Proc. Nati. Acad. Sci. USA Vol. 86, pp. 2093-2097, March 1989 Neurobiology

Visualization of chandelier cell by immunoreactivity in monkey cerebral cortex (/inhibition) J. DEFELIPE*, S. H. C. HENDRYt, AND E. G. JONESt *Section of Neuroanatomy, Cajal Institute, Madrid, Spain; and tDepartment of Anatomy and Neurobiology, University of California, Irvine, CA 92717 Communicated by Vernon B. Mountcastle, December 15, 1988

ABSTRACT Antibodies directed against the calcium- at either light or electron microscopic levels, the presence of binding protein parvalbumin label a subpopulation of y- peptides or any substance other than GABA or GAD in aminobutyric acid-releasing neurons in the cerebral cortex that chandelier cell axons. Hence, chandelier cells appear to is thought to have particular metabolic and physiological constitute a "pure" type of GABA cell. properties. The chandelier cell is a well-characterized morpho- Recent immunocytochemical studies in the rat and monkey logical type of y-aminobutyric acid-releasing cortical interneu- cerebral cortex have shown that the calcium-binding protein ron, the of which possesses very distinctive terminal parvalbumin (PV) is found almost exclusively in a subpop- portions located around the initial axon segments of pyramidal ulation ofGABA cells (18, 19), which, at least in the rat, does cells. In the pre- and postcentral gyri of the monkey, we found not contain the peptides cholecystokinin and somatostatin that these distinctive terminal portions of chandelier cell axons (20). It has also been shown that immunoreactivity for the were immunocytochemically stained for parvalbumin in a 28-kDa vitamin D-dependent calcium-binding protein calbin- manner that reveals their complete structure. The chandelier din and PV is present in separate subpopulations of GABA cell axons were identified light-microscopically as short, ver- cells in the monkey cortex and that, on the basis of somal tically oriented rows of parvalbumin-positive puncta (PV-Rs). size, the large basket cells, another well-characterized type of The PV-Rs varied in both length and complexity and were GABAergic cells, are likely to contain PV (10). Here, we located beneath unstained pyramidal cells. PV-Rs were very demonstrate that PV can also be used as a reliable marker for numerous in layers II-Ill, where most pyramidal cells ap- chandelier cell axon terminals and, thus, for the chandelier peared to have a PV-R beneath them. Fewer PV-Rs were found cells themselves. in deeper layers, and in layer VI PV-Rs were rare. With EM all PV-Rs could be seen to form multiple synaptic contacts of the symmetrical type on the initial segments of MATERIALS AND METHODS axons. Parvalbumin immunoreactivity can therefore be used as This study was made on tissue taken from the cerebral cortex a reliable marker for chandelier cell axons. of five Macaca fascicularis brains that have also been used in other studies (19, 21). All animals were normal. Colchicine There is evidence that a very large percentage of neocortical was not used. Animals were deeply anesthetized with Nem- interneurons are likely to use the inhibitory neurotransmitter butal and perfused through the heart with 0.1 M phosphate y-aminobutyric acid (GABA) (1). In the past, attempts to buffer (pH, 7.2) followed by a 0.1 M phosphate-buffered correlate cell profiles immunoreactive for GABA or its synthe- mixture of 2-3% paraformaldehyde and 0.1% or 0.2% glutar- sizing enzyme glutamic acid decarboxylase (GAD) with the aldehyde. The most complete immunoreactive staining was several morphological types of cortical interneurons have gen- obtained when the higher concentrations of paraformalde- erally relied upon indirect evidence, for the pattern of axonal hyde and glutaraldehyde were used. arborization, which is the most distinctive feature of each type The brains were removed and cut into smaller blocks that of cortical , is not revealed fully by immunocy- were postfixed for 2 hr in 2% paraformaldehyde in 0.1 M tochemistry (2-4). GABA cells can also be subdivided with phosphate buffer. The blocks used in the present study were regard to their content of neuroactive peptides (5-10) and sagittal slices 3-5 mm thick spanning the pre- and postcentral calcium-binding proteins (9), but, again, direct correlation with gyri. After postfixation they were transferred to 7% sucrose morphologically defined forms has been difficult. in 0.1 M phosphate buffer, in which they remained overnight. One of the best-studied GABAergic interneurons in the Next day the blocks were sectioned at 20 or 50 ,um on a cerebral cortex is the chandelier cell, which is distinguished Vibratome and stained immediately by immunocyto- by the terminal portions ofits axon, which form short vertical chemistry. Sections were washed repeatedly in 0.1 M phos- rows ofterminal boutons resembling candlesticks (11). These phate buffer, preincubated in 3% normal rabbit serum/0.05% terminal portions form multiple synaptic contacts exclusively Triton X-100, and then transferred to a mixture containing a with the axon initial segments of pyramidal cells and are sheep antiserum made against HPLC-purified PV (P. C. thought to be a major source of GABAergic synapses on Emson, personal communication). This antiserum was di- pyramidal cell axons (12-17). Because of this remarkable luted 1:1000 or 1:2000 in 0.1 M phosphate buffer/3% normal synaptic specificity and presumably strong inhibitory influ- rabbit serum/0.05% Triton X-100 for immunocytochemical ence on pyramidal cells, chandelier cells probably play a staining. Sections were incubated for 24 or 48 hr at 4°C. crucial role in the control of pyramidal cell excitability and, Thereafter, they were washed repeatedly in phosphate buffer thereby, in the control of cortical efferent activity. Up to the and further treated by the ABC-peroxidase (Vectastain) present, all immunocytochemical studies have failed to show, method. The bound immunoglobulin-peroxidase complexes

The publication costs of this article were defrayed in part by page charge Abbreviations: PV, calcium-binding protein parvalbumin; PV-Rs, payment. This article must therefore be hereby marked "advertisement" vertical rows of PV-positive puncta; GABA, y-aminobutyric acid; in accordance with 18 U.S.C. §1734 solely to indicate this fact. GAD, glutamic acid decarboxylase.

Downloaded by guest on September 27, 2021 2093 2094 Neurobiology: DeFelipe et al. Proc. Natl. Acad. Sci. USA 86 (1989) were visualized by staining in 0.05% diaminobenzidine thin sections were collected on Formvar-coated, single-slot tetrahydrochloride/0.002% hydrogen peroxide. grids, stained with uranyl acetate and lead citrate and Stained sections were washed several times in phosphate examined with a JEOL-1OOB electron microscope. buffer and postfixed in 2% osmium tetroxide, dehydrated, and flat-embedded in Spurr's resin (Polyscience). Serial 2-gm-thick plastic sections were then obtained with a Re- RESULTS ichert ultramicrotome, examined with the light microscope, Numerous darkly stained immunoreactive cell somata were photographed, and then resectioned at 65-70 nm (22). The found in all layers except layer I ofall cortical areas examined

'vA'lRa9,.1 e*.* 4 .99 *;t *.t ._

p- at@ . *

f,9 Aray . 9 * -49' 9

P * t e *; ; * t *t k | 5 - 4.1~~~~~~~~~~~~9 j Al

*;.j t

9: S * - ' v '----' ''111-''

799; .'t : I . Cu a..!itch; . :x~~k.)-t i ,. an '-.. ' -IC~~~~ 9, *SM -i e r 999 .. - 5$t 4'^*:i, P>>-a

.:s a v *9.6' At~ -''>i~~~,t'rjra2iBt1 t > s^A Atf /~.~ Gil }i. .f'-rt., * | j 9et .~~~\9. >t-rAi,49 t-._9. '9 *'^[ B8 S S~w ; 't 1 A .I'V *. :'

9% -~~~1 9,W. '*sese;-ift **sv _ XS.~~~~~~~~.-!

FIG. 1. PV immunostaining through area 4 (A, B, D, E) and area 3b (C). (A) Low-power photomicrograph that includes layer I and the upper part of layer IIIA. No PV-positive neurons are found in layer I. Notice the predominant labeling of PV-Rs (short, vertically oriented rows of puncta) (arrows) in the neuropil of layers I-I4A. (B) Higher magnification of the PV-R e shown in D, but at a different focal plane, in which the PV-R e is seen surrounding the presumed axon initial segment of an unstained pyramidal cell (p). (C) Three unstained pyramidal cells (p) in layer IIIA that have PV-Rs beneath them with different complexities. (D) Higher magnification of the bracketed area in A. Note the large number of complex PV-Rs. The PV-Rs a-d are also shown in Fig. 2. (E) PV-positive cell (L) in upper part of layer V. Fewer complex PV-Rs are seen (arrows) than in layer II (compare with D). An unstained pyramidal cell covered by puncta probably derived from basket cell axons is also indicated (p). [Bars = 250 glm (A), 10 ,um (B), 25 Mm (C), and 50 tum (D and E).] Downloaded by guest on September 27, 2021 Neurobiology: DeFelipe et al. Proc. Natl. Acad. Sci. USA 86 (1989) 2095 (Fig. 1A). The greatest number of PV-positive neurons were varied in both their length (from 15 um to 35 pum) and their found in layers IIIB and IV. Layers II, IIIA, and V contained complexity: there were PV-Rs made up of only one or two a moderate number of PV-positive cells, and there were few rows of about 10 puncta, whereas others were so complex PV-positive cells in layer VI. PV-positive neurons showed a that they formed tight braid-like structures (Figs. 1C and 2) variety of somal sizes and shapes (Fig. 1 A and E), but none with large numbers of puncta. Each PV-R was located were pyramidal, and where proximal dendrites were stained, immediately beneath or at a short distance (no more than 15 the morphology was invariably nonpyramidal. Small- ,tm) below a nonimmunoreactive soma that in many cases to-medium (10-15 gtm) PV-positive cells with an ovoid soma could be identified as a pyramidal cell (Figs. 1C and 2). seemed to predominate, but from layer IIIB through layer V Frequently the PV-Rs outlined shapes that suggested the there was also a considerable number of large (20-30 gm), axon initial segment of the corresponding cell located above multipolar, PV-positive cells (Fig. 1E), which closely resem- them (Fig. 1B). ble the large basket cells of the monkey cortex (3, 23). The complex PV-Rs were not homogeneously distributed A large number ofimmunoreactive processes and punctate throughout all layers; they were very numerous in layers II- elements were found in the neuropil throughout all layers III (Fig. 1D). From layer IV to layer VI there was a dramatic except layer I of both pre- and postcentral gyri. Layer I was decrease in the number of this type of PV-R (Fig. 1E), characterized by an almost complete lack of stained puncta, especially in the deep part of layer VI, where PV-Rs of any and there were only a few beaded processes (Fig. 1A), kind were only occasionally found. The general low density although these could often be followed to stained parent cells of the background PV-immunoreactive plexus in layers II- located in layer II. These processes are likely to be dendrites. IIIA caused the abundant complex PV-Rs in these layers to The density of the plexus of immunoreactive processes and be the most overt stained elements of the neuropil (Fig. 1 A puncta in other layers was related to the number of PV- and D). In any given region of layers II-III, the number of positive cells found in each layer. Thus, the densest plexus PV-Rs of both single and complex types was so high that was found in layers IIIB and IV; density was moderate in virtually every pyramidal cell appeared to have a PV-R layers Il-lIIA and low in layer VI. Numerous puncta were beneath it (Fig. 2). Close examination revealed, however, seen around the unstained somata of pyramidal cells (Fig. 1 that there were some pyramidal cells that did not have any C and E), especially in layer IIIB and V, where many PV-Rs associated with them. pyramidal cells were covered by immunoreactive puncta, In previous Golgi-EM studies, the terminal portions of the outlining their somata and proximal dendrites. These puncta chandelier cell axons have been found to form symmetrical have previously been interpreted (1, 23) as the axon terminals synaptic contacts with initial segments of pyramidal cell of basket cells. axons (12-17). Therefore, to verify that PV-Rs were indeed Apart from the above mentioned features of the immuno- the terminal portions of chandelier cell axons, we studied cytochemical staining, the most conspicuous characteristic whether the initial segments ofpyramidal cell axons were the was the labeling in layers II-VI of short, vertically oriented structures postsynaptic to the PV-Rs. First, all readily rows of puncta (Fig. 1), which appeared identical to the identifiable PV-Rs were localized in 2-,um-thick plastic semi- terminal portions of the chandelier cell axons identified in thin sections (Fig. 3 A and B), and then the same semithin Golgi preparations ofthe monkey sensory-motor cortex (17). sections were resectioned for EM (Fig. 3C). We found that As occurs with terminals of Golgi-impregnated chandelier all PV-Rs studied (n = 32) were made up of round or cell axons, the PV-positive vertical rows of puncta (PV-Rs) elongated terminal boutons, many ofthemjoined together by

4. 4 . a

St 44 I p ...

*4 * r xf 4, .4 :: *i .* lfv t le za / 44 *, Su '.e/ :- -* 7: VI . f 'qF 10 0~~~~~~ ,.0 I V % * I p p k 6 0 441. * %I. v '4 i i I .: 1 *L P.,.-

..l

* 4 A *. "e , ws

AP @. S~~~ .*1*j wX.,.^4 d s t § { >t4>§

.b .1 .1

* W 4 4 .. * * ,. .B

FIG. 2. (A, B, and C) Photomicrographs at different planes offocus, but at the same magnification, showing four PV-Rs (a-d), also indicated in Fig. 1D. The PV-Rs are located beneath unstained pyramidal cells (p). (Bar = 15 ,um.) Downloaded by guest on September 27, 2021 2096 Neurobiology: DeFelipe et al. Proc. NatL Acad. Sci. USA 86 (1989) 3-$ 2 ..1 ...2 i.j, I, i .b /ox

A B

f, 'i {C.

R. s A.. j s t..t--.

FIG. 3. Correlative light and electron micrographs of PV-Rs in layer II of area 4. (A and B) Photomicrographs of two consecutive semithin sections showing four PV-Rs (a-d) and three PV-positive cells (1-3). Arrows indicate same blood vessels. (C) Electron micrograph taken after resectioning the semithin section shown in A, which contains the PV-R b. Asterisks, the same blood vessels as indicated by arrows in A and B. (D) Higher magnification ofthe PV-R b illustrated in C showing a portion ofan unstained axon initial segment (is) and the PV-positive boutons that made up the PV-R b. (E) Serial section showing the continuation ofthe axon initial segment, is, indicated in D, and the PV-positive boutons surrounding the is. de, Same dendrite as in D. (F) Serial section showing two PV-positive boutons forming symmetrical synaptic contacts (arrows) with the is. [Bars = 30 aum (A and B), 5 ,um (C), 1 ,.m (D and E), and 0.5 tkm (F).] cytoplasmic bridges. The boutons contained numerous syn- almost exclusively in a subpopulation of GABA cells (see aptic vesicles and usually one, two, or more mitochondria, Introduction). Among the various morphological types of the external surfaces of which were covered with immuno- GABA interneurons, it has been recently suggested from cytochemical reaction product (Fig. 3 D and E). All PV-Rs correlative evidence that the large basket cells are very likely were found surrounding the axon initial segments of pyra- to be included in the PV-positive group, but others (for midal cells, forming multiple symmetrical synaptic contacts example, the double bouquet cells) are more likely to be with them (Fig. 3F). Thus, we conclude that chandelier cell included in a different subpopulation of GABA cells that are axons are indeed the sources of PV-Rs. Tracing the parent defined by their immunoreactivity for another calcium- axons back to chandelier cell somata is not possible in this binding protein, calbindin (19). In the present study we have of material, but the selective termination of the PV-Rs demonstrated at both light and electron microscopic levels type axons on axon initial segments of pyramidal cells confirms their that the distinctive terminal portions of chandelier cell chandelier cell origin. are stained for PV. This staining is as complete as that obtained with Golgi methods, but the staining for PV has the DISCUSSION advantage that, instead of labeling only selected terminal Cortical neurons displaying PV-immunoreactivity have a ramifications, as does the Golgi stain, PV immunoreactivity non-pyramidal morphology with a variety of shapes and sizes is more widespread, homogeneous, and may label virtually all therefore, constitute a heterogeneous group of interneu- chandelier cell terminations. Apart from the potential phys- and, PV GABA in rons. Colocalization studies have shown that PV is found iological significance of the coexistence of and Downloaded by guest on September 27, 2021 Neurobiology: DeFelipe et al. Proc. Natl. Acad. Sci. USA 86 (1989) 2097 chandelier cell axons, the importance of the present findings 2. Valverde, F. (1971) Int. J. Neurosci. 1, 181-197. lies in the capacity to quantify these terminations in the 3. Jones, E. G. (1975) J. Comp. Neurol. 160, 205-268. normal, experimental, and diseased cortex. It has been 4. Fairdn, A., DeFelipe, J. & Regidor, J. (1984) in Cerebral suggested, for example, that chandelier cell axons degenerate Cortex, eds. Peters, A. & Jones, E. G. (Plenum, New York), in cortical epileptic foci (24). Vol. 1, pp. 201-253. The present results indicated that the number of PV- 5. Hendry, S. H. C., Jones, E. G., DeFelipe, J., Schmechel, D., chandelier cell axons with complex morphology (i.e., those Brandon, C. & Emson, P. C. (1984) Proc. Natl. Acad. Sci. USA that give rise to a high density of axoaxonic synaptic 81, 6526-6530. contacts) are much more numerous in layers II-111 than in 6. Somogyi, P., Hodgson, A. J., Smith, A. D., Nunzi, M. G., layers V-VI. This distribution of complex PV-chandelier cell Gorio, A. & Wu, J.-Y. (1984) J. Neurosci. 4, 2590-2603. axons according to layers agrees with our previous EM study 7. Schmechel, D. E., Vickrey, B. G., Fitzpatrick, D. & Elde, of chandelier cell axons in monkey sensory-motor cortex R. P. (1984) Neurosci. Lett. 47, 227-232. (17). We found that initial segments of pyramidal cell axons 8. Lin, C., Lu, S. M. & Schmechel, D. E. (1986) J. Comp. were, on Neurol. 244, 369-383. average, contacted by more synaptic terminals in 9. Demeulemeester, H., Vandesande, F., Orban, G. A., Brandon, layers II-III than in layer V. The significance of the variabil- C. & Vanderhaeghen, J. J. (1988) J. Neurosci. 8, 988-1000. ity by layer and/of the variability in number of terminal 10. Jones, E. G., DeFelipe, J., Hendry, S. H. C. & Maggio, J. E. boutons per cell awaits functional analysis. It is also not yet (1988) J. Neurosci. 8, 1206-1224. determined to what extent the number ofterminals of a given 11. Szentdgothai, J. (1975) Brain Res. 95, 475-496. axon type is fixed for a particular recipient cell type in the 12. Somogyi, P. (1977) Brain Res. 136, 345-350. cortex or whether variability in number of synapses is 13. Fairdn, A. & Valverde, F. (1980) J. Comp. Neurol. 194, 761- reflected in different functional attributes. The present 779. method offers the opportunity of analyzing these issues. 14. Somogyi, P., Freund, T. F. & Cowey, A. (1982) Neuroscience In the present study, we also found that the number of 7, 2577-2607. perisomatic PV terminals on pyramidal cell somata was 15. Peters, A., Proskauer, C. C. & Ribak, C. E. (1982) J. Comp. higher in layers IIIB and V than in other layers. This is likely Neurol. 206, 397-416. due to the fact that the large basket cells, which are thought 16. Freund, T. F., Martin, K. A. C., Smith, A. D. & Somogyi, P. to be the major source of pericellular axon terminals on (1983) J. Comp. Neurol. 221, 263-278. pyramidal cell somata in layers III-V (3, 25), are also 17. DeFelipe, J., Hendry, S. H. C., Jones, E. G. & Schmechel, D. PV-positive (19). (1985) J. Comp. Neurol. 231, 364-384. It is thought that the GABAergic axoaxonic and axoso- 18. Celio, M. R. (1986) Science 231, 995-997. matic synapses on pyramidal cells should exert a strong 19. Jones, E. G., Hendry, S. H. C. & Emson, P. C. (1988) Soc. inhibitory effect on pyramidal cell excitability and the gen- Neurosci. Abstr. 14, 780. eration of 20. Kosaka, T., Heizmann, C. W., Tateishi, K., Hamaoka, Y. & axon potentials. PV/GABA cells appear to be Hama, K. (1987) Brain Res. 409, 403-408. electrically and metabolically more active than those GABA 21. Hendry, S. H. C., Jones, E. G., Emson, P. C., Lawson, cells that do not contain PV (see ref. 18), but the exact D. E. M., Heizmann, C. W. & Streit, P., Exp. Brain Res., in function of PV in these two kinds of cortical interneurons press. remains to be determined. 22. DeFelipe, J. & Fairdn, A. (1982) Brain Res. 244, 9-16. 23. DeFelipe, J., Hendry, S. H. C. & Jones, E. G. (1986) Neuro- 1. Houser, C. R., Vaughn, J. E., Hendry, S. H. C., Jones, E. G. science 17, 991-1009. & Peters, A. (1984) in Cerebral Cortex, eds. Jones, E. G. & 24. Ribak, C. E. (1985) Brain Res. 326, 251-260. Peters, A. (Plenum, New York), Vol. 2, pp. 63-89. 25. Marin-Padilla, M. (1969) Brain Res. 14, 633-646. Downloaded by guest on September 27, 2021