Nature Neuroscience October 2006

www.nature.com/natureneuroscience

EDITORIAL OFFICE [email protected] Varick Street, Fl 9, New York, NY 10013-1917Tel: (212) 726 9319, Fax: (212) 696 0978Editor: Sandra Aamodt Senior Editor: Kalyani Narasimhan Associate Editors: Annette Markus, Charvy Narain Assistant Editors: Hannah Bayer, Josh McDermottCopy Editors: Anita Gould, Lavanya ReddySenior Production Editor: Jessica IannuzziProduction Editor: Elizabeth JerabekIllustrators: Kimberly Caesar, Katie RisCover Design: Erin BoyleEditorial Assistant: Mark Zipkin

MANAGEMENT OFFICESNPG New York75 Varick Street, Fl 9, New York, NY 10013-1917Tel: (212) 726 9200, Fax: (212) 696 9006Publisher: Beatrice RenaultExecutive Editor: Linda MillerChief Technology Officer: Howard RatnerHead of Nature Research & Reviews Marketing: Sara GirardAssistant Marketing Manager: Amy MaurerProduction Coordinator: Diane TempranoNew Technology Director: Greg SuprockHead of Web Services: Anthony BarreraWeb Production Manager: Susan Kline

NPG LondonThe Macmillan Building, 4 Crinan Street, London N1 9XWTel: 44 207 833 4000, Fax: 44 207 843 4996Managing Director: Annette ThomasPublishing Director: Peter CollinsEditor-in-Chief, Nature Publications: Philip CampbellMarketing Director: Della SarDirector of Web Publishing: Timo Hannay

NPG Nature Asia-PacificChiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843Tel: 81 3 3267 8751, Fax: 81 3 3267 8746Publishing Director Asia-Pacific: David SwinbanksAssociate Director: Antoine E. BocquetManager: Koichi NakamuraSenior Marketing Manager: Peter YoshiharaAsia-Pacific Sales Director: Kate YoneyamaAsia-Pacific Sales Manager: Rinoko Asami

DISPLAY ADVERTISING [email protected] (US/Canada) [email protected] (Europe) [email protected] (Asia)US Head of Display Advertising: Stephen Schwartz, Tel: (212) 726 9256, Fax: (212) 696 9481Global Head of Display Advertising Sales: John Michael, Tel: 44 207 843 4960, Fax: 44 207 843 4996Global Head of Online Advertising: Matt Peskett, Tel: 44 207 843 4988, Fax: 44 207 843 4749Head of Display AdvertisingEurope: Gerard Preston, Tel: 44 207 843 4960, Fax: 44 207 843 4996Business Development Manager: Claire Hines, Tel: 44 207 843 4960, Fax: 44 207 843 4996Asia-Pacific Sales Manager: Rinoko Asami, Tel: 81 3 3267 8751, Fax: 81 3 3267 8746Western Region Sales Manager: George Lui, Tel: (415) 781 3804, Fax: (415) 781 3805Sales Executives:New England: Sheila Reardon, Tel: (617) 399 4098, Fax: (617) 426 3717New York, Mid-Atlantic, Southeast: Jim Breault, Tel: (212) 726 9334, Fax: (212) 696 9481Midwest: Mike Rossi, Tel: (212) 726 9255, Fax: (212) 696 9481Northwest: Mathieu DesJardins, Tel: (415) 781 6422, Fax: (415) 781 3805Eastern England/Scotland/Italy/Spain/Israel: Matthew Wilkinson, Tel: 44 207 843 4960, Fax: 44 207 843 4749South/West United Kingdom/Scandinavia/Holland: Marianne Boulakas, Tel: 44 207 843 4969, Fax: 44 207 843 4749Northern Germany: Gerard Preston, Tel: 44 207 843 4960, Fax: 44 207 843 4749Southern Germany/Austria/Switzerland/France/Belgium: Sabine Hugi-Frst, Tel: 41 52761 3386, Fax: 41 52761 3419

NATUREJOBS [email protected] (US/Canada) [email protected] (Europe) [email protected] (Asia)Publisher: Ben Crowe, Tel: (212) 726 9245, Fax: (212) 696 9482US Sales Manager: Peter Bless, Tel: (212) 726 9248, Fax: (212) 696 9482Japan Sales Manager: Rinoko Asami, Tel: 81 3 3267 8751, Fax: 81 3 3267 8746

SITE LICENSE BUSINESS UNITAmericas: Tel: (888) 331 6288 [email protected]/Pacific: Tel: 81 3 3267 8751 [email protected]/New Zealand: Tel: 61 3 9825 1160 [email protected]: Tel: 00 91 11 324 4186 [email protected]: Tel: 44 207 843 4759 [email protected]

CUSTOMER SERVICE www.nature.com/helpSenior Global Customer Service Manager: Gerald CoppinFor all print and online assistance, please visit www.nature.com/helpPurchase subscriptions:Americas: Nature Neuroscience, Subscription Dept., 342 Broadway, PMB 301, New York, NY 10013-3910. Tel: (866) 363 7860, Fax: (212) 689 9108Europe/ROW: Nature Neuroscience, Subscription Dept., Macmillan Magazines Ltd., Brunel Road, Houndmills, Basingstoke RG21 6XS, United Kingdom. Tel: 44 1256 329 242, Fax: 44 1256 812 358Japan: Nature Neuroscience, Nature Japan K.K., MG Ichigaya Bldg. 5F, 19-1 Haraikatamachi, Shinjuku-ku, Tokyo 162-0841. Tel: 81 3 3267 8751, Fax: 81 3 3267 8746India: Harpal Singh Gill, Macmillan Magazines Ltd, 5A/12 Ansari Road, Darya Ganj, New Delhi, 110 002 India. Tel: 00 91 11 324 4186, Tel/Fax: 00 91 11 327 2010

REPRINTS [email protected] Neuroscience Reprint Department, Nature Publishing Group, 75 Varick Street, Fl 9, New York, NY 10013-1917, USA.For commercial reprint orders of 600 or more, please contact:UK Reprints Sales Executive: Caren Collins, Tel: 44 207 843 4967, Fax: 44 207 843 4839US Reprints Sales Executive: Sharda Tulsie, Tel: (212) 726 9631, Fax: (212) 679 0843

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e

iVOLUME 9 NUMBER 10 OCTOBER 2006

Nature Neuroscience (ISSN 1097-6256) is published monthly by Nature Publishing Group, a trading name of Nature America Inc. located at 75 Varick Street, Fl 9, New York, NY 10013-1917. Periodicals postage paid at New York, NY and additional mailing post offices. Editorial Office: 75 Varick Street, Fl 9, New York, NY 10013-1917. Tel: (212) 726 9319, Fax: (212) 696 0978. Annual subscription rates: USA/Canada: US$199 (personal), US$2,160 (institution). Canada add 7% GST #104911595RT001; Euro-zone: 271 (personal), 1,860 (institution); Rest of world (excluding China, Japan, Korea): 175 (personal), 1,200 (institution); Japan: Contact NPG Nature Asia-Pacific, Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 (03) 3267 8751, Fax: 81 (03) 3267 8746. POSTMASTER: Send address changes to Nature Neuroscience, Subscriptions Department, 342 Broadway, PMB 301, New York, NY 10013-3910. Authorization to photocopy material for internal or personal use, or internal or personal use of specific clients, is granted by Nature Publishing Group to libraries and others registered with the Copyright Clearance Center (CCC) Transactional Reporting Service, provided the relevant copyright fee is paid direct to CCC, 222 Rosewood Drive, Danvers, MA 01923, USA. Identification code for Nature Neuroscience: 1097-6256/04. Back issues: US$45, Canada add 7% for GST. CPC PUB AGREEMENT #40032744. Printed by Publishers Press, Inc., Lebanon Junction, KY, USA. Copyright 2006 Nature Publishing Group. Printed in USA.

E D I TO R I A L

1195 Fighting animal rights terrorism

B O O K R E V I E W

1197 Understanding Autism: from Basic Neuroscience to TreatmentEdited by S O Moldin & J L R RubensteinReviewed by Tony Charman

N E W S A N D V I E W S

1199 Keeping the memories flowingGregory J Quirk & Ivan Vidal-Gonzalez see also p 1321

1201 Turning by asymmetric actinGuo-li Ming see also pp 1247 and 1265

1203 Hooking up new synapsesThomas Biederer see also p 1294

1205 Intercellular miscommunication in polyglutamine pathogenesisChristopher A Ross & Don W Cleveland see also p 1302

1207 Walk the line: parietal neurons respect category boundariesVincent P Ferrera & Jack Grinband

A proposed genetic mechanism for developmental dyslexia

(p 1213)

Developmental disorders such as dyslexia, autism and fragile X

syndrome can affect a wide range of cognitive and social functions

that arise in childhood. Researchers suspect that multiple genes in early

development, as well as environmental factors, may influence the course of

these disorders, but surprisingly little is known about their etiology. In this

issue, we present several perspectives on the neurobiology of developmental

disorders and their relationship to one another. This special focus is

sponsored by the March of Dimes, Cure Autism Now and Autism Speaks.

(pp 12091229)

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e

iii

VOLUME 9 NUMBER 10 OCTOBER 2006

NATURE NEUROSCIENCE

Local caspase activity directs dendritic pruning

(p 1234)

Focal adhesion kinase regulates axon growth and guidance

(p 1274)

Meningeal membranes are fundamental to cortex development

(p 1284)

I N T R O D U C T I O N : C H I L D H O O D D E V E LO P M E N TA L D I S O R D E R S

1209 Childhood developmental disordersCharvy Narain

S P O N S O R S F O R E WO R D : C H I L D H O O D D E V E LO P M E N TA L D I S O R D E R S

1211 A genetics pioneer focused on child health challengesMichael Katz

P E R S P E C T I V E S : C H I L D H O O D D E V E LO P M E N TA L D I S O R D E R S

1213 From genes to behavior in developmental dyslexiaAlbert M Galaburda, Joseph LoTurco, Franck Ramus, R Holly Fitch & Glenn D Rosen

1218 Time to give up on a single explanation for autismFrancesca Happ, Angelica Ronald & Robert Plomin

1221 Fragile X syndrome and autism at the intersection of genetic and neural networksMatthew K Belmonte & Thomas Bourgeron

1226 What developmental disorders can tell us about the nature and origins of languageGary Marcus & Hugh Rabagliati

B R I E F COM M U N I C AT I O N S

1231 Kinase activity of mutant LRRK2 mediates neuronal toxicityW W Smith, Z Pei, H Jiang, V L Dawson, T M Dawson & C A Ross

1234 Local caspase activity directs engulfment of dendrites during pruningD W Williams, S Kondo, A Krzyzanowska, Y Hiromi & J W Truman

1237 NMDA receptors are critical for unleashing consolidated auditory fear memoriesC B Mamou, K Gamache & K Nader

1240 Brain state and contrast sensitivity in the awake visual thalamusM Cano, T Bezdudnaya, H A Swadlow & J-M Alonso

1243 When sustained attention impairs perceptionS Ling & M Carrasco

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e

vVOLUME 9 NUMBER 10 OCTOBER 2006

NATURE NEUROSCIENCE

Impaired glutamate transport in glia causes neurodegeneration

(pp 1205 and 1302)

Amygdala facilitates cortical interactions to enhance memory

(pp 1199 and 1321)

A R T I C L E S

1247 Asymmetrical -actin mRNA translation in growth cones mediates attractive turning to netrin-1K Leung, F PG van Horck, A C Lin, R Allison, N Standart & C E Holt see also pp 1201 and 1265

1257 Soluble adenylyl cyclase is required for netrin-1 signaling in nerve growth conesK Y Wu, J H Zippin, D R Huron, M Kamenetsky, U Hengst, J Buck, L R Levin & S R Jaffrey

1265 An essential role for -actin mRNA localization and translation in Ca2+-dependent growth cone guidanceJ Yao, Y Sasaki, Z Wen, G J Bassell & J Q Zheng see also pp 1201 and 1247

1274 Focal adhesion kinase signaling at sites of integrin-mediated adhesion controls axon pathfindingE Robles & T M Gomez

1284 Meninges control tangential migration of hem-derived Cajal-Retzius cells via CXCL12/CXCR4 signalingV Borrell & O Marn

1294 NGL family PSD-95interacting adhesion molecules regulate excitatory synapse formationS Kim, A Burette, H S Chung, S-K Kwon, J Woo, H W Lee, K Kim, H Kim, R J Weinberg & E Kim see also p 1203

1302 Bergmann glia expression of polyglutamine-expanded ataxin-7 produces neurodegeneration by impairing glutamate transportS K Custer, G A Garden, N Gill, U Rueb, R T Libby, C Schultz, S J Guyenet, T Deller, L E Westrum, B L Sopher & A R La Spada see also p 1205

1312 Nonlinear, binocular interactions underlying flow field selectivity of a motion-sensitive neuronK Farrow, J Haag & A Borst

1321 Emotional enhancement of memory via amygdala-driven facilitation of rhinal interactionsR Paz, J G Pelletier, E P Bauer & D Par see also p 1199

1330 Functional alignment of feedback effects from visual cortex to thalamusW Wang, H E Jones, I M Andolina, T E Salt & A M Sillito

1337 A human parietal face area contains aligned head-centered visual and tactile mapsM I Sereno & R-S Huang

N AT U R E N E U R O S C I E N C E C L A S S I F I E D

See back pages

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e

NATURE NEUROSCIENCE VOLUME 9 | NUMBER 10 | OCTOBER 2006 1195

E D I TO R I A L

Fighting animal rights terrorismA researcher recently announced his decision to stop working with primates after animal rights protesters targeted his family. These events underscore the need for neuroscientists to support the Animal Enterprise Terrorism Act.

In June, the Animal Liberation Front took responsibility for attempting to firebomb the home of Lynn Fairbanks, a primate researcher at the University of California, Los Angeles (UCLA). The Molotov cocktail was accidentally left at the wrong address, on the porch of her elderly neighbor, and did not explode because its timing device failed, but the terrorists incompetence was apparently no consolation to her colleague Dario Ringach. The neuroscientist, who has young children, emailed animal rights groups in August to say that he had decided to stop doing research on monkeys, so please dont bother my family anymore.

Like Fairbanks, Ringach had been targeted by the UCLA Primate Freedom Project, whose website still lists home addresses and phone numbers for several other colleagues, including neuroscientists Joaquin Fuster, John Schlag and Madeleine Schlag-Rey. Over several years, the researchers have been subjected to a campaign of harassment that included demonstrations at their homes and pamphlets distributed to their neighbors, as well as threatening phone calls and emails. Elsewhere, targets of similar protests have had abuse shouted through bullhorns or painted on their homes or cars, doorbells rung repeatedly, and windows smashed or doors broken down while family members were in the house. Animal-rights websites post the names of scientists spouses and children, along with their ages and schools.

In the wake of these events, UCLA has announced that it is working to prepare a more effective response to future incidents, which might include lawsuits against activists. Unfortunately, this initiative comes too late for Ringach. The correct time for such reflection is before the problems become critical. Other universities would be well advised to give serious attention to the possibility that their researchers may be targeted and to develop a plan for dealing with such incidents before they occur.

Of course, activists are entitled to express their point of view vigorously through peaceful protests and to campaign for changes in the law and public opinion. However, such demonstrations against scientists at their homes are clearly aimed at intimidating individual researchers and their families, not at persuading the public. Unfortunately, some extremists within the animal-rights movement are attempting to impose their views on others through violence and threats, rather than pursuing change through the political system. I think Dario Ringach is a poster boy for the concept that the use of force or the threat of force is an effective means to stop people who abuse animals, Animal Liberation Front spokesman Jerry Vlasak told the Los Angeles Times. People who act on this philosophy deserve to be treated as criminals, not as debating partners.

In recent years, animal-rights groups have developed sophisticated tactics to exploit loopholes in the law by harassing employees of animal research organizations and the companies with which they do business.

Law enforcement agencies and universities have had difficulty protecting researchers from these attacks because even if the perpetrators are identified and convicted, offenses like trespassing and disturbing the peace are misdemeanors that carry minor penalties. These tactics, which originated in the United Kingdom (UK), have proven successful in other countries, as we have discussed previously1. The UK has also led the way in responding to such attacks by amending the Serious Organized Crime Act to define animal extremism as terrorism.

Senator James Inhofe of Oklahoma and Congressman Thomas Petri of Wisconsin recently introduced a bill to close loopholes in United States (US) law. The Animal Enterprise Terrorism Act would extend previous legislation to cover attacks on individual employees or on companies that do business with animal research organizations. Had it been passed in time, this law would have allowed federal officers to respond effectively to the protests at UCLA, by criminalizing any pattern of conduct (two or more incidents) that intentionally place a person or that persons family in reasonable fear of death or serious injury by threats, vandalism, property damage, trespass, harassment or intimidation. The act also spells out increased penalties for violations, which range from 6 months in jail for nonviolent obstruction of a business that causes no injury up to life imprisonment for the death of a person.

Civil rights groups initially raised concerns about the bills potential to restrict constitutionally protected speech, including legitimate consumer boycotts. To address these concerns, the House Judiciary Committee, in cooperation with the American Civil Liberties Union, has modified the bills language to make it explicit that the law would not prohibit any expressive conduct (including peaceful picketing or other peaceful demonstration). There is precedent for legislation that balances the need to prevent harassment of individuals against respect for civil liberties, in anti-stalking laws that aim to protect women from domestic violence.

The recent events in Los Angeles have had unintended consequences for the animal-rights movement. Although the majority of activists hope to sway public opinion away from supporting animal research, the acts of extremists have drawn media attention instead to their tactics of harassment and intimidation of individuals, which do not enjoy public support. In particular, the bomb attempt and Ringachs announcement caught the attention of Senator Dianne Feinstein of California, who in September signed on as a cosponsor of the Animal Enterprise Terrorism Act, making it a bipartisan initiative and increasing its odds of passing in the legislature. We urge all US neuroscientists to contact their senators and representatives to demand that this bill be brought to a vote before Congress adjourns for the year.

1. Nat. Neurosci. 7, 413 (2004).

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e


B O O K R E V I E W

Autism at the crossroads: determining the phenotype matters for neuroscience

Understanding Autism: from Basic Neuroscience to Treatment

Edited by S O Moldin & J L R Rubenstein

Taylor & Francis, 2006552 pp, hardcover, $159.95 ISBN 0849327326

Reviewed by Tony Charman

An ambitious and comprehensive volume, Understanding Autism: from Basic Neuroscience to Treatment summarizes the state of the art in the neuroscience of autism. It strikes a tone of promise and potential but also caution, highlighting the many things we do not yet know about autism.

Autism is more common than was previously recognized. Up to 1% of the population has autism spectrum disorders, a range of neurodevel-opmental conditions that are likely to differ in their underlying etiology but share a common phenotype, characterized by impairment in social relatedness and reciprocity. A neuroscientific understanding of autism is important in light of increasing evidence that multiple etiologies may converge to disrupt brain systems critical for social behavior.

The first chapter of the book strikes a cautionary note. Lord and Spence point out that, because of the lack of a clear neurobiological marker, autism spectrum disorders are necessarily defined by behavior, which is both intriguing and frustrating for researchers. The caveat is necessary because the categorical diagnosesas set out in psychiatric classification systemsmay not represent true categories in nature.

This problem has led researchers to look for more homogenous sub-types of autism. Geneticists have shifted from attempting to identify susceptibility genes for autism per se to studying endophenotypeselementary single aspects of the autism phenotype, such as language development, social cognition or severity of repetitive behaviors. One chapter highlights promising initial findings from such studies, but a major challenge is that we do not know which subtypes are biologically meaningful. Several chapters provide some clues by comparing autism with neurodevelopmental disorders, such as chromosome 15q11-13 disorders, fragile X syndrome and Rett syndrome, with phenotypes that overlap aspects of autism.

The book does justice to a major advance of the past decade: the recognition that autism is a multi-system disorder. Individual chapters review brain systems and structures that show abnormal development in autism: amygdala and the social brain, cerebellum, prefrontal cortex and thalamus. Each chapter contains detailed, scholarly summaries of the relevant research. The focus on one brain system at a time, however, can lead the reader to lose sight of the clinical phenotype.

Two exemplary chapters summarize evidence for whole-brain abnormalities in the structural and functional connectivity of differ-ent brain systems. Although such models are necessarily complex, they may explain both how overdevelopment of local connections leads to enhanced perception of detail in autism and how undergrowth of dis-tant connections leads to underdevelopment of higher cognitive abili-ties. Other chapters summarize early work on modeling features of autism in animals.

The study of homogeneous subtypes is less common in neuroscience than in genetics. As in other complex neuropsychiatric disorders, the traditional approach in neuroscience is to compare brain structure and function in categorically defined individuals with autism and unaf-fected controls. However, these studies have not led to a consensus on what defines or causes autism. This is not surprising, as both the etiol-ogy and the outcome (in terms of brain and behavior) are likely to be heterogeneous among people who meet clinical diagnostic criteria for autism. Variation among studies seems inevitable in this situation.

If there is one issue that receives less attention than it warrants throughout the volume, it is that autism is far more common in males than females. Several researchers have attempted to explain this sex ratio, most notably Simon Baron-Cohen and David Skuse, although their proposed explanations (respectively, differences in fetal testoster-one and an X-linked gene for social cognition) are very different from each other. This topic seems important enough that it should have received more emphasis than it does.

However, the phenotypic challenge presented by autism might go even further. There is increasing evidence that the elements of the autism phenotype might be dimensionally distributed in the popula-tion, with no clear boundary between normal population variation and disordered levels of these behaviors. Thus, genetics and neuroscience studies will have to encompass a population as well as a clinical disor-derbased approach. Several chapters include such studies, which have found promising associations between the severity of autism features and behavioral, neuropsychological or biological measures.

The volume also includes chapters on epidemiology, pharmacologi-cal and psychosocial approaches to intervention and an analysis of the considerable costs to society of autism. However, the main focus of the book is on the biology of autism. A mature neuroscience of autism will need to investigate links back from the brain to heterogeneous genetic and epigenetic etiologies, and also forward from the brain to the impairments in cognition, behavior and development that char-acterize the complex range of conditions and presentations that we currently can still only describe as the autism phenotype. This book provides a useful summary of where the science of understanding the autistic brain has arrived and provides glimpses of the advances we can anticipate in the next decade.

Tony Charman is Professor of Neurodevelopmental Disorders, Behavioural &

Brain Sciences Unit, UCL Institute of Child Health, 30 Guilford Street, London

WC1N 3JH, UK.

e-mail: [email protected]

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Keeping the memories flowingGregory J Quirk & Ivan Vidal-Gonzalez

Transmission of sensory information through the rhinal cortices is essential for hippocampus-dependent learning. In this issue Paz et al. show that amygdala activity elicited by an unexpected reward facilitates communication from perirhinal to entorhinal cortex, providing a physiological mechanism for emotional modulation of memory.

So short is love, and so long is forgetting, wrote Pablo Neruda in a testimony to the tenacity of emotional memories. How is it that emotions, positive and negative, strengthen memories for particular events in our lives? At the heart of emotional modulation of memory lies the amygdala, a structure activated by emotional stimuli. It is widely accepted that the basolateral complex of the amygdala (BLA) strengthens consolidation of emotional memories through its projections to wide-spread cortical areas1,2. Despite a wealth of neuropharmacological data, however, there is still no neurophysiological model to account for this important function. In a tour de force of behavioral neurophysiology, Paz and colleagues3 demonstrate that amygdala activity strengthens memories by facilitating transmission of sensory information from the neocortex to the hippocampus.

Amygdala neurons increase their sensory responses to cues associated with positive or negative outcomes46. Such responses could provide a means for emotional responses to influence memory, but the mechanism by which amygdala activity might strengthen learning in other systems has remained unclear. To address this important issue, Paz et al. used an appetitive trace conditioning task in cats, in which a light and food are separated by a brief interval. Over time, the cat learns to lick a food spout when the light is presented. The hippocampus is needed for the animal to learn the association between the light and food.

Sensory information reaches the hippocampus primarily through a cascade

of sequential projectionsfrom sensory cortices to the perirhinal cortex, from there to entorhinal cortex, and finally to the hippocampus. Paradoxically, there is little communication between perirhinal and ento-rhinal cortex owing to strong feed-forward inhibition, thereby diminishing the flow of information into and out of the hippocampus7 under normal circumstances.

Both perirhinal and entorhinal cor-tex receive a robust projection from the BLA, raising the possibility that the BLA might facilitate perirhinal-entorhinal transmission in the presence of an arousing stimulus, thereby enhancing memory. To test for such interactions, Paz et al. recorded simultaneously from neurons in all three structures. As previously observed, perirhi-nal cortex neurons on average only weakly excited entorhinal cortex neurons, as evi-denced by few significant cross- correlations between their spike trains. Paz et al. then modified the cross- correlation technique to examine all three structures at once, using a spike- triggered joint histogram (STJH). This approach reveals how the perirhinalentorhinal correlation is modulated by a reference event, in this case the occurrence of BLA spikes (Fig. 1), by computing the cross-correlation conditioned on the presence of the reference event. Surprisingly, this analysis revealed that highly significant correlations between perirhinal and entorhinal cortex neurons emerged within 150 ms of a spike in the BLA. In other words, BLA action poten-tials briefly enabled coordinated activity between perirhinal and entorhinal cortex. Previous studies of rhinal areas missed these interactions because they did not record simultaneously from the BLA.

The finding that BLA activity induces interactions between cortical structures is

interesting, but the crucial test of the role of these interactions in emotional modulation is whether the incidence of BLA-triggered interactions (the STJH signal) increases with arousal. Impressively, Paz et al. report that the STJH signal increased following each unexpected food reward. Furthermore, as the cat learned that the light predicted the food, the STJH signal shifted from being elicited by the food to being elicited by the light. Thus, as the presentation of the food became reliably predicted by the light, the food no longer triggered the STJH signal. This is an important observation because it implies that it is not the reward per se that triggers the rhinal interactions, but the unex-pectedness of the reward.

These results agree with classical learning theory, which says that learning rates are maximal following unexpected outcomes8. Accordingly, emotional events would be expected to modulate learning most strongly when they are surprising, so that future encounters with associated sensory stimuli will not catch the animal off-guard. Not mentioned by Paz et al. is that a similar shift of neuronal responses from the unconditioned stimulus to the conditioned stimulus occurs in brainstem dopaminergic neurons9. This suggests that dopaminergic signaling could be involved in the synchroniza-tion of BLA activity and the development of the STJH signal, an idea supported by the observa-tion that dopamine suppresses local inhibitory circuits in the BLA (ref. 10).

Perirhinal and entorhinal cortex are reciprocally connected, and transmis-sion through this bottleneck is necessary for information both entering and exit-ing from the hippocampus. Calculating a directionality index from the STJHs, Paz et al. observed that reward shifted rhinal interactions in the perirhinal-to- entorhinal

Gregory J. Quirk and Ivan Vidal-Gonzalez are in

the Department of Physiology, Ponce School of

Medicine, P.O. Box 7004, Ponce, Puerto Rico.


2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e

1200 VOLUME 9 | NUMBER 10 | OCTOBER 2006 NATURE NEUROSCIENCE


cortex direction, favoring the transfer of information from the neocortex to the hippocampal formation. This intriguing finding suggests that the BLA facilitates memory encoding, rather than memory readout, which makes sense for the acquisi-tion phase of learning. Later consolidation phases were not examined, but a testable prediction from their findings is that such phases of learning should produce a rever-sal of information flow, in the entorhinal-to-perirhinal cortex direction, consistent with storage of hippocampal memories in the neocortex.

Although one might think that BLA facilitation of rhinal interactions could be explained solely by increased firing rate in the BLA, this was not the case. STJH changes did not follow changes in firing rate but were correlated with the degree of BLA synchrony. Thus, as previously observed11, task-relevant information can be signaled by the timing of action potentials, rather than their average firing rate. Paz et al. extend this idea by demonstrating that these interactions occur across structures. The combination of multisite recording with the STJH is a powerful new tool for investigations of neuronal networks

of learning, which have traditionally used techniques with lower temporal resolution, such as c-Fos immunocytochemistry or deoxyglucose mapping. One wonders if STJH signals would be observed for other structures implicated in memory consolidation, such as the hippocampus12 and medial prefrontal cortex13.

How prolonged is the effect of the amygdala on cortical circuits? Although the increase in the rhinal STJH signal following each reward was relatively short (approximately 1 second), a previous study by this group showed that footshocks increased firing rate and syn-chrony in the BLA for as much as 2 hours14. Rhinal activity was not recorded in that study, but an important prediction following from the results of Paz et al. is that STJH signals should show prolonged increases following aversive stimuli. Persistence of STJH signals long after shock would agree with infusion studies showing that activity after training in the BLA is required for its memory-enhancing effects1. An intriguing possibility is that STJH signals may reflect rehearsal of earlier events as part of the consolidation process12,15. Future studies combining pharmacological approaches with an STJH analysis of the

BLAs influences on target structures hold great promise for explaining emotions strong influence on our memories.

1. McGaugh, J.L. Science 287, 248251 (2000).2. Cahill, L. & McGaugh, J.L. Trends Neurosci. 21,

294299 (1998).3. Paz, R., Pelletier, J.G., Bauer, E.P. & Pare, D. Nat.

Neurosci. 9, 13211329 (2006). 4. Maren, S. & Quirk, G.J. Nat. Rev. Neurosci. 5, 844

852 (2004).5. Saddoris, M.P., Gallagher, M. & Schoenbaum, G.

Neuron 46, 321331 (2005).6. Paton, J.J., Belova, M.A., Morrison, S.E. & Salzman,

C.D. Nature 439, 865870 (2006).7. de Curtis, M. & Pare, D. Prog. Neurobiol. 74, 101

110 (2004).8. Rescorla, R.A. & Wagner, A.R. in Classical

Conditioning II: Current Research and Theory (eds. Black, A.H. & Prokasy, W.F.) 6499 (Appleton-Century-Crofts, New York, 1972).

9. Schultz, W. Annu. Rev. Psychol. 57, 87115 (2006).

10. Bissiere, S., Humeau, Y. & Luthi, A. Nat. Neurosci. 6, 587592 (2003).

11. Vaadia, E. et al. Nature 373, 515518 (1995).12. Sirota, A., Csicsvari, J., Buhl, D. & Buzsaki, G. Proc.

Natl. Acad. Sci. USA 100, 20652069 (2003).13. Santini, E., Ge, H., Ren, K., Pena, D.O. & Quirk, G.J.

J. Neurosci. 24, 57045710 (2004).14. Pelletier, J.G., Likhtik, E., Filali, M. & Pare, D. Learn.

Mem. 12, 96102 (2005).15. Foster, D.J. & Wilson, M.A. Nature 440, 680683

(2006).

AmygdalaPerirhinal

a bNeocortex

Hippocampus

Entorhinal

150

150

Coincidences

Ent

Prh

150 0

0

150

150

150

(ms)

(ms)

Perirhinal

Neocortex

Hippocampus

Entorhinal

108642

c

Amygdala

Figure 1 The amygdala enhances transfer of sensory information. Sensory input flows from the neocortex to the hippocampus via the rhinal cortices (descending arrows). The hippocampus in turn assists in consolidating and storing this information through projections back to neocortex (ascending arrows). (a) During spontaneous activity, with low amygdala firing rates (blue traces on left), transfer of information from perirhinal cortex (Prh) to entorhinal cortex (Ent) is minimal (thin red arrows). (b) Following reward, the amygdala increases its firing rate and synchrony. This enables the transfer of sensory information through the rhinal cortices into the hippocampus (large red arrow). (c) The dependence of transfer through the rhinal cortices on amygdala activity is evidenced by peaks in the spike-triggered joint histogram (STJH, colored square). The STJH calculates the cross-correlation between perirhinal and entorhinal spike trains with reference to a third event, in this case a BLA spike. In the color plots, the x- and y-axes represent the latencies of entorhinal and perirhinal spikes relative to the occurrence of a BLA spike, respectively. Each cell of the grid represents the number of spike pairs that occurred with the corresponding latencies relative to a BLA spike. The red peak in the STJH indicates that shortly after a BLA spike, entorhinal cortex fires after perirhinal cortex, consistent with facilitated transmission in the perirhinalentorhinal direction. The plot on the upper right corner of the STJH is a crosscorrelogram resulting from summing the leftward diagonal rows of the STJH along the rightward diagonal (indicated in black). This represents the number of pairs of entorhinal and perirhinal spikes that occur in the vicinity of a BLA spike. The effect of amygdala activity on rhinal throughput is a plausible mechanism for facilitation of emotional memory by the amygdala.

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Turning by asymmetric actinGuo-li Ming

Axon guidance requires local protein synthesis at the growth cone. Two new studies show that guidance cues induce asymmetric targeting and translation of -actin (Actb) mRNA. Such asymmetry may be the mechanism that underlies growth cone turning.

BDNF and myelin- associated glycoprotein, requires extracellular Ca2+ and is sensitive to modulation by cyclic AMP/protein kinase A (cAMP/PKA) activity11,12. Using focal laserinduced photolysis (FLIP) of caged Ca2+, an approach previously developed in their laboratory, Yao and colleagues now show that local Ca2+ elevationinduced growth cone attraction and repulsion are both abolished by protein- synthesis inhibitors7. This excit-ing finding suggests a general requirement of new protein synthesis underlying signaling of Ca2+- dependent (Group I) guidance cues.

Regulated local protein synthesis is achieved by two sequential processestransport of mRNA to specific locations, followed by initiation of mRNA translation upon stimu-lation. Although several axonal proteins are

known to be locally translated and important for growth cone collapse and repulsion810, these two studies newly identify -actin as another axonally translated protein, and go a step further: they depict nicely how path-ways regulating -actin mRNA transport and translation act in concert to induce the asym-metry that is required for acute growth cone turning6,7. Zipcode binding protein 1 (ZBP1) regulates transport of -actin mRNA through binding to a conserved 54-nucleotide element in the 3-untranslated region (UTR) of the -actin mRNA known as the zipcode (ref. 13). Both studies by Leung et al. and Yao et al. used Xenopus neuronal cultures and provide substantial evidence that Vg1RBP, a Xenopus homolog of ZBP1, interacts with -actin mRNA within neuronal growth cones. Using

The brain is wired up during development by active axon pathfinding in response to environmental guidance cues. The neuronal growth cone, the leading tip of the growing process, makes steering decisions through rearrangement of the cytoskeleton1, often at a significant distance from the cell body. The autonomy of axons in guidance has been demonstrated in a classic experiment showing that retinal axons are capable of finding their targets in the developing Xenopus embryos after being cut off from the cell body2. Protein synthesis, a process previously believed to be limited to the cell body, occurs locally in the axons3 and is required for both acute4 and long-term adaptive5 growth cone responses to guidance cues. Since these initial discover-ies, there has been a tremendous interest in the identity of newly translated proteins and their role in growth cone guidance. In this issue, two studies by Leung and colleagues6 and Yao and colleagues7 show that guidance cues induce asymmetric targeting and translation of -actin mRNA in the growth cone. This find-ing is particularly interesting because polym-erization of actin filaments provides major driving forces for growth cone motility. Thus, asymmetric actin synthesis may actually underlie growth cone turning.

Localized protein translation allows a cell to fine-tune gene expression in both space and time. Although the existence of local protein synthesis in dendrites and synapses has been well established, the discovery that rapid axonal protein synthesis is required for acute growth cone responses (collapse and turning) to some guidance cues (such as netrin-1, brain-derived neurotrophic factor (BDNF), Sema3A and Slit2) came as a sur-prise810. Early studies of cultured Xenopus spinal neurons established that these neurons respond to a large number of guidance cues, which can be classified into two groups11. The signaling of Group I cues, including netrin-1,

Guo-li Ming is at the Institute for Cell Engineering

and in the Department of Neurology, Johns Hopkins

University School of Medicine, 733 North Broadway,

BRB731, Baltimore, Maryland 21205, USA.


-actin mRNA

Cell body

Growthcone

Vg1RBP/ZBP1

Guidancecue

Actin fiber

Receptor

Src

-actin

+ +

+

+

+

P

SrcP P

P

4EBP

Figure 1 A model for guidance cueinduced asymmetric transport and translation of -actin mRNA that initiates growth cone turning. Translocation of -actin mRNA is controlled by ZBP1. ZBP1 associates with the -actin mRNA in the nucleus and transports it into neuronal growth cones while blocking its translation initiation. A gradient of guidance cues activates yet unknown pathways to promote the asymmetric transport of granules containing ZBP1/-actin mRNA complex into the growth cone periphery. Translation of -actin mRNA is regulated by phosphorylation events. Phosphorylation of 4EBP promotes cap-dependent translation initiation, whereas Src kinase phosphorylates ZBP1 to release its block of translation initiation. Together, these cellular events lead to asymmetric distribution of newly synthesized -actin and the preferential incorporation of this nascent -actin into the cytoskeleton at one side of the growth cone.

Kim

Cae

sar

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Vg1RBP-eGFP as a reporter in live cell imag-ing, Leung and colleagues further show that ZBP1-containing vesicles are actively trans-ported from the growth cone center domain into filopodia of retinal neurons upon netrin-1 stimulation, potentially carrying -actin mRNA with them. Indeed netrin-1induced translocation of -actin mRNA into filopodia was confirmed by fluorescent in situ hybrid-ization (FISH) analysis. Furthermore, these authors observed an asymmetrical transport of Vg1RBP granules into filopodia at the near site of the gradient.

Yao and colleagues took a functional approach to examine the role of ZBP1 in growth cone guidance7. They found that a cocktail of antisense oligonucleotides to Xenopus -actin 3 UTR zipcode, but not the control oligonucleotide, attenuated BDNF-induced transport and increase of -actin mRNA levels in spinal neuron growth cones. Furthermore, they observed a significant increase in the near/far ratio of -actin mRNA/ZBP1 complex in the presence of a BDNF or Ca2+ gradient. More importantly, antisense oligonucleotide treatment blocked both growth cone attraction and repulsion but not neurite extension. These results provide the first functional evidence that ZBP1 is required for growth cone turning.

The above results establish that guidance cues induce active and asymmetric transport of -actin mRNA, but is its translation regulated by guidance cues as well? Leung and colleagues6 observed a rapid increase of total -actin protein within growth cones (by ~30%) upon global application of netrin-1. Such an increase was blocked by a protein synthesis inhibitor, or functional blocking antibodies to the netrin-1 receptor DCC, or an antisense morpholino directed against -actin mRNA. To directly examine the regulation of -actin mRNA translation, the authors developed an elegant reporter construct in which the 3 UTR of -actin was linked to the coding sequence of Kaede, a green fluorescent protein that can be effectively and irreversibly converted to a highly stable red form upon brief ultravio-let illumination. Thus, analysis of green and red fluorescence signals after photoconver-sion provides a quantitative measure of both newly synthesized (green) and pre-existing (red) proteins. Imaging studies showed that netrin-1 increased Kaede-green signal in axons separated from the soma. These experiments firmly establish that netrin-1 stimulates local translation of -actin mRNA.

Growth cone turning requires a break in symmetry. Both groups show that chemoattrac-tants lead to an asymmetric elevation of -actin within the growth cone6,7. Using a morpholino

oligonucleotide directed against -actin mRNA to block its translation, Leung and colleagues showed that the morpholino prevented the asymmetric actin distribution in response to netrin-1 and, more importantly, netrin-1induced growth cone attraction. Yao and colleagues showed that asymmetric elevation of -actin protein on the side facing the gradient required ZBP1 function and extracellular Ca2+. Together, results from both groups support the idea that asymmetric localization of -actin is essential for attractive turning responses.

What are the potential mechanisms that regulate -actin mRNA translation locally at the growth cone? ZBP1 associates with the -actin mRNA and prevents its pre-mature translation by blocking translation initiation14. Such inhibition is released by the protein kinase Src, which phosphorylates a key tyrosine residue in ZBP1 that is required for binding to RNA (ref. 14). Indeed, a gradient of BDNF led to an asymmetric distribution of phospho-Src (active form) within the growth cone6. Another level of regulation lies in the cap-dependent translation initiation for -actin mRNA. Netrin-1 stimulates rapid phosphorylation of the translation initiation factor eIF-4E binding protein 1 (4EBP)4, a marker for cap-dependent translation ini-tiation. Leung and colleagues now show that netrin-1 gradient induces asymmetric phos-phorylation of 4EBP. Together, these results suggest that guidance cueinduced phos-phorylation events may underlie translation initiation of -actin mRNA.

Whereas both studies agree on the asymmetric localization of -actin in attractive turning, they differ on growth cone repulsion. Leung and colleagues examined repulsion in response to gradients of netrin-1 in the presence of laminin and Sema3A (not a Group I cue)6 and observed a slight bias but not significant reverse asymmetry of -actin protein. Furthermore, morpholino against -actin did not seem to affect repulsion to either Sema3A or netrin-1. The authors argue that -actin translation is particularly important for directional guidance toward a positive cue. On the other hand, Yao and colleagues observed reversed asymmetry of both phospho-Src and -actin levels in response to BDNF under repulsive conditions7. Importantly, the cocktail of antisense oligonucleotides against ZBP1 blocked both BDNF- and Ca2+ gradientinduced repulsion, suggesting an essential role for ZBP1 in repulsion as well. Furthermore, these authors observed a significant reduction of -actin protein with BDNF under repulsive conditions. Thus, it is possible that differences in the target of manipulation (-actin versus ZBP1) contributes to the differential findings

from the two groups. ZBP1 may regulate transport of other mRNAs in addition to -actin mRNA. Further studies are needed to firmly determine whether attraction and repulsion differentially depend on -actin mRNA transport and translation.

The following model can be drawn from these new findings (Fig. 1). Once -actin mRNA is made in the nucleus, it becomes associated with ZBP1, which transports -actin mRNA while preventing translation initiation. A gradient of guidance cues activates an unknown pathway to promote the asymmetric transport of granules containing ZBP1/-actin mRNA complex into the peripheral region of the growth cone. The same gradient also asymmetrically activates Src kinase and 4EBP, promoting translation initiation of -actin mRNA. The preferential incorporation of nascent -actin into the cytoskeleton at one side may provide the initial driving force for growth cone steering.

These in vitro studies provide substantial evidence for the involvement of asymmetric -actin mRNA transport and translation in growth cone guidance and also raise many interesting questions. First, is asymmetric -actin translation instructive or permissive in growth cone guidance? New tools need to be developed to examine whether asymmetry of -actin mRNA translation within the growth cone is sufficient to induce turning. It is not clear why new -actin is needed at all, given the large pool of -actin present in the growth cone. Nascent -actin may be more effective in the initiation of actin polymerization, and new methodology using tetracysteine tags to differentiate new and old proteins15 may allow direct testing of this hypothesis. Second, the signaling pathways that regulate transport, sta-bility and rapid translation of specific mRNAs before and after stimulation of guidance cues remain poorly characterized. Third, we know little about the role of local protein synthesis in axonal pathfinding in vivo. For example, does midline axon guidance of commissural interneurons, a classic example of netrin-1 regulated Ca2+-dependent guidance12, require local protein synthesis? And finally, as many inhibitory factors in the adult brain signal through Ca2+-dependent pathways to regulate axonal behaviors11, could modulation of ZBP1 function and axonal -actin protein synthesis help to promote axonal regeneration after injury?

1. Kalil, K. & Dent, E.W. Curr. Opin. Neurobiol. 15, 521526 (2005).

2. Harris, W.A., Holt, C.E. & Bonhoeffer, F. Development 101, 123133 (1987).

3. Brittis, P.A., Lu, Q. & Flanagan, J.G. Cell 110, 223235 (2002).

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



4. Campbell, D.S. & Holt, C.E. Neuron 32, 10131026 (2001).

5. Ming, G.L. et al. Nature 417, 411418 (2002).6. Leung, K.M. et al. Nat. Neurosci. 9, 12471256

(2006).7. Yao, J., Sasaki, Y., Wen, Z., Bassell, G.J. & Zheng, J.Q.

Nat. Neurosci. 9, 12651273 (2006). 8. Steward, O. Cell 110, 537540 (2002).9. Martin, K.C. Curr. Opin. Neurobiol. 14, 305310

(2004).10. Twiss, J.L. & van Minnen, J. J. Neurotrauma 23, 295

308 (2006).

11. Song, H. et al. Science 281, 15151518 (1998).12. Shim, S. et al. Nat. Neurosci. 8, 730735 (2005).13. Zhang, H.L. et al. Neuron 31, 261275 (2001).14. Httelmaier, S. et al. Nature 438, 512515 (2005).15. Martin, B.R., Giepmans, B.N., Adams, S.R. & Tsien,

R.Y. Nat. Biotechnol. 23, 13081314 (2005).

Hooking up new synapsesThomas Biederer

Synapse formation requires adhesive interactions between pre- and postsynaptic membranes. A new study reports that netrin-G2 ligand (NGL-2) interactions with netrin-G2 induces excitatory synapses, expanding the range of known synapse-inducing signals.

For most of us, the heyday of synapse formation is over. After all, this process occurs most intensely in the human central nervous system during late prenatal and early postnatal development. Nevertheless, synapse formation vies for attention from both neurobiologists and cell biologists. Neurobiologists are interested in synaptogenesis simply because it is required to set up nervous systems. The cell biological challenge stems from the need to coordinate synapse formation in time and space between pre- and postsynaptic neurons. Only this close coordination allows two neurons to form the asymmetric cell-cell junctions that are synapses. A study by Seho Kim and colleagues1, in this issue, provides fresh molecular insights into the repertoire of interactions between two neurons as they form a new synapse. The authors demonstrate that NGL-2 is a postsynaptic cell-surface protein that instructs the formation of excitatory synapses.

Previous work had identified two synaptic adhesion systems that not only hold synapses together, but also instruct their formation. Neuroligins, a small family of postsynaptic membrane proteins2, were the first molecules shown to induce the assembly of functional presynaptic specializations3. They act in synaptogenesis through binding of their presynaptic partners, neurexins4. This heterophilic interaction also induces the assembly of postsynaptic protein complexes57. Another synaptic adhesion system involves the cell adhesion molecule SynCAM 1, a homophilic immunoglobulin (Ig) domain containing protein belonging to its own small gene family8. SynCAM 1 drives the forma-tion of functional presynaptic specializations

that release excitatory neurotransmitter and concurrently promotes synaptic transmission9,10. Obviously synaptogenesis is too important to be left to only two synaptic adhesion systems, so NGL-2 constitutes the third example of a trans- synaptic interaction mediating synaptogenesis.

NGL-2 may literally hook up new synapses: its extracellular domain contains leucine-rich repeats that form a horseshoe-shaped hook. (ref. 11 and Fig. 1) The extracellular sequence of NGL-2 also includes an Ig-like domain, implying that this protein could serve as adhesion molecule. However, the feature that led to the identification of NGL-2 in this new study is its intracellular carboxyl-terminal PDZ (postsynaptic density-95/Discs large/zona occludens-1) domain interaction sequence. The authors found NGL-2 in a yeast two-hybrid screen for proteins interacting with PDZ domains of the adaptor molecule postsynaptic density95 (PSD-95). They confirmed that this interaction occurs in the brain and that NGL is a postsynaptic membrane protein at excitatory but not inhibitory synapses. Synaptic membrane proteins interacting with scaffolding molecules such as PSD-95 are, of course, generally good candidates for the organization of postsynaptic membrane specializations. However, a more direct motivation probably drove the authors screening strategy as well: neuroligins bind PSD-95 (ref. 12), and thus other PSD-95 binding synaptic membrane proteins might similarly function in synaptogenesis.

Knockdown of NGL-2 in dissociated hippocampal neurons, which reduces the density of excitatory but not inhibitory postsynaptic specializations, demonstrates its role in synapse formation or stability. Furthermore, NGL-2 is involved in excitatory neurotransmission, as knockdown of the proteins expression reduces the frequency of excitatory miniature potentials (minis) but not of inhibitory mini s

in dissociated hippocampal neurons. What are the functions of NGL-2 on the cellular level? Here, the authors provide two key results. First, NGL-2 promoted the assembly of postsynaptic scaffolding molecules and receptors. Exogenous clustering of overexpressed NGL-2 in the postsynaptic membrane led to the recruitment not only of PSD-95 and other components of the postsynaptic scaffold, but also of AMPA and NMDA receptors, indicating that functional postsynaptic assemblies had been formed. Consistently, NGL-2 overexpression promoted dendritic spine protrusions. The second key result of the new work is that NGL-2 induced presynaptic assembly through a trans-synaptic interaction. Presentation of NGL-2 to growing axons induced specializations that contained not only clustered markers of excitatory presynaptic sites, but also actively recycling synaptic vesicles. This was demonstrated by uptake of an antibody to the luminal domain of the synaptic vesicle protein synaptotagmin I into these NGL-2induced specializations. The adhesive interactions of NGL-2 are sufficient to achieve this effect, as its purified extracellular sequence can be presented on beads to neurons in order to induce this presynaptic differentiation. These experiments place NGL-2 into the same group as neuroligins and SynCAM 1, synaptic membrane proteins capable of inducing functional presynaptic terminals.

An open question is which presynaptic protein the NGL-2 hook catches during synaptogenesis. The only known extracellu-lar partner of the related NGL-1 leucine-rich repeats is netrin-G1, a glycosylphosphatidylino-sitol (GPI)-anchored membrane protein that is also known as laminet-1 (ref. 13). Netrin-G1 shares extensive sequence similarities with netrins, secreted proteins that are critical regu-lators of axon outgrowth in early development. Consistent with functions in axon guidance, NGL-1 and netrin-Gs promote neurite out-growth13. No other binding partners are known

The author is in the Department of Molecular

Biophysics and Biochemistry, Yale University,

New Haven, Connecticut 06520, USA.


2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



for the extracellular sequences of NGL pro-teins; in particular, none are known for their Ig-like domains. NGLs and netrin-Gs form small protein families of three and two mem-bers, respectively. The authors present evidence that NGL-1 binds only netrin-G1, and NGL-2 interacts only with netrin-G2, whereas NGL-3 at this point has no known binding partner. NGL-2 and netrin-G2 are coexpressed during early postnatal development in the same brain regions, including the cortex and hippocampus. The authors were able to confirm the NGL-2/netrin-G2 interaction on neuronal surfaces in their system, as presentation of immobilized NGL-2 to the neurons caused clustering of the axonal netrin-G2. However, exogenous clus-tering of netrin-G2 in axons did not induce formation of presynaptic specializations. The NGL-2/netrin-G2 interaction therefore is not a simple binary postsynaptic-presynaptic adhesive interaction acting on both sides of the synaptic cleft during synaptogenesis.

Presynaptic partners of NGL-2 other than netrin-G2 are therefore likely to participate in its synaptic terminalinducing activity. One may speculate that netrin-Gs, which lack an intracellular domain, primarily function as co-receptors to confer specificity to NGL-mediated trans-synaptic adhesion, whereas the unidentified presynaptic partner could trigger the subsequent steps in synaptic dif-ferentiation, recruit intracellular proteins and demarcate developing presynaptic membrane specializations (Fig. 1).

Notably, NGL proteins share remarkable similarities with the SALM (synaptic adhesion-like molecules) proteins, a family of membrane proteins independently identified by this group and another laboratory14,15. SALMs were also identified in yeast two-hybrid screens for proteins interacting with scaffolding molecules containing PDZ domains. Both NGLs and SALMs contain leucine-rich repeats and Ig-like domains in their extracellular sequences, with SALMs having one additional fibronectin domain. Functional similarities exist too. SALM2 aggregation causes clustering of postsynaptic proteins, and moderate SALM2 overexpression increases excitatory but not inhibitory synapse density in cultured neurons and promotes spines, whereas knockdown of SALM2 reduces excitatory spine density and excitatory mini frequency15. The important difference from NGL-2 is that SALM2 does not induce presynaptic specializations. The related SALM1 promotes neurite outgrowth in earlier steps of neuronal differentiation14. NGLs and SALMs form distinct gene families, which together constitute a group of proteins important for synaptic and neuronal differentiation.

What are the key lessons of this study and the next open questions? First, the work by Kim and colleagues on both NGL-2 and SALM2 underscores the central role of scaffolding molecules such as PSD-95. The question that remains is whether PSD-95 is the major endogenous PDZ-domain adaptor molecule that binds NGL-2, and how exactly the intracellular interactions of NGL-2 promote postsynaptic membrane differentiation. Does NGL-2 need to multimerize as the neuroligins4, and do scaffolding molecules achieve such an assembly? Second, this study expands the current binary view of trans- synaptic interactions modeled on neurexins/ neuroligins and SynCAMs. In the case of NGL-2, it is possible that several proteins may function as presynaptic partners during synaptogenesis. Netrin-Gs could confer adhesive specificity, whereas other partners may be responsible for intracellular signaling. The existence of additional presynaptic partners of NGL-2 remains to be shown; if confirmed, it will provide valuable insights into presynaptic differentiation steps. Third, it seems likely that the synapse- inducing molecules NGLs, neuroligins, neurexins and SynCAMs all exist as isoforms to provide differential adhesive properties during synaptogenesis. Such differential adhesion can in turn specify the process on multiple levels and may define connections between particular neurons, determine the subcellular location of new synapses, or guide excitatory or inhibitory synapse formation. In this context, it is notable that there are multiple splice variants of the netrin-G proteins. Fourth, the finding that NGL-2 acts during synaptogenesis, whereas NGL-1 and netrin-Gs are involved in neurite outgrowth, exemplifies the idea that early and late stages of neuronal differentiation can involve related protein interactions. In summary, this characterization of NGL-2 as a synapse- inducing molecule contributes to deepening our understanding of the role of synaptic adhesion molecules in excitatory synapse formation.

1. Kim, S. et al. Nat. Neurosci. 9, 12941301 (2006). 2. Ichtchenko, K., Nguyen, T. & Sdhof, T.C.

J. Biol. Chem. 271, 26762682 (1996).3. Scheiffele, P., Fan, J., Choih, J., Fetter, R. &

Serafini, T. Cell 101, 657669 (2000).4. Dean, C. et al. Nat. Neurosci. 6, 708716 (2003).5. Graf, E.R., Zhang, X., Jin, S.-X., Linhoff, M.W. &

Craig, A.M. Cell 119, 10131026 (2004).6. Nam, C.I. & Chen, L. Proc. Natl. Acad. Sci. USA 102,

61376142 (2005).7. Chih, B., Gollan, L. & Scheiffele, P. Neuron 51,

171178 (2006).8. Biederer, T. Genomics 87, 139150 (2006).9. Biederer, T. et al. Science 297, 15251531 (2002).10. Sara, Y. et al. J. Neurosci. 25, 260270 (2005).11. Kobe, B. & Deisenhofer, J. Nature 366, 751756

(1993).12. Irie, M. et al. Science 277, 15111515 (1997).13. Lin, J.C., Ho, W.H., Gurney, A. & Rosenthal, A.

Nat. Neurosci. 6, 12701276 (2003).14. Wang, C.Y. et al. J. Neurosci. 26, 21742183

(2006).15. Ko, J. et al. Neuron 50, 233245 (2006).

Kim

Cae

sar

Figure 1 NGL-2 (red) is a postsynaptic membrane protein of excitatory synapses. Its extracellular, hook-shaped leucine-rich repeats (LRR) bind the GPI-anchored netrin-G2 (blue), likely to be its extracellular partner at nascent synapses. The extracellular interactions of NGL-2 are then sufficient to initiate formation of new presynaptic specializations and postsynaptic dendritic protrusions, marked by recycling synaptic vesicles and glutamate receptors with scaffolding molecules, respectively. The trans-synaptic interactions of NGL-2 are likely to involve additional, yet unknown, presynaptic partners (beige; see text). Inset, the extracellular domain organization of NGL-2 with LRR sequences and an immunoglobulin (Ig)-like domain suggests a function in adhesion, whereas its carboxyl terminus links it to the intracellular scaffolding molecule PSD-95.

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Intercellular miscommunication in polyglutamine pathogenesisChristopher A Ross & Don W Cleveland

A new paper demonstrates that polyglutamine pathogenesis in spinocerebellar ataxia need not be cell autonomous. Mutant ataxin-7 expression restricted to Bergmann glia was sufficient to cause Purkinje cell degeneration and neuronal pathology.

Huntington disease is a progressive, fatal neuropsychiatric condition caused by degenera-tion of neurons in the striatum and other brain regions. Huntington disease and related disor-ders are caused by triplet repeat mutations (CAG repeat expansions), which encode expanded polyglutamine repeats in the particular disease genes. Since the discovery of these muta-tions about 15 years ago, many mechanisms for pathogenesis have been proposed. These include polyglutamine protein conformation change and aggregation, interference with gene transcription, blockade of axonal transport, acti-vation of caspases and other apoptotic pathways, and interference with mitochondria1,2. Most studies have considered these mechanisms at the intracellular level. However, although selective vulnerabilities of one or more types of neurons are the hallmark of each polyglutamine disease, all of the polyglutamine proteins are expressed in many cell types. Their effects are observed in many regions of the brain, with signs and symp-toms caused by malfunction of brain circuitry.

Until recently, there have been few attempts to consider polyglutamine pathogenesis in terms of interactions between cells. An elegant new study in this issue3 by Custer et al. (from the group of Al La Spada, who, along with Kurt Fishbeck, was the initial discoverer of muta-tions by polyglutamine expansion) now shows that polyglutamine pathogenesis need not be cell autonomous. That is, damage of the neu-ron whose death or dysfunction is the primary contributor to disease can be mediated via cell interactions. Furthermore, mutant expression restricted to glia, not the target neurons, can cause striking neuronal pathology. Aside from the intrinsic interest of the broader context

for viewing these disorders, this may open up new avenues to therapeutics, as signaling molecules and receptors involved in intercel-lular communication may be more amenable to therapeutic intervention than intracellu-lar changes, such as protein conformational change and aggregation.

This discovery follows a previous study from La Spadas group4 in which the prion promoter was used to generate a transgenic mouse model of spinocerebellar ataxia type 7 (SCA7). SCA7 in humans is a rare autosomal dominant disease caused by poly-glutamine expansion in the ataxin-7 protein. It selectively causes neurodegeneration in

cerebellar Purkinje cells and retinal neurons, and causes ataxia, tremor and visual loss. The authors of that study4 used the prion promoter to drive expression of mutant ataxin-7 in all neurons of the brain, except for cerebellar Purkinje cells. Even without mutant ataxin-7 in Purkinje cells, however, these transgenic mice showed severe Purkinje cell degeneration. Because data from models of other diseases, like amyotrophic lateral sclerosis (also called Lou Gehrigs disease), indicate that glia could contribute to neuronal pathogenesis, the investigators hypothesized that expression in glial cells may have been an important contributor.

Christopher A. Ross is in the Division of

Neurobiology, Department of Psychiatry,

Johns Hopkins University School of Medicine,

CMSC 8-121, 600 North Wolfe Street, Baltimore,

Maryland 21287, USA. Don W. Cleveland is at the

Ludwig Institute, Department of Neurosciences,

University of California San Diego, 3080 CMM

East 9500 Gilman Drive, La Jolla, California

92093-0670, USA.


Glutamatestimulator

Cellular interactionsin polyglutamate

pathogenesis

Nucleus

Mitochondrion

NMDA receptorsER

GABA

Microtubules

GABA-R

Neuron

IPSP

Inhibition

Survivalsignaling

Excitotoxicoxidative and

apoptotic stress

Glutamatetransport

Glutamatetransporters

Glutamate

BDNFsynthesis

and transport

BDNF

Trkreceptor

Glial cellErkAkt

Ca2+

Figure 1 Potential involvement of cell interactions in polyglutamine pathogenesis. Right, excessive glutamate, due to decreased glutamate uptake through glial glutamate transporters, causes increased calcium influx in neurons, leading to increased oxidative and metabolic stress. Left, loss of trophic or inhibitory input renders cells more susceptible to excitotoxicity.

Kim

Cae

sar

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



In the current study, therefore, the authors produced a transgenic mouse model using the glial fibrillary acidic protein promoter Gfa2, which is selective for Bergmann glia, so that mutant ataxin-7 expression was restricted to the Bergmann glia of the cere-bellum. Bergmann glia are closely apposed to Purkinje cells and are believed to provide support to them. These mice were viable and lived a normal lifespan. The mice developed a late-onset neurological phenotype with an abnormal clasping reflex and motor incoor-dination. Unlike some polyglutamine disease mouse models, the phenotype was restricted to motor abnormalities and did not include other phenotypes such as weight loss or early mortal-ity. However, when the investigators conducted pathological analyses, they found Purkinje cell degeneration and Bergmann glia pathology. Thus, Purkinje cell degeneration and its con-sequent behavioral phenotype resulted from expression restricted not to Purkinje cells, but to the nearby Bergmann glia.

It is clear from this study that pathogenesis is not cell autonomous in this transgenic mouse model. This does not, however, quite prove that this mechanism is necessary for the human disease. Ataxin-7, like most polyglutamine proteins, is widely expressed, including in cere-bellar Purkinje cells. Expression of other mutant polyglutamine proteins, such as ataxin-1, the cause of SCA-1, can cause cell- autonomous degeneration in Purkinje cells (albeit when accumulated to extraordinary levels)5.

An interesting issue relates to the type of neuronal cell death seen in the mouse model in the current study. Cells showed dark cell degenerationthat is, increased osmium stain, condensed cytoplasm and nucleus, and lack of nuclear blebbing or apoptotic bod-ieswhich may be characteristic of excitotoxic injury, in which excessive glutamate stimu-lation or excessive sensitivity to synaptic glutamate causes cellular toxicity (Fig. 1). Dark cell degeneration has been observed in mouse models of Huntington disease and in Huntington disease postmortem tissue (though most prominently in an area different from that seen in the human disease), but had not previously been described in SCA. Excitotoxicity in vivo may be caused by loss of normal glutamate reuptake in the glia due to loss of glutamate transporter activity. This would affect the levels of glutamate avail-able at the synaptic cleft for neuronal uptake; excess glutamate could then trigger an excito-toxic reaction in the neuron. The investigators examined levels of the Bergmann glia specific glutamate transporter GLAST and found that its expression was reduced in the mouse model. Thus, one mechanism for Purkinje cell

degeneration may be excitotoxicity owing to reduced glutamate reuptake.

SCA7 is a rare disease, but the results of the current study may have wider relevance for other polyglutamine diseases, such as Huntington disease, as well as other neuro-degenerative diseases. Excitotoxicity has long been thought to be relevant to Huntington disease, especially because medium spiny neu-rons, which are most affected in this disease, receive massive glutamatergic input from the cerebral cortex. Many studies of mouse mod-els expressing mutant huntingtin indicate enhanced sensitivity to toxic stimuli, includ-ing NMDA receptorinduced excitotoxicity6,7, likely involving glial participation as well.

Just as in the ataxin-7 model, mutant hun-tingtin accumulates in glia in brains affected by Huntington disease and decreases the expression of glutamate transporters8. In addi-tion, in a neuron-glia coculture system, wild-type glial cells protect neurons from mutant huntingtin neurotoxicity, whereas glial cells expressing mutant huntingtin increase neuro-nal vulnerability, perhaps because of reduced glutamate reuptake.

Other cell interactions that could facilitate neurotoxicity might include decreased inhibi-tory transmission or loss of neuronal trophic support. Loss of inhibitory transmission may contribute to non-cell-autonomous toxic-ity in mouse models of Huntington disease. Evidence for this emerged from expression of a ubiquitously expanded Huntington trans-gene. Progressive motor deficits and cortical neuropathology were only observed when mutant huntingtin was expressed in multiple neuronal types9. Also, Huntington disease mouse models show early alterations of cor-tico-striatal signaling10.

Pyramidal neurons in the cortex, in addi-tion to using glutamate as their excitatory transmitter, also secrete brain-derived neurotrophic factor (BDNF), which provides trophic support for medium spiny neurons in the striatum. Secretion of BDNF from these cortical terminals in the striatum may be impaired in Huntington disease. One normal function of huntingtin seems to be stimulating the expression of BDNF, and mutant huntingtin may interfere with this activity via a loss-of-function or dominant-negative effect11 or via gain-of-function effects on transcriptional mediators such as CREB-binding proteins (CBP; ref. 12). Loss of BDNF signaling could also occur via interference with the transport of BDNF- containing vesicles13.

A likely model is that toxic cellular interac-tions may enhance cell-autonomous damage within neurons. For both neurons and non-neuronal cells, mutant huntingtin can directly

cause mitochondrial toxicity, and transcrip-tional effects may decrease the production of proteins related to cell survival. Additionally, huntingtin may interfere with the expression of neurotrophin receptors necessary for sig-naling via BDNF and other trophic factors.

Non-cell-autonomous interactions of mutant neurodegenerative disease proteins have resonance for other diseases aside from those caused by expanded polyglutamine. In inherited ALS caused by the ubiquitously expressed cytoplasmic superoxide dismutase (SOD1), conditional gene deletion in mice has demonstrated that mutant SOD1 provokes disease by initiating damage within the target motor neurons, whose death causes the fatal paralysis. However, mutant damage within microglia, the innate immunity cells of the nervous system, sharply accelerates disease progression14. Similarly, in a mouse model of multiple system atrophy, overexpression of alpha-synuclein in oligodendrocytes of mice results in degeneration of neurons in the CNS (ref. 15). Analogously, for Alzheimer dis-ease, because each of the gene products that cause inherited forms of the disease are widely or ubiquitously expressed, and the disease largely involves synaptic pathology, cell-cell interactions seem all but certain to contribute centrally to pathogenesis. Thus, aberrant cell interactions are likely to be a general lesson in neurodegenerative diseases.

These findings have implications for thera-peutics. Effects of polyglutamine at the level of cell-cell interactions and neuronal circuits may be relatively far downstream from the immediate effects of an expanded polyglutamine protein. However, they may represent a final common pathway for many aspects of pathogenesis. The involvement of intercellular signaling mecha-nisms, such as neurotrophic factors and their receptors, glutamate and glutamate receptors and transporters, or other well-defined signaling molecules provides opportunities for therapeu-tic development via small molecules.

1. Tobin, A.J. & Signer, E.R. Trends Cell Biol. 10, 531536 (2000).

2. Ross, C.A. Neuron 35, 819822 (2002).3. Custer, S.K. et al. Nat. Neurosci. 9, 13021311

(2006).4. Garden, G.A. et al. J. Neurosci. 22, 48974905

(2002).5. Burright, E.N. et al. Cell 82, 937942 (1995). 6. Zeron, M.M. et al. Neuron 33, 849860 (2002).7. Laforet, G.A. et al. J. Neurosci. 21, 91129123

(2001).8. Shin, J.Y. et al. J. Cell Biol. 171, 10011012 (2005).9. Gu, X. et al. Neuron 46, 433444 (2005).10. Cepeda, C. et al. J. Neurosci. 23, 961969 (2003).11. Zuccato, C. et al. Science 293, 493498 (2001).12. Nucifora, F.C. Jr. et al. Science 291, 24232428

(2001).13. Gauthier, L.R. et al. Cell 118, 127138 (2004).14. Boillee, S. et al. Science 312, 13891392 (2006).15. Yazawa, I. et al. Neuron 45, 847859 (2005).

2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Walk the line: parietal neurons respect category boundariesVincent P Ferrera & Jack Grinband

Categorization of objects has been considered a function of the temporal what pathway, but a new paper shows that neurons in the lateral intraparietal area of the where pathway show learned responses based on category boundaries.

In everyday life, we are subjected to a continuous stream of visual images, no two of which are exactly alike. We might find this barrage of input overwhelming were it not for the brains ability to rapidly extract critical features and sort stimuli into familiar categories. Categorization can be thought of as a process by which we assign importance to the similarities and differences among stimuli. These distinctions can be innate for some classes of stimuli, but often they are learned. The critical values of the stimulus dimensions that distinguish categories define the category boundaries. Freedman and Assad1 provide an important insight into how these boundaries are learned and represented by neurons in visual cortex, in a recent Nature paper.

The authors trained monkeys to categorize the direction of moving random dot patterns presented on a computer screen. If one thinks of the hour hand of a clock, all the directions between 10:30 and 4:30 were arbitrarily assigned to category 1, and the remainder belonged to category 2. Such experiments always raise a concern that the abstract categorical decision could be confounded with the motor response used to express that decision. To avoid this confound, the investigators used a match-to-category task: monkeys viewed two successive motion stimuli whose directions were chosen randomly and judged whether the two stimuli belonged to the same or different categories. Hence, the monkeys behavioral responses were correlated with the same/ different decision but were randomized with respect to any particular category. This is a crucial methodological point because there is evidence that perceptual decisions are functionally distinct from movement plans2,3.

The investigators then recorded from neurons in two cortical areas that respond to

visual motion: the middle temporal area (MT) and the lateral intraparietal area (LIP). They found that the direction tuning of neurons in LIP was systematically biased away from the category boundary to which the monkeys were trained. Specifically, the difference in response to any two arbitrary stimuli was significantly greater if the stimuli were in different categories than if they were in the same category. In con-trast, the responses of neurons in area MT, which provides direct input to LIP, did not show any such bias. These results imply that, across the population, the preferred directions of LIP neurons were shifted independently of changes in the input signals from MT.

Freedman and Assad then went a step further by retraining the same monkeys on a new category boundary that was orthogonal to the original. After retraining, they recorded a fresh population of LIP neurons whose tuning now reflected the newly learned categories. Although they did not directly show that individual neurons change their tuning during retraining, this conclusion is virtually inescapable. Their results therefore provide evidence for a form of plasticity wherein neurons can change their tuning to reflect behaviorally relevant classes of stimuli. The new data suggest that the transformation from a sensory representation of motion to a categorical representation could occur in a single synaptic layer between MT and LIP, and that this synapse could be a site of plasticity for category learning.

These results are striking for several reasons. Perhaps the most significant is that visual object recognition, which is intimately related to categorization, has long been considered a function of the temporal what pathway and not the parietal where pathway4. Damage to visual areas of the inferior temporal lobe can result in category- specific visual agnosia5. Such patients might be unable to recognize fruits and vegetables or different makes of car. The most famous form of visual agnosia is prosopagnosia or face-blindness, which is often caused by lesions of the medial occiptotemporal cortex6. In contrast, damage to parietal cortex almost never results in problems with object recognition.

Rather, parietal lesions typically result in spatial neglect7, a complex disorder that involves an inattention to stimuli in the contralesional visual space. Although visual pattern selectivity has been described in LIP (ref. 8), Freedman and Assad now provide some of the first evidence for categorical representations in parietal cortex.

Although LIP is only a few synapses removed from the retina (the shortest path from the retina to LIP comprises four synapses: retina to LGN to V1 layer IVb to MT to LIP), it is gaining recog-nition as a critical watershed for transforming stimulus-bound visual activity into more abstract representations related to spatial attention, deci-sion making and movement planning. Current theories posit that LIP is a salience map that tags important locations in the visual scene for perceptual or motor tasks9. LIP is also involved in evaluating the quality of sensory information during motion discrimination tasks10. Although the results of Freedman and Assad were not anticipated based on previous work in LIP, they nevertheless make sense in this context. After all, categorization involves attending to specific stimulus features and evaluating the relationship of those features to the category boundary.

The new results point to possible mechanisms of category learning. One such mechanism is based on reward expectation (or spatial atten-tion related to expectation of reward), which modulates neuronal responses in LIP (ref. 11). In the experiments of Freedman and Assad, stimuli that were near the category boundary were associated with lower behavioral perfor-mance in terms of percent correct. This means that these stimuli were rewarded with lower probability than other stimuli. It would be reasonable to infer that monkeys had a lower expectation of reward for these stimuli.

One possible model argues that reward expectation modulates the gain of visual neurons in LIP. We call this the reward-gain model (Fig. 1a). If one assumes that a reward expectation signal varies with behavioral performance and that this signal interacts with the stimulus-driven response of direction-tuned neurons, then the preferred directions could be biased so that they cluster around an axis orthogonal to the category boundary.

The authors are at the Mahoney Center for Brain

and Behavior, Center for Neurobiology and Behavior,

Department of Psychiatry, Columbia University,

and the New York State Psychiatric Institute, 1051

Riverside Drive, New York, New York 10032, USA.


2006 Nature Pu

blishing

Group http

://www.nature.co

m/naturen

euroscienc

e



Freedman and Assad examined their data to evaluate the reward-gain model. One prediction of the model is that overall neuronal responses to stimuli near the category boundary should be weaker than responses to stimuli far from the bound-ary. The authors found that the average population response was the same for all stimuli regardless of their proximity to the boundary. The reward-gain model also predicts that neurons with preferred directions near the boundary should be more broadly tuned than those with preferred directions far from the boundary (J. Ditterich, personal communication. This predic-tion also was not supported by the data.

The fact that some stimuli are more predictive of reward than others seems not to explain the change in tuning of LIP neurons after category learning in any straightforward man-ner. However, at some level, reinforcement must be involved in category learning. Parkinson and Huntington disease patients are impaired in certain kinds of category-learning tasks, imply-ing a role for the dorsal stria-tum12, a structure involved in reward expectation13. A model (Fig. 1b) in which the synaptic strength of MT-to-LIP con-nections is modulated by a reinforcement signal originating in the striatum might explain changes in LIP tuning better than the reward-gain model. The reinforcement-learning model tends to broaden the tun-ing curves for LIP neurons such that they respond only to stimuli in one category and do not discriminate well within that category (yellow tuning curve, Fig. 1b).

So does categorization happen in LIP? It seems clear that LIP can contribute to the representation of some category boundaries. However, categorization requires more than this. At a minimum, categorization involves comparing sensory data against a category boundary. Neurons that are involved in making this comparison should encode the relationship of any given stimulus to the category boundary. For example, a categorization neuron might respond more strongly to stimuli near the boundary than to those far away. It is not clear that this

comparison occurs in LIP, as, on average, LIP activity did not vary with stimulus direction. Other work on categorical decision making has identified several regions, including the striatum, medial prefrontal cortex and anterior insula, in which activity varies depending on the proximity of the stimulus to the category boundary14. Previous work15 showed that some neurons in ventrolateral prefrontal cortex are more strongly activated for stimuli close to the category boundary. Thus, it is possible that LIP is involved in representing categorical boundaries, whereas the frontal cortex is involved in performing the comparison between the stimulus and the boundary to categorize the stimulus.

1. Freedman, D. & Assad, J.A. Nature 443, 8588 (2006).

2. Horwitz, G.D., Batista, A.P. & Newsome, W.T. J. Neurophysiol. 91, 22812296 (2004).

3. Heekeren, H.R., Marrett, S., Ruff, D.A., Bandettini, P.A. & Ungerleider, L.G. Proc. Natl. Acad. Sci. USA 103, 1002310028 (2006).

4. Ungerleider, L.G. & Mishkin, M. in Analysis of Visual Behavior (eds. Ingle, D.J., Goodale, M.A. & Mansfield, R.J.) 549586 (MIT Press, Cambridge, Massachusetts, 1982).

5. Farah, M. Visual Agnosia: Disorders of Object Recognition and What They Tell Us about Normal Vision (MIT Press, Cambridge, Massachusetts, 1990).

6. Meadows, J. J. Neurol. Neurosurg. Psychiatry 37, 489501 (1974).

7. Hillis, A.E. Neuroscientist 12

Nature Neuroscience October 2006

Documents

Transcript of Nature Neuroscience October 2006