.

12 336

5.1

SCHÜRMANN AND JOOST

Arl4 is involved in male germ cell development

Arl4 was cloned as a GTPase that is abundantly expressed in differentiated cells of the pre-adipocyte cell line 3T3-L1, but not in undifferentiated cells (Schürmann et al., 1994). The GTPase was found predominantly in testis; lower levels of Arl4 mRNA were detected in spleen and intestine, brain, heart, fat, liver, lung and thymus (Jacobs et al., 1998). The subcellular distribution of Arl4 differs from other Arf and most Arl proteins because it is in the nucleus and nucleoli (Lin et al., 2000). On the basis of sequence similarity, Arl4, Arl6 and Arl7, form a subgroup within the ARF family (Figure 1). In addition, a striking characteristic of these three GTPases is their C-terminal NLS and unusually high guanine nucleotide exchange rate. Jacobs et al. (1999) demonstrated that the C-termini of Arl4, Arl6 and Arl7 are capable of targeting green fluorescent protein (GFP) to the nucleus of COS-7 cells, and suggested that these GTPases are involved in the regulation of protein transport to or from the nucleus. Transport through the nuclear pore complex is GTP-dependent and is regulated by Ran (Görlich and Mattaj, 1996; Nigg, 1997). Cargo proteins bind to importin-α by their NLS, and Ran regulates the translocation and subsequent dissociation of the complex inside the nucleus. Data from yeast two-hybrid and in vitro protein interaction assays indicated that Arl4 does not bind importin-α in a GTP-dependent manner (Lin et al., 2000). Thus, Arl4 appears to be a cargo protein rather that a Ran-like regulator of nuclear import. Several lines of evidence indicate that Arl4 is involved in male germ cell development. In situ hybridization showed that Arl4 is expressed in testis of pubertal and adult rats, but not in testis of prepubertal animals (Jacobs et al., 1998). In the mouse, Arl4 expression starts at day 16 when spermatogenesis proceeds to the late pachytene. In adult testis, Arl4 was detected immunohistochemically in pre- and post-meiotic cells, spermatocytes, and spermatids, but not in spermatogonia and mature spermatozoa (Schürmann et al., 2002). Deletion of the Arl4 gene produced a significant reduction in testis weight and number of spermatozoa without affecting their mobility or reproductive function. Thus, Arl4 is required for progression through meiosis, but not for the division of spermatogonia or the function of mature spermatozoa (Schürmann et al., 2002). Because of the differential expression of Arl4 in 3T3-L1 adipocytes (Schürmann et al., 1994), a defect in the development of adipose tissues was expected in the Arl4-/- mice. Therefore, the weight of epididymal fat pads and several metabolic parameters were analyzed in Arl4+/+ and Arl4-/- mice. However, no differences in epididymal fat pad weight, blood glucose levels, serum cholesterol, serum triglycerides, or blood cell count were detected.

ARF-LIKE PROTEINS

13 337

Lin et al. (2000) have shown that Arl4 is developmentally expressed during mouse embryogenesis. Arl4 mRNA appears transiently at day 8.5 to 10.5 in a rostro-caudal direction, suggesting that Arl4 is involved in somite formation and central nervous system differentiation. However, because mice lacking the Arl4 gene were viable with normal growth and no apparent abnormality it was concluded that Arl4 is dispensable for embryonic development (Schürmann et al., 2002). The organization of the Arl4 gene appears unique among the Arf family in that its coding region, its 3’ UTR, and a considerable portion of its 5’ UTR are located on a single exon. This observation suggested that the Arl4 gene has evolved by retrotransposition, and that its evolution is a relatively recent event (Jacobs et al., 1998). Interestingly, two separate promoters direct the expression of Arl4 in different tissues. The downstream promoter appeared to be active in most tissues, whereas the upstream promoter specifically drives the large expression in testicular germ cells (Jacobs et al., 1998).

6.

ARL5

Arl5 was identified by PCR cloning on the basis of its sequence similarity with members of the Arf family (Breiner et al., 1996). Arl5 exhibits 48.2% and 49.2% identity with Arf1 and Arl1, respectively, and is mainly expressed in brain, thymus and intestine. No other information (e.g., subcellular localization, nucleotide binding properties, function) is yet available.

7.

ARL6/ARF4L

Arl6, also named Arf4-like protein (Arf4L; Smith et al., 1995; accession number L38490), exhibits the highest similarity to Arl4 (61%) and Arl7 (59%) (Jacobs et al., 1999). Arl6/Arf4L binds guanine nucleotides rapidly and in a Mg2+-dependent manner. The exchange rate of Arl6/Arf4-like protein alone was considerably higher than that of Arf1 in the presence of its exchange factor Sec7-2/ARNO. Like its close relatives Arl4 and Arl7, Arl6/Arf4L contains an NLS in its C-terminus, suggesting that it too is located within the nucleus (Jacobs et al., 1999). Unfortunately, a different Arl has also been named Arl6 (accession number AF031903). The amino acid sequence of this GTPase is 41% identical with that of Arf1 (and only 34% identical to Arl6/Arf4L). Its cDNA was identified by differential display (Ingley et al., 1999). The gene

14 338

SCHÜRMANN AND JOOST

exhibits an expression pattern that is unique within the Arf and Arl family, with highest mRNA levels present in brain and kidney. In addition, this Arl6 transcript is up regulated during erythropoietin-induced differentiation of erythroid cells and down regulated during interleukin 6-induced macrophage differentiation, suggesting that it plays a role in hematopoietic development. The protein was found mainly in the cytoplasm (Ingley et al., 1999). Stimulation of cells with GTPγS produced an increase in membrane association, indicating that the protein may cycle between the cytoplasm and membranes in a GTP-dependent manner. A yeast two-hybrid screen and coimmunoprecipitation revealed that Arl6 interacts with Sec61β, a subunit of the core component of the ER translocon (Ingley et al., 1999).

8.

ARL7 IS PRESUMABLY INVOLVED IN PROTEIN TRANSPORT INTO OR OUT OF THE NUCLEUS

Arl7 was cloned by Jacobs et al. (1999), and shown to be a close relative of Arl4 (71% identical amino acids) and Arl6/Arf4L (59% identical amino acids). The C-terminal NLS of Arl7 targets a GFP fusion protein to the nucleus of transfected COS-7 cells. Like Arl4 and Arl6/Arf4L, Arl7 binds GTPγS rapidly and in a Mg2+-dependent manner. Arl7 mRNA expression was highest in brain, with lower levels seen in spleen, thymus, esophagus, stomach, intestine and uterus (Jacobs et al., 1999).

9.

ARF-RELATED PROTEIN 1 (ARFP1)

Arfrp1 exhibits the lowest similarity to Arf proteins and several differences to Arf and Arl proteins. It inhibits activation of phospholipase D (PLD) as a result of its interaction with an Arf-specific activator (cytohesin 1, an Arf GEF). Because deletion of the Arfrp1 gene resulted in marked defects in early gastrulation, it was speculated that Arfrp1 is involved in adhesiondependent morphogenesis, cytoskeletal reorganization, or development of cell polarity, occurring during this stage of embryogenesis.

9.1

ARFRP1 is essential in embryogenesis

Arfrp1 is a 25 kDa GTPase that exhibits somewhat lower similarity to Arf and Arf-like proteins (33 and 39% identical amino acids to Arf1 and Arl3, respectively) than do the Arls. Guanine nucleotide exchange on recombinant Arfrp1 is slower than most GTPases, requiring an hour to reach steady state in vitro. Arfrp1 shares two features with most Ras and GĮ

ARF-LIKE PROTEINS

15 339

proteins that distinguish it from other members of the Arf family; Arfrp1 can hydrolyze GTP in the absence of a GAP (0.093 min-1), and it lacks the site for N-myristoylation (glycine 2). Arfrp1 was detected in the plasma membrane and was absent from the cytosol (Schürmann et al., 1995) of 3T3L1 cells. The mechanism of this tight attachment to membranes is unknown. Arfrp1 appears to be involved in a pathway that can inhibit Arf activities, e.g., activation of PLD (Schürmann et al., 1999). In a two-hybrid screen designed to identity proteins interacting with Arfrp1, a partial sequence encoding the N-terminus and most of the Sec7 domain of the Arf-specific guanine nucleotide exchange factor cytohesin 1 was isolated. The interaction between Arfrp1 and the Sec7 domain was GTP-dependent and resulted in the inhibition of the Arf/Sec7-dependent activation of PLD in a system of isolated membranes. In addition, transfection of HEK-293 cells with a constitutively active mutant of Arfrp1 inhibited the stimulation of PLD by the muscarinic acetylcholine receptor-3, and the translocation of Arf from the cytosol to membranes. The closest relative of Arfrp1 in yeast is a protein with 43% identity that has been identified by Huang et al. (1999) and designated ScArl3 (see Chapter 1). Like ScArl1 and ScArf3, ScArl3 is not essential for cell viability. However, disruption of ScArl3 resulted in cold-sensitive growth. At the permissive temperature (37°C), maturation of vacuolar proteins was retarded and, at the non-permissive temperature (15°C), impairment in fluid phase endocytosis was observed. Unlike mammalian Arfrp1, ScArl3 was detected in the cytosol and in part associated with the ER-nuclear envelope structure, suggesting that the GTPase is involved in the regulation of vesicular transport pathways (Huang et al., 1999). The murine Arfrp1 gene spans approximately 7 kb and contains 8 exons. The proximal 5’-flanking regions of mouse and human Arfrp1 lack a TATA box and a CAAT box, are highly GC rich and contain potential transcription factor binding sites. Sequence analysis of human Arfrp1 showed that its 5’flanking region contains the first exon of another gene (DJ583P15.3, ensemble data base) on the opposite strand. Promoter analysis revealed that the intergenic region between both genes (54 bp) exhibits bi-directional promoter activity. However, deletion analysis demonstrated that transcription of both genes is regulated by different cis-elements. Mutational analysis and electrophoretic mobility shift assays indicated that two short cRel- and cEts1-like elements in the 5’-flanking region of Arfrp1 are critical for the regulation of Arfrp1 expression (Mueller et al., 2002a). Arfrp1 null-mutant mice, generated by gene targeting in embryonic stem cells, were the first knockout mice of an Arf or Arl protein (Mueller et al., 2002b). Heterozygous Arfrp1 mutants developed normally, whereas homozygosity for the mutant allele resulted in embryonic lethality. Cultured

16 340

SCHÜRMANN AND JOOST

homozygous Arfrp1-null mutant blastocysts were indistinguishable from wild-type blastocysts, suggesting that disruption of Arfrp1 did not alter the growth of stem cells or the development of the blastocysts. Furthermore, blastocysts implanted in vivo and formed egg-cylinder-stage embryos that appeared normal until day 5. At day 6-6.5, the time of early gastrulation, Arfrp1-/- embryos exhibited profound alterations of the distal part of the egg cylinder. Rounded pyknotic cells were found within this area of the egg cylinder in the proamniotic cavity, and some of these were only loosely attached to the ectodermal cell layer. In contrast, the proximal extraembryonic part of the embryo appeared normal (Figure 4). In order to determine whether the morphologically abnormal cells within the proamniotic cavity were apoptotic, TUNEL (terminal deoxy-nucleotidyltransferase-mediated UTP end labeling) assays were performed with sections

Figure 4. Schematic overview of the egg cylinder stage (E6.5) and the primitive streak stage (E7.5) in control and Arfrp1-/- mouse embryos. (PS = primitive streak).

adjacent to those stained for histological examination. This analysis demonstrated that the pyknotic epiblast cells in the area of the primitive

ARF-LIKE PROTEINS

17 341

streak region were TUNEL-positive and therefore apoptotic (Figure 4; Mueller et al., 2002b). Early gastrulation is a stage during embryogenesis in which a set of morphogenetic movements coupled with cell proliferation and differentiation is necessary to convert the two germ layers into three; the ectoderm, the mesoderm and the endoderm (Figure 4). A likely explanation for the embryonic lethality is that deletion of Arfrp1 prevents the proper integration of dividing cells into the ectodermal cell layer. We speculate that Arfrp1 plays a critical role in processes of adhesion-dependent morphogenesis, cytoskeletal reorganization, or development of cell polarity, which are known to occur in early gastrulation (Mueller et al., 2002b).

10.

SAR1 REGULATES VESICLE BUDDING FROM THE ENDOPLASMIC RETICULUM

Sar1 exhibits weak similarity with the Arf family (30% identity with Arf1 and Arf5). The GTPase controls protein-protein and protein-lipid interactions that direct budding from the endoplasmic reticulum (Springer et al., 1999). In S. cerevisiae, only a single isotype was identified (Nakano and Muramatsu, 1989), whereas two isotypes, Sar1a and Sar1b (91% identity) were found in mammalian cells (Kuge et al., 1994). Unlike Arf proteins, Sar1 lacks the myristoylated glycine residue at position two, and does not undergo any other lipid modification. In NRK and CHO (Chinese hamster ovary) cells, Sar1 was detected in the juxtanuclear Golgi region; in rat pancreatic insulin cells, the protein was abundant at smooth transitional elements of the endoplasmic reticulum (ER), and at vesicular profiles in the proximal Golgi cisternae facing the transitional region (Kuge et al., 1994). Sar1 is an essential component of COPII vesicle coats, and regulates the selective export of membrane-bound and soluble proteins from the ER. The COPII coat forms by the sequential recruitment of Sar1 and two large complexes, Sec23/24 and Sec13/31 (Barlowe et al., 1993; Matsuoka et al., 1998), and is necessary for the deformation of ER membranes into vesicles as well as for the selection of cargo molecules (Schekman and Orci, 1996; Rothman and Wieland, 1996). Sar1•GTP exhibits the unique ability to interact directly with both lipid membranes and Sec23/24; therefore, the switch between the GDP and GTP-bound forms controls assembly and disassembly of COPII coats. Assembly is driven by the exchange of GDP for GTP, which is catalyzed by Sec12. Sec12 is a transmembrane ER glycoprotein containing a 40 kDa cytoplasmic domain and a ∼10 kDa lumenal domain. The cytosolic domain operates independently of membranes as a Sar1-specific GEF (Barlow and Schekman, 1993;

18 342

SCHÜRMANN AND JOOST

Weissmann et al., 2001). Activation of Sar1 by Sec12 leads to the formation of ER transitional elements (Aridor et al., 2001), recruitment of cargo to prebudding complexes, and assembly of the Sec23/24 and Sec 13/31 complexes, leading to vesicle budding from the ER (Aridor et al., 1998; Kuehn et al., 1998; Balch et al., 1994). The disassembly of COPII coats requires GTP-hydrolysis on Sar1, which is stimulated by the COPII coat subunit Sec23/24 (a Sar1 GAP), to promote vesicle uncoating and the subsequent delivery of cargo to the Golgi complex through Rab1-dependent tethering and fusion factors (Allan et al., 2000; Moyer et al., 2001).

11.

THE STRUCTURE OF ARF-LIKE PROTEINS

The conformational changes that accompany the binding of GDP and GTP can alter the affinity of the GTPases for proteins, lipids, and membranes. These conformational changes are centered in two critical regions, referred to as switch I and switch II (Lacal and McCormick, 1993). Residues in these switches make direct contact with effectors and GAPs in a variety of GTPases, including Arfs (Goldberg, 1998; Goldberg, 1999). The interactions and the mechanism of regulation of Arf proteins differ from those of other GTPases of the Ras superfamily. Arf proteins have an additional nucleotide-sensitive region, the “myristoyl switch”, an extension in the N-terminus and a covalently attached myristate which cooperate to coordinate GTP binding (activation) and membrane association (Goldberg, 1999). In addition, there is evidence that the N-terminus may modify effector binding or activity (Kahn et al. 1992; Zhu et al., 2000). The Nterminal region (residues 1-17) of Arf is sufficient to confer Arf activity and increase GTPase activity to an Arl protein, despite the distance between the N terminus and the switch or nucleotide binding region (Kahn et al., 1992). Amor, et al. (2001) analyzed the structure of yeast Arf2 and Arl1 and reported that the secondary structure of GDP-bound yeast Arf2, yeast Arl1 and human Arf1 were essentially identical. Their structures include seven βstrands, six α-helices, and 12 connecting loops. The core of the protein is an invariant six-stranded β-sheet surrounded by six α-helices. The seventh strand forms an edge to the core and corresponds to the switch I region (Lacal and McCormick, 1993). The N-terminal α-helix (αNT) is wedged between the C-terminal helix α-E and loop L-2/3 which connects strands β2 and β3 (also referred to as the interswitch region, ISR). The presence of αNT is in contrast to the rather undefined secondary fold seen for the Nterminal sequence in the Arl3 model (see Figure 5; Hillig et al., 2000). Like Arl3•GDP (Hillig et al., 2000), yeast Arl1•GDP appears to form a covalent

ARF-LIKE PROTEINS

19 343

homodimer, linked by a disulfide bond between the two Cys81 residues present in switch II. However, dimerization of Arl1 does not

Figure 5. Ribbon diagram of Arl3•GDP and Arl2•GTP (Hanzal-Bayer et al., 2002), with switches in blue, interswitch region in green, N-terminus in yellow, C-terminal helix in magenta, nucleotides in orange, and Mg2+ as red sphere (see arrow).

20 344

SCHÜRMANN AND JOOST

appear to be essential for its function, since knock-in of the Arl1-S31 mutant produced no effect in S. cerevisiae. In contrast to Arf1, Arf2 and Arf6, the electrostatic surface of yeast Arl1 is characterized by an overall negative electrostatic potential. Arl1 is Nmyristoylated but, in contrast to Arf proteins, no GTP-dependent binding to membranes or lipid vesicles has been detected. The difference in the electrostatic potential led to the hypothesis that the binding of Arl1 to membranes is dependent on protein-protein interactions at the membrane (Amor et al., 2001). The comparison of the structures of Arfs with that of Arl1 revealed differences in the N-terminal region. The 17 N-terminal amino acids of Arfs, which form the αNT and loop (L-NT), are involved in membrane binding, nucleotide affinity, phospholipid dependence of GTP-binding, and nucleotide hydrolysis (Kahn et al., 1992; Kahn et al., 1996). The N-terminal helix and the length of the loop L-NT can affect the structure of switch I and the interswitch region and therefore alter the nucleotide binding (Kahn et al., 1992; Zhu et al., 2000). In comparison to Arf1 and Arf6, the loop L-NT in yeast Arl1 is much longer, and this difference was proposed to be responsible for different biophysical and biochemical properties of Arf and Arl proteins (Amor et al., 2001). Like Arl1, the structure of Arl3 predicts that the N-terminal extension folds into an elongated loop region, suggesting that Arl3 releases its Nterminus and undergoes a β-sheet register shift upon the binding of GTP. Structure as well as kinetic experiments demonstrated high affinity GDP (but not GTP) binding in the absence of Mg2+. The disturbed magnesium binding site and the lack of Mg2+-dependence of GDP binding distinguish Arl2 and Arl3 from Arf proteins. In contrast to Arf proteins, the conformational change of Arl2/3•GTP did not require the presence of membranes and might thus be energetically unfavored (Hillig et al., 2000). Recently, Hanzal-Bayer, et al. (2002) presented the crystal structure of full-length Arl2•GTP in a complex with its effector PDEδ. In this complex, Arl2•GTP shows a typical G domain fold with six-stranded β sheet surrounded by five α helices, with an additional α helix at the N terminus. Because Arl3 is the closest relative of Arl2 (53% identity, Figure 1) and also a binding partner of PDEδ (Figure 2), the authors compared the structure of Arl2•GTP with Arl3•GDP (Figure 5). During nucleotide exchange, the interswitch region of Arl2 has to undergo a β sheet register shift, with β strands β2 and β3 sliding by two residues relative to the remaining β sheet. This shift results in displacement and conformational change of the Nterminus, which appears to interact with a membrane, as observed with Arf proteins (Hanzal-Bayer et al., 2002).

ARF-LIKE PROTEINS

21 345

The interaction of Arl2•GTP and PDEδ is strikingly GTP-dependent and involves the switch region of the molecule (Hanzal-Bayer et al., 2002). It was concluded that PDEδ is an effector of Arl2. Interestingly, the authors concluded that the structure of PDEδ is closely related to Rho-GDI and features a hydrophobic pocket similar to that found in Rho-GDIs. This conclusion provides a structural explanation for the ability of PDEδ to extract the catalytic subunit of PDE from the membrane, and to generate a cytosolic pool of the enzyme (Cook et al., 2001). A GDI-like function was also supported by the identification of four new putative targets that bind to PDEδ: H-Ras, Rheb, Rho6 and Gαi1. The C-terminus of H-Ras is absolutely required for the binding to PDEδ (Hanzal-Bayer et al., 2002). Thus, PDEδ may function as a transport factor for H-Ras and other prenylated proteins, and Arl2 and Arl3 might mediate their release and/or uptake.

REFERENCES Allan, B. B., Moyer, B. D., and Balch, W. E. (2000). Rab1 recruitment of p115 into a cisSNARE complex: programming budding COP-II vesicles for fusion. Science, 289, 444448. Amor, J. C., Horton, J. R., Zhu, X., Wang, Y., Sullards, C., Ringe, D., Cheng, X., and Kahn, R. A. (2001). Structures of yeast Arf2 and Arl1: distinct roles for the N terminus in the structure and function of Arf family GTPases. J. Biol. Chem., 276, 42477-42484 Aridor, M., Fish, K. N., Bannykh, S., Weissman, J., Roberts, T. H., Lippincott-Schwartz, J., and Balch, W. E. (2001). The Sar1 GTPase coordinates biosynthetic cargo selection with endoplasmic reticulum export site assembly. J. Cell Biol., 152, 213-229. Aridor, M., Weissman, J., Bannykh, S., Nuoffer, C., and Balch, W. E. (1998). Cargo selection by the COP-II budding machinery during export from the ER. J. Cell Biol., 141, 61-70. Balch, W. E., McCaffery, J. M., Plutner, H., and Farquhar, M. G. (1994). Vesicular stomatitis virus glycoprotein is sorted and concentrated during export from the endoplasmic reticulum. Cell, 76, 841-852. Barlowe, C., d'Enfert, C., and Schekman, R. (1993). Purification and characterization of SAR1p, a small GTP-binding protein required for transport vesicle formation from the endoplasmic reticulum. J. Biol. Chem., 268, 873-879. Barlowe, C., and Schekman, R. (1993). SEC12 encodes a guanine-nucleotide-exchange factor essential for transport vesicle budding from the ER. Nature, 365, 347-349. Bhamidipati, A., Lewis, S. A., and Cowan, N. J. (2000). ADP ribosylation factor-like protein 2 (Arl2) regulates the interaction of tubulin-folding cofactor D with native tubulin. J. Cell. Biol., 149, 1087-1096. Boman, A. L., Kuai, J., Zhu, X., Chen, J., Kuriyama, R., and Kahn, R. A. (1999 ). Arf proteins bind to mitotic kinesin-like protein 1 (MKLP1) in a GTP-dependent fashion. Cell Motil. Cytoskeleton., 44, 119-132. Breiner, M., Schürmann, A., Becker, W., and Joost, H.-G. (1996). Cloning of a novel member (Arl5) of the Arf family of Ras related GTPases. Biochim. Biophys. Acta., 1308, 1-6.

22 346

SCHÜRMANN AND JOOST

Brown, H. A., Gutowski, S., Moomaw, C. R., Slaughter, C., and Sternweis, P. C. (1993). ADP-ribosylation factor, a small GTP-dependent regulatory protein, stimulates phospholipase D activity. Cell, 75, 1137-1144. Cavenagh, M. M., Breiner, M., Schürmann. A., Rosenwald, A. G., Terui, T., Zhang, C., Randazzo, P. A., Adams, M., Joost, H.-G., and Kahn, R. A. (1994). ADP-ribosylation factor (Arf)-like 3, a new member of the Arf family of GTP-binding proteins cloned from human and rat tissues. J. Biol. Chem., 269, 18937-18942. Clark, J., Moore, L., Krasinskas, A., Way, J., Battey, J., Tamkun, J., and Kahn, R. A. (1993). Selective amplification of additional members of the ADP-ribosylation factor (Arf) family: cloning of additional human and Drosophila Arf-like genes. Proc. Natl. Acad. Sci., USA, 90, 8952-8956. Cockcroft, S., Thomas, G. M., Fensome, A., Geny, B., Cunningham, E., Gout, I., Hiles, I., Totty, N. F., Truong, O., and Hsuan, J. J. (1994). Phospholipase D: a downstream effector of Arf in granulocytes. Science, 263, 523-526. Cook, T. A., Ghomashchi, F., Gelb, M. H., Florio, S. K., and Beavo, J. A. (2001). The delta subunit of type 6 phosphodiesterase reduces light-induced cGMP hydrolysis in rod outer segments. J. Biol. Chem., 276, 5248-5255. Dascher, C., and Balch, W. E. (1994). Dominant inhibitory mutants of Arf1 block endoplasmic reticulum to Golgi transport and trigger disassembly of the Golgi apparatus. J. Biol. Chem., 269, 1437-1448. Donaldson, J. G., Kahn, R. A., Lippincott-Schwartz, J., and Klausner, R. D. (1991). Binding of Arf and beta-COP to Golgi membranes: possible regulation by a trimeric G protein. Science, 254, 1197-1199. D’Souza-Schorey, C., Boshans R. L., McDonough, M., Stahl, P. D., and Van Aelst, L. (1997). A role for POR, a Rac1-interacting protein, in ARF6-mediated cytoskeletal rearrangements. EMBO J., 16, 5445-5454. Fiore, C., Trezeguet, V., Le Saux, A., Roux, P., Schwimmer, C., Dianoux, A. C., Noel, F., Lauquin, G. J., Brandolin, G., and Vignais, P. V. (1998). The mitochondrial ADP/ATP carrier: structural, physiological and pathological aspects. Biochimie, 80, 137-150. Florio, S. K., Prusti, R. K., and Beavo, J. A. (1996). Solubilization of membrane-bound rod phosphodiesterase by the rod phosphodiesterase recombinant delta subunit. J. Biol. Chem., 271, 24036-24047. Gillespie, P. G., Prusti, R. K., Apel, E. D., and Beavo, J. A. (1989). A soluble form of bovine rod photoreceptor phosphodiesterase has a novel 15-kDa subunit. J. Biol. Chem., 264, 12187-12193. Görlich, D., and Mattaj, I. W. (1996). Nucleocytoplasmic transport. Science, 271, 1513-1518. Goldberg, J. (1998). Structural basis for activation of Arf GTPase: mechanisms of guanine nucleotide exchange and GTP-myristoyl switching. Cell, 95, 237-248. Goldberg, J. (1999). Structural and functional analysis of the Arf1-ArfGAP complex reveals a role for coatomer in GTP hydrolysis. Cell, 96, 893-902. Graham, B. H., Waymire, K. G., Cottrell, B., Trounce, I. A., MacGregor, G. R., and Wallace, D. C. (1997). A mouse model for mitochondrial myopathy and cardiomyopathy resulting from a deficiency in the heart/muscle isoform of the adenine nucleotide translocator. Nat. Genet., 16, 226-234. Hanzal-Bayer, M., Renault, L., Roversi, P., Wittinghofer, A., and Hillig R. C. (2002). The complex of Arl2-GTP and PDE delta: from structure to function. EMBO J., 21, 20952106. Hillig, R. C., Hanzal-Bayer, M., Linari, M., Becker, J., Wittinghofer, A., and Renault, L. (2000). Structural and biochemical properties show Arl3-GDP as a distinct GTP binding protein. Structure Fold. Des., 8, 1239-1245.

ARF-LIKE PROTEINS

23 347

Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H., Watanabe, H., Kawamoto, K., Nakayama, K., Morris, A. J., Frohman, M. A., and Kanaho, Y. (1999). Phosphatidylinositol 4-phosphate 5-kinase alpha is a downstream effector of the small G protein Arf6 in membrane ruffle formation. Cell, 99, 521-532. Hoyt, M. A., Stearns, T., and Botstein, D. (1990). Chromosome instability mutants of Saccharomyces cerevisiae that are defective in microtubule-mediated processes. Mol. Cell Biol., 10, 223-234. Huang, C. F., Buu, L. M., Yu, W. L., and Lee, F. J. (1999). Characterization of a novel ADPribosylation factor-like protein (yArl3) in Saccharomyces cerevisiae. J. Biol. Chem., 274, 3819-3827. Ingley, E., Williams, J. H., Walker, C. E., Tsai, S., Colley, S., Sayer, M. S., Tilbrook, P. A., Sarna, M., Beaumont, J. G., and Klinken, S. P. (1999). A novel ADP-ribosylation like factor (Arl-6), interacts with the protein-conducting channel SEC61beta subunit. FEBS Lett., 459, 69-74. Jacobs, S., Schilf, C., Fliegert, F., Koling, S., Weber, Y., Schürmann, A., and Joost, H.-G. (1999). ADP-ribosylation factor (Arf)-like 4, 6, and 7 represent a subgroup of the Arf family characterization by rapid nucleotide exchange and a nuclear localization signal. FEBS Lett., 456, 384-388. Jacobs, S., Schürmann, A., Becker, W., Böckers, T. M., Copeland, N.G., Jenkins, N. A., and Joost, H.-G. (1998). The mouse ADP-ribosylation factor-like 4 gene: two separate promoters direct specific transcription in tissues and testicular germ cell. Biochem. J., 335, 259-265. Kahn, R. A., and Gilman, A. G. (1984). ADP-ribosylation of Gs promotes the dissociation of its alpha and beta subunits. J. Biol. Chem., 259, 6235-6240. Kahn, R. A., and Gilman, A. G. (1986). The protein cofactor necessary for ADP-ribosylation of Gs by cholera toxin is itself a GTP binding protein. J. Biol. Chem., 261, 7906-7911. Kahn, R. A., Kern, F. G., Clark, J., Gelmann, E. P., and Rulka, C. (1991). Human ADPribosylation factors. A functionally conserved family of GTP-binding proteins. J. Biol. Chem., 266, 2606-2614. Kahn, R. A., Randazzo, P., Serafini, T., Weiss, O., Rulka, C., Clark, J., Amherdt, M., Roller, P., Orci, L., and Rothman J. E. (1992). The amino terminus of ADP-ribosylation factor (Arf) is a critical determinant of Arf activities and is a potent and specific inhibitor of protein transport. J. Biol. Chem., 267, 13039-13046. Kahn, R. A., Terui, T., and Randazzo, P. A. (1996). Effects of acid phospholipids on Arf activities: potential roles in membrane traffic. J. Lipid Mediat. Cell. Signal., 14, 209-214. Kanoh, H., Williger, B. T., and Exton, J. H. (1997). Arfaptin 1, a putative cytosolic target protein of ADP-ribosylation factor, is recruited to Golgi membranes. J. Biol. Chem., 272, 5421-5429. Klausner, R. D., Donaldson, J. G., and Lippincott-Schwartz, J. (1992). Brefeldin A: insights into the control of membrane traffic and organelle structure. J. Cell Biol., 116, 1071-1080. Kuehn, M. J., Herrmann, J. M., and Schekman, R. (1998). COP-II-cargo interactions direct protein sorting into ER-derived transport vesicles. Nature, 391, 187-190. Kuge, O., Dascher, C., Orci, L., Rowe, T., Amherdt, M., Plutner, H., Ravazzola, M., Tanigawa, G., Rothman, J. E., and Balch, W. E. (1994). Sar1 promotes vesicle budding from the endoplasmic reticulum but not Golgi compartments. J. Cell Biol., 25, 51-65. Kuriyama, R., Gustus, C., Terada, Y., Uetake, Y., and Matuliene, J. (2002). CHO1, a mammalian kinesin-like protein, interacts with F-actin and is involved in the terminal phase of cytokinesis. J. Cell Biol., 156, 783-790. Lacal, J. C. and McCormick, F. (1993). The Ras superfamily of GTPases. Boca Raton, FL. CRC Press.

24 348

SCHÜRMANN AND JOOST

Lee, F. J., Moss, J., and Vaughan, M. (1992). Human and Giardia ADP-ribosylation factors (Arfs) complement Arf function in Saccharomyces cerevisiae. J. Biol. Chem., 267, 2444124445. Levy, S. E., Chen, Y. S., Graham, B. H., and Wallace, D. C. (2000). Expression and sequence analysis of the mouse adenine nucleotide translocase 1 and 2 genes. Gene, 254, 57-66. Lewis, S. A., Tian, G., and Cowan, N. J. (1997). Thel alpha and beta tubulin folding pathways. Trends Cell Biol., 7, 479-485. Lin, C. Y., Huang, P. H., Liao, W. L., Cheng, H. J., Huang, C. F., Kuo, J. C., Patton, W. A., Massenburg, D., Moss, J., and Lee, F. J. (2000). Arl4, an Arf-like protein that is developmentally regulated and localized to nuclei and nucleoli. J. Biol. Chem., 275, 37815-37823. Linari, M., Hanzal-Bayer, M., and Becker, J. (1999a). The delta subunit of rod specific cyclic GMP phosphodiesterase, PDE delta, interacts with the Arf-like protein Arl3 in a GTP specific manner. FEBS Lett., 458, 55-59. Linari, M., Ueffing, M., Manson, F., Wright, A., Meitinger, T., and Becker, J. (1999b). The retinitis pigmentosa GTPase regulator, RPGR, interacts with the delta subunit of rod cyclic GMP phosphodiesterase. Proc. Natl. Acad. Sci., USA, 96, 1315-1320. Lowe, S. L., Wong, S. H., and Hong, W. (1996). The mammalian Arf-like protein 1 (Arl1) is associated with the Golgi complex. J. Cell. Sci., 109, 209-220. Lu, L., Horstmann, H., Ng, C., and Hong, W. (2001). Regulation of Golgi structure and function by Arf-like protein 1 (Arl1). J. Cell. Sci., 114, 4543-4555. Marzesco, A. M., Galli, T., Louvard, D., and Zahraoui, A. (1998). The rod cGMP phosphodiesterase delta subunit dissociates the small GTPase Rab13 from membranes. J. Biol. Chem., 273, 22340-22345. Matsuoka, K., Orci, L., Amherdt, M., Bednarek, S. Y., Hamamoto, S., Schekman, R., and Yeung, T. (1998). COP-II-coated vesicle formation reconstituted with purified coat proteins and chemically defined liposomes. Cell, 93, 263-275. Meindl, A., Dry, K., Herrmann, K., Manson, F., Ciccodicola, A., Edgar, A., Carvalho, M. R., Achatz, H., Hellebrand, H., Lennon, A., Migliaccio, C., Porter, K., Zrenner, E., Bird, A., Jay, M., Lorenz, B., Wittwer, B., D'Urso, M., Meitinger, T., and Wright A. (1996). A gene (RPGR) with homology to the RCC1 guanine nucleotide exchange factor is mutated in Xlinked retinitis pigmentosa (RP3). Nat. Genet., 13, 35-42. Moyer, B. D., Allan, B. B., and Balch, W. E. (2001). Rab1 interaction with a GM130 effector complex regulates COP-II vesicle cis--Golgi tethering. Traffic 2, 268-276. Mueller, A. G., Joost, H.-G., and Schürmann A. (2002a). Mouse Arf-Related Protein 1: Genomic Organization and Analysis of Its Promoter. Biochem. Biophys. Res. Commun. 292, 113-120. Mueller, A. G., Moser, M., Kluge, R., Leder, S., Blum, M., Büttner, R., Joost, H.-G., and Schürmann, A. (2002b). Embryonic lethality caused by apoptosis during gastrulation in mice lacking the gene of the ADP-ribosylation factor-related protein 1. Mol. Cell. Biol., 22, 1488-1494. Murtagh, J. J., Jr, Lee, F. J., Deak, P., Hall, L. M., Monaco, L., Lee, C. M., Stevens, L. A., Moss, J., and Vaughan, M. (1993). Molecular characterization of a conserved, guanine nucleotide-dependent ADP-ribosylation factor in Drosophila melanogaster. Biochemistry, 32, 6011-6018. Nakano, A. and Muramatsu, M. (1989). A novel GTP-binding protein, Sar1p, is involved in transport from the endoplasmic reticulum to the Golgi apparatus. J. Cell Biol., 109, 26772691. Nigg, E. A. (1997). Nucleocytoplasmic transport: signals, mechanisms and regulation. Nature, 386, 779-787

ARF-LIKE PROTEINS

25 349

Radcliffe, P. A., Vardy, L., and Toda, T. (2000). A conserved small GTP-binding protein Alp41 is essential for the cofactor-dependent biogenesis of microtubules in fission yeast. FEBS Lett., 468, 84-88. Renault, L., Nassar, N., Vetter, I., Becker, J., Klebe, C., Roth, M., and Wittinghofer, A. (1998). The 1.7 A crystal structure of the regulator of chromosome condensation (RCC1) reveals a seven-bladed propeller. Nature, 392, 97-101. Rothman, J. E. and Wieland, F. T. (1996). Protein sorting by transport vesicles. Science, 272, 227-234. Schekman, R. and Orci, L. (1996). Coat proteins and vesicle budding. Science, 271, 15261533. Schürmann, A., Breiner, M., Becker, W., Huppertz, C., Kainulainen, H., Kentrup, H. and Joost, H.-G. (1994). Cloning of two novel ADP-ribosylation factor-like proteins and characterization of their differential expression in 3T3-L1 cells. J. Biol. Chem. 269, 15683-15688. Schürmann, A., Koling, S., Jacobs, S., Saftig, P., Krauss, S., Wennemuth, G., Kluge, R., and Joost, H.-G. (2002). Reduced sperm count and normal fertility in male mice with targeted disruption of the ADP-ribosylation factor-like (Arl4) gene. Mol. Cell. Biol., 22, 27612768. Schürmann, A., Massmann, S., and Joost, H.-G. (1995). ARP is a plasma membraneassociated Ras-related GTPase with remote similarity to the family of ADP-ribosylation factors. J. Biol. Chem., 270, 30657-30563. Schürmann, A., Schmidt, M., Asmus, M., Bayer, S., Fliegert, F., Koling, S., Massmann, S., Schilf, C., Subauste, M.C., Voss, M., Jakobs, K.H., and Joost, H.-G. (1999). The ADPribosylation factor (Arf)-related GTPase Arf-related protein binds to the Arf-specific guanine nucleotide exchange factor cytohesin and inhibits the Arf-dependent activation of phospholipase D. J. Biol. Chem., 274, 9744-9751. Sharer, J. D. and Kahn, R. A. (1999). The Arf-like 2 (Arl2)-binding protein, BART. J. Biol. Chem,. 274, 27553-27561. Sharer, J. D., Shern, J. F., Van Valkenburgh, H., Wallace, D. C., and Kahn, R. A. (2002). Arl2 and BART Enter Mitochondria and Bind the Adenine Nucleotide Transporter. Mol. Biol. Cell, 13, 71-83. Smith, S. A., Holik, P. R., Stevens, J., Melis, R., White, R., and Albertsen, H. (1995). Isolation and mapping of a gene encoding a novel human ADP-ribosylation factor on chromosome 17q12-q21. Genomics, 28, 113-115. Springer, S., Spang, A., and Schekman R. (1999). A primer on vesicle budding. Cell, 97, 145148. Stearns, T., Hoyt, M. A. and Botstein, D. (1990). Yeast mutants sensitive to antimicrotubule drugs define three genes that affect microtubule function. Genetics, 124, 251-262. Stepien, G., Torroni, A., Chung, A. B., Hodge, J. A. and Wallace, D. C. (1992). Differential expression of adenine nucleotide translocator isoforms in mammalian tissues and during muscle cell differentiation. J. Biol. Chem., 267, 14592-14597. Tamkun, J. W., Kahn, R. A., Kissinger, M., Brizuela, B. J., Rulka, C., Scott, M. P., and Kennison, J.A. (1991). The arflike gene encodes an essential GTP-binding protein in Drosophila. Proc. Natl. Acad. Sci., USA, 88, 3120-3124. Tarricone, C., Xiao, B., Justin, N., Walker, P. A., Rittinger, K., Gamblin, S. J., and Smerdon, S. J. (2001). The structural basis of Arfaptin-mediated cross-talk between Rac and Arf signalling pathways. Nature, 411, 215-219 Tian, G., Lewis, S. A., Feierbach, B., Stearns, T., Rommelaere, H., Ampe, C., and Cowan, N. J. (1997). Tubulin subunits exist in an activated conformational state generated and maintained by protein cofactors. J. Cell. Biol., 138, 821-832.

26 350

SCHÜRMANN AND JOOST

Van Valkenburgh, H., Shern, J. F., Sharer, J. D., Zhu, X., and Kahn, R. A. (2001). ADPribosylation factors (Arfs) and Arf-like 1 (Arl1) have both specific and shared effectors: characterizing Arl1-binding proteins. J. Biol. Chem., 276, 22826-22837. Weissman, J. T., Plutner, H., and Balch, W. E. (2001). The mammalian guanine nucleotide exchange factor mSec12 is essential for activation of the Sar1 GTPase directing endoplasmic reticulum export. Traffic, 2, 465-475. Zhu, X., Boman, A. L., Kuai, J., Cieplak, W., and Kahn R. A. (2000). Effectors increase the affinity of ADP-ribosylation factor for GTP to increase binding. J. Biol. Chem., 275, 13465-13475.