International Review of Neurobiology Volume 36

International Review of

NEUROBIOLOGY VOLUME 36

Editorial Board W. Ross ADEY

PAULJANSSEN

J U L ~ U SAXELROD

SEYMOUR KETY

Ross BALDESSARINI

KEITH KILLAM

SIRROGERBANNISTER

CONANKORNETSKY

FLOYD BLOOM

ABELLAJTHA

DANIELBOVET

BORISLEBEDEV

PHILLIPBRADLEY

PAULMANDEL

YURIBUROV

HUMPHRY OSMOND

Jost DELGADO

RODOLFOPAOLETTI

SIRJOHN ECCLES

SOLOMON SNYDER

JOEL

ELKES

STEPHEN SZARA

H. J. EYSENCK

MARATVARTANIAN

KJELL FUXE

STEPHENWAXMAN

Bo HOLMSTEDT

RICHARDWYATT

International Review of

Edifedby RONALD J. BRADLEY Department of Psychiatry LSU Medical Center Shreveport, Louisiana

R. ADRON HARRIS Department of Pharmacology University of Colomdo Health Sciences Center Denver, Colorado

VOLUME 36

ACADEMIC PRESS Son Diego New York

Boston London Sydney Tokyo Toronto

This book is printed on acid-free paper.

@

Copyright 0 1994 by ACADEMIC PRESS, INC. All Rights Reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc. A Division of Harcourt Brace & Company 525 B Street, Suite 1900, San Diego, California 92101-4495 United Kingdom Edition published by Academic Press Limited 24-28 Oval Road, London NW I 7DX

International Standard Serial Number:

0074-7742

International Standard Book Number:

0- 12-366836-0

PFUNTFD IN THE UNITED STATES OF AMERICA 94 95 9 6 9 7 98 9 9 B B 9 8 7 6

5

4

3 2 1

CONTENTS

Ca2+,N-Methybaspartate Receptors, and AIDS-Related Neuronal Injury

STUARTA. LIPTON ..................... I. Introduction . , . . . . . . . . . . . . . . . . . . . . . . 11. Neuronal Loss in the CNS of AIDS Pa 111. gpl20-Induced Neuronal Injury Is Ameliorated by Calci .............. Channel Antagonists . . . , . . . . . . . , . . . . . . . . . . 1V. Involvement of the NMDA Receptor in gpl20-Induced ..__....,......... Neuronal Injury . . . . . . . , . . . . . . . . . . V . Indirect Neuronal Injury Mediated by .................... Stimulated Monocytic Cells , . . . . . . . . 1 VI. Possible Involvement of Astrocytes, 0 HIV-1 Proteins in Neuronal Injury . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Overstimulation of NMDA Receptors, a Final Common Pathway . . . . VIII. Development of Clinically Tolerated NMDA Antagonists for HIVRelated Neuronal Injury . . . . . . . . . , , . . . . . . . ....’.’...’....’.’ IX. Excitatory Amino Acid Antagonist Treatments on the Horizon . . . . . . ......................... X. Conclusion . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I

1 3 4

6 8

12 15

17 21 23 24

Processing of Alzheimer Ap-Amyloid Precursor Protein: Cell Biology, Regulation, and Role in Alzheimer Disease

SAMGANDYAND PAULGREENCARD Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alzheimer Disease Is Associated with an Intracranial Amyloidosis . . . . APP Structure Gives Clues to Some of Its Functions . . . . . . . . . . . . . . . APP Is Processed via Several Distinct Enzymatic and Subcellular Pathways . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . , . . . . . . V. “Alternative” Pathways of APP Metabolism Provide Clues to the Source of A@-Amyloid . , . . , . . , . . . . . . . , . , . . . . , . . . . . . . . . . . . . . . . . . VI. AP-Amyloid Is a Normal Constituent of Body Fluids and the Conditioned Medium of Cultured Cells . . . . , . . . . . . . . . . . . . . . . . . . . . VII. Evidence Suggests the Existence of an Enzyme, P-Secretase, That Cleaves APP at the Amino Terminus of the AP-Amyloid Domain . . . . 1. 11. Ill. IV.

V

29 30 31

32 34 36

37

vi

CONTENTS

VIII. APP Mutations in Familial Cerebral Amyloidoses Occur within or near the AP-Amyloid Domain, Segregate with Disease in Affected Kindreds, and Yield APP Molecules That Display Some Proamyloidogenic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX. Signal Transduction via Protein Phosphorylation Regulates the Relative Utilization of APP Processing Pathways . . . . . . . . . . . . . . . . . . . X. Beyond Ap-Amyloid: Other Molecular Factors in Amyloidogenesis and Factors Differentiating Aging-Related Cerebral Amyloidosis from Alzheimer Disease ............................................ References ...................................................

37 39

42 44

Molecular Neurobiology of the GABAA Receptor

SUSANM. J. DUNN,ALANN. BATESON,AND IAN L. MARTIN I. Introduction ....................

................

11. Pharmacology of the GABAARece ............. 111. Biochemistry . . . . . . . . . . . . . . . . . . . . . .............. IV. Molecular Cloning of Receptor Subunits . . . . ...............

V. Characterization of the Receptor Family . . . . . . . . . . . . . . V1. The Future ................... .................

51 52 59 71 74 87 88

The Pharmacology and Function of Central GABAs Receptors

DAVIDD. MOTT AND DARRELL V. LEWIS I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pharmacology of GABAB Receptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1 I. Properties of GABABReceptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Function of GABA, Receptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References ................................................... 11.

97 98 104 126 209 210

The Role of the Amygdala in Emotional Learning

MICHAEL DAVIS I.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11. Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Electrophysiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Anatomical Connections between the Amygdala and Brain Areas

V.

Involved in Fear and Anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elicitation of Fear by Electrical or Chemical Stimulation of the Amygdala . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

225 226 227

228 23 1

CONTENTS

VI. Effects of Amygdala Lesions on Conditioned Fear ........ VII. Effects of Amygdala Lesions on Unconditioned Fear . . . . . . . . . . . . . . . VIII. Effects of Local Infusion of Drugs into the Amygdala o n Measures of Fear and Anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX. T h e Role of the Amygdala in Attention . . . . . . . . . . . . . . . . . . . . . . . . . . X. T h e Amygdala Is Critical for the Fear-Potentiated Startle Effect XI. Are Aversive Memories Actually Stored in the Amygdala? . . . . . XI!. Is the Amygdala Absolutely Essential for Fear-Potentiated Startle? . . . XIII. Can Initial Fear Conditioning Occur without the Amygdala? . . . . . . . . XIV. T h e Role of Excitatory Amino Acid Receptors in the Amygdala in Fear Conditioning . . . . . . . . . . . . . .. xv. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . .. .. ..

vii 234 235 236 238 240 24 1 244 245 250 258 259

Excitotoxicity and Neurological Disorders: Involvement of Membrane Phospholipids

AKHLAQA. FAROOQUI AND LLOYD A. HORROCKS I. 11. 111. IV. V.

VI. VII. VIII. IX.

Introduction ....................... .. Classification of Excitatory Amino Acid Receptors . . . . . . . . . . . . . . . . . Excitatory Amino Acid Receptors and Neural Membrane ........ Phospholipid Metabolism . . . Role of Enhanced Excitatory Am lipi Metabolism in Developing Brain .. .. Possible Mechanism of Cell Injur Amino Acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Excitatory Amino Acid Receptors, Phospholipid Metabolism, and Neurological Disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Excitatory Amino Acid Receptor Antagonists and the Treatment of Neurological Disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . ........ Summary . . . . . . . . . ........ References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

267 269 280 287 288 29 1 307 310 313 313

Injury-Related Behavior and Neuronal Plasticity: An Evolutionary Perspective on Sensitization, Hyperalgesia, and Analgesia

EDGART. WALTERS Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evolutionary Considerations .................................... Adaptive Behavioral Reactions to Injury . . . . . . . . . . . . . . . . . . . . . . . . . IV. Classes of Injury-Related Behavioral Modifiability . . . . . . . . . . . . . . . . . V. Injury Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I.

11. 111.

325 327 330 342 36 1

...

Vlll

CONTENTS

VI. Mechanisms of Rapid Nociceptive Sensitization .................... VII. Mechanisms of Long-Term Nociceptive Sensitization . . . . . . . . . . . . . . . VIII. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

367 386 407 412

INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONTENTS OF RECENT VOLUMES .................................

429 44 1

Ca2+, N-METHYL-D-ASPARTATE RECEPTORS, A N D AIDS-RELATED NEURONAL INJURY

Stuart A. Lipton Laboratory of Cellular and Molecular Neuroscience, Deportment of Neurology, Children’s Hospital, Beth Israel Hospital, Brigham and Women’s Hospital, Massachusetts General Hospital, and Program in Neuroscience, Howard Medical School, Boston, Massachusetts 021 15

I. Introduction 11. Neuronal Loss in the CNS of AIDS Patients 111. gpl20-Induced Neuronal Injury Is Ameliorated by Calcium Channel Antagonists

IV. Involvement of the NMDA Receptor in gpl20-Induced Neuronal Injury V. Indirect Neuronal Injury Mediated by HIV-Infected or gp 120-Stimulated Monocytic Cells VI. Possible Involvement of Astrocytes, Oligodendrocytes, and Other HIV- 1 Proteins in Neuronal Injury A. Astrocytes and HIV-Related Neuronal Damage B. gp120 Binding to the Oligodendrocyte Surface Molecule GalC C. gp120 Binding to Sulfatide and Myelin-Associated Clycoprotein D. gp120 Attenuates p-Adrenergic Stimulation of Astrocytes and Microglia E. Neurotoxicity of Other HIV- 1 Proteins F. Direct Effects of HIV-Infected Macrophages on Neurons VII. Overstimulation of NMDA Receptors, a Final Common Pathway VIII. Development of Clinically Tolerated NMDA Antagonists for HIV-Related Neuronal Injury A. Sites of Action of Potential Clinically Tolerated NMDA Antagonists B. NMDA Open-Channel Blockers C. NMDA Receptor Redox Modulatory Site X. Excitatory Amino Acid Antagonist ‘Treatments on the Horizon XI. Conclusion References

1. Introduction

A large number of adult patients and children with acquired immunodeficiency syndrome (AIDS) eventually suffer from neurological manifestations, including dysfunction of cognition, movement, and sensation. How can human immunodeficiency virus type 1 (HIV-1) result in neuronal damage if neurons themselves are not infected by the virus? This article reviews a series of experiments leading to a hypothesis that accounts at least in part for the neurotoxicity observed in the brains of INTERNATlONAL REVlEW OF NEUROBIOLOGY, VOL. 36

1

Copyright 0 1994 bv Academic Press, lnc. All rights of reproduction In any form reserved.

2

STUART A. LIPTON

AIDS patients. There is growing support for the idea of the existence of HIV- o r immune-related toxins that lead indirectly to the injury or demise of neurons via a potentially complex web of interactions among macrophages (or microglia), astrocytes, and neurons. HIV-infected macrophages/microglia or macrophages activated by HIV- 1 envelope protein gp120 appear to secrete excitants/neurotoxins including plateletactivating factor, arachidonate, and its metabolites, and possibly cysteine, nitric oxide, and superoxide anion. In addition, interferon-? (IFN-7) stimulation of macrophages induces release of the glutamate-like agonist quinolinate. Furthermore, HIV-infected macrophage production of cytokines, including TNF-a and ILl-p, contributes to astrogliosis. A final common pathway for neuronal susceptibility appears to be operative, similar to that observed in stroke, trauma, epilepsy, and several neurodegenerative diseases, possibly including Huntington's disease, Parkinson's disease, and amyotrophic lateral sclerosis. This mechanism involves the activation of voltage-dependent Ca2+channels and N-methyl-D-aspartate (NMDA) receptor-operated channels, and therefore offers hope for future pharmacological intervention. This review focuses on clinically tolerated calcium channel antagonists and NMDA antagonists with the potential for trials in humans with AIDS dementia in the near future. Our laboratory has a long-standing interest in the relationship of neuronal viability/outgrowth to intracellular Ca'+ levels (Lipton and Kater, 1989). Glutamate, o r a related excitatory amino acid (EAA), is the major excitatory neurotransmitter that controls the level of intracellular neuronal Ca2+ ([Ca2+Ii).Escalating concentrations of glutamate have been measured in vim following focal stroke and head injury (Choi, 1988; Meldrum and Garthwaite, 1990). As a result, there is an immediate elevation in [Ca2+Ii that precedes neurotoxicity by -24 h. Although the rise in [Ca2+Iimay not account by itself for the ensuing neuronal injury, several laboratories have now reported that prevention of the increase in [Ca2+Iileads to the amelioration of anticipated neuronal cell death (Choi, 1988; Meldrum and Garthwaite, 1990). Excessive intracellular Ca2+is thought to contribute to the triggering of a series of potentially neurotoxic events leading to cellular necrosis or apoptosis; these events include overactivation of the enzymes protein kinase C, Ca2+/ calmodulin-dependent protein kinase 11, phospholipases, proteases, protein phosphatases, xanthine oxidase, nitric oxide synthase, and endonucleases (Lipton and Rosenberg, 1994). There are many mechanisms involved in intracellular calcium homeostasis, and this subject is beyond the scope of this article (but see Miller, 1991). Here we will consider modes of Ca2+ entry into neurons during these pathological processes. Two major routes of entry of Ca2+ occur

Ca2+ CHANNELS, NMDA, A N D AIDS-RELATED NEURONAL INJURY

3

via ion channels that are permeable to Ca2+ and can be summarized as follows: (a) glutamate o r related EAAs trigger voltage-dependent calcium channels (VDCC) by depolarizing the cell membrane; the major VDCC subtype that is chronically activated by prolonged depolarizations is the L-type calcium channel (Lipton, 1991b). (b) glutamate or related EAAs activate ligand-gated ion channels directly; the major glutamate receptor-operated channel that is permeable to Ca2+ under these conditions is the NMDA subtype, but other types may also contribute (Choi, 1988; Lipton and Rosenberg, 1994). We have shown that activation of these channel types can control neuronal plasticity during normal development, but, in excessive amounts, our laboratory and others have shown that this stimulation can lead to neuronal death, e.g., after a stroke (Lipton and Kater, 1989). Similar mechanisms may obtain in various neurodegenerative conditions. In fact, this mechanism may represent a final common pathway of neuronal injury, although not involved in the primary pathophysiology of a neurologic disorder. Most importantly, this pathway makes the disease process amenable to pharmacotherapy. This line of reasoning led us to think that this mechanism might be involved in AIDS-related neuronal injury.

II. Neuronal Loss in the CNS of AIDS Patients

A significant number of adults and children with AIDS eventually develop neurological manifestations including dementia, myelopathy, and peripheral neuropathy; as many as 50% of infected children have neurological deficits presenting as delayed developmental milestones. These deficits occur even in the absence of superinfection with opportunistic organisms or associated malignancies (Price et al., 1988). Among the several neuropathological manifestations of AIDS in the brain is neuronal loss. In selected brains from AIDS patients, several groups (Ketzler et al., 1990; Everall et al., 1991; Wiley et al., 1991; Tenhula et al., 1992) have demonstrated loss of 18-50% of cortical neurons and retinal ganglion cell neurons. In addition, in the neocortex there is a loss in the complexity of dendritic arborization as well as presynaptic area (Masliah et al., 1992). The question remains, however, in at least this subset of patients with neurological manifestations, how can neurons be injured and yet not be infected?

4

STUART A. LIPTON

111. gpl20-Induced Neuronal Injury Is Ameliorated by Calcium Channel Antagonists

The major cell type infected with HIV- 1 in the CNS is the macrophage or microglial cell. These cells act as a reservoir for the virus and quite possibly release virus or viral proteins o r protein fragments. Possibly accounting at least in part for the injury to neurons is the observation first made in uitro by Brenneman and colleagues that picomolar concentrations of the envelope protein of HIV-1, gp120, can induce neuronal loss in rodent hippocampal cultures (Brenneman et al., 1988). Subsequently, our group demonstrated that in mixed cultures of neurons and glia, picomolar gp120 could increase [Ca2+liin rodent hippocampal neurons and retinal ganglion cells within a few minutes of application (Dreyer et al., 1990). Recently, similar findings were reported by Thayer's group (Lo et al., 1992), who were also able to resolve the increase in [Ca' +Ii into discrete oscillations by monitoring the calcium signal on a faster time scale. Within the next 24 h, neuronal injury ensues (Dreyer et al., 1990). Several groups have now confirmed that picomolar concentrations of gp120 can cause injury in a variety of neuronal preparations, including rat cortical neurons (Miiller et al., 1992; Dawson et al., 1993) and cerebellar granule cells (Savio and Levi, 1993). Both the early rise in [Ca2+Iiand the delayed neuronal injury can be largely prevented by antagonists of the L-type VDCC, including nimodipine (100 nM in 5% serum o r approximately 4 nM free drug) (Dreyer et al., 1990; Dawson et al., 1993; Savio and Levi, 1993). Other antagonists of the L-type VDCC are also effective to some degree (Lipton, 1991a) (Table I). Not only are rat retinal ganglion cells and cortical neurons in vitro partially protected by nimodipine and other voltage-dependent Ca2 channel antagonists, but also in a rat pup animal model, stereotactic injection of gp120 into the cortex produces a lesion consisting of cellular infiltrates of foamy macrophages and putative neuronal injury that is prevented by concomitant intraperitoneal administration of nimodipine (Lipton and Jensen, 1992). Additional in vivo evidence that low concentrations of gp120 are associated with neuronal injury has come from experiments of Brenneman, Hill, Ruff, Pert, and co-workers, who have found that intraventricular injections of gp120 into rats result in dystrophic neurites in hippocampal pyramidal cells as weil as behavioral deficits; moreover, cerebrospinal fluid of HIV-infected patients has gp120-like neurotoxic activity (Mervis et al., 1990; Buzy et al., 1992; Glowa et al., 1992). This evidence points to a potential role of gp120 in a neurodegenerative process. Because of these developments, the AIDS Clinical Trials Groups (ACTG) of the +

5

Ca2+ CHANNELS, NMIIA, A N D AIDS-RELATED NEURONAL INJURY

TABLE I VOLTAGE-DEPENDENT CALCIUM CHANNEL ANTAGONISTS ATTENUATEgp 120-MEDIATED NEURONAL INJURYI N VITRO _____

_____~

________

_________~

____

Amelioration"of gp 120-induced neuronal injury by voltagedependent calcium channel antagonists of the class

Dihydropyridine*

Dipheny lalkylamine piperazine derivative'

Phenylalkylamined

++++

++

+

Benzothiazepine' -

Note. Adapted from Lipton, 1992d. With permission from Hogrefe & Huber Publishers, Seattle. ' An increasing number of pluses indicates greater potency. Nimodipine and nifedipine (10-100 nM in 5% serum; -4-40 nM free drug). Flunarizine (10 p M in 5% serum). Verapamil (100 p M in 5% serum). Diltiazem (1 p M in 5% serum).

NIH Division of AIDS has asked us to begin a multicenter, randomized double-blind, placebo-controlled clinical trial to test the effects of nimodipine in adult patients with HIV-associated cognitive/motor complex (a subset of which has the more debilitating AIDS dementia complex), and this study is currently ongoing. Nevertheless, these developments do not tell us the mechanism of action of gp120 on neurons, which more recent evidence has led us to believe is an indirect pathway via macrophages/microglia (see below). For example, we noted that only neurons clustered in groups, presumably with synaptic contacts, were vulnerable to gp120, and this fact suggested that cellular interactions were necessary to produce injury. Moreover, the HIV envelope protein does not appear to act directly on calcium channels; in whole-cell and single-channel patch-clamp recordings, picomolar gp120 does not increase calcium current per se (H. S.-V. Chen, M. Plummer, P. Hess, S . A. Lipton, unpublished findings, 1990). It is possible that calcium channel antagonists ameliorate gp 120-induced neuronal injury by reducing the overall intracellular Ca" burden of the neurons. After all, Ca2+can accumulate in neurons during normal activity with each action potential fired, and nimodipine may be only indirectly beneficial by helping offset an increased calcium load due to another mechanism.

6

STUART A. LIPTON

IV. Involvement of the NMDA Receptor in gpl20-Induced Neuronal Injury

As outlined above, there is another prominent mode of Ca2+entry via channels directly coupled to EAAIglutamate receptors. T h e type of glutamate receptor subtype that is primarily (but not exclusively)involved in this regard is named after NMDA, a glutamate analog that is a selective agonist of this receptor (however, NMDA does not occur naturally in the body). We reasoned that since gp120 causes an early rise in [Ca2+Ii and delayed neurotoxicity, similar to glutamate acting at the NMDA receptor, perhaps glutamate or a closely related molecule was involved in HIV-related neuronal injury. Furthermore, it was well known that VDCC antagonists such as nimodipine could block some forms of glutamate neurotoxicity (Abele et al., 1990; Weiss et al., 1990; Sucher et al., 1991). Therefore, it was certainly possible that glutamate or a related NMDA agonist was somehow involved in gp120-induced neuronal damage. In addition, Heyes and colleagues (Heyes et al., 1989, 1991) had found that cerebrospinal fluid (CSF) levels of quinolinate, a naturally occurring (albeit weak) NMDA agonist, was correlated with the degree of dementia in AIDS patients. To test the possibility that EAAs were involved, the following experiments were undertaken. NMDA antagonists were assessed for their ability to prevent gp 120-induced neuronal injury. We found that MK-80 1 (dizocilpine), an open-channel blocker of NMDA receptor-coupled ion channels, prevented gp 120-induced neuronal injury (Lipton et al., 1990, 1991). D-2-Amino-5-phosphonovalerate (APV), a competitive antagonist at the glutamate binding site of the NMDA receptor, was partially effective in ameliorating this form of neuronal injury. Recently, other groups have obtained similar results using NMDA antagonists or inhibitors of nitric oxide synthase (nitric oxide, or NO., is believed to be involved in one of the toxic pathways activated by NMDA receptor stimulation) (Muller et al., 1992; Dawson et al., 1993; Savio and Levi, 1993). In contrast, CNQX, a non-NMDA antagonist, did not protect from gpl20-induced neuronal damage, at least to retinal ganglion cells (Lipton et al., 1990, 1991). Another possible link between the effects of gp120 and NMDA receptor activation arises from the observation that one form of neuronal injury in both the brains of AIDS patients (Masliah et al., 1992) and the brains of rats injected with gp120 (Mervis et al., 1990) or the brains of transgenic gpl20 mice (Toggas et al., 1994) involves dystrophic neurites. These neurites are excessively tortuous and display a paucity of branches. Some of these neurites may be retracting, giving them a “bald” appearance. We and others have found a similar pattern of dystrophic neurites,

Ca2+ CHANNELS, NMDA, AND AlDS-RELATED NEURONAL INJURY

7

including retraction of growth cones, in response to sublethal concentrations of NMDA or glutamate in cultured rat retinal ganglion cells and hippocampal neurons (Mattson et al., 1988; Lipton and Kater, 1989; Offermann et al., 1991). Furthermore, these effects are dependent on influx of Ca2+ into the neurons. These findings indicate that the endpoints for neuronal injury related to gp120 or excitotoxicity should include more subtle changes than death, and these alterations in neuronal cytoarchitecture could have important consequences for neuronal function and plasticity (Lipton and Kater, 1989). T h e simplest potential explanation for all of these findings is that gp120 might simulate an NMDA-evoked current, or somehow augment such currents. T o examine this idea, we used the patch-clamp technique to determine whether gp 120 affected membrane currents. However, in whole-cell recordings, using both conventional and perforated-patch techniques, no effect of picomolar gp120 was observed, even in recordings lasting tens of minutes. Similarly, no enhancement of glutamateor NMDA-evoked currents was encountered (Lipton et al., 1991). Interestingly, nanomolar concentrations of g p 120 (a 1000-fold excess over the levels used in the aforementioned experiments) have been reported to block NMDA receptor-operated ion channels, preventing NMDAevoked increases in Ca2+ influx (Sweetnam et al., 1993). This finding may account, at least in part, for the dose-response curve of gp120induced neuronal injury, which has an inverted “U” shape (Brenneman et al., 1988); that is, at high nanomolar concentrations in contrast to picomolar concentrations, gp 120 no longer induces neuronal cell injury. Nevertheless, during HIV-1 infection in the brain it appears unlikely that nanomolar concentrations of gp 120 actually occur because conventional ELISA and Western blots are sensitive to these concentrations but have failed to detect their presence. The next possible explanation that we considered is that endogenous levels of glutamate become toxic in the presence of picomolar gp120. To test this hypothesis, the enzyme glutamate-pyruvate transaminase (GPT) was used to degrade the endogenous glutamate in our retinal cultures. High-pressure liquid chromatography (HPLC) analysis of amino acids was used to verify glutamate degradation (-25 p M decreased to less than 5 F M ) . Under these conditions, the catabolism of endogenous glutamate in uitro protected neurons from gp 120-induced injury to rat retinal ganglion cells (Lipton et al., 1990, 1991). Recently, Dawson et al. (1993) have also found that 25 ,uM glutamate was necessary in order for them to observe neurotoxicity in rat cortical cultures in the presence of 100 pM gp120. Taken together, these data argue that concurrent activation of NMDA receptors is needed for neuronal injury

8

STUART A. LIPTON

by gp120 in AIDS. These experiments d o not tell us, however, whether the action of gp120 is mediated directly on neurons or indirectly via an intervening cell type, such as astrocytes or macrophages/microglia.

V. Indirect Neuronal Injury Mediated by HIV-Infected or gpl20-Stimulated Monocytic Cells

It is still not known definitively whether the adverse effects of gp120 are mediated directly on neurons, via glial cells such as microglia and astrocytes, or by a combination of these mechanisms. To determine at least some of the cell types involved in neurotoxicity, we performed the following experiment. L-Leucine methyl ester was used to deplete monocytoid cells from cultures of mixed glia and neurons. Under these conditions, gp120 no longer injured neurons, suggesting that at least under our culture conditions, macrophages/microglial were necessary to mediate the neurotoxic effects of gp120 (Lipton, 1992~).It is well known that gp120 binds to CD4 on human monocytic as well as lymphocytic cells, and, in fact, this appears to be the major (but probably not exclusive) route of entry of the virus into these cells. Human macrophages, monocytes, and microglia, but apparently not rat or mouse cells, possess the proper CD4 molecule to bind gp120; however, lack of known receptors does not, of course, rule out alternative mechanisms of binding or toxicity. In several laboratories, for instance, it has not been necessary to have human macrophages present to observe gp 120-induced neuronal injury in cultures. Along these same lines, in our laboratory’s cultures of rat retinal cells, anti-rat CD4 antibodies do not block the neuronal injury to retinal ganglion cells engendered by gp 120 which follows an EAA pathway, whereas anti-gp120 antiserum completely blocks this toxic effect (Kaiser et al., 1990). On the other hand, gp120 incubation with the human monocyte cell line THP-1 also produces the release of neurotoxins that follow an EAA pathway to cell injury; antibodies directed against the CD4-binding region of gp120, but not against the V 3 loop of gp120, block this toxic effect (Giulian, 1993). Thus, there appear to be both CD4- and non-CD4-mediated mechanisms of gp 120-induced neuronal injury, in a sense paralleling a similar situation concerning CD4- and non-CD4-mediated mechanisms for viral entry into cells. In conjunction with the data of other laboratories, the aforementioned results suggest the following model of HIV-related neuronal injury (Fig. 1). HIV-infected macrophages (Giulian et al., 1990; Pulliam et al., 1991) or gpl20-stimulated macrophages (Lipton, 1992c; Giulian, 1993) release neurotoxic products. These neurotoxins include relatively

Ca2+ CHANNELS, NMDA, AND AIDS-RELATED NEURONAL INJURY

9

small, heat stable compounds, which have recently begun to be characterized by Gendelman and colleagues (Genis et al., 1992). They found that the products released by HIV-infected macrophages include a product of phospholipase A, activity, platelet-activating factor (PAF), and arachidonic acid and its metabolites. Under their conditions, these substances are released only in the presence of astrocytes, implying some positive feedback loop between astrocytes and macrophages. HIV-infected macrophages also release the cytokines TNF-a and IL-lP, which have been shown to stimulate astrocyte proliferation (Chung and Benveniste, 1990; Selmaj et al., 1990), another feature of HIV encephalitis. In addition, the cytokines present in conditioned medium from lipopolysaccharide (LPS)-treated astrocytes can stimulate HIV- 1 gene expression in monocytic cells (Vitkovic et al., 1990). Under certain in vztro conditions, TNF-a and IL-1 can be associated with the death of oligodendrocytes and by implication, demyelination (see below) (Robbins et al., 1987). Moreover, there are multiple, complex interactions and feedback loops affecting cytokine and arachidonic acid metabolite production by macrophages and astrocytes. For example, TNF-a enhances IL- 1 production in macrophages (Morganati-Kossmann et al., 1992). Arachidonic acid metabolites can influence the production of TNF-a and IL-1P in macrophages, and, in turn, TNF can amplify arachidonic acid metabolism, in response to IL-1. PAF can enhance TNF and IL-1 production, and, in turn, PAF synthesis can be stimulated with TNF, IL-lP, or interferon-y (IFN-.)I)in human monocytes (Pignol el al., 1987; Valone and Epstein, 1988; Valone et al., 1988; Conti et al., 1989; Dubois et al., 1989; Poubelle et al., 1991).Finally, the same arachidonic acid metabolites and cytokines released by HIV-infected macrophages appear to be produced by gpl20-stimulated monocytic cells. For example, this HIV glycoprotein induces the release of arachidonate, its metabolites, TNF-a, and IL-1P from human monocytes (Merrill et al., 1989; Wahl et al., 1989; Merrill and Chen, 1991). It remains to be shown definitively however, that gp 120-stimulated macrophages also release PAF, although these experiments are currently in progress (H. Nottet, H. E. Gendelman, and S. A. Lipton, unpublished observations, 1994). T h e cytokines TNF-a and IL-IP in the amounts produced by HIVinfected or gp120-stimulated macrophages do not appear to be neurotoxic in and of themselves (Genis et al., 1992). Could, however, the arachidonic acid metabolites emanating from HIV-infected or gp 120stimulated macrophages be involved in neurotoxicity? Several arachidonic acid metabolites as well as of PAF have excitatory effects on neurons (Palmer et al., 1981; Kornecki and Ehrlich, 1988, 1991; Lindsberg et al., 1991; Manzini and Meini, 1991a,b; Meini et al., 1992), and this may

10

NOLdI7 ’V LXVnLS

STUART A. LIPTON

1-AIH

FIG. 1. Current models of HIV-related neuronal injury. Previous work has shown that HIV-infected macrophages/microglia release factors that lead to neurotoxicity. These factors include platelet-activating factor (PAF), arachidonic acid and its metabolites, as well as cytokines and other as yet unidentified substances. Macrophages and astrocytes have mutual feedback loops (signified by the reciprocal arrows). The excitatory action of the macrophage factor may lead to an increase in neuronal Ca2+ and the consequent release of glutamate. In turn glutamate overexcites neighboring neurons leading to an increase in intracellular Ca2+,neuronal injury, and subsequent further release of glutamate. This final common pathway of neurotoxic action can be blocked by NMDA antagonists. For certain neurons, this form of damage can also be ameliorated to some degree by calcium channel antagonists or non-NMDA receptor antagonists. The major pathway of entry of HIV-1 into monocytoid cells is via gp120 binding, and therefore it is not surprising that gp120 (or a fragment thereof) is capable of activating uninfected macrophages to release similar factors to those secreted in response to frank HIV infection. Cytokines participate in this cellular network in several ways. For example, HIV infection or gp120 stimulation of macrophages enhances their production of TNFa and IL-lp (solid arrow). The TNF-a and IL-lp produced by macrophages stimulate astrocytosis. Astrocytes appear to feedback (dashed arrow) onto monocytic cells by an as yet unknown mechanism to increase the macrophage production of these cytokines. TNFa may also increase voltage-dependent calcium currents in neurons. Interferon-? (IFNy), known to be elevated in the CNS of patients with AIDS, can induce macrophage/ microgliosis and macrophage production of quinolinate (an NMDA-like agonist) and PAF; in conjunction with IL-Ip, IFN-y can induce nitric oxide synthase (NOS) expression with consequent NO. production in cultured astrocytes and in this manner may potentiate NMDA receptor-mediated neurotoxicity in mixed neuronal-glial cultures. NO. has recently been shown to react with superoxide anion (02.-) to yield a neurotoxic substance, probably in the form on ONOO- (peroxynitrite). It is conceivable that such cytokine stimulation of the inducible form of NOS in macrophages or astrocytes may thereby contribute to HIVrelated neurotoxicity. In addition, the constitutive form of NOS (cNOS)has been implicated

Ca'+ CHANNELS, N M D A , A N D AIDS-RELATED NEURONAL INJURY

11

represent at least one pathway whereby PAF and these arachidonic acid metabolites evoke EAA-induced neurotoxicity. In particular, PAF has recently been shown to increase intracellular neuronal Ca2+and lead to enhanced neurotransmission, apparently by increasing the release of presynaptic glutamate (Bito et al., 1992; Clark et al., 1992). Although the elevated levels of arachidonate and metabolites released from HIVinfected macrophages may cycle up and down, the increased concentration of PAF is a persistent and potentially most important factor (H. E. Gendelman, personal communication, 1994). In collaboration with Gendelman's group, our laboratory has begun a series of experiments which suggest that under specific conditions, the elevated levels of PAF that have been measured in cultures of HlV-infected monocytic cells as well as in the cerebrospinal fluid of patients with the AIDS dementia complex (Genis et al., 1992; H. E. Gendelman, personal communication, 1994) can be toxic to neurons in vitro. It is possible that TNF-a also contributes to this process by increasing voltage-dependent Ca2+ currents (Soliven and Albert, 1992). Additional glutamate receptor activation may occur as a consequence of these events, as neurons are excited or injured and release their stores of glutamate onto neighboring neurons (Lipton et al., 1990b, 1991; Lipton, 1992; Lo et al., 1992). One line of evidence for this supposition lies in the finding, as detailed above, that NMDA antagonists or enzymatic degradation of glutamate ameliorates g p 120induced neuronal injury in mixed neuronal-glial cultures (Lipton et al., 1990, 1991). Moreover, PAF-related neuronal injury also appears to be ameliorated by NMDA antagonists. Intensive investigation in several laboratories is currently under way to study this potential pathway for neuronal injury that is triggered by HIV-infected or gpl20-stimulated macrophages. Also, as alluded to above, another possible link between HIV-1 infection and EAA-induced neurotoxicity involves quinolinate, an endogenous NMDA agonist that is increased in the cerebrospinal fluid of patients in gp120 neurotoxicity; the neuronal form of the enzyme (cNOS) is activated by a rise in intracellular Ca2+after stimulation of the NMDA receptor, and inhibitors of this enzyme have been reported to prevent gp120 neurotoxicity. The coat protein gp120 may have additional direct or indirect effects on astrocytes, e.g., to decrease growth factor production or to inhibit glutamate reuptake, for example, via arachidonic acid. Arachidonate has also been recently reported to enhance NMDA-evoked currents and therefore could contribute to neurotoxicity not only by enhancing net glutamate efflux but also by increasing its effectiveness at the NMDA receptor. Also, we have recently shown that gp120 enhances cysteine secretion from macrophages Cysteine is a known NMDA agonist and could therefore represent at least one of the neurotoxic substances released from stimulated macrophages.

12

STUART A. LIPTON

with the AIDS dementia complex (Heyes et al., 1991). Quinolinate levels are known to be influenced by cytokines that are increased after HIV1 infection. For example, it is known that IFN-y is present in the brains of patients with AIDS (Tyor et al., 1992), and human macrophages activated by IFN-y release substantial amounts of quinolinate (Heyes et al., 1992). In addition, under some conditions, e.g., following neuronal loss, quinolate can also be produced by astrocytes (Speciale et al., 1987; Kohler et al., 1988). Quinolinate, therefore, may also contribute to neuronal injury by activating NMDA receptors during HIV infection, although this scenario appears to also be true for a variety of CNS infections. Recently, we have found that another NMDA agonist, cysteine, is released from gp120-stimulated human macrophages, at least in nitro (M. Yeh and S. A. Lipton, in preparation).

VI. Possible Involvement of Astrocytes, Oligodendrocytes, and Other HIV-1 Proteins in Neuronal Injury

A. ASTROCYTES AND HIV-RELATED NEURONAL DAMAGE In at least some model systems, the presence of astrocytes is necessary for HIV-infected macrophages to release substantial amounts of their neurotoxic factors (Genis et al., 1992). In addition, astrocytes may be important in mediating HIV-related neuronal injury in other ways. For example, in murine hippocampal cultures Brenneman et al. (1988) found that gp- 120-induced neurotoxicity can be prevented by the presence of vasoactive intestinal polypeptide (VIP) or by a five amino acid substance with sequence homology t.0 VIP, peptide T. These workers also found that VIP acts on astrocytes to increase oscillations in intracellular Ca2+ and to release factors necessary for normal neuronal outgrowth and survival (Brenneman et al., 1990). Thus, these results raise the possibility that gp120 may compete with endogenous VIP for a receptor, most likely on astrocytes, that is critical to normal neuronal function. The receptor may bear some resemblance to mouse CD4 because mouse anti-CD4 antibodies blocked the toxic effects of gp120 in this system (Brenneman et al., 1988). This effect of gp120 is hypothesized to prevent the release of such astrocyte factors that are necessary to prevent neuronal injury and suggests that one pathway for neuronal damage is an indirect one that is mediated via astrocytes. Although not specifically illustrated in Fig. 1, gp120 might therefore interact with a receptor on

Ca2+ CHANNELS, NMDA, AND AIDS-RELATED NEURONAL INJURY

13

astrocytes. Hence, neurotoxicity may in part be realized by interfering with the normal function of astrocytes and their release of a neuronal growth factor(s) (Giulian et al., 1993) (see Fig. 1). It is also possible that gp120 might affect astrocytes in some other manner, e.g., to inhibit their ability to take up glutamate or even to reverse the reuptake process, resulting in a net efflux, and thus contributing to EAA-induced neurotoxicity. Such an effect would contribute to the apparent increase in sensitivity of neurons to glutaniate toxicity in the presence of gp 120 and would also help explain the requirement for some glutamate (-25 pLM) to be present in the culture medium in order to observe gpl20-induced neurotoxicity (Lipton et al., 1990, 1991; Dawson et al., 1993). Future experiments will have to be designed to explore these possibilities.

B.

gp120 BINDING TO

THE

OLIGODENDROCYTE SURFACE MOLECULEGalC

T h e envelope protein g p 120 has also been shown to bind to galactosyl ceramide (GalC), a molecule on the surface of the oligodendrocyte, which represents the cell type responsible for myelination in the CNS (Bhat et al., 1991; Harouse et al., 1991). Relatively high concentrations of gp120 (nanomolar) were necessary to observe this binding compared to the low (picomolar) concentrations of the coat protein that have been found to lead to neurotoxicity. Nonetheless, the findings concerning binding to GalC raise the possibility of participation of gp120 in myelin disruption and, therefore, a further indirect influence on the welfare of neurons. Future studies will be necessary to determine the significance of this potential pathway for cellular injury. C. gp120 BINDING TO SULFATIDE A N D MYELIN-ASSOCIATED GLYCOPROTEIN Recently, in addition to GalC, gp120 has been shown to bind to sulfatide (Gals), a sulfated glycoprotein implicated in sensory neuritis, and to myelin-associated glycoprotein ( M A G ) , an autoantigen in demyeh a t i n g neuropathy (van den Berg et al., 1992). Similar to GalC, binding became significant in the nanomolar range of gp120. The authors speculate that this binding could have implications for the peripheral nervous system, e.g., in an acute or chronic demyelinating neuropathy or a painful sensory axonal neuropathy such as that frequently observed during HIV infection. However, as alluded to above, the significance of binding to

14

STUART A. LIPTON

nanomolar gp120 levels in the nervous system remains uncertain. The same group of workers has recently published a preliminary report that gp 120 may bind to neurons in immunofluorescence assays (Latov et al., 1993), but it is not clear whether the concentration of gp120 required to see this effect is attained in the CNS during HIV-1 infection.

D. gp 120 ATTENUATES /I-ADRENERGIC STIMULATION OF ASTROCYTES AND MICROCLIA Other effects of gp120 have also been reported. Levi's group have found that picomolar gp 120 can attenuate /I-adrenergic stimulation of CAMPin astrocytes and microglia (Levi et al., 1993). When added alone, gp120 modestly enhanced the basal levels of CAMP. These effects of gp120 could also interfere with P-adrenergic modulation of cytokine production, e.g., of TNF-a. Thus, gp120 may have other, complex effects on glial cells in the CNS.

E. NEUROTOXICITY OF OTHERHIV-1 PROTEINS In addition to gp120, two other HIV-1 proteins have been reported to affect neurons o r neuronal-like cells, raising the possibility of their involvement in HIV-related neuronal injury. T h e nuclear protein tat was shown to be toxic to glioma and neuroblastoma cell lines in uitro and to mice in uiuo (Gourdou et al., 1990; Sabatier et al., 1991). The basic region of the peptide (amino acid residues 49-57) appears to act nonspecifically to increase the leakage conductance of the membrane, thus altering cell permeability. Moreover, neurotoxicity of the related Maedi-Visna virus peptide was ameliorated by NMDA antagonists or by inhibitors of nitric oxide synthase (Hayman et al., 1993),reminiscent of the pharmacology of antagonism of the neurotoxic effects of gp120 and HIV-infected macrophages. Further work will be necessary to attempt to relate these findings with the tat peptide to the neuropathology encountered in the brains of patients with HIV-l-associated cognitive/motor complex. Another HIV-1 protein, Nef, has also been shown to affect neuronal cell function. Nef shares sequence and structural features with scorpion toxin peptides; both recombinant Nef protein and a synthetic portion of scorpion peptide increase total K f current in chick dorsal root ganglion cells (Werner et al,, 199I).

Ca*+ CHANNELS, NMDA, A N D AIDS-RELATED NEURONAL INJURY

15

F. DIRECTEFFECTS OF HIV-INFECTED MACROPHAGES ON NEURONS Finally, it is possible that HIV-infected monocytoid cells may have a cytopathic effect on neurons by direct contact (Tardieu et al., 1992). This mechanism does not preclude, however, an additional mechanism of neuronal injury mediated by soluble factors leading to excessive stimulation of NMDA receptors (Lipton, 1993).

WI. Overstimulation of NMDA Receptors, a Final Common Pathway

From the foregoing, there appear to be at least two sites of potential interaction of HIV-related neurotoxins with NMDA receptors (Fig. 1). First, quinolinate emanating from macrophages may directly stimulate neurons. Second, after excitation by quinolinate, PAF, and possibly arachidonic acid metabolites, or after injury due to other toxic pathways, neurons would release glutamate onto second-order neurons and astrocytes might fail to take up the glutamate. This “bad neighbor hypothesis” is in some ways similar to the damage thought to occur in the penumbra of a stroke-glutamate released by injured neurons contributes to further injury to neighboring neurons. Moreover, NMDA antagonists ameliorate HIV-related neuronal injury induced by either HIV-infected macrophages (Giulian et al., 1990; H. E. Gendelman, personal communication, 1993) or, as mentioned earlier, gpl20-activated macrophages (Lipton et al., 1990,199 1). Furthermore, in some cases calcium channel antagonists can attenuate this form of damage [(Dreyer et al., 1990; Lipton, 1991b; L. Pulliam, personal communication, 1991) (Table 11). In general, the pharmacology of neuroprotection from noxious agents depends on the repertoire and diversity of ion channel types in a particular class of neurons (Lipton, 1991b). For example, neurons lacking NMDA receptors will obviously not be protected by NMDA antagonists. Conversely, if NMDA receptorassociated channels are the predominant channel in a specific neuronal cell type whereby Ca2+enters the cell, then the lethal effects of excessive stimulation by glutamate may be ameliorated with NMDA antagonists. Some non-NMDA receptor-associated channels are directly permeable to Ca2+,but most appear not to be (e.g., those containing the GluR2 receptor subunit). However, depolarization of neurons by stimulation of non-NMDA receptors will trigger VDCCs. If sufficient L-type calcium channels exist on a particular neuronal cell type, then the excessive influx

16

S T U A R T A. LIPTON

TABLE I1 PROTECTIVE EFFECTS OF CALCIUM CHANNEL ANTAGONISTSAND NMDA ANTAGONISTS AGAINST gp12O-INDUCED A N D HIV-INFECTED MACROPHAGE-MEDIATED NEURONAL INJURY rrv VITRO Attenuation of neuronal iniury by Insult gp120 glycoprotein or fragment HIV-infected macrophage toxin(s)

Ca2+channel anlagonists"

NMDA antagonist

non-NMDA antagonist

+

+

-

+b

+

-

Note. Adapted from Lipton, 1992d. With permission from Hogrefe & Huber Publishers, Seattle. Nimodipine or nifedipine (10-100 nh4 in 5% serum or -4-40 nM free dihydropyridine). L. Pulliam (personal communication, 199I)-results in human brain cell aggregates [but see also earlier work (Ciulian et al., 1990) that found no protective effective of high concentrations of calcium channel antagonists in chick ciliary and rat spinal cord neurons: nonetheless, it is possible that in these latter experiments the high concentration of calcium channel antagonist used was deleterious in and of itself or that these antagonists were not effective in the neuronal cell types tested].

of Ca'+ via these channels could lead to toxic consequences. Hence, in some cell types such as hippocampal pyramidal cells, cortical neurons, and retinal ganglion cells, there is evidence that calcium channel antagonists may attenuate damage due to activation of either NMDA or nonNMDA receptors (Abele et al., 1990; Weiss et al., 1990; Sucher et al., 1991). Along similar lines of reasoning, the release or action of glutamate may be involved in the final common pathway of neuronal injury by HIV-infected macrophages or by gpl20-stimulated macrophages. Thus, either NMDA or non-NMDA receptor activation may play a role in this form of toxicity depending on the exact repertoire of ion channels in a particular cell type. In fact, it has been suggested that non-NMDA receptors could also be important in contributing in the neurotoxic events triggered by gp120 (V. L. Dawson et al., 1992, 1993). Nevertheless, the majority of findings to date suggest that NMDA receptor-mediated neuronal injury plays a predominant role in the pathogenesis of the neurological manifestations of AIDS in the CNS (Lipton, 1992b; Lipton and Jensen, 1992).

Ca2+ CHANNELS, NMIIA, A N D AIDS-RELATED NEURONAL INJURY

17

VIII. Development of Clinically Tolerated N M D A Antagonists for HIV-Related Neuronal lniury

NMDA receptors may be involved in HIV-related neurotoxicity at two separate sites, located (a)on the primary neuron injured by factors released from glial cells and (b) on neurons that are secondarily affected (see above and Fig. 1).This fact has provided an impetus for our laboratory to begin a drug development program for clinically tolerated NMDA antagonists, as described below.

A. SITESOF ACTIONOF POTENTIAL CLINICALLY TOLERATED NMDA ANTAGONISTS Despite concerns about the potential complexity of EAA receptor pharmacology, we can consider currently available agents that appear to work on broad classes of these receptors. For the purposes of this review, we will concentrate mainly on NMDA antagonists that appear to be clinically tolerated and therefore can be considered for human trials. There are several modulatory sites on the NMDA receptor-channel complex that could potentially be used to modify the activity of the receptor-operated ion channels and thus to prevent the excessive influx of Ca2+ (Fig. 2). The first site is the glutamate or NMDA binding site. An antagonist acting here would be competitive in nature, i.e., competing for the site with an EAA. For both theoretical and practical reasons, a competitive inhibitor might not be as desirable an antagonist as one that is not competitive for the glutamate binding site. A competitive antagonist would perforce eliminate the normal, physiological activity of the NMDA receptor even before it would affect potentially excessive levels of glutamate. Thus, cognition and memory, thought to be related to long-term potentiation (LTP), might be compromised as well as other important functions mediated by excitatory transmission in the brain. In any event, as part of the disease process, escalating levels of glutamate might be able to overcome o r “out-compete” such an antagonist.

B. NMDA OPEN-CHANNEL BLOCKERS I n contrast, other modulatory sites should be able to inhibit the effects of high levels of glutamate in compromised areas of the brain while

18

STUART A. LIPTON

NMDA or giu glycine

J.

FIG.2. Sites of potential antagonist action on the NMDA receptor-channel complex. Competitive antagonists can compete with NMDA or glutamate (glu) for binding to the agonist site. Several antagonists to the glycine coagonist site have been described that are chlorinated and sulfated derivatives of kynurenic acid. These inhibitors do not compete with glutamate and hence are designated noncompetitive. It is not yet known whether any will prove to be tolerated clinically. H + effects are transmitted through another noncompetitive site; decreasing pH acts to downregulate channel activity. Other sites for polyamines and Zn2+ can also be used to affect receptor-channel function. Sites that inhibit channel activity by binding Mg2+or drugs such as MK-801, phencyclidine, and memantine (Mem.) are within the electric field of the channel and are only exposed when the channel is previously opened by agonist (termed uncompetitive antagonism). Finally, a redox modulatory site [probably a disulfide bond, or at least a long-lasting covalent modification of a thiol group, that can be converted to free sulfhydryl groups (S-S + 2-SH)] is affected by chemical reducing and oxidizing agents. Oxidation can favor the disulfide conformation (S-S) over free thiol (-SH) groups and thus downregulate channel activity. Several nitroso compounds can transfer the NO group to the thiol(s) of the NMDA receptor's redox site, producing RS-NO (NO' equivalents), and thus also leading to downregulation of channel activity.

leaving relatively spared the effects of normal neurotransmission in other regions of the brain (Karschin et al., 1988; Levy and Lipton, 1990; Chen et al., 1992). For example, one site that appears to have this advantageous effect is located in the channel itself. There are drugs that only block the channel when it is open; i.e., the antagonist can only gain access to the channel in the open state. On average, escalating levels of glutamate result in the channels remaining open for a greater fraction of time. Under these conditions, there is a better chance for an open-channel blocking drug to enter the channel and block it. The result of such a mechanism of action is that the untoward effects of greater (pathological) concentrations of glutamate are inhibited to a greater extent than those of lower (physiological)concentrations (Chen et al., 1992). Unfortunately, some of these open-channel blockers, which include phencyclidine (angel

Ca*+ CHANNELS, NMDA, A N D AIDS-RELATED NEURONAL INJURY

19

dust) and MK-801 (dizocilpine), have neuropsychiatric side effects and probably cannot be safely administered (Koek et al., 1988). Another concern with NMDA antagonists, such as phencyclidine and MK-801, is the development of reversible neuronal vacuolization (Olney et al., 1989). A problem with MK-801 is that once it enters an open channel, it leaves the channel only very slowly (half time >1 h). In practical terms this means that the degree of blockade builds up after MK-801 administration because each molecule of the antagonist entering a channel effectively does not leave. Several members of this open-channel blocking class of agents, however, are tolerated, such as ketamine and dextromethorphan o r the related molecule dextrorphan (Choi, 1987; Choi et al., 1987; MacDonald et al., 1987; Davies et al., 1988; O’Shaughnessy and Lodge, 1988). Unfortunately, it is not clear whether these particular drugs are sufficiently potent NMDA antagonists at clinically tolerated doses. Nevertheless, the fact that certain members of this open-channel blocker family are clinically tolerated appears to be associated with their rapid kinetics of interaction with the channel (the kinetic parameters are composed of the onrate and off-rate for channel blockade) (Rogawski and Porter, 1990; Chen et al., 1992). Most importantly, the safe drugs, such as memantine (see Bormann, 1989, and the discussion below), leave the channel promptly, with an off-rate -5 s at micromolar concentrations (Chen et al., 1992). Mg2+also blocks open NMDA channels, and this may be the basis for its antiepileptic and neuroprotective effects (Goldman and Finkbeiner, 1988; Wolf et al., 1990,1991). These beneficial effects, however, may not be robust, probably because Mg2+ leaves the channel so quickly that it may not act effectively to offset toxic levels of glutamate. In addition, these charged channel-blocking drugs act to a lesser degree when neurons are depolarized (become more positively charged), e.g., under conditions of energy compromise (Zeevalk and Nicklas, 1992). In summary, an agent that remains in the channel for at least some period of time is necessary to block the effects of glutamate overstimulation. Of the known NMDA open-channel blockers, memantine is one candidate for clinical trials to combat neurological disorders, such as HIV-associated cognitive/motor complex, with a component of NMDA receptor-mediated neurotoxicity because memantine has been used clinically in Germany for over a dozen years in the treatment of Parkinson’s disease and spasticity. Memantine is a congener of amantadine, the wellknown antiviral and antiparkinsonian drug used in the United States. Amantadine, however, is considerably less potent on NMDA receptoroperated ion channels at clinically tolerated doses (Chen et al., 1992),

20

STUART A. LIPTON

probably precluding its use for these other neurological diseases. It may be no accident that memantine both inhibits NMDA receptor responses and alleviates parkinsonian symptoms; one theory of Parkinson’s disease is that neurons die, at least in part, due to a form of NMDA receptormediated toxicity. C. NMDA RECEPTOR REDOXMODULATORY SITE

Another modulatory site on the NMDA receptor-channel complex of possible clinical utility in the near future was discovered several years ago in our laboratory and has been termed the redox modulatory sites (Aizenman et al., 1989). This site consists of one or more sulfhydryl groups; these sulfhydryl groups may possibly be in close approximation and form a disulfide bond under oxidizing conditions. Under chemical reducing conditions that favor the formation of free thiol (-SH) groups over a disulfide, the opening frequency of NMDA receptor-associated channels increases (Aizenman et al., 1989; Tang and Aizenman, 1993), and thus there is a net increase in Ca2+ influx through the channels (Reynolds et al., 1990; Sucher et al., 1990) and an increase in the extent of NMDA receptor-mediated neurotoxicity (Levy et al., 1990; Aizenman and Hartnett, 1992). Conversely, redox reagents that mildly oxidize the NMDA receptor, for example, to reform disulfide bonds o r form ligands on the free thiol groups, might prove useful in combating the myriad of neurological maladies following a final common pathway of NMDA receptor-mediated neuronal damage (Aizenman et al., 1990). Indeed, several such redox reagents have recently been reported, including quite surprisingly the common nitroso compound nitroglycerin (Lei et al., 1992). One mechanism of nitroglycerin’s action in this regard is mediated by a substance related to nitric oxide (NO.), but in a different redox state, for example, in the form of RS-NO (nitrosonium ion equivalents, N O + , with on less electron than NO.) (Stamler et at., 1992). Nitric oxide (NO-) itself can participate in reactions to form products that are toxic to nerve cells, such as peroxynitrite (ONOO-) and its breakdown products including hydroxyl radical (H0.)-like compounds (Beckman el al., 1990; Dawson et al., 1991; Radi et al., 1991; T . M. Dawson et al., 1992; Lei et al., 1992; Lipton et al., 1993). In other redox states, however, monoxides of nitrogen can interact with thiol groups, such as those constituting the redox modulatory sites of the NMDA receptor, by a nitrosylation reaction, resulting in transfer of the NO group to a thiol (Stamler et al., 1992; Lipton et al., 1993). This action results in downregulation of NMDA receptor activity and protects neurons from excessive stimulation of the receptor (Lei et al., 1992). Patients can be made tolerant

Ca2+ CHANNELS, NMDA, AND AIDS-RELATED NEURONAL INJURY

21

to the cardiovascular effects of nitroglycerin within hours of continuous therapy. Under these conditions, our laboratory has shown in animal models that the extent of NMDA receptor-mediated neurotoxicity can be markedly attenuated in the absence of behavioral or systemic side effects of the drug (Manchester et al., 1993). Nevertheless, the exact dosing regimen must be carefully worked out before this technique is applied to humans. Other promising reagents that appear to act either directly or indirectly on the NMDA redox modulatory site include oxidized glutathione (Gilbert et al., 1991; Levy et al., 1991; Sucher and Lipton, 1991) and the putative essential nutrient and redox cofactor pyrroloquinoline quinone (PQQ) (Aizenman et al., 1992). In addition, there are other important modulatory sites of the NMDA receptor, several of which are illustrated in Fig. 2. Antagonists of each of these sites could possibly be useful in the treatment or prevention of NMDA receptor-mediated neurotoxicity. For the purposes of this review, I have chosen to highlight only two of these, the ion channel and redox modulatory sites. The other modulatory sites may become therapeutically relevant, however, if clinically tolerated antagonists can be developed to interact with them. Intensive research efforts along these lines are now under way in both academic institutions and the pharmaceutical industry, which are exploring, for example, antagonists of the glycine coagonist site of the NMDA receptor. Since NMDA and non-NMDA receptor stimulation alike lead to neuronal depolarization and consequent activation of VDCCs, blockade of VDCCs might also ameliorate neurotoxicity, as discussed above. It has become apparent that different subpopulations of neurons have different repertoires of VDCCs, so it might be anticipated that an antagonist specific for a particular type of calcium channel may be effective only in certain regions of the brain o r for certain cell types (vide supra) (Lipton, 1991a,b; Regan et al., 1991). Therefore, it will be important to develop antagonists specific for these various types of calcium channels, and many investigators are working in this area. Currently available in the clinics are CNS-permeable antagonists of the L-type calcium channel, such as nimodipine. Other calcium channel antagonists that are permeable to the blood-brain barrier are also being tested in multicenter trials for entities other than the AIDS dementia complex (Lipton, 1991b).

IX. Excitatory Amino Acid Antagonist Treatments on the Horizon

Among the aforementioned classes of NMDA antagonists, the pharmaceutical industry is currently sponsoring in humans Phase 1/11 studies

22

STUART A. LIPTON

for stroke using the putative open-channel blockers CNS 1102 (Cambridge Neuroscience, Inc.) and dextrorphan (Roche); there is interest in testing the related compound dextromethorphan for amyotrophic lateral sclerosis (ALS). There is some evidence that the dextrorphan-like compounds also antagonize VDCCs as well as NMDA receptor-operated channels, which might be a helpful dual property (Carpenter etal., 1988). Other current clinical trials include a Phase I1 study (for stroke) using the NMDA competitive antagonist CGS 19755 (Ciba-Geigy). All of these trials to date have involved dose escalation and safety. Other companies are currently investigating both NMDA and non-NMDA antagonists, but for proprietary reasons information is scanty, and the indications d o not yet include the AIDS dementia complex. Based on animal testing, it is quite possible that for various forms of glutamate-related neurotoxicity, a combination of agents may be the most effective, e.g., combining calcium channel antagonists with NMDA antagonists (Uematsu et al., 1991; Hewitt and Corbett, 1992; Rod and Auer, 1992). Human clinical studies for indications other than the AIDS dementia complex are also in progress using agents that work downstream from EAA receptors. These include gangliosides (GMl),which are being tested for improvement of outcome after stroke (Rocca et al., 1992), and the 2 l-aminosteroid, tirilazad mesylate. However, recent reports that gangliosides can result in a polyneuropathy resembling Guillain-Barre syndrome have caused several authorities to conclude that clinical studies with ganglioside in humans should be suspended pending further assessment of this problem (Figueras et al., 1992; Raschetti et al., 1992). Finally, a case can be made that the proven NMDA open-channel blocker memantine (as well as its less potent cogener, amantadine) has already been in clinical use for years because it is known to ameliorate some of the symptoms of Parkinson’s disease. Furthermore, it is now known that the level of memantine (2-12 p M ) achieved in the human brain during this form of treatment (Wesemann et al., 1980) can afford protection from NMDA receptor-mediated neurotoxicity both in vitro and in vim (Seif el Nasr et al., 1990; Erdo and Schafer, 1991; Chen et al., 1992; Keilhoff and Wolf, 1992; Osborne and Quack, 1992). Recently, our laboratory and another independent group have reported that low micromolar levels of memantine can also protect neurons from damage induced by gp120 in uitro (Lipton, 1992a; Muller et al., 1992) and in vivo in an animal model (Lipton and Jensen, 1992). These preliminary findings raise the possibility that a clinically tolerated NMDA antagonist, memantine, might be useful in the treatment or prevention of the AIDS dementia complex. Therefore, it has been proposed to study the use of memantine as an adjunctive therapy with anti-retroviral drugs such as

Ca2+ CHANNELS, NMD.4, AND AIDS-RELATED NEURONAL INJURY

23

zidovudine o r didanosine, and the AIDS Clinical Trials Group of the NIH is currently considering this option.

X. Conclusion

Although it is likely that a complex web of cell interactions leads to neuronal loss in AIDS, HIV-infected macrophages or gpl20-stimulated macrophages release toxins whose action appears to be mediated by a final common pathway involving excessive stimulation of neurons by EAAs, such as glutamate and quinolinate. This represents at least one complete pathway to neuronal injury that is amenable to pharmacotherapy, although other pathways may also exist. A strong body of scientific evidence supports the premise that the mechanism for this form of HIV-related neuronal injury is similar to that currently thought to be responsible for a wide variety of acute and chronic neurological diseases (Choi, 1988; Meldrum and Garthwaite, 1990; Lipton, 1992b). EAAs apparently exert this excitotoxic effect by engendering an excessive influx of Ca2+ into neurons. Currently, there is intensive investigation to discover clinically tolerated drugs to combat the neurotoxic effects associated with the excessive stimulation of glutamate receptors or the events triggered downstream to receptor activation. One therapeutic approach has been to use glutamate receptor antagonists, and although several promising drugs are already in hand, additional agents are needed. With the possibility of a final common pathophysiology involving EAA receptors for many disorders of the central nervous system, including at least in part the AIDS dementia complex, the future development and testing of safe and effective EAA antagonists should become a high priority.

Acknowledgments

I thank my co-workers, Drs. E. B. Dreyer, N. J. Sucher, V. H . 4 . Chen, P. K. Kaiser, M. Oyola, S. Lei, J. Pellegrini, D. Zhang, and Y.-B. Choi for insightful discussions, and Dr. D. Leifer for comments on an earlier version of the manuscript. This work was supported by NIH Grants HD29587, EY05477, EY09024, NS07264; the American Foundation for AIDS Research; and an Established Investigator Award from the American Heart Association.

24

STUART A. LIPTON References

Abele, A. E., Scholz, K. P., Scholz, W. K., and Miller, R. J. (1990). Neuron 4, 413-419. Aizenman, E., and Hartnett, K. A. (1992). Bruin Res. 585, 28-34. Aizenman, E., Lipton, S. A., and Loring, R. H. (1989). Neuron 2, 1257-1263. Aizenman, E., Harnett, K.A., and Reynolds, I. J. (1990). Neuron 5, 841-846. Aizenman, E., Harnett, K. A., Zhong, C., Gallop, P. M., and Rosenberg, P. A. (1992). J . Neurosci. 12, 2362-2369. Beckman, J. S., Beckman, T. W., Chen, J., Marshall, P. A., and Freeman, B. A. (1990). Proc. Nutl. Acud. Sci. U.S.A. 87, 1620-1624. Bhat, S., Spitalnik, S. L., Conzalez-Scarano, F., and Silberberg, D. H . (1991). Proc. Nutl. Acad. Sci. U.S.A. 88, 7131-7134. Bito, H., Nakamura, M., Honda, Z., Isumi, T., Iwatsubo, T., Seyama, T . , Sgbra, A,, Kido, Y., and Shimizu, T. (1992). Neuron 9, 285-294. .) U. Bormann, J. (1989). Eur. J. Pharmaco/. 166, 591-592. Brenneman, D. E., Westbrook, G. L., Fitzgerald, S. P., Ennist, D. L., Elkins, K. L., Ruff, M., and Pert, C. B. (1980). Nature (London) 335, 639-642. Brenneman, D. E., Nicol, T., Warren, D., and Bowers, L. M. (1990).J . Neurosci. Res. 25, 386-394. Buzy, J., Brenneman, D. E., Pert, C. B., Martin, A., Salazar, A,, and Ruff, M. R. (1992). Bruin Res. 598, 10-18. Carpenter, C. L., Marks, S. S., Watson, D. L., and Greenberg, D. A. (1988). Bruin Res. 439,372-375. Chen, H . 4 . V., Pellegrini, J. W., Aggarwal, S. K., Lei, S. Z., Warach, S., Jensen, F. E., and L Lipton, S. A. (1992).J. Neurosci. 12, 4427-4436. Choi, D. W. (1987). Bruin Res. 403, 333-336. Choi, D. W. (1988). Neuron 1,623-634. Choi, D. W., Peters, S., and Viseskul, V. (1987).J. Pharmucol. Exp. Ther. 242, 713-720. Chung, 1. Y., and Benveniste, E. N. (1990).J. Zmmunol. 144, 2999-3007. C1ark.G. D., Happel,L.T.,Zorumski,C. F.,andBazan,N.G. (1992). Neuron9,1211-1216. Conti, P., Reale, M., Barbacane, R. C., Bongrazia, M., Panara, M. R.,and Fibre, S. (1989). In “Prostaglandins in Clinical Research: Cardiovascular System.” Alan R. Liss, New York. Davies, S. N., Alford, S. T., Coan, E. J., Lester, R. A,, and Collingridge, G. L. (1988). Neurosci. Lett. 92, 2 13-2 17. Dawson, T. M., Dawson, V. L., and Snyder, S. H. (1990). Ann. Neurol. 32,297-311. Dawson, V. L.. Dawson, T. M., London, E. D., Bredt, D. S., and Snyder, S. H. (1991). Proc. Nutl. Acud. Sci. U.S.A. 88, 6368-6371. Dawson, V. L., Dawson, T. M., Uhl, G. R., and Snyder, S. H . (1992). SOC.Neurosci. Abstr. 18, 756. Dawson, V. L., Dawson, T. M., Uhl, G. R., and Snyder, S. H. (1993). Proc. Nutl. Acad. Sci. U.S.A. 90, 3256-3259. Dreyer, E. B., Kaiser, P. K., Offermann, J. T., and Lipton, S. A. (1990). Science 248, 364-367. Dubois, C., Bissonnette, E., and Rola-Pleszczynski, M. (1989).J. Zmmunol. 143, 964-970. Erdo, S. L., and Schafer, M. (1991). Eur.1. Pharmacol. 198, 215-217. Everall, I. P., Luthbert, P. J., and Lantos, P. L. (1991). Lancet 337, 1119-1 121. Figueras, A., et al. (1992). Br. A4ed. J. 305, 1330- 1331. Genis, P., Jett, M., Bernton, E. W., Boyle, T., Gelbard, H. A., Dzenko, K., Keane, R. W.,

Ca2+ CHANNELS, NMDA, AND AIDS-RELATED NEURONAL lNJURY

25

Resnick, L., Mizrachi, T., Volsky, D. J., Epstein, L. G., and Gendelman, H. E. (1992). J. Exp. Med. 176, 1703-1718. Gilbert, K. R., Aizenman, E., and Reynolds, I. J. (1991).Neurosci. Lett. 133, 11-14. Giulian, D., Vaca, K., and Noonan, C. A. (1990). Science 250, 1593-1596. Giulian, D., Vaca, K., and Corpuz, M. (1993)./. Neurosci. 13, 29-37. Giulian, D., Wendt, E., Vaca, K., and Noonan, C. A. (1993). Proc. Natl. Acad. Sci. U.S.A. 90,2769-2773. Glowa, J. R., Panlilio, L. V., Brenneman, D. E., Gozes, I., Fridkin, M., and Hill, J. M. ( 1 992). Brain Res. 570, 49-53. Goldrnan, R. S., and Finkbeiner, S. M. (1988). N . Engl. J . Med. 319, 1224-1225. Gourdou, I . , Mabrouk, K., Harkiss, G., Marchot, P., Watt, N., Hery, F., and Vigne, R. (1990). C.R. Seances Acad. Sci., Ser. 3 311, 149-155. Harouse, J. M., Bhat, S., Spitalnik, S. L., Laughlin, M.,Stefano, K., Silberberg, D. H., and Gonzalez-Scarano, F. (199 1). Science 253,320-323. Hayman, M., Arbuthnott, G., Harkiss, G., Brace, H., Filippi, P., Philippon, V., Thompson, D., Vigne, R., and Wright, A. (1993). Neuroscience (Oxford) 53, 1-6. Hewitt, K., and Corbett, D. (1992). Stroke 23, 82-86. Heyes, M. P., Rubinow, D., Lane, C., and Markey, S. P. (1989). Ann. Neurol. 26, 275-277. Heyes, M. P., Brew, B. J., Martin, A,, Price, R. W., Salazqr, A. M., Sidtis, J. J., Yergey, J. A., Mouradian, M. M., Sadler, A,, Keilp, J., Rubinow, D., and Markey, S. P. (1991). Ann. Neural. 29, 202-209. Heyes, M. P., Saito, K., and Markey, S. P. (1992). Biochem. J. 283, 633-635. Kaiser, P. K . , Offermann, J. T., and Lipton, S. A. (1990). Neurology 40, 1757-1761. Karschin, A., Aizenman, E., and Lipton, S. A. (1988).J . Neurosci. 8, 2895-2906. Keilhoff, G., and Wolf, G. (1992). Eur.J. Pharmacol. 219, 451-454. Ketzler, S., Weis, S., Haug, H., and Budka, H. (1990). Acta Neuropathol. 80, 90-92. Koek, W., Woods, J. H., and Winger, G. D. (1988). Pharmacol. Exp. Ther. 245, 969-974. Kohler, C., Eriksson, L. G., Okuno, E., and Schwarz, R. (1988). Neuroscience (Oxford) 27, 49-76. Kornecki, E., and Ehrlich, Y. H. (1988).Science 240, 1792-1794. Kornecki, E., and Ehrlich, Y. H. (1991). Lzpids 26, 1243-1246. Latov, N., Apostolski, S., Quattrini, A., Lavasta, J., Lugaresi, A., McAlarney, T., Sadiq, S. A., and Hays, A. P. (1993). Neurology 43, A384. Lei, S. Z., Pan, 2. H., Aggarwal, S. K., Chen, H. S., Hartman, J.. Sucher, N. J., and Lipton, S. A. (1992). Neuron 8, 1087-1099. Levi, G., Patrizio, M., Bernardo, A., Petrucci, T. C., and Agresti, C. (1992). Proc. Natl. Acad. Sci. U.S.A. 90, 1541-1545. Levy, D. I., and Lipton, S. A. (1990).Neurology 40, 852-855. Levy, D. I., Sucher, N. J., and Lipton, S. A. (1990). Neurosci. Lett. 110, 291-296. Levy, D. I., Sucher, N. J., and Lipton, S. A. (1991). NeuroReport 2, 345-347. Lindsberg, P. J., Hallenbeck, J. M., and Feurstein, G. (1991). Ann. Neural. 30, 117-129. Lipton, S. A. (1991a). Ann. Neural. 30, 110-114. Lipton, S. A . (1991b). Ado. Pharmucol. 22, 271-291. Lipton, S. A. (1992a). Neurology 42, 1403-1405. Lipton, S. A. (1992b). Trends Neurosci. 15, 75-79. Lipton, S. A. (1992~ ).NeuroReport 3, 913-915. Lipton, S. A. (1992d). Zn “HIV-I Infection of the Central Nervous System,” pp. 251-260. Hogrefe & Huber, Seattle. Lipton, S. A. (1993). Ann. Neural. 33, 227-228. Lipton, S. A., and Jensen, F. E. (1992). Sac. Neurosci. Abstr. 18, 757.

26

STUART A. LIPTON

Lipton, S . A., and Kater, S. B. (1989). Trends Neurosci. 12, 265-270. Lipton, S. A., and Rosenberg, P. A. (1994). N . Engl. J . Med. 330, 613-622. Lipton, S. A., Sucher, N. J., Kaiser, P. K., and Dreyer, E. B. (1991). Neuron 7, 111-118.. Lipton, S. A., Kaiser, P. K., Sucher, N. J., Dreyer, E. B., and Offermann, J. T. (1990). SOC.Neurosci. Abstr. 16, 289. Lipton, S. A., Choi, Y.-B., Pan, Z.-H., Lei, S. Z., Chen, H . 3 . V., Sucher, N. J., Singel, D. J . , Loscalzo, J., and Stamler, J . S. (1993). Nature 364, 626-632. Lo, T.-M., Fallert, C. J., Piser, T. M., and Thayer, S. A. (1992). Brain Res. 594, 189-196. MacDonald, J. F., Miljkovic, Z., and Pennefather, P. (1987).J. Neur@hysiol. 58, 251-266. Manchester, K. S., Jensen, F. E., Warach, S., and Lipton, S. A. (1993). Neurology 43, A365. Manzini, S., and Meini, S. (1991a).J. Lipid Mediators 3, 361-366. Manzini, S., and Meini, S. (1991b). 8 r . J . Pharmacol. 103, 1027-1032. Masliah, E., Achim, C. L., Ge, N., DeTeresa, R., Terry, R. D., and Wiley, C. A. (1992). Ann. Neurol. 32, 321-329. Mattson, M. P., Lee, R.E., Adams, M. E., Guthrie, P. B., and Kater, S. B. (1988). Neuron 1,865-876. Meini, S., Evangelista, S., Geppetti, P., Szallasi, A,, Blumberg, P. M., and Mazini, S. (1992). Am. Rev. Respir. Dis. 146, 930-934. Meldrum, B., and Garthwaite, J. (1990). Trends Phunnacol. Sci. 11, 379-387. Merrill, J. E., and Chen, I. S. Y. (1991). FASEBJ. 5, 2391-2397. Merrill, J. E., Koyanagi, Y., and Chen, I. S. Y. (1989).J. Virol. 63,4404-4408. Mervis, R. F., Hill, J. M., and Brenneman, D. E. (1990). Int. ConJ AIDS 6, 184. Miller, R. J. (1991). Prog. Neurobiol. 37, 255-285. Morganati-Kossmann, M. C., Kossmann, T., and Wahl, S. M. (1992). Trends Pharmacol. S C ~ 13, . 286-290. Miiller, W. E. G., Schroder, H. C., UshlJima, H., Dapper, J., and Bormann, J . (1992). Eur. J . Phannacol. 226, 209-214. Offermann, J., Uchida, K., and Lipton, S. A. (1991). SOC.Neurosci. Abstr. 17, 927. Olney, J. W., Labruyere, J., and Price, M. T. (1989). Science 244, 1360-1362. Osborne, N. N., and Quack, G. (1992). Neurochem. Int. 21, 329-336. OShaughnessy, C. T., and Lodge, D. (1988). Eur. J . Pharmacol. 153, 201-209. Palmer, M . R., Mathews, W. R., Hoffer, B. J., and Murphy, R. C. (1981).J. Pharmacol. Exp. Ther. 219, 91-96. Pignol, P., Sylvie, H., Mencia-Huerta, J.-M., and Rola-Pleszczynski, M. (1987).Prostaglandins 33,93 1-939. Poubelle, P. E., Gingras, D., Demers. C., Dubois, C., Harbour, D., Grassi, J., and RolaPleszczynski, M. (1991). Immunology 72, 181-187. Price, R. W., Brew, B., Sidtis, J., Rosenblum, M., Scheck, A. C., and Clearly, P. (1988). Science 239, 586-592. Pulliam, L., Herndler, B. G., Tang, N. M., and McGrath, M. S. (199 1).J . Clin. Invest. 87, 503-512. Radi, R., Beckman, J . S., Bush, K. M., and Freeman, B. A. (1991).J. Biol. Chem. 266, 4244-4250. Raschetti, R., Maggini, M., and Popoli, P. (1992). Lancet 340, 60. Regan, L. J., Sah, D. W. Y., and Bean, B. P. (1991).Neuron 6, 269-280. Reynolds, I. J., Rush, E. A., and Aizenman, E. (1990). Br. J . Pharmacol. 101, 178-182. Robbins, D. S., Shirazi, T., Drysdale, B. E., Leiberman, A., Shin, H. S., and Shin, M. L. (1987).J . Immunol. 139,2593-2597. Rocca, W. A., Dorsey, F. C., Grigoletto, F., Gent, M., Roberts, R. S., Walker, M. D., Easton, J. D., Bruno, R., Carolei, A., Sancesario, G., and Fieschi, C. (1992). Stroke23,519-526.

Ca2+ CHANNELS, NMDA. A N D AIDS-RELATED NEURONAL INJURY

27

Rod, M. R., a n d h e r , R. N. (1992). Stroke 23, 725-732. Rogawski, M. A., and Porter, R. J . (1990).Pharmacol. Rev. 42, 223-286. Sabatier, J.-M., Vives, E., Mabrouk, K., Benjouad, A., Rochat, H., Duval, A., Jue, B., a n d Bahraoui, E. (1991).J. Virol. 65, 961-967. Savio, T., and Levi, G. (1993). J . Neuroscz. Res. 34, 265-272. Seif el Nasr, M., B. P., Rossberg, C., Mennel, H.-D., and Krieglstein, J. (1990). Eur. J . Pharmacol. 185, 19-24. Selmaj, K. N., Farooq, M., Norton, T., Raine, C. S., and Brosman, C. F. (199O).J. Immunol. 144, 129-135. Soliven, B., a n d Albert, J. (1992).J. Neurosci 12, 2665-2671. Speciale, C., Okuno, E., and Schwarz, R. (1987). Brain Res. 436, 18-24. Stamler, J. S., Singel, D. J., and Loscalzo,J. (1992). Science 258, 1898-1902. Sucher, N. J., a n d Lipton, S. A. (1991).J. Neurosci. Res. 30, 582-591. Sucher, N. J.. Wong, L. A., and Lipton, S. A. (1990). NeuroReport 1.29-32. Sucher, N. J., Lei, S. Z., and Lipton, S. A. (1991). Brain Res. 551, 297-302. Sweetnarn, P. M., Saab, 0. H., Wroblewski, J. T., Price, C. H., Larbon, W., and Ferkany, J. W. (1993).Eur. J . Neurosci. 5, 276-283. Tang, L. H., a n d Aizenman, E. (1993).J. Physzol. (London) 465, 303-323. Tardieu, M., Hery, C., Peudenier, S., Roespflug, O., and Montagnier. L. (199’2). Ann. Neurol. 32, 11-17. Tenhula, W. N., Xu, S. Z., Madigan, M. C., Heller, K., Freeman, W. R., and Sadun, A. A. (1992). Am. J. Ophthalmol. 15, 14-20. Toggas, S. M., Masliah, E., Rockenstein, E. M . , Kall, G . F., Abraham, C. R., and Mucke. L. (1994). Nature (London) 367, 18H-193. Tyor, W. R., Glass, J. D., Griffin, J . W., Becker, S., McArthur, J. C., Bezman, L., and Griffin, D. E. (1992). Ann. Neurol. 31, 349-360. Uematsu, D., Araki, N., Greenberg, J. H . , Sladky, J., and Reivich, M. (1991). Neurology 41,88-94. Valone, F. H., and Epstein, L. B. (1988).J. Immzinol. 141, 3945-3950. Valone, F. H., Philip, R., and Debs, R. J. (1988). lmmunolopy 64, 715-718. van den Berg, L. H., Sadiq, S. A., Lederman, S., and Latov, N. (1992). J . Neuru.rci. Res. 33, 513-518. Vitkovic, L., Kalebic, T., d e Cunha, A., and Fauci, A. S. (1990). J . Neuruimmunol. 30, 153- 160. Wahl, L. M . , Corcoran, M. L., Pyle, S. W., Arthur, L. 0..Harel-Bellan, A., and Farrar, W. L. (1989). Proc. Nall. Acad. Sci. U.S.A. 86, 621-625. Weiss, J . H., Hartley, D. M., Koh. J., and Choi, D. W. (1990). Science 247, 1474-1477. Werner, T., Ferroni, S., Saermark. T., Rrack-Werner, R., Ranati, R. B., Mayer, R., Steinaa, L., Kreutzberg, G. W . , and Erfle, V. (1991). AIDS 5, 1301-1308. Wesemann, W., Sturn, G., and Funfgeld, E. W. (1980).J . Neural. Trunsm. 16(Suppl.), 143- 148. Wiley, C. A., Masliah, E., Morey, M., Lemere, C.?DeTeresa, R. M., Grafe, M. R., Hansen, L. A., and Terry, R. D. (1991). Ann. Neurol. 29, 651-657. Wolf, G., Keilhoff, G., Fischer, S., and Hass, P. (1990). Neurosci. Lett. 117, 207-21 1. Wolf, G., Fischer, S., Hass, P., Abicht, K., and Keilhoff, G. (1991).Neurosctence (Oxjilrd) 43, 31-34. Zeevalk, G . D., and Nicklas, W. J. (1992).J . Neurochem. 59, 121 1-1220.

This Page Intentionally Left Blank

PROCESSING OF ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN: CELL BIOLOGY, REGULATION, AND ROLE IN ALZHEIMER DISEASE

Sam Gandy* and Paul Greengardt *Department of Neurology and Neuroscience, Cornell University Medical College, New York, New York 10021 and tLaboratory of Molecular and Cellular Neuroscience, The Rockefeller University, New York, New York 10021

I. 11. 111. 1V. V.

VI. VII.

VIII.

IX. X.

Introduction Alzheimer Disease Is Associated with an lntracranial Amyloidosis APP Structure Gives Clues to Some of Its Functions APP Is Processed via Several Distinct Enzymatic and Subcellular Pathways “Alternative” Pathways of APP Metabolism Provide Clues to the Source of AP-Am yloid A@-Amyloid Is a Normal Constituent of Body Fluids and the Conditioned Medium of Cultured Cells Evidence Suggests the Existence of an Enzyme, p-Secretase, That Cleaves APP at the Amino Terminus of the AP-Amyloid Domain APP Mutations in Familial Cerebral Amyloidoses Occur within or near the AP-Amyloid Domain, Segregate with Disease in Affected Kindreds, and Yield APP Molecules That Display Some Proamyloidogenic Properties Signal Transduction via Protein Phosphorylation Regulates the Relative Utilization of APP Processing Pathways Beyond AP-Amyloid: Other Molecular Factors in Amyloidogenesis and Factors Differentiating Aging-Related Cerebral Amyloidosis from Alzheimer Disease References

1. Introduction

Alzheimer disease (AD) is characterized by an intracranial amyloidosis that develops in an age-dependent manner, and that appears to be dependent on the production of AP-amyloid by proteolysis of its integral membrane precursor, the Alzheimer A@-amyloidprecursor protein (APP). Evidence causally linking APP to Alzheimer disease has been provided by the discovery of mutations within the APP coding sequence that segregate with disease phenotypes in autosomal dominant familial cerebral amyloidoses, including some types of familial Alzheimer disease (FAD). Although FAD is rare ((10% of all AD), the characteristic clinicopathologiINTERNATIONAL REVIEW OF NEUROBIOLOGY, VOL. 96

29

Copyright B 1994 by Academic Press. Inc. All rights of reproduction in any form reserved.

30

SAM GANDY A N D PAUL GREENGARD

cal features-amyloid plaques, neurofibrillary tangles, synaptic and neuronal loss, neurotransmitter deficits, dementia-are apparently indistinguishable when FAD is compared with typical, common, “nonfamilial,” or sporadic AD (SAD). The nature and regulation of pathways for the cellular processing of APP have been extensively characterized and recent data demonstrate that soluble AP-amyloid is released from various cells and tissues in the course of normal cellular metabolism. To date, studies of APP catabolic intermediates and soluble AP-amyloid in SAD tissues and fluids have not provided specific SAD-associated changes in APP metabolism. However, studies of some clinically relevant mutant APP molecules from FAD families have yielded evidence that APP mutations can lead to enhanced generation or aggregability of AP-amyloid, consistent with a pathogenic role in AD. In addition, genetic loci for FAD have been discovered which are distinct from the immediate regulatory and coding regions of the APP gene, indicating that defects in molecules other than APP can also specify cerebral amyloidogenesis and FAD. It remains to be elucidated which, if any, of these rare genetic causes of AD is most relevant to our understanding of typical, comtnon SAD.

II. Alzheimer Disease Is Associated with an lntracranial Amyloidosis

Amyloid is a generic description applied to a heterogeneous class of tissue protein precipitates that have the common feature of @pleated sheet secondary structure, a characteristic that confers affinity for the histochemical dye Congo red (Tomlinson and Corsellis, 1984). Amyloids may be deposited in a general manner throughout the body (systemic amyloids) or confined to a particular organ (e.g., cerebral amyloid). AD is characterized by clinical evidence of cognitive failure in association with cerebral amyloidosis, cerebral intraneuronal neurofibrillary pathology, neuronal and synaptic loss, and neurotransmitter deficits (Tomlinson and Corsellis, 1984). T h e cerebral amyloid of AD is deposited around meningeal and cerebral vessels, as well as in gray matter. In gray matter, the deposits coalesce into structures known as plaques. Parenchymal amyloid plaques are distributed in brain in a characteristic fashion, differentially affecting the various cerebral and cerebellar lobes and cortical laminae. The main constituent of cerebrovascular amyloid was purified and sequenced by Glenner and Wong in 1984. This 40- to 42-amino acid

31

ALZHEIMER AD-AMYLOID PRECURSOR PROTEIN

polypeptide, designated “P protein” (Glenner and Wong 1984a,b) [or, according to Masters and colleagues (1985), “A4”; now standardized as AP by Husby et al., 19931, is derived from a 695- to 770-amino acid precursor, termed the AP-amyloid precursor protein (APP; Fig. l ) , which was discovered by molecular cloning (Goldgaber et al., 1987; Tanzi et al., 1987, 1988; Kang et al., 1987; Robakis et al., 1987; Ponte et al., 1988; Kitaguchi et al., 1988).

111. APP Structure Gives Clues to Some of Its Functions

T h e deduced amino acid sequence of APP predicts a protein with a single transmembrane domain (Goldgaber et al., 1987; Tanzi et al., 1987, 1988; Kang et al., 1987; Robakis et al., 1987; Ponte et al., 1988; Kitaguchi et al., 1988). Isoform diversity is generated by alternative mRNA splicing, and isoforms of 751 and 770 amino acids include a protease inhibitor domain [“Kunitz-type protease inhibitor” domain (KPI) (Ponte et al., 1988; Tanzi et al., 1988; Kitaguchi e l a l . , 1988)l in the extracellular region

KPI

ox-2

OlA4

n

H

coated pn targeting

Potential phosphorylation sites (lhr-654, Ser-655)

FIG. 1. Structure of the Alzheimer A@-amyloidprecursor protein. (Courtesy of Dr. Gregg Caporaso; numbering according to APP695, Kang et al., 1987. pIA4 domain = Ap domain; see text or Husby et al., 1993.)

32

SAM CANDY AND PAUL GREENGARD

of the APP molecule. The ectodomains of the protease inhibitor-bearing isoforms of APP are identical to molecules that had been identified previously based on their tight association with proteases, and thus were designated “protease nexin 11” (PN-11; Van Nostrand and Cunningham, 1987; Oltersdorf el al., 1989; Van Nostrand et al., 1989). Identical molecules are also present in the platelet a-granules, where they were described under the name “factor XIa inhibitor” (XIaI; Smith et al., 1990; Bush et al., 1990). On degranulation of the platelet, factor XIaI/PN-II/ APP exerts an antiproteolytic effect on activated factor XIa at late steps of the coagulation cascade. Recent evidence suggests that KPIlacking isoforms may also act as regulators of proteolysis (Miyazaki et al., 1993). Another physiological role(s) for APP is yet unknown, although evidence from several independent lines of inquiry suggests that APP may play a role in transmembrane signal transduction (Nishimoto et al., 1993) and/or calcium metabolism (Mattson et al., 1993; Arispe et al., 1993). In addition, potential functional motifs in APP have been recognized by the presence of consensus sequences or by experimental implication. Some of these motifs suggest a role in metal ion binding (Bush et al., 1992),heparin binding (Schubert etal., 1989),cell-cell interaction (Konig et al., 1992), and/or functioning as a receptor for a currently unrecognized ligand (Kang et al., 1987; Chen et al., 1990). In some investigations, Saitoh and colleagues have accumulated evidence that APP may play a role in regulating cell growth (Saitoh et al., 1989). Recently, novel APPlike proteins (APLPs) have been discovered (Wasco et al., 1992,1993; Slunt et al., 1994), suggesting that APP may be a member of a larger family of related molecules. APLPs are highly homologous to APP and to each other, but APLPs lack the A@-amyloiddomain and therefore cannot serve as precursors to AP-amyloid.

IV. APP Is Processed via Several Distinct Enzymatic and Subcellular Pathways

APP is initially synthesized and cotranslationally inserted into membranes in the endoplasmic reticulum (ER). Studies of APP metabolism in the presence of either brefeldin A or monensin have, to date, failed to implicate the ER as an important site for discrete proteolytic processing of APP (Caporaso et al., 1992a). Following its exit from the ER, APP traverses the Golgi apparatus, where it is subjected to N- and O-glycosylation, tyrosyl sulfation, and

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

33

sialylation (Weidmann et al., 1989; Oltersdorf et al., 1990). APP is also phosphorylated in both the extracellular and the cytoplasmic domains and preliminary evidence implies that some of these events may occur in an early compartment of the central vacuolar pathway (Knops et al., 1993; Suzuki et al., 1994).In addition, some APP molecules are chondroitin-sulfated in their ectodomains (Shioi et al., 1992). T h e proteolytic processing steps for APP have been a subject of intense interest, in part due to early evidence that excluded the possibility that AD was frequently associated with APP gene mutations or with disordered APP transcription (Koo et al., 1990). One attractive possibility, then, was that AD might be a disorder of APP processing. This possibility was strengthened by early evidence for an APP processing pathway that precluded AP-amyloid generation (Sisodia et al., 1990; Esch et al., 1990), implying that a defect in this pathway might underlie AD. Several proteolytic cleavage products of APP processing have now been definitively identified by purification and sequencing. The first to be identified (Weidemann etal., 1989)was a fragment that results primarily from the cleavage that occurs within the AP-amyloid domain. A large amino-terminal fragment of the APP extracellular domain [protease nexin-I1 (Van Nostrand and Cunningham, 1987; Oltersdorf et al., 1989; Van Nostrand et al., 1989) or s-APP or APPs, for soluble APP (Citron et al., 1992)], is released into the medium of cultured cells and into the cerebrospinal fluid (Weidemann et al., 1989; Palmert et al., 1989; Oltersdorf et al., 1990), leaving associated with the cell, a small nonamyloidogenic carboxyl-terminal fragment. This pathway is currently designated the a-secretory cleavage/release processing pathway for APP, so-called because the (yet undiscovered) enzyme that performs this nonamyloidogenic cleavage/release has been designated “a-secretase” (Esch el al., 1990; Seubert et al., 1993). Thus, one important processing event in the biology of APP acts to preclude amyloidogenesis by proteolyzing APP within the AP-amyloid domain. Few details are available concerning the molecular nature of asecretase, although it is very likely to be a member of a class of enzymes that regulates the “shedding” of ectodomains from a wide variety of transmembrane molecules, including growth factor precursors, cell adhesion molecules, receptors, and ectoenzymes (Ehlers and Riordan, 1991). Surprisingly, these enzymes appear to act primarily at or near the cell surface and to specify cleavage of substrates at a certain distance from the plasma membrane, largely without regard for the primary sequence surrounding the cleavage site (Sisodia, 1992; also Maruyama et al., 1991; Sahasrabudhe et al., 1992). Based on studies of proteolytic

34

SAM GANDY A N D PAUL GREENGARD

processing of the TGF-a precursor and the c-kit ligand precursors [which also appear to be cleaved by similar, cell-surface proteinase activities (Pandiella and Massague, 1991; Pandiella et al., 1992)], it appears that “secretase-like” activities may be heterogeneous at the molecular level (i.e., several individual proteinase species probably exist). This conclusion is based on the observation that, depending on the substrate assayed, slightly different protease inhibitor sensitivity profiles were obtained in studies of TGF-a “secretase” in side-by-side comparison with c-kit ligand family “secretases.” Intracellular signal transduction, especially via protein kinase C, is commonly an important regulatory mechanism for processing of molecules via “secretase-like” pathways (see below). The possibility currently exists that the activities of some secretases are regulated by the phosphorylation state of the enzymes themselves; if true, this would provide the first known examples of proteases whose activities are regulated by their states of phosphorylation.

V. “Alternotive“ Pathways of APP Metabolism ProvideClues to the Source of Ap-Amyloid

Due to issues of peptide conformation, peptide aggregation, and antibody reagent insensitivity, the AP-amyloid molecule was not initially detected as a normal metabolite of APP, neither in brain, nor in cerebrospinal fluid, nor in a cell culture system. In fact, until mid-1992, APamyloid was generally, described as being an “abnormal” metabolite of APP. Instead, early clues into AP-amyloidogenesis were provided by the observation of electrophoretic microheterogeneity of carboxyl-terminal fragments of APP. Such microheterogeneity was detected in association with high-level overexpression of human APP using recombinant vaccinia viruses (Wolf et al., 1990), baculoviruses (Gandy et al., 1992b), or stable transfection (Golde et al., 1992),in association with supraphysiological levels of protein phosphorylation (Buxbaum et al., 1990), and in human cerebral vessels (Tamaoka et al., 1992) and cortex (Estus et al., 1992). Antigenic characterization of carboxyl-terminal fragments of APP in cerebral vessels (Tamaoka et al., 1992) and cortex (Estus et al., 1992),in transfected cells (Golde et nl., 1992),and in the baculoviral overexpression system (Gandy et al., 1992b) provided the evidence that supported the possibility of “alternative” cleavage of APP molecules, giving rise to carboxyl-terminal fragments containing the complete AP-amyloid sequence, which in turn might give rise to AP-amyloid. Protein sequencing of the various putative amyloidogenic carboxyl-terminal species (candi-

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

35

date intermediates in the pathway to AP-amyloid deposition) has recently provided for their definitive identification (Cheung et al., 1994). The “alternative” (i.e.. non-intra-AP) cleavage suggested by this microheterogeneity prompted a search for additional intracellular routes for APP trafficking and cleavage. The existence of trafficking routes other than the a-secretory cleavage/release processing pathway was also suggested by the estimation that only about 20% of immature molecules are recovered as released molecules (in PC-12 cells; Caporaso et al., 1992b). Since evidence failed to suggest the existence of an important degradative pathway for APP in the ER (Caporaso et al., 1992a), several groups undertook experiments to determine whether acidic (endosomal/ lysosomal or trans-Golgi network) compartments of the cell were important in APP metabolism (Cole et al., 1989; Caporaso et al., 1992a; Golde et al., 1992; Haass et al., 1992a; Knops et al., 1992). The possibility of endosomal metabolism of APP was bolstered by the discovery of a clathrin-coated vesicle (CCV) targeting motif in the LDL receptor (Chen et ul., 1990). This motif, NPXY, was required for proper internalization of the LDL receptor and was also present in the sequence of the cytoplasmic tail of APP (Fig. 1). The copurification of APP with CCVs was subsequently demonstrated directly (Nordstedt et al., 1993). T h e fact that APP contains an NPXY motif associates APP with a host of cellsurface receptors and suggests the possibility that APP may be a receptor for a yet undiscovered or unrecognized ligand. I n other efforts to dissect the process of AP-amyloidogenesis, vesicleneutralizing agents (such as chloroquine and ammonium chloride) were applied to cultured cells, and these compounds were associated with greatly enhanced recovery of full-length APP and an array of carboxylterminal fragments, including nonamyloidogenic and potentially amyloidogenic fragments (Caporaso e l ul., 1992a; Estus et al., 1992; Golde et al., 1992; Haass et al., 1992a; Knops et al., 1992). A similar array of fragments was recovered from purified lysosomes (Haass et al., 1992a). This led to the formulation that both the potentially amyloidogenic carboxyl-terminal fragments and A@-amyloidmight be generated primarily in lysosomes. However, no AP-amyloid could be recovered from lysosomes (Haass et al., 1993), making this a less likely (but not impossible) scenario. T h e likelihood that AP-amyloid is generated in lysosomes was further diminished by the observation that vesicular neutralization fails to diminish consistently AP-amyloid production in certain cell types (Busciglio et al., 1993; see also below), although neutralization-induced stabilization of the standard array of potentially amyloidogenic carboxylterminal fragments is consistently apparent.

36

SAM GANDY AND PAUL GREENGARD

VI. Ap-Amyloid Is a Normal Constituent of Body Fluids and the Conditioned Medium of Cultured Cells

Until mid-1992, the prevailing notion of A@-aniyloidwas that of an abnormal, potentially toxic species, the production of which was perhaps relatively restricted to the brain in humans (and perhaps a few other species), and which occurred primarily in association with aging and AD. This concept became obsolete with the discovery by several groups that a soluble AP-amyloid species (presumably a forerunner of the aggregated fibrillar species that is deposited in senile plaque cores) is detectable in body fluids from various species and in the conditioned medium of cultured cells (Haass et al., 1992b; Seubert et al., 1992; Shoji et al., 1992), but is not detectable in the lysates of cultured cells. This so-called “soluble AP-amyloid” is apparently generated in a cellular compartment distal to the ER since brefeldin abolishes its generation and does not result in its accumulation inside cells (Haass et al., 1993). Vesicular neutralization compounds are effective in inhibiting AP-amyloid release from some cell types (Shoji et al., 1992), but this is not true for all cell types studied (Busciglio et al., 1993). The precise cellular locus (loci) involved in the amino- and carboxyl-terminal cleavages responsible for AP-amyloid generation has(have) not yet been unequivocally established. T h e consistent inability to recover AP-amyloid from cell lysates or from purified vesicles has led to a shift in focus away from the terminal degradative compartments of the cell (i.e., lysosomes) as possible sources for the generation of AP-amyloid. One plausible scenario for AP-amyloid production is that cleavage at the AD-amyloid amino-terminus is catalyzed by p-secretase (see below) in the precell surface limb of the constitutive secretory pathway, perhaps beginning in the trans-Golgi network (TGN). Cell-type-dependent variations in sensitivity of the TGN to neutralizing compounds may explain the observed dissociability of APamyloid generation from the apparent stabilization by these compounds of potentially amyloidogenic carboxyl-terminal fragments. Still unexplained is the cellular mechanism by which the carboxylterminus of AP-amyloid is generated, since this region of the APP molecule resides within an intramembranous domain. A plausible and conventional scenario for this step might involve the trafficking of APP or a potentially amyloidogenic fragment into a multivesicular body where vesiculated APP or an APP fragment may be liberated from the bilayer (Candy et al., 1992b). This is supported by ultrastructural evidence that multivesicular bodies are immunoreactive for APP epitopes (Caporaso et al., 1994). A multivesicular body containing wholly intraluminal

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

37

AP-amyloid could effect release of A@-amyloid into the extracellular space.

VII. Evidence Suggests the Existence of an Enzyme, p-Secretase, That Cleaves APP at the Amino Terminus of the Ap-Amyloid Domain

T h e possibility of heterogeneous cleavage along the constitutive secretory pathway (i.e., cleavage in the precell surface pathway or at the cellsurface) was initially discounted (Golde et al., 1992). However, Seubert and colleagues (1993) extended this line of investigation and succeeded in preparing an antibody that was specific for the free methionyl residue that would reside at the predicted carboxyl-terminus of such an alternatively cleaved and released attenuated PN-II-like (or APPs-like) molecule. This species was successfully detected as a component of the PN-II/APPs pool of cleaved and released APP ectodomains. The importance of this activity, designated “p-secretase,” was subsequently established by the discovery that a pathogenic FAD mutation in APP results in dramatic increases in AP-amyloid generation, which is probably attributable to an increase in P-secretase-type cleavage of APP, because the mutation enhances the substrate properties for cleavage by P-secretase (Felsenstein et al., 1994).

VIII. APP Mutations in Familic Cerebral Amyloidoses Occur within or near the A@Amyloid Domain, Segregate with Disease in Affected Kindreds, and Yield APP Molecules That Display Some Proamyloidogenic Properties

Certain mutations associated with familial cerebral amyloidoses have been identified within or near the AP-amyloid region of the coding sequence of the APP gene. These mutations segregate with the clinical phenotypes of either hereditary cerebral hemorrhage with amyloidosis, Dutch type (HCHWAD or FAD-Dutch; Fig. 1 ; van Duinen e t a ) . , 1987; Levy et al., 1990; Van Broeckhoven et al., 1990) or more typical familial Alzheimer disease (FAD; Fig. 1 ; Goate et al., 1991; Naruse et al., 1991; Murrell et al., 1991; Chartier-Harlin et al., 1991; Mullan et al., 1992a), and provide support for the notion that aberrant APP metabolism is a key feature of AD.

38

SAM CANDY AND PAUL CREENGARD

In FAD-Dutch, an uncharged glutamine residue is substituted for a charged glutamate residue at position 693 of APP770.This mutated residue is located in the extracellular region of APP, within the AP-amyloid domain, where it apparently exerts its proamyloidogenic effect by generating AP-amyloid molecules that bear enhanced aggregation properties (Wisniewski et al., 1991). Mutations in APP that are apparently pathogenic for more typical FAD have also been discovered. In the first discovered FAD mutation (Goate et al., 199l),an isoleucine residue is substituted for a valine residue at position 717 of APP,,,, within the transmembrane domain (Fig. l), at a position just downstream from the carboxyl-terminus of the APamyloid domain. Although a conservative substitution, the mutation segregates with FAD in pedigrees of American, European, and Asian origins, arguing against the possibility that the mutations represent irrelevant polymorphisms. Other pedigrees have been discovered in which affected members have either phenylalanyl (Murrell et al., 1991)or glycyl (Chartier-Harlin et al., 1991) residues at position 7 17. Neuropathological examination has verified the similarity of these individuals to typical SAD neuropathology (reviewed by Rossor, 1992; see also Lantos et al., 1992; Mann et al., 1992; Ghetti et al., 1992; Cairns et al., 1993; Kennedy et al., 1993). Although the 717 mutant APPs are the most common of the FADcausing APP mutations, the mechanism by which the 717-mutant APPs exert their effects remains to be clarified. The location of the missense substitution raises the possibility either that the mutation may directly affect proteolytic cleavage (e.g., by leading to the production of extended, perhaps more hydrophobic and thus hyperaggregable AP-amyloid molecules; Cai et al., 1993) or that the mutation may otherwise influence the function, trafficking, or biology of the APP molecule. Missense mutations in other integral molecules are associated with alterations in their biological activities (e.g., the oncogene neu, Bargmann et al., 1986), their trafficking and proteolysis (e.g., T-cell receptor, Bonifacino et al., 1990), or their ability to effect functional physiological changes in response to phosphorylation of their cytoplasmic domains (e.g., CFTR, Schoumacher el al., 1987; Li et al., 1988, 1989; Hwang et al., 1989; Wagner et al., 1991). It has also been hypothesized that the FAD mutation may lead to abnormal APP translation as a result of a disturbance in the secondary structure of APP mRNA (Tanzi and Hyman, 1991; D. Goldgaber, personal communication, 1991). Which, if any, of these models accounts for the pathogenesis of APP-717 mutant FAD remains a mystery. Another FAD pedigree has been discovered and has proven to be substantially more informative in elucidating the cell biological conse-

ALZHEIMER AP-AMYLOIU PRECURSOR PROTEIN

39

quences of the pathogenic mutation. In a large Swedish kindred, tandem missense mutations occur at the amino terminus of the AP-amyloid domain (Mullan et al., 1992a). Transfection of cultured cells with APP molecules containing the “Swedish” missense mutations results in the production of six- to eightfold excess soluble A@-amyloidabove that generated from wild-type APP (Citron et al., 1992; Cai et al., 1993). This is the first (and, to date, only) example of Alzheimer disease apparently caused by excessive AP-amyloid production. Based on the models of FAD-Dutch and FAD-Swedish, an important issue for clarification in sporadic Alzheimer disease will be to establish whether hyperaggregation or hyperproduction of AP-amyloid (or neither) is an important predisposing factor(s) to this much more commonly encountered clinical entity.

IX. Signal Transduction via Protein Phosphorylation Regulates the Relative Utilization of APP Processing Pathways

As noted, the protease that cleaves APP within the A@-amyloiddomain, as part of the nonamyloidogenic cleavage/release pathway (asecretase), and the proteases that cleave APP at other sites within the molecule to generate AP-am yloid (P-secretase and perhaps others) have not yet been identified. Nevertheless, some progress has been made toward understanding the regulation of APP cleavage. For example, the relative utilization of the various alternative APP processing pathways appears to be at least partially cell-type determined, with transfected AtT20 cells secreting virtually all APP molecules (Overly et al., 1991) whereas glia release little or none (Haass et al., 1991). In neuronal-like cells, the state of differentiation also plays a role in determining the relative utilization of the pathways (Baskin et al., 1992; Hung et al., 1992), with the differentiated neuronal phenotype being associated with relatively diminished basal utilization of the nonamyloidogenic asecretase cleavage/release pathway (Hung et al., 1992). Certain signal transduction systems that involve protein phosphorylation are potent regulators of APP cleavage, acting in some cases, perhaps, by altering the relative activity of nonamyloidogenic cleavage by asecretase. The role of protein kinase C (PKC) in this process has received the most attention. In many types of cultured cells, activation of PKC by phorbol esters dramatically stimulates APP proteolysis (Buxbaum et al., 1990) and cleavage/release (Caporaso et al., 1992b; Gillespie et al., 1992; Sinha and Lieberburg, 1992) via the a-secretase pathway. PKC-

40

SAM GANDY AND PAUL GREENGARD

stimulated a-secretory cleavage of APP may also be induced by the application of neurotransmitters and other first messenger compounds whose receptors are linked to PKC (Nitsch et al., 1992; Buxbaum et al., 1992; Lahiri et al., 1992). Okadaic acid, an inhibitor of protein phosphatases 1 and 2A (Cohen et al., 1990), also increases APP proteolysis and release via thea-secretase pathway (Buxbaum etal., 1990; Caporasoetal., 1992b). Thus, either stimulation of PKC or inhibition of protein phosphatases 1 and 2A is sufficient to produce a dramatic acceleration of nonamyloidogenic APP degradation. Furthermore, this PKC-activated processing can be demonstrated to occur at the expense of amyloidogenic APP degradation, resulting in diminished generation of AP-amyloid ( S . Sinha, 1992; Buxbaum et al., 1993; Hung et al., 1993). These results suggest that defects in signal-dependent regulation of APP cleavage may contribute to the pathogenesis of AD, a possibility supported by evidence that deficits in cholinergic neurotransmission (Davies and Maloney, 19’76) and in protein kinase C activity (Cole et al., 1988; Van Huynh et al., 1989; Masliah et al., 1991) accompany AD. By extension, then, the possibility exists that pharmacological modulation of APP metabolism via signal transduction might be therapeutically beneficial in individuals with AD (Candy et al., 1991, 199237; Gandy and Greengard, 1992). Complicating these notions, however, is the observation that PKC is also a potent regulator of APP expression (Goldgaber et al., 1989), although these pleiotropic effects of PKC may be dissociable at the level of the PKC isoenzyme involved (Hata et al., 1993). In addition to the attention to regulation of the nonamyloidogenic a-secretase pathway as a source of candidate etiologic defects and therapeutic opportunities, it may also be fruitful to study the potentially amyloidogenic psecretase pathway in an analogous fashion. Further work will be required to elucidate the importance of signal transduction systems as important candidate defects or therapeutic targets in AD. The enormous pharmacological experience with compounds that affect signal transduction makes such an approach particularly attractive for targeting therapy. The probable causal relationship between aberrant protein phosphorylation and neurofibrillary tangle formation (another component of Alzheimer structural pathology) adds to the attractiveness of protein phosphorylation pathways as potential therapeutic targets in AD. The mechanism by which stimulation or inhibition of intracellular protein phosphorylation regulates the processing of APP (including evaluation of the effect of changing the phosphorylation state of APP per se) remains to be fully elucidated. Protein kinase C rapidly phosphorylates a seryl residue in the cytoplasmic domain of APP (Fig. I), using either a synthetic peptide (Gandy et al., 1988; Suzuki et al., 1992)or APP holopro-

ALZHEIMER AD-AMYLOID PRECURSOR PROTEIN

41

tein (Suzuki et al., 1992) as substrate. Moreover, APP species are phosphorylated on this and other seryl and threonyl residues in intact cells and in brain (Suzuki et al., 1994; M. Oishi, T. Suzuki, A. Czernik, A. C. Nairn, and P. Greengard, personal communication). Characterization of the various APP residues phosphorylated in intact cells is under way to determine which sites of phosphorylation are utilized and to determine the possible existence of novel APP phosphorylation sites and APP kinases (Knops et d.,1993; Hung and Selkoe, 1994; Suzuki et d.,1994; Oishi et al., personal communication). Once the sites for APP phosphorylation in intact cells are established, analysis of the processing of phosphorylation-site mutant APP molecules can be used to elucidate the role of direct phosphorylation of APP. This approach has already been applied to certain cytoplasmic phosphorylation sites in APP (da Cruz e Silva et al., 1993; Hung and Selkoe, 1994; Vitek et al., personal communication). These experiments have demonstrated that changes in the phosphorylation state of the APP cytoplasmic domain are not necessary for the phenomenon of phosphorylation-regulated a-secretory cleavage of APP to occur. These observations have led to the proposal that proteins of the processing/cleavage/release pathway may be phosphoprotein mediators of “regulated-” o r “activatedprocessing” (da Cruz e Silva et al., 1993; Hung and Selkoe, 1994). Activation of proteolysis by phosphorylation has been demonstrated for a number of integral membrane proteins, including the polyimmunoglobulin receptor (pIgR; Casanova et al., 1990), the transforming growth factor-a (TGF-a) precursor (Pandiella and Massague, 199l ) , and the receptor for colony-stimulating factor- 1 (CSFl R; Downing et al., 1989). Direct phosphorylation of pIgR appears to be crucial to the activation of its trafficking and processing; phosphorylation of the TGF-a precursor has not been demonstrated; CSFlR is known to be a phosphoprotein, but the relationship between its phosphorylation and its proteolysis is not yet established. In general terms, the possible mechanisms for activated processing of integral molecules can be conceptualized as involving either activation or redistribution of either the substrate (i.e., APP) or the enzyme (i.e., secretase). Based on the APP cytoplasmic tail mutational analyses described above (da Cruz e Silva et al., 1993; Hung and Selkoe, 1994), the “substrate activation” model (Gandy et al., 1988, 1991, 1992a,b) is inadequate to explain activated processing of APP. Furthermore, in recent immunofluorescent studies of APP in cultured cells that were incubated in the absence or presence of PKC-activating phorbol esters (Caporaso et al., 1994),no obvious phorbol-dependent redistribution of APP immunoreactivity was apparent at steady state. A more detailed

42

SAM GANDY AND PAUL GREENGARD

analysis of APP distribution following PKC activation is underway, as suggested by the model of Luini and De Matteis (1993). Along a related line of investigation, Bosenberg and colleagues (1993) have succeeded in demonstrating apparently faithful activated processing of TGF-a precursor in porated cells in the virtual absence of cytosol, and in the presence of N-ethylmaleimide or 2.5 M NaCI. The preservation of activated processing under such conditions suggests that extensive vesicular trafficking is probably not required for activated processing of TGF-a and is consistent with a model of enzyme activation by direct phosphorylation. Studies are under way to determine whether activated APP processing has similar features.

X. Beyond Ap-Amyloid: Other Molecular Factors in Amyloidogenesis and Factors Differentiating Aging-Related Cerebral Amyloidosis from Alsheimer Disease

Since APP can be metabolized along several nonamyloidogenic or potentially amyloidogenic pathways, AD might be a clinicopathological phenotype that is due to a metabolic imbalance of the relative utilization of a nonamyloidogenic pathway(s) versus a potentially amyloidogenic pathway(s). To examine a possible correlation between APP metabolic pathway utilization and AD, some investigators have sought to identify AD-related changes in APP metabolism. Diminished levels of the large amino-terminal fragment of APP have been reported in the cerebrospinal fluid from patients with AD and from patients with the cerebrovascular AP-amyloidosis HCHWAD or FAD-Dutch (van Nostrand et al., 1992a,b). According to these reports, decreased levels of the released APP aminoterminal fragment were characteristic of the CSF from AD and FADDutch patients, but not that from age-matched controls, although there was some overlap between AD patients and patients with non-Alzheimertype dementia. To date, however, AD-specific changes in the levels of potentially amyloidogenic carboxyl-terminal fragments have not been observed in AD cortex (Nordstedt et al., 1991; Estus et al., 1992). Further, as noted in a preceding section, the metabolism of some mutant APP molecules and their carboxyl-terminal fragments in transfected cells appears to proceed in standard fashion (Cai et al., 1993; Felsenstein and Lewis-Higgins, 1993) (including apparently "normal" secretory cleavage), unperturbed by the presence of either the APP7"-"' FAD mutation o r the APP693-"1"FAD-Dutch mutation (numbering according to APP,7, isoform).

ALZHEIMER A/3-AMYLOID PRECURSOR PROTEIN

43

CSF levels of soluble AP-amyloid in normal aging and AD have been investigated to determine whether a correlation exists between CSF soluble AP-amyloid levels and the predisposition to AD. An initial study failed to detect an obvious relationship (Shoji et al., 1992), and that observation has been recently confirmed (Wisniewski et al., 1993). Thus, there appear to be other important factors-perhaps downstream events in the metabolism of APP fragments or soluble AP-amyloid-that play key roles in AP-fibrillogenesis. In support of this latter possibility is the evidence that an important effect of the FAD-Dutch mutation is to accelerate A@-amyloidfibril formation (Wisniewski et al., 1991). Other factors contributing to A@-amyloiddeposition and fibril formation may include the processing of soluble AP-amyloid into an aggregated form (Burdick et al., 1992; Dyrks et al., 1992) and/or the association of APamyloid with other molecules, such as a,-ACT (Abraham et al., 1988), heparan sulfate proteoglycan (Snow et al., 1992), apolipoprotein E (Wisniewski and Frangione, 1992; Strittmatter et al., 1993), and P component (Wisniewski and Frangione. 1992). In addition, deposited AP-amyloid plaques may serve as nucleation foci and act to recruit additional APamyloid deposition (Maggio et al., 1992). Events beyond AP-amyloid deposition may also be crucial in determining the eventual toxicity of AP-amyloid plaques. Although aggregation of A@-amyloidis important for in vitro models of neurotoxicity (Mattson and Rydel, 1992; Pike et al., 1993), the relevance of these phenomena for the pathogenesis of AD is unclear, since AP-amyloid deposits may occur in normal aging, in the absence of any evident proximate neuronal injury (Crystal et al., 1988; Masliah et al., 1990; Berg et al., 1993; Delaere et al., 1993). This suggests that other events must distinguish “simple“ cerebral amyloidosis from “full-blown” AD. One intriguing possible contributing factor is the association of complement components with AD-amyloid (Rogers et al., 1992). In cerebellum, where AP-amyloid deposits appear to cause no injury, plaques are apparently free of associated complement, whereas in the forebrain, complement associates with plaques, perhaps becoming activated and injuring the surrounding cells (Lue and Rogers, 1992). Other, yet undiscovered, plaque-associated molecules may also play important roles. It is also possible that Alzheimer neuropathology may be a final end product that can develop through a host of independent initiating molecular abnormalities, analogous to the manner in which disorders of either oxygen radical metabolism (Rosen et al., 1993) or cytoskeletal protein expression (Brady, 1993; Cote et al., 1993; Xu et al., 1993) can lead to a clinicopathological picture of motor neuron disease. Similarly, in the

44

SAM GANDY AND PAUL GREENGARD

case of Alzheimer disease, it is unknown whether, for example, in some situations, cytoskeletal phosphorylation abnormalities could be initiating events, and AP-amyloid deposits could occur secondarily. In support of this possibility is the recent demonstration that toxin- or lesion-induced nerve terminal degeneration can be associated with altered, potentially amyloidogenic APP metabolism (Iverfeldt et al., 1993). Further, A@amyloid deposition may “decorate” the periphery of amyloid plaques primarily composed of prion protein (Ikeda et al., 1992). The most promising leads for furthering our understanding of the molecular pathology of AD beyond APP metabolism lie in elucidating the role of apolipoprotein E allelic variation in determining predisposition to SAD (Saunders et al., 1993) and in the eventual discovery of the gene that causes the most common form of FAD, a form caused by a gene that resides on chromosome 14 (Schellenberg et al., 1992; St. GeorgeHyslop et al., 1992; Van Broeckhoven et al., 1992; Mullan et al., 199213). T h e identity of this gene is entirely unknown: it may represent a molecule that regulates APP expression or degradation, analogous to the lysozyme protease enzyme defect that was recently discovered to underlie hereditary systemic amyloidosis (Pepys et al., 1993). Alternatively, the chromosome 14 mutant molecule may implicate neurofibrillary components or may point in an entirely unexpected direction. In any event, discovery of the chromosome 14 FAD gene may prove to be an important step toward the eventual unravelling of the molecular basis of typical, common SAD, and it is this information that offers the most promise for ultimately providing us with a full understanding of Alzheimer disease and enabling its rational treatment. Acknowledgments

This work was supported by USPHS grants AG-11508 (to S.G.), and AG-09464 and AG-10491 (to P.G.). S.G. is the recipient of a Cornell Scholar Award in the Biomedical Sciences. The authors thank Drs. S. Sisodia and D. Selkoe for their critical review of the manuscript.

References

Abraham, C. R., Selkoe, D. J., and Potter, H. (1988). Cell (Cambridge, Mass.) 52,487-501. Arispe, N., Rojas, E., and Pollard, H. B. (1993). Proc. Null. Acad. Scz. U.S.A. 90, 567-571.

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

45

Bargmann, C., Hung, M., and Weinberg, R. (1986). Cell (Cambridge, Mass.) 45, 649-657. Baskin, F., Rosenberg, R., and Davis, R. M. (1992).J . Neuroscz. Rex 32, 274-279. Berg, L., McKeel, D. W., Miller, J . P., Baty, J . , and Morris, J. C. (1993). Arch. Neurol. 50, 349-358. Bonifacino, J., Cosson, P., and Klausner, R. (1990). Cell (Cambridge, Mass.) 63, 503-513. Bosenberg, M. W., Pandiella, A., and Massague, J. (1993).J. Cell B i d . 122, 95-102. Brady, S. T . (1993). Cell (Cambrzdge, Mass.) 73, 1-3. Burdick, D., Soreghan, B., Kwon, M., Kosmoski, J., Knauer, M., Henschen, A,, Yates, J., Cotman, C., and Glabe, C. (1992).J . Biol. Chem. 267, 546-554. Busciglio, J., Gabuzda, D. H., Matsudaira, P., and Yankner, B. (1993). Proc. Natl. Acad. Sci. U.S.A. 90, 2092-2096. Bush, A. l., Martins, R. N., Rumble, B., Moir, R.,Fuller, S., Milward, E., Currie, J., Ames, D., Weidemann, A., Fischer, P., Multhaup, G., Beyreuther, K., and Masters. C. L. (1990).J. Biol. Chem. 265, 15977-15983. Bush, A. l., White, S., Thomas, L. D., Williamson, T. G., Van Tiggelen, C. J., Currie, J., Small, D. H., Moir, R. D., Li, Q.-X., Rumble, B., Monning, U., Beyreuther, K., and Masters, C. (1992). Ann. Neurol. 32, 57-65. Buxbaum, J . D., Gandy, S. E., Cicchetti, P., Ehrlich, M. E., Czernik, A. J., Fracasso, R. P., Ramabhadran, T. V., Unterbeck, A . J., and Greengard, P. (1990). Proc. Natl. Acad. Sci. U.S.A. 87, 6003-6006. Buxbaum, J. D., Oishi, M., Chen, H. I . , Pinkas-Kramarski, R., Jaffe, E. A , , Candy, S. E., and Greengard, P. (1992). Proc. Natl, Acad. Sci. U.S.A. 89, 10075-10078. Buxbaum, J. D., Koo, E. H., and Greengard, P. (1993). Proc. Natl. Acad. Sci. U.S.A. 90, 9195-9198. Cai, X.-D., Golde, T. E., and Younkin, S. G. (1993). Science 259, 514-516. Cairns, N. J., Chadwick, A., Lantos, P. L., Levy, R., and Rossor, M . N . (1993). Nrurosci. Lett. 149, 137-140. Caporaso, G. L., Gandy, S. E., Buxbaum, J. D., and Greengard, P. (1992a). Proc. Natl. Acad. S C ~ U.S.A. . 89, 2252-2256. Caporaso, G. L., Candy, S. E., Buxbaum, J. D., Ramabhadran, T. V., and Greengard, P. (1992b). Proc. Natl. Acad. Sci. U.S.A. 89, 3055-3059. Caporaso, G., Takei, K., Candy, S.,Matteoli, M., Mundigl, 0.. Greengard, P., and de Camilli, P. (1994).J . Neurosci. 14, 3 122-3 138. Casanova, J . E., Breitfeld, P. P., Ross, S. A , , and Mostov, K. E. (1990). Science 248,742-745. Chartier-Harlin. M.-C., Crawford, F., Houlden, H., Warren, A., Hughes, D., Fidani, L., Goate, A,, Rossor, M., Roques, P., Hardy, J., and Mullan, M. (1991). Nature (London) 353, 844-846. Chen, W.-J., Goldstein, J. L., and Brown, M. S. (1990).J. Biol. Chem. 265, 31 16-3123. Cheung, T . T., Ghiso, J., Shoji, M., Cai, X.-D., Golde, T., Candy, S., Frangione, B., and Younkin, S. (1994). Amyloid, in press. Citron, M., Oltersdorf, T., Haass, C., McConlogue, L., Hung, A. Y., Seubert, P., VigoPelfrey, C., Lieberburg, I . , and Selkoe, D. J . (1992). Nature (London) 360,672-674. Cohen, P., Holmes, C. F. B., and Tsukitani, Y. (1990). Trends Biochem. Sci. 15, 98-102. Cole, G., Dobkins, K. R., Hansen, L. A., Terry, R. D., and Saitoh, T. (1988). Brain Res. 452, 165-174. Cole, G . M., Huynh, T . V., and Saitoh, T. (1989). Neurochem. Res. 14, 933-939. Cote, F., Collard, J.-F., and Julien, J.-P. (1993). Cell (Cambridge, Mass.) 73, 35-46. Crystal, H., Dickson, D., Fuld, P., Masur, D., Scott, R., Mehler, M., Masdeu, J., Kawas, C., Aronson, M., and Wolfson, L. (1988). Neurology 38, 1682-1687.

46

SAM CANDY AND PAUL GREENGARD

d a Cruz e Silva, O., Iverfeldt, K., Oltersdorf, T., Sinha, S., Lieberburg, I., Ramabhadran, T . , Suzuki, T., Sisodia, S., Gandy, S., and Greengard, P. (1993).Neuroscience 57,873-877. Davies, P., and Maloney, A. J. F. (1976). Lancet 2, 1403. Delaere, P., He, Y., Fayet, G., Duyckaerts, C . , and Hauw, J. J . (1993). Neurobzol. Agzng 14, 191-194. Downing, J . R., Roussel, M. F., and Sherr, C . J. (1989). Mol. Cell. B i d . 9, 2890-2896. Dyrks, T . , Dyrks, E., Hartmann, T., Mast.ers, C., and Beyreuther, K. (1992).J. Biol. Chem. 267, 18210- I82 17. Ehlers, M. R. W., and Riordan, J . F. (1991). Biochemktry 30, 10065-10074. Esch. F. S., Keini, P. S., Beattie, E. C., Blacher, R. W., Culwell, A. R., Oltersdorf, T., McClure, D., and Ward, P. J. (1990). Science 248, 1122-1124. Estus, S., Golde, T., Kunishita, T., Blades, D., Lowery, D., Eisen, M., Usiak, M., Qu, X., Tabira, T., Greenberg, B., and Younkin, S. (1992). Science 255, 726-728. Felsenstein, K., and Lewis-Higgins, L. (1993). Neurosci. Lett. 152, 185-189. Felsenstein, K. M., Hunihan, L. W., and Roberts, S. 8.(1994).Nature Genel. 6, 251-256. Gandy, S., a n d Greengard, P. (1992). Trends Pharmacol. Sci. 13, 108-113. Candy, S., Czernik, A. J.. and Greengard. P. (1988). Proc. Natl. Acad. Scz. U.S.A. 85, 6218-6221. Gandy, S. E., Buxbaum, J. D., and Greengard, P. (1991). In “Alzheimer’s Disease, Basic Mechanisms, Diagnosis and Therapeutic Strategies” (K. Iqbal, D. R. Crapper McLachIan, B. Winblad, and H. M. Wisniewski, eds.), pp. 155-172. Wiley, New York. Candy, S. E., Bhasin, R., Ramabhadran. T. V., Koo, E. L., Price, D. L., Goldgaber, D., and Greengard, P. (1992a).J . Neurochem. 58, 383-386. Gandy, S. E., Buxbaum, J. D., and Greengard, P. (3992b). In “Alzheimer’s Disease, New Treatment Strategies” (Z. S. Khachaturian a n d ] . P. Blass, eds.), pp. 175-192. Dekker, New York. Ghetti, B., Murrell,J., Benson, M. D., and Farlow, M. R. (1992).Neuroscz. Lett. 571,133-139. Gillespie, S. L., Golde, T. E., and Younkin, S. G. (1992). Biochem. Biophys. Res. Commun. 187, 1285-1290. Glenner, G. G., and Wong, C . W. (I984a). Bzochem. Biophys. Res. Commun. 120, 885-890. Glenner, G. G., and Wong, C. W. (1984b). Bzochem. Biophys. Res. Commun. 122, 1131-1 135. Goate, A., Chartier-Harlin, M.-C., Mullan, M., Brown, J., Crawford, F., Fidani, L., Giuffra, L., Haynes, A., Irving, N., James, L., Mant, R., Newton, P., Rooke, K., Roques, P., Talbot, C.. Pericak-Vance, M., Roses, A., Williamson, R., Rossor, M., Owen, M., and Hardy, J. (1991). Nature (London) 349, 704-706. Golde, T. E., Estus, S., Younkin, L. H., Selkoe, D. J., and Younkin, S. G. (1992). Science 255,728-730. Goldgaber, D., Lerman, M. I., McBride, 0. W., Saffiotti, U., and Gajdusek, D. C. (1987). Science 235, 877-880. Goldgaber, D., Harris, H. W., Hla, T., Maciag, T., Donnelly, R. J., Jacobsen, J . S., Vitek, M . P., and Gajdusek, D. C. (1989). Proc. Natl. Acad. Sci. U.S.A. 86, 7606-7610. Haass, C., Hung, A. Y., and Selkoe, D. J. (1991).J. Neurosci. 11, 3783-3793. Haass, C., Koo, E. H., Mellon, A., Hung, A. Y., and Selkoe, D. J . (1992a). Nature (London) 357, 500-503. Haass, C., Schlossmacher, M. G., Hung, A. Y., Vigo-Pelfrey, C., Mellon, A., Ostaszewski, B. L., Lieberburg, I., Koo, E. H., Schenk, D., Teplow, D. B., and Selkoe, D. J. (1992b). Nature (Londox) 359, 322-325. Haass, C., Hung, A. Y., Schlossmacher, M. G., Teplow. D. B., and Selkoe, D. J . (1993). J . B i d . Chem. 268, 3021-3024. Hata, A., Akita, Y., Suzuki, K., and Ohno, S. (1993).J. Biol. Chem. 268, 9122-9129.

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

47

Hung, A. Y., and Selkoe, D. J. (1994). EMBOJ. 13, 534-542. Hung, A. Y., Koo, E. H., Haass, C., and Selkoe, D. J. (1992). Proc. N a l Acad. Scz. U.S.A. 89,9439-9443. Hung, A. Y., Haass, C . , Nitsch, R. M., Qiu, W. Q., Citron, M., Wurtman, R. J., Growdon, J. H., and Selkoe, D. J. (1993).,/. Biol. Chem. 268, 22959-22962. Husby, G., Araki, S., Benditt, E. P., Glenner, G. G., Natvig, J. B., and Westermark. P. (1993). Bull. W H O 71, 105-108. Hwang, T.-C., Lu, L., Gruenet, D. C., Huganir, R., and Guggino, W. B. (1989). Science 244, 1351-1353. Ikeda, S., Yanagisawa, N., Glenner. G., and Allsop, D. (1992).Neurodegeneratzon 1,28 1-288. Iverfeldt, K., Walaas, S. I., and Greengard, P. (1993). Proc. Nail. Acad. Sci. U.S.A. 90, 4 146-4 150. Kang, J . , Lemaire, H.-G., Unterbeck, A., Salbaum, J. M., Masters, C. L., Grzeschik, K.-H., Multhaup, G., Beyreuther, K., and Muller-Hill, B. (1987).Nature (London)325, 733-736. Kennedy, A. M., Newman, S., McCaddon, A,, Ball, J., Roques, P., Mullan, M., Hardy, J., Chartier-Harlin, M.-C., Frackowiak. R. S. .J,, Warrington, E. K., and Rossor, M . N. (1993). Brain 116, 309-324. Kitaguchi, N., Takahashi, Y., Tokushima, Y., Shiqjiri, S., and Ito, H. (1988).Naturr (London) 331,530-532. Knops, J., Lieberburg, I., and Sinha, S. (1992).J . Biol. Chem. 267, 16022-16024. Knops, J., Candy, S., Greengard, P., Lieberburg, I., and Sinha, S. (1993). Biochrm. Biophys. Res. Commun. 197, 380-385. Konig, G., Monning, U., Czeck, C.. Prior, R., Baniti, R., Schreiter-Gasser, U., Bauer, J., Masters, C. L., and Beyreuther, K. (1992).J. Biol. Chem. 267, 10804-10809. Koo. E., Sisodia, S., Cork, L., Unterbeck, A., Bayney, R., and Price, D. (1990). Neuron 2, 97-104. Lahiri, D. K., Nall, C., and Farlow, M. ( 1 992). Biochem. Inl. 28, 853-860. Lantos, P. L., Luthert, P. J., Hanger, D.. Anderton, B. H., Mullan, M., and Rossor, M. (1992). Neurosci. Lett. 137, 221-224. Levy, E., Carmen, M. D., Fernantlez-Madrid, 1 . J., Power, M. D., Lieberburg, I., van Duinen, S. G., Bots, G. Th. A. M., Luyendijk, W., and Frangione, B. (1990). Science 248, 1 124- 1 126. Li, M., McCann, J., Liedtke, C., Nairn, A , , Greengard, P., and Welsh, M. (1988). Nature (London) 331, 358-360. Li, M., McCann, J. D., Anderson, M. P., Clancy, J. P., Liedtke, C. M., Nairn, A. C., Greengard, P., and Welsh, M. j.(1989). Science 244, 1353-1356. Lue, L.-F., and Rogers, J. (1992). Dementia 3, 308-313. Luini, A., and De Matteis, M. A. (1993). Trends Crll Biol. 3, 290-292. Maggio, J. E., Stimson, E. R., Ghilartl, J . R., Allen, C . J., Dahl, C . E., Whitcomb, L). C . , Vigna, S. R., Vinters, H. V., Labenski, M. E., and Mantyh, P. W. (1992). Proc. Natl. Acad. Sci. U.S.A. 89, 5462-5466. Mann, D. M. A., Jones. D., Snowden, j.S., Neary. D., and Hardy, J . (1992). Neurosrl. Left. 137, 225-228. Maruyama, K., Kametani, F., Usanii. M., Yamao-Harigaya, W.. and Tanaka, K. (1991). Bzochem. Biophys. Res. Commun. 179, 1670-1676. Masliah, E., Terry, R. D., Mallory, M., Alford, M., and Hansen, L. A. (1990). A m . ] . Pathol. 137, 1293-1297. Masliah, E., Cole, G. M., Hansen, L. A,, Mallory, M., Allbright, T., Terry, R. D., and Saitoh, T . (1991)..]. Neurosci. 11, 2759-2767.

48

SAM GANDY AND PAUL CREENGARD

Masters, C. L., Simms, G., Weinman, N. A,, Multhaup, G., McDonald, B. L., and Beyreuther, K. (1985). Proc. Natl. Acad. Sci. U.S.A. 82, 4245-4249. Mattson, M. P., and Rydel, R. (1992). Neurobiol. Agmg 13, 617-621. Mattson, M. P., Cheng, B., Culwell, A. R., Esch, F. S., Lieberburg, I., and Rydel, R. E. (1993). Neuron 10, 243-254. Miyazaki, K., Hasegawa, M., Funahashi, K., and Umeda, M. (1993). Nature (London) 362, 839-841. Mullan, M., Crawford, F., Axelman, K., Houlden, H., Lilius, L., Winblad, B., and Lannfelt, L. (1992a). Nat. Genet. 1, 345-347. Mullan, M., Houlden, H., Windelspecht, M., Fidani, L., Lombardi, C., Diaz, P., Rossor, M., Crook, R., Hardy, J., Duff, K., and Crawford, F. (1992b). Nut. Genet. 2, 340-342. Murrell, J., Farlow, M., Ghetti, B., and Benson, M. D. (1991). Science 254,97-99. Naruse, S., Igarashi, S., Kobayashi, H., Aoki, K., Inuzuka, T, Kaneko, K., Shimizu, T., Iihara, K., Kojima, T., Miyatake, T., and Tsuji, S. (1991). Lancet 337, 978979. Nishimoto, I., Okamoto, T., Matsuura, Y., Takahashi, S., Okamoto, T., Murayama, Y., and Ogata, E. (1993). Nature (London) 362, 75-79. Nitsch, R. M., Slack, B. E., Wurtnian, R. J., and Growdon, J. H. (1992). Sczence 258, 304-307. Nordstedt, C., Gandy, S. E., Alafuzoff, I., Caporaso, G., Iverfeldt, K., Grebb, J. A., Winblad, B., and Greengard, P. (1991). Yroc. Natl. Acad. Sci. U.S.A. 88, 8910-8914. Nordstedt, C., Caporaso, G. L., Thyberg. J., Gandy, S. E., and Greengard, P. (1993). J. Biol. Chem. 268, 608-612. Oltersdorf, T.,Fritz, L. C., Schenk, D. B., Lieberburg, I., Johnson-Wood, K. L., Beattie, E. C., Ward, P. J., Blacher, R. W., Dovey, H. F., and Sinha, S. (1989). Nature (London) 341, 144-147. Oltersdorf, T., Ward, P. J.. Henriksson, T., Beattie, E. C., Neve, R., Lieberburg, I., and Fritz, L. C. (1990).J . Biol. Chem. 265, 4492-4497. Overly, C. C., Fritz, L. C., Lieberburg, I., and McConlogue, L. (1991). Biochem. Biophys. Res. Commun. 181,513-519. Palmert, M. R., Berman Podlisney, M., Witker, D. S., Oltersdorf, T., Younkin, L. H., Selkoe, D. J., and Younkin, S. G. (1989). P70C. Null. Acad. Sci. U.S.A. 86, 6338-6342. Pandiella, A., and MassaguC, J. (1991).Proc. Natl. Acad. Sci. U.S.A. 88, 1726-1730. Pandiella, A., Bosenberg, M. W.. Huang, E. J., Besmer, P., and Massague, J. (1992). J. Biol. Chem. 267,24028-24033. Pepys, M. B., Hawkins, P. N., Booth, D. R., Vigushin, D. M., Tennent, G. A., Soutar, A. K., Totty, N., Nguyen, O., Blake, C. C. F., Terry, C. J., Feest, T. G., Zalin, A. M., and Hsuan, J. J. (1993). Nature (London) 362, 553-557. Pike, C. J., Burdick, D., Walencewicz, A. J., Glabe, C. G., and Cotman, C. W. (1993). J . Neurosci. 13, 1676-1687. Ponte, P., Gonzalez-DeWhitt, P., Schilling, J., Miller, J., Hsu, D., Greenherg, B., Davis, K., Wallace, W., Lieberburg, I., Fuller, F., and Cordell, B. (1988). Nature (London) 331, 525-527. Robakis, N. K., Ramakrishna, N., Wolfe, G., and Wisniewski, H. M. (1987). Proc. Natl. Acad. Sci. U.S.A. 84, 4 190-4 194. Rogers, J., Cooper, N. R., Webster, S., Schultz, J., McGeer, P. L., Styren, S. D., Civin, W. H., Brachova, L., Bradt, B., Ward, P., and Lieherburg, 1. (1992). Proc. Natl. Acad. Sn. U.S.A. 89, 10016-10020. Rosen, D. R., Siddique, T., Patterson, D., Figlewicz, D. A., Sapp, P., Hentati, A., Donaldson, D., Goto, J., O’Regan, J. P., Deng, H.-X., Rahmani, Z., Krizus, A,. McKenna-Yasek,

ALZHEIMER AP-AMYLOID PRECURSOR PROTEIN

49

D., Cayabyab, A., Gaston, S. M., Berger, R., Tanzi, R. E., Halperin, J . J., Herzfeldt, B., Van den Bergh, R., Hung, W.-Y., Bird, T., Deng, G., Mulder, D. W., Smyth, C., Laing, N. G., Soriano, E., Pericak-Vance, M. A,, Haines, J., Rouleau, G. A , , Gusella, J. S., Horvitz, H. R., and Brown, R. H. (1993). Nature (London) 362, 59-62. Rossor, M. ( 1 992). I n “Bailliere’s Clinical Neurology: Unusual Dementias” (M. Rossor, ed.), pp. 5 17-534. Bailliere Tindall, Philadelphia, Pennsylvania. Sahasrabudhe, S. R., Spruyt, M. A,, Muenkel, H. A., Blume, A. J., Vitek, M. P., and Jacobsen, J. S. (1992)./. B i d . Chem. 267, 25602-25608. Saitoh, T., Sundsmo, M., Roch, J.-M., Kimura, N., Cole, G., Schubert, D., Oltersdorf, T., and Schenk, D. B. (1989). Cell (Cambridge, Mass.) 58, 615-622. Saunders, A. M., Strittrnatter, W. J . , Schrnechel, D., St. George-Hyslop, P. H., PericakVance, M. A., Joos, S. H., Rosi, B. L., Gusella, J. F., Crapper-MacLachlan, D. R., Alberts, M. J., Hulette, C., Crain, B., Goldgaber, D., and Roses, A. D. (1993). Neurology 43, 1467-1472. Schellenberg, G. D., Bird, T. D., Wijsrnan, E. M., Orr, H. T., Anderson, L., Nemens, E., White, J. A., Bonnycastle, L., Weber, J. L., Alonso, M. E., Potter, H., Heston, L. L., and Martin, G . M. (1992). Science 258, 668-671. Schoumacher, R. A., Shoemaker, R. L., Halm, D. R., Tallant, E. A,, Wallace, R. W., and Frizzell, R. A. (1987). Nature (London) 330, 752-754. Schubert, D., LaCorbiere, M., Saitoh, T., and Cole, G. (1989). Proc. Natl. Acad. Sci. U.S.A. 86,2066-2069. Seubert, P., Vigo-Pelfrey, C . , Esch, F., Lee, M., Dovey, H., Davis, D., Sinha, S., Schlossmacher, M., Whaley, J., Swindlehurst, C., McCormack, Wolfert, R., Selkoe, D.. Lieberburg, I., and Schenk, D. (1992). Nutwe (London) 359, 325-327. Seubert, P., Oltersdorf, T., Lee, M. G., Barbour, R., Blomquist, C., Davis, D. L., Bryant, K., Fritz, L. C., Galasko, D., Thal, L. J., Iieberburg, I . , and Schenk, D. B. (1993). Nature (London) 361, 260-263. Shioi, J.. Anderson, J. P., Ripelliiio, J. A., and Robakis, N. K. (1992).J . Biol. Chern. 267, 13819-13822. Shoji, M., Golde, T. E., Ghiso, J., Cheung, T. T., Estus, S., Shaffer, L. M., Cai, X.-D., McKay, D. M., Tintner, R., Fragione, B., and Younkin, S. G. (1992). Science 258, 126-129. Sinha, S., and Lieberburg, I. (19!32).Neurodegenerution 1, 169-175. Sisodia, S. S. (1992). Proc. Natl. Atad. Sci. U.S.A. 89, 6075-6079. Sisodia, S. S., Koo, E. H., Beyreuther. K.. Unterbeck, A., and Price, D. L. (1990). Science 248,492-495. Slunt, H. H., Thinakaran, G., Von Koch, C . , Lo, A . C. Y., Tanzi, R. E., and Sisodia, S. S. (1994).J . Biol. Chem. 269, 2637-2644. Smith, R. P., Higuchi, D. A., and Broze, G. J., J r . (1990). Science 248, 1126-1128. Snow, A. D., Sekiguchi, R., Nochlin, D., Kirnata. K., Schreier, W. A., and Morgan, D. G. (1992). Sac. Neurosci. Abstr. 18, 1465. St. George-Hyslop, P., Haines, J., Rogaev, E., Mortilla, M., Vaula, G., Pericak-Vance, M., Foncin, J.-F., Montesi, M., Bruni, A., Sorbi, S., Rainero, I . , Pinessi, L., Pollen, D., Polinsky, R., Nee, L., Kennedy. J., Macciardi, F., Rogaeva, E., Liang, Y., Alexandrova, N . , Lukiw, W., Schlumpf, K., Tanzi. R., Tsuda, T., Farrer, L., Cantu, J.-M., Duara, R., Amaducci, L., Bergamini, L.,Gusella, I., Roses, A., and Crapper-McLachlan. D. (1992). Nut. Genet. 2, 330-334. Strittmatter, W. J., Saunders, A. M., Schniechel, D., Pericak-Vance, M., Enghild, J., Salvesen, G. S., and Roses, A. D. (1993). Proc. Natl. Acad. Sci. U.S.A. 90, 19771981.

50

SAM GANDY A N D PAUL GREENGARD

Suzuki, T., Nairn, A. C., Candy, S. E., and Greengard, P. (19Y2).Neuroscience (Oxford)48, 755-76 1. Suzuki, T., Oishi, M., Marshak, D. R., Czernik, A. J., Nairn, A. C., and Greengard, P. (1994). E M B O J . 13, 1114-1122. 1-amaoka, A., Kalaria, R. N . , Lieberburg, I., and Selkoe, D. (1992). Proc. Natl. Acad. Sci. U.S.A. 89, 1345-1349. Tanzi, R., and Hyman, B. (1991). Nature (London) 350, 564. Tanzi, R. E., Gusella, J. F., Watkins, P. C . , Bruns, C . A. P., St. George-Hyslop, P., Van Keuren, M. L., Patterson, D., Pagan, S., Kurnit, D. M., and Neve, R. L. (1987). Science 235, 880-884. Tanzi, R. E., McClatchey, A. I., Lamperti, E. D., Villa-Kornaroff, L., Gusella, J. F., and Neve, R. L. (1988). Nature (London) 331, 528-530. Tomlinson, B. E., and Corsellis, J. A. N. (1984). In “Greenfield’s Neuropathology” Fourth edition. (J. H. Adams, J. A. N . Corsellis, and L. W. Duchen, eds.), 4th Ed. pp. 95- 1025. New York. Van Broeckhoven. C., Haan, I . , Bakker, E., Hardy, J. A,, Van Hul, W., Wehnert, A., Vegter-Van der Vlis, M., and Roos, R. A. C. (1990). Science 248, 1120-1 122. Van Broeckhoven, C., Backhovens, H., Cruts, M., De Winter, G., Bruyland, M., Cras. P., and Martin, J.-J. (1992). Nut. Genet. 2, 335-339. van Duinen, S. G., Castaiio, E. M., Prelli, F.. Bots, G. T h . A. M., Luyendijk, W., and Frangione, B. (1987). Proc. Natl. Acad. Sci. U.S.A. 84, 5991-5994. Van Huynh, T., Cole, G., Katzman, R., Huang, K.-P., and Saitoh, T. (1989). Arch. Neurol. 46, 1195-1 199. Van Nostrand, W. E., and Cunningham, D. D. (1987).J. Biol. Chem. 262, 8508-8514. Van Nostrand, W. E., Wagner, S. L., Suzuki, M., Choi, B. H., Farrow, J. S., Geddes, J. W., Cotman, C. W., and Cunningham, D. D. (1989). Nature (London)341, 546-549. Van Nostrand, W., Wagner, S.. Shankle, W. R., Farrow, J. S., Dick, M., Rozemuller, J. M., Kuiper, M. A., Wolters, E. C., Zimmerman, J., Cotman, C. W., and Cunningham, D. D. (1992a). Proc. Natl. Acad. Sci. U.S.A. 89, 2551-2555. Van Nostrand, W. E., Wagner, S. L., Haan, J . , Bakker, E., and Roos, R. A. (1992b).Ann. Neurol. 32, 215-218. Wagner, J., Cozens, A,, Schulman, H., Gruenert, D., Stryer, L., and Gardner, P. (1991). Nature (London) 349, 793-796. Wasco, W., Bupp, K., Magendantz, M., Gusella,,j. F., Tanzi, R. E., and Solomon, F. (1992). Proc. Natl. Acad. Scz. U.S.A. 89, 10758-10762. Wasco, W., Gurubhagavatula, S., Paradis, M. D., Romano, D. M., Sisodia, S. S., Hyrnan, R. T., Neve, R. L., and Tanzi, R. E. (1993). Proc. Natl. Acad. Scz. U.S.A. 5, 95-100. Weidemann, A., Konig, G., Bunke, D., Fischer, P., Salbaum, J. M., Masters, C. L., and Beyreuther, K. (1989). Cell (Cambridge, Mass.)57, 115-126. Wisniewski, T., and Frangione, B. (1992). Neurosci. Lett. 135, 235-238. Wisniewski, T., Ghiso, J., and Frangione, B. (1991). Biochem. Biophys. Res. Commun. 179, 1247-1254. Wisniewski, T., Wegiel, J., Wisniewski, H. M., and Frangione, B. (1993). Neurology 43, A422. Wolf, D.. Quon, D., Wang, Y., and Cordell, B. (1990). EMBO J. 9, 2079-2084. Xu, Z., Cork, L. C., Griffin, J. W., and Cleveland, D. W. (1993). Cell (Cambrzdge, Mass.) 73, 23-33.

MOLECULAR NEUROBIOLOGY OF THE GABAA RECEPTOR

Susan M. J. Dunn, Alan N. Bateson, and Ian L. Martin Department of Pharmacology, University of Alberta, Edmonton, Alberta, Canada T6G 2H7

I. Introduction 11. Pharmacology of the GABA,, Receptor A. T h e GABA Site B. T h e Benzodiazepine Site C. T h e Barbiturate Site D. T h e Steroid Site 111. Biochemistry A. Receptor Isolation B. Ligand Binding Sites C. Protein Modification D. Immunological Characterization IV. Molecular Cloning of Receptor Subunits A. Initial Isolation of Receptor Subunit cDNAs B. lsolation of a cDNA Encoding a y2Subunit C. Receptor Heterogeneity Revealed hy Multiple Subunits D. Further Heterogeneity Arises from Alternate Splicing V. Characterization of the Receptor Family A. Heterologous Expression Reveals Different Functional Attributes of GABA,A Receptor Subunits B. Gene Expression C. Immunocytochernical Localization D. Assembly of Subunits VI. T h e Future References

I. Introduction

The first indication that y-aminobutyric acid (GABA) played an important role in the mammalian central nervous system dates back to the late 1940s when high concentrations of this amino acid were identified in mouse brain (Roberts and Frankel, 1949, 1950). However, the initial evidence that GABA functioned as a neurotransmitter, and was not merely an essential amino acid, came from studies with invertebrates. Application of GABA to isolated muscle fibers from lobster and crayfish was found to produce an attenuation of the stretch-receptor discharge (Bazemore et al., 1957). Later, stimulation of the inhibitory nerves inINTERNATIONAL REVIEW OF NEUROBIOLOGY, VOL. 36

51

Copyright D 1994 by Academic Press. Inc. All rights of reproduction in any form reserved.

52

SUSAN M. J . DUNN et a1

nervating lobster skeletal muscle was found to result in the release of GABA into the extracellular space, whereas similar stimulation of the excitatory neurons in this preparation did not (Kravitz et al., 1962). In mammals, GABA is essentially localized to the central nervous system where it exhibits a differential topographical distribution (lversen and Bloom, 1972; Ottersen and Storm-Mathison, 1984). It is located in nerve terminals (Neal and Iversen, 1969) from which it can be released by depolarizing stimulii in a calcium-dependent manner (Bradford, 1970). Subsequent to release it is removed from the synaptic cleft by high affinity uptake systems (Iversen and Neal, 1968). These transporters are present in both neurons and glia, although the substrate specificity in the two cell types appears to be different (Iversen and Kelly, 1975). In the mammalian CNS, iontophoretic application of GABA usually results in an inhibitory hyperpolarizing response (Krnjevic and Schwartz, 1967; Obata et al., 1970), which is blocked by the alkaloid bicuculline (Curtis et al., 1970). The hyperpolarizing response is due to an increase in the chloride conductance of neuronal membranes (Curtis et al., 1968; Kelly et al., 1969) resulting in the passage of chloride ions into the cell. However, depolarizing responses have also been observed in superior cervical ganglion cells (Adams and Brown, 1975; Gallagher et al., 1978) and synaptic terminals (Schmidt, 1971). The receptor that responds to GABA application with an increase in chloride conductance of the neuronal membrane and is blocked by bicuculline is now termed the GABAA receptor. Until the late 1970s this was the only recognized response to the application of GABA. Subsequently bicuculline-insensitive GABA responses that could be mimicked by baclofen (P-p-chlorophenyl-GABA) were found, suggesting the existence of a novel type of GABA receptor (Bowery et al., 1980),now commonly known as the GABABreceptor. This receptor appears to be G-protein coupled giving rise to both decreases in calcium conductance and increases in potassium conductance ( Wojcik et al., 1989). Although the GABAA receptor of the invertebrate exhibits many properties in common with those found in vertebrates, only the GABAA receptor of the vertebrate CNS will be discussed in the remainder of this review.

II. Pharmacology of the GABAA Receptor

The GABAA receptor i s responsible for the majority of neuronal inhibition in the vertebrate CNS. It is widely distributed in the CNS and

GABA, RECEPTOR

53

it has received considerable attention as the site of action of a number of important centrally acting drugs, probably the most visible of which are the benzodiazepines. Recent electron microscope images of the receptor, purified from pig brain, show that it is structurally similar to the nicotinic acetylcholine receptor, It comprises five distinct protein subunits arranged in a pentameric array around a central pit that presumably forms the ion channel (Nayeem et al., 1994). The precise identity of the five proteins that form this structure are unknown for any GABA, receptor in vivo. Molecular cloning studies have now revealed the existence of a large family of gene products that may assemble to produce functional GABA, receptors (Section IV). Hence the GABA, receptor in the mammalian brain is not a single entity, but a family of closely related receptor oligomers with properties that are similar but distinct. Much of our understanding of the pharmacology of the GABAA receptor has been obtained from the study of neurones of the mammalian CNS. This will be reviewed briefly here to provide a backgound for the detailed discussion that will appear later in this review; for convenience we will address the subject in terms of the recognition sites on the receptor complex that have been the most extensively studied.

A. THEGABA SITE T h e ubiquitous distribution of the GABA, receptor in the mammalian CNS has been revealed by the use of [3H]GABAradioligand binding techniques (Enna and Snyder, 1975), whereas autoradiographic studies have demonstrated their distinct topographical localization (Zukin et al., 1974; Palacios etal., 1981). Subsequent use of immunocytochemical methods has allowed rather more detailed anatomical information to be obtained (Richards et al., 1987). I n the cerebellum, where the greatest detail is available, all five major cell types exhibit varying degrees of receptor immunoreactivity, with the granule cells showing the highest density. Interestingly, there is evidence that the receptor is not only expressed in regions of the cell that receive GABAergic input, but also, for example, on Golgi cell somata where no synaptic contacts are found. With the two monoclonal antibodies used in this study, no immunoreactivity was detected on axons, nerve terminals, or glial cells (Somogyi et al., 1989). The distribution of the GABAA receptor at both synaptic and nonsynaptic sites is quite distinct from that of the closely associated glycine receptor, which appears to be mainly associated with synapses (Triller et al., 1985). Initial electrophysiological studies of the GABAA receptor demonstrated that its activation by GABA resulted in an increased chloride conductance of the supporting neuronal membrane (Curtis et al., 1968;

54

SUSAN M. J. DLJNN

el

al.

Kelly et al., 1969). The agonist response exhibits positive cooperativity, consistent with the presence of two agonist binding sites on the receptor complex (Dichter, 1980; Sakmann et al., 1983; Bormann and Clapham, 1985; Akaike et al., 1985). The agonist induced current decreases on continued exposure to high agonist concentrations as a result of receptor desensitization (Adams and Brown, 1975; Akaike et al., 1987; Mathers, 1987). It is now clear that the agonist-induced conductance increase is due to the opening of the integral ion channel of the GABA, receptor. This channel is anion selective, potassium being about 20 times less permeant than chloride, and studies of anion specificity suggest that the minimum open pore size of the channel is about 0.56 nm (Bormann et al., 1987). T h e channel properties were first investigated using fluctuation noise analysis, which indicated that the channel had a conductance o f 16 pS and an average open time of about 24.5 ms (Barker et al., 1982); the mean open time, but not the channel conductance, was dependent on the activating agonist (Barker and Mathers, 1981). However, the introduction of single-channel current recording techniques (Hamill et al., 1981) has shown that the gating characteristics of the channel are much more complex. The GABA-activated channel of mouse spinal cord neurons in primary culture exhibits four conductance states of 44, 30, 19, and 12 pS, although their relative frequencies of occurrence vary; the 30-pS channel is most commonly observed, accounting for 83% of the openings, whereas the lowest and highest conductance states occur very infrequently (Bormann et al., 1987). Channel activity is a stochastic process that has been subjected to detailed statistical analysis. Single-channel currents, recorded from mouse spinal cord neurons in culture, show that the main conductance state of the GABA, receptor opens either singly or in bursts of several openings. Increasing the concentration of the agonist, GABA, results in an increase in the frequency with which the channels open and an increase in the average time for which the channels remain open, together with an increased frequency of occurrence of bursts of channel openings and a lengthening of the durations of the bursts of channel openings. It has no effect on the single-channel conductance (Macdonald et al., 1989a; Twyman et al., 1990). Although the detailed electrophysiological studies discussed above were carried out with the natural agonist GABA, the properties of the recognition site and the agonist recognition process itself have been investigated in some detail using a number of GABA analogues. Studies with [3H]GABAshow a heterogeneity in binding at equilibrium and this observation is common to all the other agonist ligands used (Wang et al., 1979; O l s e n e t a l . , 1981; Browner etal., 1981; Olsen and Snowman, 1982,

GABA, RECEPTOR

55

1983; Jordan et al., 1982; Falch and Krogsgaard-Larsen, 1982; Burch et al., 1983). T h e details of these interactions are discussed in Section III,B, 1. The recognition properties of the binding site for GABA have been explored by the systematic modification of the structure of this amino acid (see Krogsgaard-Larsen et al., 1984, 1986) and certain general criteria have been established; GABA agonists are zwitterionic with a charge separation of about 0.48 nm. A number of GABA isosteres, in which the carboxyl moiety has been conformationally restricted, such as muscimol, are more potent than the natural agonist (Krogsgaard-Larsen et al., 1979). Substitution of the amino terminal invariably leads to a decrease in the potency of the compound; N-methyl-GABA is substantially less potent than the parent amino acid and the N,N-dimethyl derivative is inactive (Krogsgaard-Larsen and Johnston, 1978). However, the conformation of the basic nitrogen may be restricted, as in, for example, imidazole-4-acetic acid, without significant loss of potency (Johnston et al., 1979). Rationalizations for these structural characteristics continue to appear in the literature (Lipkowitz et al., 1989). Bicuculline acts as a competitive antagonist at the GABAA receptor (Curtis et al., 1970; Simmonds, 1978, 1980; Barker et al., 1983), as do its methochloride and methiodide quaternary salts, which are more stable in aqueous solution (Pong and Graham, 1972; Johnston et al., 1972). The stereochemistry of the active enantiomer is known to be erythro(1S,9R) (Hill et al., 1974; Collins and Hill, 1974), and structure-activity studies have been carried out with a series of 45 bicuculline-related phthalideisoquinolines (Kardos et al., 1984). [3H]Bicuculline has been shown to bind to brain membranes in a saturable manner, and the interaction appears to take place with a homogeneous population of sites (Mohler and Okada, 1977a, 1978a). The topographical distribution of these binding sites, as determined by autoradiography, is similar, but not identical, to that found for [3H]muscimol (Olsen et al., 1990). T h e phenylaminopyridazine SR9553 I also exhibits competitive inhibition of GABAA receptor-mediated responses and is about 20 times more potent than bicuculline (Heaulme et al., 1987; Mienville and Vicini, 1987).Two additional compounds display competitive antagonism at the GABA, receptor, pitrazepin (Gahwiler et al., 1984; Kemp et al., 1985) and R5135 (Hunt and Clements-Jewery, 198 I ) , although both lack specificity for this receptor. SITE B. THEBENZODIAZEPINE T h e benzodiazepines were introduced into clinical practice in 1960. They remain the most frequently prescribed psychoactive drugs and are

56

SUSAN M. J. DUNN el a1

used largely for the treatment of anxiety and insomnia although they are also potent anticonvulsants and muscle relaxants (see for example Martin, 1990).Their mechanism of action remained unknown until some 15 years after their introduction. The first clues were obtained when intravenous diazepam was found to cause an increase in presynaptic inhibition in the spinal cord (Schmidt et al., 1967), although it was not until 1972 that this inhibition was satisfactorily shown to be mediated by GABA (Barker and Nicoll, 1972; Davidoff, 1972). Pharmacological studies demonstrated that the action of the benzodiazepines was to facilitate, in some ill-defined manner, GABA, receptor-mediated transmission (Polc et al., 1974). Subsequent studies have shown that the phenomenon is general, the benzodiazepines produce a facilitation of GABA, receptor-mediated transmission at both pre- and post-synaptic sites in the mammalian CNS (Haefely and Polc, 1983). Chlordiazepoxide was shown to produce a leftward shift of the GABA dose-response curve in chick spinal cord and sensory neurons with no increase in the maximum response (Choi et al., 1981). Subsequent fluctuation analysis indicated that this was due to changes in the gating characteristics of the GABA-activated channel; the benzodiazepines appeared to increase the frequency of channel opening with little effect on the channel open time or channel conductance (Study and Barker, 1981).These studies have now been extended by single-channel analysis, which indicates that the reason for the observed increase in channel opening frequency is due not to single-channel events but to an increased occurrence of bursting activity, although the average duration of the bursts did not increase in the presence of the benzodiazepines (Twyman et al., 1989). In 1977 saturable, specific, high-affinity binding sites for ['Hldiazepam were found differentially distributed in the mammalian brain with identical properties in all brain regions studied. (Mohler and Okada, 197713; Squires and Braestrup, 1977). The observations that the affinity of other benzodiazepines for this binding site was highly correlated with their efficiency in a variety of pharmacological tests suggested that this binding site was the pharmacological receptor through which these drugs produced their overt effects (Mohler and Okada, 1977b; Braestrup et at., 1977). The binding sites are restricted largely to the mammalian CNS where the highest densities, of around 1 pmollmg protein, can be found in cortical areas (Squires and Braestrup, 1977; Mohler and Okada, 1978b). Outside the CNS, benzodiazepine recognition sites, identical to those in the brain, have been found in chromaffin granule cells (Kataoka et a/., 1984) and the pituitary (Brown and Martin, 1984). A binding site with similar, but quite distinct, recognition properties is found in the

CABA, RECEPTOR

57

periphery (Braestrup and Squires, 1977), where it is associated with the mitochondria1 outer membrane (Anholt et al., 1986) and is thought to be involved in steroidogenesis (Baulieu, 1991). T h e distribution of the benzodiazepine binding sites in the mammalian CNS has been mapped in considerable detail (Richards et al., 1983) and, although they exhibit a similar distribution to the GABA, receptor binding sites, the match is not precise, suggesting that perhaps not all GABA, receptors have an associated benzodiazepine recognition site. However, observations that the affinity of [3H]diazepam is increased in the presence of high concentrations of GABA, an effect that is blocked by the competitive GABA,\ receptor antagonist bicuculline, provided strong evidence that there is a close association between the two recognition sites. Subsequent studies have shown that the purification of the benzodiazepine recognition site results in the copurification of the GABA, receptor, indicating that the two recognition sites are located on a common protein structure, a hypothesis that has received conclusive support from recent molecular biological studies (see below). It is now clear that several structurally disparate compounds recognize the benzodiazepine site with high affinity. T h e first of these compounds was the triazolopyridazine CL2 18872. This compound displaced diazepam o r flunitrazepam binding from the cerebellum with an apparently higher affinity than in the hippocampus (Squires et al., 1979); it was argued that these t w o brain regions contained subtly different subtypes of benzodiazepine binding sites that could be distinguished by CL2 18872, but not by diazepam or flunitrazepam. T h e site in the cerebellum with a higher affinity for CL218872 was termed the Bzl site and that with a lower affinity and present in the hippocampus, the Bz2 site. Later studies showed ethyl P-carboline-3-carboxylateto exhibit the same properties (Nielsen and Braestrup, 1980; see Martin et al., 1983); however, there was little evidence then of the remarkable complexity that this receptor would reveal later. T h e classical benzodiazepines possess a clearly defined profile of action as anticonvulsants, sedative/hypnotics, anxiolytics, and muscle relaxants. However, studies with ethyl @-carboline-3-carboxylaterevealed that the compound was a proconvulsant, having properties diametrically opposed to those of the classical benzodiazepines (Tenen and Hirsch, 1980; Cowan et al., 198 1);compounds of this type were termed inverse agonists. Subsequently compounds have been found that exhibit no overt behavioral actions but are able to block the effects of both agonists and inverse agonists (Nutt et al., 1982; Polc et al., 1982). T h e benzodiazepine recognition site is thus a modulatory site on the GABA, receptor. The inverse agonists exhibit precisely the characteristics expected: they shift the

58

SUSAN M. J. DUNN et a1

GABA dose-response curve to the right whereas the agonist benzodiazepines shift it to the left (Kemp et al., 1987). Thus the agonists display positive efficacy at the benzodiazepine recognition site of the GABA, receptor whereas the inverse agonists exhibit negative efficacy. The inverse agonists appear to produce effects on channel gating opposite to those of the agonist benzodiazepines; they decrease the apparent channel opening frequency, and produce no change in the channel open time or channel conductance (Barker et al., 1984). Both partial agonists and partial inverse agonists exist (Chan and Farb, 1985). There is evidence that the partial agonists may not produce the physical dependence apparent with the full agonists (Moreau et al., 1990), and thus be suitable targets for the development of more acceptable anxiolytics (Haefely et al., 1990). C. THEBARBITURATE SITE Anticonvulsant and anesthetic barbiturates potentiate the electrophysiological response to GABA but their mechanism of action is distinct from that of the benzodiazepines. Phenobarbital shifts the GABA dose-response curve to the left but, at higher concentrations, also increases the maximum response (Gallagher et al., 1981). High concentrations of the barbiturates are able to directly activate GABAA receptors (Nicoll et al., 1975; Simmonds, 1981; Higashi and Nishi, 1982), but at lower concentrations they appear to facilitate GABA-mediated transmission by increasing the channel open time, while having no effect on channel conductance or opening frequency (Study and Barker, 1981). Studies of channel gating characteristics have revealed that the barbiturates increase the channel burst duration but have no effect on bursting frequency (Twyman et al., 1989; Macdonald et al., 1989b). Unlike the benzodiazepines, no specific high-affinity binding sites for the barbiturates have been identified. However, in vitro the barbiturates enhance GABA, and benzodiazepine receptor agonist binding (Asano and Ogasawara, 1981; Olsen and Snowman, 1982),presumably by some allosteric interaction within the receptor complex. Although studies of the channel gating characteristics suggest that the convulsant picrotoxin acts in a manner reciprocal to that found for the barbiturates, i.e., it decreases channel burst duration (Twyman et al., 1989), it would appear that the barbiturates interact with a different site than picrotoxin. T h e evidence for this is somewhat indirect: picrotoxin competitively displaces the cage convulsant ligand tert-butylbicyclophosphorothionate (TBPS) from its specific high-affinity binding sites on brain membranes in a competitive fashion (Squires et al., 1983), whereas the interaction of the barbiturates with the TBPS binding sites is complex (Trifiletti et al., 1984).

GABA, RECEPTOR

59

D. THESTEROID SITE It has been known for some time that steroid derivatives are effective and potent anesthetics (Selye, 1942; Figdor et al., 1957). Many of the effects of steroids result from their intracellular actions on the regulation of gene expression, but their ability to induce rapidly anesthesia suggested alternative mechanisms of action. Both electrophysiological and radioligand binding experiments suggest that the action of the steroid derivative alphaxalone is the result of potentiation of GABA, receptormediated transmission (Scholfield, 1980; Harrison and Simmonds, 1984). Subsequent studies with dissociated adrenal chromaffin cells, in culture, have shown that concentrations of alphaxalone in excess of 30 nM dosedependently potentiate GABA,-mediated responses and, further, that the response is additive with that of phenobarbitone, suggesting that the effects are not mediated via the same mechanism (Cottrell et al., 1987). Indeed, studies at the single-channel level seem to suggest that the effects of both androsterone and progesterone on channel gating in mouse spinal cord neurons in primary cell culture are distinct from those of the barbiturates; the steroids appear to increase both the channel opening frequency and the channel open lifetime (Twyman and Macdonald, 1992). I n addition to the major recognition sites described above, the GABAA receptor also appears to carry binding sites for a number of other modulators including Zn2+,avermectin, and, possibly, alcohol (reviewed by Sieghart, 1992). T h e GABAA receptor of the mammalian CNS is thus, functionally, a complex protein displaying a variety of allosteric interactions that may lead to increases or decreases in GABA-mediated transmission. Biochemical studies of the GABA, receptor in its natural membrane environment and after its purification have proved invaluable in contributing to our present understanding of the structure and function of this receptor.

111. Biochemistry

A. RECEPTOR ISOLATION 1. Purification T h e density of the GABA, receptor in mammalian brain is around 1 pmol/mg protein and initial biochemical characterization of the receptor indicated that it had a molecular weight of approximately 250,000 (see below). Assuming that there is one binding site per receptor oligo-

60

SUSAN M. J. D U N N et al.

mer, this suggested that a purification factor of about 4000 would be required to isolate the protein in a pure state. At that time, the common belief was that all GABAA receptors carried a benzodiazepine binding site and this site was, therefore, selected as the basis for purification of the receptor by affinity chromatography. In these purification attempts, the ability of the isolated receptor preparations to bind radiolabeled drugs from each of the pharmacological classes described above, in addition to retention of the expected allosteric interactions between different binding sites, has been widely used to assess purification success. In this respect, the GABAergic agonist [3H]muscimol, the benzodiazepine [3H]flunitrazepam, and t.he channel blocker [35]TBPShave been particularly important in following the fate of isolated GABAA receptors. GABA, receptors have been solubilized from brain membrane preparations using a variety of nondenaturing detergents. As discussed in detail previously (Stephenson and Barnard, 1986), the use of sodium deoxycholate in the presence of a cocktail of protease inhibitors gives the greatest efficiency of solubilization as assessed by the number of high-affinity binding sites for [3H]muscimol and [3H]flunitrazepam in the extract. However, in sodium deoxycholate, not only is the receptor unstable but binding sites for picrotoxinin and related channel blockers are lost and barbiturates lose their ability to potentiate benzodiazepine binding. These receptor characteristics were, however, shown to be preserved when solubilization was carried out using the zwitterionic detergent CHAPS (Sigel and Barnard, 1984) or the nonionic detergent @octylglucoside (Hammond and Martin, 1986; Bristow and Martin, 1987) in the presence of protease inhibitors and exogenous lipid. Following solubilization, several different immobilized benzodiazepines have been used to purify GABAA receptors by affinity chromatography. After absorption of GABAAreceptors to the affinity resin, specific elution has usually been achieved using the water-soluble benzodiazepines chlorazepate or flurazepam (reviewed in Stephenson and Barnard, 1986). Using these procedures, purification factors of 2000-5000 may be achieved, although the yield of benzodiazepine binding sites is low with only 2-5% of the activity in the original membranes being recovered. In most reports, the purified preparations have been demonstrated to carry a single population of binding sites for [3H]flunitrazepam (Kdvalues of 4-28 nM) and a single high-affinity binding site for [3H]muscimol, although lower affinity muscimol binding sites have also been observed (see below). The ratio of muscimol : flunitrazepam binding sites have been reported to range from 0.35 to 3.8 (reviewed by Stephenson, 1988). Several allosteric interactions have also been demonstrated to be preserved in purified receptors including the stimulation of benzodiazepine binding by GABA (Schoch and Mohler, 1983; Sigel and Barnard, 1984)

GABA,& RECEPTOR

61

and barbiturates (Sigel and Barnard, 1984; Olsen et al., 1984). Binding sites for [35S]TBPShave also been measured in the purified protein (Sigel and Barnard, 1984; Trifiletti et al., 1984; Stephenson et al., 1986; Olsen et al., 1984) but these rapidly inactivate on storage (Sigel and Barnard, 1984; Olsen et al., 1984). It has been shown that retention of the ability of isolated receptors to bind [35S]TBPSis critically dependent on the presence of exogenous phospholipids during purification (Stephenson et al., 1986). This lipid dependence is consistent with the results of a study in which it was shown that treatment of rat brain membranes with phospholipase A, decreased the binding of [35S]TBPSand also inhibited the barbiturate stimulation of benzodiazepine binding (Havoundjian et al., 1986). In a detailed study of binding site stability, Bristow and Martin ( 1987) have shown that binding sites in CHAPS-solubilized preparations can be protected from inactivation by inclusion of a natural brain lipid extract supplemented with cholesterol hemisuccinate. Examination of purified GABA, receptors by polyacrylamide gel electrophoresis in the presence of SDS revealed a multisubunit structure. In most cases two major subunits, a (53 kDa) and p (57 kDa), were observed (reviewed by Stephenson, 1988; Sieghart, 1991). Based on an overall complex molecular weight of 220,000-240,000 determined by radiation inactivation (Chang and Barnard, 1982) or by gel filtration chromatography (Sigel et al., 1983; Martini et al., 1982), it was proposed that the purified GABA, receptor complex was a heterotetramer with a stoichiometry of a@, (Casalotti et al., 1986; Mamalaki et al., 1987). Molecular cloning has since revealed the presence of many more subunits and this has prompted more careful scrutiny of gel patterns, resulting in the observation of microheterogeneity within the bands and additional bands being observed by protein staining, photoaffinity labeling, and immunological techniques (see below). Early attempts to obtain amino acid sequences of isolated, intact subunits failed to provide sequence information, presumably due to their blocked N-termini. However, Schofield et al. (1987) obtained pxtial sequences of proteolytic fragments and used these to design oligonucleotide probes that were used in the initial cloning of an a and a p subunit from bovine brain. When coexpressed in X e n o p w oocytes, these subunits were shown to form GABA-gated chloride channels with many (although not all) of the pharmacological properties of the native receptor (Section IV,A).

2. Reconstitution Although many groups have reported the successful solubilization and purification of GABA, receptors, it was only recently that such preparations were shown to form functional GABA-gated chloride chan-

62

SUSAN M . J. DUNN et al.

nels after reconstitution into lipid vesicles. In early studies, solubilized (Hammond and Martin, 1986)and affinity-purified (Schoch et al., 1984; Sigel et al., 1985)receptors were reconstituted in liposomes and, although these preparations retained the ability to bind various radiolabeled ligands, no chloride flux responses were reported. In the first report of successful functional reconstitution, Hirouchi et ul. (1987) demonstrated GABA-stimulated %I- influx in vesicles reconstituted with solubilized and/or purified GABA, receptors and this response was blocked by bicuculline and stimulated by flunitrazepam. Bristow and Martin (1990) reconstituted purified GABAA receptors using a mixture of natural brain lipids and cholesterol hemisuccinate and showed that [3H]flunitrazepam binding to these preparations was potentiated by GABA, pentobarbital, and the pyrazolopyridine cartazolate. GABA-mediated ( 100 p M ) 36Clflux responses were also potentiated by flunitrazepam, pentobarbital, and cartazolate. Using an alternative fluorescence technique to measure chloride flux, GABA, receptors were solubilized in 0-octylglucoside and reconstituted using a mixture of asolectin and native brain membranes (Dunn et al., 1989a,b). By following changes in the fluorescence of a chloride-sensitive dye, 6-methoxy-N-(3-sulfopropyl)quinolinium (MSQ), trapped within the vesicles, muscimol was shown to stimulate chloride influx in a dose-dependent manner with half-maximal response occurring at 0.3 p M (Dunn et al., 1989b). This flux was inhibited by prior desensitization of the receptor and by bicuculline and picrotoxin but was stimulated by diazepam (Dunn et al., 1989a). These preparations did not, however, display any sensitivity to barbiturates perhaps due to the lability of their binding sites (Section III,A, 1). Similar techniques have since been extended to studies of the affinity-purified protein(s) that displays similar functional properties (Thuynsma and Dunn, 1991). Although the receptor preparations that have been used in reconstitution procedures are likely to contain more than a single receptor subtype, within the limitations of the techniques used, no functional heterogeneity has been detected. It is likely that identification of functional heterogeneity will require more sophisticated analysis, e.g., biophysical analysis of single channels after reconstitution of purified protein in lipid bilayers. In this respect, the ability to purify native receptor subtypes by, for example, immunoaffinity chromatography (Section III,D) will be important. SITES B. LIGANDBINDING

1. Radiolabeled Ligund Binding Studies There have been many studies of the binding of radiolabeled agonists, benzodiazepines, and other channel modulators to GABAA receptors in

GABA, RECEPTOR

63

purified brain membranes and these have been extensively reviewed before (Olsen et al., 1981; Olsen and Venter, 1986; Stephenson, 1988). At equilibrium, the binding of GABAergic agonists is heterogeneous, with the binding of r3H]GABA being characterized by a high-affinity component ( K d values of 10-20 nM) with one or more low-affinity sites with dissociation constants in the range 100 nM to 1 p M (Fischer and Olsen, 1986). The antagonist bicuculline inhibits high-affinity GABA binding with an IC,, of about 6 p M , and it appears to inhibit more potently the lower affinity sites with an IC,o of 0.8 p M (Olsen and Snowman, 1983). Since micromolar concentrations of agonists are required to induce chloride flux and to modulate the binding of benzodiazepines (reviewed by Fischer and Olsen, 1986), the lower affinity sites have been correlated with physiological function. N o function has been ascribed to the high-affinity sites and it is usually assumed that these reflect an inactivated, desensitized state of the receptor. It has not, however, been unambiguously determined whether high- and low-affinity sites represent binding to distinct binding sites or to interconverting states of the receptor (see Agey and Dunn, 1989). In studies using affinitypurified receptor preparations, both high- and low-affinity sites have been shown to exist (Sigel and Barnard, 1984; Schoch et al., 1984) but, since GABA, receptors are now known to be heterogeneous, this may reflect the presence of multiple receptor subtypes rather than multiple sites on the one receptor. The latter possibility is, however, supported by the results of 36Cl- flux experiments using cell-free preparations from rat brain, from which it has been suggested that the GABA binding sites involved in channel activation are distinct from those that induce receptor desensitization (Cash and Subbarao, 1987). A similar model proposing multiple sites and parallel pathways for activation and inactivation has previously been proposed for the nicotinic acetylcholine receptor (Dunn and Raftery, 1982). In contrast to the heterogeneity of [3H]GABAor [3H]muscimol binding to brain membranes or purified GABA, receptors, the binding of a classical benzodiazepine agonist such as [3H]diazepam is to an apparently single population of receptors, although the number of receptors shows a brain regional distribution (reviewed by Martin, 1987). As described above, benzodiazepine receptors can be pharmacologically subdivided into two major groups, on the basis of their sensitivity to the triazolopyridazine CL2 18872 and P-carboline derivatives (Martin, 1987). As noted above, micromolar concentrations of GABA or muscimol potentiate the binding of agonist benzodiazepines, whereas they inhibit that of inverse agonists and have no effect on the affinity for benzodiazepine antagonists (reviewed by Martin, 1987). The manner by which the agonist benzodiazepines modulate the characteristics of GABA-activated channels (Section

64

SUSAN M. J. DUNN et al.

11,A) is compatible with the suggestion that these compounds increase the affinity of the receptor for GABAergic agonists. Although such modulation of low-affinity GABA binding has been reported (Skerritt et al., 1982), this effect has not been observed consistently and thus remains somewhat controversial (see Haefely, 1990). The binding of barbiturates has not been directly measured but, in micromolar concentrations, they interact allosterically to increase the number of binding sites for GABA and to increase the affinity for benzodiazepines, responses that are blocked by picrotoxin (reviewed by Fischer and Olsen, 1986). Both [3H]picrotoxinin and the cage convulsant [35S]TBPShave been used in direct binding studies; these ligands appear to compete for the same binding site. This site may, however, be distinct from the barbiturate site (see Fischer and Olsen, 1986). 2. Thermodynamics Thermodynamic analysis of drug-receptor interactions can provide valuable information about the driving forces involved in complex formation and can sometimes be used to distinguish the modes of binding of different types of drugs. In the case of the p-adrenergic receptor, for example, it has been shown that agonist binding, but not antagonist binding, is temperature-dependent and whereas antagonist binding is largely entropy-driven, the binding of agonists occurs with a decrease in enthalpy to overcome an infavorable change in entropy (Wieland et al., 1979). With respect to GABA, receptors, the major goal in the application of thermodynamic analysis has been to identify possible differences in the modes of binding of benzodiazepine agonists, antagonists, and inverse agonists. In early studies of the binding of [3H]diazepam to rat brain membranes it was shown that binding decreased dramatically with increasing temperature (Mohler and Okada, 1977b; Braestrup and Squires, 1977). Speth et al. (1979) showed that for [3H]flunitrazepam, the rate constant for dissociation increased more than the association rate with increasing temperature, resulting in overall decrease in affinity. The van t’Hoff plot deviated from the classical relationship and was biphasic with an inflection point at 16°C. Analysis of the temperature dependence of binding of the agonist [’H]clonazepam also gave a biphasic van t’Hoff plot with an inflection point at 21”C, whereas the van t’Hoff plot for [“H]Ro15-1788 binding was linear (Mohler and Richards, 1981), suggesting differences in the modes of binding of the two ligands. To explain these results, a multistep model of binding was proposed, in which it was suggested that benzodiazepine agonists and antagonists bind to the receptor with similar thermodynamics: above 2 1”C, agonists, but not

GAHA, RECEPTOR

65

antagonists, can induce a further conformational change of the receptor-ligand complex. This was an attractive suggestion and evidence for such an agonist-induced conformational change was soon obtained in kinetic studies of radiolabeled ligand binding (Quast and Mahlmann, 1982; Chiu et al., 1983). However, subsequent more detailed thermodynamic studies appeared to contradict this suggestion (Kochman and Hirsch, 1982; Quast et al.. 1982) and it has also been pointed out that such a sequential two-step model would not predict curvilinearity in van t'Hoff plots (Doble, 1983). In a parallel study of the thermodynamics of ['H]flunitrazepam and the convulsant ['HH]ethyl P-carboline3-carboxylate binding to rat brain membranes, Doble (1983) found that the temperature dependence of the affinity constants of both ligands deviated from the simple relationship suggesting in each case an increase in enthalpic drive over the temperature range studied. After the effects of a series of different benzodiazepines and P-carbolines were studied, it was concluded that there was no simple correlation between biological effects and temperature dependence of binding (Doble, 1983). In the early studies described above, the degree of potential heterogeneity in GABA, receptors was unknown and the curvilinearity of the van t'Hoff plots that has been observed is possibly explained by the presence of multiple receptor subtypes having different temperature sensitivities. T o address this issue, Maguire et al. (1992) have recently studied the thermodynamics of [SH]Ro15-1788 binding to rat cerebellar membranes, in which the receptor population is apparently more homogenous with the drug binding to B z l receptors only (Massotti et al., 1991). The affinity for [3H]Ro15-1788was again found to decrease with temperature but the van t'Hoff plots for a series of benzodiazepine agonists, antagonists, and P-carbolines were linear and, with the exception of methyl-6,7-dimethoxy-~-carboline-3-carboxylate (DMCM) whose binding was entropy driven, the binding of all other ligands was driven by enthalpy. There, thus, appears to be no relationship between the different types of ligands and their thermodynamic properties, leading to the conclusion that binding and receptor modulation are distinct steps (Maguire et al., 1992). I n contrast to earlier results, Prince and Simmonds (1992) recently found that [3H]flunitrazepam binding to crude rat brain membranes was characterized by a linear van t'Hoff plot. Since these preparations are likely to contain multiple receptor subtypes, the possibility that different receptor subtypes have different temperature sensitivities thus requires further examination. Although the binding of benzodiazepines to GABA, receptors cannot simply be differentiated by the effects of tem-

66

SUSAN M. J. DUNN et al.

perature, it is likely that thermodynamic analysis will continue to provide valuable information on GABAA receptor interactions. I n a recent study, for example, it was reported that the anesthetics alphaxalone and propofol increased the entropy, but not the enthalpy, of [3H]flunitrazepam binding to rat brain membranes. Pentobarbitone, however, increased the enthalpy of binding and thus stimulates binding more effectively at lower temperatures (Prince and Simmonds, 1992). 3. Photoaffinity Labeling [3H]Flunitrazepam, although not a conventional photoaffinity ligand in that it lacks a specific photolabile group capable of generating a reactive species (Knowles, 1971), has been shown to be a very useful tool in photoaffinity labeling studies of benzodiazepine binding sites (Sieghart and Karobath, 1980; Mohler et al., 1980). Photoaffinity labeling of crude brain homogenates followed by analysis by SDS-gel electrophoresis resulted in labeling of mainly a 51-kDa protein (Mohler et al., 1980). Specificity of the labeling for the GABA, receptor was shown by the enhancement of labeling by GABA, an effect that was inhibited by bicuculline (Sieghart and Karobath, 1980). Detailed studies of the labeling patterns in different areas of the brain and during development, in addition to analysis of labeling patterns after tryptic cleavage, provided some of the first direct indications of receptor heterogeneity (Sieghart and Drexler, 1983). In photolabeling of affinity-purified receptor, in which only two major components were present, it was shown that [3H]flunitrazepam was incorporated into the 52-kDa a-subunit (Sigel and Barnard, 1984). More recently, however, multiple a-subunits have been detected by SDS-gel electrophoresis and [3H]flunitrazepam has been shown to label three different proteins with apparent molecular masses of 51, 53, and 59 kDa (Fuchs et al., 1988), all of which were recognized by an a-subunitspecific monoclonal antibody (see below). Later studies using subunitspecific antibodies have shown that the 51-, 53-, and 59-kDa proteins correspond to the a l - ,a*-, and a3-subunits, respectively (Stephenson et al., 1989; Fuchs et al., 1990). Although [3H]flunitrazepam labeling thus occurs on the a-subunits, these subunits are not the sole determinants of binding since, in studies of receptor expression, it has been shown that the presence of a y-subunit is required for both benzodiazepine binding and benzodiazepine modulation of GABAA receptor responses (Pritchett et al., 1989a). T h e position of the label within the a-subunit has not been determined but Olsen et al. (1991) have obtained partial sequences of proteolytic fragments of the labeled receptor to show that labeling occurs within residues 8-297, and most likely between residues 106 and 297. Using an alternative approach, Stephenson and Duggan ( 1989) have probed cyanogen bromide fragments with sequence-specific

GABA,&RECEPTOR

67

antibodies and have concluded that the binding site lies within residues 59 and 148 of the a,-subunit. [3H]Muscimol has also been used as a photoaffinity label for the GABA, receptor agonist sites although this again is not a conventional photoaffinity ligand (Cavalla and Neff, 1985; Asano et al., 1983). T h e major site of labeling in purified preparations was demonstrated to be the @-subunit (Casalotti et ad., 1986; Deng et al., 1986). More recently, several subunits of apparent molecular mass 51, 52, and 56 kDa were shown to be labeled by [3H]muscimol in purified receptors from rat brain and these were recognized by a P-subunit-specific antibody (Fuchs and Sieghart, 1989). [3H]Muscimol labeling has also been reported to occur on the a-subunits (Bureau and Olsen, 1988, 1990), suggesting either that agonist binding sites may occur at the interface(s) between subunits, as has been suggested for nicotinic acetylcholine receptors (Blount and Merlie, 1989; Middleton and Cohen, 1991), or that both a - and @-subunitscarry binding sites for [3H]rnuscimol. In this respect, possibly all subunits may be capable of binding GABAergic agonists since single subunits can be expressed to produce homoligomeric GABA-gated chloride channels (Section V,A).

C. PROTEIN MODIFICATION

Although the photoaffinity labeling studies described above have led to the identification of subunits that carry, at least in part, binding sites for receptor agonists and benzodiazepines, little is known about the amino acids within the subunit sequences that are involved in binding. These amino acids are currently being probed by sequencing of proteolytic fragments of photolabeled preparations and by site-directed mutagenesis (see below). In the absence of precise sequence information, chemical modification techniques have been used to provide some valuable information on the importance of certain amino acid residues in ligand binding. It has been shown that modification of receptors by reaction with diethyl pyrocarbonate (DEP) inhibited both benzodiazepine (Burch and Ticku, 1981; Sherman-Gold and Dudai, 1981; Burch et al., 1983) and P-carboline (Maksay and Ticku, 1984) binding without affecting the binding of other ligands. Flurazepam (Maksay and Ticku, 1984), but not a /3-carboline (Lambolez and Rossier, 1987), was able to protect the binding sites from DEP inhibition, indicating that the modified residues are close to, or are allosterically coupled to, the benzodiazepine binding site(s). Although DEP can react with both histidine and tyrosine residues, the pH dependencies of both benzodiazepine binding (Lambolez and

68

SUSAN M. J . DUNN el ul.

Rossier, 1987; Maksay et al., 1991) and DEP inactivation (Maksay and Ticku, 1984; Lambolez et al., 1989) indicate that histidine residues are the major residues involved. In the early studies, both Bzl and Bz2 benzodiazepine receptors were shown to be sensitive to DEP (Maksay and Ticku, 1984). More recently, Binkley and Ticku (1991) have investigated the effects of DEP on diazepam-insensitive sites. These sites, which are enriched in the cerebellum and are now known to be associated with a4-and a6-subunits (Luddens et al., 1990; Wieland et al., 1992), can be investigated using the imidazobenzodiazepine Ro 15-4513, which binds to both diazepamsensitive and -insensitive sites (Sieghart et al., 1987). Treatment of cerebellar membranes by DEP decreased the number of sites for ['HI flunitrazepam and [3H]Ro15-1788 as expected, but had no effect on ['H]Ro15-4513 binding (Binkley and Ticku, 1991). Thus the binding of the latter ligand to both diazepam-sensitive and -insensitive sites does not appear to involve the histidine residue that is crucial for the binding of other benzodiazepines. Site-directed mutagensis has shown that the histidine residue that is essential for high-affinity agonist binding is residue 101 in the a,-subunit. In the a4-and a6-subunits (Wieland et al., 1992; Wisden et al., 1991; Korpi et al., 1993) there is an arginine residue at the equivalent position. Maksay (1992) has recently investigated the possible importance of this arginine in ['H]Ro15-45 13 binding by studying the effects of treating cerebellar membranes with the argininemodifying reagent 2,3-butanedione. Binding to the diazepam-insensitive sites was partially inhibited with an EC,, of 1.6 mM, compared to an EC,, of 3.5 mM for binding to the diazepam-sensitive sites. Another histidine residue that is conserved in diazepam-sensitive receptors is replaced by Y214 in as. The effects of tyrosine modification have, therefore, been investigated to assess the importance of this residue in benzodiazepine binding. Modification of cerebellar membranes with the tyrosine reagent tetranitromethane partially inactivated ['HIRo 154513 binding (Maksay, 1992) similar to the partial loss of ['Hlfluntrazepam binding that was observed in earlier studies (Maksay and Ticku, 1984). However, the receptor sites could not be protected from the effects of tetranitromethane by carrying out the modification in the presence of benzodiazepines, suggesting that the susceptible tyrosine is not located directly within the binding site (Maksay and Ticku, 1984). Cysteineicystine residues have also been shown to be important structural determinants for benzodiazepine binding. It has been shown that the treatment of brain membrane preparations with either the sulfhydry1 alkylating agent iodoacetamide or the disulfide reducing agent pmercaptoethanol results in reduced affinity for both ['Hldiazepam and [3H]muscimol (Marangos and Martino, 1981). Recently, de Bengtsson et

GABA, RECEPTOR

69

al. (1993) reexamined these effects and showed that in bovine brain membranes, iodoacetamide alone had little effect on [3H]flunitrazepam binding but the affinity was much reduced if membranes were reacted with excess iodoacetamide after reduction with dithiothreitol (DTT). In another study, it was shown that whereas [3H]flunitrazepam binding is relatively insensitive to DTT reduction, the binding of the partial inverse agonist [3H]Ro15-4513 is inhibited by DTT (IC50of 4.6 mM), this being a consequence of a reduction in both affinity and number of sites. The binding sites for this latter ligand were protected if the reduction was carried out in the presence of either Ro15-4513 or flunitrazepam, suggesting that the disulfide that is important for the binding of [3H]Ro1545 13 is either located near a benzodiazepine binding site(s) or is involved in stabilizing the conformation of the site (Duncalfe and Dunn, 1993). The binding of both [3H]fli~nitrazepamand [3H]Ro15-4513 was inhibited by the sulfhydryl alkylatirig agent N-ethylmaleimide, although apparently by different mechanisms (Duncalfe and Dunn, 1993). Taken together these results suggest an important role for disulfide bonds and sulfhydryl groups in benzodiazepine binding and different benzodiazepines display differences in their sensitivity to modification of these groups, suggesting differences in their modes of binding. Protein modification studies have also been used to study the involvement of particular amino acid residues in the binding of GABA and other receptor agonists. Arginine residues have been implicated in [3H]muscimol binding, since exposure of membrane preparations or purified GABA, receptors to the arginine modification reagents 2,3butanedione or phenylglyoxal inhibited [3H]muscimol binding by up to 82% (Widdows et al., 1987). This was due to a loss of sites rather than a reduction in affinity and the sites could be protected by carrying out the reaction in the presence of GABA. It has also been reported that diazotized sulfanilate, a reagent that is fairly nonselective for histidine and tyrosine residues, inhibits the binding of [3H]GABA to its lowaffinity, but not its high-affinity, binding sites (Burch et al., 1983). The ability of GABA to potentiate benzodiazepine binding was lost in parallel, suggesting that the low-affinity sites are important for this allosteric property.

D. IMMUNOLOGICAL CHARACTERIZATION I n view of the large number of GABA, receptor subunits identified by cloning techniques (Section V), the potential number of receptor subtypes is very large. At the present time, however, we do not know the actual subunit compliment of any native receptor. In order to probe

70

SUSAN M. J. DUNN et a1

native receptor compositions and the regional distribution of receptor subtypes in the brain, panels of polyclonal and monoclonal antibodies have been raised against isolated receptors (Stephenson et al., 1986), specific subunits (Haring et al., 1985), and synthetic peptides whose sequences are unique to specific subunits (see below). In early studies, polyclonal (Stephenson et al., 1986) and monoclonal (Schoch et al., 1985; Haring et al., 1985) antibodies were raised against affinity-purified GABA, receptor preparations. These were used in immunoprecipitation and autoradiographic studies to further demonstrate the coexistence and colocalization of [3H]muscimol, [3H]benzodiazepine, and [35S]TBPS binding sites (Richards et al., 1986).Mamalaki et al. (1987) further demonstrated that a monoclonal antibody recognized both the a- and P-subunits in a purified receptor preparation, providing evidence for structural homology between these subunits as was soon after directly demonstrated by cDNA sequencing (Schofield et al., 1987). Although purified GABA, receptors were originally thought to contain only an a- and P-subunit (Section III,A,l), more detailed analysis has since revealed heterogeneity of these subunits. Monoclonal antibodies raised against partially purified receptors have been shown to recognize multiple bands on Western blots (Fuchs and Sieghart, 1989; Fuchs et al., 1990; Bureau and Olsen, 1990; Park and de Blas, 1991). In view of the structural homology of the GABA, receptor subunits, it is not surprising that such antibodies raised against the intact protein recognize multiple subunits. However, with the knowledge of subunit sequences that has been provided by molecular cloning, a number of laboratories have now succeeded in producing subunit-specific antibodies raised against either short synthetic peptides or fusion proteins containing putative intracellular receptor sequences as antigens. These antibodies are currently being used in quantitative immunoprecipitation and Western blotting analyses to identify receptor subtypes and to probe their subunit compositions. Using these techniques, a,-, a2-,a3-,and y2-subunits have been identified as integral components of affinity-purified receptors (see Stephenson, 1992). In the brain, the subunits that have been found to be most abundant are a , (Benke et al., 1991a; Duggan et al., 1991; Endo and Olsen, 1993, Liiddens et al., 1991; McKernan et al., 1991), &/p3 (Benke et al., 1991a), and y 2 (Benke et al., 1991a; Duggan et al., 1992). These subunits appear to associate to form a major receptor subtype (Benke et al., 1991a; Duggan et al., 1992), a finding that supports the conclusions drawn from in situ hybridization and immunocytochemical studies described below (see Sections V,B and V,C). Although initial immunopurification studies suggested that individual GABA, receptor complexes did not contain more than one type of subunit from the same class (Duggan and Stephenson, 1990), other stud-

GABA, RECEPTOR

71

ies have contradicted this notion. An a,-subunit-specific antibody was able to precipitate a higher percentage of receptors than expected on the basis of the abundance of Bzl receptors (Liiddens et al., 1991; Section V,A, 1) and also the sum of cortical receptors precipitated by four individual anti-a-subunit (a1,a2,a3,and a4)antibodies was greater than 150% (Endo and Olsen, 1993). Using antibodies against a-subunit ( a l ,apta3, and as)sequences, it has been shown that only a single isoform is likely to be found in combination with a p- and y-subunit (McKernan et al., 1991). Thus receptors containing only a single a-subunit isoform appear to predominate. However, a minor population of receptors may contain more than a single isoform (McKernan et al., 1991). Stephenson and COworkers have used immunoblotting in conjunction with sequential antia-subunit immunoaffinity chromatography to demonstrate the presence 3, (Duggan et al., 1992), and a,a6 pairs (Pollard et al., of a I a 2 ,( ~ l ( ~ a2a3 1993).Although these results may be compromised by receptor aggregation occurring during purification, the authors used conditions to minimize such effects (Pollard et al., 1993). Initial attempts have been made to characterize the pharmacological profiles of individual receptor subtypes that have been purified by immunoaffinity chromatography. Receptors purified using an a,-subunitspecific antibody displayed binding properties of Bz 1 receptors, whereas receptors purified using anti-a, o r anti-as-subunit-specific antibodies displayed a Bz2 profile (McKernan et al., 1991; Zezula and Sieghart, 1991). These characteristics are in general agreement with the profiles of GABAA receptors in cells transiently transfected with these subunits in combination with a PI- and a y,-subunit (Section V,A). Such studies of native receptors will be important to complement the information obtained in expression studies. T h e continued development of this antibody repertoire and their use, in combination with in situ hybridization, will undoubtedly yield important information concerning the coordinate expression of individual subunits within specific cell populations. Further analysis of native receptors will then allow us to determine their oligomeric structures, thus providing access to the central question concerning the functional diversity of the GABAA receptor family. However, before considering these issues, we will first discuss the cloning of GABAA receptor genes. IV. Molecular Cloning of Receptor Subunits

A. INITIALISOLATIONOF RECEPTORSUBUNIT cDNAs

The partial amino acid sequence, obtained as a result of the purification of the GABAA receptor from bovine brain, allowed the design of

72

SUSAN M. J . DUNN rf al.

oligonucleotide probes for the isolation of cDNAs encoding the a ] -and P,-subunits (Schofield et al., 1987). The deduced amino acid sequences of the encoded polypeptides were found to share approximately 35% identity. Comparison with the amino acid sequences of other neurotransmitter receptors, in particular the nicotinic acetylcholine and glycine receptors, also revealed sequence indentity and indicated that these receptors share certain conserved structural features, leading to the proposal that these receptors are members of a ligand-gated ion channel superfamily (Barnard et al., 1987; Schofield et al., 1987). T h e mature subunits contain four putative membrane spanning domains, the second of which is proposed to form the inner lining of the ion channel. The long N-terminal region contains a number of consensus sequences for N-linked glycosylation and is believed to be extracellular (Barnard et aE., 1987; Schofield et al., 1987). This region also contains the Cys-Cys loop, which has been proposed to play a role in agonist binding (Cockroft et al., 1990). Between the third and fourth transmembrane domains there exists a long loop that is presumed to be intracellular and can contain potential post-translational modification sites (Barnard et al., 1987; Schofield et al., 1987).This superfamily of ligand-gated ion channels currently includes the GABAA, glycine, nicotinic, and acetylcholine (both muscle and neuronal) and the recently described 5HT3 receptor families (Barnard, 1992). Further molecular biological studies revealed the existence of two other a-subunits (Levitan et al., 1988a). When an a,-, a*-,or a,-subunit was expressed in combination with the P,-subunit in Xenopus oocytes, functional receptors were produced that could be distinguished electrophysiologically by a 30-fold difference in their apparent sensitivity to GABA (Levitan et al., 1988a). A possible molecular basis for GABA, receptor heterogeneity was thereby suggested. I t was not possible, however, to demonstrate a robust benzodiazepine effect for these recombinant receptors, suggesting that an additional subunit, or factor, was necessary to produce GABA, receptors with the full range of characteristics as those found in viuo (Levitan et al., 1988b).

OF A cDNA ENCODING A y 2 SUBUNIT B. ISOLATION

In an effort to isolate further cDNA clones encoding GABAA receptor subunits, Seeburg and co-workers used a 96-fold degenerate pool of 23base oligonucleotides (Pritchett et al., 1989b; Ymer et al., 1989) designed to correspond to an octameric amino acid sequence (TTVLTMTT) that was found to be present in the second proposed transmembrane domains

C;ABA,4 RECEPTOR

73

of the GABAA receptor a,-, a2-,as-,and P,-subunits (Schofield et al., 1987; Levitan et al., 1988a) and the glycine receptor a,-subunit (Grenningloh et al., 1987). This approach led the isolation of a number of clones, including one that encoded a polypeptide, named y2,that shares approximately 40% sequence identity with the previously described a,-and P1-subunits. When coexpressed with a,-and P,-subunits, the resultant receptor displayed a high-affinity binding site for centrally acting benzodiazepines and a potentiation of the GABA-mediated response (Pritchett et al., 1989b).

C. RECEPTOR HETEROGENEXTY REVEALED BY MULTIPLESUBUNITS A rapid explosion in the number and type of GABA, receptor subunit cDNAs soon resulted in the identification of a large gene family (reviewed in Wisden and Seeburg, 1092). In the rat, there are at least 13 separate subunits, which, on the basis of sequence identity, can be divided into four classes. Thus there are six a-subunits, three P-subunits, three ysubunits, and one &subunit (Wisden and Seeburg, 1992). In general, members of different classes share approximately 30-40% sequence identity, and this can rise to approximately 70% between members of the same class. A fifth subunit class has recently been described containing two members, p , (Cutting et al., 1991) and p 2 (Cutting et al., 1992), which are predominantly found in retina. It is not clear at present, however, whether these subunits form part of a functional GABAA receptor in vivo (Section V,A,5; Cutting et al., 1991, 1992). One confusing aspect that has arisen in the literature is the different names given to the same subunit sequence by different groups. Tobin and co-workers described the sequence of a rat subunit, which they termed a4 (Khrestchatisky et al., 1989). Two other groups published descriptions of the same subunit sequence, which they both termed a5 (Malherbe el al., 1990a; Pritchett and Seeburg, 1990). To avoid confusion in this review, we will use the latter nomenclature when referring to this subunit. T h e isolation of cDNAs encoding GABAA receptor subunits from nonmammalian species has revealed that GABAA receptor sequences have been more highly conserved through evolution than other members of this ligand-gated ion-channel superfamily. For example, the chicken GABA, receptor a,subunit is 98% identical to any mammalian a,subunit (Bateson et al., 1991a),whereas the chicken neuronal nicotinic acetylcholine receptor &-subunit (Nef et al., 1988) is only 85% identical to the rat homologue (Deneris et al., 1988). At least one subunit, termed p4,has

74

SUSAN M. J. DUNN et al.

been identified in avian species (Bateson et al., 1991b), the mammalian homologue of which has yet to be found. Cloning studies in invertebrate species have also identified GABA, receptor subunits that display significant sequence identities (up to 52%)with those of vertebrates (ffrenchConstant et al., 1991; Harvey et al., 1991).

D. FURTHER HETEROGENEITY ARISESFROM ALTERNATE SPLICING A further mechanism for the generation of receptor heterogeneity has been revealed by the identification of alternate splice variants of some GABA, receptor subunits. The 7,-subunit occurs as two variants that differ by the presence ( Y ~o~r absence ) (y2s)of 8 amino acids in the presumed intracellular loop region that lies between the third and fourth membrane-spanning domains. This has been shown to occur in bovine, human (Whiting et al., 1990), mouse (Kofuji et al., 1991), and chicken brains (Glencorse et al., 1990, 1992) and the &amino acid insertion present in the yZLvariant is absolutely conserved among these species. In the case of both the mouse gene (Kofuji et al., 1991) and the bovine gene (Whiting et al., 1990), this %amino acid insert is encoded by a separate exon of 24 bp. Based on the sequence conservation in this region between different species it has been proposed that a similar situation occurs for the chicken gene (Glencorse et al., 1992). T w o splice variants of the chicken P,-subunit that differ by the presence (P,') or absence (&) of four amino acids have also been demonstrated (Bateson et al., 1991b). Similar to the 7,-subunit, this occurs between the third and fourth membrane-spanning domains. These variants are generated, however, by a different alternative splicing mechanism, that of differential choice of 5' donor splice site. T h e Drosophila GABA, receptor gene (Rdl) transcript undergoes extensive alternative splicing. Two pairs of exons encoding portions of the N-terminal extracellular domain each undergo alternative usage such that four different forms of this subunit are generated (ffrench-Constant and Rocheleau, 1993).

V. Characterization of the Receptor Family

A. HETEROLOGOUS EXPRESSION REVEALSDIFFERENT FUNCTIONAL ATTRIBUTES OF GABA, RECEPTOR SUBUNITS Intense effort has been maintained in examining the ligand-binding characteristics and functional attributes of recombinant GABA, recep-

GABA, RECEPTOR

75

tors expressed in heterologous expression systems. An initial hope was that by defining their properties, in comparison to those of in vivo receptors, equivalents could be found and thus the subunit compositions of in vivo receptors deduced. This has largely floundered in the face of the growing GABA, receptor gene diversity. If all possible subunit combinations were able to produce functional receptors, there would be over 500,000 different receptor subtypes. It is clearly impossible to characterize fully even a small percentage of this number of theoretical receptor subtypes and expect to be able to use such data to identify the composition of in vivo receptors. A number of potential difficulties should be considered when interpreting heterologous expression studies of recombinant receptors. First, although the subunit composition of the recombinant receptor is assumed to correspond to the nucleic acid species transfected or injected, no one has demonstrated that all of the subunit polypeptides expected are present and appropriately assembled. Furthermore, the question of subunit stoichiometry in recombinant receptors has not been addressed. Recently, it has been shown that certain subunit combinations may be preferentially formed over others (Angelotti and Macdonald, 1993; Angelotti et al., 1993; Verdoorn et al., 1990; Section V,D), a process that may be influenced by the particular heterologous expression system used (Angelotti and Macdonald, 1993; Angelotti et al., 1993). Macdonald and co-workers propose that some of the subunit combinations, formed in studies that have used high-level heterologous expression systems, may be somewhat artifactual in that the high levels of receptor proteins that are expressed in these cells may lead to inappropriate subunit association (Angelotti and Macdonald, 1993; Angelotti et al., 1993). It is also likely that some of the combinations used in these in vitro experiments may not exist in vivo. For example, most subunits have been shown to have the capacity to form homooligomeric GABA-gated channels in transfected cells; however, they generally display small currents, suggesting that these receptors are assembled with low efficiency. Even in cases where heterooligomeric receptors d o appear to be efficiently assembled, interpretation of apparent changes in GABA affinity andlor cooperativity is difficult, as these may arise from alterations in activation and/or desensitization kinetics (Mathers, 199 1). Despite these caveats, heterologous expression studies of GABA, receptors have clearly demonstrated the potential that exists for the generation of GABA, receptor heterogeneity and have revealed valuable information about the different contributions that subunits can make to a receptor’s pharmacological profile. When interpreted in the light of gene product localization (Sections V,B and V,C) and receptor purification studies (Section III,D), they have provided useful data to corroborate

76

SUSAN M. J . DUNN et al.

receptor subunit compositions proposed on the basis of nonfunctional approaches. A concept that is emerging from heterologous expression studies is that, in general, specific roles may be ascribed to individual subunit classes. However, there are clearly many characteristics of the receptor complex that are specified by interactions between subunits of different classes. 1. a-Subunit Class The a-subunit class is the largest, comprising six members, and it is believed that they specify the heterogeneity of the benzodiazepine binding site (see Doble and Martin, 1992, and Wisden and Seeburg, 1992, for review). Ligand-binding studies of membranes of HEK293 cells transiently transfected with a , 4 y 2 combinations revealed a correspondence between the presence of an al-subunitwith benzodiazepine Bzl pharmacology and an a,- or a,-subunit with that of Bz2 receptors (Pritchett et al., 1989a). Although the y,-subunit was required for the expression of the benzodiazepine site, changing the particular @-subunit present showed no effect. Furthermore, GABA potentiation of benzodiazepine binding was greater in a,-subunit-containing receptors (Pritchett et al., 1989a). Subsequently it was shown that the a,-subunit would also confer a Bz2 pharmacology, but one that was different from that conferred by a2- or a,-subunits with respect to zolpidem affinity (Pritchett and Seeburg, 1990). A third class of benzodiazepine pharmacology is displayed by a 4 y 2 recombinant receptors, in which the a-subunit is either a4 or a6. These receptors d o not bind benzodiazepine agonists, such as diazepam, but do bind certain antagonists and inverse agonists (Luddens et al., 1990; Wisden et al., 1991). Using site-directed mutagenesis, a single amino acid difference between al-and a3-subunits has been shown to confer Bzl or Bz2 benzodiazepine pharmacology. T h e exchange of a glutamic acid residue in the a,-subunit for a glycine residue, which appears in the corresponding position of the &,-subunit, confers on the hybrid receptor (a3G225,42y2) high affinity for CL2 18872, effectively changing the pharmacology of the receptor from Bz2 to Bzl (Pritchett and Seeburg, 1991). Similarly, a specific histidine residue present in some a-subunits has been shown (Wieland et al., 1992; Korpi et al., 1993) to be necessary for high-affinity benzodiazepine agonist binding. Homooligomers of al-,ap-,or a,-subunits form GABA-gated channels that have apparently normal channel properties which can be blocked by picrotoxin and display potentiation by barbiturates (Blair et al., 1988; Pritchett et al., 1988). T h e whole-cell currents are, however, small, suggesting that subunit assembly in these receptors may not be efficient.

GABA, RECEPTOR

77

These data are in contrast to the biochemical evidence, discussed above, which indicated that the GABA agonist site might be located on the psubunit. The particular a-subunit present in heterooligomers affects not only the binding but also the functional properties of recombinant receptors. or a3-and the PIThus, receptor complexes formed with either a l - ,a,.-, subunit displayed u p to a 30-fold differing sensitivity to GABA (Levitan et al., 1988a). Similarly, the alp1combination is less sensitive to GABA than alpI (Khrestchatisky et al., 1989), whereas the a5pIcombination is more sensitive than a l p l or a3pI (Malherbe et al., 1990a; Sigel et al., 1990). There may also be differences between species in that rat alpI o r as@, combinations display similar GABA dose-response curves (Malherbe et al., 1990a), whereas the equivalent bovine combinations show a small (threefold) difference in GABA sensitivity (Levitan et al., 1988a). Addition of the y,-subunit to many of these combinations appears to maintain or even increase the influence of the a-subunit isoform present on GABA sensitivity (Sigel et al., 1990). The potency of steroid action has been shown to depend on the ti-subunit present: a l p l or asp1combinations display greater potentiation than aePI(Shingai et al., 1991). Surprisingly, the addition of the y,-subunit to the a l p l or the a2p1combinations produced an increase in steroid potency, whereas the a3p,y2combination was potentiated to lesser degree than that of aspl (Shingai et al., 1991). T h e use of y-subunit-containing ternary combinations allows an assessment to be made of exchanging a-subunits on the efficacy of benzodiazepine potentiation. Some discrepancies appear, however, between the results obtained from two comprehensive studies, one using rat subunits expressed in oocytes (Sigel et al., 1990) and the other using human subunits expressed in transfected cells (Puia et al., 1992). Ternary complexes (spy) containing an a,-subunit display the greatest potentiation of GABA-mediated currents by diazepam with a rank order of either a3 > a2 > a 1> a5 (Puia et al., 1992) or a , > a5 > a I (Sigel et al., 1990). In transfected cells (Puia et al., 1992) the GABA responses of these complexes are also differentially modulated by other compounds that act at the benzodiazepine site. For example, in contrast to the effects of diazepam, alpidem displays a rank order of efficacy on GABA-mediated currents of a 1 = a2 > a, % a5 (Puia et al., 1992). In support of this, Wafford et al. (1993) recently reported that a variety of benzodiazepine site ligands (full, partial, and inverse agonists) display differential potencand a 3 p I y 2 L combinations. The efficacy of ies and efficacies for a1/31y2L triazolam on GABA-gated currents was also shown to be greater than that of diazepam on a l P I y , and a,P1y2 combinations but not on the a3pIy2combination (Ducic et al., 1993).

78

SUSAN M. J. DUNN et al

Only one study has reported in detail the coexpression of a number of subunit combinations that include more than one a-subunit with a pand a y-subunit (Sigel et al., 1990). They found that in contrast to a ffl(Y3(Y$I&y2 combination, omission of the q s u b u n i t (aI(~3plp2y2)resulted in a partial loss of cooperativity and affinity for GABA. In fact all combinations tested that included the a,-subunit produced receptors that displayed properties closest to those found in vivo (Sigel et al., 1990). The presence of at least one a-subunit was deemed necessary for the efficient production of functional GABA, receptors with large conductances (Sigel et al., 1990; Verdoorn et al., 1990). Significant benzodiazepine potentiation was obtained in at least one subunit combination lacking any a-subunits (p2y2),indicating that the a-subunits are not required for the formation of this site (Sigel et al., 1990). It is not clear, however, whether such a receptor is present in vivo, as other functional attributes of this particular combination do not reflect native receptor characteristics (Sigel et al., 1990; Verdoorn et al., 1990). Taken together, these data suggest that the a-subunits, in recombinant receptor complexes, play a major role (but not an exclusive one) in determining the nature of binding sites for GABA and some of the allosteric effectors. 2. p-Subunit Class There is little evidence to suggest that the particular P-subunit isoform influences the ligand-binding characteristics of recombinant receptors (Pritchett et al., 1989a; Wisden and Seeburg, 1992). Homooligomeric rat p,-subunit-containing channels can be formed in oocytes but the channels appear to open spontaneously in the absence of GABA (Sigel et al., 1989). In contrast, human P,-subunits expressed alone in HEK293 cells produced GABA-gated channels that could also be potentiated by barbiturates and blocked by picrotoxin (Pritchett et al., 1988). Analysis of heterooligomer combinations expressed in oocytes demonstrated that a @subunit is not essential for the formation of a picrotoxin site (Sigel et al., 1990). These ay complexes form functional GABA-gated channels that display GABA cooperativity and benzodiazepine sensitivity, but they exhibit relatively small currents (Sigel et al., 1990). In contrast, the a 1 y 2 combination expressed in cultured HEK293 cells produced large GABAactivated currents, equivalent to those seen for the a1P2y2combination (Verdoorn et al., 1990), highlighting the need for care in interpreting expression data from different experimental systems. In the latter study, the channel properties of alp2and a l y , combinations were found to be different, suggesting a role of the @-subunit (and the y-subunit) in determining channel characteristics (Verdoorn et al., 1990). In oocytes,

GABA, RECEPTOR

79

the exchange of p1for @, in aXpYy2combinations generally resulted in a decrease in the apparent affinity for GABA but a marked increase in potentiation by diazepam (Sigel et al., 1990). This result is also in contrast to that obtained with similar combinations (a,P,y,, alP2y,, o r a l P 3 y 2 in ) HEK293 cells where no effect of the particular P-subunit was found on diazepam potentiation (Puia et al., 1992). These apparent differences in diazepam sensitivity may be a result of the expression system used, but may also arise from the use of differing diazepam concentrations (1 mM in the oocyte studies; Sigel et al., 1990; 10 mM in the HEK293 studies; Puia et al., 1992). Indeed, the most effective diazepam concentration was shown to be 0.3 to 1 mM for @,- or @,-containing combinations in oocytes (Sigel et al., 1990). Clearly, more detailed studies are required to test the possible effects of @-subunitson receptor channel properties and benzodiazepine potentiation. 3. y-Subunit Class

The y,-subunit can form homooligomeric channels in HEK293 cells that are reversibly potentiated by barbiturates (Shivers et al., 1989). Although the y,-subunit is required for expression of a benzodiazepine site, it is insufficient alone to achieve this (Pritchett et al., 1989b; Shivers et al., 1989). Unlike the @-subunitsthere are clear differences in both the binding and the functional properties of heterooligomeric recombinant receptors containing differing y-subunits. All three y-subunits, when combined with a-and @-subunits,can confer benzodiazepine binding on recombinant receptor complexes (Knoflach et al., 1991; Pritchett et al., 1989a,b; Ymer el al., 1990). I n combination with an a- and a @-subunit,the y l subunit produces a marked decrease in affinity for the antagonist flumazenil and the inverse agonist DMCM in comparison to 7,-containing combinations (Ymer et al., 1990). In contrast, an a1P2y3combination shows a marked decrease in benzodiazepine agonist affinity, relative to that of an al@,y2combination, whereas both combinations have similar affinities for antagonists and inverse agonists (Herb et al., 1992). Interestingly for these combinations, receptors containing y 3 displayed much greater differences in affinities for certain benzodiazepine site agonists (e.g., midazolam versus zolpidem) than did the corresponding y2containing combination (Herb et al., 1992). Functional differences between the different 7-subunit-containing receptor complexes are particularly striking. Most notably, exchanging the 7,-subunit for the y,-subunit in many aPy combinations changes the action of DMCM and @-CCM from that of inverse agonist to agonist (Puia et al., 1991). These data provide an explanation for the previous

80

SUSAN M . J. DUNN

et

al.

finding that these compounds function as agonists on GABAA receptors in astrocytes (Bormann and Kettenmann, 1988) since these cells have been shown to express the y I subunit gene to higher levels than those of other y-subunits (Bovolin et al., 1992a). In general, the higher affinities of benzodiazepine ligands for 7,-subunit-containing complexes is reflected in increased modulation of GABA-gated currents (Herb et al., 1992; Knoflach ct al., 1991; Puia et al., 1991). Some benzodiazepine effects, however, may be attributed to differential cooperative interactions between y- and a-subunits. For example, pCCM displays a similar degree of inverse agonist activity on a,/3,y, and a,p,y, combinations. It is also an inverse agonist of an a3/3,y,combination but it is an agonist of the a 2 P I y Icombination (Puia et al., 1991). The close interaction suggested by these studies between the a- and the y-subunits in the formation of the benzodiazepine site and the determination of its functional attributes indicate that this site may not reside solely on the a-subunit, as had been suggested by previous biochemical studies (Section III,B,3). Currently no site-directed mutagenesis studies have been reported investigating the possible role of the y-subunit either in inducing conformational changes in the a-subunit to allow binding or in directly forming part of the benzodiazepine site. Such studies may give a clearer understanding of the molecular interactions involved in these effects. The y,,-subunit alternative splice variant has been shown to be responsible for conferring ethanol enhancement in receptors expressed in oocytes from recombinant clones and brain mRNA (Wafford et al., 1991). Furthermore, the previously proposed phosphorylation site within the 8-amino acid insert of the y,,-subunit (Whiting et al., 1990) appears to be essential for this effect (Wafford and Whiting, 1992). No differences in benzodiazepine or barbiturate potentiation of receptors containing either the yZL-or the y,,-subunit forms have been detected (Wafford and Whiting, 1992; Wafford et al., 1991).

4. &Subunit Class There is virtually no information regarding the possible contribution of the 6-subunit to GABA, receptor characteristics. This subunit can form homooligomers in HEK293 cells that display GABA-gated channels with small currents that are sensitive to picrotoxin, bicuculline, and pentobarbitol (Shivers et al., 1989). These workers also cite unpublished data indicating that the b-subunit cannot replace the ?,-subunit in the generation of receptors with high-affinity benzodiazepine binding sites (Shivers et al., 1989).

GABA, RECEPTOR

81

5. p-Subunit Class Two p-subunits have been identified and their expression is generally restricted to retina. The pl-subunit, in contrast to members of all other GABAA receptor classes, forms homooligomers that display large GABAgated currents (Cutting et al., 1991; Shimada et al., 1992). Furthermore, these channels display an unusual pharmacology that is not altered by coexpression with a-,p-, or y-subunits, picrotoxin-sensitive, barbiturate-, benzodiazepine-, bicuculline-, and baclofen-insensitive (Cutting et al., 1991; Shimada et al., 1992). I t appears likely that this subunit forms homooligomers in vivo in retina and has been classified as GABA, rather than GABAA (Shimada et al., 1992), although other workers argue for a separate classification entirely (Woodward et al., 1993).

B. GENEEXPRESSION A large body of literature attests to the intense effort that has gone into examining the expression characteristics of the GABAA receptor gene family. It was initially hoped that such studies, particularly those employing in situ hybridization techniques, would provide some insight into the possible subunit combinations that might exist in viuo. It soon became clear, however, that this gene family consisted of many members, each displaying specific temporal and spatial expression patterns. Nevertheless, these studies have been useful in defining the complexity of GABAA receptor gene expression and they have also provided some clues to the identity of possible in vivo receptor subunit combinations. Consequently, this section will deal with these latter conclusions and will not attempt a comprehensive review of this literature. 1 . Cell-Specific Expression

T h e rodent brain has been most widely used for the analysis of GABA, receptor gene expression by in situ hybridization with most studies focusing on a limited number of transcripts in particular brain regions. In addition, different groups have used a variety of probe types (oligonucleotides, DNA, or antisense mRNA) for the detection of GABAA receptor gene transcripts; consequently, although qualitative comparisons between these studies are possible, quantitative or even semiquantitative comparisons are difficult. Two groups have, however, made comprehensive studies of the distribution of GABAA receptor subunit mRNAs in the rat brain that do not suffer from these limitations. Seeburg and co-

82

SUSAN M. J . DUNN et al.

workers examined the distribution of all 13 known GABAA receptor genes (Laurie et al., 1992a; Wisden et al., 1992), whereas Richards and co-workers examined all hut those mRNAs encoding the ad-and the y3subunits (Persohn et al., 1992). The data from these studies are by and large consistent, not only with each other, but also with the more limited studies conducted by other workers. Seeburg and co-workers (Laurie et al., 1992a; Wisden et al., 1992) discuss in detail their colocalization data, with respect to previously reported ligand-binding autoradiographic and in vitro expression data, in an attempt to propose rationally possible subunit combinations of in vivo receptor subtypes. Some rat brain regions contain a large repertoire of GABAAreceptor transcripts, making any deductions about the possible subunit compositions in these areas difficult. For example, all subunit mRNAs except those encoding the a,-subunit are found in the dentate granule cells (Wisden et al., 1992). There are, however, a number of brain regions that express a more limited set of GABA, receptor genes, enabling restricted conclusions to he drawn about the subunit combinations that might give rise to specific receptor subtypes in these regions. Certain subunit genes are very often coexpressed at similar levels, such as a l p 2 , (~2p3, a& and asp,.The major GABAA receptor subtype is proposed to be a,P,y2 (Laurie et al., 1992a; Shivers et al., 1989; Wisden et al., 1992), which would correspond to the Bzl subtype (Pritchett et al., 1989a). Similarly, the major receptors constituting the Bz2 subtype are proposed to be a2p3y2,a3Pxy2 and a s p l y , combinations (Wisden et al., 1992). In certain brain regions other combinations are suggested that would represent specific receptor subtypes present at low overall levels. For example, the a,-subunit transcript is only found in cerebellar granule cells (Kato, 1990),in combination with a,-, p2-,p3-,y2-,and $-subunit mRNAs (Laurie et al., 1992a; Wisden et al., 1992). Given the evidence suggesting a pentameric structure for the GABAA receptor (Section V,D), the presence of six subunit mRNAs in these cells suggests at least two receptor subtypes there and Laurie et al. (1992a) propose a,&y2 and a,a,&y2 as likely combinations. Supporting evidence from immunoprecipitation studies for more than one a-subunit in a given receptor complex has been discussed previously (Section 111,D). Putative Bergmann glia, found in the Purkinje cell layer of the cerebellum, appear to contain only a2-and y,-subunit transcripts (Laurie et al., 1992a), a combination that has been shown to produce functional receptors in transfected cells (Verdoorn et al., 1990). In the thalamus, a region that has been proposed to contain GABAA receptors without an associated benzodiazepine binding site (Olsen et al., 1990), only a l - , a4-,and p2-subunit mRNAs are present, with the &subunit mRNA displaying a more restricted distribution pat-

GABA,4 RECEPTOR

83

tern (Wisden et al., 1992). Thus, it is possible that arla4pnoccur together with the addition of 6 in some thalamic nuclei (Wisden et al., 1992). In the spinal cord, major subtypes have been proposed to comprise combinations of a,, a,,a3,p3, and y2 (Persohn et al., 1991, 1992). T h e large size of spinal motor and ganglionic neurons allowed ultrathin sections to be used for sequential probe analysis, demonstrating colocalization of subunit transcripts a I with a,, a, with p3, and p3 with y2 in individual cells (Persohn et al., 1992). Furthermore, the combination ~ t z p 3 ~was 2 shown to colocalize to individual spinal cord motoneurons (Persohn et al., 1992). T h e various subunit combinations suggested to exist in rat brain do not take into account the presence of the y l - and 7,-subunit transcripts that are found in a number of specific brain regions, albeit at lower levels than that of the y,-subunit (Laurie et al., 1992a; Wisden et al., 1992). Thus, it is possible that a number of the subunit combinations proposed as receptor subtypes that contain the y,-subunit may also exist as further combinations with the y,-subunit substituted with either a yI- or a y 3 subunit (Laurie et al., 1992a; Wisden et al., 1992). There are a few nuclei that contain lower levels of the y,-subunit mRNA relative to that of the y,-subunit (e.g., medial amygdaloid nucleus) or the y,-subunit (e.g., medial geniculate nucleus) transcripts (Wisden et al., 1992). In these regions, however, assignment of subunit combinations is difficult because of the large number of other subunit transcripts present (Wisden et al., 1992). T h e above studies used probes that detect both the long and the short alternate splice forms of the y,-subunit gene transcript and no in situ hybridization study has been reported to date describing the possible differential distribution of these two y,-subunit isoforms in adult rat brain. We have demonstrated that these variants are differentially expressed in the brains of l-day-old chicks (Glencorse et al., 1992) using oligonucleotide probes that specifically detect each of these mRNA species. These data demonstrate that certain brain regions either express one of these two transcripts or express both. It is possible, therefore, that cells expressing both mRNAs contain either two receptor subtypes, each containing a different y,-subunit isoform, or a single receptor subtype that incorporates both ?,-subunit isoforms (Glencorse et al., 1992). It is probable that a similar differential distribution of these alternative splice forms in the rat brain will be revealed by in situ hybidization, especially given the similarity in distribution patterns of both the 7,- and the &,-subunit transcripts between chick and rat brains (Bateson et al., 1991a; Glencorse et al., 1991). In support of this, Whiting et al. (1990) used the polymerase chain reaction to show differences in the relative

84

SUSAN M. J. DUNN el al.

amounts of these two alternate splice forms between gross regions of the rat brain. More recently, it has been shown that the y,,-subunit isoform is preferentially expressed, over that of the yZL,in rat pituitary cells (Valerio et al., 1992) and in the developing rat embryo (Poulter et al., 1993). As a consequence, therefore, of the differential expression of the 7,-subunit isoforms, it is most likely that each proposed receptor subunit combination that contains a y,-subunit actually represents either two (yZLor y Z s )or three (yZL,yZs,and yZLwith Y , ~ )receptor subtypes.

2 . Developmental Expression Given the large number of GABAA receptor genes, it is perhaps not surprising that fewer studies have been conducted tracing developmental expression. Nevertheless, it is clear that each subunit gene displays a specific developmental expression profile. For example, the overall levels of the +-subunit (Laurie et al., 1992b; MacLennan et al., 1991), agsubunit (Laurie et al., 1992b),a,-subunit (Laurie et al., 1992b; MacLennan et al., 1991), and a,-subunit (Laurie et al., 1992b) transcripts are higher in the brains of developing rats than in those of adult animals. In contrast, the a,-subunit (Beattie and Siegel, 1993; Bovolin et al., 1992b; Gambarana et al., 1991; Laurie et al., 1992b; MacLennan et al., 1991; Zhang et al., 1992) and +subunit (Laurie et al., 1992b) transcripts are undetectable before birth and reach peak expression in the adult rat brain. Similarly, differential expression profiles exist for the other subunit transcripts with those encoding the &-subunit (Beattie and Siegel, 1993; Gambarana et al., 1991; Laurie et al., 1992b), y,-subunit (Beattie and Siegel, 1993; Gambarana et al., 1991; Laurie et al., 1992b), and 6subunit (Laurie et al., 1992b) being present predominantly in the adult rat brain. The alternative splicing of the ?,-subunit transcript is also developmentally regulated, with levels of yps-subunit isoform remaining fairly constant from birth to adult whereas the y,,-subunit isoform rises from virtually undetectable levels at birth to peak expression in the brains of adult rats (Bovolin et al., 1992b) and mice (Wang and Burt, 1991). In brain regions that express only a subset of GABA, receptor subunit genes, it has been possible to trace a developmental switch from one subunit combination to another. Thus, the globus pallidus changes from a a2/a3&y,-subunit combination that that of a,p2y,/y2and the medial septum changes from a a,la3&y,-subunit combination to that of a,&y, (Laurie et al., 1992b). Such changes in expression are not universal, however, with some brain regions and cell types displaying differential developmental profiles. For example, in the cerebellum the levels of a,-, &-, &-, and y,-subunit transcripts in Purkinje cells did not alter from Postnatal Day 6 (when these cells could be resolved as a monolayer) to

400 / A M )partially suppresse GABA, receptor-mediated responses (Mott and Lewis, 1991). Thus, this antagonist appears to most effectively discriminate GABAB receptors when used in low concentrations. Similar to the strategy used to produce the potent agonists 3-APPA and 3-APMA, the most effective GABAB receptor antagonists to date are based on the phosphinic analogues of GABA. The first of these to be introduced was CGP 35348 [P-(3-aminopropyl)-P-diethoxymethylphosphinic acid (Bittiger et al., 1990; Oipe et al., 1990; Seabrook et al., 1990)] (Fig. 2). Binding studies reported that this compound was a competitive antagonist; however, the affinity of this compound for GABAB receptors was weak, similar to that of 2-hydroxysaclofen. CGP 35348 appeared to be very selective for GABAB receptors, as it has been shown not to act on GABA,, glutamate, muscarinic acetylcholine, a-adrenergic, serotonin (5-HT,), histamine (H,), or adenosine (A,) receptors. T h e major advantage of CGP 35348 over phaclofen and 2-hydroxysaclofen is its ability to cross the blood-brain barrier, allowing it to be systemically administered in vivo. A variety of other phosphinic analogues of GABA have recently been introduced, some of which are significantly more potent than previous GABA, receptor antagonists (for review see Froestl et al., 1992). Two of these compounds represent the first orally active GABAB receptor antagonists (Bittiger et al., 1992a; Lingenhohl and Olpe, 1993; Olpe et al., 1993a). These are P-(3-aminopropyl)-P-"-butyl-phosphinic acid (CGP 36742) and P-(3-aminopropyl)-P-cyclohexylmethylphosphinic acid (CGP 46381) (Fig. 2). Radioligand binding studies have shown that the potency of CGP 36742 in displacing bound 3-[3H]APPA is similar to that of CGP 35348; however, CGP 46381 is about seven-fold more potent. The higher potency of CGP 46381 is also evident after oral administration in vivo, where it substantially antagonizes the effects of baclofen at a dose (30 mg/kg) at which CGP 36348 is ineffective (Olpe et al., 1993a). Radioligand binding studies show that both of these compounds are selective for GABAB receptors, as they do not displace any of a variety of other receptor ligands (Bittiger et al., 1992a; Olpe et al., 1993a). Two of the most potent of the newly introduced GABAB receptor antagonists are 3-N-[l-{(S )-3,4-dichlorophenyl}-ethylamino]-2-(S )hydroxypropyl cyclohexylmethyl phosphinic acid (CGP 54626) and its a-(S)-methyl analogue, CGP 55845 (Bittiger et al., 1992b; Froestl et al., 1992; Jarolimek et aE., 1993) (Fig. 2). These compounds were each found

I03

CENTRAL CABA, RECEPTORS 0

H.#J-i.on 0

HaN-COOH

DAVA

3-APS

tl

CI

Phaclofen

2-Hydroxysaclofen

CGP 35348

CGP 46381

C G P 36742

C G P 54626

CGP 55845 FIG. 2. Structures of some selected GABA, receptor antagonists.

to displace 3-[3H]APPA binding to GABA, receptors with an IC,o of about 7 nM, making them approximately 15,000 times more potent than phaclofen and 5000 times more potent than CGP 35348. By comparison, the affinity of these compounds for the GABA, receptor is greater than that of bicuculline for the GABA, receptor as determined by either a [3H]muscimol assay [IC,, = 2 p M (Beaumont el al., 1978)] or ['HI( +)bi-

104

DAVID D. MOTT A N D DARRELL V. LEWIS

cuculline methiodide assay[IC,o = 68 nh4 (Mohler and Okada, 1977)]. Because they have been recently introduced, little information is available regarding the selectivity of these compounds, although they have been reported to act exclusively at GABAB receptors in radioligand binding studies (Olpe et a1.,1993a). Similarly, little is known of their effects on synaptic responses, although a recent study by Jarolimek et al. (1993) reported that CGP 55845 potently antagonized postsynaptic GABAB receptor-mediated currents in hippocampal neurons. These new highly potent antagonists promise to provide many insights into the physiological role of GABAB receptors. In particular, the ability of several of these compounds to be orally active suggests that they will be useful tools for investigating the role of GABAB receptors in modulating behavior. Preliminary reports suggest that these new antagonists are not selective for GABAB receptor subtypes. As will be discussed below (see Section IV,D), GABAB receptors located postsynaptically can have different and often opposing effects on neuronal activity compared to those located presynaptically on axon terminals. Because of these opposing actions, generalized GABAB receptor antagonists often produce complex and contradictory effects on neuronal activity. Thus, antagonists that are selective for presynaptic or postsynaptic GABAB receptors are needed.

111. Properties of GABAB Receptors

A. GABAB RECEPTORBINDING GABAB receptor binding was first demonstrated in crude synaptic membranes using the tritiated form of baclofen as the prototype ligand (Hill and Bowery, 1981; Bowery et al., 1983). [3H]Baclofen was used because of its high potency and specificity for bicuculline-insensitive GABA receptors as well as its inability to be transported into neuronal tissue via the Nat-dependent GABA uptake process (Bowery et al., 1983). 3H-Baclofenbinding was stereospecific with the ( - )isomer being approximately 100 times more potent than the (+)isomer (Hill and Bowery, 1981; Robinson et al., 1989). Subsequent studies have commonly used [3H]GABA with addition of the GABAA agonist isoguvacine to prevent binding to GABAA receptors. These studies have reported that both GABA and (-)baclofen bind to a common GABAB receptor site with equal affinity (Hill and Bowery, 1981). In general, Hill coefficients close to unity and linear Scatchard plots have been reported, indicating a

CENTRAL G A B A s RECEPTORS

105

single population of GABA, receptors (Hill and Bowery, 1981; Majewska and Chuang, 1984; Bowery et al., 198513; Al-Dahan and Thalmann, 1989; Chu et al., 1990). The binding affinity ( K d )of [3H]GABA for both GABAB and GABAA receptors is similar and has typically been reported to be in the range of 30 to 80 nM (Bowery et al., 1983; Falch et al., 1986, but see Chu et al., 1990). However, several studies, using low concentrations of [3H](- )baclofen o r ['HIGABA, have reported both low-affinity (Kd 229-304 nM) and high-affinity (Kd 19-60 nM) forms of the receptor, suggesting either that the same receptor can exist in two different affinity states or that there are two populations of GABAB receptors (Karbon et al., 1983; Bowery et al., 1985b). In addition to GABA and baclofen, several other compounds have also been examined for their ability to displace 3H(-)baclofen from GABAB receptors. For example, 3-APPA is approximately 100 times more active than (-)baclofen at GABAB receptors (Pratt et al., 1989; Ong et al., 1990a). In contrast, the GABAA agonist isoguvacine and the GABAA antagonist bicuculline methobromide are devoid of any activity at the GABAB receptor, whereas the GABAA agonist muscimol is only weakly active, being over 300-fold less potent than (-)baclofen (Bowery et al., 1983). A summary of the relative potencies of several different compounds is given in Table I. Unlike binding to GABAA receptors, ligand binding to GABA, receptors is dependent on physiological concentrations of the divalent cations Ca2+ or Mg2+ (Hill and Bowery, 1981; Bowery et al., 1983; Kato et al., 1983; Majewska and Chuang, 1984). [3H]Baclofen or [3H] GABA binding is increased in a concentration-dependent fashion by divalent cations with the following order of potency Mn2+ = Ni2+ > Mg2+ > Ca2+> Sr2+> Ba" (Kato et al., 1983). T h e maximal effect of these divalent cations is not additive, suggesting that they act through a common binding site. Although these cations enhance GABA, receptor binding, other divalent cations, including Hg2+,Pb'+, Cd2+,and Zn2+, inhibit binding of [2H]baclofen to GABAB receptors (Drew et al., 1984; Turgeon and Albin, 1992). The opposing effects of these two groups suggest the existence of both excitatory and inhibitory cation sites for modulation of CABAB receptor binding. Another characteristic of CABAB receptors is their sensitivity to guanyl nucleotides. Guanyl nucleotides, such as GTP, do not affect GABAA receptor binding, but potently inhibit GABAB receptor binding (Hill et al., 1984; Asano et al., 1985; Ohmori et al., 1990). This effect is specific for guanyl nucleotides, as it is not mimicked by adenosine 5'-triphosphate (ATP). The inhibition of ligand binding produced by GTP is the result of a reduction in the affinity of the GABAB receptors rather than a

106

DAVID D. MOTT AND DARRELL V. LEWIS TABLE I POTENCY OF VARIOUSGABA ANALOGUES AT GABABSITES Analogue

GABA, binding (IC50; F M ) ”

GABA agonists GABA (-)Badofen (?)Badofen (+ )Badofen 3-APPA 3-APMA Muscimol Isoguvacine

0.026-0.04 0.03-0.08 0.18 15-33 0.00 1-0.007 0.014 2.4-12.3 > 100

GABA antagonists 3-APS 8-Aminovaleric acid Phaclofen 2-H ydroxysaclofen CGP 35348 CGP 36742 CGP 4638 1 CGP 54626 CGP 55845 Bicuculline methobromide

11 7-50 108-130 5.1-8.5 34 35-36 4.9 0.006-0.007 0.007 >lo0

Reference* 1-3 2-5 3

2, 3 4, 6, 7 4 2, 3, 7 3. 7 3 1-3 4, 5, 8 5, 8 9 10,Il 11 4, 12

4 3, 7

” Binding affinities were determined by measuring the displacement of bound [3H]baclofen (3,5,8), [3H]GABA + isoguvacine (1, 2, 6 ) , or [3H]3-APPA [CGP 27492, (4, 7, 9, 10, 11, 12)J from rat brain synaptic membranes. 1. Kristiansen et al. (1992); 2. Falch el al. (1986); 3. Bowery et al. (1983); 4. Froestl et al. (1992); 5. Al-Dahan et al. (1990); 6. Pratt et al. (1989); 7. Bittiger el al. (1988); 8. Drew et al. (1990); 9. Olpe el a/. (1990); 10. Bittiger et al. (1992a); 11. Olpe el a!. (1993a); 12. Bittiger et al. (1992b).

change in their binding capacity (Hill et al., 1984; Asano et al., 1985). Inhibition of GABAB receptor binding by GTP is blocked by pertussis toxin, suggesting that GABAB receptors are functionally coupled to inhibitory G proteins, either Gi and/or Go (see Section 111,C). Inhibition of ligand binding by guanyl nucleotides has been reported for other types of receptors, such as dopamine (Freedman et al., 19811, opiates (Pert and Taylor, 1980), a-adrenergic agonists (U’Prichard and Snyder, 1978), and serotonin (Mallat and Hamon, 1982), which are also coupled to GTP binding proteins. In addition to modulation of GABA, receptor binding by divalent cations and guanyl nucleotides, several other features also distinguish

CENTRAL GABA, RECEPTORS

107

GABA, from GABAA receptor binding sites. For example, numerous studies have shown that GABA binding to GABAA receptors may be enhanced by either benzodiazepines or barbiturates (Tallman et al., 1978; Willow and Johnston, 1980; Asano and Ogasawara, 1982; Skerritt et al., 1982). However, GABAB receptor binding is not affected by either of these classes of drugs (Doble and Turnbull, 1981; Bowery et al., 1988). Another difference in GABA, and GABAB receptors is in their reaction to receptor solubilization. Unlike GABA, reeptor binding, GABAB receptor binding is inhibited by exposure of the membranes to a variety of detergents, such as sodium deoxycholate o r Triton X-100 (Bowery et al., 1983; Ohmori et al., 1990), suggesting that GABAB receptor function is easily destroyed following solubilization. This reduction in GABAB receptor binding is produced primarily by a reduction in the number of binding sites, rather than any alteration in their affinity, suggesting that solubilization inhibits GABAB receptor binding either by removing the receptor and/or cation binding site from the membrane o r by denaturing the receptor (Bowery et al., 1983). T o date, GABAB receptor binding in both crude synaptic membranes and tissue slices has been examined exclusively using tritiated GABAB receptor agonists. However, Bittiger et al. (1992b), using crude membrane fractions from rat cerebral cortex, have recently reported the first GABA, binding study using a labeled GABAB receptor antagonist ([3H]CGP54626). Binding studies using labeled antagonist are of interest because they can potentially reveal different affinity states of the receptor or a different number of receptors or provide a more accurate determination of antagonist affinity than agonist binding studies. Using the labeled antagonist, Bittiger et al. (1992b) reported a maximum number of binding sites (BmaX)that was two to three times greater than that found for agonist binding. In addition, although the potencies of antagonists (CGP 35348 and CGP 36742) in displacing labeled antagonists were not different from those found in agonist displacement studies, agonists (GABA, (-)baclofen, and 3-APPA) were found to be much weaker. Taken together, these results suggest that GABAB receptor antagonists bind to different states of the GABAB receptor or receptor subtypes, which are only partially accessible to agonists. B. GABAB RECEPTORDISTRIBUTION GABA, receptors have been demonstrated on neurons in both the peripheral (For review see Ong and Kerr, 1990) and the central nervous system. In the central nervous system receptor autoradiography using [3H]GABA and [3H]baclofen (Gehlert et al., 1985; Bowery et al., 1987;

1 oa

DAVID D. MOTT AND DARRELL V. LEWIS

Chu et al., 1990) has demonstrated that, although most regions of the brain contain both GABA, and GABAs binding sites, GABAA sites predominated in the majority of these areas. For example, GABA, receptors were found to be in the highest concentration relative to GABAB receptors in areas such as frontal cortex, hippocampus, subiculum, amygdala, septum, the external plexiform layer of the olfactory bulb, and the granule cell layer of the cerebellum. Chu et al. (1990) reported that in these regions GABA binding to GABAA receptors accounted for about 70430% of total GABA binding. In contrast, in a few areas, including the molecular layer of the cerebellum, the interpeduncular nucleus of the brainstem, and certain thalamic nuclei, GABAB receptors were in higher concentration and in these areas GABAB sites could account for up to 90% of the total GABA binding. Finally, in other regions, such as the superficial gray of the superior colliculus and certain thalamic nuclei, the concentration of both GABAB and GABAAbinding sites was similar. The regions of the brain that exhibited the highest absolute concentrations of GABABbinding were frontal cortex, interpeduncular nucleus, the glomerular layer of the olfactory bulb, the superficial gray of the superior colliculus, and the molecular layer of the cerebellum. Intermediate levels of GABABbinding were found in the entorhinal cortex, molecular layer of the hippocampal dentate gyrus, amygdala, granule cell layer of the cerebellum, and various thalamic nuclei. Low levels of GABAB receptors were observed in hippocampus, subiculum, substantia nigra, hypothalamus, various thalamic nuclei, neostriatum, and dorsal raphe nucleus. Although the hippocampal formation contains a low level of GABAB receptors, the highest concentration of binding sites in this area is observed in the outer two-thirds of the molecular layer of the dentate gyrus, with lower levels of binding in the region of the granule cell layer. In the hippocampus GABAB sites are distributed throughout areas CA 1-CA4, with GABAB receptors being more highly concentrated in the dendritic than somatic layers (Bowery et al., 1985a, 1987; Chu et al., 1990). This result agrees well with the observation that GABAB receptor-mediated potentials are most readily produced by application of GABA in the dendritic layer (Newberry and Nicoll, 1985) (see Section III,C,5). Siniilarly, in cerebellum GABAB binding sites appear to be located almost entirely in the molecular layer (Wilkin et al., 1981). However, it should be noted that because of the small diameter of dendrites, the surface area of membrane in the dendritic layer is greater than that in the cell body layer. Thus, it is possible that even if the cell had an equal distribution of GABAB receptors over its entire surface, the receptor density determined by autoradiography may appear greater in the dendritic layer.

CENTRAL GABA, RECEPTORS

109

Although the above studies suggest that postsynaptic GABAB receptors may be preferentially located on dendrites in certain areas of the brain, receptor autoradiography coupled with selective lesions has been used to determine further the subcellular location of GABAB receptors. These studies have demonstrated that GABAB receptors can be located on nerve terminals as well as at postsynaptic sites. For example, electrolytic lesion of the habenular nucleus caused a 90% reduction in GABAB receptor binding in the interpeduncular nucleus. These results indicate that the majority of the GABAB receptors in the interpeduncular nucleus are on terminals of afferent fibers arising from the habenula (Price et al., 1984). In contrast, Bowery el al. (1985a) demonstrated that lesion of the perforant path, the primary afferent projection to the molecular layer of hippocampal dentate gyrus, caused no change in GABAB receptor binding in the molecular layer. They concluded that GABAB binding sites in the molecular layer were postsynaptic. This finding agrees with the observation that baclofen has only a minimal effect on excitatory transmission in the lateral perforant pathway (Lanthorn and Cotman, 1981; Ault and Nadler, 1982). However, it does not rule out the presence of GABAB sites on terminals of inhibitory interneurons. T h e presence of both presynaptic and postsynaptic GABAB receptors has recently been confirmed using monoclonal antibodies against the ( - )isomer of baclofen to visualize directly GABA, binding sites. Holstein et al. (1992) observed postsynaptic immunoreactivity on dendrites in the medial vestibular nucleus, indicating the presence of postsynaptic GABA, receptors. The immunostained dendrites received synapses from presumed GABAergic terminal. In addition, the authors observed immunostained terminals that synapsed onto both dendrites and axons. Similarly, Martinelli et al. ( 1992) reported postsynaptic immunoreactivity in both the molecular layer of the cerebellum and the substantia nigra, indicating the presence of postsynaptic GABAB receptors in these areas. These results suggest a morphological basis for both presynaptic and postsynaptic inhibition by GABAB receptors. It should be kept in mind that studies of GABA, receptor binding using labeled agonists are subject to a number of technical limitations. For example, because guanyl nucleotides lower the affinity of GABAB receptors for agonist, labeled agonist binding could be affected by any endogenous GTP remaining in the tissue. Similarly, endogenous GABA present in the tissue may be able to displace the labeled ligand. Thus, if different brain regions have different amounts of endogenous G T P or GABA, the ligand could bind with different kinetics in these different regions. These limitations may explain some of the variability observed between GABAB receptor binding in these studies. Perhaps, labeled antagonist binding might eliminate some of these problems.

110

DAVID D. MOTT A N D DARRELL V. LEWIS

C . GABAB RECEPTOR-EFFECTOR SYSTEMS

GABAB receptors are coupled to a variety of intracellular effector systems. These effector mechanisms are the means by which GABA through GABAB receptors can transduce messages into the neuron. The effector systems so far described include (a)inhibition of adenylyl cyclase, (b) facilitation of the transmitter-mediated activation of adenylyl cyclase, (c) inhibition of agonist-induced inositol triphosphate (IP,) synthesis, (d ) inactivation of voltage-dependent calcium channels, and ( e ) activation of potassium channels. We will discuss each of these actions below. 1. Inhibition of Adenylyl Cyclase The observation that guanyl nucleotides reduce the affinity of GABA, receptors for GABA (see Section II1,A) suggested that these receptors played a role in regulating adenylyl cyclase activity. Indeed, both ( - )baclofen and GABA were found to inhibit basal adenylyl cyclase activity as measured by the conversion of [32P]ATPto cyclic [32P]AMP (Wojcik and Neff, 1984; Nishikawa and Kuriyama, 1989). Similarly, if forskolin was first used to ensure activation of adenylyl cyclase, GABAB receptor activation resulted in a greater reduction in cAMP levels. Inhibition of basal adenylyl cyclase activity was observed in cultured cerebellar granule cells (Xu and Wojcik, 1986) as well as membrane preparations from a variety of different brain regions (Wojcik and Neff, 1984). Inhibition of adenylyl cyclase was strongest in the striatum, followed by cerebellum, cerebral cortex, thalamus, hippocampus, and hypothalamus. In contrast, studies using slices of cerebral cortex or cerebellum and measuring cAMP accumulation within the tissue found that whereas GABA, receptor activation by ( - )baclofen inhibited forskolin-stimulated adenylyl cyclase activity, it failed to alter basal cAMP levels (Hill et al., 1984; Hill, 1985; Karbon and Enna, 1985). Inhibition of basal adenylyl cyclase activity in cerebellar membranes is produced by both ( - )badofen and GABA with half-maximal concentrations (EC,,) of 4 and 17 p M , respectively. As was reported for baclofen binding, the (-)isomer of the drug is the active form, being approximately 1000 times more potent than the (+)isomer at inhibiting adenylyl cyclase activity. In support of the role of GABAB receptors, the GABAB receptor antagonist phaclofen blocked the inhibition of adenylyl cyclase activity produced by (-)baclofen (Nishikawa and Kuriyama, 1989). In contrast, the GABA, receptor agonist isoguvacine did not alter adenylyl cyclase activity. Similarly, neither bicuculline methiodide, a GABA, receptor antagonist, nor diazepam, a benzodiazepine, modified the inhibition of adenylyl cyclase activity produced by GABA or ( - )badofen.

CENTRAL GABA, RECEPTORS

111

Finally, since the effects of agonists at GABAB and adenosine A, receptors were not additive, Wojcik et al. (1985) concluded that these two receptors share a common catalytic subunit on adenylyl cyclase. Adenosine A, receptors are known to couple to adenylyl cyclase through the inhibitory G protein G, (Dunwiddie, 1985); therefore, this finding suggested that GABAB receptors may do the same. Receptor coupling to adenylyl cyclase is accomplished through one of the many types of G proteins. These G proteins are composed of three subunits, a, p, and y, the a subunit of which contains an intrinsic GTPase. The major events in this transduction pathway are as follows (for review see Neer and Clapham, 1988; Sternweis and Pang, 1990): agonist binding to the receptor promotes the interaction of the receptor with G protein, catalyzing the exchange of guanosine 5'-diphosphate (GDP), which is normally bound to the a subunit, for GTP. This exchange of GTP for GDP promotes dissociation of the G protein from the receptor, which reverts back to its low-affinity conformation. The liberated G protein then separates into a and Py subunits. The a subunit can activate intracellular effector mechanisms. The By subunit may indirectly contribute to the inhibition by binding to free a subunits from stirnulatory G proteins to form the inactive heterotrimer. Hydrolysis of GTP back to GDP by the GTPase of the a subunit promotes dissociation of the a subunit from the adenylyl cyclase and reassociation with the Py subunit, thereby terminating the signal. GABABreceptors have been shown to be coupled to adenylyl cyclase through the inhibitory G proteins Gi and G, (Asano et al., 1985; Ohmori et al., 1990). T h e coupling of inhibitory G proteins with GABAB receptors was first reported by Asano et al. (1985),who found that binding of GABA to GABAB receptors in cortical membranes was decreased following incubation with pertussis toxin. Pertussis toxin selectively inactivates the inhibitory G proteins Gi and G, but not the stimulatory G protein, G,. The toxin catalyzes the transfer of adenosine 5'-diphosphate (ADP)ribose from nicotinamide adenine dinucleotide (NAD) to the a subunit of Gi and G,, thereby preventing association of these G proteins with the receptor, and thus causing the receptor to remain in the low-affinity conformation. Likewise, the reduction of GABAB receptor binding produced by pertussis toxin was the result of reduced receptor affinity, not receptor number, indicating that more of the receptors were in the low affinity conformation. As predicted, the subsequent addition of purified Gi/Goproteins restored high-affinity binding of GABA to GABAB receptors. Closer examination of the subtype of G proteins(s) necessary to restore high-affinity binding revealed that addition of Go, GZ or Gil, but not Gi2,was sufficient (Morishita et al., 1990). In addition, Go remained

112

DAVID D. MOTT A N D DARRELL V. LEWIS

effective, and Gi2 remained ineffective, even when the Py subunit complex was altered. These results indicate that coupling of the G protein to GABAB receptors is determined primarily by the (Y subunit. Furthermore, the ability of GABAB receptors to couple to different types of G proteins suggests that GABAB receptors may use different G protein subtypes to regulate separately the effectors to which they are coupled. Further evidence that GABAB receptors were coupled to G proteins was provided by Hill et al. (1989) and Bowery et al. (1989), who reported that (-)badofen and GABA increased the specific GTPase activity in rat brain membranes preparations with the largest increase occurring in areas known to contain high concentrations of GABAB receptors. This increase in GTPase activity was attributed to the receptor-mediated exchange of GDP for GTP on the (Y subunit of the G protein and the subsequent activation of the intrinsic GTPase activity of this subunit. This effect was reduced by prior exposure of the membranes to pertussis toxin, indicating that it was produced by an inhibitory G protein. Although these findings indicate that GABAB receptors are coupled to inhibitory G proteins, they d o not directly demonstrate a link between the GABAB receptor-coupled G protein and adenylyl cyclase. This link was first reported in cultured cerebellar granule neurons by Xu and Wojcik (1986) and in membranes of cerebral cortex by Nishikawa and Kuriyama (1989). These studies reported that pertussis toxin blocked the inhibitory effect of baclofen and GABA on forskolin-stimulated adenylyl cyclase activity. These results suggest that both the inhibition of forskolinstimulated adenylyl cyclase activity and the increase in GTPase activity are mediated by the (Y subunits of the same inhibitory G proteins that regulate GABAB receptor binding.

2 . Facilitation of Transmitter-Mediated Activation of Adenylyl Cyclase In contrast to its direct inhibition of adenylyl cyclase through Gi proteins, GABAB receptor activation also potentiates the accumulation of cAMP produced by other receptors (Karbon et al., 1984; Karbon and Enna, 1985; Hill, 1985; Watlingand Bristow, 1986, for review see Karbon and Enna, 1989). For example, addition of baclofen to the incubation medium markedly enhanced the increase in cAMP accumulation produced by norepinephrine (Karbon et al., 1984; Hill, 1985). Similarly, GABA or baclofen potentiated the increase in adenylyl cyclase activity caused by other G,-coupled neurotransmitters, such as adenosine A,, histamine, and vasoactive intestinal peptide (VIP) receptors (Karbon and Enna, 1985; Watling and Bristow, 1986; Scherer et al., 1988b; Schaad et al., 1989). In each case, baclofen caused a two- to three-fold increase in the accumulation of CAMP produced by the neurotransmitter. Potenti-

CENTRAL GABA, RECEPTORS

113

ation of the receptor-mediated increase in cAMP accumulation by baclofen is similar to the effects of a-adrenergic agonists on the increase in cAMP levels produced by P-adrenergic agonists (Daly et al., 1980; LeBlanc and Ciaranello, 1984). Interestingly, potentiation of CAMP accumulation has only been observed following receptor-mediated activation of adenylyl cyclase, not direct activation of adenylyl cyclase with forskolin. As discussed previously, following direct activation of adenylyl cyclase with forskolin, GABAB receptor activation inhibits cAMP formation (Karbon and Enna, 1985). Potentiation of transmitter-mediated cAMP accumulation was observed in slices from cerebral cortex, hippocampus, and corpus striatum, but not cerebellum (Karbon and Enna, 1985). The effect is thought to be GABAB receptor mediated because both baclofen and GABA are equally efficacious, with a concentration of about 3 P M (?)baclofen necessary to yield a half-maximal potentiations. In addition, potentiation by baclofen was stereoselective with the (-)isomer more potent that the (+)isomer. The effect was not sensitive to GABA, receptor antagonists, such as bicuculline methiodide (Karbon et al., 1984). Finally, except for the cerebellum, the intensity of the response in different brain regions paralleled the density of GABA, binding sites in these areas (Karbon and Enna, 1985; Bowery et al., 1987). In the cerebellum GABAB receptor activation failed to potentiate the CAMP response to norepinephrine, despite the high density of GABABbinding sites in this area. This discrepancy may be caused by the existence of subtypes of the GABAB receptor, only one of which is coupled to the augmenting response. Alternatively, it may be caused by multiple forms of adenylyl cyclase that are modulated differently by G protein interaction (Tang and Gilman, 1992). T h e mechanism by which GABAB receptors potentiate receptormediated adenylyl cyclase activity has not been conclusively demonstrated. It appears unlikely that baclofen interacts with the G,-linked receptor recognition site, since baclofen is able to augment the cAMP response to a variety of different G,-linked receptors. Further studies confirmed this for the P-adrenergic system by demonstrating that baclofen did not alter the affinity or number of P-adrenoceptors (Karbon and Enna, 1989). One possible mechanism by which baclofen could potentiate the effect of G,-linked receptors of cAMP levels is by synergizing the interaction of adenylyl cyclase with the stimulatory G protein, C, (Bourne and Nicoll, 1993). Like its inhibitory counterparts, G, is a heterotrimeric protein composed of an a,a p, and a y subunit. In most cases, receptor-mediated stimulation of adenylyl cyclase is mediated solely by the a subunit of the G, protein (Neer and Clapham, 1988; Sternweis and Pang, 1990).

114

DAVID D. M O T T AND DARRELL V. LEWIS

However, under certain conditions, Py subunits may also carry a signal. In light of this, multiple subtypes of adenylyl cyclase, which respond differently to the presence of Py subunits, have been identified (Tang and Gilman, 1992). Adenylyl cyclase type I, when activated by a,, can be inhibited by subsequent interaction with Py subunits. In contrast, adenylyl cyclase type 11, when stimulated by the a, subunit is potentiated up to five-fold by the subsequent binding of Py subunits (Tang and Gilman, 1991; Federman et al., 1992). Because of the similarity between their Py subunits, either G, or Gi,, proteins could provide these subunits. Thus, although ail,liberated by GABAB receptor-activation of Gi/, proteins could cause direct inhibition of adenylate cyclase, Py subunits liberated by this same interaction could synergize the effect of a, on adenylyl cyclase type 11, resulting in a net increase in cAMP in the neuron. Studies have shown that both type I and type 11 adenylyl cyclase are expressed in several different brain regions, including hippocampus (Gannon and McEwen, 1992; Mons et al., 1992), suggesting that this mechanism may occur in central neurons (Andrade, 1993) (see section IV). The potentiation of receptor-mediated adenylyl cyclase activity produced by baclofen is observed only in slice preparations (Karbon et al., 1984; Hill, 1985; Karbon and Enna, 1985). In contrast, in preparations of broken cells or membranes from these same brain regions, baclofen inhibits adenylyl cyclase activity (Wojcik and Neff, 1984; Xu and Wojcik, 1986; Nishikawa and Kuriyarna, 1989) (see Section III,C,I). One possible explanation for this difference is that the GABAB receptor-mediated potentiation of cAMP accumulation is dependent on cytosolic factors, such as phospholipase A, and arachidonic acid, which are not present in the membrane preparations. Phospholipase A, is a calcium-activated enzyme that catalyzes the formation of arachidonic acid from membrane phospholipids. Arachidonic acid is then metabolized into a number of substances, including prostaglandins and leukotrienes (Piomelli and Greengard, 1990; Shimizu and Wolfe, 1990). Potentially, either arachidonic acid or its metabolites could interact with the adenylyl cyclase system to increase cAMP levels; however, the actual molecular events by which this interaction may occur remain to be determined. In support of a role for phospholipase A, in the potentiating response of baclofen, Duman et al. (1986) found that nonspecific inhibitors of phospholipase A,, such as quinacrine and corticosterone, did not affect the potentiation of cAMP levels produced by isoproterenol alone; however, they prevented baclofen from augmenting isoproterenol-mediated cAMP accumulation. Furthermore, they found that inhibitors of arachidonic acid metabolism had no effect on the augmentation by baclofen, indicating that arachidonic acid itself mediates the potentiation. These results sug-

CENTRAL GABA, RECEPTORS

115

gest that phospholipase A, plays a role in the augmenting response of GABA, receptor activation on G,-linked neurotransmitter stimulation of adenylyl cyclase. However, because the phospholipase A, inhibitors used in this study were not very specific, additional evidence is necessary to demonstrate conclusively a role for this enzyme.

3 . Inhibition of Agonist-Induced Inositol Trifihosphate (IP,) Synthesis Inositol phospholipid breakdown represents another intracellular signaling pathway that can be modulated by GABAB receptor activation. T h e mechanism of this pathway is as follows (for review see Berridge and Irvine, 1989): Binding of agonist to the receptor triggers the activation of phospholipase C through a G protein. This enzyme catalyzes the cleavage of phosphatidylinositol-4,5-bisphosphate (PIP,) resulting in the formation of diacylglycerol (DAG) and inositol l ,4,5-triphosphate (l,4,5.JP,). Both DAG and 1,4,5-IP3can then act as second messengers in the cell. 1,4,5-IP3 mobilizes calcium from intracellular stores, whereas DAG activates protein kinase C (PKC), a calcium-activated, phospholipiddependent enzyme. GABAB receptor activation does not affect basal [3H]inositol phosphate formation in cerebral cortex slices from rat or mouse (Crawford and Young, 1988, 1990; Godfrey et al., 1988,but see Hahner et al., 1991). However, in cerebral cortex both GABA and baclofen markedly inhibit the increase in [3H]inositol phosphate produced by histamine [H, receptors (Crawford and Young, 1988)] or serotonin [5-HT, receptors (Godfrey et al., 1988)] in a noncompetitive manner. Baclofen is effective at low concentrations and is stereospecific with the (-)isomer being markedly more potent than the (+)isomer. In addition, the GABA, agonist isoguvacine does not affect the histamine-induced increase in [3H]inositol phosphate formation (Crawford and Young, 1988).Similarly, bicuculline methiodide does not prevent the GABA-induced depression of the 5H T response (Godfrey et al., 1988). Taken together, these observations suggest that GABA acting on GABAB receptors may modulate both histamine and serotonin function. Unfortunately, the molecular mechanism through which GABA, receptors may modulate these receptors is currently unknown. 4. Inhibition of Voltage-Sensitive Calcium Channels In central neurons calcium influx through voltage-sensitive calcium channels has been found to be sensitive to a variety of dif€erent neurotransmitters, including norepinephrine, acetylcholine, adenosine, somatostatin, and opioid peptides (Carbone and Swandulla, 1989; Anwyl, 1991). Another neurotransmitter reported to inhibit calcium influx is

116

DAVID D. MOTT AND DARRELL V. LEWIS

GABA, acting through GABAB receptors. Electrophysiological studies reported that baclofen depressed excitatory postsynaptic potentials (EPSPs) in hippocampal neurons, possibly by inhibiting presynaptic calcium currents and thereby depressing transmitter release (Lanthorn and Cotman, 1981; Ault and Nadler, 1982). In support of this possibility, GABA and baclofen were reported to reduce the decrease in extracellular calcium concentration that occurred during repetitive stimulation in hippocampal area CA 1, presumedly by blocking calcium entry into presynaptic terminals of pyramidal neurons (Konnerth and Heinemann, 1983; Heinemann et al., 1984). Inhibition of calcium influx at presynaptic terminals was also suggested by the observation that baclofen depressed the K+-evoked release of transmitter from synaptosomes (Bonanno et al., 1989b). Similarly, direct measurements of calcium entry into cortical (Stirling et al., 1989) or cerebellar (Bowery and Williams, 1986) synaptosomes revealed that baclofen depressed calcium influx. Although these studies demonstrate that GABAB receptor activation can depress calcium entry into presynaptic terminals of central neurons, they do not reveal the mechanism of this action. Unfortunately, detailed study of calcium channels on presynaptic terminals is hampered by their electrophysiological inaccessibility. To overcome this problem voltage-sensitive calcium channels on neuronal cell bodies have been used as an experimental model in most cases. In particular, numerous studies have focused on sensory neurons in the dorsal root ganglion (DRG) because they are readily accessible and exhibit relatively large somatic calcium currents (Dunlap and Fischbach, 1978). In support of the usefulness of this model, it has been reported that many neurotransmitters, which depress calcium entry into central neurons, also decrease calcium current in somata of DRG neurons (Carbone and Swandulla, 1989; Anwyl, 1991). Examination of voltage-sensitive calcium currents in DRG neurons demonstrated that these currents were inhibited by GABA, receptor activation. The inhibition was first reported as a decrease in the calciumdependent plateau phase of the action potential in response to GABA or baclofen (Dunlap and Fischbach, 1978, 1981; Dunlap, 1981) with no change in the resting membrane potential. Subsequent studies using voltage clamp (Dunlap and Fischbach, 1981) o r whole cell patch clamp techniques (Deisz and Lux, 1985; Dolphin and Scott, 1986; Robertson and Taylor, 1986) confirmed that this apparent reduction in calcium current was not caused by an increase in an outward potassium current, but rather by a direct suppression of voltage-sensitive calcium channels. Several lines of evidence indicate that this effect was produced by GABAB receptor activation. For example, the effect of baclofen was stereospecific, with (-)baclofen being the active isomer (Dolphin and Scott, 1986; Rob-

CENTRAL GABA, RECEPTORS

117

ertson and Taylor, 1986). T h e action of baclofen was blocked by phaclofen (Huston et al., 1990), but not bicuculline (Dunlap, 1981), and it was not mimicked by isoguvacine (Desarmenien et al., 1984). Multiple types of voltage-sensitive calcium channel have been shown to exist in neuronal membranes (for review see Nowycky et al., 1985; Miller and Fox, 1990; Bertolino and Llinas, 1992). In DRG neurons at least three classes of calcium channels have been reported: a low voltageactivated (LVA) T-type channel, a high voltage-activated (HVA) N-type channel, and a HVA L-type channel (Nowycky et al., 1985). It has been proposed that these different classes of voltage-sensitive calcium channel play different roles in neuronal function (for review see Bertolino and Llinas, 1992). For example, the N-type channel has been implicated in the control of neurotransmitter release, whereas the T-type channel has been suggested to control neuronal oscillatory activity, including spontaneous repetitive and burst firing. L-type channels are thought to be involved in the generation of action potentials and signal transduction. Because of the different physiological effects of these channels, it is of interest to determine which channel type is modulated by GABAB receptor activation. Unfortunately, the evidence indicating an effect of baclofen on a specific subtype of calcium channel in DRG neurons is inconclusive. Studies have found that baclofen will depress the LVA T-type channel by amounts ranging from 25 to 55% (Deisz and Lux, 1985; Dolphin et al., 1990; Formenti and Sansone, 1991). In addition, baclofen has been reported to slow the activation phase of the HVA calcium current and reduce the amplitude of both the HVA N-type calcium channel and the HVA L-type calcium channel, with the former channel being more sensitive to the drug (Dolphin and Scott, 1987; Green and Cottrell, 1988; Huston et al., 1990). Recently, however, Tatebayashi and Ogata (1992) reported a selective effect of baclofen on the HVA N-type calcium current with no effect of the drug on either the HVA L-type channel or the LVA T-type channel. They suggested that this preferential depression of the HVA inactivating N-type calcium current, which forms the rapid rising phase of the total HVA current, is responsible for the slowing of the activation phase of the total HVA calcium current. This result is of interest in light of the proposed role of N-type channels in regulating transmitter release. T h e mechanism by which GABAB receptor activation inhibits voltagesensitive calcium currents in DRG neurons involves inhibitory G proteins (Dolphin et al., 1990). The role of G proteins was demonstrated in studies (GDP-P-S),a GDP anashowing that guanosine 5 -0-(2-thiodiphosphate) logue that competitively inhibits the binding of GTP to G proteins, reduced the effect of baclofen on voltage-sensitive calcium currents.

118

DAVID D. MOTT A N D DARRELL V. LEWIS

In contrast, guanosine 5 -0-(3-thiotriphosphate)(GTP-y-S), a nonhydrolyzable GTP analogue that irreversibly activates G proteins, enhanced the effect of baclofen and, if applied alone, mimicked the inhibitory effect of baclofen on the HVA N-type calcium current (Scott and Dolphin, 1986; Dolphin and Scott, 1987; Holz et al., 1989). Finally, it was observed that preincubation of DRG neurons with pertussis toxin blocked the inhibitory effect of baclofen on calcium influx, indicating that GABAB receptor activation inhibited voltage-sensitive calcium currents through one of the inhibitory G proteins, either Gi or Go (Dolphin and Scott, 1987; Holz et al., 1989). Using antibodies raised against the C terminus of the a subunit of Go and Gi proteins, Menon-Johansson and Dolphin (1992) found that the a subunit of Go, but not Gi, protein was responsible for the inhibition of voltage-sensitive calcium currents by baclofen. Similarly, the a subunit of the Go protein has been found to mediate inhibition of voltage-sensitive calcium channels by a variety of different neurotransmitters, including neuropeptide Y, (Ewald et al., 1988), opioid peptides (Hescheler et al., 1987), somatostatin (Kleuss et al., 1991), dopamine (Harris-Warrick et al., 1988), acetylcholine (Toselli et al,, 2989; Offermanns et al., 1991), and norepinephrine (McFadzean et al., 1989), suggesting an essential regulatory role for this G protein in calcium channel inhibition. It is not yet clear what role, if any, the dimer plays in the inhibition. Although it has been demonstrated that inhibitory G proteins are required for GABAB receptor-mediated inhibition of voltage-dependent calcium channels, it remains unclear whether these G proteins interact directly with the channel or whether another second messenger is involved. Several lines of evidence suggest the G protein may interact directly with the calcium channel (for review see Dolphin, 1991a). For example, using cell attached patches, Green and Cottrell (1988) found that baclofen inside the patch pipette was able to inhibit calcium channels, whereas external baclofen was not. These authors concluded that a second messenger capable of diffusing to calcium channels under the patch was not involved in the inhibition. A similar conclusion was reached by Dolphin et al. (1989), based on their inability to link the inhibition of voltage-dependent calcium channels in DRG neurons to any of the transduction systems known to be coupled to GABA, receptors. In contrast, in cultured spinal cord neurons Karnatchi and Ticku (1990) reported that inhibition of calcium currents by baclofen was antagonized by either activation of adenylyl cyclase by forskolin or the addition of 8-bromoCAMP.In addition, these authors found that activation of protein kinase C with phorbol esters also antagonized the action of baclofen. They concluded that both protein kinase A and C contribute to the inhibitory

CENTRAL GABA, RECEPTORS

119

effect of baclofen on calcium currents in spinal cord neurons. A similar depression of calcium currents by phorbol esters or diacylglycerol has also been reported in other studies using DRG neurons (Rane and Dunlap, 1986; Gross and Macdonald, 1988). Although these results demonstrate that PKC activation mimics the effect of baclofen, they do not necessarily demonstrate that baclofen acts through this transduction mechanism to inhibit calcium currents. Furthermore, other studies have reported phorbol esters either to have no effect or to activate calcium currents (for review see Kaczmarek, 1987). Thus, the mechanism by which G proteins regulate calcium channel activity remains to be conclusively demonstrated. Interestingly, a recent study using DRG neurons has suggested that a continuous cycle of phosphorylation and dephosphorylation may tonically regulate activation of voltage-sensitive calcium channels. Dolphin (1991b, 1992) reported that CAMP-dependent phosphorylation, produced by direct addition of CAMPor activation of adenylyl cyclase with forskolin, was able to reverse the inhibition of calcium currents produced by G protein activation with GTP-y-S. Similarly, inhibition of calcium currents was reduced by addition of the active fragment of phosphorylated inhibitor 1, which prevented dephosphorylation by inhibiting phosphatase 1. T h e author concludes that the interaction between the G protein and voltage-dependent calcium channels may be regulated by phosphorylation, suggesting that GABAB receptor-mediated inhibition of these channels would be sensitive to the phosphorylation state. The potential site of this CAMP-dependent phosphorylation includes the calcium channel itself, the phosphatase l inhibitor, or the Go protein. In support of an action on the G protein, it was recently reported that phosphorylation of G,, protein by protein kinase C blocks its ability to inhibit adenylyl cyclase (Bushfield et al., 1991). In contrast to the large number of reports from DRG neurons, few studies in central neurons have examined calcium current modulation by GABA, receptors. Initial studies of central neurons reported baclofen to have no direct effect on calcium currents (Gahwiler and Brown, 1985; Howe, 1987; Howe et al., 1987). However, recent studies in these neurons have reported inhibition of calcium currents by baclofen (Scholz and Miller, 1991; Swartz and Bean, 1992; Mintz and Bean, 1993). The lack of an effect of baclofen in the report by Gahwiler and Brown (1 985) is most likely explained by their use of small depolarizing steps (+ 20 mV) to elicit the current. Indeed, in recent reports in hippocampus (Scholz and Miller, 1991) and cerebellum (Wojcik et al., 1990) baclofen had little to no effect on calcium currents evoked by small depolarizing steps (< + 30 mV), while significantly inhibiting calcium currents evoked by

120

DAVID D. MOTT AND DARRELL V. LEWIS

larger depolarizations. This inhibition was blocked by 2-hydroxysaclofen, indicating that it was mediated by GABA, receptor activation (Scholz and Miller, 1991). As in DRG neurons, central neurons express multiple types of calcium current, including an LVA T-type calcium current and two types of HVA calcium current: an o-conotoxin-sensitive N-type current and a dihydropyridine-sensitive L-type current. In addition, central neurons also express an HVA P-type calcium current that is resistant to both oconotoxin and dihydropyridines, but is sensitive to funnel-web spider toxin (for review see Bertolino and Llinas, 1992). In hippocampal neurons, as in DRG neurons, baclofen primarily suppressed the HVA N-type current; however, some inhibition of the HVA L-type current was also noted (Scholz and Miller, 1991). In addition, baclofen depressed the LVA T-type calcium current in interneurons in stratum lacunosummoleculare (Fraser and MacVicar, 1991) (see Section IV,A,P,b). Interestingly, in area CA3 baclofen was typically much more effective at inhibiting calcium currents in pyramidal cells than in nonpyramidal neurons in this same region (Wojcik et al., 1990). In cerebellar granule cells baclofen was found to inhibit an HVA L-type current, although an effect on another current type was not ruled out (Huston et al., 1990; Wojcik et al., 1990; Marchetti et al., 1991). Alternatively, in cerebellar Purkinje cells both baclofen and GABA were reported to suppress an HVA P-type current (Mintz and Bean, 1993). The reason for these differences in the effect of baclofen is unclear; however, they may reflect differences between cell types, differences in the subtype of calcium channel mediating the current, or differences in the mechanism coupling the receptor to the calcium channels. As in DRG neurons, baclofen appears to inhibit calcium channels in central neurons through activation of G proteins. Thus, in hippocampal neurons inclusion of GTP-y-S in the patch pipette enhanced calcium channel inhibition by baclofen whereas pretreatment of the cells with pertussis toxin blocked the actions of baclofen, indicating that baclofen acted through an inhibitory G protein (Wojcik et at., 1990).Furthermore, Sweeney and Dolphin (1992), using cortical and cerebellar membrane preparations, have recently examined the subtype of G protein involved in the interaction by using antibodies directed against the C-terminus of the a subunit of Gi and Go proteins. They found that, although baclofen activates both G, and Go proteins, the L-type calcium channel interacts selectively with the C-terminus of the a subunit of the Go protein. They suggested that, as in DRG neurons, GABA, receptor-mediated activation of Go proteins leads to inhibition of L-type calcium channels, whereas activation of Gi proteins causes inhibition of adenylyl cyclase.

CENTKAL GABA, RECEPrOKS

121

In summary, baclofen inhibits voltage-sensitive calcium currents in both DRG neurons and central neurons through activation of inhibitory G proteins. Inhibition of L-type calcium currents appears to occur by direct interaction of the C-terminus of the (Y subunit of Go proteins with the calcium channels, although the contribution of another transduction system has not been conclusively eliminated. In both DRG and hippocampal neurons baclofen appears to inhibit N-type calcium channels most effectively, whereas it inhibits L-type calcium channels in cerebellar granule cells and P-type calcium channels in cerebellar Purkinje cells. Although it is tempting to speculate that inhibition of calcium currents is responsible for the GABA, receptor-mediated inhibition of transmitter release, it must be remembered that these results have been collected indirectly by using somatic calcium channels as a model of calcium channels on neuronal terminals. Thus, direct evidence relating the GABA, receptor-mediated inhibition of calcium currents on neuronal terminals with depression of transmitter release has not yet been reported.

5. Activation of Potassium Channels In addition to its affect on voltage-sensitive calcium channels, GABA, receptor activation by baclofen also hyperpolarizes and reduces the input resistance of central neurons. This effect was first observed in the hippocampal formation (Klee et al., 1981; Misgeld et al., 1982) and has since been recorded in a variety of other brain regions including hypothalamus (Ogata and Abe, 1982), neocortex (Howe et al., 1987), lateral septum (Stevens et al., 1985), dorsal raphe (Colmers and Williams, 1988), and locus coeruleus (Shefner anti Osmanovic, 1991). In contrast, baclofen appears to produce no hyperpolarization in the neostriatum (Calabresi et al., 1992; Nisenbaum et al., 1992), the supraoptic nucleus (Ogata, 1990b), or sensory neurons in the DRG (Dunlap and Fischbach, 1981). Focal application of baclofen was found to hyperpolarize consistently pyramidal neurons in the hippocampus; however, this response was much stronger when baclofen was applied dendritically than when it was applied near the soma, suggesting that baclofen may act primarily in the dendrites (Newberry and Nicoll, 1984b, 1985). The hyperpolarization produced by baclofen was shown to persist during blockade of synaptic transmission with tetrodotoxin or cadmium, indicating that this effects was postsynaptic (Newberry and Nicoll, 198413, Nicoll and Newberry, 1984). Furthermore, baclofen acted in a stereospecific fashion with the (-)isomer being 100-200 times more potent than that (+)isomer (Newberry and Nicoll, 1984b, 1985; Nicoll and Newberry, 1984; Padjen and Mitsoglou, 1990). The GABA, receptor agonist 3-APPA also hyperpolarized central neurons, indicating that this effect was not specific to

122

DAVID D. M O T T AND DARRELL V. LEWIS

baclofen (Seabrook et al., 1990; Lovinger et al., 1992). Finally, the GABA, receptor antagonists phaclofen (Dutar and Nicoll, 1988b; Olpe el al., 1988), 2-hydroxysaclofen (Hasuo and Gallagher, 1988; Lambert et al., 1989), CGP 35348 (Olpe et al., 1990), and CGP 55845 (Jarolimek et al., 1993) blocked the effect of baclofen. These results demonstrate that the hyperpolarization and conductance increase were produced through activation of GABA, receptors. The ability of baclofen to hyperpolarize and increase the conductance of central neurons suggested that GABA should produce a similar postsynaptic effect. It had been previously shown that focal application of GABA onto the somata of hippocampal pyramidal neurons produced a hyperpolarization, due to activation of a bicuculline-sensitive GABA, receptor-mediated chloride conductance. Furthermore, focal application of GABA onto the dendrites of these same neurons resulted primarily in a large GABA, receptor-mediated depolarization (Alger and Nicoll, 1982b). However, blockade of these GABA, receptor-mediated responses with bicuculline exposed an underlying hyperpolarization similar to that produced by baclofen (Newberry and Nicoll, 1984b, 1985). Like the baclofen response, this GABA-induced hyperpolarization was evoked more readily in the dendrites than near the soma and was insensitive to pentobarbitone, a barbiturate that enhances the GABA, response (Newberry and Nicoll, 1985). Furthermore, it was not additive with the baclofen response, suggesting that both the GABA and the baclofen responses shared the same effector channel (Nicoll et al., 1989). Thus, the hyperpolarizing nature of the GABA response, its insensitivity to GABA, antagonists and barbiturates, and its nonadditivity with the baclofen response suggested that it was mediated through the same GABAB receptor-coupled mechanism as the baclofen response. This conclusion was strengthened when it was demonstrated that, in addition to blocking the baclofen response, GABA, receptor antagonists, such as phaclofen (Dutar and Nicoll, 1988b), CGP 35348 (Solis and Nicoll, 1992a), or CGP 55845A (Jarolimek et al., 1993), also blocked the postsynaptic action of GABA. However, despite the many similarities between the responses to baclofen and GABA, some pharmacological differences have been reported (Ogata et al., 1987; Dutar and Nicoll, 1988a; Muller and Misgeld, 1989; Segal, 1990a, but see Solis and Nicoll, 1992a). In particular, the available GABAB receptor antagonists have been found to be more effective at blocking the postsynaptic response to baclofen than the response to GABA. This difference leaves open the possibility that, whereas baclofen may be specific for a single GABA, receptor subtype, GABA may activate more than one GABA, receptor subtype with differing antagonist pharmacology (Johnston, 1986; Bowery, 1993).

CENTRAL GABA, RECEPTORS

123

Several lines of evidence indicate that GABAB receptor activation hyperpolarizes neurons by increasing the conductance of the membrane to potassium ions. First, the bicuculline-insensitive dendritic response to both GABA and baclofen reversed at a hyperpolarized membrane potential, close to the equilibrium potential for potassium (Nicoll and Newberry, 1984; Gahwiler and Brown, 1985; Inoue et al., 1985c; Newberry and Nicoll, 1985). This reversal potential was similar to that of the postburst slow afterhyperpolarization [I,,, (Newberry and Nicoll, 1984b, 1985; Nicoll and Newberry, 1984; Inoue et al., 1985a)], which has been shown to be mediated by a calcium-activated potassium conductance (Alger and Nicoll, 1980; Schwartzkroin and Stafstrom, 1980). Second, altering the gradient of chloride ions across the membrane did not change the reversal potential of the baclofen response, as would he expected if it were mediated by an increase in chloride conductance, like the GABA, response (Newberry and Nicoll, 1984b, 1985; Nicoll and However, increasing the extracelluNewberry, 1984; Inoue et al.. 1985~). lar concentration of potassium did produce a depolarizing shift in the reversal potential of both the baclofen and the GABA responses by an amount close to that predicted by the Nernst equation (Nicoll and Newberry, 1984; Gahwiler and Brown, 1985; Inoue et al., 1985a; Newberry and Nicoll, 1985, but see Jarolimek and Misgeld, 1992).Third, external barium (Nicoll and Newberry, 1984; Gahwiler and Brown, 1985; Newberry and Nicoll, 1985; Connors et al., 1988) or internal cesium (Gahwiler and Brown, 1985), both of which block potassium channels, reduced the hyperpolarizing response. Finally, tetrahydroaminoacridine (THA), a potassium channel blocker, suppressed both the GABA- and the baclofen-induced conductance increase (Halliwell and Grove, 1989; Dutar et al., 1990; Lambert and Wilson, 1993). Taken together, these results demonstrate the GABAB receptor activation produces postsynaptic hyperpolarization by increasing the potassium conductance of the membrane. As with calcium channels, the potassium conductance activated by GABABreceptors is similar to that activated by a variety of other neurotransmitters including dopamine, serotonin, opiates, adenosine, and somatostatin (Andrade et ul., 1985; North et ul., 1987; Christie and North, 1988,Brown, 1990). In general, this conductance shows inward rectification such that the current becomes smaller on membrane depolarization (Gahwiler and Brown, 1985; Newberry and Nicoll, 1985, but see Premkumar et al., 1990a; Wagner and Dekin, 1993). This voltage dependence differentiates it from an M-type potassium current. T h e current is insensitive to potassium channel blockers such as tetraethylammonium chloride [TEA (Stevens et al., 1985)] and 4-aminopyridine [4-AP (Padjen and

124

DAVID D. MOTT AND DARRELL V. LEWIS

Mitsoglou, 1990; Solis and Nicoll, 1992a, but see Inoue et al., 1985a; Stevens et al., 1985)], but is blocked by the lidocaine derivative, lidocaine N-ethyl bromide [QX-314 (Nathan et al., 1990a; Andrade, 1991)l. In addition, several lines of evidence suggest that the GABAB receptoractivated potassium current is not dependent on calcium. First, the current is not blocked by inorganic calcium channel blockers, such as cadmium or cobalt, making it unlikely that entry of extracellular calcium is necessary (Gahwiler and Brown, 1985; Newberry and Nicoll, 1985; Stevens et al., 1985; Padjen and Mitsoglou, 1990, but see Blaxter et al., 1986). Second, intracellular injection of EGTA, a calcium chelator, fails to block the current, indicating that a rise in intracellular calcium, presumedly from intracellular stores, is also unnecessary (Thalmann, 1984; Andrade et al., 1986, but see Blaxter et al., 1986; Segal, 1990b). Finally, single-channel studies on cell-attached membrane patches from cultured hippocampal neurons have found that removal of calcium from the perfusion medium has no effect on potassium channel activity induced by baclofen or GABA (Premkumar et al., 1990a). Single-channel analysis of GABA, receptor-operated potassium channels has only recently been performed. Using cell attached patches from cultured hippocampal neurons, Premkumar et al. (1990a) found that 30- to 90-s exposure of the neuron to baclofen or GABA caused single-channel potassium currents to appear, which were sensitive to 2hydroxysaclofen, but not bicuculline. Single-channel current amplitudes varied considerably with the smallest having an amplitude of about 0.36 PA, which corresponded to a conductance of 5-6 pS. These channels were potassium-selective (P,,/P, ratio of 0.03-0.04) and opened in bursts. During these bursts, the channels flickered rapidly between subconductance states that were integral multiples of 5-6 pS. The authors concluded that GABAB receptor activation causes the opening of a single class of potassium channels that can be coupled together and open cooperatively, each cochannel having an “elementary” conductance of 5-6 pS. This conductance is three to four times smaller than the singlechannel conductance of 17-20 pS reported for GABA, receptormediated chloride channels (Bormann, 1988). GABA, receptors are indirectly coupled to potassium channels through G proteins. This conclusion is based on the observation that the hydrolysis-resistant GDP analogue, GDP-/3-S, reduced the action of baclofen, whereas the GTP analogue, GTP-y-S, mimicked the effect of baclofen (Andrade et al., 1986; Thalmann, 1988). In addition, pretreatment with pertussis toxin blocked the action of both baclofen and GABA, indicating that the coupling of GABA, receptors to potassium channels is achieved through the inhibitory G proteins, G, and/or Gi (Andrade

CENTRAL GABA, RECEPTORS

125

et al., 1986; Colmers and Williams, 1988; Thalmann, 1988).Using excised

inside-out membrane patches from cultured hippocampal neurons, VanDongen et al. (1988) found that exposure of the membrane to purified or recombinant Go, but not Gi, would activate four types of potassium channels, one of which was also activated by serotonin. Since serotonin and GABAB receptors are coupled to the same G protein (Andrade et al., 1986),these results suggest that potassium channels are also coupled to GABAB receptors by Go protein. Despite the ability of Go proteins to activate directly neuronal potassium channels (VanDongen et d., 1988), it has not yet been conclusively demonstrated whether the Go protein-mediated coupling of GABAB receptors to neuronal potassium channels is direct or is mediated through a diffusible second messenger. A role for cAMP in this interaction has been ruled out based on the inability of 8-bromo-cAMP, a membrane-permeant cAMP analogue, or intracellular injection of cAMP to affect the baclofen-induced hyperpolarization (Andrade et al., 1986; Inniset al., 1988). It is also unlikely that phospholipase C plays a role in the GABAB receptor-mediated postsynaptic response because neither calcium (see above) nor protein kinase C activation (Andrade et al., 1986; Dutar and Nicoll, 1988a) is necessary for expression of the current. However, a role for phospholipase A, has been proposed. Premkumar et al. (1990b) found that application of arachidonic acid to the inner surface of excised inside-out membrane patches from cultured hippocampal neurons mimicked the effect of baclofen on single potassium channel currents. Fluctuation analysis revealed that the potassium channels activated are similar to those stimulated by baclofen. However, these results indicate only that arachidonic acid is capable of mimicking the effect of baclofen, not that it mediates this response in the normal cell. Thus, a role for arachidonic acid has yet to be conclusively demonstrated. Interestingly, inhibitors of PLA, do not block the EPSP depression produced by baclofen in hippocampal slices (Dunwiddie et d., 1990), suggesting that if presynaptic GABAB receptor-operated potassium channels play a role in the transmitter release process, they may not be modulated by arachidonic acid. GABAB receptors may also be coupled to other types of potassium channels. For example, Saint et al. (1990) reported that a high concentration of GABA or baclofen could alter the voltage dependence of an A-type potassium current in cultured hippocampal neurons. They concluded that by shifting the voltage dependence of inactivation of these channels to more positive potentials, GABAB receptor stimulation can enhance activation of the A-current at resting membrane potential. The authors suggest that by enhancing the A-current in presynaptic terminals,

126

DAVID D. MOTT AND DARRELL V. LEWIS

GABAB receptor activation could shorten action potential duration and thereby decrease transmitter release. Alternatively, Wagner and Dekin ( 1993) have reported that in cultured premotor respiratory neurons, GABAB receptors are coupled to a barium-insensitive outwardly rectifying potassium conductance. These channels are distinct from those underlying the response to GABA and baclofen described above, but have properties similar to the S-channel, which regulates transmitter release from Aplysia sensory neurons. In conclusion, GABAB receptors on a variety of central neurons are linked through an inhibitory G protein, presumedly Go protein, to an inwardly rectifying potassium conductance. It is not yet known whether Go proteins directly couple GABA, receptors to potassium channels o r whether an additional diffusible second messenger is involved; however, a role for phospholipase A, has been proposed. The Go protein that couples GABAB receptors to potassium channels is most likely the same G protein responsible for coupling GABAB receptors to voltage-sensitive clacium channels. In contrast, GABAB receptors are coupled through Gi protein to inhibition of adenylate cyclase and inhibition of transmitterstimulated adenylyl cyclase activation. The ability of a single cell to express all of these effector systems raises the possibility that either GABAB receptors are coupled to more than one subtype of G protein in that cell or that distinct subtypes of receptors are involved.

IV. Function of GAB& Receptors

The known actions of GABAB receptors can be attributed to the effector systems to which GABAB receptors are coupled. At present, a functional role for each of these effector systems has not been determined. However, several recent studies have suggested a potential physiological role for GABAB receptor-mediated potentiation of neurotransmitter-stimulated adenylyl cyclase activity. Andrade ( 1993) observed that baclofen enhanced the ability of norepinephrine, through P-adrenergic receptors, to increase intracellular cAMP and thereby reduce the calcium activated slow afterhyperpolarization (I,,,) in hippocampal neurons. However, in the absence of norepinephrine, baclofen had no effect on the reduction in the AHP produced by fJ-bromo-cAMP, a membrane-permeable cAMP analogue. Thus, the author concluded that GABAB receptor activation enhanced the action of norepinephrine on the AHP by potentiating the P-adrenergic stimulation of adenylyl cyclase. Similarly, Burgard and Sarvey (1991) found that baclofen was

CENTRAL GABA, RECEPTORS

127

able to enhance the induction of long-lasting potentiation (LLP) of the excitatory response in the dentate gyrus produced by P-adrenergic agonists. They observed that a low concentration of isoproterenol had no effect; however, when isoproterenol and baclofen were applied together, at concentrations that alone produced no LLP, significant LLP was induced. Taken together, these results demonstrate that coactivation of GABAB receptors and G,-linked receptors can synergistically enhance several physiological effects, possibly by potentiating the stimulation of adenylyl cyclase (see Section III,C,2). In addition to interacting through adenylyl cyclase with G,-coupled receptors, GABAB receptors may also interact through inhibitory G proteins with GABAA receptors. Binding studies in whole cells from cerebellar primary culture have demonstrated that baclofen markedly reduces [3H]muscimol binding to GABAA receptors (Kardos and Kovacs, 1991). This effect was blocked by pretreatment with pertussis toxin, indicating that an inhibitory G protein was required. Similarly, Hahner et al., (199 1) found that in membrane vesicles from mouse cerebellum (and to a lesser extent from cerebral cortex) baclofen inhibits muscimol-stimulated 36Cluptake through nondesensitizing GABAA receptor/channels. This action of baclofen was stereoselective, calcium-dependent, and blocked by 2-hydroxysaclofen, indicating that baclofen was operating through GABAB receptors. The inhibitory action of baclofen was mimicked by GTP-y-S, but not GDP-P-S, suggesting that G proteins played a role in the inhibition. Furthermore, the action of baclofen was blocked by U73 122, an inhibitor of phospholipase C. These results suggest that GABAB receptors can inhibit the function of GABA, receptors through a G protein-mediated activation of phospholipase C and the subsequent phosphorylation of the GABAA receptor (Sigel and Baur, 1988; Browning et al., 1990; Hahner et al., 1991). Thus, it now appears that, through their transduction mechanisms, GABA, receptors can integrate with, and modulate the activity of other neurotransmitter systems. In contrast to this biochemical interaction, the majority of identified electrophysiological actions of GABAB receptors are mediated through modulation of potassium channels and voltage-sensitive calcium channels. These actions will be discussed in detail in the remainder of this review. We will first discuss postsynaptic GABAB receptor-mediated effects and then the presynaptic effects of GABAB receptors on both excitatory and inhibitory terminals. Because the majority of the studies of GABAB receptor function have been performed in the hippocampal formation, we will concentrate on the effects of GABAB receptors in this region and include results from other brain regions where appropriate.

128

DAVID D. MOTT A N D DARRELL V. LEWIS

A. POSTSYNAPTIC GABA, RECEPTORS 1. Characteristics of GABA, Inhibit09 Postsynaptic Potentials

a. Late Inhibitory Postsynaptic Potentials. Orthodromic stimulation in the hippocampus evokes a synaptic response consisting of an EPSP followed by a biphasic inhibitory postsynaptic potential (1PSP) (Fig. 3). The early component of this IPSP, which peaks 10-20 ms after the stimulus, results from a GABAA receptor-mediated increase in the conductance of the membrane to chloride ions. It is blocked by GABAA antagonists, such as bicuculline or picrotoxin, and reverses at a membrane potential of about -70 mV, close to the equilibrium potential for chloride (for review see Bormann, 1988). In contrast, the late component of the biphasic IPSP is resistant to GABA, blockers. It peaks 150-200 ms after the stimulus and lasts for about 1 s (Fig. 3). A similar late IPSP has been recorded from neurons in many different brain regions, including neocortex (Howe et al., 1987; Connors et al., 1988; McCormick, 1989), thalamus (Soltesz et al., 1989; Crunelli and Leresche, 1991), locus coeruleus (Olpe et al., 1988), and septum (Stevens et al., 1987), indicating that the presence of these potentials is not restricted to the hippocampus. 6. Pharmacology and Conductance Mechanism of the Lute IPSP. The late 1PSP has many characteristics in common with the hyperpolarizing response to GABA or baclofen, suggesting that all are mediated through GABA, receptors. One similarity is that, like the hyperpolarizing GABA, receptor-mediated response, the late IPSP is associated with an increase in membrane potassium conductance (Nicoll and Alger, 1981; Thalmann and Ayala, 1982; Alger, 1984; Newberry and Nicoll, 1984b; Thalmann, 1984; Kehl and McLennan, 1985b; Hablitz and Thalmann, 1987). This conclusion is based on several observations. First, the late IPSP reverses at a membrane potential of about -90 mV, close to both the equilibrium potential for potassium and the reversal potential of the calcium-activated potassium conductance (Fig. 3). Second, the equilibrium potential of the late IPSP is sensitive to changes in the extracellular concentration of potassium, but not chloride. Finally, the response is blocked by agents know to block potassium channels, such as extracellular barium (Gahwiler and Brown, 1985; Newberry and Nicoll, 1985), intracellular cesium (Gahwiler and Brown, 1985; Otis and Mody, 1992), QX-314 (Nathan et al., 1990a; Otis and Mody, 1992),and THA (Halliwell and Grove, 1989; Lambert and Wilson, 1993). Finally, the hyperpolarization produced by baclofen and that produced by the late IPSP are not additive, indicating that the potassium conductance responsible for each of these responses is one and the same (Ogata, 1990b).

CENTRAL GABA, RECEPTORS

129

The late IPSP is also similar to the GABAB receptor-mediated postsynaptic response in its dependence on G proteins. Thus, addition of GTPy-S mimics the hyperpolarization and conductance increase produced by the late IPSP. This effect is not additive with the late IPSP (Thalmann, 1988; Mott et al., 1993b). Furthermore, pertussis toxin blocks the late IPSP, indicating that as with the GABA, receptor-mediated hyperpolarization, the required G protein is either Goo r Gi (Thalmann, 1987, 1988; Dutar and Nicoll, 1988a). No conclusive evidence has been presented to demonstrate a requirement for a diffusible second messenger in the mechanism underlying the late IPSP; however, several potential candidate systems have been ruled out. For example, it is unlikely that G protein-mediated alterations in cAMP produce the late IPSP, since increasing the intracellular concentration of cAMP by applications of forskolin (Hablitz and Thalmann, 1987) or direct addition of the niembrane-permeant cAMP analogue 8bromo-CAMP (Newberry and Nicoll, 1984a) has no effect on the late IPSP in CA3 neurons. In addition, G protein activation of phosphiolipase C is unlikely to be required since neither of the reaction products of this enzyme mimicked the late IPSP. For example, 1,4,5 IP,, does not appear to be involved, since intracellular injection of EGTA has no effect on the late IPSP, indicating that the response is calcium-independent (Thalmann, 1984; Hablitz and Thalmann, 1987, but see Blaxter et al., 1986; Segal, 1990b). Alternatively, activation of protein kinase C with phorbol esters does not mimic, but rather blocks, the late IPSP (Baraban et al., 1985; Dutar and Nicoll, 1988a). Although these results suggest that phospholipase C is not responsible for the late IPSP, they do not exclude a modulatory role for protein kinase C. Thus, like the GABAB receptormediated response, the late IPSP appears to be mediated by a direct Go protein interaction with potassium channels, although a role for phospholipase A, has not been ruled out. T h e similarities in appearance and underlying mechanism between the late IPSP and the GABA, receptor-mediated postsynaptic response suggested that the late IPSP was produced by activation of GABABreceptors by synaptically released GABA. This conclusion was confirmed when it was demonstrated that the late IPSP was blocked by phaclofen (Dutar and Nicoll, 1988b; Karlsson et al., 1988; Solteszetal., 1988).Subsequently, studies in many different brain regions with a variety of GABABreceptor antagonists, including 2-hydroxysaclofen (Lambert et al., 1989), CGP 35348 (Olpe et al., 1990), and CGP 55845 (Jarolimek et al., 1993), have supported this finding (Fig. 3). c. Current-Voltage Relationship of IPSPIC,. The current-voltage (Z-V) relationship of the late IPSP (IPSPB) has generally been reported to show

A

1o m j

e

0

IPSP, IPSP, 100 ma

B 20

Em(mV)

-96 0

0

7 4 100 ms

c

20 pM

DNQX

+

50 P M D-APV

20

pM DNQX

60 p M D-APV

+

+

-

’

-15

-20

-

+ 50

p M Plcrotoxln

+ 400 p M 2-OH Saclofen

400

pM

2-OH Saclofsn

Wash

1 OmJv 200 ma

FIG. 3. (A) A typical perforant path-evoked response recorded from a dentate gyrus granule cell at resting membrane potential ( - 79 mV) showing the EPSP (solid triangle), the IPSP, (open circle), and the IPSPB (solid circle). Because of the hyperpolarized resting potential of granule cells the IPSP, is typically depolarizing. (B) Voltage dependence of the IPSP. (Left) Sample IPSPs recorded from a granule cell at the membrane potentials shown on the left. Responses were evoked by direct stimulation of interneurons in the dentate gyrus molecular layer. This monosynaptic IPSP was pharmacologically isolated by blocking the EPSP with DNQX and D-APV. (Right) Monosynaptic IPSP amplitude plotted against membrane potential for this cell. The amplitude of IPSP, (open circle) and IPSPB (solid circle) was measured at 14 and 200 ms, as indicated by the circles on the waveforms on the left. (C) Pharmacological characterization of the monosynaptic IPSP. (Top) Depolarization of a granule cell to - 60 mV caused the IPSP, to become hyperpolarizing. Addition

CENTRAL GABA, RECEPTORS

131

no rectification (Thalmann, 1984; Hablitz and Thalmann, 1987; Rovira et al., 1990). Linear I-V curves over a range of membrane potentials from about - 110 to -50 mV have been reported. In granule cells of the dentate gyrus we found that, although IPSPB appeared to show inward rectification, this could largely be accounted for by the voltagedependent properties of the membrane (Mott et al., 1993b, also see Alger, 1984; Thalmann, 1984; Rausche et al., 1989; Crunelli and Leresche, 1991). Similarly, Otis et al. (1993) reported a linear current-voltage relationship for the isolated IPSCB in dentate granule cells. However, they also found that IPSCB displayed outward rectification in some cells when the extracellular potassium was raised. In contrast, other studies have reported that the late IPSP inwardly rectifies (Newberry and Nicoll, 1985; Soltesz et al., 1989). The reason for this discrepancy is not known; however, some of the apparent nonlinear behavior of the response may be caused by the voltage-dependent properties of the membrane. It is of interest to note that the current-voltage relationship generally reported for the response to GABAB receptor agonists shows inward rectification (see Section IlI,C,5).This raises the possibility that the population of potassium channels activated by the agonist may not be identical to those activated synaptically. d. IPSPIC, Kinetics. The kinetics of IPSPB are considerably slower than those of IPSPA, reflecting the G protein coupling of the GABAB receptor with the potassium channel. In general, the latency to onset of the GABAB receptor-mediated potential ranges from 30 to 50 ms (Connors et al., 1988; Davies et al., 1990; Mott et al., 1993b), whereas the onset latency of the underlying current is 12-35 ms (Hablitz and Thalmann, 1987; Otis et al., 1993). This latency is considerably longer than that of the GABAA receptor-mediated current [