The neural network of the basal ganglia as revealed by the study of ...

The neural network of the basal ganglia as revealed by thestudy of synaptic connections of identified neuronesA. David Smith and J. Paul BolamThe study of synaptic connections in the electronmicroscope has established an 'elementary' circuit forthe neostriatum which consists of a pathway fromcortical areas (neocortex, hippocampus, amygdala) tomedium spiny neurones of the striatum that also receiveconverging synaptic input from midbrain dopamineneurones. The striatal medium spiny neurones areprojection neurones and they form synaptic contactswith output neurones in the globus pallidus andsubstantia nigra reticulata. In this way, dopaminergicafferents can directly modulate the flow of informationfrom cortical areas through the striatum to the 'premotor'areas of the brains/em and to the thalamus. It isproposed that certain parts of the striatum can themselvescontrol the activity of midbrain dopamine neuronesand so one part of the striatum can 'gate' the flowof information through another part.Studies of the fibre connections of the basal ganglia 1 . lhave shown that the neostriatum (caudate nucleus,putamen, nucleus accumbens and olfactory tubercle)receives input from the entire cortical mantle, includingthe allocortex, but that its major output is to justtwo regions - the substantia nigra and pallidum. Therecognition that information originating in differentparts of the cortex may remain segregated in severalparallel pathways that pass through the striatum tothe pallidum or substantia nigra 3 • 4 (see also article byG. Alexander and M. D. Crutcher, this issue) haschallenged the traditional view that the striatumserves as a kind of funnel through which infom1ationfrom the cortex converges onto a lin1ited number ofoutput targets. Accordingly, when we consider theorganization of connections at the synaptic level withinthe striatum, we no longer have to look for a networkthat leads to a high degree of convergence. Instead,we can ask the question, is there a fundamental neuralnetwork that occurs in each of the parallel channels?In other words, is the striatum organized as a numberof parallel 'central processing units', each of whichcarries out essentially the same operations on informationfrom different parts of the cortex? We shallsee how far it is possible to answer this question fromultrastructural studies on the synaptic connections ofidentified neurones and will go on to develop aworking hypothesis that goes beyond current knowledgeof synaptic connections.Importance of the study of identified neuronesI t is nearly two decades since the publication of aseries of papers by Kemp and Powell that havebecome a landmark in our knowledge about the basalganglia s . In these papers, the organization of thestriatum was described on the basis of studies at thelight microscopic level by the Golgi method and at thesynaptic level by electron microscopy. The commonestcell type in cat striatum, also described byRaman y Cajal in 1911, was a medium-sized neuronewhose dendrites were densely covered in spines(medium spiny neurone); this neurone comprised 96%of all Golgi-in1pregnated neurones in the caudatenucleus of the cat (Fig. 1). Because the Golgiprocedure led to in1pregnation of extensive local axonarbors of these neurones but did not reveal any axonsleaving the striatum, Kemp and Powell consideredthat the medium spiny neurone was a local-circuitneurone in the striatum and that the main projectingneurones were larger in size and comprised less than3% of the total population of impregnated neurones.On the other hand, their electron microscopic studies,in which anterograde degeneration was used to,7J. Paul Bolam are atthe MRC AnatomicalNeuropharmacologyUnit, UniversityDepartment ofPharmacology. SouthParks Road, OxfordOXI3QT. UK.A. David Smith and.. i\lFig. 1. A typical medium-sized, densely spiny neurone of the neostriatum asrevealed by Golgi impregnation in the rat. This example shows the morecommon type of very densely spiny neurone that has been shown to project tothe substantia nigra and to receive synaptic input from the cortex and fromdopaminergic terminals. A second type of less densely spiny neurone receivesinput from the parafascicular nucleus of the thalamus in the rat.Reprinred/romTrends in Neurosciences - July 1990Vo!. 13, No. 7 © 1990, Elsevler Science Publishers Ltd , (UK) 0166 - 2236/901102 .00 259

Spines 59 0 1035 0 10dopaminergicaxonSubstantia nigraFig. 2. The distribution of tyrosine hydroxylase·immunoreactivesynaptic boutons in contact with identifiedstriatonigral medium spiny neurones in the rat. It can beassumed that these terminals, which form symmetricalsynaptic contacts, originate from midbrain dopaminergicneurones, and so the results indicate that one of the mainfunctions of dopamine in the striatum is to influence theflow of information from neurones that terminate withasymmetrical contacts on the same dendritic spines. It hasbeen shown that such terminals originate in corlical areas,including the neocorlex and hippocampus. (Data fromRef. 17.)detennine the origin of synaptic boutons, led them toconclude that the medium spiny neurone was themajor target of inputs from the cortex and thalamus.Because the medium spiny neurones were consideredto be intemeurones, it was assumed that they relayedthis input to the larger neurones that were believed tobe projecting neurones. Kemp and Powell concluded'consequently, the information passing to the globuspallidus and substantia nigra represents a highlyintegrated and modified version of the input to the[striatum],. This conclusion was fully in agreementwith the authoritative view of C. and O. Vogtexpressed in their magnum opus of 1920 (for anexcellent historical review see Ref. 6). However, aswe have already indicated, this view is difficult toreconcile with the new evidence of a series of parallelcorticostriato-pallidonigral pathways.The conclusion reached by Kemp and Powe1l 5 wasbased upon a comparison of the results of electronmicroscopic studies with the results of separate Golgistudies; at that time, it was not possible to examineGolgi-impregnated neurones in the electron microscope.Another problem, common to most contemporarystudies of synaptic contacts in the electronmicroscope, was that the identity of the postsynaptictarget neurone could only be deduced by analogy withthe appearance of neurones in the light microscope.Two technical advances have now made it possible toreveal which type of Golgi-impregnated neurone projectsfrom a brain area and to study these identifiedneurones in the electron microscope (for reviews seeRefs 7,8). Application of these methods has establishedthat the medium spiny neurone is the major'projecting neurone of the striatum: a combination ofretrograde transport of HRP and Golgi impregnationin the same material showed that almost all striatonigralneurones are of medium size and have dendritesdensely covered in spines 9 . 1O • A similar conclusionwas reached for the striatopallidal pathway by applicationof a different method - tracing individual axonsafter the intracellular injection of HRP into striatalneurones 11. The finding that the commonest cell typein the striatum is the major projecting neurone ismuch easier to reconcile with multiple parallel pathwaysand allows us to begin to formulate the basiccircuit of the neostriatum.The medium spiny neurone as the target ofstriatal inputsKemp and Powell s showed that temMals in catstriatum that degenerated after placement of lesionsin the neocortex formed asymmetric synaptic contactsmainly with dendritic spines, and they suggested thatthese spines belonged to the medium spiny neurone.Direct evidence of such an input was provided by thedemonstration of degenerating boutons in contactwith the dendritic spines of Golgi-impregnatedFig. 3. Topography of synaptic inputs to the medium spinyneurone. The distribution of inputs on the striatal mediumspiny neurone is far from being random. On the whole,inputs from outside the stria tum terminate on more distalparls of the dendritic tree and in parlicular on the dendriticspines, while inputs from local neurones (probably othermedium spiny neurones) that contain substance P andCA BA, terminate on proximal parls of the dendritic shaftand on the cell body. Inputs from the large cholinergicintemeurones terminate in an intermediate position.260TINS, Vol. 13, No. 7, 1990

medium spiny neurones in catstriatum after unilateral ablation ofthe cortex l2 . Application of a combinationof three procedures inthe same material (anterogradedegeneration, Golgi impregnationand retrograde transport of HRP)made it possible to show synapticinput from rat sensorimotor cortexto the dendritic spines of identifiedmedium spiny neurones that projectto the substantia nigra 13. Thelatter result could only have beenobtained by using a procedure thatrevealed the distal dendrites of theprojecting striatonigral neuronesbecause synaptic input from thecortex was only found on distaldendrites. These two examplesshow that input from the neocortexterminates mainly on the dendriticspines of medium spiny neuroneslocated in the dorsal striatum; thesynaptic contacts are invariably ofthe asymmetric type. If such aconnection forms part of a basicstriatal circuit, it should alsobe found in the ventral striatumthrough which pass parallel pathwaysfrom limbic regions. Thistype of synaptic contact has nowSensorimotor striatumlimbic striatumbeen demonstrated in rat medial nucleus accumbensafter destruction of hippocampal efferent fibres in thefimbria-fornix: degenerating boutons were found inasymmetric synaptic contact with dendritic spines ofGolgi-impregnated medium spiny neurones l4 .Kemp and Powe1l 5 described degenerating boutonsin asymmetric synaptic contact with dendritic spinesin the striatum after placement of large lesions in catthalamus; they also argued, on the basis of Golgistudies, that such inputs converge on the samedendrites as cortical inputs. In the rat, a differentresult was found after injection of HRP-WGA (wheatgerm agglutinin) in the parafascicular nucleus: anterogradelylabelled boutons were found in asymmetriccontact with the dendritic shafts of a distinct type ofGolgi-impregnated medium spiny striatal neurone thathad a lower density of spines than the more commontype l 5 . In the same study, anterograde degenerationwas used to identify boutons originating from neuronesin the sensorimotor cortex and these were found toterminate on the more common type of densely spinyneurone; no evidence of convergence of cortical andthalamic input onto the same medium spiny neuronecould be found.There is no direct evidence, by anterograde tracingfrom the substantia nigra or ventral tegmental area,about the nature of the target neurone of thesemesencephalic inputs to the striatum. However, useof antibodies to tyrosine hydroxylase to revealcatecholaminergic boutons gives strong evidence thatsuch inputs terminate extensively on medium spinyneurones. A technique for combining Golgi impregnationwith immunocytochemistry has been devised l 6 ,and application of this procedure reveals a strikingdistribution of tyrosine hydroxylase-immunoreactiveboutons on Golgi-impregnated medium spiny neur-1Globus pallidus1! dopamine neuronspallidonigral- - ---- - --- - - - 1Substantia nigraOmgrolhalam,c (VM) 1~----~~. • 1mgrotectal1- - - - - - - - - - - - I I 0 mgrolegme~la l I: Ventral pallidum ~ ; L - - - - - - - - - - - - - - - - - - IIpal/JdOlha~ Thalamus (MD)'-- __ _ __ _ __ ____ 1Fig. 4. Identified targets of striatal output neurones. The targets of the projection neurones of theneostriatum have been identified mainly by looking for degenerating terminals in synaptic contactwith characterized neurones in the target zones. In some cases, anterograde transport of HRP hasbeen used (e.g. in the demonstration of a striato-nigrostriatalloop35). There are four classes ofextrinsic neurone that receive input from the medium spiny neurones of the striatum: thoseprojecting to the thalamus; those projecting to subcortical 'pre-motor' areas such as the tectumand tegmentum; those projecting from the external segment of the pal/idum to the substantianigra ; and, lastly, nigrostriatal dopaminergic neurones. Abbreviations: MD, mediodorsal nucleus ofthe thalamus; VM, ventromedial nucleus of the thalamus.ones that have also been retrogradely labelled fromthe substantia nigra (Fig. 2). The major targets ofthese boutons, which form symmetrical synapticcontacts, are distal dendritic spines of striatonigralneurones, but only those spines that also receiveinput from another, non-immunoreactive, boutonreceive dopaminergic input 17. Such an anatomicalarrangement implies that one of the main functions ofthe dopaminergic input is to interact with the otherinput received by the same spine. The nondopaminergicboutons form asymmetrical synapticcontacts and so probably derive from neurones in thecortex. Evidence of convergence of input in thestriatum from the cortex 18 or from the hippocampus 14onto dendritic spines that also receive dopaminergicsynaptic input has been obtained. In the nucleusaccumbens, identified Golgi-impregnated mediumspiny neurones have been shown to receive hippocampalinput on spines arising from dendritic shaftsthat also receive doparninergic input l4 . It is likely,therefore, that circuits through the dorsal (sensorimotor)striatum and those through the ventral (limbic)striatum are basically similar with regard to theorganization of dopaminergic input.We can conclude that a fundamental striatal circuitwill probably include a densely spiny medium-sizedneurone that projects from the striatum and thatreceives input on its distal dendritic spines from thecortex and input on its distal dendritic spines andshafts from dopaminergic terminals. A second class ofmedium-sized spiny neurone, with a lower density ofspine so, may form part of another circuit; this type ofneurone receives input on its dendritic shafts from theparafascicular nucleus of the thalamus and may also bea projecting neurone and also receive dopaminergicinput (see discussion in Ref. 15).11TINS. Vol. 13, No. 7, 1990261

StriatumCortical pyramidal neuronTransmitters of striatal medium spinyneuronesThe chemical nature of the transmitters used bymedium spiny neurones has been difficult to establishby rigorous techniques that include identification ofthe neurone by a method that reveals its dendriticarbor. Combination of Golgi impregnation withimmunocytochemistry has shown that medium spinyneurones contain substance P and enkephalin-likepeptides l 9 . There is much evidence that GABA ispresent in striatal output pathways (see Ref. 20 forreview) and also in medium-sized neurones within thestriatum 21 , and a preliminary report has describedGABA immunoreactivity in Golgi-impregnatedmedium spiny striatonigral neurones 22 . On the otherhand, uptake of radiolabelled taurine has been demonstratedin identified striatonigral neurones 23 and inGolgi-impregnated medium spiny neurones 24 . We canconclude that likely transmitters of medium spinyneurones as a class are a tachykinin, opiate peptides,GABA and taurine. There is evidence that in individualstriatal neurones GABA co-exists with eithersubstance P or an opiate peptide 21 , but such coexistencehas not yet been demonstrated in identifiedmedium spiny neurones. No distinction has yet beenCortexmade between the transmitters of the commondensely spiny type of neurone and the less denselyspiny type that receives input from the thalamus.The striatal medium spiny neurone as a targetof chemically defined local circuitsThe medium spiny neurone has extensive localaxon collaterals within the striatum and, indeed, thesehave been directly demonstrated for identifiedmedium spiny striatonigral neurones that receivecortical input l 3 . Although the synaptic contacts (symmetrical)of local axon collaterals of medium spinyneurones have been demonstrated n 2 5, there is nodirect evidence that other identified medium spinyneurones are one of the targets, although this is verylikely.Immunocytochemistry has been used to characterizethe synaptic inputs of Golgi-impregnated mediumspiny neurones: boutons immunoreactive for eithersubstance p 26 or glutamic acid decarboxylase (GAD j27have been found in symmetric synaptic contact withthe proximal dendritic shaft. Although the majortarget is the dendritic shaft, occasional boutons arefound on cell bodies. Any, or all, of the above boutonscould originate from other medium spiny neurones butThalamusadditional sources are possible.Thus, the striatum contains a localGABAergic aspiny interneuronethat has an extensive axon field 27 • 28 ,and a small population of substanceP-containing neurones may beinterneurones 29 . Another striatalinterneurone, the large aspinyneurone that contains choline acetyltransferase3u , is the likely sourceof the CM T -immunoreactiveboutons found 111 symmetricalsynaptic contact with the dendriticshafts and spines of Golgi-impregnatedmedium spiny neurones 31 .SNrentopeduncular n.(GPi) Ior ventral pallidumBrainstem}-__.-_________-'-___. 'premoto( areasI(e .g. tectumI -------Ipedunculopontine n.)Fig. 5. Th e fundamental circuit of the basal ganglia is shown here in a highly simplified form. Thekey features are the monosynaptic link from cortical areas (neocortex, hippocampus and amygdala)through stria tal medium spiny neurones to the main output zones of the basal ganglia (pallidumand substantia nigra). The operation of this monosynaptic link is likely to be strongly and specificallyinfluenced by the input from dopaminergic midbrain neurones that terminates on the samedendritic spines. Note that distinct medium spiny neurones project from the stria tum to thesubstantia nigra and to the external segment of the globus pallidus and that it is suggested thatthese neurones are laterally connected through axon collaterals of medium spiny neurones. It is alsosuggested that the striatopallidal neurones receive input from the cortex and from midbraindopamine neurones. (There is no evidence of these connections as yet.) Note that output neuronesin the substantia nigra reticulata may receive converging input directly from the striatum andindirectly via the striato-pallidonigral loop. We have omitted from this scheme another similarcircuit originating in the parafascicular nucleus of the thalamus that terminates on another type ofmedium spiny neurone (see text). We have also omitted the subthalamic nucleus, which probablyplays an important role in the control of output neurones in the globus pallidus and substantianigra 49. Abbreviations: DA, dopaminergic; CPi, internal segment of the globus pallidus; SNr,substantia nigra pars reticulata.Topography of synaptic inputsto the medium spiny neuroneA notable feature of the distributionof synaptic inputs on differentparts of the medium spinyneurone is the preponderance ofextrinsic inputs on the distal dendritesand the concentration ofintrinsic inputs on the proximaldendrites and cell body (Fig. 3).Two types of local afferent inputcan be distinguished - that presumablyfrom other medium spinyneurones, which terminates proximallyon the cell body and proximaldendrites, and that from the largecholinergic interneurone, whichterminates rather more distally onthe dendritic shaft and on somespines. The afferent synaptic'topography' of the medium spinyneurone provides an anatomicalbasis for several possible levels ofinteraction or integration of afferentinformation. Level one can bedefined as the distal dendritic262 TINS, Vo!. 13, No. 7, 1990

spine, of which two types havebeen identified, a spine that has asingle synaptic bouton in asymmetriccontact, comprising 53% ofthe total in the study of Freund etal. 17 and a second type that receivesdual input, i. e. from onebouton fonning an asymmetriccontact and from another formingsymmetric contact (47% of thetotal). The first type of spinepresumably allows input from thebouton to reach the dendritic shaftunimpeded, while the second typeof spine, as pointed out by Kempand Powe 1l 5 , might form the anatomicalsubstrate for interactionbetween the two types of input.Since by far the largest proportionof such 'dual-input' spines receivesinput from dopaminergic terminals,these spines provide a site for theinteraction of dopaminergic inputwith that from cortical regions!7.The second level of interactionbetween inputs occurs on the distaldendritic shaft, where the changesin the electrical properties of themembrane induced by the inputsfrom both types of spine will interactwith each other and with synapticinput directly onto the shaft,such as that from dopaminergic orcholinergic afferents. The thirdlevel of interaction is the proximalpart of the dendritic shaft and thecell body, where the inputs originatingfrom within the striatum caninteract with each other and alsoinfluence the propagation of theinformation from the distal denclrites.In principle, it seems verylikely that the actions of transmittersreleased from boutons incontact with the proximal dendriticshafts and cell body will be modulatoryin character, setting thelevel of excitability of the neuroneand so altering its response toinputs arriving more distally on thedendritic tree.Many questions remain to beanswered about the significanceand the operation of the topographicallyspecific inputs to themedium spiny neurone. For example,we have not considered thefunctional significance of the lateralinteractions between mediumspiny neurones that involve synapticactions on the proximal partof the neurone: Groves 32 hassuggested that one function ofthese interactions is to providelateral inhibition. We have notconsidered the question of whatALimbic cortexLimbicstriatumBDA-neuronsStriatal patchor striosomeMidbrain dopamine neuronsNeocortexDA-neurons1---------,IIIIIIIIIIII1--------IIIIIIIIIIIIISensorimotorstriatumThalamusBrainstemSN r or'pre-motor' areasen topeduncular n. (GPi)Striatalmatrixfrom GPeFig. 6. Circuits that link one part of the striatum with another through midbrain dopamineneurones. The right-hand sides of (A) and (8) are simplified versions of the circuit shown in Fig. 5,and the striato-pallidonigralloop is not shown in full. (A) The limbic striatum might be able to 'ga te 'the flow of information through the semorimotor striatum because 50me of the output from thelimbic striatum terminates on nigrostriatal neurones that innervate the sensorimotor striatum 35 . Inthis way, areas of the brain concerned with emotion and motivation could influence the motorfunctiom of the stria tum. (B) Illustrates a hypothesis in which the scheme shown in (A) isgeneralized to show how information flow through the patch (striosome) compartment of thestriatum can influence the flow of information through the matrix compartment via a dopaminergicpathway originating in the substantia nigra. In this way, particular striosomal compartments mightbe able to 'gate ' the flow of information through particular matrix compartments. Inputs to therespective patch and matrix compartments may either originate in distinct parts of the cortex orfrom different layers of the same cortical area. Abbreviatiom: 0 , and O 2 , different classes ofdopamine receptor; GPe and GPi, external and internal segments of globus pallidus, respectively;SNc, substantia nigra pars compacta; SNr, substantia nigra pars reticulata ; VTA, ventral tegmentalarea.TINS, Vol. 13, No. 7, 1990263

controls the activity of the striatal interneurones thatinnervate the proximal region of the medium spinyneurone (see Ref. 20 for review). We will considerthe important question of what controls the dopaminergicinfluence on the distal parts of the medium spinyneurone at the end of this article.Output function of the striatal medium spinyneuroneAs we have already indicated, there is now directevidence that the medium spiny neurones carry themajor output from the striatum to the substantia nigraand external segment of the globus pallidus. It is verylikely that medium spiny neurones will be found to bethe origin of the striatopallidal output to the internalsegment of the globus pallidus as well. However, thechemical nature of some of the transmitters in thesedistinct output pathways appears to be different.Thus, the striatonigral pathway in the rat containsGABA, a tachykinin, and dynorphin, while the stria topallidalpathway contains GABA and enkephalin, andthe striato-entopeduncular pathway contains GABAand a tachykinin 2o .If we are to define the output function of thestriatum we must identify the nature of the neuronesthat are targets of the medium spiny neurone in thepallidum and substantia nigra. Ultrastructural studieshave established that striatonigral neurones terminateon identified nigrothalamic 1o , nigrotectal·13 , nigrotegmental34 , and nigrostriataP5 neurones, and onidentified nigral dopaminergic neurones 36 . Stria topallidalneurones terminate on pallidonigral neurones:;7,while one target of striato-entopeduncular neuronesis the entopedunculo-thalamic neurone 38 . Finally,neurones in the striatal part of the olfactory tuberclehave been shown to form synaptic contacts withneurones in the ventral pallidum that project to thethalamus 39 (see also A. Parent, this issue). It can beseen that four distinct types of remote neuronereceive input from the medium spiny neurones of thestriatum (Fig. 4): (1) those projecting to the thalamus,which presumably form part of the return loop to thecortex (see G. Alexander and M. O. Crutcher, thisissue); those projecting to subcortical premotor areas(tectum, tegmentum); those projecting from theexternal segment of the pallidum to the substantianigra; and the dopaminergic neurones that projectback to the striatum. It will clearly be crucial todetermine if different types of medium spiny neurone(defined chemically or morphologically) participate inthese different pathways and to identify the postsynapticresponses in the different target neurones.For example, the direct striatonigral pathway usingGABA as transmitter could mediate the inhibitoryresponse of nigrotectal neurones to striatal stimulation(see article by G. Chevalier and j. M. Oeniau,this issue), whereas the indirect striato-pallidonigralpathway could mediate disinhibition of nigrotectalneurones, since it has been shown that the pallidonigralpathway is GABAergic 40 .Towards the fundamental neural networkWe have shown how the study of the synapticconnections of identified neurones allows us topostulate a basic network for the operation of thestriatum (Fig. 5). Such a network could operate ineach of the major subdivisions of the striatum, thelimbic striatum and the sensorimotor striatum, and itcould operate for each of the many segregated parallelpathways that pass through both parts of the striatum ..For each pathway the operation of the network will bethe same: its functions appear to be to allowinformation from different parts of the cortex toinfluence subcortical premotor areas and also toreturn to the cortex in a loop circuit. However, theoperation of this network is likely to be cruciallydependent upon the gating or modulation exerted bythe dopaminergic input from the midbrain 41 . 42 • It isthus particularly significant that output from thestriatum can itself influence the dopaminergic neuronesof the substantia nigra. We would like to concludeby suggesting that control of the dopaminergicneurones by the striatum is exerted by anatomically,and perhaps chemically, distinct parts of the striatum(Fig. 6).Nauta et al. 43 discovered that a prominent outputfrom the nucleus accumbens terminates in the medialpars compacta of the substantia nigra. Subsequently,this pathway was confirmed at the ultrastructural levelwhere it was shown that terminals from neurones inthe nucleus accumbens form symmetric synapticcontacts on identified nigrostriatal neurones 3S Sincethe nucleus accumbens is part of the limbic striatumand receives input from hippocampus, amygdala andprefrontal cortex, this monosynaptic pathway providesa way in which information passing through thenetworks in the limbic striatum could influence thedopaminergic modulation of information passingthrough the networks in the more dorsal, sensorimotorstriatum (Fig. 6A). As pointed out by Nauta and00mesick 44 , this pathway might provide an interfacebetween areas of the brain concerned with 'motivation'and those concerned with motor behaviour.The control of nigrostriatal neurones by the nucleusaccumbens could be just one example of a moregeneral type of control exerted by those striatonigralneurones that terminate on dopaminergic neurones(Fig. 6B). The striatonigral projection originating instriosomes (see article by A. M. Graybiel, this issue)preferentially innervates the pars compacta 45 : themedium spiny neurones in the striosomes, whichcontain a tachykinin as well as GABAl9, mightpreferentially terminate on nigrostriatal neuroneswhile the dynorphin-containing striatonigral neuronesin the matrix might preferentially terminate on outputneurones in the pars reticulata.It has been reported that inputs to the striosomesand the matrix arise either from different parts of thecortex or from different layers of the same corticalarea (see A. M. Graybiel, tlus issue and also Ref.46). The fundamental network of the basal gangliamight therefore be extended to include a striatonigrostriatalloopthat would permit particular corticalareas or layers to influence the flow of informationthrough the basal ganglia from other cortical areas orlayers (Fig. 6B). Such an arrangement means that theoperations of the basal ganglia nught be even moreintimately linked with those of the cortex than hithertosuspected. The proposed circuit also highlights therole of dopamine, which is involved at two stages inthe circuit, and of the two major classes of dopaminereceptor (0 1 and O 2 ; see article by G. Sedvall, thisissue). It is suggested that dopamine operates in thefirst stage within striosomes via O 2 receptors and in264TINS, Vol. 13, No. 7, 1990

the second stage within the matrix via DJ receptors.Such a complex role for dopamine in the functioning ofthe basal ganglia is consistent with many studiessuggesting that motor behaviour requires the synergisticactivation of both DJ and D2 receplors (see, forexample, reviews in Refs 47,48) and might provide abasis for understanding why the type of movementmost affected in Parkinson's disease is that with amotivational or cognitive element.The circuit proposed in Fig. 6B, which provides amechanism for lateral interaction between the patch(striosome) and matrix compartments of the striatumvia a monosynaptic link in the substantia nigra, is putforward as a working hypothesis; further work will berequired to establish whether or not such synapticconnections exist.Selected referencesGraybiel, A M . and Ragsdale , C. W . (1979) Prog. Bra in Res.51 , 239-2832 Heimer, L., Alheid, G. F. and Zaborszky , L. (1985) in The RatNervous System, 1.- Forebra in and M idbrain (Paxinos, G.,ed .), pp. 37-86, Academic Press3 Heimer, L., Switzer, R. D. and Van Hoesen , G. W . (1982)Trends Neurosci. 5, 83-874 Alexander, G. E., DeLong, M . R. and Strick, P. L. (1986)Annu. Rev. Neurosci. 9, 357-3815 Kemp, J. M . and Powell , 1. P. S. (1971) Philos. Tra ns. R. soc.London ser. B 262, 383-4576 Pasik, P., Pasik, T. and DiFiglia, M . (1979) in The Neostriatum(Divac, I. and Oberg, R. G. E., eds) , pp. 5- 36, PergamonPress7 Somogyi , P. and Freund, T. (1989) in Neuroanatomlcal TracttracingMethods 2 (Heimer, L. and Zaborszky, L., ed s), pp.239-264, Plenum Press8 Bolam , J. P. and Ingham, C. (1990) in Handbook of ChemicalNeuroanatomy Vol. 8 (Bjiirklund, A., Hiikfelt, T., Wouterlood ,F. G. and van den Pol , A, eds), pp. 125-198, Elsevier9 Somogyi , P. and Smith , A D. (1979) Brain Res. 178,3-1510 Somogyi, P., Hodgson, A. J. and Smith. A D. (1979)Neuroscience 4, 1805- 185211 ( hang, H. T., Wilson , C. J. and Kit." S. T. (1981) Science 213,915-91812 Frotscher, M ., Rinne, U. , Hassler, R. and Wagner, A. (1981 )Exp. Bra in Res. 41 , 329-33713 Somogyi, P .. Bolam , J. P. and Smith, A D. (1981 ) 1. Comp.Neurol. 195, 567-58414 Totterdell, S. and Smith , A D. (1989) 1. Chem. Neuroanat. 2,285-29815 Dube, L., Smith , A D. and Bolam , J. P. (1988) 1. Comp.Neurol. 267, 455--47116 Freund, T. F. and Somogyi, P. (1989) in NeuroanatomicalTract-tracing Methods 2 (Heimer, L. and Zaborszky, L. , ed s),pp. 201 - 238, Plenum Press17 Freund , T. F., Powell, J. and Smith, A. D. (1984) Neuroscience13 , 1189-121518 Bouyer, J. J., Park, D. H., Joh, T. H. and Pickel , V. M . (1984)Brain Res. 302 , 267- 27519 Izzo, P. N., Graybiel, A. M . and Bolam, J. P. (1987) Neuroscience20, 577-58720 Bolam, J. P. and Smith, A D. in Handbook of ChemicalNeuroanatomy (Bjiirklund, A., H6kfelt, T. and Swanson,L. W. , eds), Elsevier (in press)21 Penny, G. R., Afsharpour, S. and Kitai, s. T. (1986) Neuroscience17,1011-104522 Dub", L. and Descarries, L. (1987) Neuroscience 22 , S79823 (Iarke, D. J., Smith , A D. and Bolam, J. P. (1983) Brain Res.289, 342-34824 Della (orte, L., (Iarke, D. J. , Bolam, J. P. and Smith, A D.(1987) in The Biology of Taurine (Huxtable, R. J., Franconi, F.and Giotti, A., eds), pp. 285-294, Plenum Press25 Wilson, C. J. and Groves, P. M. (1980) 1. Comp. Neurol. 194,599-61526 Bolam, J. P. and Izzo, P. N. (1988) Exp. Brain Res. 70,361-37727 Bolam , J. P., Powell, J. F .. Wu, J-Y. and Smith , A D. (1985)1. (omp. Neurol. 237, 1-2028 Bolam, J. P. , (Iarke, D. J. , Sm ith, A D. and Somogyi , P.(1983) 1. Comp. Neurol. 213, 121- 13429 Bolam , J. P., Somogyi, P., Takagi, H., Fodor, I. and Smith,A. D. (1983) 1. Neurocytol. 12, 325-34430 Bolam , J. P., Wainer, B. H. and Smith , A D. (1984) Neuroscience12 , 711-71831 Izzo, P. N. and Bolam, J. P. (1988) 1. Comp. Neurol. 269,219-23432 Groves, P. M . (1983) Bra in Res. Rev. 5, 109-13233 Williams, M . N. and Faull, R. L. M . (1985) Neuroscience 14,991-101034 Tokuno, H., Moriizumi, T., Kudo, M., Kitao, Y. andNakamura, Y. (1989) Brain Res. 485, 189-19235 Somogyi, P., Bolam , J. P. , Totterdell, s. and Smith, A D.(1981 ) Brain Res. 217, 245-26336 Wassef, M., Berod, A. and Sotelo, C. (1981) Neuroscience 6,2125-213937 Totterdell, 5. , Bolam, J. P. and Smith, AD. (1984) 1. Neurocytol.13, 593-61638 Moriizumi, T. , Nakamura, Y., Okoyama, s. and Kitao, Y.(1987) Neuroscience 20, 797-81639 Zahm, D. s. and Helmer, L. (1987) Brain Res. 404, 327-33140 Smith, Y. and Bolam, J. P. (1989) Bra in Res. 493, 160-16741 Steven s, J. R. (1973) Arch. Gen. Psychiatry 29,177-18942 Bjiirklund, A (1986) in Handbook of Physiology (Sect. 1: TheNervous System; Vol. IV: Intrinsic Regulatory Systems of theBrain) (Mountcastle, V. B., Bloom, F. E. and Geiger, S. R.,eds), pp. 155-235, American Physiological Society43 Nauta, W. J. H., Smith, G. P., Faull, R. L. M. and Domesick,V. B. (1978) Neuroscience 3, 385--40144 Nauta, W . J. H. and Domesick, V. B. (1984) in Fun ctions ofthe Basal Ganglia, Ciba Foundation Symposium 107 (Evered,D. and O '(onnor, M .. eds), pp. 3-23, Pitman45 Gerfen, C. R. (1985) 1. Comp. Neural. 236, 454-47646 Gerfen, C. R. (1989) Science 246, 385-38847 Clark, D. and White, F. J. (1987) synapse 1,347- 38848 Waddington, J. L. and Oboyle, K. M . (1989) Pharmacal Ther.43 , 1-5249 Kita, H. and Kitai , S. T. (1987) 1. Comp. Neural. 260,435--452TINS, Vol. 13, No. 7, 1990265