Stimulation of hindlimb, trunk, or forelimb activated primary sensory cortex In a localized columnar pattern, indicating activation of somatosensory receptors and.
Proc. Nad. Acad. Sci. USA Vol. 89, pp. 7403-7407, August 1992 Neurobiology
Somatotopic organization in rat striatum: Evidence for a combinational map (deoxyglucose/caudate putamen/forelimb/hldHmb/neocortex)
Lucy L. BROWN The Saul R. Korey Department of Neurology and Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY 10461
Communicated by Louis Sokoloff, May 6, 1992
ABSTRACT 2-Deoxy-D-[14Clglucose autoradiography was used in awake rats to map neural activity in the sensorimotor sector of striatum. Stimulation of hindlimb, trunk, or forelimb activated primary sensory cortex In a localized columnar pattern, indicating activation of somatosensory receptors and a discrete cortical func l unit. In sensorimotor striatunM, an image analysis detection technique revealed regions of maximal activity, or features, that formed a patchy pattern of activation reminiscent of the known anatomic patterns of corticostriate terminals. Ipsilateral as weD as contralateral activation was observed. The activated areas revealed a body map in striatum that was organized in a manner consistent with cortical topography (dorsoventrally: hindlimb, trunk, forelimb) at most anteroposterior levels, similar to that found in other species. However, at other levels, a different organization (e.g., trunk, hindlimb, forelimb) was observed. Furthermore, the arrangements of body region and side were also unique at different anteroposterior levels. Thus, functional activity showed multiple, different juxtapositions of body elements-i.e., a combinational map. The data suggest that striatum may provide an anatomic substrate for different combinations of inputs necessary to select and integrate movement.
arm/face (4, 9, 11), which has been confirmed by electrophysiological studies (12, 13). But the somatotopy is not always distinct (4); it shows shifts (9, 10), and electrophysiological findings show some "intermingling of arm neurons with leg and orofacial neurons" (13). A dorsolateral "sensorimotor zone" is also seen to extend AP through most of rat striatum, and anatomical and electrophysiological data indicate a general, but also apparently intermingling, somatotopy (5, 6). The DG autoradiographic technique is especially useful to identify simultaneous changes in activity in
this extensive map. Because rat striatum demonstrates the basic features of neurochemical compartmental organization and somatosensory representations, this extensively studied species is an excellent model for the study of striatal functional organization with DG autoradiography.
MATERIALS AND METHODS Male Sprague-Dawley rats (300-400 g) were used. Somatosensory stimuli were applied to the left side of the body of four groups: hindlimb (n = 5), trunk (n = 5), forelimb (n = 5), and hindlimb and forelimb combined (n = 4). A control group did not receive stimuli (n = 5). All animals, including controls, were placed in a Plexiglas restrainer in which one or two limbs were extended and held in place by a cuff of soft surgical tape. To minimize auditory and visual stimuli, ears were blocked with wax, and opaque Plexiglas blocked the stroking apparatus from view. Electromyographic activity was recorded through electrodes implanted a week before the experiment in four hindlimb, four forelimb, and all control animals. Stroking stimuli were applied for 10 min before the experiment began and during the entire 45-min experimental period. Stimuli (three or four per sec) were applied by a nylon bristle (diameter, 1.0 mm) that exerted 2.0-2.5 g of force. Both cutaneous and deep receptors may have been affected (14). The bristle, mounted on a wheel driven by a variablespeed motor, traversed a path that was intentionally varied between 0.3 and 2.0 cm. The length of the stimulus path for forelimb was within the receptive field shown for individual cells in perigranular SI (14). Forelimb strokes were applied between wrist and elbow; in controls, the bristle did not touch the animals; hindlimb strokes were between patella and pelvis; trunk strokes were on the upper half of the torso. DG experiments followed described procedures (15, 16). Briefly, catheters placed in the external iliac artery and vein under halothane anesthesia were drawn out at the back of the neck. Two to 4 hr after surgery, the fully awake and active animals were adapted to the restraining device for at least 20 min before stimulus onset. An intravenous injection of 50 gCi (1 Ci = 37 GBq) of DG (New England Nuclear) in 0.5 ml of saline was given and 11 arterial samples were taken over the 45-min experimental period. Animals were sacrificed with pentobarbital and brains were removed and processed for
The motor symptoms of striatal pathology-for example, Huntington chorea-suggest that striatum plays a role in motor and sensorimotor integration. A region with wellcharacterized chemoarchitectonic compartments, striatum receives a major afferent component from the cortex. Intrinsic receptors and transmitters form patches, or striosomes, in a matrix that contains clusters of terminals from cortex (e.g., see refs. 1 and 2). However, the functional significance for motor integration of the cortical terminal patterns and chemoarchitectonic compartments is unclear. To develop a functional probe of striatum, this study used 2-deoxy-D-[14C]glucose (DG) autoradiography, a technique successfully used by others to delineate functional units in the central nervous system (e.g., visualization of columns in striate cortex) (3). The DG autoradiographic technique can reveal the effect of a stimulus throughout the nervous system. Literally hundreds of simultaneously activated regions can be identified. First-order information, somatotopic representation of cutaneous input, was used to probe the complex organization of striatum because primary somatosensory cortex (SI) projects to striatum (4, 5), cell firing rate in striatum can be altered by somatosensory stimulation (6, 7), and sensory-motor deficits are seen following striatal lesions (8). Data from primates and cats suggest that corticostriate terminals from SI form a complex somatotopic map that extends anteroposteriorly throughout most of striatum (4, 9, 10). Anatomical studies in primates have shown a general dorsal-ventral organization of corticostriatal inputs for leg/ The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Abbreviations: AP, anterior-posterior; DG, 2-deoxy-n-[14C]glucose; SI, primary somatosensory cortex.
7403
7404
Neurobiology: Brown
Proc. Nati. Acad. Sci. USA 89 (1992)
quantitative autoradiography. Serial sections (20 gm) were saved through striatum. Analysis of digitized autoradiograms was done on a Quantimet 970 image analysis system (Leica-Cambridge, Deerfield, IL). The system was programmed to detect regions of contiguous pixels ("features") at preset levels of sensitivity ("grey levels"). Threshold was defined as the sensitivity at which the most optically dense pixels (minimum, 5; maximum, 10) in a section were detected. A feature was defined as an area of 30 or more contiguous pixels (100 pLm2) detected at 10-12 grey levels above threshold. As many as 25 features (typically 10) were detected in a single coronal section. A previous study has shown that the detection procedure used preserves the heterogeneity of striatal glucose utilization and identifies independent zones of maximal striatal activity (17). Area (mm2) and position were calculated for each feature. Position was defined as the center of mass (centroid) relative to the midline (xO) and dorsal edge of striatum (yo). The largest feature for a given coronal section is referred to as the primary feature; it is the region of greatest activation for that anterior-posterior (AP) level. The next two largest features are referred to as secondary features. Primary and secondary feature area and position were analyzed every 200 ,um, from the anterior pole of striatum (+2.2 mm) to 1.5 mm caudal to bregma (18). All AP levels are referred to bregma (AP 0 mm), and follow the atlas ofPaxinos and Watson (18). The error in coordinate assignment caused by brain shape differences and angle of cut were rarely as much as 0.2 mm. Statistical analyses were carried out on mean mediolateral and dorsoventral positions, and areas of primary and secondary features at each AP level (ANOVA followed by t test); AP x mediolateral and dorsoventral position were analyzed separately (ANOVA; repeated measures). Right-left comparisons were made within groups (paired t test). RESULTS Animals were generally quiet during the experimental period and there was no bursting of electromyographic activity that correlated with the stroke stimuli. Stimulation of different body regions activated nonoverlapping regions of contralateral SI cortex. The activation included layers III, IV, and V, creating a columnar appearance (Fig. 1) at locations that were predictable from electrophysiological studies (14, 19). Alignment of stained alternate sections with the autoradiograms suggested that activation was in granular and perigranular SI. The activation extended anteroposteriorly from 200 to 500 Am. Ipsilateral cortical activation was seen in addition to contralateral in two hindlimb and all trunk animals, but not in forelimb animals (Fig. 1). In the case of the forelimb, for which sensory and motor cortex are known to be separate at some points (e.g., see ref. 19), the motor cortex did not show any columnar activation nor did the sensorimotor overlap region. The control group did not show columnar patterns of activation in SI. The location of contralateral, maximal activation (features) in dorsolateral striatum depended on the body region stimulated (stimulation group effect for mediolateral position, P < 0.02; stimulated vs. control, P < 0.01; ANOVA, repeated measures). Also, feature position (centroid) within each group varied from anterior to posterior; each animal in the group showed similar shifts in position from anterior to posterior, and the statistical interaction effects were significant (mediolateral x AP position, P < 0.02; dorsoventral x AP position, P < 0.01). For example, at AP +0.2, primary features for each group achieved maximal separation (Fig. 2) with hindlimb most dorsomedial (3.4 0.2 mm lateral; 1.0 + 0.1 mm ventral), trunk ventrolateral to hindlimb (3.8 0.3 mm lateral; 1.2 0.1 mm ventral), and forelimb ventrolateral ±
±
±
A
HL
B
TR
HL/FL
FIG. 1. Autoradiograms showing localization of increased glustimulation. (A) Hindlimb (HL) stimulation. A column of high density can be seen on the right. (B) Trunk (TR). A column of high density is seen on the right, with a shadow in homotopic left cortex. (C) Forelimb (FL). A column of high density is seen on the right, lateral to the column in A. (D) Left hindlimb and forelimb (HL/FL) stimulated simultaneously. Two columns of high density are seen on the right in the same regions as for the individually stimulated animals. (Bar = 1.0 mm.) cose utilization in cortex during left-side somatosensory
to trunk (4.4 + 0.1 mm lateral; 1.3 ± 0.1 mm ventral). Thus, at this AP level, the distance between hindlimb and forelimb features was >1.0 mm. However, posteriorly, at bregna, the primary feature for hindlimb was ventral to its position at AP
+0.2, and the forelimb feature was dorsal, making the separation between forelimb and hindlimb features minimal (Fig. 2). Animals that received simultaneous stimulation of both hindlimb and forelimb had features that were consistent with the foregoing: distinct features were seen in both the forelimb and hindlimb areas defined by the individually stimulated animals, maximally distant at AP +0.2, minimally at bregma (Fig. 2). A plot of each animal's primary feature onto an outline of striatum at four different AP levels shows the consistent localization of activated regions within groups and how the features shift their relative positions at different AP levels (Fig. 3). The group means are shown graphically in Fig. 4. Although a different organization of hindlimb, trunk, and forelimb is seen at each AP level, these are deviations from a general organization in which hindlimb is dorsal and forelimb is ventral. Secondary features also distinguished the different groups at many AP levels and shifted position as a function of AP level. For example, in trunk-stimulated animals, the secondary feature at AP +0.6 was dorsomedial to the primary
Neurobiology: Brown
Proc. Natl. Acad. Sci. USA 89 (1992)
7405
x: 3.8 y: 0.9
x: 4.0 y: 1.4
111124
*X: 4.0
y: 1. 5
FIG. 2. Autoradiograms with primary and secondary features at different AP levels. Features are highlighted as white masks covering regions ofmaximal glucose utilization. The Cartesian coordinates are indicated (mm) for the centroid of each primary feature shown: xO = midline, yo = dorsal edge of striatum; y coordinates indicate mm ventral to yo. (A-C) Features seen at AP +0.2 in a hindlimb (HL)-, trunk (TR)-, and forelimb (FL)-stimulated animal, respectively. At this AP level, the features for body regions were most separated. (D-F) The same animals as described above, at AP 0 (bregma). Note that the features for hindlimb and forelimb have changed position. (G) In an animal stimulated on hindlimb and forelimb simultaneously, two juxtaposed features (arrows) can be seen at AP 0, similar in position to those from the individually stimulated animals shown above in D and F.
feature (P < 0.01; Fig. 3C), but at AP + 1.0 it was ventral (Fig. 3D). At AP +0.2, secondary features were scattered, with a tendency for forelimb features to intermingle with trunk and for hindlimb to intermingle with trunk and forelimb (Fig. 3B). Even the sixth largest features for forelimb and hindlimb differed significantly (P < 0.01) at this level and were closely situated to their primary features. At AP +0.6 the secondary feature positions paralleled the primary feature positions of hindlimb, trunk, and forelimb (Fig. 3C) and were distinct (P < 0.03). In ipsilateral striatum, primary feature position differed among groups and from controls as a function of AP level (P < 0.04; repeated measures). Within a group, at a given AP level, the ipsilateral feature centroid was either in the homotopic region of the contralateral striatum, or it was displaced by 300-500 um, dorsoventrally or mediolaterally (P values between 0.01 and 0.03; matched t test). For example, at AP +0.2, the ipsilateral and contralateral features were within 250 ,um of each other; however, at AP +1.0 the ipsilateral forelimb feature was 400 gm dorsal to the contralateral (P < 0.02; matched t test), and the ipsilateral hindlimb feature was 400 ,um ventral to the contralateral (P < 0.02; matched t test;
0 FIG. 3. Plots of centroids of primary (solid symbols) and secondary (open symbols) features from individual animals at four AP levels. (A) At bregma (AP 0), forelimb and trunk features are intermingled over a wide area and close to hindlimb features. Secondary features from forelimb animals are spread out over the whole lateral field. (Secondary trunk features are not shown.) (B) At this AP level the greatest separation of primary features for the three body regions is seen. Secondary hindlimb and trunk features are found ventrally, within either the forelimb or trunk primary feature field, while secondary forelimb features are found dorsally, within the trunk field. (C) At 400 pm anterior to the previous section, the hindlimb primary features are found ventrolateral to their previous position, and trunk features are dorsomedial. The forelimb primary features have also shifted dorsomedially and are proximal to hindlimb features. Secondary features have similar relationships to each other as do their primary features. (D) At AP +1.0, the hindlimb primary features are shifted dorsomedially to their previous position, the forelimb features are in similar positions, and trunk features are ventrolateral. This results in an intermingling of forelimb and trunk features, which are separated from hindlimb features. The secondary features for hindlimb and trunk show separate clustering.
7406
Proc. Natl. Acad. Sci. USA 89 (1992)
Neurobiology.- Brown
FIG. 4. Three-dimensional graphical representations of mean positions of primary features in striatum following hindlimb (a), trunk (A), or forelimb (e) somatosensory stimulation. (A) Mean coetralatml activation points for each body region shift their positions from anterior to posterior reltve to each other. Arrows highlight forelimb shifts. The effect was reliable from animal to animal (P < 0.01; dorsoventral position x AP position). A medum spiny cell is drawn for size context. (B-D) Mea coaterl (solid lines) and ipsilateral (dotted lines) aivation points are plotted for each body region separately. Arrows indicate AP +0.2 mm, where mean contralateral and ipsilateral points are closely juxtaposed for each body region. Lines connecting the points are used only to emphasize the shifting positions and do not necessarily indicate continuous activation along the path of the line. Asterisks indicate ipsilateral/ contralateral comparisons; P < 0.05.
Fig. 4). Such significant dorsoventral differences were also found for each group at AP -0.5 (Fig. 4). Finally, mediolateral differences of 500-600 pm between ipsilateral and contralateral feature centroids were found for the forelimb and trunk at AP +0.6 (Fig. 4). Thus, ipsilateral activation is distinct from contralateral and reflects a second body map on each side of striatum. Primary features were approximately circular or had complex shapes; largest diameters ranged from 460 94 ,um for hindlimb to 662 125 Am for forelimb. In the forelimb group, diameter decreased significantly from anterior to posterior (662 ± 125 to 300 64 pm; P < 0.04; ANOVA), mean trunk diameter increased (280 ±63 to 468 94 pm; P < 0.05), while hindlimb did not change. Mean glucose utilization rates in dorsolateral striatum of stimulated animals were 20% higher than in controls in the regions of their primary feature, but the difference was not statistically significant (85 5 Imol per 100 g per min for controls compared to 102 5 for all stimulated animals). Left/right differences in stimulated animals were small (916%) but reliable (P < 0.03). Mean glucose utilization rates in cortex and striatum were similar. Layer IV in the contralateral forelimb region of controls was 81 3; in stimulated animals it was 102 6 Amol per 100 g per min, a 26% ±
±
±
±
±
±
±
increase.
DISCUSSION Somatosensory stimulation of different body regions altered glucose utilization patterns in rat striatum, revealing a map that changed as a function of AP level. Also, independent contralateral and ipsilateral representation was seen. The complex nature of the map, with its AP variations that were reliable across animals, is the most important finding of this study. However, these were variations from a somatotopic framework that has a dorsal-to-ventral representation of hindlimb/trunk/forelimb similar to that found in cat caudate (10) and primate putamen (4, 9, 11, 13). The shifting relationships of body region representation at different AP levels are consistent with corticostriate projection patterns in other species: in cat, "a gradual rotation in the axis of orientation of the map at different anteroposterior levels" (10); in primate, relationships of SI projections to striatum differed from
anterior to posterior (9). The striatal region in which features were found has been described as the "sensorimotor sector" in rats, known to receive projections from sensorimotor cortex (5, 20). In DG studies, electrical stimulation of motor cortex causes activation in rat dorsolateral striatum (21, 22), and electrophysiological studies have found cells responsive to movement and to stimulation of individual body parts throughout its AP extent (6).
"'Features" objectively identified regional differences in neural activity among groups. Because glucose utilization appears to reflect presynaptic axon terminal activity (e.g., see ref. 23), the primary features identified in the striatum probably reflect the areas of greatest afferent input from cortex (with some contribution from axon collaterals of principal neurons). Glucose utilization is heterogeneous and patchy in striatum, even under control conditions (17), and it is difficult for the eye to detect differences from control. Accordingly, a detection technique was necessary to achieve reliable and objective identification of features as the loci of changes related to somatosensory input. Note further that somatosensory stimulation did not increase glucose utilization rates in striatum by more than 20% above control because the stimuli were physiologically normal. This small percentage difference does not diminish the significance of the effect, however. For example, in layer IV of the cortex, where activity changes are robust and focused, the glucose utilization rate was 26% above controls. The size of the features may depend on factors such as the number or rate of discharge of active axon terminals. Feature size varied significantly within groups, decreasing AP for forelimb and increasing for trunk. This may reflect greater input anteriorly for forelimb and posteriorly for trunk. Note also that although the data in Fig. 4 have continuous lines connecting features in the AP plane, they were often not continuous for more than 300 gm, with discontinuities of 100-200 gm. Somatosensory stimuli produced columnar cortical activation limited to SI, and electromyograms did not show increases associated with the stimuli. Although motor effects cannot be entirely ruled out, the major effects observed are very likely the result of somatosensory activity. In addition, trunk stimuli activated cortex bilaterally, but forelimb stimuli did not; this is consistent with anatomical data which show
Neurobiology: Brown
Proc. Natl. Acad. Sci. USA 89 (1992) Medial
CORTEX SI
T
F
Lateral
HMedial STRJATUMT
F
It
Anterior -Posterior
FIG. 5. Diagrammatic representation of three AP levels in striatum to show changes in localization of activation related to body
region. Organization of cortical activation is also represented. At one AP level (to the right in the diagram), the most laterally activated cortical region was associated with the most laterally activated striatal region; the most medial cortical region was associated with the most medial striatal activation; and each was close to ipsilateral activation. However, at other AP levels, hindlimb, trunk, and forelimb were juxtaposed in different combinations or sets. In the middle set, leg could interact with trunk and forelimb. The major axes of the ovals are 400 Am in diameter and indicate the area of hypothetical interactions on or among medium spiny cells. Small circles within ovals indicate contralateral activation; rectangles indicate ipsilateral activation. Arrow pairs show examples of different juxtapositions of forelimb points with other points. F, forelimb; H, hindlimb; T, trunk.
that forelimb regions do not project across the corpus callosum to homotopic cortex (24). The asymmetry of ipsilateral and contralateral activation in striatum was unexpected. Bilateral projections from motor cortex to striatum are well-documented in primate and rat (5, 11), but they were not described as nonoverlapping until recently (25). In rats, primary sensory cortex also projects bilaterally to striatum (26). The asymmetry may not have been detected in anterograde tracing studies because the large injection sites were relatively nonspecific compared to the specific functional activation in this study, and differences might be difficult to detect without computerized measurements.
One hypothesis to explain varying juxtapositions of region and side is that the striatal body map provides the substrate to compute all of the permutations and combinations of spatiotemporal input necessary for carrying out motor plans. Consistent with this hypothesis are electrophysiological data that find some striatal cells to be nonspecific or responsive to stimuli from several body regions (6, 12, 13). Such responses could be the result of cortical afferent convergence onto single striatal cells or onto juxtaposed striatal cells that can then interact through axon collaterals or intemeurons. Most striatal neurons are medium spiny cells, with dendritic fields of 250-400 pum and axon collateral fields of up to 1200 gm (e.g., see ref. 27). With such wide axon collateral fields, cells within 300-500 tkm could directly influence each other, and information could be combined while also maintaining a separate identity; but also, afferent activation points within 400 um of each other could affect the same dendritic field. In this study, primary feature centroids associated with different body regions could be found within 400 ,um of each other, forming apparent pairs. Thus, striatal afferents from different cortical regions representing different body areas might affect the same striatal cells, or juxtaposed cells and interneurons, effectively combining inputs. The results from the simultaneously stim-
7407
ulated hindlimb and forelimb animals suggest that these inputs do not overlap but are on juxtaposed cell clusters. In conclusion, the results confirm the functional significance of somatosensory input to striatum, which may be useful to probe striatal pathology (7, 8). The results also confirm a complex body map in dorsolateral striatum (6), especially when both primary and secondary activation regions are considered. Sensorimotor integration in striatum may depend on combinations of somatosensory, other sensory, and probably motor inputs from different body regions bilaterally. The arrangement observed (Fig. 5) would be appropriate for decision-making, sequencing, and any other functions that require integration of information that also retains distinct representational identity. Like the cerebellum, such a combinational map, providing the substrate to compute a wide range of possible movements, would need a large number of neurons in comparison with its target nuclei. Indeed, estimates of the striatal-pallidal convergence ratio are on the order of 100:1 (28). Finally, these specific functional maps may be related to neurochemical alterations and compartments in future studies. I thank Diane Smith and Sandra Salinas for technical assistance and Dr. Samuel Feldman for assistance in preparation of the manuscript. This work was supported by U.S. Public Health Service Grant NS21356. 1. Gerfen, C. R. (1992) Trends Neurosci. 15, 133-138. 2. Graybiel, A. M. (1990) Trends Neurosci. 13, 244-254. 3. Hubel, D. H., Wiesel, T. N. & Stryker, M. P. (1978) J. Comp. Neurol. 177, 361-379. 4. Kunzle, H. (1977) Exp. Brain Res. 30, 481-492. 5. McGeorge, A. J. & Faull, R. L. M. (1989) Neuroscience 29, 503-537. 6. Carelli, R. M. & West, M. 0. (1991) J. Comp. Neurol. 309, 231-249. 7. Lidsky, T. I., Manetto, C. & Schneider, J. S. (1985) Brain Res. Rev. 9, 133-146. 8. Marshall, J. F. (1979) Brain Res. 177, 311-324. 9. Flaherty, A. W. & Graybiel, A. M. (1991) J. Neurophysiol. 66, 1249-1263. 10. Malach, R. & Graybiel, A. M. (1986) J. Neurosci. 6, 3436-3458. 11. Kunzle, H. (1975) Brain Res. 88, 195-209. 12. Alexander, G. E. & DeLong, M. R. (1985) J. Neurophysiol. 53, 1417-1443. 13. Crutcher, M. D. & DeLong, M. R. (1984) Exp. Brain Res. 53, 233-243. 14. Chapin, J. K. & Lin, C.-S. (1984) J. Comp. Neurol. 229, 199-213. 15. Crane, A. M. & Porrino, L. J. (1989) Brain Res. 499, 87-92. 16. Sokoloff, L., Reivich, M., Kennedy, C., DesRosiers, M. H., Patlak, C. S., Pettigrew, K. D., Sakurada, 0. & Shinohara, M. (1977) J. Neurochem. 28, 897-916. 17. Brown, L. L., Wolfson, L. I. & Feldman, S. M. (1987) Brain Res. 411, 65-71. 18. Paxinos, G. & Watson, C. (1986) The Rat Brain in Stereotaxic Coordinates (Academic, San Diego). 19. Hall, R. D. & Lindholm, E. P. (1974) Brain Res. 66, 23-38. 20. Cospito, J. A. & Kultas-Ilinsky, K. (1981) Exp. Neurol. 72, 257-266. 21. Coffins, R. C. (1978) Brain Res. 150, 503-517. 22. Sharp, F. R. (1984) J. Comp. Neurol. 224, 259-285. 23. Mata, M., Fink, D. J., Gainer, H., Smith, C. B., Davidsen, L., Savaki, H., Schwartz, W. J. & Sokoloff, L. (1980) J. Neurochem. 34, 213-215. 24. Jones, E. G., Coulter, J. D. & Wise, S. P. (1979) J. Comp. Neurol. 188, 113-136. 25. Flaherty, A. W. & Graybiel, A. M. (1991) Neurosci. Abstr. 17, 1299. 26. Canteras, N. S., Shammah-Lagnado, S. J., Silva, B. A. & Ricardo, J. A. (1988) Brain Res. 458, 53-64. 27. Kawaguchi, Y., Wilson, C. J. & Emson, P. C. (1990) J. Neurosci. 10, 3421-3438. 28. Wilson, C. J. (1990) in The Synaptic Organization of the Brain, ed. Sheperd, G. M. (Oxford Univ. Press, New York), 3rd Ed., pp. 279-316.
