Distribution of GABAergic Neurons and Axon Terminals in the Macaque Striate Cortex

Transcription

1 THE JOURNAL OF COMPARATIVE NEUROLOGY 264:73-91(1987) Distribution of GABAergic Neurons and Axon Terminals in the Macaque Striate Cortex D. FITZPATRICK, J. S. LUND, D.E. SCHMECHEL, AND A.C. TOWLES Department of Anatomy and Psychology (D.F.) and VA Medical Center and Department of Medicine (D.E.S.), Duke University, Durham, North Carolina 27710; Department of Psychiatry (J.S.L.), University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15261; and Department of Neurology (A.C.T.), Cornell University Medical Center, New York, New York ABSTRACT Antisera to glutamic acid decarboxylase (GAD) and y-aminobutyric acid (GABA) have been used to characterize the morphology and distribution of presumed GABAergic neurons and axon terminals within the macaque striate cortex. Despite some differences in the relative sensitivity of these antisera for detecting cell bodies and terminals, the overall patterns of labeling appear quite similar. GABAergic axon terminals are particulariy prominent in zones known to receive the bulk of the projections from the lateral geniculate nucleus; laminae 4C, 4A, and the cytochrome-rich patches of lamina 3. In lamina 4A, GABAergic terminals are distributed in a honeycomb pattern which appears to match closely the spatial pattern of geniculate terminations in this region. Quantitative analysis of axon terminals that contain flat vesicles and form symmetric synaptic contacts (FS terminals) in lamina 4C, and in lamina 5 suggest that the prominence of GAD and GABA axon terminal labeling in the geniculate recipient zones is due, at least in part, to the presence of larger GABAergic axon terminals in these regions. GABAergic cell bodies and their initial dendritic segments display morphological features characteristic of nonpyramidal neurons and are found in all layers of striate cortex. The density of GAD and GABA immunoreactive neurons is greatest in laminae 2-3A, 4A, and 4C,. The distribution of GABAergic neurons within lamina 3 does not appear to be correlated with the patchy distribution of cytochrome oxidase in this region; i.e., there is no significant difference in the density of GAD and GABA immunoreactive neurons in cytochrome-rich and cytochrome-poor regions of lamina 3. Counts of labeled and unlabeled neurons indicate that GABA immunoreactive neurons make up at least 15% of the neurons in striate cortex. Layer 1 is distinct from the other cortical layers by virtue of its high percentage (77-81%) of GABAergic neurons. Among the other layers, the proportion of GABAergic neurons varies from roughly 20% in laminae 2-3A to 12% in laminae 5 and 6. Finally, there are conspicuous laminar differences in the size and dendritic arrangement of GAD and GABA immunoreactive neurons. Lamina 4C, and lamina 6 are distinguished from the other layers by the presence of populations of large GABAergic neurons, some of which have horizontally spreading dendritic processes. GABAergic neurons within the superficial layers are significantly smaller and the majority appear to have vertically oriented dendritic processes. These results provide support for the idea that GABAergic neurons make up a significant proportion of the neurons within ALAN R. LISS, INC. Accepted April 8, 1987.

2 74 D. FITZPATRICK ET AL. the macaque striate cortex and that there are laminar differences in the number and the types of GABAergic neurons-differences that may be relevant for understanding the contributions of GABA-mediated inhibition to striate cortex function. Key words: GABA, GAD, inhibition, aspinous neurons, cytochrome oxidase, immunocytochemistry Neurons that utilize the neurotransmitter y-aminobutyric acid (GABA) appear to play an important role in generating many of the receptive field properties that distinguish neurons in the striate cortex from those in the lateral geniculate nucleus. When the inhibitory effects of GABA are blocked by the administration of the potent GABA antagonist bicuculline, the selective responses of cortical neurons for the orientation, length, direction of movement and velocity of a visual stimulus are abolished or diminished (Sillito, '75, '77a,b, '79; Sillito and Versiani, '77; Pate1 and Sillito, '78; Tsumoto et al., '79). GABAergic neurotransmission within the striate cortex, as in other regions of neocortex, is thought to be mediated largely by a morphologically diverse class of non-pyramidal neurons that share the feature of having relatively spinefree dendritic processes. Evidence for the selective role of aspinous neurons in cortical inhibition is derived from the observations that neurons with this morphology (1) accumulate JH-GABA (Chronwall and Wolff, '78; Hamos et al., '83; Hendry and Jones, '81; Somogyi et al., '84b); (2) are hmunoreactive for glutamic acid decarboxylase (GAD), the synthesizing enzyme for GABA (Ribak, '78; Peters and Fairen, '78; Freund et al., '83; Houser et al., '83; Somogyi et al., '83b); (3) are immunoreactive for GABA itself (Ottersen and Storm-Mathisen, '84; Somogyi and Hodgson, '85); and (4) have terminals that contain flat vesicles and make symmetric synaptic contacts, characteristics thought to be indicative of inhibitory transmission (Uchizono, '65; Peters and Fairen, '78; Somogyi and Cowey, '81). While a general consensus has been reached on the morphological class of neurons responsible for GABAergic transmission in the striate cortex, considerably less is known about the numbers of these neurons and their distribution within the macaque striate cortex. The existing information on the precise laminar and columnar segregation of cell types and connections within the monkey striate cortex provides a valuable structural and functional framework within which to examine the distribution of GA- BAergic neurons. For example, the projections from the magnocellular and parvicellular layers of the lateral geniculate nucleus (LGN) to the striate cortex terminate in separate tiers of lamina 4C (4C,, and 4C,, respectively; Hubel and Wiesel, '72; Hendrickson et al., '78). Differences in the number and types of GABAergic neurons in these layers may have important implications for understanding the way striate cortex processes the information derived from these two pathways. The discovery that lamina 3 has distinct cytochrome-oxidase-rich patches that are distinguished from the surrounding regions in their response properties and in their connections (Horton and Hubel, '81; Livingstone and Hubel, '82, '84; Fitzpatrick et al., '83a) provides another set of functional landmarks for examining the distribution of GABAergic neurons. Indeed, there is already some evidence that the cytochrome-oxidase-rich re- gions of lamina 3 are particularly rich in GABAergic neurons and terminals (Hendrickson et al., '81; Carroll and Wong-Riley, '85). In the present study, immunocytochemical markers for GAD and GABA were used to characterize the morphology, number, and distribution of presumed GABAergic neurons and axon terminals in the macaque striate cortex. The main purpose of this study was to establish how the normal distribution pattern of GABAergic neurons and terminals relates to the well-defined laminar and columnar subdivisions of the monkey striate cortex. It is our view that this information constitutes a useful first step in understanding how neurons and terminals involved with GABAergic neurotransmission are arranged with respect to the terminal fields of functionally different geniculocortical pathways and to the cortical cells that give rise to extrastriate and brainstem projections. MATERIALS AND METHODS Immunoc ytoc hemistry A total of five macaque monkeys (Mucaca fasciculuris) were used in these experiments. Animals were deeply anaesthetized and perfused through the heart with a 0.9% saline solution followed by one of the following fixatives: 4% paraformaldehyde, 0.1% glutaraldehyde in 0.1 M phosphate buffer, ph 7.4 (two animals); 0.5% zinc salicylate dissolved in 10% formalin in water as described by Mugnaini and Dahl('83) (one animal); or a mixture of paraformaldehyde, lysine, and sodium periodate (PLP) made according to the protocol of McLean and Nakane ('74) (two animals). Except for the animals perfused with the PLP mixture, the fixative was flushed out with a solution of 10% sucrose in 0.1 M phosphate buffer, ph 7.4. Sections from the striate cortex were cut at pm on an Oxford vibratome and collected in 0.01 M phosphatebuffered saline (PBS), ph 7.6. Selected sections were processed using GAD or GABA immunocytochemistry, cytochrome oxidase histochemistry (Wong-Riley, '79) or Nissl stains (either thionin or toluidine blue). The antiserum developed by Oertel et al. ('81a) was used for the demonstration of GAD immunoreactivity. Details of the purification of GAD and the preparation of the antiserum have been published (Oertel et al., '81a,b). The PLP fixative was superior to other fixatives for labeling cortical cell bodies with this antiserum, but labeled terminals were found with all three fixatives. In some cases, prior to immunocytochemical processing, sections were run through an increasing and decreasing series of alcohol solutions (10, 25, 50%) for 3 minutes each. This procedure significantly increased the penetration of the antiserum; in the thinner sections, terminals appeared through the depth of the section. However, immersion in alcohol solutions significantly

3 GABAERGIC NEURONS IN STRIATE CORTEX 75 reduced the intensity and number of labeled cell bodies, so this method was of limited value in assessing the distribution of GAD immunoreactive neurons. Sections were incubated in primary antiserum diluted 1:1,500-1:2,000 in PBS containing 1% normal rabbit serum for 2-24 hours. Linking antiserum (biotinylated antigoat) was used at a dilution of 150 and the avidin-biotin complex (ABC, Vector Laboratories) was localized with 0.1%, 3,3'- diaminobenzidine tetrahydrochloride (DAB) in 0.1 M phosphate buffer, ph 7.4, containing 0.001% H202. For controls, adjacent sections were incubated in preimmune serum at a dilution of 1:1,000 and processed in the same way as the sections incubated in the anti-gad serum. Control sections showed no labeling of cell bodies or terminals. The antiserum produced by Lauder et al. ('86) was used for the demonstration of GABA immunoreactivity. This antiserum was raised in rats against GABA conjugated to keyhole limpet hemocyanin with glutaraldehyde and it shows no cross-reactivity with glutamate, aspartate, 0-alanine, taurine, or glycine. Immunoreactive staining was completely abolished by absorption with the GABA conjugate. The most robust staining of cell bodies and dendrites was achieved by fixation with the 4% paraformaldehyde, 0.1% glutaraldehyde mixture. The intensity of GABA immunoreactivity in cell bodies was not affected by alcohol pretreatment, and this was routinely used to enhance penetration. Sections were incubated in primary antiserum diluted at 1:2,000 in PBS containing 1% normal rat serum (NRS) for hours. Linking antiserum (biotinylated antirat) was used at a dilution of 150 and the ABC complex was localized with DAB as described above. In some cases, the silver intensification procedure described by Liposits et al. ('84) was used to enhance the staining of GABA immunoreactive terminals. Electron microscopy One macaque monkey was deeply anaesthetized and perfused with 0.9% saline followed by a solution of 2.0% paraformaldehyde, 1.0% glutaraldehyde in 0.1 M phosphate buffer, ph 7.4. The brain was postfixed in the same solution overnight. Blocks of tissue (1-2 mm) from area 17 were rinsed in phosphate buffer and placed in 2% osmium tetroxide in 0.1 M phosphate buffer for 1 hour. Following a brief rinse in phosphate buffer, the tissue was dehydrated and embedded in Polybed 812. In order to compare the size of axon terminals that contained flat vesicles and made symmetric synaptic contacts (FS terminals) in lamina 4C and in lamina 5, 1-2-pm-thick sections were cut, stained with toluidine blue, and viewed with the light microscope. Drawings were made to illustrate the 4C-5 border, and blocks were trimmed to include 4C, and lamina 5. Rows of thin sections were cut, placed on 0.25% Formvar-coated 2-mm slot grids, and stained with uranyl acetate and lead citrate. Measurements of the size of FS terminal profiles from lamina 4C, and lamina 5B were made from electron micrographs taken at a magnification of 15,000 with the aid of a graphics tablet and computer. Analysis of the distribution of GABAergic neurons The laminar distribution of GAD and GABA immunoreactive neurons was analyzed by drawing all of the labeled neurons in a 500-pm-wide strip extending from the pia to the white matter at a magnification of 540 x with a camera lucida. This strip of cells was then divided into sequential 65- or 70-pm horizontal sectors and the number of cells in each of these rectangular sectors (500 x 65 or 70 pm) was counted. Multiple nonoverlapping strips were drawn from nearby areas of the same section and the cell counts from all the sectors of a given depth were added together to achieve a reasonable sample size from each depth. Sample collection continued until the bin with the smallest number of cells (usually from the vicinity of lamina 4B) exceeded 20 neurons. Only those neurons in which a nucleus could be discerned were counted. The proportion of GABAergic neurons as a function of depth was determined by counting GABA immunoreactive and non-immunoreactive neurons from pm counterstained sections by using procedures identical to those described above. This was done using only the sections that had undergone alcohol pretreatment to enhance antiserum penetration. Since alcohol pretreatment is incompatible with GAD immunoreactive labeling of cell bodies, this issue was not addressed with the GAD antiserum. The relationship between the distribution of GABA and GAD immunoreactive neurons and the cytochrome-rich patches of lamina 3 was analyzed with the aid of a Videometrics M-100 graphics system attached to an IBM PC-XT. A custom-designed program superimposed a square box (0.08 pm2 in area) on a video image of the tissue section. This box served as the sample window in which cell counts were made and it could be positioned to include the cytochrome-rich or cytochrome-poor regions of the tissue section. The size of the sample window was automatically adjusted to the objective magnification so that it enclosed 0.08 pm of tissue area regardless of magnification. This allowed the borders of the cytochrome-rich regions to be seen at low power (where the boundaries appear sharper), a sample region to be chosen, and counts to be made of the number of GAD or GABA immunoreactive neurons in the sample region at higher power (so that cell nuclei could be detected). Typically, the process consisted of (1) choosing a sample site on a section treated for cytochrome oxidase with a 4 x objective and drawing both the borders of the region to be sampled ana the pattern of blood vessels on the video image with the graphics tablet; (2) replacing the cytochrome oxidase section with an adjacent section treated for GAD or GABA immunocytochemistry and aligning the blood vessel patterns; (3) changing the objective to 20 x and enlarging the sample window proportionally; and (4) counting the number of cells within the borders of the sample region. Cells falling on the border of the sample region were included in the sample. The cross-sectional area of the neurons within a given layer was also measured with the computer graphics system. The borders of the neurons were drawn at a magnification of 1,050 x on the video monitor. To ensure an accurate representation of the size distribution of cells within a layer, samples were taken throughout the depth of each layer. Statistical methods for comparison of samples were taken from Hays and Winkler ('71). RESULTS GABAergic axon terminals Sections processed for GAD immunoreactivity and those processed for GABA immunoreactivity show a distinct punctate pattern of labeling that is distributed throughout

4 76 D. FITZPATRICK ET AL. 4c 5 Fig. 1. Darkfield photomicrograph showing the distribution of GAD immunoreactive axon terminals in a coronal section through macaque striate cortex. GAD immunoreactive axon terminals are prominent in zones that receive direct projections from the LGN: laminae 4C, 4A and the patches in lamina 3. Scale bar = 150 pm. the neuropil or directly apposed to cell bodies. While some lamina 3A that have been processed for cytochrome ox<dase of these immunoreactive elements may be dendritic pro- and for GAD immunoreactivity. cesses or fibers cut in cross section, the vast majority are The precision with which the prominent GAD immunolikely to be axon terminals, and we have chosen to apply reactive terminals match the pattern of geniculate termithis term to the punctate immunoreactive staining. In gen- nations is exemplified by the intricate pattern of tereral, the GAD antiserum produced more robust staining of minations that is found in lamina 4A. The projections from axon terminals than the GABA antiserum, but both anti- the LGN to lamina 4A have been shown to terminate in a sera produced comparable patterns of labeled terminals weblike or honeycomb pattern when viewed in tangential throughout the cortex. sections (Hendrickson et al., '78; Blasdel and Lund, '82), GAD and GABA immunoreactive axon terminals are and this pattern is reflected in the distribution of cytofound in all layers of striate cortex from layer 1 to the deep chrome oxidase staining in this region. Tangential sections aspect of layer 6, but are particularly striking in the zones through lamina 4A show that GAD immunoreactive terthat have been shown to receive the bulk of the projection minals are also distributed in a honeycomb pattern and from the LGN; laminae 4C and 4A (Fig. 1). GAD and GABA that this pattern is strikingly similar to the pattern of immunoreactive terminals are also more prominent in the cytochrome oxidase staining (Fig. 3). While the photomicrogeniculate recipient, cytochrome-rich patches of lamina 3, a graphs of Figure 3 are from adjacent sections, the honeyfinding first reported by Hendrickson et al. ('81).The coin- comb pattern of cytochrome oxidase staining appears to be cidence between the cytochrome-rich and GAD immuno- shifted to the left of the prominent GAD terminal distribureactive patches can be seen in Figure 2, which shows pho- tion. This shift is due to the difficulty in achieving perfectly tomicrographs from adjacent tangential sections through tangential sections through a layer as thin as 4A. Never-

5 GABAERGIC NEURONS IN STRIATE CORTEX 77 Fig. 2. Comparison of the distribution of prominent GAD immunoreactive terminal staining and cytochrome oxidase staining in adjacent tangential sections through lamina 3 of macaque striate cortex. a: GAD immunocytochemistry (darkfield illumination). b: Cytochrome oxidase histochemistry (brightfield illumination). Cytochrome-oxidase-rich patches are distingushed by prominent GAD immunoreactive terminal staining. Scale bar = 300 pm. theless, the similarity in the size and shape of the two patterns provides strong evidence that they are coincident. The phrase prominence of terminal staining has been used to refer to the difference in the quality of terminal staining that distinguishes the geniculate recipient regions from the others. At first glance, it might be assumed that this difference is simply a reflection of the density of GA- BAergic terminals. But when viewed at higher-power magnification, GABAergic axon terminals appear to vary in size, and many of the GAD and GABA immunoreactive terminals in the geniculate-recipient regions appear larger in caliber than those in regions that do not receive geniculate terminals (Fig. 4). This apparent difference can be seen between lamina 4C and the immediately adjacent laminae (4B and 5) and between the cytochrome-rich and cytochrome-poor regions of laminae 3 and 4A. A difference in size of this magnitude would be difficult to quantify at the light microscopic level, so the size of putative GABAergic terminals was examined with the electron microscope. It is generally accepted that GAD and GABA immunoreactive terminals in the cortex have a distinctive morphology. They contain flattened or pleomorphic vesicles and are associated with symmetric synaptic thickenings (Ribak, 78; Hendry et al., 83; Somogyi et al., 83b). For the sake of con- venience, profiles with this morphology will be referred to as FS (flat-symmetrical) profiles. If the population of GA- BAergic axon terminals in the geniculate-recipient regions are larger than those in other regions, there should be a detectable difference in the cross-sectional area of the FS profiles in these regions. To test this idea, the cross-sectional area of FS profiles in lamina 4C, were measured and compared with the cross-sectional areas of FS profiles in lamina 5. Typical electron micrographs of FS profiles and profiles with round vesicles and asymmetric contacts (RA profiles) from lamina 4C, of macaque striate cortex are shown in Figure 5. The distributions of area measurements for 158 FS profiles from lamina 4C, and 124 FS profiles from lamina 5B are shown in Figure 6 (a and b). Large profiles (> 0.6 pm2) appear to make up a greater proportion of the FS profiles in lamina 4C, than their counterparts in lamina 5, and conversely, small profiles (< 0.4 pm2) make up a larger proportion of the FS profiles in lamina 5 than their counterparts in lamina 4C,. The difference between the two distributions is best seen in Figure 6c, where the values for each bin in the lamina 5 distribution have been subtracted from the corresponding bins in the lamina 4C, distribution. Statistical analysis supports the interpretation that the profiles in lamina 4C, are significantly larger than their

6 78 D. FITZPATRICK ET AL. Fig. 3. Comparison of the distribution of prominent GAD immunoreactive terminal staining and cytochrome oxidase staining in adjacent tangential sections through lamina 4A of macaque striate cortex. a: GAD immunocytochcmistry (darkfield illumination). b: Cytochrome oxidase histochemistry (brightfield illurnination). Note the similarity in the honeycomb distribution of GAD immunoreactive terminals and the pattern of cytochronie oxidase activity. Scale bar = 150 pm. counterparts in lamina 5 (lamina 4C,: mean = 0.49 pm2, SD = 0.3; lamina 5: mean = 0.35 pm2, SD = 0.2; Kolmogorov-Smirnov two-sample test: P <.001). These distributions probably underestimate the actual size of FS boutons since many of the profiles with small cross-sectional areas are parts of larger boutons that have not been sectioned through their centers. Despite this limitation, the results show that FS profiles in lamina 4C, are, on average, 40% larger in cross-sectional area than their counterparts in lamina 5. GABAergic cell bodies Laminar differences in size and morphology. Both antisera labeled neurons that were distributed across all layers of striate cortex. However, the staining for GABA was much more robust and this staining often continued into the proximal portions of dendritic processes. Examples of the morphology of the GABA immunoreactive neurons from different layers of striate cortex are shown in Figures 7 and 8. GABA immunoreactive neurons have smooth or occasionally beaded dendrites that are displayed in multipolar or bitufted dendritic arrangements. Each layer contains a morphologically heterogeneous population of GABAergic neurons that differ in cell size as well as dendritic configuration. In general, the GABA immunoreactive neurons in the superficial layers appear smaller in size than those in other layers and have predominantly vertically oriented dendritic fields. Larger GABA immunoreactive neurons with horizontally oriented dendritic processes are most common in lamina 4C, and in lamina 6. General impressions about the size distribution of GA- BAergic neurons were confirmed by measurements of the cross-sectional area of GAD and GABA immunoreactive cell bodies in different layers (Table 1). The GABA immunoreactive population as a whole shows a considerable variation in size, from 25 to 200 pm2 in cross-sectional area. While each layer has a range of cell sizes, the largest GAD and GABA immunoreactive neurons are found in lamina 4C, and in lamina 6. Indeed, these are the only two layers that contain immunoreactive neurons with soma areas greater than 140 pm2. At the same time, neurons in laminae 2 and 3A are distinctly smaller than those in the other layers; in our sample of 214 neurons from 2-3A, none were found to exceed 110 pm2 in area. The sizes of GABA and GAD immunoreactive neurons in the portions of lamina 4C that receive their projections from the magnocellular and parvicellular layers of the LGN were examined in more detail. Figure 9 shows histograms of the soma area of GABA and GAD immunoreactive neurons in 4C, and 4C, from two different macaque monkeys

7 GABAERGIC NEURONS IN STRIATE CORTEX 79 Fig. 4. Comparison of GAD and GABA immunoreactive axon terminals in lamina 4C, and lamina 5. a: GAD immunoreactive axon terminals in lamina 4C,. b: GABA immunoreactive axon terminals in lamina 4C,. c: GAD immunoreactive axon terminals in lamina 5. d: GABA immunoreactive axon terminals in lamina 5. Many of the immunoreactive axon terminals in lamina 4C, appear larger than those in lamina 5. The staining of the GABA immunoreactive terminals in the material used for this comparison was enhanced by the silver intensification procedure described by Liposits et al. ('84). Scale bar = 5 pm. perfused with fixatives that were optimal for each antiserum. In both cases, the distribution of GABAergic neurons in 4C, has a broader range than that of 4C, and includes large neurons that are not found within 4C,. These differences are statistically significant (Kolmogorov-Smirnov two-sample test P <.001) The largest GAD and GABA immunoreactive neurons in lamina 4C, are not distributed evenly throughout the layer but appear more frequently near the lamina 4B border. The size of the neurons in 4C was plotted as a function of their distance from the top of 4C, (the top of 4C being defined as the upper boundary of the zone of prominent GABA immunoreactive axon terminals), and the results of this analysis are presented in Figure 10 (total 316 neurons). This plot shows that a population of large GABA immunoreactive neurons (> 120 pm2) is clustered in the upper 20% of lamina 4C (or the upper half of 4CJ. These large GABA immunoreactive neurons make up a small percentage of the total population of GABA immunoreactive neurons in lamina 4C, (roughly 7%; nine out of 135), but their restriction to the upper half of 4C, is significant since there are other lines of evidence that the upper part of 4C, is distinct from the lower part (see Discussion). Density andproportion of GABAergic neurons across cortical layers. While GABA and GAD immunoreactive neurons are found in all layers of striate cortex, there are obvious differences in their density across the cortical depth. GAD and GABA immunoreactive neurons are more dense in laminae 2-3,4A, and 4C, than in the other laminae. The density of GABA and GAD immunoreactive neurons as a function depth is depicted in the histograms of Figures lla and 12. A low-power photomicrograph showing the laminar distribution of GABA immunoreactive neurons is shown in Figure 13. The fluctuation in the density of GABAergic neurons across cortical layers resembles the fluctuation in the density of unlabeled neurons and this relationship is depicted in Figure llb. In general, regions with a high density of

8 80 D. FITZPATRICK ET AL. Fig. 5. Examples of profiles with flat vesicles and symmetric synaptic contacts (FS) and those with round vesicles and asymmetric synaptic from lamina 4C,. Arrows point to sites of synaptic contact. Scale bar = 0.2 pm. GABA immunoreactive neurons also appear to have a high density of unlabeled neurons (4C,, 4A, 2-3A) while zones with a low density of GABA immunoreactive neurons (4B and 5) contain fewer unlabeled neurons. The similarity in the fluctuations in the density of labeled and unlabeled populations across the cortical depth is reflected in the fact that layers that differ markedly in their total density of neurons have similar percentages of GABA immunoreactive neurons. For example, lamina 4C,, which contains the greatest density of neurons, has a percentage of GABA immunoreactive neurons that is not significantly different from lamina 4B or 4Ccy- layers that contain far fewer total neurons (Fig. llc, Table 2). However, the fluctuation in the density of the GABA immunoreactive population is not a simple function of the density of the unlabeled population. If this were true, then the percentage of GABA immunoreactive neurons should be constant across all cortical layers, and, as shown in Figure 8c and Table 2, this was not the case. GABA immunoreactive neurons make up approximately 15% of all the neurons in striate cortex, but layer 1 is clearly distinct from the other layers in containing a very high proportion (> 75%) of GABA immunoreactive neurons. This finding is consistent with previous reports in other species and in other cortical areas (Schmechel et al., '84; Gabbott and Somogyi, '86; Lauder et al., '86; Lin et al., '85). Differences between other layers in the proportion of GABA immunoreactive neurons are modest by comparison with lamina 1. In general the proportion of GABA immunoreactive neurons is highest in the superficial layers, decreases through the middle layers, and is lowest in the deep layers. In both monkeys that were examined, lamina 3A had the greatest proportion of GABA immunoreactive neurons (roughly 20%), and this value was almost twice that found in lamina 5 and lamina 6. These differences are statistically significant (lamina 3 vs. lamina 5: x2 = 10.9, d.f. = 1, P <.Ol); lamina 3 vs. lamina 6: x2 = 8.8, d.f. = 1, P <.01). Areal distribution of GABAergic neurons within lamina 3. One question that remains to be addressed is whether the patchy distribution of GABergic axon terminals in lamina 3 is associated with a patchy distribution of GABAergic cell bodies. This issue was examined by comparing adjacent tangential sections through lamina 3 that had been reacting for cytochrome oxidase and for GABA immunocytochemistry (Fig. 14). There are no obvious signs of a clustered or patchy distribution of GABA immunoreactive neurons that might correspond to the pattern of cytochrome-oxidaserich patches. A similar result was found with the GAD antiserum, and an example of the relationship between GAD immunoreactive cell bodies and the cytochrome-oxidase-rich patches is shown in Figure 15. From the photo-

9 GABAERGIC NEURONS IN STRIATE CORTEX 81 LAMINA % 14- m 12- LAMINA 4CO , 3' $!? 26- m c E I : ; Q) 14- m 12- c. c 10-0 LAMINA C FS Cross-sectional Area Ipm21 Fig. 6. Cross-sectional area of FS profiles from lamina 4C, (a) and lamina 5 (b) of macaque striate cortex, The distributions are significantly different and the nature of this difference is best seen in c, where the bins from the lamina 5 distribution have been subtracted from the lamina 4C, distribution. Lamina 4C, has a greater percentage of FS terminals that are larger than 0.5 pm2 in cross-sectional area. Fig. 7. Camera lucida drawings of GABA immunoreactive neurons from different layers of macaque striate cortex. All neurons were drawn at the same magnification but the widths of the layers are not drawn to scale.

10 82 D. FITZPATRICK ET AL. Fig. 8. Photomicrographs of GABA immunoreactive neurons in different laminae of macaque striate cortex. a-e: Lamina 2-3. f,g: Lamina 4C,,. h,i: Lamina 4C,. j-i: Lamina 6. Scale bar = 20 pm.

11 GABAERGIC NEURONS IN STRIATE CORTEX 83 TABLE 1. Cross-Sectional Area Measurements of GABA and GAD Immunoreactive Neurons in Macaque Striate Cortex' Laver No. Mean SD Range GABA neurons 2-3A 3B-4A 4B 4C," 4C6 5 6 GADneurons 2-3A 3B-4A 4B 4c 4c;; 'Measurements (pm2) are from two different macaque monkeys perfused with fixative optimal for each antiserum. micrographs and the plots, it becomes apparent that if there is any difference in the density of GABAergic neurons in the cytochrome-rich and cytochrome-poor regions of lamina 3, this difference must be small. To test this possibility, the number of GABA immunoreactive neurons in equal-size samples of cytochrome-rich and c tochrome-poor regions was determined. A total of 1.3 mm x of the cytochrome-rich patches and 1.3 mm2 of the cytochrome-poor zones were examined and 1,595 GABA immunoreactive neurons were counted. There was considerable variation in the number of GABA immunoreactive neurons found in the sampies from both regions. The average density of GABA immunoreactive neurons was 586 neurons /mm2 (SD = 90) in the cytochrome-rich regions and 624 neurons/mm2 (SD = 107) in the cytochrome-poor zones. The slightly greater number of GABA immunoreactive neurons found in the cytochrome-poor regions (6%) is not statistically significant (Kolmogorov-Smirnov two-sample test: P =.49). Thus, while there is a difference in the GABAergic axon terminals in DISTRIBUTION OF SOMA SIZE IN LAMINA 4C GABA Neurons in 4Ca GAD Neurons in 4Ca N = N=176 X = 87.2 * X = 98.7 c C C SD = 34.1 SD = 34.7? 2 n a GABA Neurons in 4Cp N = L c C C a, SD= a n GAD Neurons in 4cp N=196 K=72.0 SD = Soma Area vm2 Soma Area pm2 Fig. 9. Distribution of cross-sectional area measurements for GAD and GABA immunoreactive neurons in laminae 4C,, and 4C,, of macaque striate cortex.

12 84 D. FITZPATRICK ET AL. DISTRIBUTION OF GAEA IMMUNOREACTIVE NEURONS t I A I A I C0 ]4Cp I 70. = : I.... * : *. -..._ ~. 4 v. m....., a. :....- * Percent Distance from Top of 4C 6 0. E U DlSTRlEUTlON OF GAEA AND UNLABELED NEURONS Fig. 10. Cross-sectional area of GABA immunoreactive neurons as a function of their depth from the top of lamina 4C,,. the cytochrome-rich and cytochrome-poor regions, this difference is not correlated with a difference in the density of GABAergic cell bodies. DISCUSSION In this study antisera to GAD and GABA have been used to show that GABAergic neurons constitute a significant proportion of the neurons in macaque striate cortex and that there are differences between laminae in the number and sizes of GABAergic cell bodies and axon terminals. While GAD and GABA immunoreactivity was not examined in tissue from the same animal, the laminar distribution of labeled neurons was found to be quite similar in tissue that was fixed optimally for each antiserum (compare Figs. lla and 12). Both antisera also showed similar laminar differences in the size of labeled neurons and in the size and/or density of immunoreactive terminals. However, there were differences in the total density of labeled cell bodies and terminals stained by these antisera; the number of cell bodies labeled and the intensity of cell body labeling seemed greater with the GABA antiserum than with the GAD antiserum. Conversely, axon terminal labeling generally appeared more robust with the GAD antiserum than with the GABA antiserum. These differences in staining intensity do not appear to be unique to the particular antisera that were used in this study. Somogyi et al. ('84a) used a different set of GAD and GABA antisera in the study of GABAergic neurons in cat striate cortex and also noted the increased effectiveness of GABA antiserum for labeling cell bodies. Hendry and Jones ('86) applied still-another GABA antiserum to macaque striate cortex, and their photomicrographs show cell body and terminal staining levels comparable to the GABA antiserum used in the present study. Whatever the reasons for these differences, the overall similarities in the pattern of GAD and GABA immunoreactivity strengthen the argument that the labeled cells and terminals are involved in GABAergic neurotransmission. GABAergic axon terminals In the present report we have noted the coincidence in the distribution of prominent GABAergic axon terminals b g 220 g 200 I f 140 E 120 = C ff PERCENTAGE OF GABA IMMUNOREACTIVE NEURONS GABAIIGABA +Unlabeledl Depth in 65 pm Intervals Fig. 11. a: Distribution of GABA immunoreactive neurons across the depth of macaque striate cortex. The number of neurons in lamina 1 has been collected in a single bin. The remaining bins reflect the number of neurons in successive 65pm steps through the cortex. Approximate laminar boundaries are indicated at the top. The number of neurons in each bin is expressed as a percentage of the total population of GABA immunoreactive neurons. The actual number of GABA immunoreactive neurons in each bin is indicated in Figure llb. b: Comparison of the distribution of GABA immunoreactive neurons (white) and unlabeled neurons (black) across the depth of macaque striate cortex. The height of each bar represents the total number of neurons in each sample (GABA + unlabeled). c: Percentage of neurons that are GABA immunoreactive as a function of depth. The number of GABA immunoreactive neurons in each bin was divided by the total number of neurons in the bin to arrive at a percentage value.

13 GABAERGIC NEURONS IN STRIATE CORTEX DISTRIBUTION OF GAD IMMUNOREACTIVE NEURONS Depth in 70 pm Intervals I I I 4C0 14Cp 5 I 6 I l n l 1 Fig. 12. Distribution of GAD immunoreactive neurons in 70-pm steps through macaque striate cortex. and the pattern of termination of axons from the LGN. A prominence of GABAergic axon terminals in the geniculate-recipient regions of the striate cortex has also been found in the cat and squirrel monkey (Fitzpatrick et al., '83b; Bear et al., '85). At the very least, this suggests that there is some particular specialization in the inhibitory circuitry associated with the geniculate-recipient regions that may play a crucial role in the transfer of information from LGN axons to neurons within the striate cortex. The specialized nature of the GAD immunoreactive terminals within the geniculate-recipient zones of macaque striate cortex was first noted by Hendrickson et al. ('81) and interpreted as an increase in their density. While the present results do not contradict the idea that GAD-immunoreactive axon terminals in these regions may be more numerous, they do suggest that, as a population, they are larger in size than the axon terminals in the regions that do not receive geniculate afferents. A difference in the size and density of the GAD immunoreactive terminals has also been noted between lamina 4 and other layers of macaque somatic cortex (Houser et al., '84). 85 Fig. 13. Photomicrograph showing the distribution of GABA immunoreactive neurons in a coronal section through macaque striate cortex. Scale bar = 150 pm.

14 86 TABLE 2. Numbers of GABA Immunoreactive Neurons in Different Layers of Macaque Striate Cortex' No. of GABA and No. of % unlabeled GABA GABA Layer neurons neurons neurons A B A B c,, C, Total 3, (Avg.) A B A B C, c,i Total 2, (Avg.) 14.9 'Figures are from two different macaque monkeys perfused with the optimum fixative for GABA immunoreactive cell bodies (see Materials and Methods). D. FITZPATRICK ET AL. In the original description of GAD immunoreactivity in macaque striate cortex (Hendrickson et al., %1), attention was focused on the similarity in the distribution of GAD immunoreactivity and the patchy pattern of cytochrome oxidase staining in lamina 3. However, it is not clear from that report whether the increased GAD immunoreactivity in the cytochrome-rich patches consisted of GAD immunoreactive terminals alone or an increase in both labeled cell bodies as well as terminals. Another study using the uptake of 3H-GABA as a marker for GABAergic neurons presented evidence for an increased density of GABA-accumulating neurons within the cytochrome-rich patches (Carroll and Wong-Riley, '85). The present report demonstrates that the cytochrome-rich regions of lamina 3 are distinguished by an increased size and, perhaps, density of GAD immunoreactive terminals. However, these regions do not differ from the rest of lamina 3 in their density of GABAergic neurons. These results are in agreement with a previous report of the distribution of GAD immunoreactive neurons in lamina 3 of squirrel monkey striate cortex (Fitzpatrick et al., '83b) and the recent description of GABA immunoreactive neurons in macaque striate cortex (Jones et al., '86). Since the cytochrome-rich patches in lamina 3 are the targets of a direct projection from the LGN (Livingstone and Hubel, '82; Fitzpatrick et al., '83a), it seems reasonable to view the distinctive pattern of GAD immunoreactive terminals as another example of the prominent role that Fig. 14. Photomicrographs showing the distribution of cytochrome-oxidase-rich patches and GABA immunowactive neurons in adjacent tangential sections through lamina 3 of macaque striate cortex. a: Cytochrome osidasc histochemistry. b: GABA immunocytochemistry. Scale bar = 200 pm.

15 GABAERGIC NEURONS IN STRIATE CORTEX 87 a b Fig. 15. a: Distribution of GAD immnnoreactive neurons in a tangential section through lamina 3 of macaque striate cortex. Each dot represents a single labeled neuron. Circles represent blood vessels used as landmarks. b: Comparison of the distribution of GAD immunoreactive neurons shown in Figure 15a with the distribution of cytochrome oxidase activity in an adjacent section. Cytochrome-oxidase-rich regions are depicted with dark gray stipple. Scale bar = 250 pm.

16 88 D. FITZPATRICK ET AL. GABAergic inhibition plays in regions that receive direct projections from the LGN. The coincidence in the spatial distribution of large GA- BAergic axon terminals and terminals derived from the LGN is further emphasized by their correspondence in the honeycomb matrix of lamina 4A. The exact significance of the honeycomb distribution of geniculate terminations in lamina 4A remains unknown. Presumably it reflects some form of segregation between different afferents to this region, such as those from the LGN and from lamina 4C (Fitzpatrick et al., 851, or some specialization in the arrangement of the dendrites that are postsynaptic to the geniculate terminations. Whatever the basis for this pattern of terminations, the close spatial correspondence of GABAergic axon terminals and LGN axon terminals raises the question of whether some of the GAD immunoreactive axon terminals might be derived from the LGN. Most lines of evidence do not support the idea that GABAergic neurons in the LGN project to striate cortex. For example, GAD or GABA immunoreactive neurons within the LGN are not labeled following injections of retrograde tracers into cat (Fitzpatrick et al., 84; Montero and Zempel, 85) or macaque (Montero, 86) striate cortex, and GAD immunoreactive terminals within layer 4 are not noticeably diminished following the undercutting of striate cortex (Bear et al., 85). Furthermore, most studies report that LGN axon terminals have round vesicles and establish asymmetic synaptic contacts (Garey and Powell, 71; LeVay and Gilbert, 76; LeVay, 86; Einstein et al., 87), in contrast to GAD immunoreactive profiles that have flat vesicles and make symmetric synaptic contacts (Ribak, 78; Hendry et al., 83; Somogyi et al., 83b). While Einstein et al. ( 87) have recently reported that some of the profiles within layer 4 of cat striate cortex that are labeled following injections of 3H-amino acids into the LGN have flattened vesicles and make symmetric synaptic contacts, there is, at present, no evidence that these terminals are immunoreactive for GAD or GABA. If one accepts the prevailing view that GABAergic terminals within cortex have a local origin, then it is natural to wonder whether the large GABAergic terminals in the geniculate-recipient zones originate from a particular type of neuron. One possible source for the prominent endings in the geniculate recipient regions is a class of small aspinous neurons that has been given a variety of names including spider cell, clewed cell, and more recently, clutch cell or B1 variety (Ramon y Cajal, 1899; Valverde, 71; Lund 73, 87; Lund et al., 77; Mates and Lund, 83; Kisvarday et al., 85, 86). This suggestion is consistent with the fact that these neurons (1) appear to be restricted to layers 4C and 4A, where they are the most frequently impregnated smooth dendritic neurons; (2) have axons characterized by a dense, highly branched field close to the soma and by the presence of conspicuously large boutons; and (3) are immunoreactive for GABA. Kisvarday et al. (1986) have also argued for a unique association between the axon terminals of this cell class and those from the LGN, citing the similarity in the sizes of the axonal fields of clutch cells and LGN afferents and the correspondence in their postsynaptic targets within lamina 4C,. In the macaque the major targets for both clutch cell axon terminals and LGN axon terminals are spines and dendritic shafts, presumed to be from spiny stellate neurons. It would be of some interest to determine the spatial relationship between the synaptic contacts of LGN terminals and clutch cell terminals along the dendrites of individual postsynaptic neurons. Perhaps the matching distribution of GABAergic and LGN terminals in lamina 4A signifies an underlying similarity in the postsynaptic targets of these two populations of boutons. Whether or not the large GABAergic axon terminals are derived from a single cell type, the question remains as to the functional significance of prominent terminal staining in the geniculate-recipient regions. One major difference between neurons in the LGN and those in striate cortex lies in their level of spontaneous activity. Neurons in the LGN have spontaneous activity levels at least three times higher than neurons in the striate cortex (Creutzfield and Ito, 68). This difference that can be detected in the cortical termination zones of geniculate afferents (low spontaneous discharge rates characterize the neurons in lamina 4C and lamina 4A in monkey striate cortex; Blasdel and Fitzpatrick, 84). Presumably, this difference reflects the fact that neurons within layer 4 are under the influence of tonic inhibition derived from interneurons within the cortex. Perhaps the larger-caliber swellings of GABAergic neurons in the geniculate-recipient regions are specializations that act to balance the strong excitatory drive of the geniculate afferents. For example, large-caliber endings provide a larger surface area that could give rise to a greater number of synaptic contacts. Against this suggestion is the evidence from Beaulieu and Colonnier ( 85) that type 2 synaptic contacts are not more numerous in layers 4A and 4B in cat striate cortex-layers also distingushed by large-caliber GAD immunoreactive profiles. Another possibility is that the more frequent release of neurotransmitters which would be expected in a tonically active system might require a larger pool of synaptic vesicles or a larger number of mitochondria-needs that could be met by a larger axon terminal. In this respect it is interesting to note that Kageyama and Wong-Riley ( 86) have described a distinct type of largecaliber axon terminal in layer 4 of cat striate cortex that establishes symmetric synaptic contacts and contains a large number of mitochondria that are rich in cytochrome oxidase activity. Also, the recent observation that the axonal swellings of clutch cells contain large groups of mitochondria is consistent with this idea (Kisvarday et al., 85). Distribution of GABAergic cell bodies The results of this study support the idea that GABAergic neurons constitute at least 15% of the neurons in monkey striate cortex. Previous attempts to estimate the percentage of GABAergic neurons in macaque neocortex have produced a wide range of values. For example, Mates and Lund ( 83) used the observation that Golgi-impregnated aspinous neurons are distinguished from spiny neurons by the presence of both type 1 and type 2 contacts on their somata (LeVay, 73) to estimate that roughly 6% of the neurons in lamina 4C, were aspiny stellate neurons. In contrast, Hendry and Jones ( 81) suggested that up to 40% of the neurons in monkey somatosensory and motor cortex might be GA- BAergic since they appeared to selectively accumulate 3H- GABA. The present results are closer to the values obtained with GABA immunocytochemistry by Gabbott and Somogyi ( 86) in cat striate cortex (20%) and recently by Jones et al. ( 86) in monkey striate cortex (20%). While GABA immunocytochemistry appears to be a reliable method for identifying GABAergic neurons, it does

17 GABAERGIC NEURONS IN STRIATE CORTEX 89 have several limitations that may influence the accuracy of quantitative analysis. For example, even though pm vibratome sections were used for cell counts, and alcohol pretreatment was used to increase the depth of penetration, it is still possible that our figures underestimate the total number of GABA immunoreactive neurons because a certain percentage of small cell bodies that lie in the center of the section were not detected. Likewise, it is possible that some GABAergic neurons are not detected because the levels of GABA within their somata fall below the detection threshold of the antiserum. Recent evidence that the number of GABA immunoreactive neurons detected in layer 4 of macaque striate cortex is reduced by enucleation or visual deprivation (Hendry and Jones, '86) suggests that GABA levels within neurons can vary and opens up the possibility that at any given time, some percentage of the GABA-containing population may be below the threshold for detection. Given these considerations, it seems prudent to view cell counts as minimal estimates since a variety of factors, physiological and/or methodological, may affect the number of cells labeled in any given animal. Both the GAD and GABA antisera show consistent laminar differences in the density of labeled neurons. The greatest density of labeled neurons is found in laminae 4C,, 4A and in laminae 2-3. The increased density of labeled neurons in the geniculate-recipient zones (4C, and 4A) is consistent with the observations made by Gabbott and Somogyi ('86) in cat visual cortex. In the cat, the greatest number of GABAergic neurons is found in laminae 4 and 3B, and since these regions are the major targets of geniculate afferents, they argued that this distribution reflects the allocation of a large share of cortical GABAergic inhibition to the early stages of visual processing. However, it should be pointed out that laminae 4 and 3B in the cat and lamina 4C, and 4A in the monkey also contain the greatest density of non-gabaergic neurons. As a result, the ratio of GABA immunoreactive neurons to the unlabeled population in these layers is not greater (and may be less) than that found at other levels within the superficial layers. In addition, lamina 4C, and the cytochrome-rich patches of lamina 3 in macaque are prominent geniculate-recipient regions, but they do not have an increased density of GABAergic neurons. There does seem to be a significant difference in the percentage of GABA immunoreactive neurons in the superficial and deep layers of striate cortex. The present study suggests that almost 20% of the neurons in the upper part of lamina 3 are GABAergic, while roughly 12% of the neurons in laminae 5 and 6 are GABAergic. A difference in the proportion of GABAergic neurons in the superficial and deep layers has been reported also in the visual cortex of the cat (Gabbott and Somogyi, '86). This difference may reflect the fact that there is a greater number of potential synaptic contact sites for the GABAergic neurons within lamina 3 since they have access to the processes of neurons whose somata lie within the layer as well as the apical dendrites of pyramidal neurons whose somata lie in the deeper layers. This seems unlikely to be the whole explanation, however, since at least some GABAergic neurons in the superficial layers project to the deep layers and vice versa (Valverde, '71; Lund, '73; Somogyi et al., '81, '83a). Morphology of GABAergic neurons The results of this study are consistent with the evidence derived from a variety of sources that GABAergic neurons in macaque striate cortex are non-pyramidal neurons with smooth or beaded dendritic processes. In addition, these results demonstrate conspicuous laminar differences in the soma size and the orientation of the initial dendritic segments of GABAergic neurons. Perhaps the most distinctive neurons observed in this study are the large GABA and GAD immunoreactive neurons found in the upper part of 4C, and in lamina 6. Jones ('75) and DeFelipe et al. ('86) have argued that the largest of the smooth dendritic neurons in somatic and motor cortex of macaque belong to the class of cells referred to as basket cells. Basket cells are distinguished from most other classes of smooth dendritic neurons by axon arbors that form terminal nests around the somata of pyramidal neurons and other cell types (Marin-Padilla, '69; Szentagothai, '73; Jones '75; Somogyi et al., '83~). If this difference in size applies to the smooth dendritic population in macaque striate cortex, then the large neurons in the upper part of 4C, and in lamina 6 would be good candidates for basket cells. A recent Golgi study (Lund, '87) has described a population of aspinous neurons restricted to the upper part of 4C, that has horizontally spreading dendrites much like the GABA immunoreactive neurons described here. These neurons also have axon arbors that extend horizontally for long distances (over 1 mm) within upper 4C, and 4B and give rise to short vertical collaterals resembling those identified as basket cell terminations by Somogyi et al. ('83c) and DeFelipe et al. ('86). Long-distance lateral connections are also characteristic of basket cells in somatosensory and motor cortex (Marin-Padilla, '69; Jones, '75). Large aspinous neurons with horizontally displayed dendrites in lamina 6 have received only casual attention in previous studies using the Golgi method. Jones ('75) noted that basket cells in layer 6 of somatosensory cortex could be extended in a horizontal form, and both Lund ('73) and Tombol('78) mentioned the presence of large aspinous neurons with horizontally directed dendritic processes in lamina 6 of monkey striate cortex. Horizontally directed axonal processes with basketlike terminations have been observed in lamina 6 of macaque striate cortex, but the source of these axon arbors remains to be determined (Lund, unpublished observations). If the large GABA immunoreactive neurons in lamina 4C, and lamina 6 are, indeed, basket cells, then these regions may be expected to have long-distance lateral connections that are, at least in part, inhibitory. Long-distance lateral connections are also characteristic of the superficial layers of striate cortex, but the bulk of the evidence suggests that these projections originate in pyramidal neurons and that their effect is excitatory (Gilbert and Wiesel, '83; Rockland and Lund, '83; Rockland, '85; Ts'o et al., '86). Differences in the lateral extent of inhibitory connections could obviously contribute to laminar differences in receptive field properties. In this context it is worth noting that lamina 4C,, 4B and lamina 6 contain neurons with directionally selective receptive fields and that neurons in 4B and 6 are the source of projections to the middle temporal area (MT) which contains a high proportion of directionally selective neurons (Dow, '74; Rockland and Pandya, '79; Tigges et al., '81; Livingstone and Hubel, '84; Hawken and Parker, personal communication). Whether long-distance inhibitory connections contribute to the directionally selective responses of these neurons remains to be determined. Finally, the presence of a population of large GABA immunoreactive neurons in the upper portion of 4C, provides

18 90 D. FITZPATRICK ET AL. further evidence for a functional subdivision within the cortical target of the magnocellular layers of the LGN. The original subdivision of lamina 4C into two tiers, a and 0, was based on differences in the dendritic morphology and projections of spiny stellate neurons: those in 4C, were shown to send their axons into lamina 4B, while the spiny stellate neurons in 4C3 were shown to send their axons vertically through 4B, without arborization, to terminate in lamina 3B (Lund,'73). This distinction was reinforced by the demonstration that the magnocellular and parvicellular layers of the LGN had largely separate terminal fields within 4C,, and 4C,, respectively (Hubel and Wiesel, '72; Hendrickson et al., '78). More recent evidence suggests that 4C, and 4C,, may not be uniform throughout their depth, and that both contain sublaminae with different patterns of connections (Blasdel et al., '85; Fitzpatrick et al., '85). The upper part of 4C, appears to be distinguished from the lower part by (1) the termination of a distinct type of geniculocortical axon (Blasdel and Lund, '82); (2) the presence of laterally spreading axon arbors that terminate in patches in umer 4C- and in 4B (Rockland and Lund. '83: Blasdel et II,, al., '85); and (3) the presence of orientation-se~ective neurons (Blasdel and Fitzpatrick, >84). our finding of a class of large GABAergic in upper 4cm adds to the evidence for further subdivision of 4C, and complements Lund'S observation that upper 4c, contains aspinous neurons that are distinct from those in lower 4C, and 4C, (Lund, '87). ACKNOWLEDGMENTS by grants EY and EY- Dow, B.M. (1974) Functional classes of cells and their laminar distribution in monkey visual cortex. J. Neurophysiol. 37: Einstein, G., T.L. Davis, and P. Sterling (1987) The ultrastructure of synapses from the A laminae of the lateral geniculate nucleus in layer IV of cat striate cortex. J. Comp. Neurol. 260: Fitzpatrick, D., K. Itoh, and LT. Diamond (1983a) The laminar organization of the lateral geniculate body and the striate cortex in the squirrel monkey (Saimiri sciurens). J. Neurosci. 3t Fitzpatrick, D., J.S. Lund, and D.E. Schmechel(1983b) Glutamic acid decarhoxylase immunoreactive neurons and terminals in the visual cortex of monkey and cat. Soc. Neurosci. Abstr. 9:616. Fitzpatrick, D., G.R. Penny, and D.E. Schmechel (1984) Glutamic acid decarboxylase-immunoreactive neurons and terminals in the lateral geniculate nucleus of the cat. J. Neurosci. 4r Fitzpatrick, D., J.S. Lund, and G.G. Blasdel (1985) Intrinsic connections of macaque striate cortex: Afferent and efferent connections of lamina 4C. J. Neurosci Freund, T.G., K.A.C. Martin, A.D. Smith, and P. Somogyi (1983) Glutamate decarboxylase-immunoreactive terminals of Golgi-impregnated axoaxonic cells and of presumed basket cells in synaptic contact with pyramidal neurons of the cat's visual cortex. J. Comp. Neurol. 221: Gahbott, P.L.A., and P. Somogyi (1986) Quantitative distribution of GABAimmunoreactive neurons in the visual cortex (area 17) of the cat. Exp. Brain Res. 61: Garey, L.J., and T.P.S. Powell (1971) An experimental study of the termination of the lateral geniculo-cortical pathway in the cat and monkey. Roc. R. Soc. Lond. [Biol.] 179; Gilbert, C.D., and T.N. Wiesel(1983) Clustered intrinsic connections in cats visual cortex, J, ~ ~ 3; , ~ ~ ~ ~ ~ i Hamos, J.E., T.L. Davis, and P. Sterling (1983) Four types of neuron in layer IVab of cat cortical area 17 accululate "H-GABA. J. Comp. Neurol. 21 7: Hays, W.L., and R.L. Winkler (1971) Statistics: Probability, Inference and Decision. New York Holt Rinehart. Hendrickson, A.E., J.R. Wilson, and M.P. Ogren (1978) The neuroanatomical organization of pathways between dorsal lateral geniculate nucleus and visual cortex in old and new world primates. J. Comp. Neurol. This research was from the National Eye Institute, and VA Career Development Award and National Institute on Aging 182: ADRC grant AG David Fitzpatrick is an Alfred P. Hendrickson, A.E., S.P. Hunt, and J.Y. Wu (1981) Immunocytochemical sloan research fellow. we thank ill ~ i ~ for her ~ help- t ~ i localization ~ of glutamic acid decarboxylase in monkey striate cortex. Nature 292: ful comments' Jay for his with the Hendry, S.H.C., and E.G. Jones (1981) Sizes and distribution of intrinsic quantitative 'Ox and Tom Harper for neurons incorporating tritiated GABA in monkey sensory motor cortex. their help with the histology and Susan Havrilesky for J. Neurosci. 1: preparing the manuscript. Hendry, S.H.C., and E.G. Jones (1986) Reduction in number of immunostained GABAergic neurons in deprived-eye dominance columns of mon- LITERATURE CITED key area 17. Nature , Bear, M.F., D.E. Schmechel, and F.F. Ebner (1985) Glutamic acid decarbox- Hendry, S.H.C., C.R. Houser, E.G. Jones, and J.E. Vaughn (1983) Synaptic ylase in the striate cortex of normal and monocularly deprived kittens. organization of immunocytochemically identified GABA neurons in the J. Neurosci. 5: monkey sensory-motor cortex. J. Neurocytol. 12: Beaulieu, C., and M. Colonnier (1985) A laminar analysis of the number of Horton, J.C., and D.H. Hubel (1981) Regular patchy distribution of cytoround-asymmetrical and flat-symmetrical synapses on spines, dendritic chrome-oxidase staining in primary visual cortex of macaque monkey. trunks, and cell bodies in area 17 of the cat. J. Comp. Neurol. 231:180- Nature Houser, C.R., S.H.C. Hendry, E.G. Jones, and J.E. Vaughn (1983) Morpho- Blasdel, G.G., and D. Fitzpatrick (1984) Physiological organization of layer logical diversity of immunocytochemically identified GABA neurons in 4 in macaque striate cortex. 3. Neurosci. 4: the monkey sensory-motor cortex. J. Neurocytol. 12: Blasdel, G.G., and J.S. Lund 11982) Termination of afferent axons in ma- Houser, C.R., J.E. Vaughn, S.H.C. Hendry, E.G. Jones, and A. Peters (1984) caque striate cortex. J. Neurosci. 3: GABA neurons in the cerebral cortex. In A. Peters and E.G. Jones (eds): Blasdel, G.G., D. Fitzpatrick, and J.S. Lund (1985) Intrinsic connections of Cerebral Cortex Vol. 2: Functional Properties of Cortical Cells. New macaque striate cortex: Axonal projections of cells outside lamina 4C. J. York: Plenum, pp Neurosci. 5: Hubel, D.H., and T.N. Wiesel (1972) Laminar and columnar distribution of Carroll, E.W., and M. Wong-Riley (1985) Correlation between cytochrome geniculo-cortical fibers in the macaque monkey. J. Comp. Neurol. oxidase staining and the uptake and laminar distribution of triated 146: aspartate, glutamate, y-aminobutyrate and glycine in the striate cortex Jones E.G. (1975) Varieties and distribution of non-pyramidal cells in the of the squirrel monkey. Neuroscience 25: somatic sensory cortex of the squirrel monkey. J. Comp. Neurol. 260:205- Chronwall, B.M., and J.R. Wolff (1978) Classification and location of neu rons taking up 13HJ-GABA in the visual cortex of rats. In F. Fonnum Jones, E.G., S.H.C. Hendry, J. Yan, and H.D. Schwark (1986) Consistency (ed): Amino Acids as Chemical Transmitters. New York: Plenum, pp. in numbers of GABA cells across areas of monkey cerebral cortex. SOC Neurosci. Abstr. I2t350. Creutzfied, D.D., and M. Ito (1968) Functional synaptic organization of Kageyama, G.H., and M. Wong-Riley (1986) Laminar and cellular localizaprimary visual cortex neurons. Exp. Brain Res. 6: tion of cytochrome oxidase in cat striate cortex. J. Comp. Neurol. 245: DeFelipe, J., S.H.C. Hendry, and E.G. Jones (1986) A correlative electron microscopic study of basket cells and large GABAergic neurons in the Kisvarday, Z.F., K.A.C. Martin, D. Whitteridge, and P. Somogyi (1985) monkey sensory-motor cortex. Neuroscience 17: Synaptic connections of intracellularly filled clutch cells, a type of small

19 GABAERGIC NEURONS IN STRIATE CORTEX 91 basket cell in the visual cortex of the cat. J. Comp. Neurol. 241t Kisvarday, Z.F., A. Cowey, and P. Somogyi (1986) Synaptic relationships of a type of GABA-immunoreactive neuron (clutch cell), spiny stellate cells and lateral geniculate nucleus afferents in layer IVC of the monkey striate cortex. Neuroscience 19t Lauder, J.M., V.K.M. Han, P. Henderson, T. Verdoorn, and A.C. Towle(1986) Prenatal ontogeny of the GABAergic system in the rat brain: An immunocytochemical study. Neuroscience 19: LeVay, S. (1973) Synaptic patterns in the visual cortex of the cat and monkey. Electron microscopy of Golgi preparations. J. Comp. Neurol. 150: LeVay, S. (1986) Synaptic organization of claustral and geniculate afferents to visual cortex of the cat. J. Neurosci. 6: LeVay, S., and C.D. Gilbert (1976) Laminar patterns of geniculo-cortical projections in the cat. Brain Res. 113tl-19. Lin, C.-S., S.M. Liu, and D.E. Schmechel(l985) Glutamic acid decarboxylase immunoreactivity in layer IV of barrel cortex of rat and mouse. J. Neurosci Liposits, Z., G. Setalo, and B. Flerko (1984) Application of the silver-gold intensified 3,3'-diaminobenzidine chromogen to the light and electron microscopic detection of the luteinizing hormone-releasing hormone system of the rat brain. Neuroscience 13: Livingstone, M.S., and D.H. Hubel (1982) Thalamic inputs to cytochrome oxidase-rich regions in monkey visual cortex. Proc. Natl. Acad. Sci. U.S.A. 79: Livingstone, M.S., and D.H. Hubel (1984) Anatomy and physiology of a color system in the primate visual cortex. J. Neurosci. 4t Lund, J.S. (1973) Organization of neurons in the visual cortex, area 17, of the monkey (Macaca mulata). J. Comp. Neurol. 147r Lund, J.S. (1987) Local circuit neurons of macaque striate cortex. I: Neurons of laminae 4C and 5A. J. Comp. Neurol. 257t Lund, J.S., R.G. Boothe, and R.D. Lund (1977) Development of neurons in visual cortex of the monkey ( M~c~cR nemestrina). A Golgi study from fetal day 127 to postnatal maturity. J. Comp. Neurol. 176: Marin-Padilla, M. (1969) Origin of the pericellular baskets of the pyramidal cells of the human motor cortex: A Golgi Study. Brain Res. 14: Mates, S.L., and J.S. Lund (1983) Neuronal composition and development in lamina 4C of monkey striate cortex. J. Comp. Neurol. 221t McLean, I.W., and P.K. Nakane (1974) Periodate-lysine-paraformaldehyde fixative: A new fixative for immunoelectron microscopy. J. Histochem. Cytochem. 22r1077. Montero, V.M. (1986) Interneuronal nature of GABAergic neurons in the lateral geniculate nucleus of the rat and rhesus monkey: A combined HRP and GABA immunocytochemical study. Soc. Neurosci. Abstr. 12:lO. Montero, V.M., and J. Zempel (1985) Evidence for two types of GABAcontaining interneurons in the A laminae of the cat lateral geniculate nucleus: A double-label HRP and GABA immunocytochemical study. Exp. Brain Res Mugnaini, E., and A.L. Dahl (1983) Zinc-aldehyde fixation for light microscopic immunocytochemistry of nervous tissues. J. Histochem. Cytochem. 31t Oertel, W.H., D.E. Schmecbel, M.L. Tappaz, and I.J. Kopin (1981a) Production of a specific antiserum to rat brain glutamic acid decarboxylase by injection of an antigen-antibody complex. Neuroscience 6t Oertel, W.H., D.E.'Schmechel, E. Mugnaini, M.L. Tappaz, and I.J. Kopin (1981b) Immunocytochemical localization of glutamate decarboxylase in rat cerebellum with a new antiserum. Neuroscience 6: Ottersen, O.P., and J. Storm-Mathisen (1984) Glutamate- and GABA-containing neurons in the mouse and rat brain as demonstrated with a new immunocytochemical technique. J. Comp. Neurol. 229: Patel, H.H., and A.M. Sillito (1978) Inhibition and velocity tuning in the cat visual cortex. J. Physiol. (Lond.) 284:113P-l14P. Peters, A,, and A. Fairen (1978) Smooth and sparsely-spined stellate cells in the visual cortex of the rat: A study using a combined Golgi-electron microscopic technique. J. Comp. Neurol. 181t Ramon y Cajal, S. (1899) Estudios sobre la corteza cerebral humana. Corteza Visual. Revue Trimest. Microgr. 4t1-63. Ribak, C.E. (1978) Aspinous and sparsely spinous stellate neurons in the visual cortex of rats contain glutamic-acid decarboxylase. J. Neurocytol. 7: Rockland, R.S. (1985) Intrinsically projecting pyramidal neurons of monkey striate cortex: An EM-HRP study. Soc. Neurosci. Abstr. IIt17. Rockland, K.S., and D.N. Pandya (1979) Laminar origins and terminations of cortical connections of the occipital lohe in the rhesus monkey. Brain Res. 179t3-20. Rockland, K.S., and J.S. Lund (1983) Intrinsic laminar lattice connections in primate visual cortex. J. Comp. Neurol. 216t Schmechel, D.E., B.G. Vickrey, D. Fitzpatrick, and R.P. Elde (1984) GA- BAergic neurons of mammalian cerebral cortex: Widespread subclass defined by somatostatin content. Neurosci. Lett. 47t Sillito, A.M. (1975) The contribution of inhibitory mechanisms to the receptive field properties of neurons in the striate cortex of the cat. J. Physiol. (Lond.) 250t Sillito, A.M. (1977a) Inhibitory processes underlying the directional specificity of simple, complex and hypercomplex cells in the cat's visual cortex. J. Physiol. (Lond.) 271t Sillito, A.M. (1977b) The spatial extent of excitatory and inhibitory zones in the receptive field of superficial layer hypercomplex cells. J. Physiol. (Lond.) 273t Sillito, A.M. (1979) Inhibitory mechanisms influencing complex cell orientation selectivity and their modification at high resting discharge levels. J. Physiol. (Lond.) 289t Sillito, A.M., and V. Versiani (1977) The contribution of excitatory and inhibitory inputs to the length preference of hypercomplex cells in layers I1 and 111 of the cat's striate cortex. J. Physiol. (Lond.) 273.5' Somogyi, P., and A. Cowey (1981) Combined Golgi and electron microscopic study on the synapses formed by double bouquet cells in the visual cortex of the cat and monkey. 3. Comp. Neurol Somogyi, P., and A.J. Hodgson (1985) Antisera to y-amino butyric acid. 111: Demonstration of GABA in Golgi impregnated neurons and in conventional electron microscopic sections of cat striate cortex. J. Histochem. Cytochem. 33t Somogyi, P., A. Cowey, N. Halasz, and T.F. Freund (1981) Vertical organization of neurons accumulating 3H-GABA in visual cortex of rhesus monkey. Nature 294t Somogyi, P., A. Cowey, Z. Kisvarday, T.F. Freund, and J. Szentagothai (1983a) Retrograde transport of y-amin~[~]butyric acid reveals specific interlaminar connections in the striate cortex of monkey. Proc. Natl. Acad. Sci. U.S.A. 80: Somogyi, P., T.F. Freund, J.Y. Wu, and A.D. Smith (1983b) The section-golgi impregnation procedure. 2. Immunocytochemical demonstraton of glutamic acid decarboxylase in Golgi impregnated neurons and in their afferent synaptic boutons in the visual cortex of the cat. Neuroscience 9t Somogyi, P., Z.F. Kisvarday, K.A.C. Martin, and D. Whitteridge (1983~) Synaptic connections of morphologically identified and physiologically characterized large basket cells of the striate cortex of cat. Neuroscience 10t Somogyi, P., A.J. Hodgson, A.D. Smith, M.G. Nunzi, A. Gorio, and J. -Y. Wu (1984a) Different populations of GABAergic neurons in the visual cortex and hippocampus of cat contain somatostain- or cholecystokinin-immunoreactive material. J. Neurosci. 4: Somogyi, P., Z.F. Kisvarday, T.F. Freund, and A. Cowey (1984b) Characterization by Golgi impregnation of neurons that accumulate 3H-GABA in visual cortex of monkey. Exp. Brain Res. 53t Szentagothai, J. (1973) Synaptology of the visual cortex. In R. Jung (ed): Handbook of Sensory Physiology, Vol. IU3: Central Visual Information, Part B. Berlin: Springer-Verlag, pp Tigges, J., M. Tigges, S. Anschel, N.A. Cross, W.D. Ledbetter, and R.L. McBride (1981) Areal and laminar distribution of neurons interconnecting the ventral visual cortical areas 17, 18, 19, and MT in squirrel monkey (Sairniri). J. Comp. Neurol Tombol, T. (1978) Some Golgi data on visual cortex of the rhesus monkey. Acta. Morphol. Acad. Sci. (Hung.) 26tll Ts'o, D.Y., C.D. Gilbert, and T.N. Wiesel (1986) Relationships between horizontal interactions and functional architecture in cat striate cortex as revealed by cross-correlation analysis. J. Neurosci. 6: Tsumoto, T., W. Eckart, and O.D. Creutzfeldt (1979) Modification of orientation sensitivity of cat visual cortex neurons by removal of GABAmediated inhibition. Exp. Brain Res. 34t Uchizono, D. (1965) Characteristics of excitatory and inhibitory synapses in the central nervous system of the cat. Nature 207t Valverde, F. (1971) Short axon neuronal subsystems in the visual cortex of the monkey. Int. J. Neurosci. It Wong-Riley, M.T.T. (1979) Changes in the visual system of monocularly sutured or enucleated cats demonstrable with cytochrome oxidase histochemisty. Brain Res. 171:ll-28.