Reversible deactivation of cerebral network components

REVIEW R c B

d R P

S

e o G eL

A

o ce at E V

n a lro J

v m

B

ee

a

y

e

tm

uis

p

Reversibledeactivationtechniqueshave shownthat the cerebralnetwork:(1) is dynamic,its functionsdependingon contemporaneousprocessingelsewherein the network;(2) is composed of singlenodesthat contributeto severalbehaviors;(3)possesses an inherentplasticitythattends to minimizelesion-induceddeficits;and (4) comprisesfeedforwardand lateralconnectionsthat contributein differentwaysto networkoperations.The next major advancesin understanding network operationswill probably be made by applyinga combinationof behavioral,neuronrecordingand deactivationtechniques. The greatestnea~term gainsare likelyto be made in understandingthe contributionsthat feedbackprojectionsmaketo cerebralnetworkfunction. N (1996)19, 535-542 e u r o s Trends

s

TUDIESOF CEREBRALLOCALIZATION of function and interactions between components in the cerebral network can be traced back to the second half of the 19th century to: (1) the observations made by Hughlings Jackson of patients with lesions; (2) the work of Goltz, Munk, Schafer and Ferrier and the experimental lesions they made in animals; and (3) the electrical-stimulation experiments of Fritsch and Hitzig, and Ferrier’. Aside from the regions with motor-relatedfunctions, the approachof using lesions has dominated examination of cerebral functions during the intervening century. In the initial studies, large lesions were employedthat correspondedto the rather coarse models of cerebralfunction prevalent at that timel. However,at the end of the 20th century our knowledgeof cerebral organization,based largely on recently appliedelectrophysiologicalmappingand sensitive pathway-tracing studies, is much more refined, and we know there is a richly and specifically interconnected network of multiple areas in the cerebral cortex>s. Moreover, we know that each area is unique in terms of its representationof the periphery, connectional signature, inventory of receptive-field properties, and contributions to cerebral functionb. Thus, basedon a multiplicity of features,each areacan be considereda node in the cerebralnetwork. Studies of the contributions made by cortical areas to network interactions and cerebral function require the deactivation of individualareas. For many investigatorsthe lesion technique is the method of choice to silence a region of cortex because the limits of the lesion can be defined with a reasonablelevel of accuracy. However,there are drawbacksto this method: (1) lesions can only be defined post mortem; (2) it is not possible to reinstate the damaged region in the network to obtain control measuresthat bracket the lesion-inducedeffects; (3) comparisonsmust be made between animals, and internal double dissociations are not possible; and (4) the mature CNS is plastic and connections can be activated, strengthened or modified following lesions (Ref. 7, and see below). Therefore, it is most desirableto use reversibledeactivation techniques to investigate the functional roles of, and interactions between, components in the cerCopyight

@ 1996, Elsevier Science Ltd. All rights reserved. 0166-

2236/96/$15.00

c

ebral network.These techniques largelyovercome the drawbacksidentified for the lesion technique. Macro-andmicrodeactivation techniques There are two major classes of reversible deactivation techniques: (1) cooling plates or probes; and (2) pharmacological agents using GABA,its agonists, and local anesthetics such as lidocaine. Applicationof the pharmacological agents results in microdeactivations of about 1-3 mm3of neuropil,which are likely to induce subtle influences on behavior. In contrast, cooling is a macrodeactivatorand a single probe can deactivate a large expanse of cerebral cortex ranging from <10mm3 to >100mm3, which has a greater likelihood of influencing behavior. Moreover, cooling deactivation can be induced within minutes, held constant for periods >1 h, and reversedwithin minutes. Also, the extent of the deactivated cortex can be defined accurately by making thermal or electrophysiologicalmeasurements” or by using the reduced uptake of 2-deoxyglucoseto define the areas of the cortex that have been cooled (Fig. 1). Both types of technique permit the elucidation of interactions between nodes in the networkby deactivating one node and examining the effects of the inactivity on neurons at other nodes. Cooling has the advantagethat it can be used to test neuronal interactions and behaviors simultaneously,whereaspharmacological techniques are preferred for deactivating deep structuresor modules such as subcompartments in thalamic nuclei, individual cortical layers, local circuits, and the regions of areas VI and V2 rich and poor in cytochrome oxidase. However,the relationship of the deactivatedregion to anatomically defined modulesrequirespost-mortemanalysisof tissue,which is an obviousdrawback.Studieshave been carriedout on a varietyof sensorysystems,althoughvisualregions of catsandmonkeyshaveundergonethe closestscrutiny and will be the focus of this article. Networkcontributions to behavior Studieswithreversibledeactivation

Only a limited number of behavioral studies have been carried out using modern cooling probes to PII: S0166-2236(96)1

OO61-8

TINS Vol. 19, No. 12, 1996

B RP a S G L a a t L f a V P D a C o A a N e B U S o M B M 0 U A E V i a t I o n P F h o M U o n L a R d B 7 C L S w J B i a C e V I 3 1 A d D L F B C F

535

B

P

e a

– Cerebra

:

a

.

y

l

R

electrophysiologically defined borders (for example, Ref. 14) are not available for individual cats. Even so, rough approximations can be made between anatomical and physiologically mapped regions: for example, vPS cortex - area 20; MS cortex - areas AMLS, PMLS, ALLS, PLLS; posterior MS (ph4S) cortex is largely limited to areas PMLS and PLLS. Also see Refs 6,9 for more details.] In these cats visual fibers crossing the midline, in the optic chiasm and corpus callosum, were severed. This permits visual stimuli to be directed to one hemisphere when an opaque occluder is placed over the contralateral cornea. A battery of tests was applied to animals prepared in this way that encompassed motion discrimination, detecting and orienting to stimuli, and formrecognition capabilities of the two regions. Except for the orienting task, testing was carried out under monocular viewing conditions whereby visual stimuli were directed to either the right hemisphere (vPS probe) or left hemisphere (MS probe). In this battery of tasks striking differences in the cooling effects of each region were observed (Table 1). For example, cooling of MS cortex, but not vI’S cortex, impaired the ability both to discriminate between moderate differences in the direction of movement of fields composed of non-systematic shapes, and to detect and orient toward stimuli. In contrast, cooling of vPS cortex, but not MS cortex, impaired the recall of recently learned object-discrimination and highly familiar pattern- and object-discriminations. Moreover, the cooling blocked learning of new objectdiscriminations and the relearning of highly familiar pattern-discriminations. The g cooling of F 1 Diminisheduptake of 2-deoxyg/ucosein the vicinity of ai coo/- complex MS or vPSe cortices ing probe. Coronalsectiot showst t c o t f middle either o h o impaired discrimination l h l of suprasylvionSUICUS (M5s) i I 1“Cby coolingprobes(blackdots)reduces masked patterns when the mask was in motion. the uptake of 2-deoxyg/u[me (2DG) in this area. Note the very /ow Importantly, the deficits caused by the cooling did not 2DG uptakethroughout a or most of the thicknessof cortexbounding attenuate responses in the same session or over many the MS~,and the h;gh 2, K uptake throughout much of cortex a testing sessions separated by several months and >100 Fort deltlih, seeRef.10.Abbreviations: e Aud,auditory c h n i t h a / a m intervening cooling sessions. These observations areu c cortex;IGN, lateralgenicti~te nucleus;17, area 17; 18, area 18. S the first to demonstrate a double dissociation, in the bar, 5 mm. T w p b B P ir W V e : aa s r p n a y s dat same animal, of the neural operations of visuoparietal a S L t o e m n p b h e and visuoternporal cortices and show that the two regions make fundamentally different contributions i then ceret ml network. v This is because e cool- to s visually guided t g behaviors. i Similar dissociations ing plates, which were adopted from electrophysiohave been made between monkey prefrontal and cortices’(’]’, and between lateral and ventral Iogical studies, arc c ~mbersome to apply to the behav- narietal . ing animal, and als(J because the highly effective and regions of temporal cortex’ 8i’~. straightforward co )ling technique pioneered by Cooling has also revealed dynamic features of the Horel [ “ in the e:irly 1980s has not been widely cerebral network (Fig. 2). For example, deactivation of adopted. Here]’s me: hod permits cooling probes made the pMS cortex in one hemisphere of cats induces a of hypodermic tub ng to be chronically implanted virtually complete neglect of stimuli introduced into and cooled by the circulation of chilled methanol. the contra-cooled hemifield (Fig. 2Bi,ii; Ref. 21). This Animals can be con !lected readily, and as needed, to neglect contrasts with the normal orienting towards stimuli introduced into the ipsi-cooled hemifield. tubes carrying the c jolant, We summarize a series of recent studies to illustrate Base(i on these observations it might be safe to predict the advantages of cc~oling deactivation for investigat- that bilateral cooling of pMS cortex would virtually ing cerebral networ}. contributions to behavior. These eliminate orienting to stimuli introduced anywhere in studies concentrate~l on visuotemporal and visuo- the visual field. In fact, the exact opposite result is obtained, for there is virtually complete restoration of parietal regions of . at cortex and the contributions they make to perc ption, attention, cognition and orienting throughout the visual field (Fig. 2Biii; Ref. 20). A similar observation was made by Sprague” mnemonic processi rig’). To compare the behavioral deficits induced by .ieactivation of these two regions, following a massive ablation of visual areas in one one cooling probe Mas implanted in contact with ven- hemisphere, and a subsequent ablation of the upper tral posterior supra.ylvian (vPS) cortex of the right layers of the contralateral superior colliculus. However, hemisphere (visuotempora] cortex), and a second the results from cooling show that the crucial region probe was inserted :nto the middle suprasylvian (MS) is mluch more localized in the cortex than we would SUICUSof the left bc nisphere (visuoparietal cortex) of have been led to believe by the ablation studies. But split-brain cats’ [AI.ltomical terms are used to describe more importantly for the” purposes of the present regions deactivated Iv cooling. This is necessary because article, we conclude from the cooling observations

n

h

i

a

B

m, S

T

o d

i v

g

b u

ei

ib c

eM a

T

a M

M O M F

o a c

F

o

v

r p

a

o p

a

o

a

b

t a

Novel objects

A

+b d

– n db

M

e

m er s

Rm R Rm R R L

mt e j e

e ev

uv

that the effect of deactivatingthe pMS node modifies the prevailingfunctioning of the network and greatly altersthe outcome of cooling the second pMSnode in the opposite hemisphere. From a practical point of view these observations show that it is clearly unsafe to predict the effect on behavior of deactivating two nodes simultaneously,even when the effects of deactivating each node individuallyare known.

p f v

vm u fr

R

n is c

v

i

e

t

n p

i

+ + + + —

l e

e

l

l

e

+ +

a D roeu id a f

R a i p 9dt s

n

e e c

li

c

ci

r a

r

a

r

c

P c snr iS

B

O

5

c

yt aor

0

1

-.

3

3

6

6

a

9 1

9 5

0

5

1

0°

(ii)

30°

30°

60°

60°

90° m 1

90° 50%

o%

5

1

5

1

0°

(iii) 6

o

90” 1

5

O

ea

a

a t

n

u“

(i)

a

a r

o

9

r

o

6

5

o

n

0°

1

e

e

Reversibledeactivationversuslesions

Comparison of the sequelae of circumscribed lesions of visuoparietal and visuotemporal cortices with cooling deactivationsof identical regionssuggest that the cerebral network is plastic and capable of substantial change following damage. For example, the neglect evident during unilateral cooling of cat pMS cortex21is not evident several weeks following the surgical removal of the same region23when the standardmethod of testing is used, and the stimulusis introduced into the cat’s visual field when the cat is @.tS for tempora~ 1esiOnstationary22 . More-specific induced deficits in orienting to stimuli following lesions of pMScortex have recently been carriedout24. These show a virtually complete neglect of the left visual hemifield initially following a lesion of right pMS cortex (Fig. 3A). This neglect attenuated rapidly and was no longer detectable three days after the lesion was made. The similarity of the initial deficits induced by lesion and cooling is attested by the capacity of cooling left pMS cortex to reversethe lesioninducedneglect(Fig.3B).Moreover,asthe lesion-induced compensations restorednormal orienting into the left hemifield, cooling of left pMS cortex induced a progressiveneglect of the right hemifield. This coolinginduced neglect of the right hemifield reached completion (Fig. 3B) at the same time as the restitution of orienting into the left hemifield (Fig. 3A). Similar shortlived deficits, indicative of substantial compensations in the cerebral network, also follow lesions of monkey areaV5 complex25-27, a possiblehomologueof cat MS cortexb. Comparison of lesion and cooling deactivation effects also provides evidence for plasticity in visuotemporal circuits. For example, deactivation of vPS cortex results in an inability of cats to learn new three-dimensional object discriminations (Table 1; Ref. 15). This type of discrimination is relatively straightforward for the cat to learn and it survives ablation of the same region of cortex28’29. The difference in these resultsindicates that existing, but poorly used, secondary circuits are activated or strengthened

ui da

c

in the absenceof vPScortex and contribute to the successful learning of object discriminations. Such circuits cannot be used and strengthened during short cooling periods of -1 h. A similar plasticity becomes

A

o

me

i

i

l

ip s f

i

t t

— — e —

ho

—

m

e

REVIEW e a

c

c

n

i

e

e a – C

COtteX

+ + +

e

i

P

F 2. Effect of bilateral cooling prabes in the posterior middle i suprosylvian(pMS) cortex of a cat. Iconsrepresentdorsalviewsof cot brain to indicateposition of coolingprobes.Theoutlinedregionin cortexindicatesposition of a probe thot wasnot operational;so/idregions indicate the coolingof the probe to between3 and ~C. Thetwo concentricsemicirclesin the radialplots represent50% and 700%response levels,and the length of eachbe/d radiusrepresentsthe percentageof responsesat eachlocationthat werecorrect.( Allcortexis warm. ( Coolingof (i) right pMS cortex,(ii) left pMS cortexand (iii) both pMS cortices.Note the neglect of the contra-cooledvisual field to stimuli during unilateralcooling,and the restorationof orienting to virtuallyall a f R 2 id positionsby bilateral cooling. F

TINSVO1. 19, No. 12,1996

5

i d a l tns , i e r

B

P

e a

C

a

o

.

y

evidc>n\ in monkq nfkrotemporal cortex when the effects of c-ooling de ctivation are compared with the effects f)f lesions (see Ref. 1S for citations to studies on monkevs). Why; arc neural ompensations not induced by cooling? The domin nt factor is likely to be the dumtio[l of the deacti~ tion. Following lesions, aninlals live permtrn~>ntlywi in the neural defect and there is Considt’rable opport! nit} fol prolonged interactions betwmn the animal vith the defect and the envir(Jnrnent. “I”hese intc’ra tions might result in nlodificatit)[ls 1n the rem. inirrg circuitry that reduce the severit> of the bar .iical). In contrast, these same influc’li{es have Iittl~ time to act on the nervous system during coolirlg, which occupies only 5–1 O(XJof each da}7, and ther(. is Iittle or no strengthening of seconcia]”v circuits. urthcrrnore, any compensatory changes initiated d!lring brief cumulative cooling peri(xts do not seem {) accumulate over repeated cooling sessions, becaut either the changes are very minor or they are 1 versed during the much longer int(’1’Vill f when the brain is functioning normal Iy. IIovvin(:t, it is imlm t;int to recognize that there arc exce[)tlons to this ;enerzrlization that have been scientific’cl in stud ie of learning and memory. For exanip[t’, if the engl m for certain type’s of form Liiscriminations is firm ~~ established by long periods of train ing prior to CO(~ing, it is possible for the cats to re-cstat)l ish high 1(vels of performance on the discrim irlations during ooling (Table 1; Ref. 1!5). In this irrstan~c, the initial xml ing deficit might bc one of and it is this initial access diffiaCCL’>5 [ O th(’ L’I]AJrii]l cult~)which is ovcro me with training during cooling, and it sho LIId not c confused with new learningr pc~.SC,of the term di: .rirninations. Regardless of interpretation, these resu ts provide evidence for plasticity in t I]( Cerebral nctw )rk. Moreover, it is clrar that to undc’l”standthe fLIntCiomd bases of this p[asticity, it wilI iw iinportant t compare the effects of cooling an d aI)lat ing Lfist:: It nodes in the network on neurot Ial properties. N

ie

i tn u n

t

Nt’UrOll–Ilt’LlrO1l ill: eractions are based on rich intr!rconnec’tions betwec components, or nodes, in the cerehra I network an i no node can be treated in isolatiol~. Ll)nsequenti> a firm grounding in the anatomy and t’unctional sigl~ of connections is essential for adequate interpretat ans of the effects of deactivating one notie on anothv These effects might be mediated directlv via primary’ ~~rojcctions or indirectly via one m rn[)le secondarv r ,utes involving connections with other nodes in th{ network. Moreover, there are multi [he local ci rcuif. within each node that have profound influences on uansmittecf signals (for example, see Rc’i\ 30-33). ((~ lsequently, influences identified usinx deactivation t, chniques are net influences that reprew’nt the sum ot vffects on the composite pathway ]inking two nodes. A

p

si ~ t r c

network c o

e

j

}f knowledge of the composite pathw:]v is hi~hligh ed by deactivation of gcniculocortlca 1’projc>ction~. the prir~-lary7 input to the visual cortlc~l network. In ~non keys, deactivation of rnagnoce[luia~”, but not p, !woceliular, Iayel”s in the iaterai geniculate nLlcieus (LC,N”)strongly attenuates the rmp(lnse> of neuron ~~1 in area V5 (Ref. 34). Anatomical ‘1’lle importance

l

R

studies have shown that this region is connected polysynaptically with the magnocellular layers of LGN. F.quivalcnt studies on area V4, which was thought to receive signals only from the parvocel]ular layers of LCTN,produced surprising results because deactivation of either parvocellular or magrrocellular layers clecre:iscd the activity of area V4 neurons”. These resu~t!s argue against the suspected straightforward mapping of magnoccllular and parvocellular pathways on to the parietal and temporal streams of processing in cortex. Moreover, they are supported by anatomical studies showing the c~nvergerlce of pathways emanating from both magnocellular and parvocellular layers in the L(IN on to neurons in areas VI and V2 (see Ref. 35), and confirmed by electrophysiological studies demonstrating the convergence of magnocellular and parvoccllular signals on to neurons in area V1 (Refs 36–:38), ‘1’he irnpor-tance of knowledge of composite pathways is emphasized further by studies of the geniculocortic:il system in the cat, where a different arrangement of connections exists between the LGN and the visual cortical network”. In cats, deactivation studies on individual layers in the LGN showed that layers A and A1 are necessary for activity in layers IV and VI of area 17, but not for activity in layers 11–111,where the receptive-field properties were largely normal even in the absence of activity in the deeper layers3’).This is surprising because these deeper layers were thought to be the primary input to the superficial layers~()~l. Subsequent deactivation studies revealed that the sustaining signals to layers 11–111in area 17 were transmitted via area 18 (Ref. 42), which also receives input from the LGN (Ref. 6). Feedforward projections within the network

Anatomically, feedforward projections within the cerebral network originate from superficial cortical layers alone or in combination with deep layers, but not layer IV. They terminate primarily in layer IV. The influence of these projections can vary from substantial to negligible depending upon the weighting of the composite neural pathway from one node to another r , o e : n : -t relative to other pathways that converge on the same target node. For example, there are differences in the subcortical visual circuitry leading to monkey parietal and temporal cortices, and silencing of area V1 has different repercussions on the two regions. For example, cooling deactivation of area V1 silences virtually every neuron in the retinotopically matching parts of areas V2, V3, V4 and infemtemporal cortices[’~]-~s. These results show that there are no secondary subcortical pathways bypassing area VI that transmit sufficient signals to activate neurons in areas V2, V3, V4 and irrfcrotemporal regions in monkeys. The opposite conclusion is reached for some neurons in area V3A and many neurons in area V5. In these two areas only a fraction of neurons are silenced by tlhe deactivation of area VI (Refs 44,45). These results show that subcortical inputs bypassing area VI neuronsbi i e are capableer of activating n c some e extrastriate d ht independently of area V 1. Moreover, the residual components are capable of generating direction selectivity, independent of signals that normally enter the cerebral network via area V 1. The secondary, subcortical circuits probably involve the superficial layers of the superior colliculus and the pulvinar nucleus(’q~~s. In cats, cooling of area 17 silences 25(XJof neurons in

n

B

P

e a – C

e a REVIEW

c

area 21a and depressesthe activity of many others4G.These results are understandablebecausearea21a receivesprojections from visualthalamus and a host of cortical areas besides area 17 (Refs 6,35). Lateralconnections withinthenetwark Modulationof ongoingactivityis a characteristicof lateralconnections. Anatomically, lateral connections originate from superficialand deep Ablation layers, and terminate as columns spanning severalcortical layers.On thesegrounds,areas17 and 18 in the cat can be consideredto be laterally and reciprocally connected, a conclusion validatedby the substantial parallel inputs from the LGN to both areasb.Becauseof the parallel inputs, cooling deactivation of area 17 has only minor modulatory influences on neuronal activity in area 18 (Refs47,48) and vice versa39. Similar resultshave been identified for interactions between the S1and Cooling Ablation L R L R L R L R “ L R S11areas connected by parallel inHemifield puts in cats and rabbits4950,which Day 1 Day 2 Day 3 Day 17 Pre contrast with the substantial effects of deactivating area S1on the F Ability of contralateral cooling to reverseablation-induced i neglect. ( H s t pi o and on sequentialdays(days1–17)followingablation P of right orientingresponses before o( r r pMScortexalone feedforwardprojection to area S11 c (greenhemisphere).Thereis a high proficiencyin orientinginto both right (red)and left (green)visualhemifieldsprior to in monkeys51. The transcallosal pathway of cat ablation. On day 1, thereis a profound neglectof the left hemifield,whereasorienting into the right (red) hemifieldis areas 17 and 18 links homotopic norrna/.Byday 3 themagnitudeof the defecthad attenuatedgreatl~ and remainedlow on subsequentdays(for example, regions in the two hemispheres, on day 17). (B) Withconcomitantcoolingof /eftpMS cortex(bluehemisphere)on day 1, orientinginto the left hemifield was restored(greenbar) and there was a s/ight decreasein the number of trials in which the cat oriented to stimuli and it has numerouscharacteristics presentedin the right hemifield (blue bar). Thisnumber decreasedsubstantiallyon day 2 and becamea prafound of lateral connections52.In this sys- neg/ecton day 3, whichpersistedon all subsequentdaysof testing(for example,day 17). Data takenfrom Ref.24. tem, cooling or pharmacological deactivationsmodulate,but do not silence, neuronal activity in the cortex of the opposite responsesto stimulation by different types of stimuli hemisphere that receives the callosal fiber projec- activating both the center and surround5G.This sugtionssz’53.Presumably,neuronal activity is modulated geststhat the main effect of the feedbackfrom areaV2 only because a major, parallel driving input through is to increase selectivity of neurons in area W for small stimuli activatingthe receptive-fieldcenter. This the LGNremains54. is in keepingwith the process of adaptivefiltering to Feedback projections Anatomically, feedback projections originate from increase the salience, or highlighting, of features in deep layers alone or with a contribution from the the activitymap in areaV1. This type of processmight superficiallayers. However,the major defining char- be a feature common to many parts of the cerebral acteristic is terminations in layers I or VI, and occa- network, and it might be essential for redirecting sionally other layers, but not layer IV. Also, contrary attention and retrieving image codes from memory. to feedforwardprojections, feedback projections are Even though these types of studies promise to reveal not visuotopically organized35.To our knowledge valuableinformation about the roles of feedbackpaththere have been only two studies of feedbackprojec- ways,it is important to recognizethat so far they have tions in visual cortex. In one of these, cooling deacti- been performed on anesthetized monkeys. Because vation of monkey area V2 revealedexcitatory as well higher-order areas in the network are the origin of as suppressiveinfluences on area VI neuronsss.In the feedbackprojections, it is likely that major advances second study, which focused on center-surround in understanding feedback projections will only be interactions, the main effect of GABAapplication in made when studies combine deactivation, neuron monkey area V2 was to increase the response of the recordingand behavioraltasksto examine the roles of surroundand to reduce that of the center, thus modi- attentional, mnemonic and mental-imageryfunctions @ing the response of the neuron to simultaneous on neurons in receipt of feedbackprojections. center and surround stimulation (Fig. 4). This effect DescendingProjectionsfrom the cerebral network Anatomically, descending projections from the appears to be mediated by a push-pull mechanism: excitatory input from V2 neurons with activating cerebral network have their origins in deep cortical regions overlapping the receptive-fieldcenter of the layers, and they terminate in thalamic nuclei, midVI neuron, and inhibitory inputs from V2 neurons brain or more caudalstructuresalong the neuraxis.The with activating regions located in the surround. In magnitude of the influences mediated by these progeneral, deactivating V2 increased the similarity of jections varies from small to substantial, depending TTNSVO1. 19, N0. 12,1996

539

o

REVIEW

B

P

e

a –C

a

cooling

.

y

l

R

n !,,,,.,,..

70

Control GABA

L:3 60 50 ‘* 40 m & $ 30 20

J

LJ

L

0

500

o

—

500

500

o

t(ins)

ON

——.

———

————

———— ——

0— ———— ———

—

——

El— ————

—

— — . .

F 4. Peristimuluxtime histograms of the responseof a s i orea W neuron befare, during iand after the application of GABAi a V2. Theresponseswere((’cordedin an anesthetizedparalysedmonkeybefore(control, r blue line), during (GABA,red line and hatching) and aftw (recovery,greenline) injectianof 60 nl of 100 mM GABAfrom eoch of threepipetteslocatedwithin 1.5m of eachother in area V2. Thepipetteswereplacedin a regionof area V2 that correspondedretinotopicallyto the recordingsite III area VT. Responses and the effectsof GABAinjectionsare shown for threestimuli flashedON at time zero and OFFS00ms later. During testing,stimuhwererandomlyinledeaved.( Stimuluswasa singleshort bar of optimal orienta[)onpositionedwithin the receptive-fieldcenter.( A setof oriented short bars locatedin the surroundalone, (C) A set of orientedshort barspositionedin the centerand surround.In A the centerresponseis suppressedstranglyby appkcationof GABAin area V2; whereosin B responseta :rtimulationof the surroundis substantiallyincreasedduring GA6Aapplication(red hatching). In C responseto stimulation of center and surroundbecomesdominated by siqnolselicit(~dby the surroundduring GABAapplicationin area V2. Bin width, 2 R f R e 56. 0 p r r eo m d

.

ceptible effect on neuronal activity in the superficial collicular layers, even though there are massive projections linking the two structures”c.These results imply that the collicular activity is maintained by the massive number of signals arriving from the retinae. However, in the same preparations many intermediate and deep-layer neurons are silenced. Because these layers do not receive direct projections from area VI, it is likely that the effect of the cooling on the superior colliculus is mediated via cortical areas silenced by the deactivation of area VI and which project to the intermediate and deep collicular layers. By way of comparison, cooling of cat primary visual cortex silences neither other cortical areas nor superficial colIicular layers’’f’7. These observations reflect the greater emphasis in cats of the parallel ascending projections to extrastriate regions that bypass primary co;texe’s. g visual n C

r

e

g

o

The great complexity of the cerebral network demonstrated by anatomical studies shows us that it will take a great deal of time and effort to evaluate the contribution individual nodes make to network activity, cerebral functions and visually guided behavior. Moreover, u s o fc the network is dynamic, and interupon whether the nucleus receives substantial signals actions between nodes are not fixed’but are influenced by contemporaneous processing elsewhere in the netfrom other sources. Numerically, prim.u’y visual cortical projections to work. It is likely that the next substantial advances in understanding the cerebral network will be made in the LGN outweigh the retinal inputs by approximately studies that combine electrophysiological assays of tenfold. Even so, {.ooling deactivation of cortex appears to have o]]ly limited influence on LGN the effects of distant deactivations while the animal is in the’ process of attempting to carry out stereotyped neuronss7-59. This apparent weakness of the cortical projection might be artificial and reflect the use of or learned behaviors. Such a complex experimental paradigm is within technical reach and will be particustimuli that are not potent activators of the cortical feedback system. Other studies suggest that global larly useful for investigating neuronal contributions to context, which is gei~erated by cortex, modifies local perception, motor behavior, attention, cognition and encoding to improve and synchronize signal trans- learning, which are all mediated via cerebral network mission through the LGN by groups of neuronseO)”l. interactions. Subsequent advances will rely on the use Similar conclusions ilave been reached for auditory of multiple deactivating probes either alone or in and somatosensory cortical feedback projections to combination with multisite neuronal recording to thalamic neurons8[’z”: For example, cooling primary tease apart subcircuit contributions to network operauditory cortex both modifies responses of neurons in ations. Such studies will provide powerful data, identify important nodes, and lead to a more global view the medial geniculate nucleus (MGN) to white-noise bursts (Fig. 5A; Refs 8,62) and alters the strength of of cerebral network interactions. coupling and synchroi~izationbetween neurons (Fig. 5B). Selected references These observations Imply that the cortex mediates A History o 1 Finger, S. (1994) in Origim of Nearmcie~r~”e: Explorationsinto Brain Farrctian,pp. 38-41, 86-89, 126–129 and an adaptive filterin~ of signals passing through the 195-199,OxfordUniversity I’less thalamus’z’”~. 2 Rwkland, K.S. and l D.N. (1979) B %R 1 Finally, studies of the effects of deactivating de3 _ scending projections from the cerebral network to the 3 F D.J. and Van e Essen,D.C. (1991) l Cereb.Cortex 1, l 1 superior collliculus re-emphasize the importance of 4 RouiUer,E e a ( .Exp.B R 8 4 . knowledge of composite neural circuits when interJ.W. andYoung, C B 3 a1 1 5S cM ( preting the results of deactivations. For example, cool6 P B ( Cereb.Cortex3 1a 1 . 7K J ( i Brain a M M a1 a ) o .m ing deactivation of a[-eaVI in monkeys has little per-

540

TINS Vol. 19, N(I. 12, 1996

B

P

e a

C

e a REVIEW

c

o

A . (

i 125

I 1

‘i’ CO 0

2

3! “

(ii)

0 125

0

u“-”” . . .... ... .... ] During )

~

p

7

I j “

(iii) 0 125 -1

0

o

l

-

(“f$’d

500

t

........ ..... \R

I

1

. ●*

-

.

e

‘o& f

c

o

(ins)

38°C

38°C 5°C

F

5

C

m

o

i

t

c ro m

g

t

ind

e(

i

( e Peristirru.dus u r c ut

histogramsof l rMo a sing/eneuron d ga cfi h in the a MGfdutoc tpresentationof ( li i g a t i

200 m (horizonta/bars)binaura/ white-noiseburst 20dB abovethresfro/dfor neuronactivation.(i) Complexpattern of responseto both onsetand offsetof noiseburst. (ii) Complexchangesin responseduring deactivationof primary auditory cortex.Bothlong-latencyON and short-latencyOFFresponsesare increasedby coolingof audishowthe alterationsi couplingbetweenpairsof spantory cortex.(iii) Thereis a /arge/ynorma/return of activityfo//owingterminationof cooling.( Crosscorre/ograms taneoudy activeneuronsin the MGN before,during and after coolingblockadeof neurona/activity in primary auditory cartex.Themost significantchangesare indic i i p / c tg o a / o for i the cuh c os rm/ s h t o a oneurona/ tc chl sactivity dmand ri uthe coupling u oi n i r cated by snra//arrows. T r t i betweenneuronsT1and T2. Thediagramsincludea commonsourceinput (opensymbof)and an inhibitory element(Jilledsymbof).In the three m elementand its connectionsrepresentdifferentpopulationsof neurons.(i) Thenarrow uni/atera/peak on the /etl sideof time zero indicatesa direct excitatoryinf/uence of neuron T2 an neuron T1. Thislinkageis not affectedby coo/ingdeactivationof primary auditory cortex,However,the increasedsynchronizationof the neuronpair during cooling (arrow) suggestsa tonic inhibition of a commoninput sourceto both neuronsT1 and T2 when the cortexis normally active. ( Theappearanceof a unilateral peak during cooling (verticalarrow) suggeststhat neuron T2 exertsa direct excitatoryinfluenceon neuron T1 which is dampenedby a corticallycontrolled h greater activity of neuron T1 increasesthe activity ofi and sustainedpresynapticinhibition. T presumedpresynapticinhibition is reducedby cortica/coo/ing,and neuron T2. Note that a commoninput sourceis not affectedby corticaldeactivation(horizontal arrows).(iii) Directexcitatorycouplinq of neuron Tl with neuron T2, c oo r o which is dampenedby corticallymediateddisynapticinhibition.”Theinhibitory blockadeis reducedduring c

8 9 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2

o f ( N J L W e M Nu a e L r Gc e p 2 O U 0 P dn x 6 e a ( E Brain Res.86, 1 V A i .5 Cereir. L S e a ( o Cortex6 6 m Comp.N n 3 1 V W e a a( d H J ( P P 1 o2 h 1 s H J ( i Lesionsa To r1 M ai N ( e 7 ( P Mu c p 9 r A Cv P r M R ( J C P Py 4 41 oh T R e a ( i C SensoryOrganization:Multiple 1o u VisualAreas(Woolsey, C.N., c p 1 H P L S e a ( Po N A S U S A1 9m 1 6 5 G P a A o G (l Nl 2 e 6 1 3 Q J a F uJ ( Cereb.Cortex3 i u 1 e a ( B B R 1 e H J o 2 2 B R 6o 1 H J ( B 1 e L S a P Bo ( V N a 1m 1 1 4 P N Aa S U S A 9 1 P B e a ( S J ( S p 1 1 1 cr 5 H S a S B ( a / Comp. NeuroL t 2 5 2 7 P B ( E J N Sa 9 4 1e 5 8 N W e a ( e ] N w e 1 Y D a W a R ( ] mN u6 e 6 5 1

o

2 oi P Wa .b a ( e . e tn u4 p T Gan yM Za ss Cerefr.Cortex i t n r r1 c .g ige i 2 r –v sf p 4 e 2 ore 7 ) 2 s 2 C i 0 E A ( x J9 aC P m h9 s - op 6 l Pl .9 41 a p Py 9 y 2 C P a W b oJ ( 3 l J Comp.P a em P G 1 h wr r s . . 7 9 6e 3 5 6u –e r f 0 5 u1 f C (s 9 Nc ie3 3o 8 i 2 A- h 1 l a 6l o . 3. 5 By r J ay G9 o4 2 bl J Pp 3 y l6e t ya1 9 3. E nr U e eas9 ( a9n 1 hA n . 5 sh tl o r nl o3 C 7 J c des a ( oN c oa–r 4 1e i . 1 . u rdep 1 . e 9 K a S e A (r E Brain i Res.87,521-529 i 1s .l . e 3 G M J 0r ( a s 5Ar c ERn s34. M e 7 W 9 ay r ye 8 - u .. s i i1 h5 6 u r 9 o . N r1 3 9 e t s 8J i c 3 S d P ua - B J ( r P a Ru 7m 1. p 3h 1 A- l ie c0 a y k o0 3 M . r J a e ac ( b 9 Ja N cl 4 e1 o el 1t uG 9a T 1. a r wp 3rt 34 N T a M eJ ( 1a ] N 1a 6u. . e 1 5n l 2 . 0 69 6 tn d1 x a . m a n a S t9 C d 7 n E u3 S A a C - a E ( aN w 3 6 4 l C a1 f a 1 4 l . 31 nM Js (2 . Jt N a 1 lu 9n 4 e t5 a2 n c r.e99 - Mop a o 9o a S 9 i Whe ( i a 371-394 4. M r r U a 9E e BraintRes.33, 8A 3a 1 z s d l n l v 9l o ts 4 91ni 5ua 9 1 5A 41 5 . 6Lr r J e a9 ( he J Comp.N e1 a - e e. m1 lo 29 tu R3ru a . 4 M M. b y aiJ ( n 9 Science as2 r 1eg .1 e iM a nG P a M- c J ( Brain ha 366-369 1N 1i . l I .5 l 43S 1 Res.126, 44. G N9a , .o 9 o e pt 0 9 r 3y C c ( ce l2r n4 2 u o 1 ar9 R 9 c o H I Ve 9 6 4 .B Ja 4e ia ( 9 5 i Cerebral u 4C g ortex( e 1 ( 1l6- A aM u n P1l 3 t kSl 8 N K S e . p eo 3 Pn9 d P c 70 r i 1.R r iCd y1 E J P 4 c1 h1 3F 4 M A e a (-i h S 5 4ay 4 . S uuy H (u 9 J N 4 e 2h pn 1o u 9 r a0 sN r R r e pe f o 9 6 9 5s 8s u 4 C . 2 us C e a ( a 9 V r N o 1e m 8Ti- a o5 e n s o ui o r Gu e a ( s. 1 pu rt S k . r9h0 6H 4 aM 1. rr] Nu n9 o6a e 7 r 0 . er A e a (J Nu 7 e1 6 1 ur om 5 T 7 71NS Vol. 19, No, 12, 1996

5

REVIEW

B

e a – Cerebral

P

a

oolmg

.

y

l

R

n ,,,

5

Pcrns,

5 5

a (1 / N 9 B B eta ( 1 i“ N B ( 1990) L’is{d N J a S aP J. e a / Pby.\Jo/. Puris etal. ( B E J a M

‘1’.1’. et

P

c

B

S P

.

S 56 Bullicr,

G M 1 S

a

S

R

6

6

S

,

f

e6

1

11, 189-197

a4 N c

( (in

press]

R

2 R

(1

/ 6

5

4 ( B

E

eu

i e n4

: e

3

P R

1

3

S

1l

Mr V .S y

a

We e

W

t

I t

r

m

a 5

(

etac

]

N

e(

f 38f

V

A

4 ] l 8 ~ 9y N lC 11, i a 6 4 -l f i C .fn 1 .n l ’ f4 S m v x [m An d b C w Puma R l p Hl l) 100, 2 r 3 e f , i N : 6 e s 3 c1 n 5 e1 w 9

t

A H

o t

f

J

(. d l[ fZ d pT [ e ~ m h .Imprivite wd i o ne E B Rl 63 Ghosh,S. etd ( Y N 8 r ce a ( a D b F e a ( Cy eta ( N S P e a (1 N C

r

Sylvie R

A9

,,,...

e

E

xh

ea g

i g

o E s

a

a

ni e er e

o an

a o sl

tl gt

o

t

t

m

m r

n p ac g definesthe rostrocaudal axisofe h

the tectum.Various experimentsthat causeectopic Engrailedexpressioncause predictable readjustmentsof the retinotectalmap.The newlydiscovered ‘realisators’of the retinotopicmap, suchasreceptortyrosinekinaseIigandsELF-l andRAGS couldbecontrolleddirectlybyEngrailed. Indeed,recentresultsshowthat Engrailed regulatesthe expressionof theseligands.The Engrailed gradientitselfappearsto be setup by signalsincludingFGF8 andWNT 1,allowingusto beginto tracethe molecularcascade that isresponsible for the correctwiringof the visualprojectionback into the earlyembryo. T

,

(

r

9 5N

e

w1

4

r

r

29 l

m

c

- c

9

i

-

D

U

pression of downstream ‘realisator’ target genes which, in some cases, are cell-surface molecules and signaling factors that give local molecular character to the tissues expressing them. Indeed some of the targets of homeotic genes in both flies and vertebrates are membraneassociated molecules that affect axonal growth and guidanceA-7.Over the past few years, the discovery of several patterning transcription factors (such as the ones :schematized in Fig. 1A) expressed specifically in the vertebrate brain have been described (for review, see Ref. 8). The expression of these genes in different subregions of the forebrain and midbrain has strengthened the hypothesis that the chordate brain arises from an ancestral regionalized structureglo. The mosaic of the expression pattern of these transcription factors covers the brain like a multicolored map of the neural territories, and provides a rich source of information which could in principle control the local cues that guide axons, in the brain (Fig. 1A). Intriguingly, it has been noticed that boundaries of homeobox gene expression in the embryonic vertebrate brain often coincide with the tracts of early projection neurons in the brain] 1-l:l.In fact, in zebrafish, i t the where expression pattern of these homeobox genes has been experimentally or genetically altered, the pioneering tracts of brain follow new routes along the changedi boundaries r of homeobox gene expres-/i a sion 141’.Recently, the idea that homeobox genes conr b trol local information responsible 0
542

TINS \“ol. 1‘L ?4[>.12, 1996

Ritmu is IIt lb(>Labor(ztofre de NewwchimieA?tatomie,lmtitut de.sNwams(ienm, , Berrlard, 9q 75[M5Paris, Fran(-e,and S

W

A H

i L t

IIept

B

9

G

D

Utliwrsityo C

S

D (

L [ 9

EVELOPING RFi!’INAL GANGLION CELLS send out axons that ll~vigate over a variety c)f different brain regions to arri~e at their distant specific targets in the optic tectum w thalamic nuclei (Fig. 1A). When they arrive, they obe’. a general rule of organization of connectivity based 01 topography. Thus, nasal retinal axons project to cau, ial tectum while temporal axons project to rmtral teclum (Fig. IB). In the orthogonal dimension, dorsal relinal axons project to the lateral tectum while ventral axons project to the medial tectum. That axons foil )Wtheir normal course along the optic tract when th tectal primordium is ablated], and are predictable} deflected when the neuroepithelium of the presul nptive optic tract is rotatedz, suggests that optic axon< follow local positional information rather than lon~-range diffusible cues in the brain while navigating to t!le tectum. Similarly, experiments in which the retina 01-parts of the tectum are displaced or rotated argues that topography is established by retinal axons responding to local cues within the target, findings that originally led Roger Sperry to formulate his theory of chemospe(ificity;. Because retinal axons respond to local positional information in the neuroS a epithellum,u both on the way to and within their targets, it follows that there are differentially expressed molecular-guidance c{les on the surfaces of the cells over which i the retinal axonsa grow. do [ A key question is how l r these guidance molecldes get expressedat the right place?

(.kpyright

f{> 19Y6, I:,lsebvcv Suvuce

[Id

,111 (ri{;hls rmtvwd.

0166

22.36/96/$

15.(10

1111:Sol 66-22:36(96)

10062-X

w

i