Morphological taxonomy of the neurons of the primate striatum.

THE JOURNAL OF COMPARATIVE NEUROLOGY 313273494 (1991)

Morphological Taxonomy of the Neurons of the Primate Striatum JEROME YELNIK, CHANTAL FRANCOIS, GERARD PERCHERON, AND DOMINIQUE TANDE Laboratoire de Neuromorphologie informationnelle et de Neurologie experimentale du mouvement, INSERM U106, HBpital de la SalpgtriBre, 75013 Paris, France

ABSTRACT

A quantitative taxonomy of primate striatal neurons was elaborated on the basis of the morphology of Golgi-impregnated neurons. Dendritic arborizations were reconstructed from serial sections and digitized in three dimensions by means of a video computer system. Topological, metrical, and geometrical parameters were measured for each neuron. Groups of neurons were isolated by using uni- and multidimensional statistical tests. A neuronal species was defined as a group of neurons characterized quantitatively by a series of nonredundant parameters, differing statistically from other groups, and appearing as a separate cluster in principal component analysis. Four neuronal species were isolated: (1) the spiny neuronal species (96%of striatal neurons) characterized by spine-free proximal dendrites (up to 31 p,m) and spine-laden distal dendrites, which are more numerous, shorter, and less spiny in the human than in the monkey, (2) the leptodendritic neuronal species (2%) characterized by a small number of long, thick, smooth, and sparsely ramified dendrites, (3) the spidery neuronal species (1%) characterized by very thick dendritic stems and a large number of varicose recurrent distal processes, and (4)the microneuronal species (1%)characterized by numerous short, thin, and beaded axonlike processes. All striatal neurons give off a local axonal arborization. The size and shape of cell bodies were analyzed quantitatively in Golgi material and in materials treated for Nissl-staining, immunohistochemical demonstration of parvalbumin and histochemical demonstration of acetylcholinesterase. Only three types were distinguishable: small, round cell bodies corresponding to either spiny neurons or microneurons, medium-size elongated cell bodies, which were parvalbumin-immunoreactive and corresponded to leptodendritic neurons, and large round cell bodies, which were acetylcholinesterase-positive and corresponded to spidery neurons. Thorough analysis of previously elaborated classifications revealed that spidery neurons do not exist in rats and cats and that large cholinergic neurons in these species correspond to leptodendritic neurons. From this, it can be assumed that the dendritic domain of striatal cholinergic neurons is considerably smaller in primates than in other species. Computer simulations based on both the frequency of each neuronal species and their three-dimensionaldendritic morphology revealed that the striatum consists of two intertwined dendritic lattices: a fine-grain lattice (300-600 pm) formed by the dendritic arborizations of spiny, spidery, and microneurons, and a large-grain lattice (1,200 pm) formed by the dendritic arborizations of leptodendritic neurons. This suggests that cortical information can be processed in the striatum through two different systems: a fine-grain system that would conserve the precision of the cortical input, and a large-grain system that would blur it. Key words: dendrites, quantitative morphology, three-dimensionalanalysis, neuronal taxonomy

The striatum is made up of different types of neurons which are present in highly different proportions and contain several different neuroactive substances. In the classification schemes previously published (e.g., Kolliker, 1896; Dejerine, '01; Ramon y Cajal, '11; Leontovich, '54; Fox et al.7 '71ajb; Kemp and Powell, '71; DiFiglia et al*,'76; Braak and Braak, '82; Chang et al., '82; Graveland et al.,

o 1991 WILEY-LISS, INC.

'85), there is general agreement that the vast majority of striatal neurons (96%;Kemp and Powell, '71) belong to a Accepted July 11,1991 Address reprlnt requests to Jerijme Yelnk, INSERM U106, Pavlllon Claude Bernard, HBpltalde la SalpStrlere, 47, Bd de I'hBpital, F-75651 Pans Cedex 13, France.

274 single population of medium-size spiny neurons. Although they were first considered as local circuit neurons (Dbjerine, '01; Ram6n y Cajal, '11; Fox et al., '71a; Kemp and Powell, '71), it is now well established that they project to the globus pallidus and substantia nigra. They are also known to contain gamma-aminobutyric acid (Pasik et al., '88) and substance P or enkephalin (Izzo et al., '87) and to be identifiable by their calcium-binding, protein calbindinD28k content (DiFiglia et al., '89). The existence of large cholinergic neurons is also well documented, but they have been correlated with several different morphological types (Armstrong et al., '83; Eckenstein and Sofroniew, '83; Levey et al., '83; Bolam et al., '84a,b; Phelps et al., '85; DiFiglia and Carey, '86; DiFiglia, '87). The morphological classification of the remaining neurons is even less clear. Striatal neurons have been classified into either two or three types (small, medium, and large neurons) in Nissl material (see Pasik et al., '79, for review) and into three (Fox et al., '71a,b) to eight types (Chang et al., '82) in Golgi material. This might explain why recent analyses of the basal ganglia circuitry have tended to focus on medium-size spiny neurons, whereas interneurons have been rather neglected (e.g., Goldman-Rakic and Selemon, '90). The goal of our study was thus to clarify the taxonomy of striatal neurons and to evaluate their participation in the internal circuitry of the striatum. The first stage in this analysis consisted in classifying the neurons of the primate striatum into a quantitative taxonomic system satisfying a series of theoretical and methodological rules (Tyner, '75; Rowe and Stone, '77; Yelnik et al., '87; Rowell, '89). We used a quantitative approach (Yelnik et al., '81, '83) by which dendritic arborizations can be studied in three dimensions and after reconstruction from serial sections. Our classification comprises four neuronal species that we have already presented in abstract form in Yelnik et al. ('89). In this study they have been confronted with the neuronal types previously isolated. This has revealed important differences between primates and nonprimates. The second step in the analysis consisted of determining the distribution and frequency of each of the four neuronal species. This could not be done in Golgi material since this method only reveals a small proportion of neurons and moreover preferential impregnation of certain types of neurons can occur (Fox et al., '71a; Pasik et al., '79). For this reason, we made a quantitative analysis of the size and shape of cell bodies. This was first done in Golgi material to determine whether or not neurons previously classified on the basis of their dendritic morphology could be identified on the basis of the morphology of their cell body alone. Nissl preparations, which do not reveal dendritic morphology but in which all cell bodies are stained, were then analyzed comparatively. Histochemical demonstration of acetylcholinesterase and immunohistochemical demonstration of the calcium-binding proteins calbindin-D28k and parvalbumin were also investigated because they are known to mark different types of striatal neurons (DiFiglia, '87; DiFiglia et al., '89; Celio, '90; Cowan et al., '90). Nissl staining, acetylcholinesterase, and parvalbumin were able to be correlated with the neuronal species defined in Golgi material and were thus used to determine their distribution and frequencies. The third part of the study was an analysis of the dendritic lattice formed by each neuronal species of the

J. YELNIK ET AL. striatum. We made use of a computer simulation similar to the one we used to study the dopaminergic neurons of the retina (Savy et al., '89) and in which both the threedimensional morphology of dendritic arborizations and the frequency of each neuronal species are taken into account. Several architectural features of the internal circuitry of the striatum were thus determined, the functional implications of which were analyzed.

MATERIALS AND METHODS Golgi material The main part of this study was carried out on Golgiimpregnated preparations of our collection, which comprises 43 brains of adult primates: 21 human brains, 18 Macaca irus, and 4 Papio papio. Human brains were obtained at autopsy from patients who did not have any neurological disease. The deaths were related to cardiopulmonary failure. Brains were removed within 48 hours and left in a 10%formalin solution for 2 to 8 months. Monkeys, anesthetized with Fluothane, protoxide, and oxygen by using an intratracheal tube connected to a Fluotec respirator (Cyprane, England), were perfused transcardially with a formalin solution. A detailed description of histological procedures has been published in previous reports (Yelnik et al., '84, '87). Two brains ofPapiopapio (P1 and P3), two brains of Macaca irus (I9 and 1131, and two human brains (H14 and H18) in which the striatum was particularly well impregnated were selected for this study. As a whole, they contained several thousand neurons which were analyzed qualitatively on a Leitz Orthoplan microscope. Neurons exhibiting typical characteristics were photographed on a Leitz Vario-orthomat.

Nissl, histochemical, and immunohistochemical material These techniques were used to study the morphology and distribution of cell bodies in the striatum. Monkeys were perfused transcardially under Fluothane, protoxide and oxygen anesthesia, with 0.5 liter of a 0.1 M phosphate buffer (pH = 7.2) solution followed by 5 liters of a paraformaldehyde-glutaraldehyde solution mixed in the same buffer. All brains were cut serially into 50-k.m thick sections oriented with reference to the standard ventricular frontal plane. A Macaca mulatta brain (MM16) was embedded in celloi'din and stained by the standard Nissl method. The other brains were cut on a freezing microtome. A Macaca irus brain (MI15) was processed according to the DFPhistochemical technique (Poirier et al., '77) to reveal the distribution of neurons rich in acetylcholinesterase (AChE). The brain of another Macaca irus (MI40) was used for immunohistochemical detection of parvalbumin (PV). Sections were incubated for 48 hours at 4°C with a mouse monoclonal anti-parvalbumin antibody (BioMakor) diluted 15,000. After thorough washing, antibodies were localized with the avidin-biotin complex method (Hsu et al., '81). They were visualized by incubation with a mixture of 3'3-diaminobenzidine tetrahydrochloride (0.05%), cobalt chloride (I%), and hydrogen peroxide (1%).All sections were counterstained with neutral red or cresyl violet.

Quantitative analysis of striatal neurons Golgi-impregnated neurons of the striatum were first drawn using the camera lucida and the x40 or the xl00 oil

TAXONOMY OF PRIMATE STRIATAL NEURONS

275

immersion objective. Each dendritic arborization (defined neuronal species was analyzed by comparing the cell bodies as the set of the dendritic trees of a given neuron) was of Golgi-impregnated neurons previously identified on the reconstructed from serial sections (Yelnik et al., ’81). A basis of their dendritic morphology with cell bodies obpreliminary classification into qualitatively defined neu- served in Nissl, histochemical, and immunohistochemical ronal types was elaborated, and for each type, a sample of materials. The largest contour of each cell body was drawn complete neurons was constituted. Neurons which had a with the aid of the camera lucida (x100 oil immersion part of their arborization hidden by silver precipitates, glial objective) and digitized on a Summagraphics XY digitizer. processes or blood vessels, and neurons whose dendrites Its area (ss) was computed in the formula of Pullen (’84) ended abruptly instead of tapering progressively were and its biggest and smallest diameters were determined. A discarded. shape-index (sh) was obtained by dividing the biggest Each selected neuron was analyzed quantitatively on a diameter by the smallest one. Tissue shrinkage, calculated computer-aided system (Yelnik et al., ’81) with which a for each brain by comparing its in vivo to its histological series of 18 parameters is measured (see Yelnik et al., ’84, dimensions, was 31.3% for the Nissl-stained brain MM16 ’87 for details). Topological parameters (Percheron, ’79a,b, and 15%for the other brains. Cell body sizes were corrected ’82) comprised the number of stems (S), tips (F) and accordingly. The number of cell bodies per mm3 was segments (U = 2F-S), the branching index (F/S), and the measured in several samples of striatal tissue and corrected maximal stature (Hm). Metrical parameters comprised the for tissue shrinkage. The correction factor of Abercrombie mean length of all segments (Ln), stems (Ls), internodes (’46) was also used. Mean values were defined as the (Li), and twigs (Lp), the total dendritic length (L), and the frequency of each neuronal species. highest dendritic length (Lm). The thickness of dendritic segments was measured by using calibrated circles superLattice analysis posed on each dendritic segment under the camera lucida ( ~ 1 0 0oil immersion objective). The mean thickness of The dendritic lattice formed by the whole set of interdendritic stems (dS) was calculated for each neuron. Pro- twined dendritic arborizations was analyzed quantitatively cesses which were thinner than 0.5 pm and bore swellings by considering both the dendritic morphology and the were defined as thin axonlike processes (FranGoiset al., ’84; frequency of each neuronal species. The total number of Yelnik et al., ’87). The density of dendritic spines was neurons was obtained by multiplying the frequency of each determined by dividing the number of visible spines by the neuronal species by the total volume of the striatum. The three-dimensional length of individual dendritic segments. latter was measured on the macaque brain MM16 with the Geometrical parameters, determined from the principal aid of a computer-aided three-dimensional cartographic axes of each arborization (Yelnik et al., ’831,comprised the method (Percheron et al., ’86). Corresponding data for length (11, width (w), and thickness (t), and the indices of human were taken from Schroder et al. (’75). The total axialization (a)and flatness (p). length of dendrites was obtained for each neuronal species by multiplying the total dendritic length (L) of individual Taxonomy of striatal neurons neurons by the total number of neurons. The dendritic Each group of neurons was characterized by its mean density was estimated by dividing the total length of (m), median (Md), standard deviation (sd), coefficient of dendrites by the total volume of the striatum. The “grain” variation (V = sd/m), and index of skewness (K = 3 x of the lattice was expressed as the length of dendritic (m-Mdisd)) (Blalock, ’72). Correlation between pairs of arborizations. Dendritic overlap was expressed as the numparameters were computed using both the parametric ber of cell bodies present in the space delimited by a single coefficient of correlation and the nonparametric rank-order arborization (length x width x thickness). A computer coefficient of Spearman. Comparisons between groups were graphic simulation of the dendritic lattice was obtained by computed using the Student’s difference-of-means test and using the neurons already analyzed and stored in computer the Mann-Whitney nonparametric test. The whole set of data files. A rectangular volume (400 x 400 x 10 pm) was quantitative data was also submitted to a principal compo- defined as a sample of striatal tissue. The number of nent analysis or PCA (Cooley and Lohnes, ’71). In PCA, a neurons present in this volume was determined from the neuron is considered as an individual localized in a multidi- frequency of each neuronal species. Neurons were distribmensional space defined by the 18 morphological parameuted randomly in this volume. Neurons located in the ters studied. Since projection on the principal plane show vicinity of the rectangle were also analyzed. Only the the individuals in their greatest spread, neurons which portions of dendrites present in the rectangle were drawn. were grouped together in a well-delimited cluster could be considered as similar neurons, whereas neurons located in separate clusters could be assumed to come from different RESULTS neuronal populations. A “neuronal species” was defined as a group of neurons characterized quantitatively by a series In Golgi material we identified qualitatively four types of of nonredundant parameters, differing statistically from neurons that we referred to as spiny neurons, leptodenother groups, and appearing as a separate cluster in PCA. dritic neurons, spidery neurons, and microneurons (Figs. 1, Striatal neurons were compared statistically with pallidal 2). In a series of 12 successive serial sections, which and nigral neurons previously analyzed quantitatively (Yel- contained 2,564 impregnated neurons, there were 2,460 nik et al., ’84, ’87). spiny neurons (96%), 41 spidery neurons (1.6%),30 leptodendritic neurons (1.2%),and 25 microneurons (1%).Only Identification, distribution, and frequency of 8 neurons (0.3%) could not be identified because their striatal neuronal species dendrites were not impregnated beyond the proximal stems. As silver impregnation does not reveal all the neurons of A neuron of each type, reconstructed from serial sections, is a given piece of striatal tissue, the distribution of each illustrated at the same magnification in Figure 3.

276

Fig. 1. Golgi preparations of the striatum of a monkey (Pupiopapzo) photographed using the x 40 oil objective. Scale bar: 50 pm. Same magnification for all four pictures. A. Spiny neuron characterized by a high density of dendritic spines. Spines are typically absent up to 31 pm from the cell body. The cell body is elongated. An axon hillock is visible. B. Leptodendritic 'neuron characterized by thick, sparsely branched,

J. YELNIK ET AL.

and smooth dendrites. The cell body is fusiform. An axon hillock is visible. C. Spidery neuron characterized by thick dendritic stems, highly branched and varicose distal processes, and a voluminous, globular cell body. I). Microneuron characterized by very thin, beaded processes and a small round cell body.


277

Fig. 2. Golgi preparations of the striatum of a monkey (Papiopupio) photographed using the x 10 objective. Scale bar: 200 pm. Same magnification for all three pictures. A. Leptodendritic neuron characterized by sparsely branched, long, thick, smooth dendrites and a large fusiform cell body. Spiny neurons present in the same field have numerous, short and thin dendrites and small, round cell bodies. Spines

are not discernible at this magnification. B. Spidery neuron characterized by numerous, thick varicose dendrites which curve back toward the soma and a voluminous, round cell body. Note the strikingly different aspect of the spiny neuron located above. C. Microneuron characterized by short very thin processes and a small, round cell body. At the right is a spiny neuron.

Quantitative analysis of striatal neurons

human, but only 5-34 in the monkey (p < 0.001). The number of segments (U) was also higher (78 versus 63); the branching index (FIS) and the maximal stature (Hm) were the same in the two species. Among the metrical parameters, the mean length of dendritic segments (Ln) was shorter in the human (75 p m ) than in the monkey (94 pm), whereas the total dendritic length (L = 5,900 pm) and the highest dendritic length (Lm = 275 pm) were the same in both species. To sum up, dendritic segments of spiny neurons were more numerous but shorter in the human than in the monkey. The mean thickness of dendritic stems was smaller in the human than in the monkey (Table 1).The threedimensional geometry of dendritic arborizations was remarkably similar in the two species. The mean geometrical module was an ellipsoid with dimensions of 430 x 330 x 230 pm (see Fig. 71, but spherical and discoidal shapes were also observed. The location of the first spine was not related

Spiny neurons. Dendrites, except on their proximal portion, were typically covered with spines (Fig. 1A). They extended radially and often overlapped with those of neighboring neurons (Fig. 2A). Cell bodies were generally round (Fig. 4A), but they could also be elongated (Figs. 1A and 4B). In most cases they did not have any spines. Their cross-sectional area was smaller in human than in monkey (Table 1).The dendritic arborizations of 28 spiny neurons were reconstructed under the xl00 oil immersion objective. Two to four serial sections were necessary to reconstruct the entire arborization of each neuron. For none of the tested morphological parameters was there a significant difference between the spiny neurons of the putamen (n = 24) and the spiny neurons of the caudate nucleus (n = 41. Conversely, there were significant differences between monkey (n = 20) and human (n = 8 ) spiny neurons. The dendritic formula (S - F) was 6-42 in the

278

Fig. 3. Camera lucida drawings of the four neuronal species of the primate striatum: spiny neuron (A), leptodendritic neuron (B), spidery neuron (C), and microneuron (D). Golgi method. All arborizations, reconstructed from serial sections, are illustrated at the same magnifi-

J. YELNIK ET AL.

cation. Note that the leptodendritic neuron has extremely long dendrites with several thin axonlike distal processes and that its cell-body size is intermediate between that of spiny neuron and that of spidery neuron.


Fig. 4. Camera lucida drawings of spiny striatal neurons. Golgi method. Original drawings were made using the x 100 oil immersion objective from two 240-pm-thick serial sections of the baboon (A) and four 120-~m-thicksections of the human (B). Note the spine-free

279

proximal dendrites and the spine density, which is higher in baboons than in humans. For the purpose of comparison, camera lucida drawings of Figures 4-6 are illustrated at the same magnification, which is about two times higher than that of Figure 3.

J. YELNIK ET AL.

280 TABLE 1. Moruholoeical Characteristics of Striatal Neurons of Primates' m SPINY n = 20

h SPINY n = 8

LEPTO

INT

n

n=4

=

18

SPIDERY n=2

MICRO n=2

Soma SS

dS Topological S

F U FIS Hm Metrical L Lm Ln LS Li LP Geometrical 1 W

t a P

219 10.13) 3.3 (0.18)

170 (0.29) 2.2 (0.17)

641 (0.28) 4.5 (0.231

5 11 (0.27) 3.6 (0.25)

1,164 4.1

182 1.4

5.2 (0.19) (0.13) (0.13) (0.22) (0.17)

6.5 (0.15) 42.4 (0.08) 78.4 (0.08) 6.6 (0.12) 7.0 (0.12)

4.1 (0.24) 20.1 (0.25) 36.1 (0.26) 5.0 (0.24) 5.8 (0.15)

4.5 (0.11) 48.0 (0.13) 91.5 (0.14) 10.8 (0.19) 7.5 (0.07)

12.0 ?8.5 245.0 10.9 10.0

6.5 64 122 10.9 7.5

5,943 (0.11) 275 (0.12) 93 (0.09) 11(0.79) 17 (0.20) 160 (0.09)

5,905 (0.20) 275 (0.11) 75 (0.15) 9 (0.18) 19 (0.21) 124 (0.18)

6,879 (0.29) 900 (0.17) 196 (0.27) 87 (0.94) 155 (0.43) 246 (0.31)

10,490 (0.13) 725 (0.05) 118 (0.23) 38 (0.40) 83 (0.23) 154 (0.30)

427 (0.09) 337 (0.14) 246 (0.22) 0.26 (0.41) 0.51 (0.25)

446 (0.12) 329 (0.19) 233 (0.15) 0.30 (0.33) 0.50 (0.14)

1,187 (0.21) 829 (0.25) 531 (0.27) 0.34 (0.43) 0.60 (0.19)

977 (0.13) 653 (0.16) 440 (0.31) 0.37 (0.48) 0.64 (0.15)

34.6 64.0 7.0 6.6

23,391 543 95 8 48 142 613 540 396 0.19 0.42

4,686 159 39 12 17 58 240 211 162 0.24 0.44

'Each morphological parameter is expressed as the mean value and coefficient of variation (in brackets) for monkey imSPINY) and human (h SPINY) spiny neurons and for leptodendritic neurons (LEPTO), intermediate neurons (INT), spidery neurons (SPIDERY), and microneurons (MICRO) of monkey and human.

to any topological ordering of dendritic segments but to a distance from the soma which was the same in both human (m = 30.3 pm; sd = 5.6; n = 16) and monkey (m = 31.1 pm; sd = 3.5; n = 27). The number of spines per 10 ym length, as measured on the neurons which exhibited the highest spine density, was significantly lower in the human (m = 2.6; sd = 0.8; n = 11)than in the monkey (m = 8.0; sd = 1.8; n = 10). Neurons which had a lower density of spines were also analyzed quantitatively. They were topologically, metrically, and geometrically indistinguishable from typical spiny neurons. Some of them had a few spinelike appendages on their proximal stems or, more rarely, on their cell body. All spiny neurons in our material had an axon-hillock which stemmed generally from the cell body (Fig. 1A). When it was impregnated, the axon gave rise to a densely ramified local arborization (see Fig. 8B), which largely overlapped the dendritic arborization. The number of axonal segments varied from 37 to 73 (n = 3 neurons) with mean lengths from 58 to 103 pm. They were very thin but relatively straight. The overall dimensions of the axonal arborization were 540 x 480 x 420 pm. Leptodendritic neurons. They had few but very thick dendritic stems that emerged from a fusiform or triangular cell body (Fig. 1B). The cross-sectional area of cell bodies was significantly bigger than that of spiny neurons (Table 1). Dendritic segments tapered progressively and were characteristically stiff and thick. They were typically devoid of spines, although occasional spines or spinelike appendages could be observed. A few dendritic twigs were less than 0.5 p m thick and bore swellings (Fig. 3), thus corresponding to the thin axonlike processes as defined in Methods. Dendrites of leptodendritic neurons clearly differed from those of spiny neurons in that they were smooth (compare A and B in Fig. 11, thicker, less ramified, and far longer (Fig. 2A). The dendritic arborizations of 18 leptodendritic neurons, which spread over 7-12 serial sections, were reconstructed using the x40 oil immersion objective. After reconstruction, they proved to be distinctly larger than the dendritic arborizations of all other striatal neurons. This can be seen in Figure 3 in which a leptodendritic neuron, a

spiny neuron, a spidery neuron, and a microneuron are illustrated at the same magnification. Leptodendritic neurons were characterized by sparsely branched dendritic arborizations. Their dendritic formula (4-20) was the smallest of all striatal neurons (Table 1). Conversely, their dendritic segments were the longest (Ln = 200 pm). Their highest dendritic length (Lm) and the length of their dendritic arborization (1) reached 1.3 and 1.6 mm, respectively. Statistical tests demonstrated that there was no difference between monkey and human leptodendritic neurons (p < 0.05). All quantitative parameters, except the total dendritic length (L in Table l ) , were significantly different for leptodendritic and spiny neurons (p < 0.001). The dendritic arborizations of leptodendritic neurons were either ellipsoidal (see Fig. 7), spherical, discoidal, or cylindrical with varying dimensions (1 = 7301,450 ym, w = 590-1,400 Fm, t = 280-800 pm). All the leptodendritic neurons of our series had an axon-hillock which originated either from the cell body (Fig. 1B) or a dendritic stem. In comparison with the axon of spiny neurons (see Fig. 8A), the axonal branches of leptodendritic neurons (see Fig. 8B) were less numerous (19-23; n = 2 neurons) but thicker and longer (146 pm). The o v e r a l l d i m e n s i o n s of t h e l o c a l a r b o r i z a t i o n (680 x 350 x 270 pm) were smaller than those of the dendritic arborization (Table 1). Spidery neurons. Spidery neurons had extremely thick stems (up to 10 pm thick) the mean thickness of which was similar to that of leptodendritic neurons (Table 1).They were profusely ramified with successive branches becoming significantly thinner a t each bifurcation (Fig. 1C). Distal branches were very numerous and varicose. They often curved back toward the soma thus forming a very dense arborization (Fig. 5). The cell bodies of spidery neurons were globular and voluminous (Figs. lC, 3C, 51, their cross-sectional area being in fact twice the size of that of leptodendritic neurons (Table 1). Since their dendritic arborizations were highly ramified, spidery neurons had to be drawn using the xl00 oil immersion objective. Only two


A

281

PlGC18PU19

5Dum

B

Fig. 5 . Camera lucida drawings of spidery neurons of the babocn (A) and the human (B) striatum. Golgi method. Original drawings were made using the x 100 oil immersion objective from two 120-pm-thick

serial sections. Note the huge dimensions of the cell body, the thick proximal stems, and the thin varicose distal processes, which curve back toward the soma.

J. YELNIK ET AL.

282

A

f'

P1 G C 1 3 P U 5 3

1 - 1

5Opm

,

B

P3DE16CD59

,

D Fig. 6. Camera lucida drawings of microneurons of the baboon (A, B), macaque (0, and human (D) striatum. Golgi method. Original drawings were made using the x 100 oil immersion objective. Note the round cell bodies and the thinness of both dendritic stems and distal beaded axon-like processes.

of them were completely reconstructed from four serial sections. The topological parameters of spidery neurons were significantly different from those of all the other neurons of the striatum (p < 0.001). Their dendritic formula was 12-129 for the two neurons reconstructed, which was by far the highest found in the striatum (Table 1).The number of successive bifurcations (Hm in Table 1) was also the highest. The mean length of dendritic segments was the same as that of spiny neurons but the total dendritic length was far longer (23,000 vs 5,900 pm). The highest dendritic length (Lm = 540 pm) and the overall dimensions of the ellipsoidal arborization (600 x 550 x 400 pm) were only slightly larger than for spiny neurons (see Fig. 7). In most spidery neurons no axon or axon hillock was visible. However, one of the reconstructed neurons had a well-developed axonal arborization (see Fig. 8). The axon emerged from the cell body as a thick, 21-pm-long stem, which resembled a dendrite but which divided into two thin axonal branches. Each of them gave off several branches which formed a densely ramified local arborization (92

axonal segments; 600 x 300 x 300 pm). One branch, 700 pm long, exceeded the local arborization. Microneurons. Microneurons were characterized by numerous, short, thin processes which bore swellings (Figs. lD, 2C). They had round cell bodies whose cross-sectional area (Table 1) was not significantly different from that of spiny neurons (p < 0.01). Dendritic stems were thinner than those of spiny neurons (Table 1). They branched rapidly into even thinner processes, which were beaded and closely resembled axonal branches (Fig. 6). The number of these thin axonlike processes varied from one microneuron to another, but it was not possible to determine whether these differences were due to incomplete impregnation or to actual differences between neurons. The two microneurons which were reconstructed (from two serial sections) had respectively 52 and 76 twigs, whereas the three other neurons in Figure 6 had only between 32 and 37. Metrical and geometrical parameters were very low (Table l ) , especially the dimensions of arborizations (240 x 210 x 160 pm), which were only two-thirds of those of spiny neurons (Fig. 7).


A-

Spidery

Spiny

Micro

Fig. 7. Computer drawings of a leptodendritic neuron (Lepto), a spidery neuron (Spidery), a spiny neuron (Spiny), and a microneuron (Micro). Each dendritic arborization is represented within its three principal planes in perspective view. The principal axes are proportional to the length, width and thickness of the arborization. Same magnification for all four drawings.

In microneurons it was often difficult to distinguish a dendritic stem from an axon hillock and their thin and beaded processes from axonal branches (Fig. 6). In one microneuron, however, an axon giving rise to a local arborization (43 segments) was clearly visible (Fig. 8D). The dimensions of the axonal arborization (300 x 150 x 150 km) were almost equal to those of the dendritic arborization (Fig. 6A) and both arborizations overlapped considerably.

283 have been shown to belong to the same neuronal species (Yelnik et al., ’871, they were all classified as a single striato-pallido-nigral ‘‘leptodendritic neuronal species.” Spidery neurons and microneurons, which formed two distinct clusters (4 and 5 in Fig. 9), were classified as the “spidery neuronal species” and the “microneuronal species.” Unexpectedly, four neurons which had been first identified qualitatively as leptodendritic neurons appeared in a separate cluster, which lay between the leptodendritic and the spiny neurons (6 in Fig. 9). They were referred to as intermediate neurons. Two of them are illustrated in Figure 10. In comparison with leptodendritic neurons, they had more numerous (92 vs 36) but shorter (118vs 196 Fm) and thinner dendritic segments and a higher total dendritic length (10,500 vs 6,900 km). They also had more numerous dendritic spines or spinelike appendages and beaded distal processes. We looked for intermediate neurons in the whole striatum, but only five of them could be identified near the external medullary lamina of the lenticular nucleus. They were statistically indistinguishable from the intermediate neurons previously observed near the medullary laminae in the globus pallidus (FranGois et al., ’84; Yelnik et al., ’84). Intermediate neurons were therefore considered as a neuronal species belonging to the lenticular nucleus as a whole rather than as a purely striatal neuronal species. To conclude, the striatum was considered to consist of four distinct neuronal species: spiny neurons, leptodendritic neurons, spidery neurons, and microneurons.

Identification, distribution and frequency of striatal neuronal species

In Nissl-stained material (Fig. 11A,B), the striatum consisted mainly of small achromatic cell bodies among which were scattered a few larger chromatic cells. Glial cells appeared as very small dark cell bodies. In material treated for PV-immunoreactivity (Fig. 1lC,D), a small proportion of neurons was labeled. They were found throughout the striatum and often had a fusiform or triangular cell body. AChE-histochemistry (Fig. 11E,F) also labeled a small number of neurons. They were also found throughout the Taxonomy of striatal neurons striatum but had larger cell bodies, which were round. The The 54 reconstructed neurons, each characterized by 18 morphology of 225 cell bodies selected from this material quantitative parameters, were submitted to a principal and of 51 cell bodies of Golgi-impregnated neurons was component analysis. Projection of the data on the principal analyzed on correlation diagrams (Fig. 12) in which cell plane is illustrated in Figure 9. The most discriminant body size (cross-sectional area or ss) was plotted against cell parameters were the length of segments (Ln), the total body shape (biggest/smallest diameter or sh). dendritic length (L), the number of tips (F), and the surface In Golgi material (Fig. 12A),spiny neurons and microneuof cell bodies (ss). The other topological parameters (S, U, rons could not be identified on the basis of their cell bodies Hm, F/S) correlated with F, whereas the other metrical since they were both small and round. Leptodendritic parameters (Ls, Li, Lp, Lm) and the geometrical parame- neurons and spidery neurons had medium-size elongated ters (1, w, t) correlated with Ln. Shape indices (a and p) were and large round cell bodies, respectively. The same three not discriminant. types were also identified in Nissl, PV, and AChE materials, Neurons grouped together into clearly distinct clouds of although cell bodies were about half the size as those in points. The spiny neurons of monkey and human largely Golgi material. PV-immunoreactive cell bodies were meoverlapped to form a particularly dense and homogeneous dium-size and elongated, whereas nonlabeled neurons were cluster (1and 2 in Fig. 9). They were classified as the “spiny either smaller or larger (Fig. 12C).AChE-reactivity marked neuronal species.” Leptodendritic neurons formed a larger the largest cell bodies, whereas nonlabeled neurons were cloud (3 in Fig. 9), which reflected a higher morphological smaller and more elongated (Fig. 12D). Therefore, the variability (see also the coefficients of variation in Table 1). small round achromatic cell bodies of Nissl material were Comparison with previously isolated neuronal species dem- identified as spiny neurons and microneurons, PV-immunoonstrated that they were statistically indistinguishable reactive neurons as leptodendritic neurons, and AChE(p < .05) from the large pallidal neurons bearing thin positive neurons as spidery neurons. These identification processes (referred to as group 1in Yelnik et al., ’84).As the criteria were used to determine the frequency of each large neurons of the globus pallidus and substantia nigra neuronal species.

J. YELNIK ET AL.

284

A

C

L-,

D

Fig. 8. Camera lucida drawings of the local axonal arborization of a leptodendritic neuron (A), a spiny nel-r-,n (B), a spidery neuron ( C ) ,and a microneuron (D).Golgi method. Same magnification for all four d .swings.

The frequencies of spiny neurons and microneurons were estimated from the density of achromatic neurons (20,800/ mm3) and from their relative proportion in Golgi material (1/96), which gave 20,000 spiny neurons (96%) and 200

microneurons (1%) per mm3. The frequency of leptodendritic neurons, estimated from the density of PV-immunoreactive neurons, was 402/mm3 (2%). That of spidery neurons, estimated from the density of AChE-positive

285


t ;

\

;’\

‘

1 Spiny (m) 2 Spiny (h) 3 Lepto 4 Spidery 5 Micro 6 Intermediate

\ \

\

\ 3

I

1

/

P1

Fig. 9. Principal component analysis of 54 striatal neurons of human, macaques, and baboons, each reconstructed from serial sections and characterized by 18 quantitative morphological parameters. The figure is a projection of the data on the principal plane defined by the first two components P1 and P2. The most discriminant parameters, mean length of dendritic segments (Ln), cross-sectionalarea of cell bodies (ss), total dendritic length (L),and number of tips (F) are

indicated by arrows. The number of segments (U) is strongly correlated with F so that these parameters project very close to each other. See text for other abbreviations. Spiny neurons of monkey (1) and human (2),leptodendritic neurons (3), spidery neurons (4), microneurons (51, and intermediate neurons (6) segregated into clearly distinct clusters (broken lines).

neurons, was 150/mm3(1%). These percentages were close to those observed in Golgi material.

rather fine (600 km). The overlapping index (20) was 10 times smaller than for leptodendritic neurons. Microneurons were more numerous than spidery neurons: 316,000 in macaques, 1,000,000in humans. However, as they had less numerous and shorter processes, they formed a very loose lattice (Fig. 13). The dendritic density was 1.1m/mm3. The overlapping index (2) was 10 times smaller than for spidery neurons so that, characteristically, some portions of striatal tissue did not contain any microneuron processes (Fig. 13).

Lattice analysis Spiny neurons formed an extremely dense dendritic lattice (Fig. 13). Their total number was 30,400,000 in macaques and 110,000,000 in humans. As the total dendritic length of each neuron was 5,900 km, the total length of dendrites in the striatum of one hemisphere was 179 km in macaques and 649 km in humans with a dendritic density of 118 m/mm3. The grain was fine (430 km) with a high overlapping index (653). Leptodendritic neurons were 611,000 in macaques and 2,170,000 in humans. They intertwined with spiny neurons but formed a far less dense dendritic lattice (Fig. 13).The dendritic density was 2.8 m/mm3.The grain of the leptodendritic lattice (1,200 pm) was far larger than for the spiny lattice. The overlapping index (219)was three times smaller than for spiny neurons. Each spidery neuron gave rise to a densely ramified arborization (Fig. 5), but as they represented only 1%of the striatal neurons, they formed a loose lattice (Fig. 13). Their total number was 228,000 in macaques and 799,000 in humans. The dendritic density (3.5 m/mm3) did not differ greatly from that of the leptodendritic neurons. However, as dendritic arborizations were smaller, the grain was

DISCUSSION The conceptual bases of neuronal taxonomy The classification and naming of neurons form a fundamental step in any study of the nervous system. Tyner (’75) and Rowe and Stone (’77) have analyzed the conceptual bases to be respected in the process of neuronal classification. Of particular interest in the context of the present study is the opposition they make between classification and identification. Classification consists in elaborating a taxonomic system which is supposed to include any neuron of a given neuronal population and, ideally, of the whole nervous system. For doing this they recommend a “polythetic approach” based on a large number of different features instead of the “essentialist approach” by which

286

J. YELNIK ET AL.

Fig. 10. Camera lucida drawings of intermediate neurons of the human striatum. Golgi method. Original drawings were made through the x 40 oil immersion objective from three 240-pm-thick serial sections. Note that dendrites are shorter but more numerous than for the

leptodendritic neuron shown in Figure 3 at the same magnification, that they bear spines or spinelike appendages, and that the distal processes are very thin.

neurons are classified on a small number of particular features. Identification, which consists of placing a given individual into a previously elaborated taxonomic system, relies on a limited number of identification keys, which may vary with the technique used. In previous studies of striatal neurons, no clear distinction was made between classifica-

tion and identification. For example, neurons have often been classified on the basis of cell body size (Kemp and Powell, '71; Lu and Brown, '77; Chang et al., '82). This feature is an important identification criterion since it can be used with different techniques such as Nissl staining, electron microscopy, retrograde tracing techniques, and


287

Fig. 11. Sections of the striatum as revealed using the Nissl technique (A, B), parvalbumin-immunohistochemistry (C, D), and AChE-histochemistry (E, F). Scale bars are 100 km for A, C, E and 20 &m for B, D, F. In Nissl-stained sections, striatal neurons consist of numerous small round achromatic cell bodies (open circles), and of a few chromatic cell bodies, which are either medium-size elongated

(large black circles) or large round (stars). Glial cells are also visible (small blacks circles). Note that PV-immunoreactive neurons (C) and AChE-positive neurons (E) are far less numerous than achromatic neurons (A) and that PV-immunoreactive neurons have smaller and more elongated cell bodies than AChE-positive neurons.

immunohistochemical procedures. However, it does not constitute an adequate basis for a general classification. Other classifications (Kolliker, 1896; Dbjerine, '01; Ramon y Cajal, '11; Leontovich, '54) have been based on the length of axons (i.e., projection versus local circuit neurons), a feature which has an obvious functional meaning but which can be biased by incomplete Golgi impregnation (see Pasik et al., '79 for review). Finally, other studies have focused on the presence of dendritic spines (Fox et al., '71a,b; DiFiglia et al., '76). However, this feature can vary with different

experimental conditions (Fox et al.,'71b; DiFiglia et al., '80; Graveland et al., '85) and this can make it difficult to identify individual neurons as spiny or aspiny. Recently, Rowel1 ('89) has proposed that the normal procedural rules of classical organism taxonomy would be highly suitable for the classification of neurons. He insists on several rules that should be used in neuronal taxonomy, among which is the need for taxonomy to reflect phylogenetic relationships. A phylogenetic transformation of a given neuronal system can appear as species differences in

J. YELNIK ET AL.

288

.

A 3-

. *

* . o-*;

*

4l 3

**** ** * *

9

... . '

0.

.

*

oQooBooo

.*. 9 ' .

&g:oOo

**

1.-

0

.

,

1

I

1,000

500

.

.

%$2-

B

.' . .. .

c a

..

41*

D

4-

. .. ... ... . * *0

8

1

100

.

.

3-

.

.

* **'

* 400

* ** SS

700

Fig. 12. Correlation diagrams of cell body-size (ss in pm2) and cell-body shape (sh). In Golgi material (A), spiny neurons (open circles), leptodendritic neurons (large black circles), spidery neurons (stars), and microneurons (small black circles) were identified on the basis of their dendritic morphology. Note that microneurons do not segregate from spiny neurons. In Nissl-stained material (B), cell bodies were

identified as small achromatic (open circles) or large chromatic (black circles). In material treated for PV-immunohistochemistry (C), both PV-immunoreactive neurons (large circles) and nonlabeled neurons (small circles) were analyzed. In material treated for AChE-histochemistry (D), both AChE-positive neurons (stars) and AChE-negative neurons (small circles) were analyzed.

neuronal morphology. Unfortunately, such differences have never been investigated in previous studies of striatal neurons, even when different classifications were correlated (Pasik et al., '79; Chang et al., '82; Bolam et al., '84a; Takagi et al., '84). Another requirement that Rowel1 ('89) formulates is the recognition of individual variations and their expression by quantitative and statistical means. This requires that neurons are entirely reconstructed and analyzed in three dimensions, thus making the process much more time-consuming. Nevertheless, we consider this condition as the most crucial one for elaborating a reliable taxonomy. We also assume that a taxonomy based on the morphology of dendritic arborizations has a higher functional value than a classification based on the morphology of cell bodies. Our previous theoretical (Percheron, '79a,b, '82; Yelnik et al., '81, '83) and experimental (Yelnik and Percheron, '79; Hammond and Yelnik, '83; Yelnik et al., '84, '87) studies have shown that topological, metrical, and geometrical parameters are powerful criteria for differenciating neuronal groups on statistical bases. Such groups can be classified quantitatively in a multidimensional and hierarchical taxonomy for which we have proposed the neuronal species as the basic taxon (Yelnik et al., '87).

further morphological, physiological, and pharmacological analyses of the striatum. In order to correlate these neuronal species with previous classifications, it is necessary to examine in some details the criteria upon which neuronal types were defined. The spiny neuronal species. Neurons covered with dendritic spines have been mentioned in all previous studies (Table 2). We demonstrate in this study that spiny neurons are also characterized by other dendritic features (topological, metrical, geometrical) which are quantitatively remarkably stable. We even observed neurons with a low spine density which were statistically indistinguishable from typical spiny neurons after reconstruction and quantitative analysis. The existence of a second type of spiny neurons characterized by a lower spine density (spiny type I1 of DiFiglia et al., '76; medium with fewer dendritic spines of Rafols and Fox, '79; Type I11 of Braak and Braak, '82) is in fact questionable. These neurons have also spines on their soma and proximal dendrites, but this feature also exists in type I neurons (DiFiglia et al., '76; Fig. 5 in Graveland et al., '85). Variations in spine density could be attributed to different delays in postmortem fixation and to random reactions to silver impregnation. However, it has been demonstrated that spine density increases progressively from birth to 16 weeks (DiFiglia et al., '80), whereas it can decrease with aging (Graveland et al., '85). We thus conclude that spiny neurons form a homogeneous neuronal species characterized by stable topological, metrical, and geometrical parameters, and to which belong 96% of stri-

Taxonomy of primate striatal neurons Our classification, based on quantitative morphological features, considerably simplifies the description of striatal neurons since it comprises only four strongly contrasted neuronal species. It is hoped that this will be useful for


289

Lept o

Spiny

.-------

Micro

Spidery Fig. 13. Computer drawing of the lattice formed by the dendritic arborizations of spiny neurons (Spiny), leptodendritic neurons (Lepto), microneurons (Micro), and spidery neurons (Spidery) of the primate striatum. For each neuronal species, all the dendrites present in a rectangular portion of striatal tissue (400 x 400 x 10 pm) were drawn.

atal neurons. Typically, they have spine-free somata and proximal dendrites and spine-laden distal dendrites, a distribution whose functional significance is still unknown. Natural, experimental, or age-related variations in spine distribution can be observed. The leptodendritic neuronal species. This name (from Leptos: slender) was coined by Ramon Moliner (’62, ’69) to

designate “spindle-shaped neurons with scanty, extremely long, and relatively rectilinear dendrites.” This author was the first to point out the characteristic morphology of these neurons, their distribution in various cerebral regions, and their general significance. In the striatum, they are indisputably different from spiny neurons (Fig. 3). However, due to the length of their dendrites, which often exceeds 1 mm,

J. YELNIK ET AL.

290

TABLE 2 ComDarison of Striatal Neuronal Suecies of Primates (Spiny, Leptodendritic, Spidery, Microneurons) with Previous Classifications’

Kolliker Ramon y Cajal Leontovich

Fig. 713 Fig. 325A,B Long axon densely ramified arbory

Fox et al. DiFiglia et al Pasik et al. Graveland et al. Rafols and Fox

Spiny Spiny I and Spiny I1 Medium with many dendritic spines Type I Medium

Braak and Bra&

LEPTO

SPIDERY

Fig. 712 Fig. 325C Lone axon sDarselv ramified reticu;1 . ” Large aspiny (Figs. 3, 16 in ’71b) Large version of spiny I1 Large with fewer dendrites

-

Type 111 Large multipolar with extended dendrites

their dendritic arborizations are never contained in a single Golgi section. This has probably hampered their identification in previous studies. Their spiny or aspiny nature was also a disputed question. For Ramon y Cajal (’ll),dendrites were “bristled with a multitude of spines,” whereas for Leontovich (’541, spines were more sparsely distributed than with spiny neurons. Fox et al. (’71b) considered leptodendritic neurons as aspiny neurons and concluded: “the spines Ramon y Cajal observed on the large neurons with long axons in the immature striatum may be transitory.” DiFiglia et al. (’76) classified leptodendritic neurons as spiny neurons (the large version of spiny type 111,but they were working on the brains of young monkeys ( 3 4 months of age). Rafols and Fox (’79)described leptodendritic neurons as bearing “a few scattered spines,” whereas for Braak and Braak (’821, they were aspiny neurons. Our conclusion is that leptodendritic neurons form an homogeneous neuronal species which is topologically, metrically, and geometrically distinct from the spiny neuronal species. Although occasional spines or spinelike appendages can be observed, particularly in young individuals, these neurons are fundamentally of the aspiny type as are the large neurons of the globus pallidus (Yelnik et al., ’84) and substantia nigra pars reticulata (Yelnik et al., ’87), which belong to the same neuronal species. Quantitative analysis of cell bodies suggested that leptodendritic neurons of the striatum were PV-immunoreactive neurons. As leptodendritic neurons of the globus pallidus and substantia nigra pars reticulata are also PV-immunoreactive in both rats (Gerfen et al., ’85; Celio, ’90) and monkeys (present material, unpublished data), parvalbumin could be a selective marker of the leptodendritic neuronal species. In the rat striatum, however, PV-immunoreactive neurons are interneurons with varicose dendrites (Cowan et al., ’go), which do not resemble leptodendritic neurons. This indicates that a phylogenetic transformation is likely to occur. The spidery neuronal species. This name was first used by Fox et al. (’71b)to describe an aspiny neuron in monkeys having “a fanciful resemblance to spiders” but also, later, for either spiny (Lu and Brown, ’77) or aspiny neurons (Danner and Pfister, ’79a; Bolam et al., ’84b) in the rat. Nevertheless, we have retained this name since it refers, albeit subjectively, to the “spidery” appearance of these neurons (Fig. 5). Spidery neurons, which were not mentioned by Kolliker (1896) or Ramon y Cajal (’111, were probably seen by Leontovich (’54, Figs. 4 B2, 5D; see also Leontovich, ’75, Fig. 21Ia). The first indisputable illustrations were given by Fox et al. (’71b, Figs. 4, 81, but, surprisingly, they classified both these neurons and leptodendritic neurons (their Figs. 3 and 16) as large aspiny

MICRO

Large with many dendrites

Fig. 325D Short axon smooth and radiated dendrites Small with somatic spines Aspiny I and Aspiny 111 Small aspiny

Type IV large aspiny

Type V small to medium aspiny

Short axon smooth and suiral dendrites Large aspiny (Figs. 4,8 in ’71b) Aspiny I1

neurons. Spidery neurons were isolated as a particular neuronal type, the aspiny type 11,by DiFiglia et al. (’76)and have since been described under different names (Table 2). In all studies including the present, it is reported that spidery neurons have the largest cell bodies of the striatum. As the cell bodies of AChE-positive neurons are larger than those of unlabeled neurons (Fig. 12D),as the distribution of AChE closely corresponds to that of choline-acetyltransferase ( C U T ) in the striatum (Mesulam et al., ’841, and as CUT-positive neurons constitute a single population of large neurons (Nagai et al., ’83; Mesulam et al., ’84; Smith and Parent, ’84; Satoh and Fibiger, ’85;DiFiglia, ’871,it can be concluded that the cholinergic neurons of the primate striatum are the spidery neurons of Golgi material. The microneuronal species. This name was first used by McLardy (’63) to describe local interneurons of the thalamus whose cell bodies were definitely smaller than those of thalamo-cortical neurons. Although microneurons of the striatum have the same cell body size as projection spiny neurons, we retained this name, which refers to their very short and thin processes. Neurons of this species, which were not shown by Kolliker (1896), have been described in all subsequent studies (Table 2). Fox and Rafols (’71) first considered the dwarf neurons of Ramon y Cajal (’11, Fig. 325D) as oligodendrocytes. However, they finally distinguished microneurons having cell bodies “either in or smaller than the size-range of spiny neurons,” from oligodendroglia having smaller cell bodies (Fox et al., ’74). DiFiglia et al. (’76) also distinguished glial cells from small aspiny neurons, but they subdivided the latter into an aspiny type I having varicose processes and an aspiny type I11 having irregular processes. Apart from this feature, all other characteristics were similar (Table 1 in Pasik et al., ’79). Rafols and Fox (’79) considered that the aspinous I and I11 neurons both fell into a single category. Microneurons were correlated with the aspiny type 111by Braak and Braak (’82) and with the aspiny type I by Graveland et al. (’85). The absence of varicosities on some microneurons may be explained either by the incomplete impregnation of their distal varicose processes, or by the natural variation within local circuit neurons (FranGois et al., ’79). We observed, as did DiFiglia et al. (’76), Pasik et al. (’771, and Braak and Braak (’82), that many dendrites were in fact thin axonlike processes. Microneurons having two s o n s were even reported (Braak and Braak, ’82). The most realistic interpretation is thus that the topologicalorganization of axonlike processes can vary from neuron to neuron: they may emanate from a single process which would appear as an individualized axon, or from two or more

TAXONOMY OF PRIMATE STRIATAL NEURONS different processes. In the latter case, the microneuron could be considered as an amacrine cell. Intermediate neurons. This name refers to quantitative morphological features which are intermediate between those of spiny and leptodendritic neurons. Neurons having the qualitative characteristics of intermediate neurons, i.e., a triangular or elongated cell body with numerous, long, thin, and beaded dendrites, have rarely been described. The large spiny I1 neuron of Pasik et al. ('76, Fig. 1C; '79, Fig. 3C), the medium neuron with radiating beaded dendrites of Rafols and Fox ('79, Fig. 6) and the aspiny I neuron shown in Figures 15 and 17 of Graveland et al. ('85) could be intermediate neurons. We have observed intermediate neurons near the medullary laminae of the lenticular nucleus in both the striatum (present study) and globus pallidus (FranGois et al., '84; Yelnik et al., '84). Given this distribution, they are probably a perilaminar neuronal population belonging to the lenticular nucleus as a whole. We suggested that intermediate neurons of the globus pallidus could be pallido-habenular neurons (FranGois et al., '84), but striatal neurons projecting to the habenula were not reported by Parent ('79). A detailed analysis would be necessary to analyze the exact distribution and efferent target of intermediate neurons.

Species differences Classifications of striatal neurons of rats, cats, and mice have been based predominantly on the size of cell bodies. The classes obtained-small, medium, and large neuronshave been further subdivided according to different dendritic features into five (Danner and Pfister, '79a,b), six (Kemp and Powell, '71; Dimova et al., '80; Tanaka, '80; Rafols et al., '89), seven (Lu and Brown, '77), or eight (Chang et al., '82) neuronal types. Two of the neuronal species of the primate striatum, the spiny neurons and the microneurons, are present in these classifications. Both primate and nonprimate spiny neurons account for 94%96% of the total population, have a round or oval cell body, spine-free proximal stems, spine-laden distal dendrites, and a richly ramified local axonal arborization. Neurons with short, thin, and beaded axonlike processes have been classified as either small neurons (Fig. 7C of Chang et al., '82) or medium neurons (Fig. 2B of Kemp and Powell, '71; Fig. 7B of Changet al., '82; Fig. 2 of Rafols et al., '89), but as they clearly differ from the glial cells shown by Kemp and Powell ('71; Fig. 4B), Dimova et al. ('80; Fig. 16) and Rafols et al. ('89; Fig. 6), they are likely to be homologous to primate microneurons. Comparison of leptodendritic neurons and spidery neurons with the striatal neurons of nonprimate species has brought out unexpected results with regard to the phylogenetic evolution of the striatum. Morphologically, the large striatal neurons (i.e., those having the largest cell bodies in the striatum) are the spidery neurons in primates, whereas in nonprimate species they have large elongated cell bodies with a small number of thick, long, and sparsely ramified dendrites (Dbjerine, '01, Fig. 279; Leontovich, '54, Fig. 3B1, B2; Kemp and Powell, '71, Fig. 4A; Danner and Pfister, '79b, Fig. 6; Dimova et al., '80, Fig. 14; Tanaka, '80, Figs. IA, 2A; Bishop et al., '82, Fig. 4B; Changet al., '82, Figs. 1, 2; Changand Kitai, '82, Fig. lB,C; Takagi et al., '84, Fig. 22; Rueda et al., '86, Fig. 3; Rafols et al., '89, Fig. 7). They are therefore, indisputably, leptodendritic neurons and not spidery neurons as in primates. Conversely, neurons resembling the spidery neurons of primates, i.e., with a volumi-

291 nous cell body, numerous, thin, varicose, and curved-back processes, have never been shown in rats, mice, or cats. In primates, cholinergic neurons are the spidery neurons (DiFiglia, '87; present results). In rats, Lynch et al. ('72), Butcher and Bilezikjian ('75), Butcher and Hodge ('76), Henderson ('811, and Armstrong et al. ('83) reported two types of AChE-containing neurons, namely medium-size and large fusiform multipolar. However, by studying ChATimmunoreactivity, Eckenstein and Sofroniew ('83) observed a continuous range of sizes between small and large neurons and they noted that small profiles in 20-pm-thick sections could simply represent angled sections of larger neurons. Phelps et al. ('85, '89) also concluded that the apparent heterogeneity of ChAT-positive neurons might be accounted for by different orientations of elongated perikarya with respect to the plane of section. The ChATpositive neurons of rats would thus form a single population of neurons with large elongated cell bodies that Butcher and Bilezikjian ('75), Armstrong et al. ('831, Satoh et al. ('831, Bolam et al. ('84b), and Phelps et al. ('85) have correlated with leptodendritic neurons (i.e., the large neurons of Kemp and Powell, '71). Moreover, Bolam et al. ('84a,b) demonstrated, by using a combination of Golgiimpregnation, electron microscopy, AChE-histochemistry, and ChAT-immunohistochemistry, that the cholinergic neurons of rats were leptodendritic neurons and not spidery neurons. It therefore must be concluded that spidery neurons do not exist in rats and cats and that, in these species, large cholinergic neurons are leptodendritic neurons and not spidery neurons as in primates.

Functional implications Differences of neuronal typology between primates and nonprimates suggest that the internal circuitry of the striatum has changed during phylogenesis. This implies specificproperties in terms of information processing which can be analyzed at the level of both individual neurons and neuronal sets. I n individual neurons, the number and connectivity of branching nodes (topological parameters) and the length and thickness of dendritic segments (metrical parameters) determine how postsynaptic potentials will be processed (Rall, '64; Percheron, '82; Bras et al., '87). In spiny neurons, for example, an important summation of afferent inputs is likely to occur since they have numerous branching nodes and short dendritic segments. The difference that was observed between human and monkey spiny neurons (Table 1)implies a n even higher summation in human. The spine density, which is lower in humans, could reflect a more abundant local circuitry (Graveland et al., '85). It is worth mentioning that the dendrites of spiny neurons are similar in number, length, and spine density to the basal dendrites of the pyramidal neurons of the cortex. This morphological similarity, which implies common properties in terms of information processing, is probably an important characteristic of the cortico-striatal system. In addition the dimensions of dendritic arborizations of spiny neurons are similar in the monkey and the human (Table 1) and in the rat (Kitai et al., '79; Preston et al., '801, which suggests that they represent a morphological and functional module of information processing. Leptodendritic neurons are highly different. As they have long dendritic segments and few branching nodes postsynaptic potentials are likely to propagate with a marked decrement and with little summation. Their morphology and frequency, which

292 are similar to those of the leptodendritic neurons of the globus pallidus and substantia nigra, indicate that they also form a convergent system of information processing (Percheron et al., ’87). The role of spidery neurons is less clear. Their processes are characteristically varicose (Fig. 5) and DiFiglia and Carey (’86) have shown that up to 240 pm from the cell body varicosities are mainly postsynaptic. However, as the longest processes are 540 pm in length (Table l), the most distal varicosities could also be presynaptic and participate to the local output of spidery neurons. An important finding of this study is that the striatal cholinergic neurons are spidery neurons in primates, whereas they are leptodendritic neurons in nonprimates. As the former have far smaller dendritic arborizations than the latter (Fig. 71, cholinergic neurons of nonprimates could process information from wide striatal regions, whereas those of primates could not. Microneurons resemble the local circuit neurons of other cerebral regions (FranCois et al., ’79). In the striatum they participate to a profuse local circuitry, which is also formed by the local axonal arborizations of other neurons and presumably some processes of spidery neurons. In a neuronal set, dendritic arborizations intertwine with each other to form a dendritic lattice. The total number of neurons and the total length of dendrites determine the “capacity” of a neuronal set (Percheron et al., ’871, which is to be compared for different neuronal sets and different animal species. It is a noteworthy fact that the capacity of the spiny lattice had to be expressed in millions of neurons and in kilometers of dendrites. The three-dimensional extent of a single arborization determines the “dendritic dwelling” (Percheron et al.,’871, i.e., the portion of space within which a neuron can receive afferent inputs. Its dimensions determine the “grain” of a neuronal set: the larger the grain, the wider the cerebral region from which individual neurons can receive axons. Neurons whose cell bodies are close to each other have almost the same dwelling and are in position to receive similar afferent inputs. This redundancy, which is likely to ensure a reliable transmission of nervous information, can be expressed as the index of dendritic overlap. The dendritic lattice formed by the spiny neurons of the striatum is a dense, fine-grain, highly redundant lattice. These characteristics are very similar to those of the lattice formed by the pyramidal neurons of the cortex, which have almost the same frequency (unpublished data). In addition, both neuronal species have a local axonal arborization through which many neighboring neurons can establish synaptic contacts with each other. Since the terminal axonal arborizations of pyramidal neurons fall within the size range of the dendritic arborizations of spiny neurons (less than 300 km wide in DiFiglia et al., ’78), the preciseness of cortical information can be preserved by the spiny striatal neurons. Conversely, the leptodendritic neurons of the striatum form a loose, large-grain lattice, which is only slightly redundant. Since their dendritic arborizations are very large, they can receive converging information from wide cortical areas. The efferent information they transmit is thus likely to have only a crude topographical and functional specificity. The spidery neuronal species forms a fine-grain but little redundant lattice. Since the density of spiny neurons is very high (2,592 spiny neurons in the dendritic dwelling of one spidery neuron), the role of spidery neurons could be to synchronize the activity of groups of neighboring spiny neurons and by this means to

J. YELNIK ET AL. reinforce their action on pallidal and nigral neurons. Microneurons, which have very short processes and form a discontinuous nonredundant lattice, are likely to participate to very local circuits. Finally, the striatum is composed of an intertwining of different dendritic lattices. With regards to dendritic density, there is a marked contrast between the very dense redundant spiny lattice and the loose lattices formed by the other neuronal species (Fig. 13). With regard to the size of dendritic arborizations, there is another striking contrast between a fine-grain lattice and a large-grain lattice: the fine-grain lattice, formed by the dendritic arborizations of spiny neurons, spidery neurons and microneurons, falls within the size range of the striosomal compartment (300600 pm; Graybiel and Ragsdale, ’78, ’831, whereas the large-grain lattice, formed by the dendritic arborizations of leptodendritic neurons, is 2-4 times larger (1,200 pm). The striatum could thus process cortical information through two different systems: a fine-grain system, which would conserve the precision of cortical inputs, and a large-grain system, which would blur it. One or other of the systems, or both together, could be activated at a given moment depending on the nature and level of afferent signals.

LITERATURE CITED Abercrombie, M. (1946) Estimation of nuclear population from microtome sections. Anat. Rec. 94:239-247. Anderson, A.D. (1954) Alcoholic thionin counterstaining following Golgi dichromate-silver impregnation. Stain Technol. 29: 179-184. Armstrong, D.M., C.B. Saper, A.I. Levey, B.H. Wainer, and R.D. Terry (1983) Distrihution of cholinergic neurons in rat brain: Demonstrated by the immunocytochemical localization of choline acetyltransferase. J. Comp. Neurol. 216:53-68. Bishop, G.A., H.T. Chang, and S.T. Kitai (1982) Morphologicaland physiological properties of neostriatal neurons: An intracellular horseradish peroxidase study in the rat. Neuroscience 7t179-191. Blalock, H.M., J r . (1972) Social Statistics. New York McGraw Hill. Bolam, J.P., C.A. Ingham, and A.D. Smith (1984a) The section-Golgiimpregnation procedure-3. Combination of Golgi-impregnation with enzyme histochemistry and electron microscopy to characterizeacetylcholinesterase-containingneurons in the rat striatum. Neuroscience 12t687709. Bolam, J.P., P. Somogyi, S. Totterdell, and A.D. Smith (1981) A second type of striatonigral neuron: A comparison between retrogradely labelled and Golgi-stained neurons at the light and electron microscopic levels. Neuroscience 6 2 1 4 1-2157. Bolam, J.P., B.H. Wainer, and A.D. Smith (1984b) Characterization of cholinergic neurons in the rat neostriatum. A combination of choline acetyltransferase immunocytochemistry, Golgi-impregnation and electron microscopy. Neuroscience 12:711-718. Braak, H., and E. Braak (1982) Neuronal types in the striaturn of man. Cell Tissue Res. 227:319-342. Bras, H., P. Cogan, and S. Tyc-Dumont (1987) The dendrites of single brain-stem motoneurons intracellularly labelled with horseradish peroxidase in the cat. Morphological and electrical differences. Neuroscience 22947-970. Butcher, L.L., and L. Bilezikjian (1975) Acetylcholinesterase-containing neurons in the neostriatum and substantia nigra revealed after punctate intracerebral injection of diisopropylfluorophosphate. Europ. J. Pharmacol. 34:115-125. Butcher, L.L., and G.K. Hodge (1976) Postnatal development of acetyicholinesterase in the caudate-putamen nucleus and substantia nigra of rats. Brain Res. 106:223-240. Celio, M.R. (1990) Calbindin D-28k and parvalbumin in the rat nervous system. Neuroscience 35375-475. Chang, H.T., and S.T. Kitai (1982) Large neostriatal neurons in the rat: An electron microscopic study of gold-toned Golgi-stained cells. Brain Res. Bull. 83331-643. Chang, H.T., C.J. Wilson, and S.T. Kitai (1982) A Golgi study of the rat neostriatal neurons: Light microscopic analysis. J. Comp. Neurol. 208: 107-126.

TAXONOMY OF PRIMATE STRIATAL NEURONS Cooley, W.W., and P.R. Lohnes (1971) Multivariate Data Analysis. New York John Wiley & Sons. Cowan R.L., C.J. Wilson, P.C. Emson, and C.W. Heizmann (1990) Parvalbumin-containing GABAergic interneurons in the rat neostriatum. J. Comp. Neurol. 302197-205. Danner, H., and C. Pfister (1979a) Untersuchunger zur Struktur der Neostriatum des Ratte. J. Hirnforsch. 20:285-301. Danner, H., and C. Pfister (1979b)Zur Struktur des Neostriatum der Ratte. Gegenbaurs Morphol. Jahrb. 12513):S.349-364. Danner, H., and C. Pfister (1981a) Spine-haltige Neurone im Caudatusputamen-komplex der Ratte. J. Hirnforsch. 2275-84. Danner, H., and C. Pfister (1981b) 4 Spine-lose Neurontypen im Neostriatum der Ratte. J. Hirnforsch. 2 2 4 6 5 4 7 7 . Danner, H., and C. Pfister (1981~)Darstellung Spine-haltiger Zwergneurone im Caudatus-putamen-komplex der Ratte nach Pargylinappliation. Z. Mikrosk. Anat. Forsch., Leipzig 95(2/:S.210-216. Davenport, H.A., and C.M. Combs (1954) Golgi’s dichromate-silver method. 3. Chromating fluids. Stain Technol. 29t165-173. Dbjerine, J. (1901) Anatomie des centres nerveux. Paris: Rueff. DiFiglia, M. (1987) Synaptic organization of cholinergic neurons in the monkey neostriatum. J. Comp. Neurol. 255:245-258. DiFiglia, M., and J. Carey (1986) Large neurons in the primate neostriatum examined with the combined Golgi-electron microscopic method. J. Comp. Neurol. 244t36-52. DiFiglia, M., S. Christakos, and N. Aronin (1989) Ultrastructural localization of immunoreactive calbindin-D,,, in the rat and monkey basal ganglia, including subcellular distribution with colloidal gold labeling. J. Comp. Neurol. 279:653-665. DiFiglia, M., P. Pasik, and T. Pasik (1976) A Golgi study of neuronal types in the neostriatum of monkeys. Brain Res. 114t245-256. DiFiglia, M., P. Pasik, and T. Pasik (1980) Early postnatal development of the monkey neostriatum: A Golgi and ultrastructural study. J. Comp. Neurol. 19Ot303-331. DiFiglia, M., T. Pasik, and P. Pasik (1978) A Golgi study of afferent fibers in the neostriatum of monkeys. Brain Res. 152341-347. Dimova, R., J. Vuillet, and R. Seite (1980) Study of the rat neostriatum using a combined Golgi-electron microscope technique and serial sections. Neuroscience 5: 1581-1596. Eckenstein, F., and M.V. Sofroniew (1983) Identification of central cholinergic neurons containing both choline acetyltransferase and acetylcholinesterase and of central neurons containing only acetylcholinesterase. J. Neurosci. 3:2286-2291. Fox, C.A., and J.A. Rafols (197111972) Observations on the oligodendroglia in the primate striatum. Are they Ram6n y Cajal’s “dwarf’ or “neurogliaform” neurons? J. Hirnforsch. 13:331-340. Fox, C.A., I.J. Lu Qui, and J.A. Rafols (1974) Further observations on Ram6n y Cajal’s “dwarf’ or “neurogliaform” neurons and the oligodendroglia in the primate striatum. J. Hirnforsch. 15517-527. Fox, C.A., A.N. Andrade, D.E. Hillman, and R.C. Schwyn (197111972a)The spiny neurons in the primate striatum: A Golgi and electron microscopic study. J. Hirnforsch. 13r181-201. Fox, C.A., A.N. Andrade, R.C. Schwyn, and J.A. Rafols (197111972b) The aspiny neurons and the glia in the primate striatum: A Golgi and electron microscopic study. J. Hirnforsch. 13t341-362. Fransois, C., G. Percheron, J. Yelnik, and S. Heyner (1979) Demonstration of the existence of local circuit neurons in the Golgi-stained primate substantia nigra. Brain Res. 172160-164. Franqois, C., G. Percheron, J. Yelnik, and S. Heyner (1984) A Golgi analysis of the primate globus pallidus. I. Inconstant processes of large neurons, other neuronal types, and afferent axons. J. Comp. Neurol. 227:182-199. Gerfen, C.R., K.G. Baimbridge, and J.J. Miller (1985) The neostriatal mosaic: Compartmental distribution of calcium-binding protein and parvalbumin in the basal ganglia of the rat and monkey. Proc. Natl. Acad. Sci. U.S.A. 82:8780-8784. Goldman-Rakic, P.S., and L.D. Selemon (1990) New frontiers in basal ganglia research. TINS 13:241-244. Graybiel, A.M., and C.W. Ragsdale, Jr. (1978) Histochemically distinct compartments in the striatum of human, monkey and cat demonstrated by acetylcholinesterase staining. Proc. Natl. Acad. Sci. U.S.A. 7557235726. Graybiel, A.M., and C.W. Ragsdale, Jr. (1983) Biochemical anatomy of the striatum. In P.C. Emson (ed): Chemical Neuroanatomy. New York: Raven Press, pp. 427-504.

293 Graveland, G.A., R.S. Williams, and M. DiFiglia (1985) A Golgi study of the human neostriatum: Neurons and afferent fibers. J. Comp. Neurol. 34r317-333. Hammond C., and J. Yelnik (1983) Intracellular labelling of rat subthalamic neurones with horseradish peroxidase: Computer analysis of dendrites and characterization of axon arborization. Neuroscience 8:781-790. Henderson, Z. (1981) Ultrastructure and acetylcholinesterase content of neurons forming connections between the striatum and substantia nigra ofrat. J. Comp. Neurol. 197:185-196. Hsu, S.M., L. Raine, and H. Fanger (1981) Use of avidin-biotin-peroxidase complex (ABC) in immunoperoxidase techniques: A comparison between ABC and unlabeled antibody (PAP) procedures. J. Histochem. Cytochem. 29:577-580. Izzo, P.N., A.M. Graybiel, and J.P. Bolam (1987) Characterization of substance P- and [Metlenkephalin-immunoreactive neurons in the caudate nucleus of cat and ferret by a single Golgi procedure. Neuroscience 20577-582. Kemp, J.M. (1968) Observations on the caudate nucleus of the cat impregnated with the Golgi method. Brain Res. 11.467470. Kemp, J.M., and T.P.S. Powell (1971) The structure of the caudate nucleus of the cat: light and electron microscopy. Philos. Trans. R. SOC.Lond. [Biol.]262t383-401. Kitai, S.T., R.J. Preston, G.A. Bishop, and J.D. Kocsis (1979) Striatal projection neurons: Morphological and electrophysiological studies. In Poirier, L.J., T.L. Sourkes, and P.J. Bedard (eds): Advances in Neurology. New York; Raven Press, Vol. 24, pp. 45-51. Kolliker, A. (1896) Handbuch der Gewebelehre des Menschen. Leipzig: Engelmann. Leontovich, T.A. (1954)Fine structure of subcortical ganglia (in Russian). 2. Nevropat. Psikh. 54:168-178. Leontovich, T.A. (1975) Quantitative analysis and classification of subcortical forebrain neurons. In M. Santini (ed): Golgi Centenial Symposium Proceedings. New York: Raven Press, pp. 101-122. Leontovich, T.A. (1983) Spatial arrangement of tissue elements of nucleus caudatus in dog (in Russian). Neurophysiol. 1 5 4 7 4 4 8 4 . Levey, A.I., B.H. Wainer, E.J. Mufson, and M.M. Mesulam (1983) Colocalization of acetylcholinesterase and choline acetyltransferase in the rat cerebrum. Neuroscience 9:9-22. Lu, E.J., and W.J. Brown (1977) The developing caudate nucleus in the euthyroid and hypothyroid rat. J. Comp. Neurol. 171:261-284. Lynch, G.S., P.A. Lucas, and S.A. Deadwyler (1972) The demonstration of acetylcholinesterase containing neurones within the caudate nucleus of the rat. Brain Res. 45:617-621. McLardy, T. (1963)Thalamic microneurones. Nature 1992320-821. Mesulam, M.-M., E.J. Mufson, A.I. Levey, and B.H. Wainer (1984) Atlas of cholinergic neurons in the forebrain and upper brainstem of the macaque based on monoclonal choline acetyltransferase immunohistochemistry and acetylcholinesterase histochemistry. Neuroscience 12: 669-686. Nagai, T., T. Pearson, F. Peng, E.G. McGeer, and P.L. McGeer (1983) Immunohistochemical staining of the human forebrain with monoclonal antibody to human choline acetyltransferase.Brain Res. 265:300-306. Parent, A. (1979) Identification of the pallidal and peripallidal cells projecting to the habenula in monkey. Neurosci. Lett. 15:159-164. Pasik, P., T. Pasik, G.R. Holstein, and J. Hamori (1988) GABAergicelements in the neuronal circuits of the monkey neostriatum: A light and electron microscope immunocytochemical study. J. Comp. Neurol. 270: 157-170. Pasik, P., T. Pasik, and M. DiFiglia (1976) Quantitative aspects of neuronal organization in the neostriatum of the macaque monkey. In M.D. Yahr (ed.): The Basal Ganglia. New York: Raven Press, pp. 57-90. Pasik, P., T. Pasik, and M. DiFiglia (1979) The internal organization of the neostriatum of mammals. In I. Divac, and R.G.E. Oberg (eds): The Neostriatum. New York: Pergamon Press, pp. 5-36. Pasik, T., P. Pasik, and M. DiFiglia (1977) Interneurons in the neostriatum of monkeys. In J. Szentagothai, J. Hamori, and E.S. Vizi (eds): The Neuron Concept Today. Budapest: Akademai Kiado, pp. 153-162. Percheron, G. (1979a) Quantitative analysis of dendritic branching. I. Simple formulae for the quantitative analysis of dendritic branching. Neurosci. Lett. 14:287-293. Percheron, G. (1979b) Quantitative analysis of dendritic branching. 11. Fundamental dendritic numbers as a tool for the study of neuronal groups. Neurosci. Lett. 14t295-302. Percheron, G. (1982) Principles and methods of the graph theoretical analysis of natural binary arborescences. J. Theoret. Biol. 99r509-552.

294 Percheron, G., C. Franqois, and J. Yelnik (1987) Spatial organization and information processing in the core of the basal ganglia. I n M.B. Carpenter and A. Jayaraman (eds): The Basal Ganglia 11. Structure and Function-Current Concepts. New York: Plenum Press, pp. 205-226. Percheron, G., J. Yelnik, and C. Frangois (1986) Systems of coordinates for stereotactic surgery and cerebral cartography: Advantages of ventricular systems in monkeys. J. Neurosci. Methods 17:69-88. Phelps, P.E., C.R. Houser, and J.E. Vaughn (1985) Immunocytochemical localization of choline acetyltransferase within the rat neostriatum: A correlated light and electron microscopic study of cholinergic neurons and synapses. J. Comp. Neurol. 238t286-307. Phelps, P.E., D.R. Brady, and J.E. Vaughn (1989) The generation and differentiation of cholinergic neurons in rat caudate-putamen. Develop. Brain Res. 46:47-60. Poirier, L.J., A. Parent, R. Marchand, and L.L. Butcher (1977) Morphological characteristics of the acetylcholinesterase-containing neurons in the CNS of DFP-treated monkeys. Part 1. Extrapyramidal and related structures. J. Neurol. Sci. 31r181-198. Preston, R.J., G.A. Bishop, and S.T. Kitai (1980) Medium spiny neuron projection from the rat neostriatum: An intracellular horseradish peroxidase study. Brain Res. 183:253-263. Pullen, A.H. (1984) A structured program in Basic for cell morphometry: Its application to the spinal motoneurone. J. Neurosci. Methods 12155178. Rafols, J.A., H.W. Cheng, and T.H. McNeill(1989) Golgi study of the mouse striatum: Age-related dendritic changes in different neuronal populations. J. Comp. Neurol. 279t212-227. Rafols, J.A., and C.A. Fox (1979) Fine structure of the primate striatum. Appl. Neurophysiol. 4213-16. Rafols, J.A., and T.H. McNeill(1987)Age-related dendritic changes of spiny and aspiny neurons in the rodent striatum. In M.B. Carpenter and A. Jayaraman (eds): The Basal Ganglia 11. Structure and FunctionCurrent Concepts. New York Plenum, pp. 227-239. Rall, W. (1964) Theoretical significance of dendritic trees for neuronal input-output relations. I n R.F. Reiss (ed): Neuronal Theory and Modeling. Stanford, CA Stanford Univ. Press, pp. 73-97. Ramon Moliner, E. (1962) An attempt at classifying nerve cells on the basis of their dendritic patterns. J. Comp. Neurol. 1 I9t211-227. Ramon-Moliner, E. (1969) The leptodendritic neuron: Its distribution and significance.Ann. N.Y. Acad. Sci. 16735-70. R a m h y Cajal, S. (1911) Histologie du systhme nerveux de l’homme et des vertebres. Paris: Maloine. Rowe, M.H., and J. Stone (1977) Naming of neurons. Classification and naming of neurons of cat retinal ganglia cells. Brain Behav. Evol. 14: 185-2 16. Rowell, C.H.F. (1989) The taxonomy of invertebrate neurons: A plea for a new field. TINS 12:169-174. Rueda, J., J. Prieto, J. Juiz, and A. Angulo (1986) A Golgi study on the nucleus accumbens septi of the rat. J. Hirnforsch. 5515-520.

J. YELNIK ET AL. Satoh, K., and H.C. Fibiger (1985) Distribution of central cholinergic neurons in the baboon (Pap20 pupio). I. General morphology. J. Comp. Neurol. 236:197-214. Satoh, K., W.A. Staines, S. Atmadja, and H.C. Fibiger (1983) Ultrastructural observations of the cholinergic neuron in the rat striatum as identified by acetylcholinesterase pharmacohistochemistry. Neuroscience 4:11211136. S a y , C., J. Yelnik, E. Martin-Maxtinelli, I. Karpouzas, and J. NguyenLegros (1989) Distribution and spatial geometry of dopamine interplexiform cells in the rat retina: I. Developing retina. J. Comp. Neurol. 289t99-110. Schroder, K.F., A. Hopf, H. Lange, and G. Thorner (1975) Morphometrischstatistische Structuranalysen des Striatum, Pallidum and Nucleus subthalamicus beim Memschen. J. Hirnforsch. 16:333-350. Smith, Y., and A. Parent (1984) Distribution of acetylcholinesterasecontaining neurons in the basal forebrain and upper brainstem of the squirrel monkey (Saimiri sciureus). Brain Res. Bull. 12t95-104. Takagi, H., P. Somogyi, and A.D. Smith (1984) Aspiny neurons and their local axons in the neostriatum of the rat: A correlated light and electron microscopic study of Golgi-impregnated material. J. Neurocytol. 13t239265. Tanaka, D., J r . (1980) Development of spiny and aspiny neurons in the caudate nucleus of the dog during the first postnatal month. J. Comp. Neurol. 192t247-263. Tyner, C.F. (1975) The naming of neurons. Applications of taxonomic theory to the study of cellular populations. Brain Behav. Evol. 12:75-96. Wilson, C.J., and P.M. Groves (1980) Fine structure and synaptic connections of the common spiny neuron of the rat neostriatum: A study employing intracellular injection of horseradish peroxidase. J. Comp. Neurol. 194:599-616. Yelnik, J., and G. Percheron (1979) Subthalamic neurons in primates: A quantitative and comparative analysis. Neuroscience 4t1717-1743. Yelnik, J., G. Percheron, and C. Franqois (1984) A Golgi analysis of the primate globus pallidus. 11.Quantitative morphology and spatial orientation of dendritic arborizations. J. Comp. Neurol. 227200-213. Yelnik, J., G. Percheron, and C. Franqois (1989) Taxonomy of neurons of the neostriatum of primates: Classification into four neuronal species. IBAGS. 3rd. Congress. Abstr. p. 184. Yelnik, J.,C. Franqois, G. Percheron, and S. Heyner (1987) Golgi study of the primate substantia nigra I. Quantitative morphology and typology of nigral neurons. J. Comp. Neurol. 265t455-472. Yelnik, J., G. Percheron, C. Franqois, and Y. Burnod (1983) Principal component analysis: A suitable method for the three-dimensional study of the shape, dimensions and orientation of dendritic arborizations. J. Neurosci. Methods 9t115-125. Yelnik, J., G. Percheron, J. Perbos, and C. Frangois (1981) A computer-aided method for the quantitative analysis of dendritic arborizations reconstructed from serial sections. J. Neurosci. Methods 4:347364.

Modular organization of projection neurons in the matrix compartment of the primate striatum.

Neurons in the primate dorsal striatum signal the uncertainty of object-reward associations.

The skin primates. XLIV. Numerical taxonomy of primate skin.

Primate taxonomy: species and conservation.

Cellular Taxonomy of the Mouse Striatum as Revealed by Single-Cell RNA-Seq.

Responses of neurons in the primate taste cortex to glutamate.

Collaterals of primate spinothalamic tract neurons to the periaqueductal gray.

Birds have primate-like numbers of neurons in the forebrain.

Morphological assessment of neuronal aggregates in the striatum of the rat.

New neurons in the adult striatum: from rodents to humans.

Morphological characterization of respiratory neurons in the pre-Bötzinger complex.

Representation of the body by single neurons in the dorsolateral striatum of the awake, unrestrained rat.

The effects of neuropeptide urocortin 2 on the spontaneous discharge and glutamatergic neurotransmission of striatum neurons.

Interneurons in the rat striatum: relationships between parvalbumin neurons and cholinergic neurons.

Inhibition of primate spinothalamic tract neurons by stimulation in the region of the nucleus raphe mahnus.

The non-human primate striatum undergoes marked prolonged remodeling during postnatal development.

Brain stem projections to the striatum: experimental morphological studies in the rat.

Functional and morphological aspects of hypothalamic neurons.

Degeneration of the nigral dopamine neurons after 6-hydroxydopamine injection into the rat striatum.

Parkinson׳s disease-related modulation of functional connectivity associated with the striatum in the resting state in a nonhuman primate model.

Unsupervised lineage-based characterization of primate precursors reveals high proliferative and morphological diversity in the OSVZ.

The synaptology of parvalbumin-immunoreactive neurons in the primate prefrontal cortex.

Integrative taxonomy in the Liolaemus fitzingerii complex (Squamata: Liolaemini) based on morphological analyses and niche modeling.

Behavioral modulation of sensory responses of primate putamen neurons.