freund-buzsaki1994

HIPPOGIMPUS 6.347-470 (1996)

Interneurons of the Hippocampus T.F. Freundl and G. Buzsi5ki2

IInstitute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, Hungary and 2Center for Molecular and Behavioral Neuroscience, Rutgers University, Newark, New Jersey

KEY WORDS: pal cells, inhibition, inhibitory neurons

GABA, GABAergic cells, nonpyramidal cells, nonprinci-

Much of our current knowledge about the neuronal organization of the cerebral cortex comes from studies of principal neurons. Only recent work points to the crucial role of inhibitory interneurons in regulating the complex interactions among principal cells, including population oscillations, plasticity, epileptic synchronization, hormonal effects, and cortical development. Perhaps the best illustration of this point is the pivotal role interneurons play in population oscillations (theta, gamma, 200-Hz ripples) and memory-related plasticity in the hippocampal formation. The very ba- sics of neuronal cooperation is a subject of intensive research (“integrate- and-fire” vs. coincidence detectors; rhythmic vs. stochastic activity; Gernstein and Mandelbrot, 1964; Shadlen and Newsome, 1994; Softky, 1995). Even if a complete identification of the molecular and biophysical properties of single cells eventually becomes possible, such knowledge is insufficient to predict the behavior of large neuronal aggregates in complex integrative areas of the brain, such as the hippocampal formation. Population interactions of neuronal ensembles underlying behavioral control cannot be revealed without a comprehensive understanding of the dialogue between interneuronal networks and principal cell populations.

It has been known for over 100 years (Ram6n y Cajal, 1893, 191 1) that neurons constituting any cortical area are far from being uniform with regard to their morphology and connectivity, suggesting that they possess the capacity to interact with each other in a complex and diverse fashion. Descriptions of increasing numbers of cell types with distinct dendritic morphology and axonal targets result in a multitude of possible interactions. Even if only a fraction of these connections are active at one time, a network with virtually endless numbers of possible activity patterns

The authors wish to dedicate this monograph to the memory of their late mentors, Janos Szentdgothai and Endre Grastyin.

.Accepted for publication July 1, 1996. Address correspondence and reprint requests to Tam& F. Freund, Institute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, P.O. Box 67, H-1450, Hungary; e-mail: [email protected]; or Gyorgy Buzsdki, Center for Molecular and Behavioral Neuroscience, Rutgers University, Newark, NJ 071 02; e-mail: [email protected].

0 I996 WILEY-LISS, INC.

emerge. Recent results provide evidence that electrophysiology of unitary cellular interactions and of large cell populations in conjunction with precise structural data on the synaptic architecture may eventually prove to be realistic experimental approaches to study the complex questions of interneuron function (Buzsiki et al., 1992; Gulyis et al., 1993a,b; Buhl et al., 1994a; Sik et al., 1995; Ylinen et al., 1995a,b; Miles et al., 1996). Significant steps have been made toward this goal in the past two decades (Schwartzkroin and Mathers, 1978), in which advantages of the marriage between structural and functional analyses of neuronal networks have been realized, and the painstaking labor involved in this approach has begun to provide critical information not ac- cessible by other means. The multitude of recently aquired data representing several different levels of analysis still awaits synthesis. The outcome of such a synthesis may result in the generation of realistic hypotheses for the specific roles of hippocampal microcircuits and activity patterns and ultimately for the function of the hippocampal formation itself. The primary objective of this review is to provide a preliminary synthesis of much of the anatomical, cellular physiological, pharmacological, and systems physiological data focused on hippocampal interneurons and to propose specific unifying hypotheses for their roles in the control of network activity.

Why do interneurons represent a key to the understanding of network operations? In contrast with the rather uniform population of principal cells in any of the hippocampal subfields, the afferent and efferent connectivity of interneurons shows great variation (Ram6n y Cajal, 1893, 1911; Lorente de N6, 1934), thereby enabling them to carry out multiple tasks (Sections VI-XIII). Inhibition is critical in shaping response properties in single cells and in assisting cooperativity in large cell populations (Lytton and Sejnowski, 1991; Buzsiki et al., 1992; Traub et al., 1996; Ylinen et al., 1995a,b).

The basic cell types of the cerebral cortex were described in the pioneering Golgi impregnation studies of Ram6n y Cajal in 1893. For over half a century, neu- roanatomists referred simply to two major cell types, “pyramidal” and “nonpyramidal” neurons, largely on the

348 FREUND AND BUZSAKI

Abbreviations:

tACPD AMPA

5-HT

APV BDNF CA1-3

CB CCK ChAT CNQX CR DAB ENK EPSP, EPSC CABA GAD gamma

CluR HICAP cell

osci I lations

HlPP cell

HRP I A ~ A H P

IC IH I K IPSP, IPSC LM cell

LTD LTP mGluR MOPP cell

NADPH NCF Ni-DAB

NMDA N O N PY NT3 0 -LM cell

PCa, PNa PHAL PP PV ripple

SBA SOM SP SPR SPW T channel theta VIP VVA

5- hydroxy-tryptam i n, serotonin trans-1 -amino-cyclopentane-1,3-dicarboxylic acid amino-3-hydroxy-5-methyl-4-isoxazolepropi- onic acid 2-amino-5-phosphonovaleric acid brain-derived neurotrophic factor fields of the hippocampus (Cornu ammonis) according to Lorente de N 6 calbindin D28k cholecystokinin choline acetyltransferase 6-cyano-7-nitroquinoxal i ne-2,3-dione ca I ret i n i n 3,3’-diaminobenzidine 4HCI leu- and metenkephalin excitatory postsynaptic potential, current gamma aminobutyric acid glutamic acid decarboxylase 40-1 00-Hz field oscillatory waves

glutamate receptor hilar neuron with axon distributed in the commissural/associational pathway termination zone hilar neuron with its axon distributed in the perforant path termination zone horseradish peroxidase transient, rapidly inactivating current cyclic AMP-dependent potassium current calcium-activated potassium current hyperpolarization-activated current sustained current (potassium) inhibitory postsynaptic potential, current interneurons with cell bodies in stratum lacunosum-moleculare long-term depression long-term potentiation metabotropic glutamate receptor molecular layer cell with axon arborizing in the perforant path termination zone nicotinamide adenine dinucleotide phosphate nerve growth factor nickel-intensified 3,3‘-diaminobenzidine 4HCI (chromogen for peroxidase reaction) N-methy I-D-aspartate nitric oxide neuropeptide Y neurotrophin 3 interneuron with soma and dendrites in stratum orines, and axons in strata lacunosum-moleculare and oriens calcium or sodium permeability fhaseolus vulgaris-leucoaggl uti n i n perforant path parvalbumin fast (120-200 Hz), transient field oscillation in association with sharp waves soybean agglutinin somatostatin substance P substance P receptor hippocampal or entorhinal sharp waves low threshold calcium channel rhythmic slow wave activity (4-1 2 Hz) vasoactive intestinal polypeptide Vicia villosa agglutinin

basis of Ram6n y Cajal’s descriptions. A more detailed characterization of nonpyramidal cell types is found in the work of Lorente de N6 (1934), particularly for the hippocampal formation. I t was only in 1966, however, when Colonnier (1966) brought attention to the large variety of interneurons with distinctive dendritic and axonal fields described first by Ram6n y Cajal (1893, 1911). The word “inhibition” was not mentioned in any of the early Golgi studies. The assumption that interneurons may be responsible for local inhibition arose from Gray’s (1959) description of two types of synapses, asymmetrical (type I) and symmetrical (type 11), which motivated Eccles 1‘1964) promptly to pronounce type I as excitatory and type I1 as inhibitory (Andersen et al., 1963). An important step was again taken by Colonnier (1968), who demonstrated that the somata of pyramidal cells, where the basket type of interneurons terminate, were covered almost exclusively by type 11, presumably inhibitory, synapses. After Szentigothai’s (1962, 1965a,b) demonstration that specific basket types of axon arborizations remain intact in chronically isolated cortical slabs, the view of basket cells (and other types of nonpyramidal cells) as local inhibitory interneurons became well established. Further evidence for this conclusion has been provided by the selective interneuronal localization of glutamic acid decarboxylase (GAD; Ribak et al., L978), the synthesising enzyme of gamma-aminobutyric acid (GABA), the major inhibitory neurotransmitter in the cerebral cortex (Krnjevic and Schwartz, 1967).

Debates over nomenclature have surfaced in more recent years, particularly when neurons with local axons were shown to establish asymmetrical (presumably excitatory) synapses. Examples of these types of cells include mossy cells of the dentate gyrus (Laurberg and Sorensen, 1981; Ribak et al., 1985) and spiny stel- late cells of the neocortex (Somogyi, 1978). Confusion also ex- ists concerning what to call “interneurons” with axon collaterals that project to distant brain areas (Seress and Ribak, 1983; Ribak et al., 1986; T6th and Freund, 1992; T6th et al., 1993). These findings clearly indicated that the correspondence among Golgi’s type 2 short axon cells, the “nonpyramidal” cells of Ramon y Cajal and Lorente de N6, and inhibitory interneurons of Eccles, Colonnier, and Szentigothai was not at all straightforward. For the hippocampal formation, the term “nonprincipal cell” was sufficiently simple and correct to designate neurons mostly involved in local synaptic circuits, but some of them, in addition to their local collaterals, may have an extrahippocampal or commissural projection. Neurons with well-established local excitatory output (e.g., mossy cells) are not considered interneurons in this review, even if their axon remains restricted to che hippocampal formation. Given that most, if not all, nonprincipal cells use GABA as a transmitter (Section 111. l), the definition “GABAergic nonprincipal cells” appears to be the most precise term for the sub- jects of this review. However, the use of the term “interneurons” should not be discouraged in spite of the debate related to its con- notations because it carries an important and descriptive message about the major contribution of these cells to the wiring of local networks. Furthermore, the term “interneuron” is sufficiently simple, widely used, and will continue to be used in the future in spite of efforts to change it. Thus, we propose that the term

INTERNEURONS OF THE HIPPOCAMPUS 349

hippocampal “interneuron” should now be redefined as being synonymous with “GABAergic nonprincipal cell.”

In the following sections, the currently available data on the connectivity and function of interneurons will be reviewed in terms of their morphological characteristics, inputs, and outputs (Fig. 17 and 18), neurochemical features (Fig. 26), physiological and pharmacological properties in isolation and in networks, and in in vitro slices and in the behaving animal (Tables 2-4). An at- tempt will be made to synthesize these data to reach meaningful hypotheses for the roles of interneurons in the control of network patterns associated with plasticity and memory formation. The possible roles of interneurons in the pathomechanisms of disease states will also be discussed. The review is largely restricted to rodents (mainly rats) due to the considerable variability among species, especially regarding the neurochemical characteristics of apparently identical cell types. Nevertheless, a section is devoted to highlight major similarities and differences among primates, rodents, and submammalian species.

The types, characteristics, and connectivity of principal cells in the hippocampus (cornu ammonis, divided into three subfields, CA1-3, according to Lorente de No, 1934) and the dentate gyrus (fascia dentata and hilus) have been well known since the studies of Ramon y Cajal (1893) and Lorente de N6 (1934) and have been reviewed extensively (Amaral and Witter, 1989, 1995; Lopes da Silva et al., 1990). For sake of convenience, we will use the term “hippocampal formation” to refer to the hippocampus and dentate gyrus collectively, although it is acknowledged that the term is sometimes used to include the subicular complex and the entorhinal cortex (Amaral and Witter, 1995). The present overview is limited to the description and interpretation of data on interneurons in the hippocampus and the dentate gyrus (i.e., the hippocampal formation). The hippocampal formation therefore consists of a complex of three main subfields (Fig. l ) , the dentate gyrus, the CA3 region, and the CA1 region. The CA2 subfield is less well defined in the rat and appears to lack specific features regarding interneurons and will therefore be largely ig- nored here.

The dentate gyrus is considered to be the first stage of the intrahippocampal trisynaptic loop. It is the target for the majority of entorhinal afferents (Fig. 1) by carrying sensory information of multiple modalities about the external world. Merents from the lateral entorhinal cortex terminate in the outer one-third and those from the medial entorhinal cortex terminate in the middle one-third of the molecular layer, where dendrites of dentate principal cells arborize. Principal cells of the dentate gyrus are the granule cells, which number almost 1 million in the rat and 5 million in the monkey (Claiborne et al., 1986, 1990; Seress, 1988) and the mossy cells of the hilus (Amaral, 1978). The relatively small cell bodies of granule cells (8-12 p m in diameter) form a

densely packed layer called stratum granulosum (or granule cell layer), which is 4-8 somata in thickness. These cells characteristically have two main radially oriented dendrites emitting several fine branches, which reach the pial surface or the hippocampal fissure. Except for the most proximal shafts, all dendrites are densely spiny. The entire dendritic tree of granule cells is confined to stratum moleculare, the layer adjacent to stratum granulosum. Basal dendrites in the rat are extremely rare and mainly appear in the ventral hippocampus (L. Seress, personal communication). The axons of granule cells, called mossy fibers, originate at the opposite pole of the soma and enter the hilus, where they give rise to several local collaterals that largely remain in the hilar region (Claiborne et al., 1986). Recurrent collaterals peri- odically enter the granule cell layer, climb along the cell bodies and dendrites of presumed basket cells, and form multiple synapses only on these interneurons in normal animals (Ribak and Peterson, 1991). Occasional mossy fiber collaterals may also contact granule cell dendrites, particularly in stratum moleculare of the ventral dentate gyrus, but the physiological role of monosynaptic recurrent excitation among granule cells is likely to be negligible. The main axon of the granule cells leaves the hilar region and courses adjacent to the pyramidal cell layer (in stratum lucidum, see below) of the CA3 subfield, where they form giant en passant boutons, the characteristic mossy terminals, on the proximal dendrites of pyramidal cells. The hilus, or polymorphic zone of the dentate gyrus, is located subjacent to the granule cell layer and is bordered on the other side by the dendritic layer of CA3c that lies between the upper (suprapyramidal) and lower (in- frapyramidal) blades of the dentate gyrus. The principal and most numerous cell type of the hilus is the mossy cell, which has densely spiny dendrites and several thornlike excrescenses on both the cell body and proximal dendritic shafts. The dendrites of mossy cells are typically confined to the hilar region (Amaral, 1978), but some of them subsequently have been found to have a single dendritic branch penetrating stratum moleculare (Soltksz and Mody, 1994; Scharfman, 1995b). The axon of mossy cells innervates the inner one-third of the molecular layer of the dentate gyrus both ipsi- and contralaterally and also emits collaterals within the hilus (Amaral, 1978; Laurberg and Sorensen, 1981; Rbak et al., 1985; Buckmaster et al., 1996). Although their primary postsynaptic targets are the dendrites of granule cells, mossy cell collaterals also terminate on unidentified dendritic shafts in the hilus and occasionally on dendrites of interneurons (Frotscher and Zimmer, 1983a,b; Frotscher et al., 1984; Ribak et al., 1985; Buckmaster et al., 1996). Whether mossy cells should be considered as modified displaced pyramidal cells of the hippocampus or whether they represent a unique and specific cell type of the dentate gyrus is still a debated question (Frotscher et al., 1991). The recent demonstration of a back projection of ventral hippocampal CA3c pyramidal cells to the inner one-third of the molecular layer of the dentate gyrus, similar in distribution and termination pattern to mossy cell projections, provides strong support for the former interpretation (Li et a]., 1994). Both cell types make asymmetrical, presumably glutamatergic, excitatory (Soriano and Frotscher, 1994; Scharfman, 1995b) connections. Moreover, no major differences between CA3c pyramidal cells and mossy cells are ap-

B

FIGURE 1. A: Main excitatory connections in the hippocampal formation. Layer I1 of the entorhinal cortex forms a longitudinally widespread projection to granule cells and CA3 pyramidal cells via the perforant path. (The direct entorhinal cortex to CA3 connection is not shown.) The next stage is the mossy fiber projection, organized in a lamellar fashion, from the granule cells to the CA3 pyramidal cells. CA3 pyramidal neurons are strongly interconnected by a longitudinally projecting recurrent associational system. The CA3-CAI associational projection (Schaffer collaterals) is, again, longitudinally widespread. The extent of CA3-CA3 and CA3-CA1 projections in the septotemporal direction is similar, but more cal- laterals are present in CAI than in CA3. In contrast with the diver- gent multisynaptic system, layer I11 pyramidal cells of the entorhinal cortex provide a direct and spatially restricted innervation (in

the transverse dimension) of CAI pyramidal cells, which in turn project back to the same columns in the entorhinal cortex (deep layers). In essence, the entorhinal cortex is mapped onto the CAI region and the trisynaptic, intrahippocampal “diffuse” system is superimposed on this organized topography. Another major output from the CA1 region is to the subicular complex (not shown). Superimposed on the excitatory projections are the locally projecting and widely projecting inhibitory interneurons that form an interneuronal network (not shown). B: Coronal section through the dorsal hippocampus, immunostained for parvalbumin. 0, CA1 stratum oriens; p, stratum pyramidale; r, stratum radiatum; lm, stratum lacunosum-moleculare; m, dentate molecular layer (stratum moleculare); g, granule cell layer (stratum granulosum); h, hilus proper; M a - c , subregions of the CA3 field. From Buzsdki (1995).

parent in terms of their inputs or their electrophysiological or neurochemical features. For these reasons, mossy cells will be considered here as excitatory principal cells that provide long-range ipsilateral and commissural associational projections into the den- late gyrus (Amaral, 1978).

The CA3 subfield represents the second stage of the trisynap-

tic loop (Fig. 1 ) . Pyramidal cells of this region are the principal targets of granule cell axons, the mossy fibers. According to Lorente de N6 (1334), the region is subdivided further: the area penetrating the dentate hilus is called CA3c, the segment adjacent to CA2 is called CA3a (the curved segment), and the area located between is called CA3b. CA3 pyramidal cells have char-


acteristic large somata arranged in a layer of 60-120 p m thickness. One or two prominent apical dendrites emerge from the soma and often branch proximally into equally large-diameter segments (Ishizuka et al., 1995; Turner et al., 1995). The apical dendrites are radially oriented in strata radiatum and lacunosum moleculare, where they give rise to additional thin side branches that reach the hippocampal fissure or the border of the hilus proper. Basal dendrites are more numerous. They run obliquely in stratum oriens toward the alveus. The entire dendritic tree is densely covered with thin spines. In addition, the proximal dendrites also bear large complex spines, called thorny excrescenses, which form synaptic complexes with the large mossy terminals in a narrow layer called stratum lucidum (Ram6n y Cajal, 1893; Lorente de N6, 1934; Blackstad and Kjaerheim, 1961). The dendritic tree of CA3c pyramidal cells, situated between the blades of the den- rate gyrus, is somewhat different. The so-called basal and apical dendrites are rather uniform and relatively short, extending until the hilar border, and dendrites originating at both poles are covered by large numbers of thorny excrescences, implying a high density of mossy fiber input. Mossy synapses on basal dendrites are rare in CA3a and CA3b of the rat. Similarly to inputs of the dentate gyrus, the other major excitatory afferents of principal cells of this subfield are also strictly laminated. In addition to the mossy fiber projection in stratum lucidurn, the commissural and associational collateral input from other CA3 pyramidal cells terminate in strata radiatum and oriens. CA3 pyramidal cells also receive a direct projection from the entorhinal cortex, which innervates the most distal dendrites in stratum lacunosum moleculare (Nafstad, 1967; for review, see Amaral and Witter, 1989, 1995; Lopes da Silva et al., 1990; Tamamaki and Nojyo, 1993, 1995).

The axons of CA3 pyramidal cells arise from the lower pole of the soma or often from a primary basal dendrite. They give rise to extensive axonal arborizations within CA3 and, in addition, project even more densely to the CAI subfield, as originally described by Schaffer (1 892). The axon arbor of individual CA3 pyramidal cells may span two-thirds of the longitudinal extent of the hippocampus and may give rise to 20,000-60,000 axon terminals (Sik et al., 1993; Li et al., 1994). Pyramidal cells in CA3a appear to innervate both strata radiatum and oriens and have a bias for the adjacent segments of CA1, and C u b cells terminate in stratum radiatum and to a smaller extent in oriens, predomi- nating in intermediate areas of CA1, whereas CA3c pyramidal cells innervate stratum radiatum with a preference for distant segments of CA1 close to the subicular border (Ishizuka et al., 1990; Li et al., 1994). CA3c pyramidal cells in the ventral hippocampus also send several collaterals back into the dentate gyrus, where they terminate in the inner one-third of the molecular layer along with the hilar mossy cells. CA3 pyramidal cells also project to the contralateral hippocampus (commissural pathway) and to the lateral septa1 nucleus (Swanson et al., 1980; Tamamaki et al., 1984).

The CA1 subfield represents the third and last stage of the intrahippocampal trisynaptic loop and is the major target of CA3 pyramidal cell axons, the Schaffer collaterals (Fig. 1). Pyramidal cells of this region, which have somewhat smaller cell bodies than those in CA3, form a narrow layer (stratum pyramidale) of 50-100 p m thickness. CAI pyramidal cells typically give rise to

a single, radially oriented apical dendrite that emits several processes in stratum radiatum (Ishizuka et al., 1995). These dendritic processes ultimately terminate in a tuft of thin branches in stratum lacunosum-moleculare, usually reaching the hippocampal fissure. Basal dendrites are numerous, arborize in stratum oriens, and often reach the alveus. The dendritic tree is densely covered with spines, but thorny excrescenses are absent in the CA1 region. The axon emerges from the region of the soma adjacent to the apical dendrite or occasionally from a basal dendrite before entering the alveus. Local collaterals and consequently the excitatory interactions among CAI pyramidal cells are relatively sparse (Lorente de NO, 1934; Amaral and Witter, 1989; Amaral et al., 1991; Radpour and Thomson, 1992) compared with the CA3 region. These collaterals travel parallel to the alveus in stratum oriens and remain restricted to this layer (Ram6n y Cajal, 1893; Lorente de N6, 1934; Tamamaki and Nojyo, 1990). The main extrinsic projections of CAI pyramidal cells are to the subiculum and entorhinal cortex (Fig. I), but other limbic cortical areas and the lateral septum, the nucleus accumbens, and the olfactory bulb are also among the targets of the CAI subfield (for a detailed description see, Van Groen and Wyss, 1990). The CA1 region (together with the subiculum) should therefore be considered as the major output structure of the hippocampus back to the entorhinal cortex and indirectly to neocortical areas. In addition to the ipsi- and contralateral Schaffer collaterals, the other excitatory inputs of CAI pyramidal cells include entorhinal afferents that terminate in stratum lacunosum-moleculare.

The possible functional significance of morphological differences of various hippocampal cell types was first addressed by Ram6n y Cajal (1893). He and, later his pupil, Lorente de N6 (1934) demonstrated that characteristic laminar distributions of dendritic trees predicted the possible sources of afferent input, whereas the pattern of axon arborization provided a strong indi- cation of postsynaptic target selection. Their approach exempli- fied the utility of functional neuroanatomy, i.e., a method to predict function. A modern version of this approach, allowing direct conclusions to be made based on strict criteria, is the characterization of the input-output properties by combinations of tracing and multiple labeling techniques (for review, see Freund and Somogyi, 1989; Somogyi and Freund, 1989; Freund, 1993). Visualization of neurons with their dendritic and axonal processes can be achieved in different ways, including ( I ) classical Golgi impregnation methods, (2) intra- or extracellular dye injections, and (3) immunocytochemical staining for transmitters, their synthesizing enzymes, neuropeptides, calcium-binding proteins, cell surface markers, etc. The second approach, involving intracellular labeling, has the major advantage of combining direct physiological characterization and morphology. However, in immunocytochemical studies, most if not all cells that belong to the


same “neurochemically” identified class, are simultaneously visualized. This finding may provide crucial information about the frequency of certain cell types and the consistency of morphological features found in single-cell labeling studies. Importantly, it may help decide whether a new morphological finding obtained by single-cell labeling techniques is a peculiarity (e.g., a pathological or developmental abnormality, fixation artifact, etc.) or may represent a consistent feature of the network. Obviously, once the physiological effects of substances (e.g., neuropeptides, calcium-binding proteins) are revealed, it will add a new dimension to the immunocytochemical data. A major aim of this review is to establish a unifying classification scheme that integrates morphological, neurochemical, and physiological features. The morphological characteristics, which are based on single-cell labeling studies, will be presented first. Additional cell types, described to date only by immunostaining, will also be summarized here.

111.1. Chandelier or Axo-axonic Cells This cell type was discovered in the neocortex by Szentrigothai

and Arbib (1974) and described later in the hippocampus (Kosaka, 1983a; Somogyi et al., 1983a,b, 1985a) and dentate gyrus (Soriano and Frotscher, 1989). Since then, their presence and characteristic features have been investigated by several laboratories (Kawaguchi and Hama, 1987a,b; Soriano et al., 1990; Li et al., 1992; Gulyis et al., 1993a; Han et al., 1993; Buhl et al., 1994a,b). The major distinguishing feature of chandelier cells is the characteristic termination of their axon, which forms rows of 2-30 boutons in the principal cell layers and in stratum oriens and the hilus. These rows of boutons are aligned parallel to the axon initial segments of pyramidal and granule cells. Somogyi (1977) was the first to demonstrate in the neocortex that the exclusive postsynaptic elements of chandelier cells were axon initial segments of pyramidal cells, a finding that was later confirmed in the hippocampus and dentate gyrus (Kosaka, 1983a; Somogyi et al., 1983a,b, 1985a; Soriano and Frotscher, 1989; Soriano et al., 1990; Halasy and Somogyi, 1993b). This highly specific termination is unique among the interneurons and prompted Somogyi (1977) to endow these neurons with a functionally more descriptive name of axo-axonic cell.

Dentate gyms Chandelier cells are located within or immediately adjacent to

the granule cell layer. They have a dendritic tree with a tuft of several radially running branches in the molecular layer, most of which reach the outer one-third of the layer or even the hippocampal fissure (Soriano and Frotscher, 1989; Soriano et al., 1990; Han et al., 1993; Buhl et al., 1994a). Dendrites crossing the granule cell layer, directed toward the hilus, are small in number or occasionally even absent (Soriano et al., 1990). This arrangement of the dendritic wee suggests that the dominant excitatory input to these cells is feed-forward, originating from the perforant path and the commissural-associational projection. Nevertheless, the rather sparse basal dendrites appear to be suff- cient to provide the cell with a feedback drive because antidromic activation of granule cells was shown to discharge the axo-axonic

cell (Han et al., 1993). Short thick spines are occasionally present on the dendrites, but they are insufficient in frequency to call the cell sparsely spiny.

The axon originates from the soma or a proximal dendritic shaft, arborizes profusely in stratum granulosum (Soriano and Frotscher, 1989; Buhl et al., 1994a), and in the case of the cell reported by Han et al. (1993) also in the deep hilus or even in CA3c. The complex vertical rows of boutons following individual axon initial segments frequently have a horizontal or oblique orientation, especially at the top of the granule cell layer, where initial segments are trying to find their way to the hilus by traveling around granule cell bodies in their tightly packed layer. In the hilus, axon collaterals of chandelier cells also have an apparently random orientation similar to the position of initial segments of principal cells of this region (possibly mossy cells or displaced CA3c pyramidal cells). In the transverse dimension, the axon may span a distance well over a millimeter along the granule cell layer. Although there are no definite data about the longitudinal extent of these axons, the axon of a single chandelier cell was shown to arborize in all sections cut from a 400-pm-thick slice. Thus, 400 p m should be considered a minimum value for this dimension (Han et al., 1993). The distribution of terminal segments (i.e., rows of boutons) is rather homogeneous; patchy innervation, if it occurs at all, is exceptional. The number of bouton rows, which approximately equals the number of innervated target cells, is more than 1,000 for a completely visualized axo-axonic cell (for a fully reconstructed cell within a 400-pm slice, see Han et al., 1993). The terminal segments typically consist of 4-12 boutons (Han et al., 1993), but single boutons contacting axon initial segments can also be found (Soriano et al., 1990). Correlated electron microscopy of Golgi-impregnated or intracellularly filled chandelier cells demonstrate that the overwhelming majority of postsynaptic elements are the shafts or spines of axon initial segments of granule cells and hilar neurons (Soriano et al., 1990; Halasy and Somogyi, 1993b; Buhl et al., 1994a). Cell bodies of granule cells, especially regions close to the axon hillock, or unidentified processes are innervated extremely rarely (Soriano et al., 1990). The synaptic contacts formed by chandelier cells are always symmetrical, and the boutons contain several pleomorphic vesicles and mitochondria. Data derived from labeling of single axo-axonic cells suggest convergence of axons from several axo-axonic cells onto the axon initial segment of a single target neuron. In the neocortex, the axons of approximately four axo-axonic cells terminate on a single target cell (Freund et al., 1983).

Ultrastructural features of the soma-dendritic compartment of chandelier cells are similar to those of other interneurons. Their cell bodies have large invaginated nuclei, with occasional intranuclear rods, a cytoplasm rich in mitochondria and rough endoplasmic reticulum, and a dense perisomatic innervation by both asymmetrical and symmetrical synapses (Soriano et al., 1990). The dendrites are covered by a large number of primarily asymmetrical synapses on the distal segments and mixed with approximately an equal number of symmetrical synapses on the proximal shafts. Dendrites in the hilus often receive asymmetrical synapses from boutons showing characteristics of small mossy fiber terminals (Soriano et al., 1990).


A

B FIGURE 2. The axonal and dendritic arbors of an axo-axonic Note the numerous vertically oriented axon terminal segments in the cell (A) and a basket cell (B) in the CA3 region intracellularly filled case of the axo-axonic cell (A), which are rare in the basket cell ar- with biocytin in a 400-pm guinea pigslice in vitro and reconstructed bor (B). From Gdy& et al. (1993a). Scale bars = 50 pm. from serial 60-pm Vibratome sections. Insets illustrate the extent Reproduced from GulyAs et al. (1993a) “Precision and variability in and gross position of the axon arbors in CA3. The dendritic trees of postsynaptic target selection of hippocampal nonpyramidal cells.” both cell types are bitufted and span all layers, whereas the axon ar- Eur. J . Neurosci., 51729-1751 by permission of Oxford University bors are limited to strata pyramidale and proximal oriens. A small Press. number of basket cell collaterals also penetrate stratum radiatum.


Hippocampus

Chandelier cells of the hippocampus are very similar to those in the dentate gyrus. The cell bodies are located within or immediately adjacent to the pyramidal cell layer and possess radially oriented dendrites spanning all layers (Fig. 2). The dendrites are smooth, often varicose, and spines are only rarely present on a few branches. Three to six main dendritic trunks extend toward the hippocampal fissure, emitting few if any side branches in stratum radiatum. In stratum lacunosum-moleculare however, they form a tuft of fine, slender processes (Li et al., 1992; Gulyis et al., 1993a; Buhl et al., 1994b). There is a rich arbor of basal dendrites in stratum oriens, which extends up to, or occasionally penetrates, the alveus. Thus, according to the distribution of the dendritic tree, chandelier cells are in a position to receive excitatory input from all major sources of afferents in both the CA1 and CA3 subfields, with some preference for the perforant path input. The column occupied by the dendrites has a diameter of 200-300 pm.

The axon, which originates from the soma or a primary dendrite, forms a dense arbor in strata pyramidale and proximal oriens and consists of vertical or oblique rows of boutons (Figs. 2, 3). The only axo-axonic cell with a complete axonal and dendritic tree intracellularly labeled in vivo was described by Li et al. (1 992). They reported an axon arbor occupying an elliptical area of 600 by 850 pm, elongated in the septotemporal direction, within which 40-50-pm groupings of terminal segments could be observed in some sections. Three chandelier cells filled in the CA3 subfield of 400-pm-thick slices in the guinea pig had a transverse axonal spread of 800 pm, 1,300 pm, and 1,700 p m (Gulyis et al., 1993a), whereas one reported in CA1 extended for 950 p m (Buhl et al., 1994b). Each of these cells had axon collaterals in all sections cut from the 400-pm slice. The main axonal branches of axo-axonic cells run horizontally above the pyramidal cell layer and give rise to collaterals descending into stratum pyramidale, where they form characteristic bouton rows climbing on axon ini-

FIGURE 3. Light and electron micrographs of axo-axonic (A,C) and basket cell (B,D,E) axon collaterals. A: Axon collaterals (arrowheads) of a Golgi-impregnated axo-axonic cell in the monkey hippocampal CAI region climb along the axon initial segments of two unstained pyramidal cells (N, and N2). B: Basket cell collaterals (arrows) also form rows of boutons in strata pyramidale, proximal oriens, and radiatum, but these segments have no preferred orientation, unlike axo-axonic cells, which have numerous radially oriented terminal segments. Basket cell axons pass among the tightly packed pyramidal cell bodies and occasionally follow apical or basal dendrites into strata radiatum or oriens. The layers of the CA3 subfield are indicated. s.o., stratum oriens; s . ~ . , stratum pyramidale; s.]., stratum lucidum. C: Electron micrograph of an axon terminal of a Golgi-impregnated axo-axonic cell from the rat hippocampus, forming a symmetrical synapse (black arrow) on a spinelike appendage of an axon initial segment (ais). Ultrastructural characteristics of axon initial segments include an electron-dense membrane undercoating (arrowheads) and lamellar bodies (open arrows). D,E: Axon terminals of an intracellularly filled basket cell in the guinea pig hippocampal CA3 region form symmetrical synapses (arrows) on a cell body (s in D) and on a dendritic shaft (d in E). Data from Gulyds et al. (1993a) and Somogyi et al. (1983a, 19853. Scale bars = 10 pm for A,B, 0.2 pm for C-E.

tial segments. These rows are often obliquely oriented and follow the trajectory of axon initial segments. Each row consists of 2-15 axon terminals. The number of terminal segments closely corresponds to the number of innervated pyramidal cells, which was estimated to be approximately 1,200 for the completely reconstructed chandelier cell in CA1 (Li et al., 1992). This number may be even larger for CA3 cells, with an axon arbor twice as long in the coronal plane in the guinea pig (Gulyds et al., 1993a). Based on the average number of synapses on pyramidal cell axon initial segments and on the contribution of a single axo-axonic cell to these synapses, a convergence of 4-10 axo-axonic cells onto a single pyramidal cell has been estimated (Li et al., 1992).

Electron microscopy of chandelier cell axon terminals confirmed the absolute selectivity of this cell type for termination on axon initial segments of pyramidal cells in rat (Fig. 3), cat, and monkey (Somogyi et al., 1983a,b, 1985a; Gulyis et al., 1993a; Buhl et al., 1994a,b). Ultrastructural features of chandelier cell axons, dendrites and somata are indistinguishable from those described for dentate chandelier cells.

111.2. Basket Cells The term “basket cell” originated from Ram6n y Cajal’s de-

scription of the cerebellar cortex, where the Purkinje cells are surrounded by a “basket” of axons and terminals. The cell type that gave rise to this plexus was called a “basket cell.” In the hippocampus, the term “basket cell” identifies a population of interneurons with heterogenous afferent connections. The precise distribution of their axonal arbors and neurochemical characteristics (e.g., the expression of calcium binding proteins or neuropeptides; Section IV) also differs. Nevertheless, an important unifying morphological feature of these neurons is the predominant innervation of the perisomatic region (i.e., cell body and most proximal dendrites) of principal cells. Most of the basket cell types were initially observed by Ram6n y Cajal (1893) and described in more detail by Lorente de N6 (1934). However, the true extent of the axonal and dendritic trees and the source of afferent inputs was revealed only recently by intracellular labeling studies and by combined immunocytochemical and anterograde tracing techniques.

Dentate gyms Two types of basket cells can be distinguished on the basis of

significant differences in the laminar distribution of dendritic arbors, which in turn appears to reflect differences in afferent excitatory drive. However, if variations in the location and shape of the cell bodies are also considered, which may eventually prove important, the types of basket cells in the dentate gyms increases to five or six (Ribak and Seress, 1983). Basket cells can be further subdivided by the presence of calcium-binding proteins and neuropeptides (Section IV).

One of the two types to be discussed here is the so-called pyramidal-shaped basket cell (Lorente de N6, 1934), which, as suggested by its name, has a dendriric tree reminiscent of pyramidal cells in shape but is free of spines. Cell bodies of these neurons are located among deep granule cells at the hilar border but may

356 FREUND AND SUZSAKI

occasionally be found just outside stratum granulosum either below or above the layer (Amaral, 1978; Seress and Pokorny, 198 1; Scharfman, 1995a). A prominent apical dendrite emerges from the soma, which may divide proximally, and runs across stratum moleculare to reach the pial surface or the hippocampal fissure (Seress and Pokorny, 1981; Ribak and Seress, 1983; Seress and Ribak, 1983, 1990a,b). Basal dendrites, which may vary in number from two to five, always enter the hilus for a considerable length. Thus, pyramidal basket cells are in a position to be activated in both a feed-forward manner by entorhinal cortex, the commissural-associational path, or even CA3 pyramidal cells and a feedback manner by mossy fiber collaterals (Ribak and Peterson, 1991; Kneisler and Dingledine, 1995a,b). Direct evidence for these afferent connections was provided by combining anterograde tracing and single-cell labeling techniques (Frotscher and Zimmer, 1983a,b; Ribak and Seress, 1983; Seress and Ribak, 1984; Zipp et al., 1989). Variations of this cell type include neurons with an upper and lower bouquet of dendrites, with larger or smaller, multipolar or fusiform cell bodies, or with a soma located in deep stratum moleculare or in the hilus. However, the relative distribution of dendrites in the different input layers of the dentate gyrus is similar for all these variations, which justifies pooling them into one type at this level of analysis.

The distinguishing feature of the other type of basket cell is its dendritic tree, which is largely, if not completely, restricted to the hilus. This type of neuron was described by Ram6n y Cajal (1893) and Lorente de N6 (1934), but in recent Golgi and intracellular labeling studies basket cells with similar dendritic arbors have not been described. Neurons with similar dendritic trees in the hilus have been extensively reported, but their axons ram- i+ in stratum moleculare rather than in the granule cell layer, and therefore should not be called basket cells (Han et al., 1993; Buckmaster and Schwartzkroin, 1995a,b; Sik et al., submitted). A hilar dendritic arbor precludes feed-forward activation by entorhinal afferents. The major excitatory drive to these cells appears to be from mossy fiber collaterals and possibly also from CA3 pyramidal cells.

The axonal arbors of the two basket cell types are indistinguishable. It originates from the soma or a primary dendrite and ascends through the granule cell layer, distributing long horizontal branches within deep stratum moleculare, or within stratum granulosum. These main axon trunks emit a large number of collaterals that descend into the granule cell layer, where they form dense pericellular arrays of synaptic boutons (Ribak and Seress, 1983; Seress and Ribak, 1990a,b; Han et al., 1993; Sik et al., submitted). Unlike in the cerebellum, axons of individual basket cells do not form typical baskets around the somata of their target cells; they only contribute to the pericellular array of boutons. The transverse extent of the axon of an intracellularly injected basket cell (within a 400-pm-thick slice) was estimated to be 900 pm; the density of bouton-laden collaterals within this arbor is ex7 tremely high (Han et al., 1993). A basket cell intracellularly recorded and filled in the dentate hilus in vivo had an axon arbor that covered the entire suprapyramidal blade (longer than 1 mm transverse spread) and extended more than 1.5 mm along the septotemporal axis (Sik et al., submitted). More than 11,000 boutons were estimated to be present along the 44-mm-long axon,

which largely terminated in stratum granulosum (48.6%) and to a lesser extent in the deep portion of stratum moleculare (32.6%) and stratum pyramidale of the CA3c region (18.8%). Axonal branches crossing the upper blade of the dentate gyrus to the CA3c region were free of synaptic varicosities but started to emit varicose collaterals upon entering stratum pyramidale. One main axonal branch was observed to cross from area CA3c to innervate some parts of stratum granulosum of the ventral blade.

Electron microscopy. Axon terminals of basket cells, studied by electron microscopy, contain a large number of synaptic vesicles, one or two mitochondria, and form symmetrical synaptic contacts with the cell bodies, proximal dendrites (Ribak and Seress, 1983; Halasy and Somogyi, 1993b), and occasionally with axon initial segments of granule cells (Halasy and Somogyi, 1993b). Somata of basket cells also receive several symmetrical (GABAergic) synapses, suggesting that granule cells are not the sole targets of basket cells. However, there are two other possible sources for these boutons: the septohippocampal GABAergic pathway (Section V. 1) and a specific class of interneurons (Section 111.5), both of which selectively innervate somata and dendrites of GABAergic cells. Thus, the existence and proportion of interneuronal targets of basket cells remains to be established.

The ultrastructural features of the somata and dendrites of basket cells (Ribak et al., 1978; Ribak and Seress, 1983; Seress and Ribak, 1990a,b) are similar to those of the chandelier cells. The cell bodies are twice as large as granule cells, which is mainly due to the thick perinuclear cytoplasm containing numerous mitochondria and large amounts of rough endoplasmic reticulum. The euchromatic nucleus is invaginated and often contains an intranuclear rod. The soma and the entire dendritic tree are covered by a large number of both symmetrical and asymmetrical synapses. The asymmetrical ones are especially dense on shafts penetrating the hilus. These terminals are packed with round clear vesicles and contain several mitochondria and dense core vesicles, suggesting that most of them may originate from mossy fibers. Direct evidence for a granule cell input to hilar basket cell dendrites is also available (Ribak and Seress, 1983; Seress and Ribak, 1990a,b). The asymmetrical synapses on basket cell dendrites in stratum niolec- dare include those originating from commissural and entorhinal afferents (Zipp et al., 1989; Seress and Ribak, 1990a,b).

Hippocampus Axon arborizations of basket cells show similar features to those

described in the dentate gyrus, but the laminar distribution and orientation of the dendritic tree is more heterogeneous, at least as can be seen on the drawings of Ram& y Cajal (1893) and Lorente de N6 (1934). Recent Golgi and intracellular labeling studies visualized basket cells showing much less variation in soma location or dendritic arborization patterns (Figs. 2, 4. 6). However, this may be partly due to the biased sampling of neurons from strata pyramidale and oriens for intracellular recordings (Kawaguchi and Hama, 1987a,b, 1988; Gulyhs et al., 1993a; McBain et al., 1994; Buhl et al., 1994a; Sik et al., 1995). The predominant dendritic morphology of basket cells in these layers is pyramidal-shaped or bitufted, as in the dentate gyrus, hence


the name pyramidal basket cell. One to three dendrites originate from the apical pole of the triangular or fusiform soma, which then branch proximally, ascend through stratum radiatum, and often penetrate stratum lacunosum-moleculare. Primary basal dendrites are more numerous. They also branch close to the soma and fan out toward the alveus, spanning the entire depth of stratum oriens (Gulyis et al., 1993a; McBain et al., 1994; Buhl et al., 1994a; Sik et al., 1995). All dendrites are spine-free, but occasionally a small number of short spine-like appendages can be observed. Thus, this type of basket cell, both in CA3 and CA1, is likely to receive input from all major sources of excitatory afferents. These include mossy fibers (in the CM region), Schaffer collaterals (Figs. 7, 8), commissural and entorhinal afferents, and recurrent collaterals of local principal cells (in strata radiatum and oriens in CA3, Figs. 6, 8; and in stratum oriens in CA1). Multiple lines of evidence for the existence of these afferent inputs are available from tracing studies (Frotscher and Zimmer, 1983a; Frotscher at al., 1984; Frotscher, 1985, 1989) and from intracellular recordings coupled with afferent stimulation and visualization of the recorded neurons (Gulyis et al., 1993b; Sik et al., 1993, 1995; McBain et al., 1994; Buhl et al., 1994a; Miles et a]., 1996).

According to the drawings of Ram6n y Cajal (1893) and Lorente de N6 (1934), some neurons with a basket type of axon plexus have horizontally oriented dendrites largely restricted to stratum oriens in the CA1 subfield. This finding may imply a predominant input from recurrent collaterals of local pyramidal cells, which selectively arborize in this layer. However, in addition to the axonal arbor in the pyramidal cell layer, some of these cells appear to have a main ascending axon reaching stratum lacunosum-moleculare (Ram6n y Cajal, 1893). There are indeed neurons with similar projections to stratum lacunosum-moleculare and horizontal dendrites in stratum oriens (Gulyis et al., 1993a,b; McBain et al., 1994; Sik et al., 1995), but these do not appear to innervate the perisomatic region of pyramidal cells (Section III.3.2.a). Thus, unless further examples of such cells are visualized with intracellular labeling techniques, basket cells with dendrites that selectively arborize within stratum oriens, described by Ram6n y Cajal and Lorente de N6, will be considered immature (the early Golgi studies were usually done on juvenile animals) or incompletely stained. Cell bodies of some basket cells may be located relatively far from the pyramidal cell layer in stratum radiatum, but their dendrites are oriented radially in a manner similar to basket cells with somata located within or around the pyramidal layer. Such cells were also described by Ram6n y Cajal (1893) and Lorente de N6 (1934) and visualized in recent immunocytochemical studies with antisera against cholecystokinin (Nunzi et al., 1985; Section IV.3.c) or labeled intracellularly (Maccaferri and McBain, 1996). Whether their dendrites extend into stratum oriens in CA1 and whether they are con- victed by recurrent collaterals of local pyramidal cells are currently unknown. We tentatively conclude, however, that most basket Gells in the hippocampus are activated in both feed-forward and feedback manners. The existence of cells with exclusively feed- forward excitation is also likely (e.g., the cholecystokinin [CCKI- containing basket cells in stratum radiatum of CAI; Section IV.3c), but the possibility of selective feedback activation of some basket cells is remote.

The axon arbors of basket cells in the hippocampus also show some variation (Figs. 2, 4, 6). The horizontal extent and laminar distribution of the arbors are similar to those described for chandelier cells. This similarity has occasionally resulted in the mis- classification of chandelier cells as basket cells in the literature. Basket cell axons fill the entire depth of stratum pyramidale and proximal stratum oriens, whereas chandelier cell axonal arbors show some preference for stratum oriens, reflecting the slight difference between the location of cell bodies and axon initial segments of pyramidal cells. An additional difference is that basket cells rarely form characteristic vertical bouton rows typical of axo- axonic cells, although some of the varicose collaterals around pyramidal cell bodies are similar to chandelier terminal segments (Fig. 3).

The axon arbors of basket cells appear to respect the subfield boundaries (i.e., CA1-3), although in most intracellular labeling studies neurons close to these boundaries have not been sampled. There has been only one report of a basket cell (in the guinea pig) with an axon crossing from CA3 to the CAI subfield, providing almost equal number of boutons in both regions (Fig. 2; Gulyis et al., 1993a). The transverse extent of the axon arbor in a 450- p m slice was between 900 and 1,300 p m (Gulyis et al. 1993a; Buhl et al., 1994a). The axon arbor was densest near the soma and became somewhat diffuse distally. The number of boutons generated by this cell was in the range of 3,000 to 10,000. The number of synapses established by a single basket cell on one of its targets varied between 2 and 10 (Figs. 4,5; Gulyis et al., 1993a; Buhl et al., 1994a; Miles et al., 1996). Thus, taking an average of six synapses per connection, one basket cell may innervate 500-1,600 postsynaptic neurons in a 400-pm slice (Miles et al., 1996). In a recent in vivo intracellular labeling study, five basket cells were completely reconstructed in the CA1 region. The total axon length for each cell was between 40 and 50 mm and had 9,000 to 12,000 boutons (Sik et al., 1995). The estimated number of target cells of a single basket cell is 1,500-2,500 pyramidal cells. A basket cell may also contact interneurons. Basket or chandelier cells visualized by immunostatining for parvalbumin (PV; Section IV.2a) were shown to be among the postsynaptic targets (about 1%). Atypical basket cells. In a recent account of interneurons in the CA3 subfield of the guinea pig hippocampus, Gulyis et al. (1993a) reported two neurons described as atypical basket cells. One had an unusually wide axon arborization (wide-axonal basket cell). Although the majority of the collaterals were still in stratum pyramidale, several axons penetrated deep into strata oriens and lucidum. The axon of the other type of atypical basket cell had a stronger bias for stratum lucidum, but again the pyramidal cell layer was heavily innervated (stratum lucidum basket cell). The dendritic trees of these neurons were indistinguishable from those of other basket cells.

Elechon microscopy of the synaptic contacts made by basket cell axon terminals confirmed the predictions of light microscopy, i.e., the majority of the postsynaptic elements were the somata and proximal dendrites of pyramidal cells (Figs. 3,5; Seress and Ribak, 1985, 1990a,b; Gulyis et al., 1993a; Buhl et al., 1994a; Miles et al., 1996). The proportion of postsynaptic dendrites and cell bod-

358 FREUND AND S U Z S k l

FIGURE 5. Correlated light and electron micrographs of synaptic contacts between a basket cell and a pyramidal cell in the CA3 region of the guinea pig hippocampus, both recorded and filled intracellularly in vitro. By using paired intracellular recording, an action potential in the basket cell evoked, on average, a 1.67 mV IPSP in the postsynaptic pyramidal cell (see Fig. 4 for other pairs). A: An axon collateral of the basket cell climbs along the proximal basal

dendrite of the pyramidal cell and has several varicosities contacting it. Two of them (bl, b,) make symmetrical synapses on the dendritic shaft (asterisks in the low power micrographs) in B and D (b,) and in C and E (bz). Arrowhead in A and C indicates a small bouton of the same basket cell collateral, which is in contact with an unstained dendritic shaft. Scale bars = 5 pm in A, 0.5 pm in B,C, 0.2 pm in D,E.

FIGURE 4. A,B: Postsynaptic potentials evoked by interneurons that terminate in the perisomatic (A) or in the dendritic region (B) of pyramidal cells, as revealed by paired intracellular recording. An action potential (upper traces) triggered by intracellular current pulse in the interneurons evoked IPSPs in the postsynaptic pyramidal cells (lower traces). IPSPs at the somatic location had a larger amplitude and faster time to peak than did the dendritic IPSPs. The number and location of synaptic contacts responsible for the electrical interaction were identified by light and electron microscopy, following intracellular biocytin filling. C-E: Reconstructions of the

cell pairs (pyramidal cells are drawn in red, interneurons in black) made by a drawing tube. The basket cell in D that produced the IPSP shown in A formed three synaptic contacts on the postsynaptic pyramidal cell, as shown by arrows at higher magnification in C. The interneuron in E, responsible for the IPSP shown in B, arborized in stratum oriens and radiatum and is a bistratified cell. It formed two synapses on the apical and three on the basal dendrites of the pyramidal cell (arrows). All contacts identified at the light microscopic level were confirmed by correlated electron microscopy (see Figs. 5, 8). Scale bars in D,E = 50 pm.

359


FIGURE 6. A: Camera lucida reconstruction of the axonal trees of two presynaptic pyramidal cells (black and blue) and a postsynaptic basket cell (red) simultaneously recorded and filled with biocytin in the CA3 region of the hippocampus. The axon of both pyramidal cells established a single contact on a mid-distal dendrite of the basket cell in stratum radiatum (arrows). Correlated light and

ies varied considerably. Typically, somatic contacts accounted for 30-70% of the postsynaptic elements. The innervated dendrites were mostly proximal primary branches, which may be considered electrotonically equivalent to the soma. A small number of axon initial segments and spines have also been reported to receive input from basket cells in the hippocampus (Seress and Ribak, 1990a,b). The wide-zonal and the stratum lucidum basket cells in CA3 innervated a considerably larger proportion of dendritic shafts (60-80%), some of which were considered to be distal dendrites (Gulyis et al., 1993a). As in the dentate gyrus, the existence and proportion of nonpyramidal targets of basket cells in the hippocampus still need to be investigated.

Ultrastructural features of basket cell bodies and dendrites are very similar to those described for chandelier cells and dentate basket cells. The large number of asymmetrical synapses on the dendrites and somata originate from local collaterals of pyramidal cells (Fig. 8; Gulyis et d., 1993b; Sik et al., 1993; Buhl et al., 1994a), from mossy fibers in the CA3 region (Frotscher, 1985), from Schaffer collaterals (Fig. 8), commissural fibers (Frotscher and Zimmer, 1983a; Seress and Ribak, 1985; Sik et al., 1993; Deller et al., 1994), and from entorhinal afferents (Kiss et al., 1996). Thus, basket cells in the hippocampus are activated in both feed-forward and feedback manners.

Paired intracellular recordings and subsequent biocytin filling of the connected neurons showed that pyramidal cells form single synapses with each of their interneuron targets (Figs. 6, 8;

electron micrographs of both contacts are shown in Figure 8. B: Synaptic transmission between one of the filled pyramidal cells (1) and the postsynaptic basket cell (2). Single pyramidal cell action potentials evoked variable amplitude EPSPs in the inhibitory cell and occasionally were associated with transmission failures (second trace). Data from Gulyds et al. (1993b). Scale bar = 200 pm.

Gulyis et al., 1993b). This finding has been confirmed by in vivo intracellular labeling of a CA3 pyramidal cell (Fig. 7) reconstructed from serial Vibratome sections that were double stained for PV, a calcium-binding protein present in basket and axo-axonic cells. The intracellularly filled pyramidal cell contacted 220 PV-positive cells, 85% of them via a single bouton (Fig. 8). In 12% of the pyramidal cell-PV cell connections, two boutons were involved, and in 3% of the connections, three boutons were involved (Sik et al., 1993). These data suggest that the more than 2,000 asymmetrical synapses on the basket cell dendritic tree are likely to originate from more than 2,000 pyramidal cells. Thus, both convergence and divergence is extremely high in the pattern of efferent connectivity of pyramidal cells to interneurons (Gulyis et al., 1993b; Sik et al., 1993).

111.3. Interneurons Innervating Principal Cell Dendrites (Dendritic Inhibitory Cells)

Interneurons that synapse on principal cell dendrites (dendritic inhibitory cells) show great variability in location of their soma and the pattern of their dendritic and axonal arborizations. The unifying feature that distinguishes them from basket and chandelier cells is the predominant, or for some cell types exclusive, innervation of different segments of the pyramidal or granule cell dendritic tree. Several of these cells are also specialized to terminate in conjunction with specific sets of excitatory afferents in


well-defined laminae (Figs. 9, 11, 12, 17, 18; Han et al., 1993; Gulyk et al., 1993a,b; Sik et al., 1995, submitted). Further di- vision is introduced by the specific laminar distribution of the interneuron dendrites, allowing their selective activation by a single pathway or a combination of afferent pathways.

111.3.2. Dentate gyms The dendritic arbors of practically all major cell types known to-

day from the category of “dendritic inhibitory cells” have been described by Rarn6n y Cajal (1893) and Amaral (1978). Their drawings, based on Golgi impregnation, revealed only portions of the axonal arbors. Recent intracellular labeling studies, however, have provided a nearly complete picture of the processes (Han et al., 1993; Buckmaster and Schwartzkroin, 1995a,b; Sik et al., submitted). The following classification is based on the distinct dendritic and axonal distributions presented in these more recent studies.

III3.l.a. Interneurons with hihr dendrites and ascending axons (HIPP cellj). The most charactersitic type of cells innervating granule cell dendrites has a dendritic tree limited to the hilar region and an extensive axon arborization in the outer two-thirds of the molecular layer (Fig. 9; HIPP, hilar perforant path-associated cell of Han et al., 1993). The dendrites emerge from a hsiform cell body located subjacent to the granule cell layer and branch prohsely. They are covered with long thin spines, almost as dense as mossy cells in this region (Sik et al., submitted). Thorny excrescenses never occur on the soma and dendrites of the HIPP cell, thus providing another distinguishing feature from mossy cells (Amaral, 1978). Thus, the dendritic tree perfectly coincides with the distribution of mossy fiber collaterals and avoids the dentate molecular layer where entorhinal and commissural-associational afferents terminate.

The axon of HIPP cells originates from the soma, and several main axon trunks cross stratum granulosum giving rise to long

.... . . . . .......... ......

. . . .

. . . .

........ :I:::;:: >::>:: . . . . . . . . . . . . ........

FIGURE 7. A: The axond arborization of a Neurobiotin-labeled CA3 pyramidal cell reconstructed from the whole rat hippocampus. The axon spread through stratum radiatum and oriens of 32 (60 pm) longitudinal sections, mostly arborizing in the C AI subfield. The axon had over 15,000 boutons. The same sections were double striaeed for PV. Of the 324 boutons in contact with PV-positive cells, 85% made single contacts (see Fig. 8E,F), 12% made double contacts, and 3% made triple contacts. B: Top view of the bouton & tribution of the labeled pyramidal cell shown in A, obtained by computer reconstruction aod rotation of 32 sections. The darkness of the rectangles are proportional to the number of boutons found within that area. Three patches of boutons appear to emerge from

.... ............ . . . . :;>:;;:: . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

.... .... .... . . . . . . . . ............ .. ........

.................. .....

.....

C

A

I ....

. . . . . . . . . . . . . . . . .........

the plot, one in CA3 close to the soma, and two others in CAI, about 600-800 pm firom the soma. The broken lines indicate the approximate border region between the CAl and CA3 regions, and the closed circle the position of the soma. C: Distribution of the contacts established by the labeled pyramidal cell on PV-immunoreactive interneurons, reconstructed and viewed as in B. Note that the locations and peak densities coincide with those in B, i.e., with the peaks in total bouton number. Scale bar = 200 pm in A. Reproduced from Sllr et al., (1993) “Complete ason arborization od a single CA3 pyramidal cell in the hippocampus.” Eur J Neurosci, 51719-1728 by permission of Oxford University Press.


horizontal collaterals with several fine branches that cover the entire width of the outer two thirds of the molecular layer (Fig. 9). Occasional collaterals can be followed across the hippocampal fissure, where they form small local arbors in stratum lacunosum- moleculare of the CA1 region and in the subiculum (Han et al., 1993; Buckmaster and Schwartzkroin, 1995a,b). The collaterals are studded with a large number of small to medium-sized (0.2-1 pm) axon terminals, resulting in a density of 34 ? 6/1OO p m (Sik et al., submitted). Four cells, completely visualized by intracellular biocytin injections in vivo, have been reported to date, three in the rat (Buckmaster and Schwartzkroin, 1995a; Sik et al., submitted) and one in the gerbil (Buckmaster and Schwartzkroin, 1995b), thus allowing a detailed description of the three-dimensional extension of the axon arbor. Each of the four cells arborized in the outer two-thirds of the molecular layer, almost completely covering both the upper and lower blades of the dentate gyrus in each of several coronal sections. A HIPP cell reconstructed from a 400-pm-thick slice had an axonal plexus that occupied the entire suprapyramidal and half (toward the apex) of the infrapyra- midal blade (Han et al., 1993). Two cells appeared to have a bias toward the outer one-third (Buckmaster and Schwartzkroin, 1995a; Sik et al., submitted), where the lateral entorhinal afferents terminate, and another toward the middle one-third of the molecular layer, where the medial entorhinal afferents terminate. Several collaterals were also present in the inner one-third (Buckmaster and Schwartzkroin, 1995a). However, other examples of these cells have an almost equal innervation of the termination layers of the medial and lateral entorhinal cortex (Han et a]., 1993; Sik et al., submitted), with a minor proportion of the axon in the inner molecular layer. The septotemporal extent of the axon collaterals was also remarkably long. The cell reported by Sik et al. (submitted) was reconstructed from 39 coronal sections (80 p m thick), which corresponds to 3.1 mm in the septotemporal direction. Those cells described by Buckmaster and Schwartzkroin (1995a) were reconstructed from 3.5-mm- and 1.7-mm-thick tissue slabs. With regard to the number of axon terminals, the estimated two-dimensional axon length was 226 mm and 291 mm, respectively, for the two best-filled neurons,

FIGURE 8. Correlated light and electron micrographs of synaptic contacts established by two presynaptic pyramidal cells onto a basket cell. The cells were intracellularly recorded, the electrical interaction characterized, and the cells were subsequently filled with biocytin (see Fig. 6). A spiny dendrite of the pyramidal cell is visible in A (on the right), and the basket cell dendrite (left) is smooth. Each pyramidal cell established a single contact on the dendrites of the basket cell in stratum radiatum (A,C), and at the electron microscopic level both contacts (bl, bz) were found to be conventional synapses (large arrows in B and D). The pre- and postsynaptic membranes are indicated by white arrows in B and D and show a characteristic widening of the synaptic cleft. E,F: An intracellularly filled CX3 pyramidal cell (the reconstruction is shown in Fig. 7) established mostly single synapses on PV-containing interneurons visualized in the same sections. At the light microscopic level (in E) a vari- case pyramidal cell axon formed one bouton (b) on a PV-positive dendrite (dz). This contact was shown to be an asymmetrical synapse (arrow in F) in the electron microscope. Scale bars = 5 pm in A,C,E, 0.5 pm in B,D, 0.2 pm in F.

which corresponds to a total of 76,800 and 98,900 axon terminals, if one assumes an average density of 34 boutons per 100 p,m (Sik et al., submitted).

Neuropeptide Y (NPY) was shown to be contained in one of these neurons (Sik et al., submitted), but on the basis of axonal and dendritic patterns, they are likely to contain somatostatin as well (Sections IV.3.a, IV.3.b).

At the electron microscopic level, boutons of HIPP cells form symmetrical synapses, even at their nonvaricose segments, with spiny dendritic shafts and spines of presumed granule cells (Fig. 10; Halasy and Somogyi, 1993b). Occasionally, they also form symmetrical synapses on other interneurons (Sik et al., submitted). The same spines frequently receive an additional synapse of the asymmetrical type, most likely from entorhinal afferents. In the hilus, the dendrites and long thin spines of this interneuron type are covered by numerous asymmetrical synapses, most of which showed characteristic features of mossy fiber terminals (Halasy and Somogyi, 1993b). In summary, the HIPP cells are likely to mediate largely feedback inhibition of granule cells by terminating on distal granule cell dendrites and spines in conjunction with entorhinal afferents (Halasy and Somogyi, 1993b; Han et al., 1993; Sik et a]., submitted).

IIL3.I.b. Hilar border neurons with ascending axom and dendrites ( H I W cells). Han et al. (1993) named this cell type hilar commissural-associational pathway related (HICAP), indicating their major distinguishing features, i.e., the predominant innervation of the inner molecular layer (Fig. 9). The triangular cell body of these neurons is located in the polymorphic zone of the hilus or within stratum granulosum at the hilar border (Han et al., 1993; Soriano and Frotscher, 1993a; Buckmaster and Schwartzkroin, 199513; Sik et al., submitted). The primary dendrites have a multipolar origin and give rise to several smooth or sparsely spiny branches that invade both the hilus and the rnole- cular layer, with a small bias for the molecular layer in terms of total dendritic length.

The main axon runs in the hilus, and gives rise to ascending collaterals that branch in a Y-shaped manner in stratum moleculare, just above the granule cell layer. The dense axon arbor possesses a large number of boutons that are generally confined to the inner one-third of the molecular layer (Han et al., 1993; Buckmaster and Schwartzkroin, 1995b). However, the HICAP cell reported by Sfk et al. (submitted) had a significant number of collaterals within stratum granulosum (22.6%), near the parent cell body. In a frontal section through the center of the arbor, the axon occupies approximately half of the dentate gyrus, either the upper or the lower blade. The axon of the cell injected in vivo in the gerbil had the richest arbor near the soma and spread over 1,500 ,urn in the longitudinal direction, reaching 22% of the total length of the hippocampus (Buckmaster and Schwartzkroin, 1995b). The axon of the other cell (from the rat) spanned 2.6 mrn in the septotemporal direction (Sik et al., submitted) and was particularly dense around the parent cell body and gradually diminished toward the mediolateral and the septotemporal edges of the arbor. The total two-dimensional axonal length of this cell was 91 rnrn, which corresponded to 26,300

lNTERNEURONS OF THE HlPPOCAMPUS 365

axon terminals when calculated on the basis of a bouton density of 29 boutons per 100 p m axonal length. In Golgi impregnated material, Soriano and Frotscher (1993a) described neurons with a similar dendritic tree, but the axonal plexus expanded to innervate the outer molecular layer.

Apart from a few exceptions, axon terminals of the HICAP cells establish symmetrical synapses mostly with large caliber spiny dendrites and less often with small dendrites and dendritic spines (Fig. 10; Halasy and Somogyi, 199315). The vast majority of postsynaptic dendrites appears to belong to granule cells on the basis of ultrastructural features, the presence of spines (Soriano and Frotscher, 1993a), and the lack of GABA immunoreactivity (Halasy and Somogyi, 1993b); occasional nonprincipal cells were also among the targets (Sik et al., submitted). The axon terminals are densely packed, with clear pleomorphic vesicles and contain one or two mitochondria. The dendrites of HICAP cells receive a large number of symmetrical and asymmetrical synapses in all layers. Thus, they appear to be involved in both feed-forward and feedback circuits and innervate dendrites of granule cells largely in conjunction with commissural-associational afferents (Han et al., 1993; Sik et al., submitted).

III.3.l.c. Neurons with axons and dendrites in stratum moIecu- hre (MOM cells). The axonal and dendritic trees of this cell type are largely limited to the outer two-thirds of the dentate molecular layer, and thus was named molecular layer perforant path- associated cell (MOPP cell) by Han et al. (1993). Smooth dendrites that originate from a soma in deep stratum moleculare ascend to reach the hippocampal fissure and span an area over 800 p m in transverse length. The axon of the MOPP cell is even more extensive. The majority of collaterals run perpendicular to granule cell dendrites and arborize profusely in a terminal cloud that extends beyond the upper blade of the dentate gyrus (Fig. 9). The number of varicosities along the fine terminal branches of the MOPP cell is somewhat fewer than that for the two types of hilar interneurons described above.

Electron microscopy demonstrated that axon terminals of this cell type form symmetrical synapses mostly with spiny distal dendrites of granule cells (Fig. 10; Halasy and Somogyi, 1993b). The dendrites of these interneurons were densely covered by both symmetrical and asymmetrical synapses. Thus, this interneuron type is likely to be driven in a feed-fonvard manner, largely by en-

FIGURE 9. Costratification of the axons of three distinct types of GABAergic neuron with the major glutamatergic pathways to the dentate gyrus. The hilar neuron, which has its axon distributed in the perforant path termination zone (HIPP cell), and a molecular layer cell, which also has its axon arborizing in the perforant path termination zone (MOPP cell), have terminals that avoid the inner one-third of the molecular layer. In contrast, the hilar neuron, which has its axon distributed in the commissurallassociational pathway termination zone (HICAP cell), has terminals that avoid the outer two-thirds of the molecular layer. GABA immunoreactivity was demonstrated in the terminals of each of these three cell types. These cells were recorded intracellularly in vitro and visualized with bio- +n. Results are based on Han et al. (1993), Halasy and Somogyi (1993, 1996a,b), and Halasy et al. (1996). This figure was kindly prepared by Peter Somogyi. Scale bar = 100 pm.

torhinal afferents, and innervates granule cell dendrites in conjunction with the same pathway (Halasy and Somogyi, 199313).

1II.3.1.d. H i h r neurons with aprojection to the h+pocampus and subiculum. Two cells with these characteristics have been described recently, one in the rat (Sik et al., submitted) and another in the gerbil hippocampus (Buckmaster and Schwartzkroin, 1995b) by intracellular injection. The soma of both cells are located at the border of the hilus and stratum radiatum of the CA3c subfield. Their dendrites leave the soma in all directions in a stellatelike fashion to penetrate the deep hilus and stratum moleculare of the dentate gyrus and stratum radiatum of the CA3c region. Apart from a few collaterals that enter the hilus and stratum moleculare, the majority of axon collaterals follows the path of mossy fibers to subsequently innervate strata radiatum, lucidum, oriens, and pyramidale of the CA3 subfield. Thus, because of their dendritic distribution and afferent input, it is difficult to determine whether these neurons should be considered a part of the dentate gyrus or of the hippocampus. Their output, however, is clearly associated with the CA3 region. The total axon length of this cell in the rat is about 100 mm, with 28,000 boutons, and has a septotemporal extension of 4.3 mm (Sik et al., submitted). In this cell, one main branch could be followed to the subiculum, but varicose collaterals were not seen to originate at this distant site. The neuron labeled in the gerbil also had numerous collaterals in stratum radiatum of the CA1 region (Buckmaster and Schwartzkroin, 1995b). Electron microscopic studies have not been reported; therefore, the types of synapses and postsynaptic elements of this neuron type are still unknown. According to the laminar distribution of its axon, this cell type may be considered a CA3c-specific variety of bistratified or trilaminar cell, described elsewhere in the hippocampus (Section III.3.2.b).

III.3.1.e. H i h r neurons with unknown projection. There are interneurons in the hilus with a great variability in dendritic morphology as revealed by Golgi and immunocytochemical studies (Amaral, 1978; Gulyis et al., 1992). Due to the lack of axonal staining, these cells cannot yet be reliably classified because, as shown above, cells with different dendritic morphology may belong to the same type if classified on the basis of postsynaptic elements (e.g., basket cells). Nevertheless, there is a cell type with unique dendritic features, which deserves mention. Its dendrites are limited to fields densely innervated by mossy fibers, i.e., to the hilus of the dentate gyrus and to stratum lucidum of CA3 (Section IV.2.c). They have been previously described in Golgi preparations (Amaral, 1978; Soriano and Frotscher, 1993b) and visualized in large numbers by immunostaining for the calcium- binding protein, calretinin ( C R GulyPs et al., 1992) and heat shock protein after ischemia (Hsu and Buzsiki, 1993). Amaral (1 978) named these cells “long-spined multipolar neurons” according to the characteristic hairlike shape of the spines covering their dendrites and cell bodies. The entire dendritic tree receives abundant asymmetrical synapses from terminals or thin preter- mind segments of axons displaying ultrastructural features of mossy fibers (Gulyds et al., 1992; Soriano and Frotscher, 1993b). The dendritic spines of these cells penetrate into bundles of mossy


FIGURE 10. Immunocytochemical demonstration of GABA in boutons of dentate gyrus interneurons that innervate stratum moleculare. The electron micrographs show biocytin-filled boutons of intracellularly recorded HIPP (A-C), MOPP (D,E), and HICAP (F,G) cells. A, D, and F show that the boutons establish type two synapses (arrowheads) with the dendritic s h a h (d) of granule cells. C, E, and G show boutons in sections that were immunoreacted for GABA by the postembedding silver-intensified immunogold method. The high density of electron-dense immunoparticles over the bio-

cytin-labeled boutons demonstrates their high GABA content. B and C are adjacent sections. Open arrows point at boutons immunonegative for GABA, and unidentified GABA-immunoreactive boutons are marked by asterisks. The peroxidase reaction product is less electron dense in the immunoreacted sections due to the removal of osmium. Data based on Halasy and Somogyi (1993a,b) and on unpublished results for the HIPP cell. This figure was kindly prepared by Katalin Halasy. Scale bars = 0.2 pm for A-G. B,C and D-G are at the same magnification.

fibers, where each spine may receive as many as six synapses from small-caliber preterminal mossy fibers (Gulyis et al., 1992). The hilar variety of these cells are similar to neurons that project to the outer molecular layer (HIPP cells), which also have spiny dendrites limited to the hilus (Han et al., 1993; Buckmaster and Schwartzkroin, 1995a,b). However, the axons oflong-spined mul-

tipolar cells have never been visualized by Golgi or immunostaining techniques, probably due to myelination. This suggests that these long-spined multipolar neurons either belong to a novel class of hilar interneurons or, according to combined immunocytochemical and tracing studies (Seress and Ribak, 198.3; Bakst et al., 1986; Miettinen et al., 1992; Deller et al., 1995b; Sections


FIGURE 11. Reconstructions of 0-LM cells intracellularly recorded and filled in vivo in the CAI region (A) and in vitro in the C43 region (B). The axon of this cell type arborizes primarily in stratum lacunosum-moleculare in their respective regions and also have a minor projection to stratum oriens. However, the dendritic trees show a different laminar specificity. Dendrites of the CA1 0-

LM cell are horizontally oriented and are restricted to stratum oriens, whereas the dendrites of the 0-LM cell in CA3 span all layers except stratum lacunosum-moleculare. The laminar distribution of the dendrites in both subfields overlaps that of recurrent collaterals of local pyramidal cells. The nonvaricose main axon trunks are not drawn for the CAI cell. Scale bars = 100 pm in A, 50 pm in B.

IV and VI for details), may represent a commissurally projecting variety of HIPP neurons. The CA3 variety of long-spined multipolar cells may send its axon to stratum radiatum (M. Frotscher, personal communication) or to stratum lacunosum-moleculare of the CA3 region and back to the outer stratum moleculare of the dentate gyrus (C. McBain, personal communication). For more details, see Section IV.2.c.

111.3.2. Hippocampus With respect to the relative distribution of dendritic and ax-

onal arbors as compared with the laminar pattern of major excitatory afferent pathways, the cell types innervating principal cell dendrites in the hippocampus are very similar to those described in the dentate gyrus. The corresponding types are therefore described in the same sequential order.

III.3.2.a. Cells terminating in conjunction with entorhinal afferents (0-LM ceZls). The major distinguishing features of these neurons, as described by Ram6n y Cajal (1893) and Lorente de N6 (1934), are a dense axon arbor restricted to stratum lacunosum-moleculare and a dendritic tree localized to layers occupied by recurrent collaterals of local principal cells. In this respect, they .Ire similar to the HIPP cells of the dentate gyrus (see above), whose s o n collaterals also terminate in conjunction with perforant path afferents and whose dendrites, which are located exclusively within the hilus, are innervated by mossy fiber collaterals. Further similarity is found in their colocalization of neuropeptides because both HIPP and 0-LM cells contain somatostatin, and some of them

also contain NPY (Sections IV.3.a, IV.3.b). The axons of 0 -LM cells ascend from the soma directly to stratum lacunosum-molecdare to form a dense cloud of fine varicose collaterals (Fig. 11; Gulyh et al., 1993a,b; McBain et al., 1994; Sik et al., 1995). Occasional branches are also directed toward stratum oriens. The axon arbor in stratum lacunosum-moleculare is highly focused, having a limited transverse and longitudinal spread. In 400-pm- thick slices, the arbor extends 300%400 p m in contrast with the same cell type in the dentate gyrus, which covers more than two- thirds of the entire transverse length of the molecular layer in slices of similar thickness (Han et al., 1993). The axons of HIPP cells intracellularly filled in vivo were shown to innervate the entire upper and lower blades of the dentate gyrus (Sik et al., submitted). There is a single example in the literature of an in vivo filled 0- LM cell with a completely reconstructed axonal field. The dendrites of this cell were confined to the stratum oriens-alveus border. Its axon gave rise to a large dense cloud in stratum lacunosum-moleculare (91 S%), whereas a smaller arbor was distributed within stratum oriens (7%) (Fig. 11; Sik et al., 1995). Importantly, the size of the axon arbor was not significantly larger than those reconstructed from 400-pm-thick slices. The axon arbor had a central core of approximately 500 p m in diameter, which gradually diminished for another 100-200 p m in the septa1 and temporal directions. The total two-dimensional axon length was 63.4 mm, and the calculated number of boutons, based on a density of 26.6 boutons per 100 pm, was 16,800.

The dendritic tree of 0-LM cells located in the CA1 and CA3 regions is different, perhaps reflecting the difference in the par-


\ '

\ \


tern of recurrent collaterals of pyramidal cells in these two subfields. In the CAI region, where local collaterals of the pyramidal cells are limited to deep stratum oriens, the dendrites of 0- LM cells are also restricted to this zone. The dendrites originate from an oval soma in stratum oriens and extend in parallel with [he alveus for several hundred micrometers (McBain et al., 1994). Even the stratum-pyramidale-directed primary dendrites curve back toward the alvear border, suggesting a preference for inputs characteristic of this sublayer. In the CA3 region, where recurrent collaterals are found in all layers except stratum lacunosum- moleculare, the dendritic trees are not limited to stratum oriens, although such cells can also be observed (Gulyis et al., 1993a,b). The more typical variety has a pyramidal-shaped soma and ascending and descending dendrites spanning all strata except lacunosum-moleculare (Fig. 11; Gulyis et al., 1993a). These characteristic differences in dendritic arbors between CAI and CA3 0-LM cells obtained from single cell labeling studies have been confirmed by immunocytochemical data for somatostatin visualizing large numbers of neurons of these types (Section IV.3.a).

Electron microscopy reveals that axon terminals of this cell type form symmetrical synapses, with distal dendrites and dendritic spines of presumed pyramidal neurons (Gulya’s et al., 1993a; Sik et al., 1995). The innervated spines often also receive an asymmetrical synapse from an unlabeled terminal. The synaptic boutons of 0-LM axons are filled with clear pleomorphic, occasionally large dense-core vesicles; and, in contrast to basket and chandelier cell terminals, these boutons rarely contain mitochondria.

The dendritic tree is covered by numerous symmetrical and asymmetrical synapses. The majority of asymmetrical synapses were shown, by using ischemia to damage CAI pyramidal cells selectively, originate from recurrent collaterals of local pyramidal cells (Blasco-Ibanez and Freund, 1995). Similar to the HIPP cells of dentate gyrus, we regard this cell type as an interneuron that is primarily involved in feedback circuits and that innervates distal principal cell dendrites in conjunction with entorhinal afferents. Importantly, there is a striking correlation between the spatial ex-

FIGURE 12. A: A bistratified cell in the CA3 region intracellularly filled with biocytin in a 400-pm-thick guinea pig slice in vitro and reconstructed from serial 60-pm Vibratome sections. The dendritic tree is bitufted and does not reach stratum lacunosum-moleculare. The axon spans all layers except stratum lacunosum-moleculare, but varicose collaterals are relatively sparse in strata pyramidale and lucidum. The arbor is broad in the tangential direction and limited to the proximal half of the layer in stratum radiatum. The in- serts on the top show the area occupied by the axon in the slice and (on the right) that the bouton density along the axon collaterals is high in strata oriens and radiatum but low in strata pyramidale and lucidum. From unpublished work of R. Miles, N. Hajos, A.I. Gulyds, K.T6th, and T.F. Freund. B: A backprojection neuron in CA1 intracellularly recorded and filled with biocytin in vivo in the rat. The dendrites remain in stratum oriens, but the axon, in addition to its local collaterals in CAI, forms large arbors in CA3 stratum radiatum and in the hilar region of the dentate gyrus. The insert shows the sep totemporal distribution of the axon. The axon length/bm is indicated on the ordinate, showing a peak near the level of the soma. Reprinted with permission from Silt A, Penttonen M, Ylinen A, Bum& G (1994) “Inhibitory CAI-CA3-hilar region feedback in the h i p pocampus.” Science 265:1722-1724. American Association for the Advancement of Science. (1994). Scale bars = 100 pm.

tent of entorhinal afferents and the axon arbors of single 0-LM and HIPP cells in the hippocampus and the dentate gyrus. Anterograde labeling of a small group of entorhinal cells (Tamamalu and Nojyo, 1995) or a single-layer I1 neuron (Tamamaki and Nojyo, 1993) gives rise to axonal staining throughout the dorsal and ventral blades of the dentate gyrus. In contrast, the diameter of the terminal cluster in stratum lacunosum-moleculare of the CA1 region is smaller than 1 mm. Thus, it appears that the spatial extent of the axon arbors of single 0-LM neurons in the CA1 region and of HIPP cells in the dentate gyrus closely resemble the size of the terminal fields of perforant path afferents in these regions. The functional significance of this hypothesized match is still unknown but may be related to inhibitory flanking in the CA1 region, carv- ing out slabs of pyramidal cells with similar representations of entorhinal-cortex-mediated neocortical information (Cohen and Eichenbaum, 1993; Buzsiki et al., 1995).

lII.3.2.b. Interneurons innervating pyramidal cell dendrites in strata radiatum and oriens (bistratiJied and trihminar cells). The first descriptions of bistratified and horizontal trilaminar cells were given by Buhl et al. (19943 and Sik et al. (1995). Cell bodies of these neurons are localized within or near to stratum pyramidale or at the stratum oriens-alveus border (Fig. 12). The dendritic trees of bistratified cells are spine free, often varicose, and multipolar. Several primary dendrites bifurcate close to the soma. These dendrites have a predominantly radial orientation but frequently also emit horizontal branches. In contrast with the dendritic tree of basket and chandelier cells, the vertical branches of bistratified cells do not reach stratum lacunosum-moleculare (Miles et al., 1996; Halasy et al., 1996).

The dendrites of the horizontal trilaminar cell run horizontally at the stratum oriens-alveus border, extending over several hundred micrometers in the septotemporal and transverse directions. This cell has a large soma, with prominent primary dendrites, and occasional spines on more distal segments (Sik et al., 1995). Radial trilaminar cells are very similar to bistratified cells with regard to their axonal and dendritic trees, but there are also some consistent differences. In contrast with bistratified cells, dendrites of radial trilaminar cells often penetrate stratum lacunosum-moleculare (A.I. Gulyk and R. Miles, unpublished observations). This is especially true for those cells located in distal stratum radiatum (i.e., portion of stratum radiatum closer to stratum lacunosum-moleculare) .

The axon of bistratified cells forms a dense arbor of fine varicose collaterals both above and below the pyramidal cell layer, occupying the entire width of stratum oriens and a zone of 200-300 p m of “proximal” stratum radiatum (i.e., toward the pyramidal cell layer) in the CA1 and CA3 regions (Buhl et al., 1994a; Sik et al., 1995; Miles et al., 1996). The transverse spread of the axon was found to be between 800 and 1,100 p m in cells filled in vitro, whereas in the only example of a bistratified cell filled in vivo the axonal arbor was approximately 2 mm in diameter in the transverse and septotemporal directions (Sik et al., 1995). The two-dimensional axon length of this cell was 79 mm, carrying a total of 16,600 synaptic varicosities. Immunocytochemical double-staining techniques revealed that this cell type contained the calcium-binding protein calbindin (see Section IV.2.b).


Horizontal trilaminar neurons have a similar axon arbor, but unlike the axon arbor of bistratified cells, a considerable number of varicose collaterals also distribute within stratum pyramidale (16.7%). In addition, the distribution of the trilaminar cell axon is more strongly biased toward proximal stratum radiatum (68.4%). Another important difference between bistratified and trilaminar cells is that trilaminar cells can be antidromically activated by stimulating the fimbria (Sik et al., 1995, submitted). This activation demonstrates that trilaminar cells project outside of the hippocampus, possibly to the medial septum (T6th and Freund, 1992). The axon extended 2.6 mm in the septotemporal and 2.45 mm in the mediolateral direction. The total two-dimensional axon length amounted to 56 mm and possessed a calculated 15,800 axon terminals.

The axon arbor of radial trilaminar cells is sparse and consists mainly of long, radially running collaterals, which are uniformly varicose along their entire length, regardless of the layer they cross (A.I. Gulyis and R. Miles, unpublished observations). This fact suggests that a considerable proportion of the postsynaptic elements are likely to be cell bodies in stratum pyramidale, unlike bistratified cells.

Electron microscopy of a large sample of axon terminals demonstrated that bistratified cells established symmetrical synaptic contacts with proximal dendrites (79%) and dendritic spines (17%) of pyramidal cells (Halasy et al., 1996). In contrast, cell bodies (4%), mostly of pyramidal neurons, were rarely innervated. Axon terminals of bistratified cells were found to be significantly smaller than basket cell boutons in two independent studies, one measuring cross-sectional area at the synaptic junction in the electron microscope (Halasy et al., 1996), the other using camera lucida at the light microscopic level (Miles et al., 1996). The absolute values calculated in the two studies are different, but both demonstrate that basket cell boutons, on average, are more than two times larger in cross-sectional area than boutons of bistratified cells. The latter often lack mitochondria, whereas the basket cell boutons usually contain one or even two mitochondria and appear to have somewhat longer synaptic active zones (Miles et al., 1996). These differences suggest that the probability of release is higher at perisomatic synapses, which is supported by findings that most, if not all, spontaneous inhibitory postsynaptic potentials (IPSPs) recorded from CA1 pyramidal cells in vitro are of perisomatic origin (Miles et al., 1996). In paired intracellular recording and labeling experiments in vitro, the three bistratified cells reported to date were shown to establish five, six, and nine electron microscopically verified synaptic contacts with their target pyramidal cell (Buhl et al., 1994a; Miles et al., 1996). With an average of 6.7 contacts per target cell, a bistratified neuron (with 16,600 boutons calculated by reconstructing an in vivo injected cell; Sik et al., 1995) may innervate approximately 2,500 pyramidal cells. Interestingly, the contacts were distributed on different branches of the postsynaptic pyramidal cell dendritic tree, which suggests that these synapses are not designed to “ampu- tate” any dendritic branch (Miles et al., 1996). The trilaminar cell also made symmetrical synapses predominantly with dendritic shafts of unidentified origin (Sik et al., 1995).

The laminar distribution of the dendritic tree of bistratified

and trilaminar neurons enables them to receive input from commissural-associational fibers (Schaffer collaterals) and from local recurrent collaterals and may therefore be driven in feed-forward and feedback manners. However, the dendrites of the horizontal trilaminar cell are confined to the stratum oriens-alveus region, suggesting that axons restricted to this zone (i.e., the recurrent collaterals of CA1 pyramidal cells) may represent their major source of excitatory input, as shown for the 0 -LM cells (Blasco- Ibanez and Freund, 1995). The laminar distribution of the bistratified cell axon coincides with Schaffer collateral and commissural-associational input, as both may span the entire width of strata radiatum and oriens. There are examples of bistratified cells having axons limited to stratum oriens and the proximal one-third or half of stratum radiatum (Buhl et al., 1994a; Halasy et al., 1996). In addition to a projection to stratum oriens, Schaffer collaterals arising from CA3a also terminate in inner stratum radiatum near the CA2 border. Those originating from C A ~ C , however, innervate the entire stratum radiatum farther away from the CA2 border (Ishizuka et d., 1990). Thus, if bistratified cell axons indeed follow the distribution of Schaffer collaterals, one would expect that CA1 bistratified cells located closer to the CA2 border would predominantly arborize in proximal stratum radiatum, whereas those located near the subicular pole would span the entire stratum radiatum. This possible correlation has not been investigated so far. The axon arbor of the trilaminar cell shows no perfect overlap with the laminar distribution of any known afferent pathways. It emits relatively large numbers of bouton-laden collaterals to stratum pyramidale; in stratum radiatum, its axons are limited to the proximal half of the layer. Another possible reason for this laminar specificity of both bistratified and trilaminar cell axons is an overlap with the distribution of voltage-gated calcium channels located along pyramidal cell apical dendri tees. These calcium channels may have a role in facilitating back propagation of somatic action potentials (Magee and Johnston, 1995a,b), a function that may be modulated by inhibitory cells acting on these dendritic membrane segments.

111.3.2.c. Neurons with axon and dendrites in stratum radia- turn. Somata of this cell type may be located in strata radiatum and pyramidale, from which several smooth and varicose dendrites ascend and descend radially to form a tree largely confined to stratum radiatum (Kawaguchi and Hama, 1988; Gulyds et al., 1993a; Kauer and McMahon, 1995; Miles et al., 1996). Some of the cells may have a multipolar stellatelike appearance, The dendrites, even the radially running branches, rarely enter stratum lacunosum-moleculare. The axon branches in close proximity to the soma and forms a rather sparse arbor that extends throughout the entire width of stratum radiatum. Only a small number of collaterals enter strata pyramidale or oriens. Relatively small en passant varicosities are evenly distributed along the axon. The transverse extent of the axon cloud is smaller than that of perisomatic (basket and chandelier) cells, never exceeding 600 p m in a 400-pm-thick slice. Some of the radially running collaterals, which enter stratum pyramidale, may have been cut during the slicing procedure, and the possibility that some of these neurons are partially visualized bistratified cells cannot be excluded.

INTERNEURONS OF THE HIPPOCAh4PUS 371

Electron microscopy confirms that axon terminals of these neurons form symmetrical synapses with proximal and distal pyramidal cell dendrites and dendritic spines (Gulytis et al., 1993a). One interneuron (an axo-axonic cell) was also shown to be innervated by four contacts. Numerous pyramidal cells were also intracellularly injected within the axonal territory of this stratum radiatum interneuron. Six of them received one contact each, two received two, three cells received four, and others received five, seven, and nine contacts each. The multiple contacts were distributed on different dendritic branches of the target cells (Gulyds et al., 1993a). The dendrites of this cell type received both symmetrical and asymmetrical synapses at a somewhat lower frequency than did other interneuron types (see Section IV.2.b). The majority of axon terminals forming asymmetrical synapses may originate from Schaffer and commissural collaterals in the CA1 region and from commissural-associational fibers and recurrent collaterals in the C M region. In CAI, therefore, these neurons are activated mostly in a feed-forward manner, whereas in C M , they may also receive recurrent collateral input from the local pyramidal cells. O n the basis of input and output characteristics, these cells may be classified as bistratified cells that occasionally lack a stratum oriens arbor.

IIL3.2.d. Interneurons in stratum hnosum-molecuhre (Uf,). Cell bodies of these neurons are within stratum lacunosum-moleculare or at the border of this layer with stratum radiatum. These cells were first described by Ram6n y Cajal (1873, 171 1) and Lorente de N6 (1934) and were recently studied by intracellular recording and subsequent visualization at the light and electron microscopic levels (Kunkel et al., 1788; Lacaille and Schwartzkroin, 1988a; Williams et al., 1794). The dendritic tree is typically bitufted with a predominantly horizontal orientation as opposed to similar dendritic trees of neurons in stratum pyramidale. Some branches, however, often descend to stratum pyramidale, whereas others may travel across the hippocampal fissure and terminate in stratum moleculare of the dentate gyrus. All dendrites are spine free and often varicose. The axon originates from the soma or a main proximal dendrite. Most of the collaterals are also horizontally oriented and arborize predominantly in stratum lacunosum-moleculare or in the bordering region of stratum radiatum. Similar to dendrites, some axon branches may also cross the hippocampal fissure and terminate in the dentate gyrus. All published examples of this cell type have only partially recovered axons; thus, the true extent of the axon arbors remains to be determined.

The ultrastructural features of interneurons in stratum lacunosum-moleculare are similar to those of other nonpyramidal cell types, i.e., infolded nuclei, cytoplasm rich in organelles, and synaptic input of both asymmetrical and symmetrical type covering the dendrites and at a lower density also the soma (Kunkel et al., 1988). With anterograde degeneration, at least some of the asymmetrical, presumed excitatory, synapses were shown to originate from the ipsi- and contralateral CA3 region and from the ipsilateral entorhinal cortex. Axon terminals of these interneurons were shown to make symmetrical synapses with spiny dendritic shafts of presumed pyramidal cells in strata lacunosum-moleculare and radiatum. Contacts on dendritic spines of pyramidal cells

or on shafts of nonpyramidal cells were less common. Axon collaterals in stratum moleculare of the dentate gyrus made symmetrical synapses with dendritic shafts, which showed the ultrastructural features of granule cell dendrites (Kunkel et al., 1988).

III.3.2.e. Interneurorzsprojectingacross subjeld boundaries. The most striking examples of interneurons with a considerable proportion of their axons in more than one hippocampal subfield are the “back-projection’’ neurons of Sik et al. (1774, 1775). Only two such cells have been visualized to date by intracellular injections in vivo, but their regular occurrence has been confirmed by NADPH-diaphorase histochemistry and anterograde tracing (Sik et al., 1994). These cells typically occur at the stratum oriens- alveus border of the CA1 subfield and have horizontal sparsely spiny dendrites, which largely remain in stratum oriens (Fig. 12). The axon forms a local arbor in stratum oriens but also extends into strata pyramidale and radiatum of the CAI region. A main branch of the axon travels to stratum radiatum of CA3, giving rise to a large axon cloud in this layer and also in stratum oriens. Other main axons cross the hippocampal fissure and arborize in stratum radiatum of CA3c and the subjacent hilus of the dentate gyrus. One of the cells had a collateral that became myelinated and disappeared as it entered the fornix, thus suggesting an extrahippocampal or commissural projection. The cell that had a completely filled axon extended for 3.1 mm in the septotemporal direction. The two-dimensional axon length was 101 mm, and the total number of boutons was approximately 25,000. The majority of the boutons were in CA3 (61.5%), whereas the CA1 subfield (24.3%) and the hilar-CA3c region (14.2%) received a smaller proportion of the synapses.

Electron microscopy has been done on a limited sample of axon terminals, and all of them were found to establish symmetrical synaptic contacts with spiny dendritic shafts and cell bodies of presumed pyramidal cells. Interestingly, axospinous contacts have not been observed (Sik et al., 1974). Several boutons were found in the immediate vicinity of capillaries but never came in direct contact with the endothelium.

With the exception of the CA3 projection, these cells are remarkably similar to the trilaminar neurons, particularly in regard to their dendritic morphology, laminar distribution of local (CAI) axons, and a projection through the fimbria, which may prove to terminate in the medial septum (T6th and Freund. 1972). As to their neurochemical identity, they appear to correspond to cells that colocalize calbindin-D28k (CB), somatostatin (SOM), NPY, and nitric oxide synthase (NOS) (Sections IV.3.a, IV.3.b), but direct evidence is not currently available. According to the laminar specificity of their dendritic tree, they are likely to be driven primarily by the local collaterals of CA1 pyramidal cells in a feedback manner. They exert their inhibitory effects on the dendritic tree of pyramidal cells in the CA1 region and most notably in area CA3. The inhibition mediated by back-projection cells, therefore, is in a direction opposite to the excitatory dentate gyrus-CA3-CAl axis. A cross-regional timing of action potentials by these interneurons may be important to secure population synchrony of principal cells in distributed networks and may allow a coordinated induction of synaptic plasticity (Sik et al., 1994).


111.4. Morphological Differences Between Interneurons Terminating in the Dendritic or Perisomatic Region of Principal Cells

The most striking difference between these two major classes of interneurons lies in the laminar distribution of their axon (Figs. 17 and 18) and the resulting difference in postsynaptic elements, i.e., the dendrites or the perisomatic region (soma, axon initial segment, proximal dendrites) of principal cells. However, in addition to this obvious distinguishing feature, there are numerous, more subtle but consistent, differences. Some of these were briefly mentioned in the preceding sections, and they will be collectively described here.

Perisomatic interneurons have large cell bodies with an abundant cytoplasm and larger and more numerous mitochondria than do interneurons innervating principal cell dendrites. This fact suggests that the former have a higher metabolic rate, and it may also explain why their dendrites are frequently studded with large varicosities containing clumps of mitochondria. In contrast, interneurons that distribute their axon arbor to dendritic layers have smooth dendrites with occasional flat varicosities.

The main orientation of the dendritic tree of perisomatic cells is radial. Dendrites cross several (in most cases all) layers of the subfield, thus allowing these cells to sample excitatory input from the majority of local and extrahippocampal sources. In contrast, interneurons innervating principal cell dendrites frequently have horizontal or oblique dendritic trees limited to one or two laminae. These cell types specifically receive excitatory input from certain sources and avoid others. Accordingly, interneurons activated selectively in a feedback or feed-forward manner are often found among dendritic cells, whereas perisomatic interneuron types can be driven both ways (Section X.l).

Important differences also exist in the axonal branching pattern. The perisomatic interneurons usually have long, tangentially running, main axon trunks that are mostly myelinated. These main axon branches emit regular unmyelinared collaterals that arborize densely and form multiple synapses in the perisomatic region of principal cells. Interneurons with an axon arborizing in the dendritic layers often have unmyelinated main axons, which typically divide into branches of similar diameter, particularly upon reaching their target laminae. These morphological characteristics may point to differences in the reliability and velocity of action potential conduction to the terminals.

Members of both interneuron classes usually form multiple contacts with their target principal cells (unlike the pyramial cell-interneuron connection, which is mediated by single synapses; Section 111.2). However, whereas boutons of perisomatic cells are concentrated in a relatively small area of the postsynaptic cell membrane, the dendritic interneurons typically distribute single boutons to different branches. This distribution suggests that even single perisomatic cells may be able to influence the firing of target principal cells, whereas activity of dendritic inhibitory cells may have to be synchronized to modulate dendritic conductance efficiently (Miles et al., 1996; Section XIII.4).

Another consistent difference is observed in the size and mitochondrion content of axon terminals of perisomatic and dendritic interneurons, as described in Section III.3.2.b (Miles et al.,

FIGURE 13. A: Camera lucida drawings of three types (IS-1, IS- 2, IS-3) of interneuron that selectively innervate other interneurons (interneuron-selective cells) in the hippocampus (somata and dendrites in black, axons in red). IS-1 (cells 3, 4, 6): CR-containing interneurons with axonal arbors in stratum radiatum have dendrites located primarily in the same layer and often form axodendritic and dendrodendritic contacts with other CR-positive neurons. For example, the axon of cell 3 forms multiple contacts (open arrows) with the dendrite (dotted outline) of cell 5. The sites and extent of dendrodendritic contacts are labeled with arrowheads and dotted outlines for cells 3, 4, and 6 (see B). One of the dotted dendrites is labeled as cell 5, and it is also indicated in the cluster shown in B. Arrows label those points where CR-positive axon terminals contact the dendrites of cells 3,4, and 6. IS-2 neurons are VIP immunoreactive (cell 1). The dendritic arbor is confined to stratum lacunosum-moleculare, where entorhinal afferents terminate. The axons of these neurons arborize mainly in stratum radiatum. IS-3: B i d e d neuron containing VIP (cell 2). Its axon innervates the stratum oriens-alveus border. B: Schematic drawing of a cluster of CR-immunoreactive interneurons (15 cells) forming dendrodendritic contacts with each other. Parallel red lines and red crosses indicate long (50-200 pm) and short (1-5 pm) dendritic appositions, respectively. Cells 3 and 5 are also shown in A. Because the duster was reconstructed only from seven consecutive sections, the true size of the cluster is likely to be larger. C-F: Multiple contacts (arrows, arrowheads) are formed by axons of CR-containing (C,D) or VIP-containing (E,F) interneurons on the soma and dendrites of CB- immunoreactive (C,E,F) dendritic inhibitory cells and on a VIP-positive basket cell (D). The presynaptic processes are visualized by Ni- DAB (in black) and the postsynaptic cells by DAB (in brown) in each micrograph. From Ascidy L, G6rc.s TJ, and Freund TF (1996) “Different populations of VIP-immunoreactive interneurons are specialized to control pyramidal cells of interneurons in the hippocampus.” Neuroscience 73:317-334 by permission of Elsevier Science Ltd and from Gulyh AI, Hajos N, Freund TF (1996) “Interneurons containing calretinin are specialized to control other interneurons in the rat hippocampus.” J Neurosci 16:3397-3411 by permission of the Society of Neuroscience. Scale bars = 100 pm in A,B, 10 pm C-F.

1996; Halasy et al., 1996). Axon terminals of perisomatic interneurons are larger, usually possess a mitochondrion, and have on average larger synaptic active zones. Interneurons innervating principal cell dendrites have smaller boutons, which frequently lack mitochondria and form synapses with small active zones. In the case of perisomatic interneurons, these differences may underlie a higher tonic firing rate and more reliable transmission. This notion is indeed supported by the findings that spontaneous IPSPs recorded from pyramidal or granule cells have a perisomatic origin (Solttsz et al., 1995; Miles et al., 1996).

Although these differences exist between the “typical” types of the two main classes, one should be aware of cells that show tran- sitory features in one or several of the characteristics listed above. Such examples include the wide-zonal basket cells, trilaminar cells, and HICAP cells.

Members of the third main class of interneurons, which will be described in Section 111.5 do not innervate principal cells but only other interneurons.

111.5. Interneurons Specialized to Innervate Other Interneurons (IS-1, IS-2, IS-3)

Interneurons of this class, named interneuron-selective (IS) cells, have been described most recently. These cells may have been ob-

served in early Golgi studies, but their selective termination on other types of interneurons was not recognized. Hippocampal microcircuits have been considered for over a century as a neuronal network built up from two basic components: a profusely interconnected ensemble of excitatory principal cells and the different subsets of interneurons that govern their activity. A third component should be added to this scheme. A systematic analysis with double-immunostaining techniques identified specialized interneurons that innervate specific subsets of interneurons. GABAergic interneurons belonging to this component of hippocampal microcir-

cuits constitute at least three types (IS-1, IS-2, IS-3) based on their connectivity and neurochemical characteristics.

III.5,a. IS-1 neurons These interneurons are visualized by immunostaining for CR

and occur in all subfields of the hippocampus, with the largest number occurring in the CA1 region, followed by the CA3 region and the dentate gyrus. Their somata are primarily located in strata radiatum, oriens or pyramidale of the hippocampus, and in


the hilus or stratum granulosum of the dentate gyrus. They have a smooth dendritic tree that arborizes most extensively in stratum radiatum but may also invade other layers (Fig. 13). The dendrites of IS-1 neurons in the dentate gyrus occur in all layers. The most characteristic feature of these dendrites is that they form long dendrodendritic junctions with each other, in which typically two to three dendrites are intermingled for more than 100 p m (Fig. 14). Varicose axon collaterals of other IS-I cells are also often involved in these braids. Reconstructions from five serial Vibratome sections reveal that IS-I cells form clusters of 10-15 dendrodendritically connected neurons spanning an area of 500-800 ,um in the transverse direction. Although the axon arbors of these cells can only be partially reconstructed from immunostained sections, basic axonal characteristics can nevertheless be identified. The main axons ramify in stratum radiatum, where they emit several collaterals that course in all directions. These collaterals carry both en passant and drumsticklike axon terminals, which are distributed unevenly along the axon. These collaterals frequently become varicose upon encountering an interneuron soma or dendrite in stratum radiatum. The striking association of the CR-positive axons with interneuron dendrites and somata becomes even more apparent in sections double stained for calbindin, which labels interneurons terminating on pyramidal cell dendrites (Section IV). In addition to the large number of multiple climbing fiberlike contacts on calbindin cells, the CR- positive axons often form similar connections with other CR-positive neurons (Acsidy et al., 1996b; Gulyh et al., 1996).

Electron microscopy confirmed that all boutons of IS-I cells established multiple symmetrical synapses with the dendrites and somata of other interneurons, which were identified with postembedding immunogold staining for GABA. The CR-positive neurons and axon terminals were also GABA-positive. The postsynaptic GABAergic neurons were mostly calbindin-containing interneurons (Section IV.2.b), other CR-positive IS- 1 cells, and less frequently the vasoactive intestinal polypeptide (VIP)-positive basket cells (Section IV.3.d). PV-containing interneurons were not innervated. Interestingly, one subset of basket cells containing VIP was innervated by IS-1 CR-positive cells, whereas the other subset containing PV was avoided. This behavior may suggest that functional differences exist between the two neurochemically distinct basket cell types (Sections IV.2.a, IV.3.d).

Although numerous zonula adherencia were found between adjacent CR-positive dendrites that were attached for several hundred micrometers, the presence of gap junctions, although very likely, could not be unequivocally demonstrated. The dendrites of IS-1 cells are covered by an unusually high density of asymmetrical and symmetrical synapses.

The major excitatory input to these neurons likely originates from commissural-associational fibers (Schaffer collaterals) and perhaps from the entorhinal cortex. Their activity may be synchronized by the multiple dendrodendritic and axodendritic contacts they form with each other. In turn, they innervate mostly interneurons terminating on principal cell dendrites (Fig. 16; Acsidy et al., 1996b; Gulyis et al., 1996). Synchrony of dendritic inhibition appears to be essential for an efficient control of electrogenesis and plasticity in principal cell dendrites (Miles et al., 1996; Section XIII.4).

III.5.b. IS-2 neurons

This cell type is visualized by immunostaining for VIP. To date, IS-2 neurons have been characterized in the hippocampus alone, where their somata ace found in stratum radiatum adjacent to the border of stratum lacunosum-moleculare (Fig. 13). The dendritic tree consists of a tuft of smooth or sparsely spiny dendrites restricted to stratum lacunosum-moleculare, where they profusely arborize. The axon leaves the other pole of the soma and emits several horizontal collaterals, which after some distance turn to descend toward stratum pyramidale, thus forming a weep- ing willowlike arbor. En passant varicosities and drumsticklike boutons are unevenly distributed along these fine axonal branches throughout stratum radiatum. Occasionally, neurons with a dendritic tree typical of IS-3 cells (see below) also have an axon arborization characteristic of this VIP-positive cell group. Correlated electron microscopy reveals that all radially or obliquely running varicose fibers follow the dendrites of individual interneurons, establishing multiple symmetrical synaptic contacts with them. As with the CR-containing (IS-I) neurons, both the VIP-containing axon terminals and their postsynaptic elements are GABA positive. The type of innervated interneuron is also the same. The majority of VIP-positive fibers originating from IS-2 cells make multiple contacts with calbindin-containing cells in stratum radiatum of both CA1 and CA3 (Acsidy et al., 1996a,b). And, as with IS-I cells, the axons of IS-2 cells avoided PV-positive interneurons. Another group of VIP-positive cells with axonal distribution and postsynaptic targets similar to those of IS-2 cells has also been observed. Their dendritic tree is bitufted or bipolar, which spans all layers, and has long branches in stratum lacunosum-moleculare. These cells are considered a subtype of IS- 2 cells (Acsidy et al., I996a,b).

Based on the location of the dendritic tree, IS-2 cells may be driven by entorhinal afferents. Moreover, IS-2 cells innervate interneurons responsible for dendritic inhibition of pyramidal cells in the Schaffer collateral and commissural-associational termination zone (Fig. 16). Inhibitory interneurons terminating on pyramidal cell dendrites may therefore be under dual interneuronal control, one driven by entorhinal afferents (IS-2) and the other by Schaffer collaterals or commissural-associational fibers (IS-1).

III.5.c. IS-3 neurons

This subpopulation of interneurons has also been identified in VIP-immunostained material, both in the hippocampus and in the dentate gyrus (Acsidy et al., 1996a,b; Hijos et al., 1996). Their fusiform cell bodies are located in strata radiatum or pyramidale of the hippocampus and in strata moleculare or granulosum of the dentate gyrus. The dendritic tree of IS-3 neurons is bipolar or bitufted; the apical dendrite ascends straight to stratum lacunosum-moleculare, where ir forms a profuse tuft of fine, sparsely spiny branches (Fig. 13). A small number of cells have a similar tuft in deep stratum oriens, where the branches run horizontally for some distance. The majority of these neurons (except for those with an IS-2-like axon arbor, see above) have a main axon descending to stratum oriens, where it emits several horizontally running varicose collaterals to form a dense plexus at the stratum oriens-alveus border. A similar terminal plexus is

FIGURE 14. A,B: Two light micrographs demonstrate the extent of dendrodendritic contacts between CR-immunoreactive dendrites in the C41 area. Pairs of radially oriented dendrites often cross several laminar boundaries and run attached for more than 200 pm. Besides the extensive dendrodendritic contacts (arrowheads), axodendritic contacts (arrows in B) were also observed. C: Zonula adherentia (arrowheads) are formed between contacting CR-immunoreactive dendrites. One of the CR-positive dendrites receives an additional axodendritic contact (white arrow, symmetrical

synapse), adjacent to the punctum adherens, from CR-immunoreactive axon terminals. D,E: CR-immunoreactive axons form symmetrical synapses (arrow) on GABA-immunoreactive dendrites (d-A) both in the CA1 area (D) and in the dentate gyms (E). As demonstrated in D, in several cases multiple axon terminals innervated the same postsynaptic dendrite. F: A CR-immunoreactive axon terminal (bCR) forms a symmetrical synapse (arrow) on the soma of a VIP-positive basket cell (&p), weakly labeled by the DAB precipitate. Scale bars = 10 pm in A,B, 0.25 pm in C, 0.5 pm in D-F.

FIGURE 15. A,B: The plexus of VIP-immunoreactive axons originating from IS-3 cells (A) shows a perfect overlap with dendrites of mGluR1-positive neurons (B) at the border of strata oriens and alveus in the CA1 region. C: Electron micrograph shows two VIP- positive boutons in synaptic contact (white arrow: the other synapse is visible only in adjacent sections) with an mGluR1-immunoreactive dendrite, which also receives an unlabeled synapse. D: A VIP- immunoreactive bouton forms a symmetrical synaptic contact (arrow) with a GABA-immunoreactive soma (note the accumulation of colloidal gold particles) in stratum radiatum of the CAI region. The

GABA labeling of the VIP-positive bouton is poor in this section due to the dense DAB precipitate; however, it is evident in consecutive sections (not shown), where mitochondria in the bouton are larger. E,F: Correlated light and electron micrographs of VIP-positive boutons forming multiple contacts with a CB-positive interneuron in CA1 stratum oriens. One of the VIP-positive boutons (b,) was shown at the electron microscopic level in F to establish a conventional synaptic contact (arrow) with the CB-positive dendritic shaft (d). Scale bars = 30 pm in A,B, 0.5 p m in C,D,F, 10 pm in E.


formed in the hilus of the dentate gyrus. This terminal field shows a perfect overlap with dendrites of 0 -LM and HIPP cells, which innervate the most distal dendrites of principal cells in stratum lacunosum-moleculare of the hippocampus and in the outer molecular layer of the dentate gyrus. Electron microscopy of sections double stained for VIP and metabotropic glutamate receptor (mGluR1 a), which selectively visualizes dendrites of SOM-containing cells in stratum oriens (Baude et al., 1993), reveal that most VIP-positive terminals in this layer form symmetrical synapses with mGluRla-immunostained dendrites (Fig. 15). Multiple contacts are found also on calbindin-containing neurons of this zone. These latter cells also have horizontal dendrites. However, according to recent results, at least some of these calbindin-positive neurons may be identical to 0-LM cells (Katona et al., in press; Section IV.2.b).

IS-3 cells appear to be driven mainly by entorhinal afferents, although they also have some nonbranching dendrites in other layers. In turn, they selectively innervate interneurons at the stratum oriens-alveus border or in the hilus, most of which are responsible for feedback inhibition of principal cell dendrites in the layers innervated by entorhinal afferents (Figs. 16; Acsidy et al., 1996a,b; Hijos et al., 1996).

The possible roles of interneuron-selective interneurons in the control of population oscillations and/or disinhibition in hippocampal networks will be discussed in Sections X and XIII.

Neurotransmitter content is clearly the most fundamental chemical characteristic of neurons used for functional classification. Whether a neuron exerts an inhibitory or excitatory effect on its target cells largely depends on the nature of substance released from its synaptic terminals (Lloyd, 1946; Kuffler, 1960; Andersen et al., 1963; Eccles, 1964). Ever since GABA was identified as the major inhibitory neurotransmitter in the brain, in particular in the cerebral cortex (Krnjevic and Schwartz, 1967), a large number of studies by using different approaches have been done with the aim of identifying cell types that use GABA as neurotransmitter. Subsequently, excitatory transmitter amino acids have been localized in a complementary cell population, in principal neurons of the neocortex (Storm-Mathisen et al., 1983). The discovery of selective expression of neuropeptides (SOM, CCK, VIP, NPY, enkephalins, substance P [SP]) as possible cotrans- mitters or modulators (for reviews, see Hokfelt et al., 1980; Swanson, 1983) found most often in GABAergic neurons (Somogyi et al., 1984) resulted in further divisions of inhibitory neurons and prompted a search for a correlation between morphology and input-output features of these cells.

Calcium-binding proteins, PV, CB, and CR have also been lbund to be primarily associated with GABAergic neurons in cerebral cortical areas (Celio, 1986, 1990; Baimbridge et al., 1992). ‘Their localization in distinct subsets of interneuron classes pro-

vides an ideal tool to classify and study the connectivity of these neurons further. Greater value is likely to be gained by these chemical neuroanatomical data as the roles of calcium-binding proteins become elucidated.

Highly specific antibodies against neurotransmitter receptors and subunits are now available for immunocytochemical analysis at the light and electron microscopic levels. Studies with these antibodies reveal striking differences between principal cells and interneurons and among morphologically distinct types of interneurons. They provide further evidence for a functional specialization of interneurons with different connectivities and also allows predictions to be made about the action of different transmitters on GABAergic cell types.

The presence of neurotrophic factors (nerve growth factor, NGF; neuotropin 3, NT3; brain-derived neurotrophic factor, BDNF), either synthesized or accumulated, and the presence of nitric oxide (NO)-synthase activity in interneurons of the cerebral cortex implies that interneurons are also engaged in functions other than those derived from synaptic connectivity. This engagement may involve the regulation of synaptic plasticity, sprouting and local blood flow through diffusible messengers, and trophic factors. However, even these roles appear to be coupled to “classical” functions as suggested by correlation to input-output features.

Here we describe the neurochemical characteristics of interneurons according to the questions outlined above and focus on the correspondence between morphological and neurochemical classification schemes (Table 1 .). The functional implications, if available, of the presence or absence of these markers in particular sets of neurons will be described in Sections VII-XVI.

IV.l. Classical Neurotransmitters of Hippocampal Interneurons

IV.1.a. Most if not all interneurons are GABAergic Antisera against the synthesizing enzyme of GABA, glutamic

acid decarboxylase (GAD), were developed in the early 1970s and provided the first tool to localize GABAergic neurons in the central nervous system (Ribak et al., 1978). Some years later, antisera against GABA (i.e., against its conjugates to various proteins) also became available and confirmed the GAD immunocytochemical data describing the strictly nonpyramidal localization of this inhibitory neurotransmitter and its synthesizing enzyme. There are no consistent data demonstrating that other inhibitory neurotransmitters, such as glycine, taurine, or monoamines, may also be present in normal adult cerebral cortical neurons. Choline acetyltransferase (ChAT), the synthesizing enzyme of acetylcholine, is localized in a small number of interneurons in the CA1 region of the hippocampus and the dentate gyrus of the rat (Frotscher et al., 1986), but even these cells are also likely to contain GABA (see below). Thus, apparently GABA, glutamate and acetylcholine are the only three “classical” neurotransmitters (ex- cluding neuropeptides and purines) that are present in local hippocampal neurons. Although transmitter and metabolic pools of glutamate are not distinguishable, glutamate immunoreactivity has been observed in a type of interneuron in a single isolated study (Soriano and Frotscher, 1993a). Thus, GABA, and for a


INTERNEURON-SELECTIVE INHIBITORY CELLS PRINCIPAL CELLS INHIBITORY CELLS

INTERNEURON-SELECTIVE INHIBITORY CELLS PRINCIPAL CELLS INHIBITORY CELLS

INTERNEURONS OF THE HlPPOCAMPUS 379

very small number of cells, acetylcholine, remain the only transmitter candidates for interneurons. The available evidence supporting the view that all morphologically distinguishable interneuron types and only the interneurons (according to the new definition) are GABAergic will be discussed below.

In the hippocampus (and in several other cortical areas), GAD has been localized by immunostaining in interneurons (Ribak et al., 1978, 1981; Somogyi et al., 1983b), but the details of the dendritic and axonal morphology of GAD-immunoreactive cells is limited, even in colchicine-treated animals. Another line of studies took advantage of the presence of a high-affinity GABA uptake system on GABAergic axons by applying tritiated GABA intracortically to characterize neurons capable of taking up the labeled transmitter. Golgi-impregnated cells in the vicinity of the ‘H-GABA injection site were processed for autoradiography, and the selectivity of uptake by interneurons was confirmed (Somogyi et al., 1981). A few years later, antisera against conjugates of GABA also became available and confirmed earlier results obtained by immunocytochemical localization of GAD (Storm- Mathisen et al., 1983; Somogyi et al., 1985b). However, the lack of considerable dendritic staining and the disconnection of the terminal fields from the parent cell bodies put considerable con- straints on the identification of interneuron types immunoreactive for GABA or GAD. Techniques that combined immunostaining with either Golgi impregnation or intracellular recording, and labeling procedures were then developed (for reviews, see Freund and Somogyi, 1989; Somogyi and Freund, 1989), thereby allowing a direct demonstration of GABA or GAD in neurons visualized with mosr of their processes. Moreover, some of these double-labeling techniques enabled the simultaneous identification of morphological and physiological properties of GABA- or GAD-positive cells. Recently, in situ hybridization was used to localize mRNAs encoding two different forms of GAD. This technique appears to be the most sensitive by visualizing the largest number (possibly all) of GABAergic neurons in the hippocampus (Houser and Esclapez, 1994). However, staining is limited to the cell bodies; therefore, this technique suffers from the same limi-

tation as immunostaining when GABAergic cell types are to be identified.

The proportion of GABAergic neurons in the total neuron population of the hippocampus has been studied by immunostaining for GABA and was found to be between 7% (Aika et al., 1994) and 11% (Woodson et al., 1989). These measurements may prove to be underestimates due to false-negative staining, particularly of those GABAergic cells with distant projection (Section VI). Even the GABAergic projection neurons appear to be visualized by in situ hybridization for GAD (Houser and Esclapez, 1994; Esclapez and Houser, 1995), but a quantitative analyzis of this materal has not been reported to date. O n the basis of the presently available information, we tentatively suggest that 10% of all neurons in the hippocampus and dentate gyrus are GABAergic interneurons, with the assertion that this number differs in the different subregions and along the longitudinal axis of the hippocampal formation.

Badet and chandelier cells. The first indirect evidence for the GABAergic nature of basket and chandelier cells derives from immunostaining experiments using antisera against GAD. Somata (Ribak et al., 1978) and axon initial segments (Somogyi et al., 1983b) of hippocampal and neocortical (Freund et al., 1983) pyramidal cells are ensheathed by GAD-positive axon terminals making symmetrical synapses. At tissue levels well penetrated by antisera, practically all synapses that contact somatic and axon initial segment membranes are GAD positive, thus suggesting that interneurons known to innervate these regions of pyramidal cells are GABAergic. The first direct evidence that supports this observation is the demonstration of GABA immunoreactivity in the somata of Golgi-impregnated chandelier cells by using a postembedding immunostaining procedure developed for osmium- treated semithin sections (Fig. 19; Somogyi et al. 1985a). Subsequently, these results have been confirmed for both basket and chandelier cells by a combination of intracellular labeling and postembedding immunogold staining for GABA in the dentate gyrus (Halasy and Somogyi, 1993b; Soriano and Frotscher, 1989; Halasy et al., 1996).

FIGURE 16. Schematic diagrams of circuits involving the three types of GABAergic interneurons that selectively innervate other interneurons (iterneuron-selective cells IS-1, IS-2, IS-3). A: IS-1 cells (in red) contain CR, are in multiple dendrodendritic and axodendritic contact with each other, and innervate CB- and VIP-CCK- containing interneurons (in blue) responsible for dendritic and perisomatic inhibition of pyramidal cells, respectively (i.e., bistratified and basket cells). B: IS-2 cells contain VIP and ihnervate other VIP- positive cells and the CB-containing interneurons (bistratified cells) that synapse on pyramidal cell dendrites in stratum radiatum. IS-3 cells are also VIP immunoreactive and selectively target the SOM- containing horizontal cells of the stratum oriens (0-LM cells). In conjunction with entorhinal afferents, 0-LM cells are responsible for the innervation of the most distal dendritic segments of pyramidal cells in stratum lacunosum-moleculare. IS-1, IS-2, and IS-3 nenrons may have a role in synchronizing inhibitory cells that con- vecge onto a particular group of pyramidal cells. Alternatively, they may disinhibit pyramidal cell dendrites targeted by particular excitatory inputs, e.g., IS-3 cells are likely to reduce feedback inhibition exerted by 0-LM cells in the entorhinal termination zone.

Interneurons innervating principal cell dendrites. It is clear from the early GAD-immunostaining experiments in the hippocampus and dentate gyrus that the dendritic layers (i.e., strata radiatum, oriens, moleculare, and lacunosum-moleculare), where basket and axo-axonic cells do not terminate, also contain a high density of GABAergic axon terminals (Ribak et al., 1978; Somogyi et al., 1983b; Katsumaru et al., 1988b; Halasy and Somogyi, 1993a). The implications of these findings (i.e., that interneurons arborizing in these layers are also GABAergic) has been confirmed by the demonstration of GABA immunoreactivity in neuropeptide or calcium-binding-protein-containing cell types known to terminate on principal cell dendrites (Somogyi et al., 1984; Kosaka et al., 1985, 1988b; T6th and Freund, 1992; for further details, see Sections IV.2, IV.3). Intracellular labeling combined with postembedding immunogold staining for GABA confirmed these findings for additional “dendritic” interneuron types (bis-

380 FREUAJD AND BUZSAKI

Hippocampus LAYER-SPECIFIC

INPUT

s.m.

S.Q.

ho-axonic Basket Bistratified Horizontal Radial 0-LM cell cell cell cell Trilaminar Trilarninar

cell cell

Dentate gyrus LAYER-SPECIFIC INPUT:

I

I hilus

4

4-

entorhid afferents

commissurat/ associational afferents

Mossy fiber +- cotlaterals

Basket HICAP MOPP HlPP cell cell cell cell

Axo-axonic cell

IAJTERNEURONS OF T H E HIPPOCAMPUS 381

FIGURE 17. Summary diagram of the morphological classification scheme. Interneuron types are identitied according to dendritic and axonal arborization patterns in the hippocampus (top) and in the dentate gyrus (bottom). Filled circles mark the cell body location. The cell bodies give rise to thick horizontal and/or vertical lines indicating the predominant orientation and laminar distribution of the dendritic tree. The hatched boxes cover those laminae where the axon of each interneuron typically arborizes. The transverse extension of the axons or dendrites are not indicated. Principal cells in the background provide an idea of which membrane domains (somatic, proximal, or distal dendritic regions) are innervated by the different interneuron types. The laminar distribution of different excitatory aEerents (indicated on the right margin) often shows a perfect match with that of the axon arbor of an interneuron type.

tratified, HICAP, HIPP, and MOPP cells) at the single-cell level (Fig 10; Halasy and Somogyi, 1993b; Halasy et al., 1996).

IV.1 .b. Are there cholinergic interneurons in hippocampus?

Interneurons immunoreactive for ChAT, the synthesizing enzyme of acetylcholine, have been reported in different areas of the rat cerebral cortex by several investigators (Eckenstein and Thoenen, 1983; k e y et al., 1984; Houser et al., 1985; Frotscher et al., 1986), although they may not exist in all mammalian species (Stichel and Singer, 1985). The staining intensity of these clearly nonpyramidal cell types was weaker than that of cholinergic septal or striatal neurons, but cells with the same morphology consistently occurred in specific locations from animal to animal. The most detailed description of ChAT-positive cells in the hippocampus has been provided by Frotscher et al. (1986). In this report, Frotscher et al. counted approximately 50 ChAT-positive cells in 63 Vibratome sections, which on average corresponds to fewer than 1 cell per section. Our recent findings (N. Hijos, K. T6th, Zs. Horvith, and T.F. Freund, unpublished observations) stemming from experiments using a more sensitive antiserum against ChAT (Cozzari et a]., 1990) show that these cells are much more frequently encountered. Up to 20 ChAT-positive cells were found in a single 60-pm-thick Vibratome section. Moreover, the morphology and location of ChAT-positive cells in this study agrees well with the description provided by Frotscher et al. (1 986). The largest number of cells occur in stratum lacunosum- moleculare of the CAI subfield, but they are also found in strata radiatum and pyramidale of CAI and in strata granulosum and deep moleculare of the dentate gyrus. These cells are rarely present in the CA3 region, and they always have small cell bodies (average diameter about 10 pm), which are mostly fusiform or round. Those in stratum lacunosum-moleculare of CAI have two to three horizontal dendrites and at least one long radially oriented dendrite that extends up to the stratum pyrmidale. All dendritic branches are free of spines. ChAT-positive cells in other layers have primarily vertical dendritic trees similar to those in the neocortex.

Electron microscopy provided further evidence that ChAT- immunoreactive neurons in the hippocampus are interneurons. Their spine-free dendrites receive numerous asymmetrical and

symmetrical synapses. Their nuclei display several deep infold- ings. The soma membrane is also engaged in synaptic contacts, including asymmetrical synapses, although at a much lower frequency than the dendritic shafts (Frotscher et al., 1986).

In the neocortex, ChAT-positive cells were found to contain VIP (Eckenstein and Baughman, 1984; Chedotal et al., 1994). Because VIP is present in GABAergic bipolar cells (DeFelipe, 1993), acetylcholine and GABA could coexist in this cell type. Evidence for the colocalization of ChAT and GABA in a small number of cortical neurons has been provided (Kosaka et al., 1988a). In the hippocampus, most of the ChAT-positive cells also show morphological characteristics of VIP-positive GABAergic interneurons (Section IV.3.d), but a direct demonstration of coexistence of ChAT with both GABA and VIP is yet to be shown, Whether acetylcholine is synthesized and released from axon terminals of these neurons together with GABA remains uncertain. The lack of ChAT in the same cell types of the cat cerebral cortex implies that acetylcholine may not be vital for the normal operations of these neurons.

IV.2. Calcium-Binding Proteins Are Present in Largely Nonoverlapping Subsets of GABAergic Interneurons

IV.2.a. PV-containing neurons The selective occurrance of PV in GABAergic neurons of the

cerebral cortex was first shown by Celio (1986) and subsequently confirmed in the hippocampus by Kosaka et al. (1987), who also provided quantitative data on the laminar distribution of PV-positive neurons. In the hippocampus, the distribution of somata, dendrites, and axon terminals of PV-containing neurons show a characteristic pattern (Figs. 23, 26). Practically all cell bodies, which are generally larger than those of other interneurons, are located in strata pyramidale and oriens of the hippocampus and in stratum granulosum and the hilus of the dentate gyrus. In the apical dendritic layers of principal cells, PV-positive somata are extremely rare, and even those are close to the principal cell layers. In the dentate gym, more than 50% of the somata are in stratum granulosum, slightly less in the hilus, and only 2-3% in stratum moleculare. In the CAI and CA3 subfields, the quantitative data are similar. More than 50% of the PV-immunoreactive cell bodies are in stratum pyramidale, 3 0 4 0 % in stratum oriens, and only 3 4 % in stratum radiatum. There were no PV- positive cells in the stratum lacunosum-moleculare, whereas stratum lucidum of CA3 contained somewhat more cells (7-10%) than did the other dendritic layers (Kosaka et al. 1987). The overall proportion of GAD-positive neurons containing PV is 20-24% in the hippocampus, with somewhat higher values for the CAI (22-29%) and the CA3 (20-25%) regions as compared with the dentate gyrus (1 5-2 1 %). Because the majority of PV- positive cell bodies are within or adjacent to the principal cell Iay- ers, they account for approximately half of the GABAergic neurons in these areas. They represent 20-30% of the GABAergic cells in stratum oriens of CA1 and CA3 and 1 4 2 0 % of GABAergic cells in the hilus of the dentate gyrus. Their share in


FIGURE 18. Composite drawing of characteristic interneuron types of the hippocampus and dentate gyms assembled from reconstructions of in vivo (5, 10-13) or in vitro (1-3, 6-9) intracellularly labeled or immunostained (4, 14, 15) neurons in rat (1-5, 10-15) and guinea pig (6-9). Original camera lucida drawings were used, with slight modifications to match the curvature of the subfields. The exact location of the neurons within a particular subfield may also be slightly different from the original to obtain the best fit of all cells, but the laminar distribution of axons and dendrites has been preserved. Thick lines represent the dendritic trees. 1: HICAP cell. 2: HIPP cell. 3 MOPP cell. 4: VIP-containing basket cell. 5: Trilaminar cell of CA3c. 6: Axo-axonic (or chandelier) cell. 7: 0-LM cell of CA3. 8: Bistratified cell of CA3. 9 Basket cell with axon in CA3 and CAl. 10: Bistratified cell of CA1. 11: Basket cell of CAI. 12: 0-LM cell of CAI. 13: Horizontal trilaminar cell of CA1. 14,15: IS-2 (interneuron selective) VIP-containing cells. a, alveus; s.o., s.P., sx., s.1-m., strata oriens, pyramidale, radiatum, and lacunosum-moleculare of the hippocampus; s.m., s.g., h, strata moleculare, granulosum, and hilus of the dentate gyrus; CA1, CA3, subfields of the hippocampus (cornu ammonis) according to Lorente de N6 (1934). The original reconstructions derive from GulyAs et al. (1993a,b), Han et al. (1993), Sik et al. (1995, submitted), Acsddy et al. (1996b), and Hdjos et al. (1996).

the apical dendritic layers is negligible (Kosaka et al., 1987; Aika et al., 1994).

An interesting observation in several PV-GABA colocalization studies is the weak GABA immunoreactivity of PV-positive somata compared with other GABAergic cell bodies lacking this calcium-binding protein (Kosaka et al., 1987; Katsumaru et al., 1988b; GulyAs et al., 1991b; Aika et al., 1994). The reason for this is still unknown. It may be due to high spontaneous firing and an extensive release of transmitter, which requires a rapid transport of the synthesizing enzyme to the terminals and allows little transport of GABA back to the soma.

The dendrites of PV-positive cells span all layers and are oriented largely in the radial direction parallel to the principal cell dendrites (Kosaka et al., 1987; Sloviter, 1989; Celio, 1990; Nitsch et al., 1990b; GulyAs et al., 1991b). The general appearance of the dendritic tree of most cells is bitufted, with frequent branching close to the soma, and a few additional branches at more distal segments (Fig. 20). In the hippocampus, dendrites become relatively sparse in stratum lacunosum-moleculare, although a small number of cells still emit weakly stained branches in this layer. They have a rather even density in stratum moleculare of the dentate gyrus, and most of the branches terminate near the hippocampal fissure or the pial surface. The dentate hilus and stratum oriens of the CAI and CA3 regions also contain a small number of multipolar PV-positive neurons. They have oblique dendrites that may not reach the layers containing the most distal dendrites of principal cells. However, PV-positive neurons having a horizontal dendritic tree confined to stratum oriens or the hilus have not been observed. The majority of PV-immunoreactive dendrites are varicose, especially regions near more distal segments. Large dendritic varicosities are often a sign of accumulation of numerous mitochondria, which may suggest a high metabolic rate. They are generally spine free, but occasional sparsely

spiny segments can also be detected, particularly in stratum ra- diarum of CA1 and CA3. The axon initial segment originates either from the soma or from a proximal primary dendrite and may be directed toward the apical or basal dendrites. The axons cannot be followed beyond the initial segment due to the thick myelin sheat that begins very close to the soma (30-50 pm).

Axon terminal fields immunoreactive for PV also have a characteristic laminar distribution (Fig. 20). They are extremely dense in the cell body layers and proximal stratum oriens. In contrast, practically no PV-positive axon terminal staining can be detected in the dendritic layers fi.e., in stratum moleculare of the dentate gyrus and strata radiatum and lacunosum-moleculare of the CAI and CA3 regions). Although the majority of PV-positive varicosities surround the somata and axon initial segments of principal cells (Fig. 20), several boutons are also found to contact PV-positive cell bodies. This characteristic laminar distribution of PV-positive axon terminal fields overlaps with that of basket and chandelier cells, suggesting that one or both of these morphologically identified cell types selectively contain this calcium-binding protein Kawaguchi et al., 1987. Direct evidence for this notion has been provided by Katsumaru et al. (1988b), who demonstrated that the majority of PV-positive axonal varicosities formed symmetrical synapses with the somata, proximal dendrites, and axon initial segments of pyramidal neurons in the CA1 and C.43 regions (Fig. 20). Similar observations were made by Soriano et al. (1990) in the dentate gyrus. In a recent in vivo study, intracellularly recorded and filled basket cells were shown to be immunoreactive for PV but not for calbindin or CR (Fig. 21; Sik et al., 1995).

Thus, taken together, all PV-positive neurons can be classified as basket or chandelier cells, but the question arises as to whether all basket and chandelier cells contain PV. The answer is no. Ribak et al. (1990) demonstrated that PV-positive boutons do not account for all symmetrical synapses on the perisomatic and axon initial segment membrane of granule and pyramidal cells. Recent light and electron microscopic studies of neuropeptide-containing cells (Nunzi et al., 1985; Acsidy et al., 1996b) and their colocalization with calcium-binding proteins (Gulyh et al., 1991b; Acsidy et al., 1996a) have provided further evidence for the existence of basket cells that do not contain PV. This different subset of basket cells are visualized with immunostaining for CCK and VIP (Sections IV.3.c, IV.3.d).

The ultrastructural features of PV-positive somata and dendrites have also been studied by several laboratories (Kosaka et al., 1987; Katsumaru et al., 1988b; Sik et al., 1993), and the results agree with those obtained for identified basket and chandelier cells in single cell labeling studies (Sections 111.1, 111.2.) and will not be repeated here. One feature that has not been emphasized in Section 111 is the occurrence of gap junctions between PV-positive dendrites andlor somata and between PV-positive dendrites and shafts or spines of unknown origin (Katsumaru et al., 1988a). The overall frequency and selectivity of gap junctions in this neurochemically characterized group of GABAergic neurons remains to be established (see Section III.5.a for CR-containing neurons with dendrodendritic junctions), but in general, this membrane specialization appears to be common among interneurons (Kosaka, 1983b,c; Kosaka and Hama, 1985).

384 ERELIND AND SUZShU

SR

so

FIGURE 19. Demonstration of GABA immunoreactivity in a Golgi-impregnated axo-axonic (chandelier) cell in the cat hippocampus. A Low power light micrograph of the axo-axonic cell showing numerous vertically running rows of boutons (arrows), each wrapping around the axon initial segment of a pyramidal cell. B: Camera lucida drawing of the same cell, with the axon confined to stratum pyramidale and proximal oriens and a radially oriented dendritic tree spanning all layers. C: High power light micrograph of the cell body (open arrow) as seen in the 80-pm-thick Golgi section. D,E: Serial semithin (1 pm) sections cut from the soma of the

Golgi-impregnated axo-axonic cell, one immunostained for GABA (D) and the other with an anti-GABA serum preadsorbed with GABA (E). A small capillary (c) and a GABA-positive but not Golgi-impregnated soma (s) serve as landmarks to identify the axo-axonic cell in each figure. The axo-axonic cell (arrow) is positive for GABA in D; in the control section (E), only the silver deposit deriving from Golgi impregnation is visible as an outline of the cytoplasm. Pyramidal cells (P) are GABA-negative and indicate background level. Scale bars = 50 pm in A, 100 pm in B, 10 pm in C-E.

The major excitatory synaptic inputs of PV-containing neurons can be predicted from their dendritic distribution, and from earlier tracing studies of morphologically identified basket and axo-axonic cells (Section 111). Intracellular labeling studies have

visualized several chandelier cells with a dendritic tuft in stratum lacunosum-moleculare, whereas basket cells described in the same studies had shorter dendrites in this layer (Li et al., 1992; Gulyb et al., 1993a; Han et al., 1993; Buhl et al., 1994a). This finding


suggests that dendrites belonging to axo-axonic cells may have a larger share of the relatively smaller number of PV-positive dendrites in stratum lacunosum-moleculare. Thus, axo-axonic cells may receive a larger proportion of their input from the entorhinal cortex than do basket cells in the same region. Direct evidence for associational-commissural and entorhinal inputs to PV-positive cells is available (Zipp et al., 1989; Deller et al., 1994; Kiss et al., 1996). The subcortical innervation of these and other interneuron types will be described in Section V.

IV.2. b. CB-containing interneurons

Unlike PV, the other calcium-binding protein, CB, has been shown to be present in both principal cells and interneurons of the hippocampal formation (Baimbridge and Miller, 1982; Baimbridge et al., 1982; Sloviter, 1989; T6th and Freund, 1992). Strong CB staining is found in the granule cells of the dentate gyrus. The cell bodies, the entire dendritic tree, and axon (the mossy fiber projection) of practically all granule cells are strongly immunoreactive for CB. MispIaced granule cells, often mistaken for interneurons, are frequently found in stratum radiatum of the CA3c subfield, close to the end of the upper blade of the dentate gyrus. These cells have granule cell morphology, in which their dendrites enter stratum moleculare of the upper blade; nonethe- less, they are always negative for GABA (T6th and Freund, 1992). ‘The other set of CB-positive principal cells are the “superficial” pyramidal cells of the CA1 subfield (i.e., 2-3 rows of cell bodies toward stratum radiatum; Baimbridge and Miller, 1982). These superficial pyramids are rather faintly stained in the dorsal hippocampus and more intensely stained in the ventral hippocampus (Fig. 20). Their axons, which form a plexus of horizontally running fibers at the stratum oriens-alveus border, are not stained for the calcium-binding protein. All other CB-positive neurons in the hippocampus and dentate gyrus should be considered interneurons. These CB-positive cells were generally shown to be tmmunoreactive for GABA, with the exception of a small set of neurons in stratum oriens (T6th and Freund, 1992). These latter neurons, however, project to the medial septum, where their axon terminals were found to be immunoreactive for GABA (T6th et al., 1993). Thus, this cell type represents a group of GABAergic interneurons with distant projection (Section VI), which have somatic GABA levels below the immunocytochemical detection threshold. Data about the proportion of GABAergic neurons that contain CB is not available with a precise regional and laminar resolution as for PV, but the value of 10-12% should be considered a close estimate. The coexistence of CB with the other TWO calcium-binding proteins is negligible (Gulyis et al., 1991b; Miettinen et al., 1992; for neuropeptide coexistence, see Section IV.3).

The majority of CB-positive interneurons in the CA1 subfield are in the distal one-third of stratum radiatum, adjacent to the border of stratum lacunosum-moleculare. They are also present in strata oriens, pyramidale, and lacunosum-moleculare, but at lower frequencies (Figs. 20, 23). The number of CB-positive interneurons is consistently higher toward the subicular end of the CA1 region than in areas near the CA2 border and are also somewhat higher in the ventral than in the dorsal hippocampus. Cell-

rich and cell-poor segments of stratum radiatum alternate often in an unpredictable fashion, making quantification of these cells very difficult. Their dendrites and somata always stain stronger than those of pyramidal cells and can therefore be distinguished even if they are located in stratum pyramidale (Fig. 20). Their dendrites can be followed for a considerable length. The reconstruction of even partially stained dendritic trees reveals diverse morphologies. Dendrites may originate in any direction but often turn radial at some point; thus, the predominant dendritic trajectory remains radial. Exceptions to this rule are found in strata oriens and lacunosum-moleculare, where CB-positive neurons often have a horizontally oriented dendritic tree. Dendrites of the former type may extend for more than a millimeter without leaving a narrow zone at the oriens-alveus border. The majority of these cells contain the neuropeptide somatostatin (Section IV.3.a). The horizontal cells at the stratum radiatum-lacunosum-moleculare border often have a considerable proportion of their dendritic tree in stratum lacunosum-moleculare. In contrast, dendrites of radially oriented bitufted cells in stratum radiatum appear to avoid this layer. Their ascending branches course horizontally at the stratum lacunosum-moleculare border and may run in this direction for a considerable distance without penetrating the perforant path termination zone (Section III.3.2.b).

Axons of CB-positive interneurons are difficult to visualize. In those few animals with successful axonal staining, the arbors were found in the dendritic layers of pyramidal cells, most notably in stratum radiatum, but also in strata oriens and lacunosum-moleculare (Figs. 20, 26; Gulyis and Freund, in press). Axons frequently crossed the stratum pyramidale without giving rise to varicose collaterals in this layer (but see the CA3 subfield below). The axonal branches in strata radiatum and oriens were evenly covered with en passant varicosities; drumsticklike terminals, however, were rarely observed. CB-positive axons were rarely seen in stratum lacunosum-moleculare.

CB-positive neurons are more numerous and more easily visualized in the CA3 subfieId than in the CA1 region. The dendritic orientation here is truly multipolar, and most cells have a stellatelike appearance in all layers. The largest number of CB- positive somata is found in stratum radiatum of CA3a-b, but cells are also common in stratum oriens. They gradually diminish in number in CA3c and in the hilus of the dentate gyrus. Only short axonal segments can be reconstructed in this subfield, but those were highly varicose and arborized in stratum radiatum. Large neurons with triangular somata had axon initial segments that became myelinated close to the soma and were seen to enter the white matter. These cells are likely to be among those projecting to the medial septum (T6th and Freund, 1992).

The dentate gyrus contains the smallest number of CB-positive interneurons, and even those few cells are barely visible due to the strong immunoreactivity of granule cells. In stratum moleculare, their number varies from 3 to 20 per single coronal section (60 pm, dorsal hippocampus), and their major dendritic orientation may be either horizontal or vertical. Interneurons in stratum granulosum cannot be distinguished, whereas those in the hilus number 3-10 per section. Axons of CB-positive interneurons in the dentate gyrus have not yet been reconstructed.

Electron microscopy of CB-positive axon terminals, originat-

386 FREUND A N D BUZSAKI


FIGURE 20. Light and electron micrographs of immunocytochemically visualized interneuron types containing different neurochemical markers, which terminate in complementary layers of the CAI subfield. A,B: PV-immunoreactive neurons (arrows) have dendrites spanning all layers, but their axon arborization is limited to stratum pyramidale and proximal oriens. The axon terminal distribution is shown in B at higher magnification. PV-positive boutons surround the cell bodies (P) and axon initial segments (arrowheads) of unstained pyramidal cells. C,D: CB immunoreactivity is present in the superficially located pyramidal cells in the CAI subfield and in several interneurons (arrows) mostly in stratum radiatum but also in strata pyramidale and oriens. Interneurons always stain stronger for CB than do pyramidal cells. Axons of CB-positive interneurons are faintly stained, arborize in s t r a t u m radiatum, and are at a lower density in stratum oriens. The framed area is shown in D at higher magnification, where the faintly labeled axon collaterals (arrowheads) are more visible. E,F: SOM-positive interneuron somata (arrows) are present in stratum oriens, whereas their axon arborizes in stratum lacunosum-moleculare. The framed area is shown at higher magnification in F. The punctate axon terminal labeled begins at the stratum radiatum-lacunosum-molecdare border. G,H: Electron micrographs of PV-positive boutons in synaptic contact (arrows) with an axon initial segment (IS in G) and with a soma (S in H) of a pyramidal cell. Arrowheads in G label the membrane undercoating characteristic of axon initial segments. A PV-negative bouton (asterisk) also synapses on the same cell body shown in H. I: A CB-positive bouton, also immunoreactive for GABA as shown by the accumulation of colloidal gold particles (deriving from postembedding immunogold staining), forms symmetrical synapses (arrows) on GABA- negative, most likely pyramidal cell dendrites in stratum radiatum of the CA1 subfield. J,K: SOM-immunoreactive boutons are in symmetrical synaptic contact (thick arrows) with spinelike processes or thin dendrites of presumed pyramidal cells in stratum lacunosum- moleculare of the CAI region. The same spine, shown in K, also receives an asymmetrical synapse from an unstained axon terminal (asterisk). Thin arrows in J indicate strongly immunoreactive large dense-core vesicles. Scale bars = 50 pm in A,C,E, 10 pm in B,D,F, 0.5 pm in G,H, 0.2 pm in I-K.

ing from somata in stratum radiatum, has been done in combination with postembedding immunogold staining for GABA (Fig. 20). The axon terminals in stratum radiatum establish symmetrical synapses and are all GABA positive. Their postsynaptic elements, which include dendritic shafts and spines, are negative for GABA. This result confirms that CB-positive interneurons are involved in the GABAergic innervation of pyramidal cell dendrites in strata radiatum and oriens. The ultrastructural data on somata and dendrites of CB-positive neurons is rather limited (Danos et al., 1991; Gulyis and Freund, in press). Their indented nuclei and asymmetrical synaptic input on somata are characteristic of interneurons, but the relatively sparse afferent synapses along the dendritic shafts is a feature that distinguishes them from other interneurons.

According to the dendritic and zona l distribution and location, CB-positive neurons correspond to cell types described in Section 111.3.2 as those innervating pyramidal cell dendrites. They appear to include the bistratified cells, other interneurons with dendrites and axons arborizing in stratum radiatum (Sections 1II.3.2.b, II1.3.2.c), and the horizontal cells located at the border (of strata radiatum and lacunosum-moleculare (Section III.3.2.d). 4n intracellularly injected trilaminar or bistratified cell located at

the CA3c stratum oriens-hilus border and a bistratified cell in the CAI region, both of which arborize in strata radiatum and oriens of the same subfields, have been shown to contain CB (Fig. 21; Sik et al., 1995, submitted). Horizontal 0-LM cells, the trilaminar cells and the back-projection neurons in the CA1 region were found to be immunonegative for CB (Sik et al., 1995). Importantly, however, such negative findings, especially when immunostaining of intracellularly injected cells is attempted, should be taken with great caution. Basket and chandelier cells are clearly not among the CB-positive neurons because of the complementary laminar distribution of axon arbors (perisomatic vs. dendritic innervation of pyramidal cells).

IV.2.c. CR-containing interneurons form spiny and spine-free subpopulations with distinct connectivities

Immunostaining for CR reveals a large number of interneurons in all layers and subfields of the hippocampus and dentate gyrus (Figs. 23, 26; Jacobowitz and Winsky, 1991; Gulyis et al., 1992; Miettinen et al., 1992; Resibois and Rogers, 1992; Rogers, 1992). Two major types can be distinguished on the basis ofden- dritic morphology and distribution.

Spiny CR-immunoreactive neurons. The spiny variety is present mostly in regions where mossy fibers have a high density, i.e., in the hilus of the dentate gyms and in stratum lucidum of the CA3 subfield. Their dendrites and frequently the somata possess numerous long hairlike spines that penetrate into bundles of mossy fibers. These thin preterminai mossy fibers establish numerous (3-6) synaptic contacts on these dendrites. In the dentate gyrus, the dendrites are restricted to the hilus and never enter stratum granulosum. In CA3 stratum lucidum, the dendrites run parallel with the pyramidal cell layer, precisely follow its curve, and rarely enter strata pyramidale or radiatum. The axon of these neurons cannot be visualized by immunostaining, probably because of myelination. However, neurons of this type located in stratum lucidum of CA3 have been intracellularly injected with biocytin in vitro. Their axons arborized in stratum lacunosum-moleculare of the CA3 region, and a large arbor also formed in the outer molecular layer of the dentate gyms (C. McBain, personal communication). In another unpublished study, spiny cells with the same location and dendritic morphology sent axon collaterals to stratum radiatum (M. Frotscher, personal communication). These cells apparently form symmetrical synapses with presumed pyramidal cell dendrites; however, the types of synaptic contacts are difficult to identify in sections from the surface of immersion- fixed slices due to poor ultrastructural preservation (M. Frotscher, personal communication). Thus, these neurons are likely to be among the interneurons that innervate principal cell dendrites but have the distinguishing feature in that they receive a predominant mossy fiber input (also Section III.3.1.e). It is interesting to note here that pyramidal cells in CA3c of the ventral hippocampus show transitional features of mossy cells and CA3 pyramidal cells in that ventral CA3c pyramids may project to the inner molecular layer of the dentate gyrus and to the CA1 region (Li et al.,

388

A.

FREUND AND SUZSAKl

basket cell (CAI),

B. trilaminar cell ( CA3) ---c _- 1

-”%

-“-=..

C. HIPP cell (hilus)

INTERNEURONS OF THE H I P P O C M P U S 389

FIGURE 21. Neurochemical identification of intracellularly recorded and biocytin-filled interneurons. A: Basket cells are PV immunoreactive and arborize mostly in the pyramidal cell layer. Photomicrograph shows the biocytin-labeled basket cell and the fluorescent immunostaining for PV in the same section. B: Trilaminar interneuron. Reconstruction of the dendritic arbor and partial reconstruction of the axon collaterals (from 7 of 53 60-pm-thick sections, centered at the cell body). Photographs depict the CB immunoreactivity of the biocytin-filled cell, as in A. C: Hilar interneuron with axon terminals associated with the perforant path (HIPP cell). Reconstruction of the dendritic arbor and partial reconstruction of the axon collaterals (from 7 of 39 80-pm-thick sections, centered at the cell body). The dendrites remained in the hilus, and the axon collaterals densely innervated the outer molecular layer of dentate gyms. The light microscopic pictures show the fluorescent NPY immunoreactivity (FITC) and biocytin-labeling of the cell body in the same section. Arrows in the photographs point to the biocytin-filled cell bodies. p, pyramidal layer; r, stratum radiatum; Im, stratum lacunosum-moleculare; g, granule cell layer; f, hippocampal fissure: h, hilus; m, molecular layer. Data are from Sik et al., 1996, submitted.

1994). Spiny CR cells may represent a similar transitional cell type showing characteristics of HIPP cells of the dentate hilus and of 0-LM cells of the CA3 region, in which they innervate principal cell dendrites in the entorhinal termination zone in CA3 and in the dentate gyrus. However, inhibition produced by this cell type in CA3 stratum lacunosum-moleculare is of the feed-forward type, which is in contrast with its effect in the dentate gyrus or with the inhibition likely exerted by conventional 0-LM and HIPP cells.

The GABAergic nature of the spiny CR-positive cells remains controversial. Neurons of this type, visualized by either CR immunostaining or Golgi impregnation, are negative for GABA or stain close to background level (Miettinen et al., 1992; Soriano and Frotscher, 1993b). However, negative somatic staining for GABA does not necessarily mean that the neuron is not GABAergic (see Section IV.2.b). For example, although some earlier immunostaining studies have suggested that a large proportion of the SOM-positive hilar cells are negative for GABA or GAD, all hilar SOM-synthesizing neurons were shown by in situ hybridization to colocalize GAD65 mRNA (Esclapez and Houser, 1995). Spiny CR cells may have an extensive projection, which would explain the low levels of GABA in the soma. A small number of spiny CR-positive neurons in the hilus have been found to project commissurally, and a single cell has been retrogradely labeled from the medial septum (Miettinen et d., 1992). Glutamate immunoreactivity has been observed in the soma of a Golgi-impregnated cell of this type (Soriano and Frotscher, 1993b), but whether it should be considered as glutamate dedicated to transmitter or merely as a reflection of the low level of GAD activity in the soma (which would convert all glutarnate into GABA) is unknown. Nevertheless, the symmetrical synapses they appear to form on dendritic shafts in stratum radiatum may argue against the glutamatergic nature of these neurons (see above).

Spine-fee CR-immunoreactive neurons. The majority of these neurons belong to the class of cells described as “interneurons specialized to innervate other interneurons” (Section III.5.a). The CR-positive axons with a clustered distribution of varicosities

(both en passant and drumsticklike) in stratum radiatum and axons forming a horizontal plexus at the stratum oriens-alveus border are those shown to selectively innervate other interneurons (Gulyis et al., 1996). These axons originate from spine-free CR- positive cells with primarily radial (bipolar or bitufted) dendritic trees, several of which form multiple dendrodendritic junctions with other CR-positive cells of the same type, as described in Section 111.5.

Additional CR-immunoreactive interneurons not included among those described in Section III.5.a are horizontal neurons located near the hippocampal fissure. The soma of these neurons are located on either side of the fissure. Their axons could not be visualized. According to developmental studies of calcium-binding-protein-containing cells, these neurons appear to correspond to Cajal-Retzius cells (Soriano et al., 1994). Main axon trunks and fine varicose fibers with a predominant horizontal orientation and unidentified origin are also found in stratum lacunosum-moleculare. Electron microscopy showed that most of the varicosities along these axons establish asymmetrical synapses with GABA-negative dendrites or spines, and they themselves were also GABA negative. These findings suggest that most of the CR-positive axon terminals in stratum lacunosum-rnoleculare have an extrahippocampal origin, probably from the nucleus reuniens. This thalamic nucleus was shown to send a projection specifically to this layer (Wouterlood et al., 1990) and has CR-irnmunoreactive neurons in large numbers (Jacobowitz and Winsky, 1991).

IV.3. Neuropeptides Are Contained in Subsets of GABAergic Interneurons With Distinct Connectivities

IV.3.a. SOM-immunoreactive neurons

Prosomatostatin is a 28-amino-acid-long peptide that is cleaved to produce a leader peptide (SOMI-12) and the biologi- cally active peptide (SOM14-28). Antisera raised against either of these peptides have been used in immunocytochemical studies and were found to visualize similar cell populations. We will not distinguish between data obtained by antisera directed against one or the other peptides in the present review.

Immunostaining for SOM visualizes a large number of neurons in all subfields of the hippocampus, with characteristic laminar distribution (Figs. 24, 26; Kohler and Chan-Palay, 1982; Morrison et al., 1982; Johansson et al., 1984; Roberts et al., 1984; Somogyi et al., 1984; Bakst et al., 1986; Kohler et al., 1987; Sloviter and Nilaver, 1987; Kosaka et al., 1988b; Kunkel and Schwartzkroin, 1988; Milner and Bacon, 1989b; Leranth et al., 1990; Nitsch et al., 1990a; Finsen et al., 1992; Buckmaster et al., 1994). All SOM-positive neurons have been characterized as interneurons on the basis of morphological features, frequency of occurrence, and distribution.

Colocalization studies have demonstrated that practically all SOM-positive neurons in the hippocampus and dentate gyrus are immunoreactive, or contain the mRNA, for GAD (Somogyi et al., 1984; Kosaka et al., 1988b; Esclapez and Houser, 1995). According to Kosaka et al. (1988b), SOM-positive cells account

FIGURE 22. Photomicrographs of fluorescent confocal images of sections double-immunostained for calretinin (A,C,E) and NO synthase (B,D,F) in the CAI region of the rat hippocampus. The same areas are shown in pairs of photographs in A-B, C-D, and

E-F. There are numerous double-labeled cells (arrows), but neurons immunoreactive for NO synthase (arrowheads) or CR (open arrows) alone are also frequently seen. This figure was kindly prepared by Toshio Kosaka. Scale bars = 100 p m in A,B, 20 p m in C-F.


for 14% of all GAD-positive neurons in the hippocampus, being somewhat higher in the dentate gyrus (16%) than in the CAI (12.5%) and the CA3 (13%) subfields.

Dentate gyms. In the dentate gyrus, SOM-positive somata are restricted to the hilus, where they tend to accumulate in the subgranular polymorphic zone, but are also numerous in the deep hilus and in areas bordering CA3c stratum radiatum. Unfortunately, the dendritic trees of these neurons are poorly visualized by SOM antisera. Only the most proximal segments are visible in the majority of published studies; therefore, their laminar distribution cannot be determined from SOM immunostaining alone. Neverthe- less, indirect lines of evidence from intracellular labeling studies (Section 111.3.1 .a) and from immunostaining against receptors associated with SOM-containing interneurons (mGluR1 and SP receptor [SPR]; Sections IV.6.a, IV.6.f) strongly support the notion that dendrites of SOM-positive neurons, apart from a few exceptions (Leranth et al., 1990), remain in the hilus. Neurons in the subgranular zone have a fusiform soma, and the dendrites appear

Parvalbumin

to run parallel to the granule cell layer. Those cells located deeper in the hilus may have multipolar dendritic trees. The spiny or spine-free nature of dendrites is impossible to establish from SOM immunostaining alone. However, when visualized by immunostaining for SP or mGluRl receptors, all probable SOM-containing dendrites are covered by numerous long spines (Sections IV.6.a, IV.6.f). In summary, the dendritic tree of SOM-positive neurons in the dentate gyms is typically restricted to the hilus and is covered by numerous long spines.

The density of axon terminal staining differs considerably from animal to animal and with the type of antisera and staining procedure used. However, the laminar distribution of SOM-positive axons is consistent. There is a terminal plexus in the outer two- thirds of the molecular layer, the outer one-third being more dense than the middle one-third. The inner molecular layer and the granule cell layer have a small number of radially running axons, which are mostly nonvaricose. Main axon collaterals are often seen to cross the hippocampal fissure. Varicose collaterals in the hilus are sparse but occur consistently. The majority of axon collater-

Calretinin

\

v

Calbindin

. * .. . FIGURE 23. Camera lucida drawings illustrate the distribution of neurons in the hippocampus that contain one of the calcium- binding proteins or NO synthase. Drawings represent single 60-pm- thick Vibratome sections, cut from approximately the same anteroposterior level of the hippocampus, immunostained for the different calcium-binding proteins or histochemically stained for NADPH-di-

NAD PH-diaphorase /.--~----.- - +.

aphorase, to visualize neuronal NO-synthase-containing cells. Each dot represents one cell. The relative distribution and numerical density of PV-, CB-, CR-, and NADPH-diaphorase-positive neurons can be compared in all subfields and layers. Straight dashed lines indicate the CA1-CA2 border.

392 FREUND AND l3UZSAKl

VIP Somatostatin

CCK NPY

FIGURE 24. Camera lucida drawings illustrate the distribution of neuropeptide containing neurons. Single 60-pm-thick Vibratome sections cut from approximately the same anteroposterior level of the hippocampus as that shown in Figure 23 were immunostained

for different neuropeptides. Same conventions as in Figure 23. The relative distribution and numerical density of VIP-, CCK-, SOM-, and NPY-immunoreactive neurons can be compared in all subfields and layers.

als in the molecular layer originate from ipsilateral hilar neurons, as demonstrated by the disappearance of most SOM-fiber labeling from this layer after hilar kainate injections (Bakst et al., 1986). Hilar SOM-positive neurons can be retrogradely labeled by tracer injections into the contralateral dentate gyrus, suggesting that a smaller proportion of axons originates from the contralateral hilus (Bakst et al., 1986; Deller et al., 199Sb). Finally collaterals may also derive from ipsilateral CA1 neurons because axons passing through the hippocampal fissure are also found (Bakst et al., 1986).

Electron microscopy of SOM-immunoreactive axon terminals show that the majority of postsynaptic elements in the outer molecular layer are spines and spiny dendritic shafts, most likely belonging to granule cells (Leranth et al., 1990). The synapses are always symmetrical and are accompanied by an unstained asymmetrical synapse when terminating on a spine. Occasional synapses are formed on deep hilar neurons and on granule cell bodies. At the electron microscopic level, SOM-positive dendrites were found to be spiny and received numerous asymmetrical synapses from mossy fiber terminals both along the shaft and onto

the spines (Leranth and Frotscher, 1987). The cell body and somatic spines were also contacted by mossy terminals, although at a lower frequency. Only a negligible number of dendrites penetrate the stratum moleculare, where they were shown to receive input from the entorhinal cortex (Leranth et al., 1990). The cell bodies showed typical features of interneurons, i.e., invaginated nuclei, asymmetric synaptic input, intranuclear rods, and numerous cytoplasmic organelles. The cytoplasm, however, was not always as abundant as in some other types of interneurons (C;ulyis et al., 1990; Leranth et al., 1990).

According to the characteristic soma location and axonal and dendritic distribution, hilar SOM-positive cells are likely to be identical to HIPP cells of the morphological classification scheme (Section III.3.1.a). The correlation of morphology and neurochemical character for this cell type is perhaps the most straightforward of all such attempts; thus, “SOM cells” and “HIPP cells” in the dentate gyrus can be considered as synonymous.

Hippocampus. SOM-immunoreactive cell bodies are primarily confined to stratum oriens in the CA1 region, with only a small

FIGURE 25. Light micrographs show the colocalization in the rat DG of GluR subunits and calcium-binding proteins or SOM. By applying the mirror technique with paired surfaces of adjacent Vibratome sections, the colocalization of GluR2/3 (al) with PV (az), GluR2/3 (b,) with CB (bz), and GluR4 (c1) with SOM (cz) is demonstrated. Immunoreactivity for GluR2/3 is associated with granule cells and neurons (most likely mossy cells) in the deep hilus. None of the PV-positive and only a small number of hilar CB-positive cells (probably misplaced granule cells, numbered 1-3 in bl and bz), contain GluFW3 immunoreactivity. Large GluRW3-positive cells (probably mossy cells) in bl (long arrows) are CB-negative, as shown in

the adjacent section. The silhouettes of the negative somata are visible in bz. Immunostaining for the GluR4 subunit in the dentate gyrus was limited to hilar neurons, with occasional cells stained in stratum granulosum. AU SOM-immunoreactive cells in the hilus (numbered arrowheads in c2) were also immunoreactive for GluR4 (arrowheads in c1). From Leranth C, Szeidemannz, Hsu M, Bum& G (1996) “AMPA receptors in the rat and primate hippocampus: a possible absence of Glur 213 subunits in most interneurons.” Neuroscience 70:631-652 by permission of Elsevier Science, Ltd. This figure was kindly prepared by Csaba Leranth. Scale bars = 50 P n .

393

394 FREUND AND S U Z S k I

number of cells in stratum pyramidale, and practically none in strata radiatum and lacunosum-moleculare (Fig. 20). They are more widespread in the CA3 region and frequently occur in strata pyramidale, lucidum, and proximal radiatum. Some cells can be found embedded in the myelinated axon bundles of the alveus in both regions but especially in CAI. As in the dentate hilus, SOM- immunostaining alone visualizes only the somata and the most proximal dendrites. However, SOM-positive cells are immunoreactive for the metabotropic glutamate receptor mGluRl, which stains their dendritic tree in great detail (Baude et al., 1993; BIasco-Ibanez and Freund, 1995). The dendrites of SOM-positive neurons in the CA1 region run horizontally and are restricted to a narrow zone at the border of stratum oriens and the alveus. In the CA3 subfield, SOM-positive cells have a multipolar dendritic tree that penetrates all layers except stratum lacunosum- moleculare.

Thus, there is a remarkable correlation between the laminar distribution of SOM cell dendrites and the local axon collaterals of principal cells both in the hippocampus and in the dentate gyrus. Pyramidal cells in the CAI region have horizontally running local collaterals, which distribute to the border between stratum oriens and alveus, whereas collaterals of CA3 pyramidal cells arborize extensively throughout strata oriens, radiatum, and pyramidale (Section 11). Local collaterals of granule cells in the dentate gyrus (i.e., mossy fiber collaterals) in normal animals are known to arborize predominantly in the hilar region in a manner similar to the dendrites of SOM-positive neurons. These data predict that the major excitatory input of SOM cells will derive from local principal cells, i.e., they are likely to mediate feedback inhibition (Section X. 1). Direct evidence for this hypothesis has been observed in the CA1 region, where degenerating boutons of pyramidal neurons selectively damaged by ischemia were shown to account for more than 80% of all asymmetrical synaptic input along the dendrites of SOM cells in stratum oriens (Blasco-Ibanez and Freund, 1995).

As in the dentate gyrus, the laminar distribution of SOM-positive axon terminal fields shows a perfect overlap with the termination zone of entorhinal afferents, densely covering stratum lacunosum-moleculare in both CAl and CA3 (Fig. 20). The axons are highly varicose in this layer, whereas only sparse nonvaricose main axons are found in stratum radiatum. Occasionally, beaded collaterals are observed in stratum oriens.

Electron microscopic data on SOM-immunoreactive neurons in the hippocampus are scarce (Kunkel and Schwartzkroin, 1988; Gulyas et al., 1990; Danos et a]., 1991; Baude et al., 1993; Blasco- Ibanez and Freund, 1995). The available data suggest that SOM cells have no distinguishing ultrastructural features as compared with other interneurons. Their dendrites receive both asymmetrical and GABAergic symmetrical synapses, most of the former originating from CA1 pyramidal cell collaterals and the latter from VIP-containing interneurons (Section III.5.c and below). SOM- positive axon terminals in stratum lacunosum-moleculare make symmetrical synapses, largely with spines and spiny dendritic shafts of presumed pyramidal cells, in conjunction with asymmetrical synapses derived from the entorhinal cortex (Fig. 20; Sik et al., 1995; Katona et a]., 1996).

The correlation with morphologically identified cell types is

similarly straightforward for most of the SOM-containing cells, as in the dentate gyrus. In the CA1 region SOM cells correspond to the horizontal 0-LM cells, whereas in CA3 SOM cells correspond to the LM-projecting cells with multipolar dendritic tree. This correspondence was confirmed by the identical laminar distribution of dendritic and axonal arbors (Section III.3.2.a). The vast majority (87%) of calbindin-containing neurons located at the stratum oriens-alveus border also contain SOM; this corresponds to 32% of SOM cells in stratum oriens (Katona et al., 1996). This observation suggests that at least some of these SOM- positive cells project to the medial septum (T6th and Freund, 1992; T6th et al., 1993; Section VI.2).

IV.3.b. NPY-immunoreactive neurons The anatomical localization of NPY in the rat is less well

known than that of other neuropeptides. The types and distribution of neurons containing NPY mRNA were found to be the same as that of NPY-immunoreactive cells (Finsen et al., 1992). The GABAergic nature of NPY-containing neurons in the hippocampus has not been studied directly. However, the coexistence of NPY with SOM and other markers known to be present exclusively in GABAergic neurons provides circumstantial evidence that NPY-positive neurons contain GABA. The general distribution of NPY-immunoreactive cells in the hippocampus is very similar to that of SOM-positive neurons, but there are also important differences (Figs. 24,26; Chronwall et al., 1985; de Quidt and Emson, 1986; Kohler et al., 1986, 1987; Deller and Leranth, 1990; Milner and Veznedaroglu, 1992).

Dentate gym. The majority of NPY-immunoreactive neurons are in the hilus or within the row of granule cells bordering the hilus (i.e., base of the granule cell layer). A small number of multipolar neurons can also be found in stratum moleculare. The NPY-positive fusiform neurons in the subgranular zone and the large multipolar cells located deep in the hilus are similar to SOM- immunoreactive cells and do not send dendrites into the molecular layer. NPY-positive cells at the base of the granule cell layer have a pyramidal shape and a prominent apical dendrite that ascends through stratum moleculare often up to the hippocampal fissure or the pial surface. Their basal dendrites ramify within the hilus. This group is clearly different from SOM-containing neurons but similar to PV- or CCK-positive cells. The hilar NPY- positive cells with hilar dendritic arbors (i.e., the SOM-type) account for the majority (60-70%) of NPY-positive cells in the dentate gyrus (Types 1 and 2 cells of Kohler et al., 1987; Deller and Leranth, 1990). Those with “pyramidal basket cell” morphology represent 2O%, and multipolar cells in stratum moleculare constitute 7% of dentate NPY cells (Types 3 and 4 of Deller and Leranth, 1990).

The distribution of NPY-positive axons shows characteristic similarities and differences compared with that of SOM-positive fibers. A dense plexus is found in the outer one-third of the molecular layer, which gradually diminishes in density in the middle one-third. Only scarce labeling is observed in the inner one-third of the molecular layer, most of which are ascending main axons.


Terminal density is moderately high within the granule cell layer and even higher in the hilus (Deller and Leranth, 1990). Several main axons cross the hippocampal fissure. Axonal staining in the molecular layer is reminiscent of SOM immunostaining, whereas in stratum granulosum, and to some extent in the hilus, NPY- positive terminals are more numerous. The majority of fibers con- tributing to the axon plexus in the molecular layer likely originates from ipsilateral hilar neurons containing both NPY and SOM (Kohler et al., 1987). Some fibers may derive from the contralateral dentate hilus, as shown by retrograde labeling of hilar NPY-positive neurons (Deller and Leranth, 1990), which is similar to the SOM-positive cells of this region. These contralaterally projecting cells likely represent the same cell population containing both neuropeptides. The occurrence of fibers crossing the hippocampal fissure suggests that ipsilateral CA1 interneurons contribute to the NPY-positive axonal field in the molecular layer and the hilus of the dentate gyrus (back-projection neurons; Section III.3.2.e). The origin of NPY-positive axon terminals in the hilar region is uncertain; they may belong to axon collaterals of pyramidal-shaped NPY neurons in the granule cell layer or could partly derive from noradrenergic neurons in the locus coeruleus, some of which colocalize NPY (Wilcox and Unnerstall, 1990; Section V.4).

The ultrastructural features of NPY-positive neurons of all types are typical of interneurons, i.e., they have an infolded nu- clear membrane, thick perinuclear cytoplasm rich in organelles, and a mixture of symmetrical and asymmetrical synaptic inputs on the soma membrane. Dendrites in the hilus are covered by numerous asymmetrical synapses, most of which appear to originate from mossy fibers. The pyramidal-shaped NPY cells have a dendritic tuft in stratum moleculare, which is also covered by asymmetrical synapses. Combined anterograde-degeneration- NPY-immunostaining studies have demonstrated that some of these synapses originate from the entorhinal cortex and from the commissural-associational pathway (Deller and Leranth, 1990). NPY-positive axon terminals in all layers form symmetrical synaptic contacts (Deller and Leranth, 1990; Milner and Veznedaroglu, 1992). Their postsynaptic targets are somata and dendrites of presumed mossy cells in the hilus, occasional granule cell bodies in stratum granulosum, and dendritic shafts or spines of granule cells in stratum moleculare.

The correspondence between NPY-containing neurons and morphologically identified cell types is quite complex. SOM-positive hilar cells with hilar dendritic arbors project to the outer molecular layer and thus most likely correspond to HIPP cells. Direct evidence for the prediction that hilar NPY cells also correspond to HIPP cells has been recently provided by the immunocytochemical demonstration of NPY in an intracellularly filled HIPP cell (Fig. 21; Sik et al., submitted). Based on classical Golgi anatomy, the pyramidal-shaped NPY-positive cells at the base of the granule cell layer should be considered basket cells, but the moderate to low numbers of NPY-containing axon terminals in stratum granulosum argue against this interpretation. SOM-positive terminals are sparse in the hilus compared with the NPY-positive boutons. This suggests that NPY-positive cells that lack SOM (i.e., the pyramidal-shaped neurons) are likely to give

rise to the hilar NPY terminals. Perhaps these neurons represent a basket cell type with a predominant innervation of hilar mossy cells (see Section 1V.3.c) because NPY-positive terminal density in the hilus is higher than in the granule cell layer (but see Section V.4 about NPY-containing noradrenergic afferents).

Hippocampus. The majority of NPY-positive neurons in the CA1 subfield are located in strata oriens and pyramidale. Small groups of cells are also found in stratum radiatum. The cell density in strata oriens, pyramidale, lucidum, and proximal radiatum of the CA3 region is rather uniform. Stratum lacunosum-molecdare does not contain NPY-positive somata. Most of the NPY- immunoreactive neurons are partially stained, even in colchicine- treated animals, and only the most proximal dendrites can be visualized. However, there is a small set of cells in the CA1 and CA3 subfields that stain for NPY in a Golgi-like manner: the dendrites can be reconstructed almost to their natural ends, and their axons can also be followed occasionally, but in most cases it becomes myelinated (T.F. Freund, unpublished observations; also NOS-containing neurons, Section IV.4). The dendrites of some fusiform, horizontally oriented NPY-containing cells appear to arborize primarily in stratum oriens, whereas others may have a multipolar dendritic tree, in particular those in CA3, which do not respect laminar boundaries. However, for most neurons, the exact distribution of the dendritic trees in different layers is unknown. NPY-positive axon terminals were found in all layers, although at different densities. Strata lacunosum-moleculare, distal radiatum, and oriens contain a relatively large number of varicose collaterals, whereas in strata pyramidale and lucidum mostly nonvaricose main axons are observed at relatively low densities.

The ultrastructural features of NPY-positive neurons in the hippocampus are similar to that described for NPY-positive cells in the dentate hilus or for most other interneuron types (Milner and Veznedaroglu, 1992). A thick layer of cytoplasm, which contains large amounts of rough endoplasmic reticulum and Golgi cisternae, surrounds the infolded nucleus. The soma membrane receives numerous symmetrical and asymmetrical synaptic contacts similarly to NPY-positive dendritic shafts. NPY-positive axon terminals form symmetrical synapses with dendritic shafts, spines, and a small number of somata, most of which show characteristics of pyramidal cells.

The correspondence between morphologically identified cell types and NPY immunoreactivity is unequivocal only for those neurons that contain both NPY and SOM, i.e., the 0-LM cells (40-60% of all NPY-positive cells in the CA1 region; Kohler et al., 1987). As in the dentate gyrus, the noradrenergic projection from the locus coeruleus may also contribute to the numerous NPY-positive terminals in other layers, particularly in strata oriens, radiatum, and the hilus. 0-LM cells do also have collaterals in stratum oriens (Gulyis et al., 1993b; Sik et al., 1995), but it is unlikely that this cell type accounts for the relatively large number of NPY-positive boutons in this layer. Indirect evidence suggests that back-projection neurons in the CA1 region, which have dense axon terminal fields in the dentate hilus and CA3 stratum radiatum, may also contain NPY (NOS-containing cells, Section IV.4). This cell type also sends collaterals to strata oriens


and radiatum of both the CA1 and CA3 regions (Sik et al., 1995) and could be a source of NPY-positive axon arbors in these layers.

IV.3.c. CCK-containing neurons CCK is present in the brain largely in the form of an oc-

tapeptide, and antisera raised against this sequence have been widely used to visualize CCK-containing elements in several brain areas, including the hippocampus (Greenwood et al., 1981; Handelmann et al., 1981; Harris et al., 1985; Kosakaet al., 1985; Nunzi et al., 1985; Sloviter and Nilaver, 1987; Gulyis et al., 199 1 b). CCK-immunoreactive neurons occur in all layers and subfields of the hippocampus and dentate gyrus (Figs. 24, 26); they have nonpyramidal morphology and are immunoreactive for GABA or GAD (Somogyi et al., 1984; Kosaka et al., 1985). GABAergic neurons containing CCK represent approximately 10% of all GABAergic cells in the hippocampus, CCK-positive cells being most frequent in the CA1 region (12.5%) and somewhat less frequent in CA3 and the dentate gyrus (10% and 9%, respectively; Kosaka et al., 1985).

Dentate gym. The majority of CCK-positive neurons in the dentate gyrus are located within or just below the row of granule cell bodies bordering the hilus. They have a pyramidal-shaped soma, a prominent apical dendrite running toward the pial surface, and basal dendrites penetrating the hilus. The apical dendrites often branch when they reach the stratum moleculare but still remain radial in orientation. CCK-positive neurons are extremely rare in stratum moleculare and the deep hilar region. However, cells with a pyramidal-like or bitufted morphology can often be found in the subgranular polymorphic zone. Some cells in this region may have an initially horizontal orientation of the main dendrites, but they eventually turn at a right angle and penetrate strata granulosum and moleculare. Independent of their location, practically all CCK-positive neurons in the dentate gyrus have a bitufted dendritic tree, i.e., one tuft of dendrites (apical) in stratum moleculare, and another (basal) in the hilus, thus allowing them to receive inputs both from the perforant path and from the local mossy fiber collaterals. This neurochemically characterized cell type also has a characteristic distribution of axon terminals. There is a dense band of terminals in the upper one-third of the granule cell layer and the adjacent narrow (20-50 p m thick) zone of the molecular layer. Lower parts of the granule cell layer and the polymorphic zone are relatively poor in CCK-positive varicosities. The deep hilus also contains a larger number of terminals, most of which surround large unstained somata. With the exception of the narrow proximal zone, the molecular layer contains practically no CCK-immunoreactive axon terminals.

Electron microscopy of CCK-containing cell bodies and dendrites revealed no unique ultrastructural features that would distinguish them from other interneuron types. Their dendrites in the hilus receive numerous synapses from mossy fiber terminals, whereas those in stratum moleculare are contacted both by asymmetric and symmetric synapses (Leranth and Frotscher, 1986). CCK-positive boutons in stratum granulosum and in the adja-

cent region of stratum moleculare establish multiple symmetrical synaptic contacts with somata and proximal dendrites of presumed granule cells. They also occasionally contact the perisomatic region of other interneurons. The postsynaptic elements of terminals in the deep hilar area are somata showing ultrastructural features of mossy cells. Hilar neurons retrogradely labeled from the contralateral dentate gyrus were shown to receive synapses from CCK-positive terminals, suggesting that mossy cells with commissural-associational projections are among the targets of CCK-positive boutons (Leranth and Frotscher, 1986).

Comparison with morphologically established cell classes suggests that CCK-positive neurons in the dentate gyrus correspond to a specific subset of basket cells that have axon arbors in the upper granule cell layer and in the hilus. Interneurons with similar morphology have been intracellularly injected and visualized (Seay-Lowe and Claiborne, 1992; Sik et al., submitted), as described in Section 111.2.

Hippocampus. The distribution of CCK-immunoreactive neurons in the CA1 and CA3 regions is complementary to that of SOM- and NPY-positive cells. They occur in large numbers in stratum radiatum, less frequently in strata pyramidale and oriens, and only seldom in stratum lacunosum-moleculare. Most of the CCK-positive neurons, regardless of the laminar position, have a bitufted dendritic tree with a predominant radial orientation. Several of them have a pyramidal-shaped soma in stratum pyramidale, with an apical dendrite extending through stratum radiatum. Many others in stratum radiatum show an inverted pyramidal morpholgy, in which a major dendritic trunk runs toward stratum pyramidale, and a tuft of dendrites emerge in the opposite direction. Multipolar cells with no obvious orientation are frequently observed in strata radiatum and lucidum of the CA3 region. Axon terminal fields of CCK-positive neurons are largely limited to strata pyramidale and proximal radiatum in both the CA1 and CA3 regions, where they surround the somata and proximal dendrites of pyramidal cells. Apart from a few radially running collaterals in stratum radiatum, most of them having no varicosities, the other layers contain only negligible numbers of CCK-positive axon terminals. Thus, apparently all CCK-positive cells, whether located distantly near the stratum radiatum-lacunosum-moleculare border or within stratum pyramidale send their axon collaterals into stratum pyramidale to form perisomatic arrays of boutons that primarily surround pyramidal cell somata.

Electron microscopy of cell bodies and dendrites revealed no unique ultrastructural features that might distinguish CCK-positive cells from other types of interneurons in the hippocampus. CCK-positive axon terminals make symmetrical synapses with pyramidal cell bodies and occasionally with somata and dendrites of other interneurons (e.g., other CCK-positive cells), as shown in the rat and cat hippocampus (Harris et al., 1985; Nunzi et al., 1985).

According to the known axon termination site for CCK-positive terminals, all CCK-immunoreactive neurons can be identified as basket cells. This implies that basket cells are present also in distal stratum radiatum and in stratum oriens (for cat data, see Nunzi et al., 1985). Stratum radiatum contains no 1’V-positive


cells; thus, CCK-immunoreactive basket cells are PV negative (Gulyis et al., 1991b). Direct evidence for two neurochemically distinct types of basket cells has recently been provided (Acsidy et al., 1996a,b; see Section IV.3.d).

IV.3.d. VIP-immunoreactive neurons

The distribution, synaptic connectivity, and GABA immunoreactivity of VIP-immunoreactive neurons has been extensively studied in the hippocampus and dentate gyrus (Loren et al., 1979; Sims et al., 1980; Kohler, 1982, 1983; Leranth et al., 1984; Roberts et al., 1984; Kosaka et al., 1985; Sloviter and Nilaver, 1987; Acsidy et al., 1996a,b; Hijos et al., 1996). VIP- positive cells are present in all layers and subfields with an uneven distribution (Figs. 24, 26). They are of various morphologies, but all show characteristics typical of interneurons. The GABAergic nature of VIP-positive cells was first reported by Kosaka et al. (1985) by using antisera against GAD and VIP. They reported a colocalization of GAD and VIP in approximately 50% of the VIP-positive cells. In recent studies using pre- and postembedding double staining with antibodies against GABA and VIP, a much higher degree of colocalization has been reported. Practically all VIP-positive cells were found to express GABA immunoreactivity (Acsidy et al., 1996a,b; Hijos et al., 1996).

Dentate gyms. VIP-positive cells are present in all layers of the dentate gyrus, with a slightly higher frequency within or close to the granule cell layer. Smaller number of cells are found in the outer molecular layer and the deep hilus. The morphology and laminar distribution of dendrites are highly variable. The most frequent rype of VIP-positive cell has a bitufted dendritic tree, with a predominant radial orientation originating from somata in the inner molecular layer. These neurons always have a prominent upper dendritic tuft that extends toward the pial surface; the lower tuft, however, is sometimes missing or is limited to a single or a few long slender branches. Cells within or just below the granule cell layer may have similar morphologies or more often have a pyramidal-shaped soma with a rich basal dendritic arbor and a sparse apical dendritic tuft. In the deep hilus, a small number of cells are found with multipolar dendrites that may be very long and often cross stratum granulosum to enter the molecular layer.

The largest number of VIP-positive axon terminals is found in the hilus. Their density is higher in the ventral than in the dorsal hippocampus. The size of the boutons is relatively small, and they are diffusely distributed in the neuropil without an obvious association with cell bodies. Another type of VIP-positive axon is observed in stratum granulosum; it is studded with large varicosities and appears in clusters of varying diameters (100-200 pin). There are often long segments of granule cell layer without this type of axon arbor, but a cluster is always present at the apex of the dentate gyrus. A smaller number of VIP-positive varicose collaterals are observed in the molecular layer, where they often contact the somata or proximal dendrites of VIP-positive or other interneuron types (Hijos et al., 1996). Reconstruction of VIP-

positive neurons shows that the hilar terminal fields originate from bitufted cells located in the inner molecular layer. These cells have a main descending axon. Some VIP cells with soma and dendrites in the hilus also appear to contribute to the hilar axon collaterals. Several neurons that give rise to the clusters of large varicosities in stratum granulosum may have a soma-dendritic morphology and location similar to the cells with hilar projection.

The ultrastructure of cell bodies and dendrites of VIP-positive neurons is similar to that of most other interneuron types. The axon terminals immunoreactive for VIP always make symmetrical synapses and are positive for GABA. The postsynaptic targets of the large varicose axon type are granule cell bodies and dendrites in stratum granulosum (GABA-immunonegative profiles). The neurons that project to the hilus innervate mostly GABAergic interneurons, as was described in Section 111.5. The majority of targets of VIP-positive cells that project to stratum moleculare also appear to be other interneurons, although only short axon segments have been visualized in this layer.

For VIP-positive neurons, there is a clear correlation between input-output features and neurochemical characteristics in the dentate gyrus. The basket type of VIP cells contains CCK but not CR, whereas those projecting to the hilus are immunoreactive for CR but not for CCK (Hijos et al., 1996). The VIP-positive cells that project to the hilus have a target selectivity (Section III.5.c) complementary to that of some CCK-containing basket cells. Hilar collaterals of these CCK-positive basket cells predominantly innervate the mossy cells (Leranth and Frotscher, 1986; L. Acsidy and T.F. Freund, unublished observations), whereas the VIP-CR-positive cells with hilar projection innervate mostly GABAergic interneurons (specifically cells immunoreactive for SPR Section IV.6.0 in the same region. VIP-containing basket cells account for only a small proportion of CCK-positive basket cells in this region, and both the CCK- and the VIP-CCK- containing neurons represent basket cell populations that show no overlap with the PV-containing neurons (Gulyis et al., 1991b; Acsidy et al., 1996a).

Hippocampus. VIP-positive neurons are present in all layers of the CA1 and CA3 subfields, with higher frequency in strata pyramidale, lacunosum-moleculare, and the bordering radiatum. The somata are relatively small and fusiform, and the predominant dendritic orientation is vertical. The dendritic trees of cells in strata pyramidale, radiatum, and oriens are mostly bipolar or bitufted, and the majority appear to span all layers. The ascending branches often form a profuse tuft upon entering stratum lacunosum-moleculare, but those of another type with a larger soma in stratum pyramidale branch proximally and do not reach stratum lacunosum-moleculare. VIP-positive neurons in strata lacunosum-moleculare or distal radiatum usually have only an upper tuft, which is confined to the stratum lacunosum-moleculare.

Three major types of axon collaterals can be observed in the hippocampus, each originating from different VIP-positive neurons (Acsidy et al., 1996a,b). The first type innervates the horizontal 0-LM cells at the stratum oriens-alveus border, whereas the second type forms multiple contacts on interneurons in stratum radiatum (Section 111.5). Both of these types are known to

398 FREWND AND BUZSAKI

Hi ppocam pus LAYER-SPEC1 FIC

INPUT

PV CCK CB SOM VIP CR "PY)

Dentate gyrus

s.m.

S.Q.

hilus

entorhinal afferents

SchafYer collaterals (in CAI) a u r r e n t (only in CA3)/

commissural c o I I a t e r a I s

Massyfibers (only in CA3)

*recunent collaterals (restricted to this zone in CA1)

LAYER-SPECIFIC INPUT:

4

t

t

t

entorhinal affwentS

cornrnissural/ asscciational

afferents

Mossy fiber collaterals


FIGURE 26. Summary diagram shows the laminar distribution of dendritic and axonal arborizations of different types of calcium- binding-protein- and neuropeptide-containing interneurons in the hippocampus (top) and in the dentate gyrus (bottom). Filled circles mark the cell body location of each interneuron type, which gives rise to thick horizontal and/or vertical lines indicating the predominant orientation and laminar distribution of the dendritic tree. The hatched boxes represent the laminae where the axon of each interneuron typically arborizes. The vertical striped boxes indicate that other interneurons, rather than principal cells, are the primary targets. The transverse extension of the axons or dendrites are not indicated. Principal cells in the background provide an idea of which membrane domains (somatic, proximal, or distal dendritic regions) are innervated by the different interneuron types. The laminar distribution of different excitatory afferents (indicated on the right margin) often corresponds precisely to that of the axon arbor of specific interneuron type. Cell groups identified here on the basis of neurochemical marker content show a striking correspondence to the different types of interneurons categorized according to the morphological classification scheme as shown in Figure 17.

contain the calcium-binding protein CR and lack CCK. The third type gives rise to axon terminals in stratum pyramidale, which surround the cell bodies of pyramidal neurons. These cells contain CCK but not CR and have a proximally branching bitufted dendritic tree that does not reach stratum lacunosum-moleculare. Electron microscopy confirms that VIP-positive boutons in stratum pyramidale are positive for GABA and establish symmetrical synapses with somata and proximal dendrites of GABA-negative, presumed pyramidal cells (Acsidy et al., 1996a,b).

Thus, VIP-positive neurons in the hippocampus are of three morphologically and neurochemically distinct types, one of which is a basket cell that also contains CCK but not PV. This observation confirms that there are at least two neurochemically different types of basket cells: one contains CCK and VIP, and the other contains the calcium-binding protein PV. The remaining two types of VIP-positive cells do not contact pyramidal cells. Instead, they selectively innervate well-defined classes of other interneurons, either at the stratum oriens-alveus border (0-LM cells) or in stratum radiatum (Section 111.5).

IV.3.e. Enkephalin-immunoreactive neurons

Numerous studies have been carried out in the hippocampal formation to localize different opioid peptides, including members of the proenkephalin, prodynorphin, and beta-endorphin families. With regard to interneuronal localization, only the enkephalins appear to be of interest. Dynorphin A and B immunoreactivities were shown to be confined to granule cells and mossy fibers (Gall et al., 1981; Chavkin et al., 1985). Leu- and met-enkephalin (ENK)-immunoreactive neurons have been visualized and described in detail in the rat hippocampal formation (Gall et al., 1981). Immunoreactivity is not confined to incerneurons but is also present in dentate granule cells (Gall, 1988). ENK-positive interneurons occur primarily in strata radiatum and pyramidale of the hippocampus. They stain weakly for ENK, but colchicine treatment improves their visualization. The cells have fusiform somata and bipolar or bitufted dendritic trees, which run

largely parallel with the pyramidal cell dendrites. Another group of ENK-positive interneurons is present at the border of strata radiatum and lacunosum-moleculare and have horizontal or oblique dendrites running primarily in stratum lacunosum-moleculare. Axonal staining is observed in stratum lacunosum-moleculare of the CAI and CA3 regions and in the outer one-third of the dentate molecular layer. The origin of the latter is likely to be the entorhinal cortex. The axons of local ENK-positive cells are difficult to visualize. They ramify primarily in stratum radiatum, where their boutons are arranged in clusters around nonprincipal cell bodies and dendrites. In a recent study in the guinea pig, hamster and rat, all ENK-positive neurons in the CA1 region were also shown to contain GABA and VIP. However, only a subset of VIP-positive cells (i.e., those innervating interneurons in stratum radiatum) was also imunoreactive for ENK U.M. Blasco- Ibanez, F.J. Martinez-Guijarro, and T.F. Freund, unpublished observations). Electron microscopic and postembedding immunogold staining confirm that ENK-positive boutons are GABAergic, and make symmetrical synaptic contacts primarily on GABA-positive dendrites or somata U.M. Blasco-Ibanez, F.J. Martinez- Guijarro, and T.F. Freund, unpublished observations).

Thus, we conclude that ENK-containing neurons represent a subset of VIP-positive cells, which selectively innervates other interneurons in strata radiatum and oriens and has a dendritic tree that either spans all layers or arborizes predominantly in stratum lacunosum-moleculare.

IV.3.f. Neurokinin-immunoreactive neurons

SP is one of the neurokinins that has been reported to occur in hippocampal neurons in the monkey and in the rat (Ljungdahl et al., 1978; Roberts et al., 1984; Davies and Kohler, 1985; Iritani et al., 1989). SP immunoreactivity is largely restricted to varicose axons at the border of the hilus and stratum granulosum and in stratum pyramidale of the CA1 and CA3 regions. They are particularly dense in strata pyramidale, radiatum, and oriens of the CA2 and CA3a areas. At least some of these fibers are likely to have a local origin because after colchicine treatment (Davies and Kohler, 1985) or acrolein fixation (Zs. Borhegyi, L. Seress, and C. Leranth, unpublished observations) SP-positive cell bodies become visible in strata oriens and pyramidale of the CA1 and CA3 subfields and at the stratum granulosum-hilus border. The strongly stained axons in the CM-CMa region are likely to have an extrinsic origin; they apparently reach the region via the fimbria-fornix and disappear after fimbria-fornix transsection (Zs. Borhegyi and C. Leranth, unpublished observations). Axons of local origin mostly occur within or adjacent to cell body layers and form symmetrical synapses. Numerous SP-positive interneurons, most of which were shown to correspond to SOM-containing cells in the hilus of the dentate gyrus, have been visualized in the monkey by using acrolein fixation (Seress and Leranth, 1996).

Neurokinin B (NKB) is also present in a small number of hippocampal interneurons. Rather little information is available, however, about their morphology, distribution, GABA content, and correspondence to other interneuron types. Large pyramidal-


shaped somata are often found to be immunoreactive for NKB at the base of the granule cell layer, and interneurons with short dendritic segments can also be visualized in strata pyramidale, radiatum, and lacunosum-moleculare of the CAI subfield. Faint NKB immunoreactivity is also present in mossy fibers, which can be upregulated by seizure activity (Marksteiner et al., 1992).

IV.4. Neuronal NO Synthase (or NADPH- diaphorase) Is Selectively Present in Interneurons in the Hippocampal Formation

NO represents a new class of neuronal messengers. I t is a water- and lipid-soluble gas, which freely diffuses across cellular membranes and the extracellular space. It is likely to have both short- and long-term modulatory effects on neurons mostly by influencing cGMP formation (Garthwaite, 1991; Snyder, 1992). The neuronal (nNOS), endothelial (eNOS), and inducible (iNOS) forms of N O synthase, the enzymes synthesizing NO from arginine, are present in the brain (Paakkari and Lindsberg, 1995). In the hippocampal formation, the endothelial form is localized in pyramidal cells (Dinerman et al., 1994), whereas nNOS immunoreactivity appears to be restricted to interneurons, which accounts for practically all staining obtained by NADPH-diaphorase enzyme histochemistry (Hope et al., 1991; Vincent and Kimura, 1992; Valtschanoff et al., 1993; Dun et al., 1994). All neurons showing NADPH-diaphorase reactivity were shown to be immunoreactive for GABA, thus providing direct evidence that they represent interneurons (Valtschanoff et al., 1993).

Dentate gyrus. Interneurons immunoreactive for nNOS (or staining for NADPH-diaphorase) in the dentate gyrus are most abundant in the hilus (Fig. 23). Numerous cells are also present in stratum granulosum, where they are concentrated near the hilar border (Valtschanoff et al., 1993). Cells in the deep hilus have generally large cell bodies and bitufted or multipolar dendritic trees, which are moderately spiny, and rarely enter the granule cell layer. Large hilar cells at the border of the granule cell layer display the typical morphological features of basket cells known from Golgi and immunocytochemical studies (CCK, PV). They have prominent apical dendrites crossing stratum moleculare and several basal dendrites in the hilus. Cells in stratum granulosum have smaller, fusiform, or pyramidal-shaped somata and a radially running apical dendrite, which forms a tuft that often reaches the fissure or the pial surface. Basal dendrites also emerge from the soma and run in the subgranular polymorph zone, emitting only a small number of branches. Multipolar or radially oriented bitufted cells are also present in stratum moleculare. Dendrites of the multipolar and radial bitufted cells are varicose and spine free. The staining intensity of small neurons is generally weaker than that of large cells.

Axonal staining is remarkably dense in stratum moleculare, especially in the outer two-thirds, whereas the inner one-third displays a strong homogeneous “background (fine granular) staining. Thick main axons originating from nNOS-positive cells in CA1 often cross the hippocampal fissure to terminate in the dentate hilus (Sik et al., 1994). The hilus contains a smaller number of stained axons. Collaterals in all layers are densely varicose.

Hippocampus. A large number of interneurons immunoreactive for nNOS (or staining for NADPH-diaphorase) are found also in the CAI and CA3 regions, with an uneven distribution, heterogeneous arborization patterns, and staining intensity (Figs. 22, 23). They are most abundant in strata radiatum and pyramidale. Moderate numbers are present in stratum oriens and only a few in stratum lacunosum-moleculare. The dendritic trees of the faintly labeled cells are radially oriented bipolar or bitufted and originate from a relatively small ( 1 0-1 5 p m in diameter) fusiform or round soma. There is a congregation of this cell type in CA3a and CA3b stratum radiatum, in stratum oriens at the CA3a-CA2 border, and in stratum radiatum of the CA1 region, near the subicular border.

NOS-positive cells of the other group, i.e., those with large somata (20-35 p m in diameter) and intense staining of the entire dendritic tree, are located mostly in stratum radiatum, near stratum Iacunosum-moleculare. These cells are typically found close to the subicular border. Another common location of these cells is stratum oriens of the CA1 region, where they have largely horizontal dendrites. A few branches often penetrate through stratum pyramidale to enter stratum radiatum. Neurons in stratum radia- turn give rise to a wide upper tuft of dendrites, with long horizontal branches that penetrate stratum lacunosum-moleculare. Only one or two primary dendrites emerge toward the alveus. Similar strongly stained large neurons are found occasionally in stratum pyramidale, with a bitufted dendritic tree characteristic of interneurons.

Dense axon arborizations can be seen in stratum lacunosum- moleculare and to some extent in distal stratum radiatum. Except for the main axon trunks, which usually course radially, all collaterals are densely varicose, with a predominant horizonxal orientation. Varicose fibers are sparse but consistently present in other layers. Axons can be followed only from the strongly stained large neurons. Neurons in stratum oriens have axon collaterals directed toward stratum lacunosum-moleculare. Others may have a descending initial segment, but the main axons immediately curve back and give rise to horizontal collaterals that join the dense plexus in stratum lacunosum-moleculare. Several branches cross the hippocampal fissure and terminate in the hilar region (Sik et al., 1994).

Colocalization with other markers. In the dentate gyrus, the majority (95%) of hilar SOM-containing neurons were also shown to be immunoreactive for NOS. In the same region, only a negligible number (5%) of PV-positive cells contained NOS immunoreactivity (Dun et al., 1994; Divila et al., 1995). NOS-positive cells located at the base of the granule cell layer have pyramidal-shaped somata and prominent apical dendrites that penetrate stratum moleculare. These cells are never immunoreactive for SOM. Although they have morphological features typical of pyramidal basket cells, they lack PV.

In the hippocampus, the correspondence between NOS-containing and other neurochemically defined cell groups is even less clear. The majority of SOM- and PV-positive cells in stratum oriens of the CA1 and CA3 regions appears to be negative for NOS (Dun et al., 1994; Divila et al., 1995). Because there are no data available on the coexistence of NOS with other neuropeptides and calcium-binding proteins, one can only speculate


about the identity of these cells on the basis of morphology and laminar distribution. The large intensely stained NOS-positive cells in distal stratum radiatum or in strata pyramidale and oriens may correspond to a subset of NPY-positive cells, which project primarily to stratum lacunosum-moleculare. Their axons may cross the hippocampal fissure to terminate in the dentate gyrus. The small faintly stained fusiform cells with bipolar or bitufted dendritic tree in strata radiatum and pyramidale may correspond to several interneurons types, e.g., VIP-, CR-, CCK-, and CB- positive neurons (Fig. 22). Double-labeling studies are required to establish the precise correspondence of NOS-positive cells with any of these cell groups.

IV.5. Interneuron-Specific Cell Surface Glycoproteins Revealed by Lectin Binding or Immunocytochemistry

A cell surface molecule characteristically present on nonpyramidal cells of the cerebral cortex is N-acetylgalactosamine. This molecule is covalently bound to glycoproteins and can be selectively visualized by the binding with lectins (Woodward et al., 1985; Nakagawa et al., 1986), such as Kcia villosu agglutinin ( W A ) or soybean agglutinin (SBA). Monoclonal antisera have also been made against cell surface markers, and some of them, e.g., VC1.l, VC5.1 (Arimatsu et al., 1987; Naegele et al., 1788), and 4F4 (Yamamoto et al., 1988), appear to recognize glucuronic acid-containing glycoproteins specifically associated with some types of GABAergic neurons. Each of these markers label only a subset of GABAergic neurons. This raises the possibility that functionally distinct inhibitory neurons that differ in cell surface molecules can be distinguished. These subgroups may be differentially visualized even in living slices (Naegele and Katz, 1990; Drake et al., 1991). To this end, several studies have been done to investigate whether any of the calcium-binding-protein- or neuropeptide-containing GABAergic cell types specifically correspond to interneurons that express any of these cell surface markers. The majority of the data have been produced with W A binding, which can easily be combined with immunocytochemistry. These studies demonstrate that in the hippocampal formation the majority of WA-labeled cells are GABAergic. The only principal cell population strongly marked by the lectin is a group of CA2 pyramidal cells. Most of the labeled interneurons correspond to PV-containing cells both in hippocampus (Drake et al., 1991) and in neocortex (Kosaka and Heizmann, 1989), whereas SOM-, CCK-, and CB-immunoreactive neurons were not labeled. Quantitative data are available from the subicular region, where more than 90% of WA-labeled cells are PV-immunoreactive, but only 34% of all PV-immunoreactive cells are WA-positive (Drake et al., 1971). In the neocortex, the relationship is similar: approximately 90% of the W A - or SBA-labeled neurons are also PV-positive, and the PV-positive cells considerably outnumber the lectin-labeled cells (Kosaka and Heizmann, 1989).

JV.6. Neurotransmitter Receptors Enriched on the Surface of Specific Interneuron Types

In this section, only the distribution of immunocytochemi- cdly detected transmitter receptors (which show a characteristic

association with various interneuron subpopulations) will be described. These data should not be taken as evidence for particular transmitter actions on those interneuron types because several examples of transmitter-receptor mismatch can be quoted from the literature, and there is no guarantee that those receptors are utilized under physiological conditions. Nevertheless, these immunocytochemical descriptions can be used in physiological and pharmacological studies to predict interactions between transmitters via given receptors and may also serve as a guide for using some of these receptor antisera as markers of specific interneuron dendrites for connectivity studies.

IV.6.a. Glutamate receptors

AMPA receptors. Antisera against the four types of subunits (GluR1-4 or GIuRA-D) of this glutamate receptor class have been produced recently and applied successfully to study cellular and subcellular distribution (Petralia and Wenthold, 1992; Martin et al., 1993; Molnar et al., 1993; Baude et al., 1995; Leranth et al., 1996). In the context of interneurons, the GluR2/3 and GluR4 subunits are of particular interest, whereas GluRl appears to have a rather uniform distribution in all cell types. Inimunoreactivity for GluR4 is weak in pyramidal and granule cell dendrites, but it produces an intense staining of PV-, CB-, CR-, and most notably of SOM-containing interneurons (Fig. 25; Leranth et al., 1996). By using another polyclonal antiserum against the GluR4 subunit, Baude et al. (1995) showed a particularly strong staining of interneurons in strata pyramidale and oriens of the hippocampus. In contrast, GluR213 immunoreactivity is present in all principal cells, but appears to be absent from interneurons (Fig. 25), with the exception of some CB-containing cells in stratum radiatum of the CA3 subfield (Leranth et al., 1996). This finding may be of particular interest because amino-3-hydroxy-5-methyl-4-isox- azoleproprionic acid (AMPA) channels that lack the G l u m subunit were suggested to be permeable to calcium (Hollmann et al., 1991; Burnashev et al., 1992), and cells equipped exclusively with this type of AMPA receptor may be predisposed to calcium-dependent excitotoxic processes (Section XIV).

Metabotropicglutamate receptors. The mGluRl a subunit of the metabotropic glutamate receptors is particularly enriched in a well-defined type of GABAergic interneuron in the hippocampus. Several immunocytochemical studies show a striking staining of horizontally running dendrites at the stratum oriens-alveus border (Martin et al., 1792; Gorcs et al., 1993), but the types of immunoreactive structures remain unidentified. Subsequently, Baude et al. (1993) provided direct evidence that all mGluRla- positive neurons in this location were also immunoreactive for somatostatin, the marker of horizontal 0 -LM cells (Section III.3.2.a). The staining of the dendritic tree of mGluRla-positive cells is nearly complete and can be used to study the synaptic input of this particular cell type (Blasco-Ibanez and Freund, 1995), which has been impossible previously due to the lack of dendritic staining with SOM antisera. At the electron microscopic level, mGluRla receptors are localized perisynaptically around the edges of the postsynaptic density of asymmetrical (presumed glutamatergic) synapses (Baude et al., 1993). The ultrastructural lo-


calization of mGluR5 is similarly perisynaptic, but this receptor is present on both pyramidal and interneuron dendrites and dendritic spines (Lujan et al., 1996).

IV.6.b. GABA receptors Because markers for studying the distribution of GABA-B re-

ceptors at high resolution are not available, the present description is restricted to GABA-A receptor subunits and their relationship to interneuron types. Interneurons in the hippocampal formation have been shown to express high levels of GABA-A receptors (Richards et al., 1987; Houser et al., 1988; Zimprich et al., 1991; Nusser et al., 1995). Antisera against the alphal and the beta 2\3 subunits provide a particularly strong staining of interneurons. In a recent double-labeling study (Gao and Fritschy, 1994), a remarkable selectivity was observed in the immunoreactivity of interneuron types for GABA-A-alphal. AH PV-positive, 50% of the CR-positive, most of the NPY-positive, and a subset of the SOM-positive interneurons showed immunoreactivity, whereas the CB-, VIP-, and CCK-containing neurons were negative for GABA-A-alphal. Double in situ hybridization in the dentate gyrus confirms that all neurons strongly labeled for the alphal subunit also contain GAD65 mRNA, but none of them show labeling for the pre-prosomatostatin mRNA probe, which is present in a population of hilar neurons (Esclapez et al., 1996). Evidence for the GABAergic nature of cells strongly labeled with the beta 2\3 subunit is also available (Nusser et al., 1995).

In a recent study, postembedding immunogold staining was employed to localize the alphal and beta 2\3 subunits at the subsynaptic level in the dentate gyrus (Nusser et al., 1995). The results provide direct evidence that GABA-A receptor subunit immunoreactivity is enriched at subsynaptic membrane compartments of the target elements. Furthermore, no qualitative difference was found among chandelier cell, basket cell, and dendritic inhibitory cell synapses in terms of receptor subunit type or density on the postsynaptic membrane. This finding argues against the hypothesis that GABA-B receptors occupy most of the distal dendrites and confirms paired recording data demonstrating that IPSPs elicited by basket, chandelier, and dendritic inhibitory neurons on single postsynaptic pyramidal cells are mediated by GABA- A receptors (Buhl et al., 1994a; Miles et al., 1996). Nevertheless, it should be remembered that IPSPs evoked by strong stimuli to activate dendritic inhibitory fibers always have a GABA-B component, whereas those evoked by strong stimuli in somatic regions do not (Alger and Nicoll, 1979, 1982a,b; Miles et al., 1996).

IV.6.c. Monoamine receptors No immunocytochemical data have yet been published on the

localization of noradrenergic receptors in the hippocampal formation. However, two types of serotonergic receptors have now been visualized, both of which have important implications for GABAergic interneuronal function.

5-bydroxy-tryptamin-2 receptors (5HT-2). An antiserum against an N-terminal epitope of the 5HT-2 receptor has been developed

recently and used successfully in immunohistochemical studies (Morilak et al., 1993). Although double-labeling experiments have not been reported, the laminar distribution, morphology, and frequency of the immunoreactive cells in the hippocampal formation and neocortex suggests that GABAergic interneurons are selectively labeled (Morilak et al. 1993). In the dentate gyrus, 5HT-2-positive neurons with large, fusiform somata are present in the hilus, often in the region just below the granule cell layer. In the hippocampus, the majority of 5HT-2-positive cells are found in strata pyramidale and oriens. They have a large soma and bitufted or multipolar dendritic trees with a predominant radial orientation. The number of cells is larger in the dorsal than in the ventral hippocampus, and the CA3 region contains more 5HT-2-immunoreactive cells than the CA1 region, although there is a marked increase in cell density near the subicular end of the CA1 region even in stratum radiatum (Morilak et al., 1993). It is difficult to predict the type(s) of neurons expressing 5HT-2 on the basis of limited morphological data. According to their distribution and soma size, they may include NPY-SOM-containing cells, and/or some PV-positive neurons.

5HT-3 receptors. Both immunocytochemical and in situ hybridization data are available for the localization of 5HT-3 receptors in the hippocampal formation. The first description was provided by Tecott et al. (1993), who reported on the clistribu- tion of 5HT-3 mRNA. In the hippocampus, labeling appears to be clearly confined to interneurons, which have a distribution complementary to that of 5HT-2-containing interneurons. 5HT- 3-positive cells are more frequent in the ventral than in the dorsal hippocampus, and they occur predominantly in strata lacunosum-moleculare and radiatum and less frequently in strata oriens or pyramidale. In the dentate gyrus, the predominant cell type is the interneuron at the granule cell layer-hilar border, which has a large, often pyramidal-shaped, soma.

Immunohistochemical studies to date have been reported only in abstract form (Battenberg et al., 1994; Morales et al., 1994). They confirm the distribution data derived from in situ hybridization and provide evidence for the colocalization with GABA. Furthermore, 5HT-3 mRNA is not expressed in SOM- immunoreactive neurons in the hippocampus. This finding supports the idea of complementary distribution of 5HT-2 and 5HT- 3 receptors among interneurons because 5HT-2 receptors are likely to be localized in SOM-NPY-positive cells. O n the basis of cell distribution and morphology, the 5HT-3-containing interneurons may include the CB-, VIP-, CCK-, and CR-immunoreactive subpopulations. Double-labeling studies are re-- quired to establish which of these cell types indeed express 5HT-2 or 5HT-3 receptors. The striking correlation of receptor localization with electrophysiological data is described in Section IX.5.

IV.6.d. Acetylcholine receptors Several lines of morphological data demonstrate that both ma-

jor classes of cholinergic receptors (muscarinic and nicotinic) are present in high concentrations in hippocampal interneurons. However, the types of interneurons expressing a particular re-


ceptor type or subunit are still largely unknown, and therefore we have to be satisfied at this point with indirect correlations.

Mzlscarinic receptors. The first detailed immunocytochemical study of the distribution of four different muscarinic receptor types was published by Levey et al. (1995). Three of these receptors (ml, m3, and m4) are expressed primarily by principal cells, whereas m2 appears to be present predominantly in interneurons. The majority of m2-positive cells in the CA1 region are located in stratum oriens and form a characteristic cluster of neurons, all having a bipolar or bitufted horizontal dendritic tree restricted to the stratum oriens-alveus border. Occasionally, single dendritic branches can be seen in stratum radiatum. In the CA3 region, the m2-positive neurons are multipolar and are present in strata oriens, pyramidale, and radiatum, with most cells in stratum oriens. Interestingly, stratum lucidum appears to be devoid of immunoreactive cell bodies or dendrites. Only a few dendrites cross this layer without emitting any branches. Dendrites immunoreactive for m2 are practically absent from stratum lacunosum-moleculare in the CA1 and CA3 regions. In the dentate gyrus, a small number of multipolar neurons are scattered within the hilus, with their dendrites restricted to the hilar region.

Axonal staining in the hippocampus is particularly strong in stratum pyramidale, where the terminals of both basket and chandelier cells are immunoreactive for m2 and form a plexus around the somata and axon initial segments of pyramidal cells (Hijos et al., submitted). Throughout the hippocampus, particularly in CAI-CA3 stratum oriens, somata of some interneurons are covered by large boutons of m2-positive fibers. Occasionally, the surrounded cells themselves are also m2 immunoreactive. Some fibers involved in the basket formation around interneurons may belong to basket cells, which have occasional collaterals in stratum oriens, but an extrahippocampal origin cannot be excluded. In addition, there are several collaterals with different orientations in proximal strata radiatum and oriens that can be followed back to the horizontal cells at the stratum oriens-alveus border, although most of the main axons are lost due to myelination (Hijos et al., submitted). Thin, horizontally running axons, faintly stained for m2, are present at the border of strata radiatum and lacunosum-moleculare and often appear to cross the hippocampal fissure. These faint axons are particularly numerous in stratum radiatum of CA3a-CA3b. Some of the reconstructed cells at the stratum oriens-alveus border had occasional collaterals in this zone; however, due to myelination and partial reconstruction, it is impossible to determine whether they account for all the observed axons. Drumstick-like boutons were often seen on m2-positive axon collaterals in all layers and subfields. In the dentate gyrus, m2-positive axons are relatively sparse and occur in largest numbers in the hilus. In the outer stratum moleculare, the axons are faintly stained, whereas at the border of strata granulosum and nioleculare intensely stained fibers are also found. The m2-positive horizontal neurons at the CAI stratum oriens-alveus border are very similar to the SOM-containing 0-LM cell population. Efowever, imrnunostaining for SOM and m2 labels nonoverlapping cell populations (Eijos et al., submitted). According to the

laminar distribution of those few m2-positive axon arbors, which could be traced for a reasonable distance in spite of the heavy myelination, the m2-positive horizontal cells at the stratum oriens-alveus border are likely to correspond to the horizontal trilaminar cells or back-projection neurons (Sections III.3.2.b, III.3.2.e).

Nicotinic receptors. Hippocampal interneurons are preferentially labeled by imrnunostaining against the beta2 subunit of nicotinic receptors and by radiolabeled alpha-bungarotoxin, one of the selective ligands of nicotinic receptors (Freedman et al., 1993; Hill et al., 1993). The general distribution of interneurons expressing the receptor was similar in the two studies. They are numerous at the stratum oriens-alveus border and in strata radiatum, lucidum, and lacunosum-moleculare of the hippocampus. In the dentate gyrus, most labeled cells are found at the hilar and molecular layer borders of stratum granulosum and less frequently in the hilus. As expected from their location, most of these neurons are immunoreactive for GABA and constitute approximately one-fifth of all GABA-immunoreative neurons in the dentate gyrus (Freedman et al., 1993). Colocalization of alpha- bungarotoxin binding and different neuropeptides demonstrate that no single neurochemically identified subpopulation can account for the “nicotinic receptor-positive’’ neurons because CCK, SOM, and NPY are all present in some radiolabeled cells (Freedman et al., 1993).

IV.6.e. Opiate receptors

Powerful opiate effects on hippocampal inhibition have been repeatedly demonstrated in electrophysiological and pharmacological experiments (Section IX.6), but the morphological and neurochemical types of interneurons mediating these effects are still largely unknown. Immunohistochemical studies on the localization of opiate receptors are scarce, and apparently none of those have revealed any enrichment in interneurons of the hippocampal formation. However, naloxone, a ligand of mu-type opiate receptors, binds selectively to interneurons in certain layers and subfields of the hippocampus (A.I. Gulyis and R. Miles, personal communication).

W.6.f. SPRs A specific antibody against SPR has been developed and has

been reported to label interneurons in the cerebral cortex of the rat (Kaneko et al., 1994; Nakaya et al., 1994). In the hippocampus, immunostaining for SPR labels numerous morphologically heterogeneous interneurons in all subfields (Nakaya et al., 1994). Principal cells (granule, mossy, and pyramidal cells) are consistently devoid of SPR immunoreaction. However, pyramidal cells of the ventral subiculum show intense SPR immunostaining. The immunoprecipitate is membrane bound and outlines cell bodies and proximal and distal dendrites in a Golgi-like manner. In spite of the detailed visualization of dendritic morphology, the classification of SPR-positive interneurons is very difficult due to the lack of axonal staining. However, colocalization with neurochemical markers that label interneuron types with known affer-

TAB

LE 1.

Ove

rlap

Am

ong

Difi

eren

t N

euro

chem

ical

Mar

kers

of

Hip

poca

mpa

l In

tem

euro

nsl

CR

C

R

non-

V

IP-

VIP

G

ABA

sp

iny

spin

y CB

PV

N

OS

CCK

IC

S ba

sket

SO

M

NPY

SP

R

mG

lu

M2

GA

BA

CR s

piny

12

CR n

on-s

piny

12

CB

PV

NO

S CC

K

VIP

-ICS2

' V

IP b

aske

t2'

SOM

N

PY

SPR

Z3

mG

lu

M2I

1

11

.1~

~3

10

0~

~7

96

%15

97.1

Yn4

+ + +2

4

97%

5 10

0%

100%

9O

Yo7

89

Y0

~~

-9

0°/04

2 93

%

13Yo

2

6.2%

12

0%

0%

S%47

0%

++

29

0%

0 Yo

SY

P

0%

7%

25%

20

~~

43

-11y

O3

S.1%

16

+ 21

0~

~3

0

0 Yo

32Y0

3'

6%

4%

22.4

Y04

0 Yo

0%

+ 1' + 22

0%

0%

+ 33

+ +3'

3%

++

47

10

.5%

' 1

0%

~~

5.6%

19

- < 1

~~

47

+ 31

92%

+ +

+33

+ 39

3%

-9%

6 1%

6 14

Y07

100%

2O

o%l4

+ 2

0

+ 22

+ + + /

+25

11

~~

~~

+ +34

+ 44

+ +40

3%

2%

10

0~

~3

6

+ +4

6

+ 23

10

0%

~~

o

w4

10

0~

4

++

34

+

+3

5

10

0%

~~

+ +

45

+ +41

'Val

ues

in e

ach

row

ind

icat

e w

hat

port

ion

of n

euro

ns c

onta

inin

g a

give

n m

arke

r (i

ndic

ated

at

the

begi

nnin

g of

the

row

) ar

e al

so im

mun

orea

ctiv

e fo

r th

e m

arke

rs i

ndic

ated

at

the

top

of

each

col

umn.

The

deg

ree

of ov

erla

p is

giv

en i

n pe

rcen

tage

s, w

here

acc

urat

e da

ta a

re a

vaila

ble.

How

ever

, in

sev

eral

cas

es, o

nly

a se

miq

uant

itativ

e es

timat

e ca

n be

pro

vide

d, w

here

0%

m

eans

no

over

lap;

+ m

inim

al o

verl

ap;

+ +, m

ediu

m o

verl

ap;

+ + +,

str

ong

over

lap;

100

%, t

otal

ove

rlap

. The

sym

bol - m

eans

est

imat

ed v

alue

s fr

om th

e ra

tio o

f co

exis

tenc

e wit

h ot

her

mar

kers

. If

the

perc

enta

ge o

f ov

erla

p co

nsid

erab

ly d

iffe

rs a

mon

g hi

ppoc

ampa

l su

bfile

ds, t

his

is in

dica

ted

in th

e no

tes.

'M

ietti

nen

et a

l. (1

992)

. 3E

stim

ated

from

CB

/PV

rat

io b

ased

on

Gul

yas

et a

l. (1

991b

). "O

saka

et

al.

(198

7).

"Osa

ka

et a

l. (1

985)

. "E

stim

ated

from

Kos

aka

et a

l. (1

985)

and

Acs

idy

et a

l. (1

996a

); V

IP/C

CK

cel

ls a

re b

aske

t cel

ls.

7Kos

aka e

t al.

(198

8b);

30%

of G

AD

cel

ls in

the

hilu

s an

d C

A3-

1 or

iens

are

SO

M, 5

10

% in

CA

1 an

d C

A3

stra

tum

pyr

amid

ale,

and

2%

in th

e gr

anul

e ce

ll la

yer.

8Ded

uced

from

dat

a on

SO

M-N

PY c

oloc

aliz

atio

n, i.

e., S

OM

-NPY

cel

ls re

pres

ent

the

sam

e pr

opor

tion

of S

OM

cel

ls a

nd o

f N

PY c

ells

, whi

ch m

eans

that

the

ir r

atio

is -

1:1.

9E

stim

ated

from

CR

/SPR

rat

io f

rom

Acs

ady

et a

l. (s

ubm

itted

).

"Tha

t is

, mG

lu c

ells

per

fect

ly o

verl

ap w

ith S

OM

cel

ls (B

aude

et a

l., 1

993)

. 'lH

&jo

s et a

l. (s

ubm

itted

).

lLM

ietti

nen e

t al.

(199

2); t

he 5

% o

verl

ap o

f CB

and

CR

mig

ht a

rise

fro

m th

e sl

ight

cro

ss re

activ

ity o

f th

e CR

ant

ibod

y w

ith C

B u

sed

in th

is s

tudy

. I3

The

lack

of

GA

BA

im

mun

orea

ctiv

ity i

n th

ese

cells

may

be

due

to th

e ra

ther

ext

ensi

ve a

xon

abro

biza

tion

of t

hese

cel

ls (m

any

of t

hem

als

o ha

ve c

omm

issu

ral a

nd/o

r cr

oss-

regi

onal

pro

- je

ctio

n).

1-

''Acs

6d>

et a

l. (i

n pr

ess)

. 15

Val

ues o

nly

for r

egio

ns th

at d

o no

t con

tain

spi

ny c

ells

. l6

Va1

ue is

for

spi

ny/n

onsp

iny

cells

cum

ulat

ivel

y.

17T6

th an

d Fr

eund

(19

92).

1sG

uly5

s et a

l. (1

991b

); 0%

exc

ept i

n C

A1

stra

tum

ori

ens,

whe

re 2

1% o

f CB

cel

ls c

onta

in P

V.

19G

ulya

s et a

l. (1

991b

); th

ere

are

smal

l dif

fere

nces

bet

wee

n ar

eas

and

laye

rs.

2oK

aton

a et a

l. (1

996)

; 87%

in C

A1

stra

tum

ori

ens

and

0% in

oth

er a

reas

. 21

Gul

yas e

t al.

(199

1b);

O"/o

exce

pt in

CA

1 st

ratu

m o

rien

s, w

here

9.6

% of

PV

cel

ls c

onta

in C

B.

22D

un et

al.

(199

4).

23A

~sii

dy et a

l. (1

996;

SPR

), 90

% in

DG

and

5%

in

othe

r ar

eas.

24

Val

tsch

anof

f et a

l. (1

993)

; 100

% ex

cept

for

som

e ce

lls in

str

atum

pyr

amid

ale,

whi

ch m

ake

up 5

% of

all

NO

S ce

lls.

25D

un et

al.

(199

4); h

ilus/

othe

r ar

eas.

26

Gul

yas e

t al.

(199

1b).

27K

osak

a et a

l. (1

985)

, Acs

ady

et a

l. (1

996a

); 10

0% fo

r VIP

bas

ket c

ells

and

+ fo

r ot

her

VIP

cel

ls.

28A

csad

y et a

l. (1

996a

). 2y

VIP

type

3: lo

o%, V

IP t

ype

2a: 4

1%, V

IP ty

pe 2

b: 0

%.

300n

e per

cent

in

CA

1 an

d 0%

in

CA

3 fo

r bo

th V

IP p

opul

atio

ns.

31V

IP ty

pe 3

: O%,

V

IP ty

pe 2

a: 1

370,

VIP

type

2b:

0%

. 32

Kat

ona e

t al.

(in

pres

s).

33D

un et

al.

(199

4).

34K

ohle

r et a

l. (1

987)

; a r

atio

of

appr

oxim

atel

y 1 ca

n be

ded

uced

for

SO

M/N

PY.

"Acs

ady

et a

l. (s

ubm

itted

); 7

3% in

hilu

s an

d 8%

in C

A1.

36

Bau

de et

al.

(199

3).

37Th

e rem

aini

ng 1

1% w

ere

spin

y ce

lls in

hilu

s, p

ossi

bly

GA

BA

ergi

c neu

rons

with

dis

tant

pro

ject

ion,

whi

ch h

ave

low

som

atic

GA

BA

lev

els.

38

Fift

y-ni

ne pe

rcen

t in

DG

and

5%

in C

A3-

I. 3y

Four

to 1

9%; p

erce

ntag

e di

ffer

s am

ong

area

s.

400n

e hun

dred

per

cent

in h

ilus

and

+ in

CA

3.

41T

wen

ty-e

ight

perc

ent i

n D

G, 9

2% in

hilu

s an

d lu

cidu

m, a

nd 3

6% el

sew

here

. 42

Ded

uced

from

val

ue f

or S

OM

. 43

Fift

een p

erce

nt in

CA

1, 3

6% in

CA

3 st

ratu

m o

rien

s, 1

6% in

CA

3 st

ratu

m r

adia

tum

, and

19'4

~ in

the

hilu

s of

DG

. 44

0ne p

erce

nt a

nd 2

% i

n C

A3

and

CA

1 an

d 0%

in D

G a

nd h

ilus.

45

Acs

ddy e

t al.

(sub

mitt

ed);

80%

in D

G a

nd h

ilus

and

51%

and

55%

in C

A1

and

CA

3, r

espe

ctiv

ely.

4h

Acs

ady e

t al.

(sub

mitt

ed);

10%

and

14%

in C

A1

and

CA

3; n

o da

ta f

rom

DG

or

hilu

s.

47L.

Ser

ess (

pers

onal

com

mun

icat

ion)

.


ent and efferent connections has been successfully used to identify SPR-positive neurons (Table 1; Acsidy et al., submitted).

O n the basis of location and dendritic morphology, four different cell types may be distinguished in the dentate gyrus and the hippocampus: (I) pyramidal-shaped or fusiform cells, with the soma located in stratum granulosum of the dentate gyrus and with radially oriented aspiny dendrites spanning all layers; (2) fusiform cells in the hilus and stratum lucidum of the CA3 region, which have spiny dendrites running parallel to laminar boundaries and which remain confined to the same layers; (3) aspiny multipolar cells in strata pyramidale, proximal radiatum, and oriens of the CA1-CA3 regions, which have thin, radially running primary dendrites branching close to the soma; and (4) large, more robust aspiny or sparsely spiny multipolar cells with thick, distally branching dendrites. These cell types are found in the same layers and strata of the hippocampus as are the type 3. Several SPR-positive cells show transitional features of cell types 3 and 4.

Markers for perisomatic inhibitory cells, PV, and CCK, colocalize with SPR in pyramidal-like basket cells in the dentate gyrus (type 1 cells, described above) and in large multipolar cells of the hippocampus (type 4 cells). The dense meshwork of SPR-immunoreactive spiny dendrites in the dentate hilus and in stratum lucidum of the CA3 region belong to inhibitory cells that innervate the distal dendritic region, as suggested by their SOM and NPY content. In addition, in the hippocampus, SPR and NPY colocalize in numerous small multipolar interneurons, with dendrites branching close to the soma. There is a 25% overlap between the SPR-immunoreactive cells and CR-positive interneurons, suggesting that interneurons specialized to contact other GABAergic cells (Section III.5.a) also contain SPR. Thus, on the basis of the known termination pattern of the colocalized markers, SPR-positive interneurons may be functionally heterogeneous and participate in different inhibitory circuits, i.e., in perisomatic inhibition of principal cells (CCK-containing cells in all subfields and PV-positive cells in the dentate gyrus), in feedback dendritic inhibition in the entorhinal termination zone (SOM- and NPY- containing cells), and in the innervation of other interneurons (CR-containing cells).

IV.7. Interneurons Synthesizing or Accumulating Neurotrophic Factors

NGF Immunocytochemical studies have shown no association of

NGF immunoreactivity with any interneuron types. There is a homogeneous intense staining of the neuropil, which is particularly dense in areas innervated by mossy fibers (Conner et al., 1992). However, a recent study has demonstrated that NGF mRNA is expressed in GABAergic interneurons in the hippocampal formation (Lauterborn et al., 1993), indicating that interneurons are the major source of NGF production. The neurochemical identification of the NGF-producing interneuron subpopulations has been done in a study combining in situ hybridization for NGF mRNA and immunostaining for calcium- binding proteins PV, CB, and CR (Rocamora et al., 1996). The majority of PV-positive neurons (82%) express NGF mRNA,

which corresponds to 71% of all NGF-positive cells in the hippocampus. Only a small subset of the CB- (24%) and CR-immunoreactive (23%) interneurons show hybridization, which account for only a small proportion of NGF-synthesizing neurons. The NGF mRNA content of various neuropeptide-containing subpopulations is currently under study. Preliminary observations suggest that some SOM- and CCK-positive cells in the hilus and stratum oriens, and a small proportion of VIP- and NPY-containing cells express NGF mRNA (M. Pascual, N. Rocamora, L. Acsidy, T.F. Freund, and E. Soriano, unpublished observations).

NT3 Interneurons expressing this neurotrophin are less numerous.

Colocalization of NT3 mRNA with calcium-binding proteins reveal that only a small number of PV- and CR-immunoreactive neurons show hybridization (Rocamora et al., 1996). According to their laminar distribution, the majority of NT3-positive cells are likely to contain one or several neuropeptides. Immunostaining studies with antisera against NT3 protein have not been published.

BDNF In situ hybridization studies have shown that interneurons do

not express BDNF mRNA under normal conditions, in spite of heavy labeling in the principal cell somata (Hofer et al., 1990; Isackson et al., 1991; Berzaghi et al., submitted; Rocamora et al., 1996). In contrast with NGF immunoreactivity, immuno-staining for BDNF protein is weaker in the neuropil, but a small population of cell bodies is faintly labeled (Berzaghi et al., submitted). Immunocytochemical double labeling reveals that most if not all BDNF-positive somata belong to PV-containing interneurons in both the hippocampal formation and neocortex. The lack of BDNF mRNA in all these cells suggests that I’V-containing neurons do not synthesize but accumulate this neurotrophic factor. Interestingly, the level of BDNF protein can be remarkably enhanced afier seizure activity for several days (Nawa et al., 1995).

In a recent study by Marty et al. (1996), GABAergic stimulation was shown to increase the size and NPY content of interneurons in dissociated hippocampal cultures at a developmental stage when GABA is still enhancing calcium influx. This trophic effect of GABA was found to be mediated by an enhancement of synthesis and release of BDNF from nearby neurons. GABAergic drugs had no such effects in cultures prepared from BDNF knockout embryos. At a later stage, when GABA is inhibitory and reduces BDNF synthesis, GABA agonists induce a reduction in size and NPY immunoreactivity of neurons.

Numerous small subcortical nuclei, with neurons containing different neurotransmitters, send a diffuse innervation to the majority of neo- and archicortical areas. These pathways include GABAergic and cholinergic projections from the basal forebrain,


serotonergic and nonserotonergic fibers from the dorsal and median raphe nuclei, noradrenergic projections from the locus coeruleus, dopaminergic projections from the ventral tegmental area, histaminergic projections from the tuberomammillary nucleus, and projections containing as yet unknown transmitters from the supramammillary nucleus and the nucleus reuniens of the thalamus. One common feature of all these pathways is that they originate from a relatively small number of neurons. Nevertheless, they are able to exert a powerful control over electrical activity patterns in the entire cerebral cortex. This efficiency is due to the selective innervation of GABAergic interneurons (Fig. 27), which in turn exert GABAergic inhibition onto large populations of principal cells (Buzsiki, 1984; Freund and Antal, 1988; Freund et al., 1990c; Freund, 1992). In this section, the patterns of innervation and the neurochemical identity of postsynaptic targets will be presented mostly for those subcortical pathways that- show specif-ic relationships to hippocampal interneurons.

V.l. The GABAergic Septohippocampal Pathway By the early 1970s, it had already been shown that antero-

gradely transported radiolabeled amino acids and adenosine, injected into the medial septum-diagonal band of Broca complex (msdBb), translocate, most likely through synaptic contacts, into hippocampal interneurons (Rose and Schubert, 1977). Antero- grade labeling with the lectin PbaeoLus vulgaris leucoagglutinin (PHAL) revealed two types of septohippocampal fibers (Fig. 27). One has large boutons that occur in clusters along thick collaterals (type I), whereas the other has small boutons and arborizes more diffusely (type 2; Nyakas et al., 1987). Type 1 axons are immunoreactive for GABA, whereas type 2 axons are GABA negative (Fig. 28; Freund and Antal, 1988). Septa1 GABAergic terminals (i.e., type 1) always terminate on GABAergic interneurons (Fig. 28) in the rat (Freund and Antal., 1988; Gulyh et al., 1990) and in the monkey (Gulyhs et al., 1991a). Double immunostaining for PHAL and markers for different subsets of interneurons have revealed that all examined subpopulations, including neurons immunoreactive for PV (Fig. 27), CB, CR, SOM, NPY, CCK, and VIP, receive input from GABAergic (type 1) septohippocampal afferents (Freund and Antal, 1988; Gulyis et al., 1990; Miettinen and Freund, 1992a,b; Acsidy et al., 1993). The typical innervation pattern was multiple “climbing fiber-like” contacts on the soma and proximal and distal dendrites of the postsynaptic interneurons. All these contacts have proved to be symmetrical synapses.

The area where most of the interneurons appear to receive septal GABAergic input is the CA3 subfield, particularly strata oriens and pyramidale (Rose and Schubert, 1977; Nyakas et al., 1987; Freund and Antal, 1988; Gaykema et al., 1990). The hilus is also heavily innervated, but in the CA1 region, a relatively smaller proportion of the interneurons receive multiple synaptic input. The topography of the septohippocampal projection may in part account for the lower density of innervated interneurons in CAI. No systematic differences have been observed in the termination pattern of septal afferents between the dorsal and ventral hippocampus.

V.2. The Cholinergic Septohippocampal Pathway Detailed studies examining the target selection of cholinergic

septohippocampal afferents have been done by using immunostaining for C U T , the synthesizing enzyme of acetylcholine (Frotscher and Leranth, 1983, 1986; Leranth and Frotscher, 1987). The majority ofpostsynaptic elements are principal cells in the dentate gyrus and in the hippocampus. Only about 5-10% of all postsynaptic elements showed ultrastructural features of interneurons. This 5-10% appears to correspond to the proportion of occurrence of interneurons in the neuropil, suggesting that the interneurons are contacted in a “quasi-random” fashion. Direct evidence that interneurons are indeed innervated by cholinergic fibers is provided by double immunostaining that shows that ChAT-positive synapses terminate on hilar GAD- and/or SOM-containing neurons (Leranth and Frotscher, 1987; Frotscher and Leranth, 1986).

V.3. The Serotonergic Raphe-Hippocampal Projection

The serotonergic innervation of the hippocampus originates largely from the median raphe nucleus; a less robust projection arises from the dorsal raphe nucleus. The serotonergic projection consists of two types of fibers (Lidov et al., 1980; Kosofsky and Molliver, 1987; Tork, 1990). The most numerous are thin axons with small, evenly distributed varicosities that occur predominantly in stratum lacunosum-moleculare and adjacent regions of stratum radiatum of the hippocampus and in the subgranular zone of the dentate gyrus, although they are also present in other layers. The other fiber type has larger boutons that form clusters along secondary, but relatively large-diameter, dendritic branches. They are most abundant in stratum radiatum of the CAI region and at the hilar border of stratum granulosum. Electron microscopy of serial sections taken from material immunostained for serotonin has revealed that only about 25% of serotonergic varicosities establish morphologically identifiable synaptic contacts (Oleskevits et al., 1991). The targets of these synaptic varicosities could not be identified with simple light and electron microscopy.

When visualized by anterograde transport of PHAL, the large- caliber axons appear more numerous, probably because they transport the lectin better (Freund et al., 1 9 9 0 ~ ; Freund, 1992; Halasy et al., 1992). Double immunostaining for PHAL and different calcium-binding proteins and neuropeptides present in hip- pocampd interneurons reveals a striking association between the large-caliber raphe-hippocampal afferents and specific types of interneurons (Fig. 27). Up to 40 labeled boutons can be observed climbing along the dendrites and cell bodies of interneurons in stratum radiatum, the region bordering lacunosum-moleculare, and in stratum oriens. In the hippocampus, the most heavily innervated type of interneuron is the CB-containing cells in CA1 stratum radiatum, increasing in number toward the subicular border. Large subsets of CR-, SOM-, and NPY-positive neurons are also innervated, the latter in stratum oriens (Miettinen and Freund, 1992a,b), the former in strata radiatum, oriens, and pyramidale (Acsidy et al., 1993). Multiple contacts formed by PHAL-labeled raphe-hippocampal axons are sparse on VIP- positive neurons (Acsidy et al., 1994) and not seen at all on PV-immunoreactive basket and chandelier cells. In the dentate

FIGURE 27. Photomicrographs show septohippocampal (A,C) and raphe hippocampal (B) aEerents labeled with the anterograde tracer PHAL (black, Ni-DAB) and their postsynaptic targets. Postsynaptic targets were identified by immunostaining for PV and CB by using DAB as the chromogen (brown). A: Multiple contacts (arrows) are formed by septal fierents on a PV-positive interneuron (S,) in a climbing fiber-like manner in stratum oriens of the CA3 region. Somata of additional interneurons &) are also contacted. C: A PHAGlabeIed septal axon that traversed stratum moleculare of the

dentate gyrus forms synaptic varicosities exclusively on PV-positive dendrites of interneurons (arrows). B: Varicose axons from the median raphe establish multiple contacts (arrows) on the soma and proximal dendrites of a CB-positive interneuron in stratum radiatum of the CAl subfield. Varicosities of both septal and raphe origin that made multiple contacts with interneurons were synaptic boutons in all examined cases (see Figs. 28, 29). Data from Freund and Antal (1988) and Freund et al., (1990~). Scale bars = 10 pm.

INTERNEURONS OF T H E HIPPOCAMPUS 409

gyrus, GABAergic interneurons receive multiple innervation from median raphe fibers (Halasy et al., 1992). The target cells include mostly pyramidal-shaped neurons at the granule cell layer-hilar border and fusiform cells in the hilus. CB-positive interneurons are hardly visible due to the heavy staining of granule cells. Nevertheless, some fusiform CB-containing interneurons in the subgranular zone are surrounded by PHAL-labeled fibers. PV- positive neurons were not seen to receive multiple contacts.

All PHAL-labeled boutons in close apposition with interneurons form conventional synaptic contacts (Fig. 29; Acsidy et al., 1993). Serial ultrathin sections cut from PHAL-labeled raphe- hippocampal axons, immunostained for GABA with the postembedding immunogold procedure, reveal that the majority (over 70%) of the synaptic varicosities terminate on GABA-immunoreactive dendrites or somata (Halasy et al., 1992). The PHAL-labeled boutons themselves are always negative for GABA.

The results obtained with anrerograde tracing have been confirmed by double immunostaining for serotonin and the interneuron markers (Freund et al., 1990~). The same types of interneurons as in the PHAL study were shown to be innervated, demonstrating that the nonserotonergic component of the raphe-hippocampal projection (Kohler and Steinbush, 1982) has a negligible, if any, contribution to this innervation pattern (Freund et al., 1990c; Halasy et al., 1992; Acsidy et al., 1993). In sections with relatively good quality staining for serotonin, approximately 70% of CB-positive neurons in the CAl region and 50% in the CA3 region receive multiple innervation. Contacts on PV-positive cells are sparse and limited to 1-3 varicosities per connection. Electron microscopy of serotonin-immunoreactive axons confirm, in accordance with earlier findings (Oleskevits et al., 1991), that only about 25% of the varicosities make synaptic contacts. These synapses are established mostly with GABA-positive dendrites or somata in the hippocampus and in the dentate gyms (Freund, 1992).

In conclusion, the two types of serotonergic afferents may have different mechanisms of action in the hippocampus. More than 70% of the varicosities (probably those originating from thin dorsal raphe fibers; Kosofsky and Molliver, 1987) release serotonin at nonsynaptic sites, which may diffuse to different target cells having 5-HT 1-2 receptors, to exert a slow, tonic, G-protein-mediated action. The other fiber type with large boutons (mostly of median raphe origin; Kosofsky and Molliver, 1987) always forms synaptic contacts. The postsynaptic elements are specific interneuron types, in particular those exerting dendritic inhibition (Gulyis and Freund, in press; Miles et al., 1996). The receptor at these synapses is likely to be of the 5-HT3 type, which allows a fast excitation of these GABAergic cells, and an enhanced GABA-A receptor-mediated inhibition (Ropert and Guy, 199 1; Section IX.5).

V.4. Noradrenergic Innervation of the Hippocampal Formation

Noradrenergic afFerents to the hippocampal formation originate from the locus coeruleus. In the hippocampal formation, noradrenergic innervation is particularly dense in areas receiving mossy fiber input, i.e., in the hilus of the dentate gyrus and in stratum

lucidum of the CA3 subfield. Stratum lacunosum-moleculare of the CA1 region is also heavily innervated, but other regions and layers receive relatively sparse input (Loy et al., 1980; Morrison et al., 1979; Oleskevich et al., 1989; Moudy et al., 1993).

The majority of TH-positive (mostly noradrenergic) varicosities do not make conventional synaptic contacts. However, those that form synapses often terminate on dendritic shafts and somata of putative interneurons (Frotscher and Leranth, 1988; Milner and Bacon, 1989a). These synapses are symmetrical, whereas others terminating on spines of presumed pyramidal cells are asymmetrical (Frotscher and Leranrh, 1988). The postsynaptic elements have been confirmed to be GABAergic for a small sample in strata lucidum and radiatum of the CA3 subfield and in the hilus of the dentate gyrus by immunostaining for GAD (Frotscher and Leranth, 1988) or for GABA (Milner and Bacon, 1989a). The types and frequency of GABAergic neurons innervated in these and other regions remain to be established.

V.5. Other Subcortical Projections The supramammillary projection to the hippocampal forma-

tion has a remarkable laminar selectivity. It forms a dense plexus in the border region of strata granulosum and moleculare of the dentate gyrus and innervates stratum oriens of the CA2-CA3a regions (Dent et al., 1983; Stanfield and Cowan, 1984; Vertes, 1992; Magl6czky et al., 1994). The postsynaptic targets of PHAL- labeled supramammillo-hippocampal afferents studied by postembedding immunogold staining for GABA have revealed that the majority of targets are GABA-negative dendritic shafts, spines, and somata, most likely of granule cell origin (Maglbczky et al., 1994). Another study using a different approach has shown that some PV- or CB-positive interneurons located adjacent to the granule cell layer or within the hilus may occasionally be among the targets (Nitsch and Leranth, 1996).

Another projection to the hippocampus that displays an extreme laminar specificity originates from the nucleus reuniens. These fibers terminate in stratum lacunosum-moleculare of the CA1 region, where they form asymmetrical synapses (Wouterlood et al., 1990). The target elements include mostly spines or spiny shafts of distal dendrites, which likely belong to pyramidal cells.

Occasional histaminergic fibers originating from the tubero- mamillary nucleus (Panula et al., 1989), dopaminergic axons from the ventral tegmental area (Loy et al., 1980; Verney et al., 1985), and vasopressin-containing afferents from the medial amygdala (caffe et al., 1987) have been described in the hippocampus, but their synaptic connections and interneurons remains to be established.

An unconventional feature of some GABAergic nonprincipal neurons is that they project out of the ipsilateral hippocampal formation. According to classical definitions, therefore, such


FIGURE 28. Light and electron photomicrographs show the termination pattern of septohippocampal axons. A: PHAL-labeled s e p tohippocampal aEerents with large boutons (type 1) surround cell bodies and dendrites of interneurons (arrows), whereas those with faintly labelled small boutons (type 2) terminate diffusely in the neuropil (arrowheads). B,C: Adjacent ultrathin sections of a type 1 PHALlabeled septohippocampal bouton (b,) making a symmetrical synaptic contact (arrow) with a cell body (S). The section in C was immunostained for GABA by using the postembedding immunogold procedure. The accumulation of colloidal gold particles indicates that the septal bouton and the postsynaptic cell body are immunoreactive for GABA. Axon terminals with round vesicles (asterisks), which form asymmetrical synapses, are always negative for GABA. Other GABA- positive profiles (stars), not labeled by PHAL, are also visible. Reproduced from Freund and Antal (1988) “-A-containing neurons in the septum control inhibitory interneurons in the hippocampus.” Nature 336:170-173 by permission of Macmillan Magazines Limited. Scale bars = 20 pm in A, 0.2 Fm in B,C.

GABAergic projection neurons should not be referred to as “interneurons” or “local circuit neurons” (but see Section I). The most common are those that project from the hilus to the contralateral dentate gyrus and those in stratum oriens of the hippocampus that innervate the medial septum. They most likely have local collaterals as well, by which they contribute to (ipsilateral) intrahippocampal inhibitory networks (Sections III.3.2.b, IV.2.b, IV.3.a, IV.3.b). The GABAergic nature of these interneurons that have extrahippocampal or commissural projections is difficult to establish. It has been shown in several brain areas that GABAergic neurons with distant projections (e.g., the striatonigral, nigrothalamic, nigrocollicular, pallidonigral, septohippocampal neurons, the Purkinje cells) have very low levels of GABA in their cell bodies, levels that do not reach the immuno- cytochernical detection threshold. However, anterograde tracing combined with postembedding immunogold staining for GABA has shown that axon terminals of GABAergic cells with distant projections are immunoreactive for GABA (Figs. 28, 30; Freund and Antal, 1988; T6th et al., 1993).

VI.1. Interneurons With Commissural Projection The first hint of a possible contribution of GABAergic in-

terneurons to the hilar commissural projection was published by Seress and Ribak (1983). They showed that more than 60% of hilar neurons are GAD positive, whereas approximately 80% of neurons in the same region project commissurally. This finding suggests that at least some of the hilar projection neurons must be GABAergic. Direct evidence for this notion has been provided by combining retrograde tracing from the contralateral dentate gyms and GAD immunocytochemistry and by anterograde degeneration and horseradish peroxidase (HRP) transport (Ribak et al. , 1986). The presence of numerous double-labeled cells in the hilus confirmed that GABAergic neurons indeed participate in the commissural pathway. Furthermore, numerous anterogradely labeled or degenerating boutons were found in the contralateral side making symmetrical synapses, which is characteristic of GABAergic axon terminals. Similar combinations of retrograde tracing and immunocytochemistry show that most of the hilar in-

terneurons that give rise to a comrnissural projection belong to the SOM-containing subpopulation of GABAergic neurons (Zimmer et al., 1983; Bakst et al., 1986; Leranth and Frotscher, 1987). As one might expect from double-labeling studies, NPY- containing cells (Deller and Leranth, 1990) and spiny CR-positive neurons (Miettinen et al., 1992) are also involved in these commissural connections (Sections IV.2.c, IV.3.a, IV.3.b).

There appears to be a discrepancy between the results of retrograde and anterograde labeling studies. SOM-positive cells def- initely contribute to the commissural projection according to the retrograde labeling data, and numerous immunostaining studies have shown that SOM-containing axons arborize mostly in the outer two-thirds of the molecular layer of the dentate gyrus (Section IV.3.a). However, anterograde tracing techniques using radioactive tracers, degeneration, or HRP visualized axons only in the inner one-third of the molecular layer of the contralateral dentate gyms (Blackstad, 1956; Gottlieb and Cowan, 1973; Laurberg, 1979). A recent experiment has solved this apparent conflict of data by demonstrating that hilar commissural fibers also terminate in the outer molecular layer (Deller et al., 1995b), although these axons still appear sparse compared with the relatively large number of SOM-containing neurons participating in the projection.

These data suggest that there is a component of direct inhibition in the feed-forward inhibitory response evoked in the dentate gyrus by commissural stimulation (Buzsdci and Eidelberg, 1981; Buzsiki, 1984), but its specific function remains to be established.

VI.2. Interneurons With Hippocamposeptal Projections

In addition to pyramidal cells, a specific subset of nonpyramidal cells, most of which are located in stratum oriens of the hippocampus and in the hilus of the dentate gyrus, participate in the hippocamposeptal projection (Alonso and Kohler, 1982). Alonso and Kohler also showed that, although large septal injections of retrograde tracers labeled both pyramidal and putative nonpyramidal cells, a more focused injection into the medial septum gives rise to specific retrograde labeling of the latter cell type alone. Transection of the firnbria-fornix result in the accumulation of GABA in fibers located on the hippocampal side of the cut (Shinoda et al., 1987), suggesting the existence of hippocampofugal GABAergic projections. By using a combination of retrograde HRP transport and immunostaining for interneuron markers, the majority of interneurons (80%) projecting to the medial septum were shown to be the CB-containing horizontal cells at the stratum oriens-alveus border (Fig. 31; T6th and Freund, 1992). In some sections with good retrograde labeling, more than 50% of the horizontal CB-positive cells were found to project to the medial septum. A smaller proportion of the retrogradely labeled cells (approximately 20%) contain NPY and/or SOM, although immunostaining for these neuropeptides in that study was less sensitive, probably resulting in considerable false- negative staining. This is indeed most likely in light of recent colocalization data showing that more than 90% of the horizontal CB-containing neurons at the stratum orien-salveus (0-A)

412 FREUND AND BUZSAKl


FIGURE 29. Photomicrographs show the postsynaptic targets of PHAL-labeled raphe hippocampal afFerents. Postsynaptic targets were identified by using preembedding immunostaining for CB (CaBP) and postembedding immunogold staining for GABA A: Light micrograph of a CaBP-positive interneuron that received multiple contacts (bl-3) from PHAL-labeled afFerents arising from the median raphe. B: A correlated low power electron micrograph showing PHAL-labeled axon terminals (bl-3) making direct membrane contact with a CB-positive dendrite. Two of these juxtapositions (bz, b3) are conventional synaptic contacts (arrows) shown at higher magnification in C and D. E: An adjacent ultrathin section stained for GABA with the immunogold procedure shows that the CB-positive dendrite is GABA immunoreactive, whereas the PHAL-labeled raphe bouton (bz), which made a synaptic contact (arrow) with a dendrite, is GABA-negative. Scale bars = 10 pm in A, 1 pm in B, 0.5 pm in C-E.

border are also immunoreactive for SOM (Katona et al., in press). The GABAergic nature of these projection neurons was equivo- cal because GABA immunoreactivity of their cell bodies did not exceed background level (T6th and Freund, 1992). However, anterograde tracing with PHAL combined with postembedding immunogold staining for GABA confirmed that axon terminals of these neurons in the medial septum are GABA positive and form symmetrical synaptic contacts (Fig. 30). The postsynaptic targets have been examined by double immunostaining for PHAL and markers of cholinergic (ChAT) and GABAergic (PV) septal neurons. In some experiments, additional retrograde tracer injections into the hippocampus allowed the identification of cholinergic and GABAergic septohippocampal neurons in the same material. The predominant targets of hippocamposeptal GABAergic fierents are the PV-containing GABAergic neurons in the medial septum and the vertical limb of the diagonal band of Broca (Fig. 31), whereas cholinergic neurons are seldom innervated (T6th et al., 1993). The targets also include numerous retrogradely labeled GABAergic and a small number of cholinergic septohippocarnpal neurons.

It is important to note here that the cells of origin of this GABAergic hippocamposeptal feedback projection are interneurons, which appear to be driven primarily by recurrent collaterals of local pyramidal cells, as suggested by the stereotyped distribution of their dendritic trees (Sections 111, IV.2.b, IV.3.a; T6th and Freund, 1992; Baude et al., 1993; Blasco-Ibanez and Freund, 1995). The dendritic trees of CB- and/or SOM-positive cells of this type in CAI stratum oriens have a horizontal orientation and are restricted to the stratum oriens-alveus border in a manner similar to the local collaterals of CA1 pyramidal cells. These collaterals were shown to account for more than 80% of the excitatory input to the dendrites of horizontal cells (Blasco-Ibanez and Freund, 1995). In the CA3 subfield, the dendrites of CB- and/or SOM-containing neurons are not restricted to stratum oriens. They often penetrate stratum radiatum and overlap in distribution with recurrent collaterals of local pyramidal cells. Because each pyramidal cell is likely to establish only a single synapse with interneurons (Gulyh et al., 1993b), the activity of these neurons i s likely to reflect synchrony in the activity of large local pyramidal cell populations (T6th et al., 1993). The functional implications of such a synchrony-dependent GABAergic feedback directly

to septal GABAergic projection neurons is discussed in Sections XII.3 and XIII.2.

The wealth of anatomical knowledge about interneurons discussed above may be contrasted with the paucity of information available about their distinctive physiological properties (Tables 2-4). A rational taxonomy of interneurons should be based on the premise that the anatomically/chemically distinguishable classes have different functions, whereas neurons of the same class should have the same properties.

Prior to the widespread application of intracellular staining methods, researchers had already recognized that spontaneous activity, evoked responses, and passive membrane properties of neurons outside the pyramidal and granule cell layers are distinct from those of principal cells, and they used those criteria for interneuron identification (Andersen et al., 1963, 1964, 1969; Ranck, 1973; Fox and Ranck, 1981; Buzsiki and Eidelberg, 1981, 1982; Ashwood et al., 1984; Lacaille et al., 1987). In the intact hippocampus, pyramidal cells spontaneously discharge bursts of 2-1 0 action potentials of decreasing amplitude and increasing duration (“complex-spike” bursts). Under several circumstances, these features alone may distinguish them from interneurons of the hippocampal formation (Ranck, 1973). However, under most conditions, a set of criteria are needed for the positive identification of interneurons. These typically include their high spontaneous firing rate (5-80 Hz), short duration action potentials, short latency, and repetitive discharge in response to afferent inputs. Interneurons, in general, have a significantly lower threshold for action potential generation in response to afferent stimulation than do principal cells (Buzs&i and Eidelberg, 1982; Ashwood et al., 1984; Lacaille, 1991). Several possible reasons for this low threshold include their larger size shaft synapses as opposed to spine synapses in principal cells, different leak conductances, activity and density of Na+-K+ pumps, and hypothesized dendritic generation of fast action potentials (Section IX. 1). Furthermore, the absence of G l u m and enhanced expression of GluR4 subunits ofAMPA-gated channels in interneurons (Jonas et al., 1994; Baude et al., 1995; Geiger et al., 1995; Jonas and Burnashev, 1995; Leranth et al., 1996) may be responsible for the low threshold of spike generation in interneurons. Recombinant AMPA channels that contain the GluR4 but lack the GluR2 subunit when inserted into oocytes possess the fastest kinetics (Hollmann and Heinemann, 1994; Seeburg, 1993).

The passive electrical properties of most hippocampal interneurons are characteristically different from those of principal cells. Early in vitro experiments with anatomically characterized cells listed a number of criteria for the physiological verification of interneurons, including (1) a relatively high “spontaneous” firing rate, (2) short duration spikes (<1.2 ms), (3) large spike afterhyperpolarization, (4) weak spike frequency accommodation in response to depolarizing current injection, and ( 5 ) high input resis-

414 F R E W D AND BUZSAKI

B

FIGURE 30. Photomicrographs and line drawing show the distribution and termination of the hippocamposeptal projection labeled with PHAL. A: PHAL injection site in stratum oriens of the CAI region of the hippocampus. In addition to numerous pyramidal cells, interneurons with horizontal dendrites (white arrows) at the stratum oriens-alveus border also accumulated the lectin. B: Drawing of the distribution of PHAL-labeled hippocampal afferents in the septal region. In addition to the well-known pyramidal cell projection to dorsal parts of the lateral septum (LS), numerous afferents to the medial septum (MS), and the vertical limb of the diagonal band of Broca (VDB) were also visualized. LV, lateral ven- tricle. C-E: Electron micrographs of three serial sections of a PHAL-labeled hippocampal afferent bouton (b,) in the MS. C: In

contrast to pyramidal cell boutons in the LS, hippocampal afferent terminals in the MS and VDB regions (e.g., as bl, shown here) form symmetrical synapses (arrows in C-E). D,E: Postembedding immunogold staining of adjacent sections for GABA show that hippocampal afferents (bl) in the MS and VDB regions are GABAergic. Other GABA-positive boutons (small asterisks) and GABA-negative profiles (large asterisks) are also indicated. From T6th K, Borhegyi Z, F reud TF (1993) “Postsynaptic targets of GABAergic hippocampal neurons in the medial septum-diagonal band of broca complex.” J Neurosci 13:3712-3724 by permission of Society of Neuroscience. Scale bars = 50 p m in A, 0.5 pm in C,D, 0.75 pm in E.


tance (Fig. 32; Schwartzkroin and Mathers, 1978; Schwartzkroin and Kunkel, 1985; Lacaille et al., 1987; Kawaguchi and Hama, 1987a,b). These criteria, however, do not adequately describe all known interneuron types.

Afferent activation elicits postsynaptic potentials in CA1 interneurons that may be composed of four components: (1) a fast AMPA-receptor-mediated excitatory postsynaptic potential (EPSP), (2) an early GABAA receptor-mediated IPSP, with an equilibrium potential of -70 to -75 mV and 40% decrease in input resistance, (3) a late GABAB receptor-mediated IPSP, peaking at 100-130 ms, with an equilibrium potential of -105 to - 11 0 mV, and (4) a delayed, slower EPSP peaking at 50-80 ms (Lacaille, 1991; Buhl et al., 1994a,b; Miles 1991; Miles et al., 1996; Perouansky and Yaari, 1993; Williams et al., 1994; Sik et al., 1995; Khazipov et al., 1995a,b). The late EPSP is present only in a portion of the interneurons examined. Repetitive afferent stimulation at 1 Hz or higher decreases the early IPSPs, abolishes the late IPSPs, and reveals the delayed EPSPs. In whole-cell patch- clamp studies, the early and late excitatory postsynaptic currents (EPSCs) have been analyzed pharmacologically and correspond to the activation of AMPA and N-methyl-D-aspartate (NMDA) currents, respectively. The average time constant of the AMPA (34 ms) and NMDA (210 ms) responses is quite similar to that observed in pyramidal cells (Perouansky and Yaari, 1993).

The entire trajectory of a single action potential and associated afterhyperpolarizations is complete in 200 ms as opposed to > I s in principal cells. In addition, interneurons do not show a prominent spike after depolarization. Because most interneurons selectively possess members of the family of Kv3 (Shaw-related) potassium channel subunits (Weiser et al., 1994), the short duration of the action potentials in interneurons may be explained by the faster kinetics of K+ currents (Perney et al., 1992; Du et al., 1996). It was recently suggested that the presence of voltage-gated potassium channel subunit Kv3.1 contributes to the rapid repolarization and fast spiking nature of PV-immunoreactive basket cells (Du et al., 1996). Because other interneuron types with horizontally running dendrites in stratum oriens and the alveus, which lack subunit Kv3.1, possess shorter action potentials than do basket cells (Lacaille et al., 1987; Buhl et al., 1994a; Sik et al., 1995), it is likely that the neuron-specific phenotype of the action potential is determined by the interplay between several temporally overlapping potassium currents (Weiser et d., 1994). Interneurons, like many central neurons, possess a transient, rapidly inactivating current, IA, and at least two sustained “delayed-rectifier’’ IK current phenotypes. The time to peak, rate of inactivation, voltage dependence, and rate of recovery from inactivation are quite similar to the IA of pyramidal cells (Zhang and McBain, 1995b). However, the sustained current IK is quite distinct. IK of interneurons in stratum oriens-alveus of the CAI region showed virtually no inactivation over a wide range of volt- ages and showed minimal time-dependent inactivation. The lack of time- and voltage-dependent inactivation of IK may ensure ad- equate spike repolarization without an accumulation of inactivation during prolonged episodes of fast action potentials (Zhang and McBain, 1995b). In addition to these voltage-dependent currents, the action potential repolarization and the fast afterhyperpolarization in interneurons may also be affected by the Ca2+-

activated K+ current, Ic. However, the firing rate of the interneuron may be limited by the slow component of spike afterhyperpolarization (Zhang and McBain, 1995a).

Unique combinations of the above physiological features or ex- plicit absence of some of them have been used in attempts to group interneurons into subclasses. Although the exact relationship between physiological and anatomical features of interneurons has yet to be worked out, the available physiological information already indicates that the different anatomical subgroups may have distinct physiological properties (Schwartzkroin and Mathers, 1978; Kawaguchi and Hama, 1987a,b; LacaiIle et al., 1987; Lacaille and Schwartzkroin, 1988a,b; Lacaille 1991; Buhl et al., 1994a,b, 1995; Sik et al., 1995, submitted; Khazipov et al., 1995a). CAI interneurons in the pyramidal layer (putative basket cells and chandelier cells) display some spike frequency adaptation in response to depolarizing current injection and have smaller spike afterhyperpolarization. Whole-cell recordings indicate that basket cells have a membrane resistivity of 7-66 k a cm2 and specific cytoplasmic resistivity of 52-484 ka cm2 (Thurbon et al., 1994). Basket and chandelier cells are less responsive to mGluR agonists than are oriens-alveus interneurons (McBain et al., 1994). In addition, basket and chandelier cells can be monosynaptically activated by both CA3 and CA1 pyramidal cells, whereas the majority of neurons in the “horizontal” group are innervated by CAI but not CA3 pyramidal neurons (Blasco-Ibanez and Freund, 1995; Maccaferri and McBain, 1995; Sik et al., 1994, 1995). Chandelier cells were suggested to fire repetitive “doublet” action potentials in response to depolarizing current pulses (Buhl et al., 1994b). The dendritic tree of basket cells and chandelier cells are somewhat distinct, suggesting that at least some of their afferents are nonoverlapping (Sections 111.1, 111.2). Nevertheless, probably both cell types may be discharged by the direct entorhinal afferents (Fig. 33).

O n the basis of the anatomical variability of interneurons in the oriens-alveus region (Sections 111.1-II1.3), one might expect further physiological differences among them. Large inward current responses to mGLuR activation and an expressed H-current (depolarizing “sag”; McCormick and Pape, 1990) in response to hyperpolarizing current pulses appear to be characteristic features of 0-LM cells but not of all interneurons with similar horizontal dendritic trees (McBain 1994; McBain et al., 1994; Sik et al., 1994). These neurons contain the peptide SOM and have an ex- ceptionally high density of mGluRla receptors (Sections IV.3.a, IV.6.a; Baude et al., 1993). The sag current, however, is weak in back-projection cells, with dendritic morphology similar to 0-LM cells (Sik et al., 1994). Back-projection neurons are likely NPY immunoreactive (Section IV.3.b), and immunocytochemically labeled NPY neurons in the hilar region also displayed a weak sag current (Sik et al., submitted). Another common feature of the NPY-SOM-immunoreactive family is a very prominent spike afterhyperpolarization and the selective presence of the Kv3.2 potassium channel subunit (Weiser et al., 1994). The trilaminar interneuron type can be distinguished from the above cells by the absence of the late IPSP and by antidromic activation from the fimbria-fornix (Sik et al., 1995, submitted). In contrast to 0-LM, trilaminar, and backprojection cells, bistratified interneurons in the CA1 region can be monosynaprically activated by stimulation of the Schaffer collaterals (Buhl et al., 1994a; Sik et al., 1995).

416 FREUND AND B U Z S k I


FIGURE 31. Photomicrographs illustrate immunocytochemically identified targets of the hippocampal projection to the medial septum (ms). A: Light micrograph of a medial septal section double immunostained for ChAT (brown profiles indicated by white arrows) and PHAL (blue profiles indicated by black arrowheads). ChAT-positive elements were visualized with DAB (brown), and PHAL-labeled fibers and terminals were visualized with Ni-DAB (blue). The majority of PHAL-labeled hippocampal afferents were located medial (to the right) to the group of cholinergic cells. B: A section from the same region of the medial septum is shown here; the orientation mir- rors that in A (medial is to the left). Double staining for ChAT (blue profiles indicated by white arrows) and PV (brown profdes indicated by black arrows) shows that the region located medial to the group of cholinergic cells (i.e., the region receiving the bulk of the hippocampal input, as shown in A) is occupied by PV-positive cells that correspond to GABAergic projection neurons. The lateral septal nucleus (Is) is also labeled in A and B. C: Cell bodies and dendrites of PV-positive neurons (brown, DAB) in the MS receive multiple contacts (arrows) from PHAL-labeled hippocampal derents (black, Ni- DAB). D: Light micrograph of stratum oriens of the CA1 subfield taken from a section reacted for retrogradely transported HRP (black granules, Ni-DAB) after medial septal injection and immunostained for CB (brown, DAB). There are two CB-positive neurons (black arrows) at the stratum oriens-alveus border, which also contain granules of the reaction product of retrogradely transported HRP. Thus, these neurons represent hippocamposeptal nonpyramidal neurons immunoreactive for CB. Another CB-positive interneuron, not labeled for HRP, is indicated by an open arrow. Some CB-negative pyramidal cells are marked by asterisks. From Thth K, Freund TF (1992) “Calbindin D28k-containing nonpyramidal cells in the rat hippocampus: their immunoreactivity for GABA and projection to the medial septum.” Neuroscience 49:793-805 by permission of Elsevier Science Ltd and from T6th K, Borhegyiz, Freund TF (1993) “Postsynaptic targets of GABAergic hippocampal neurons in the medial septum-diagonal band of broca complex.” J Neurosci 13: 3712-3724 by permission of Society of Neuroscience. Scale bars = 50 pm in A,B, 15 pm in C,D.

Interneurons with cell bodies in CAI stratum lacunosum-moleculare (LM cells) have been treated as a relatively homogeneous population by physiologists; some LM cells have characteristics similar to CAI pyramidal cells and other properties more like other interneurons (Kawaguchi and Hama, 1987a,b; Lacaille and Schwartzkroin, 1988a,b; Khazipov et a!., 1995a,b). Some of these LM cells have extensive local axon collaterals. Similarities with CA1 pyramidal cells include broad and slowly decaying action potentials (2.0 ms), a relatively slow time constant (8-9 ms), no or very little spontaneous firing, and little evidence of spontaneous synaptic potentials. The lack of “spontaneous” EPSPs and action potentials in vitro, however, may simply reflect the lack of afferent drive because the somata giving rise to afferents to these interneurons are lost when using the slice procedure. The precise inputs to these LM interneurons are not known, but they likely include the perforant path, intrahippocampal, thalamic, and subcortical afferents, most of which are lost during slicing of the brain. Stimulation of dendritic layers often induces a short lasting oscillation (20-40 Hz) of the membrane potential of LM cells. As with other interneurons, cells in stratum-lacunosum moleculare have a high input resistance (40-100 M a , with sharp electrodes), a prominent afterhyperpolarization, and only limited spike frequency accommodation in response to depolarizing pulses (Kawaguchi and Hama, 1987a; Lacaille and Schwartzkroin,

1988a,b; Kunkel et al., 1988). LM cells usually display anodal break excitation and “burst” firing when the membrane is released from hyperpolarization (Lacaille and Schwartzkroin, 1988a; Fraser and MacVicar, 1991), indicating the presence of low- threshold calcium channels (T channels; Jahnsen and Llinas, 1984). Such features, however, have also been observed in some interneurons of the CAI stratum oriens-alveus (Lacaille and Williams, 1990; Zhang and McBain, 1995a).

The low-frequency discharges and the relatively wide action potentials of LM interneurons suggest that they are endowed with currents distinct from other interneurons. LM interneurons conspicu- ously lack potassium channels of the Kv3 family (Weiser et al., 1994). However, tissue culture experiments indicate that LM cells possess delayed rectifier currents different from those described in 0-LM cells and basket cells (Chickwendu and McBain, in press).

Classification of interneuron subgroups by physiological means in the CA3 region (Gulyis et al., 1993a; Arancio et al., 1994; Miles et al., 1996; Poncer et al., 1995) and the dentate gyrus (Han et al., 1993; Scharfman 1995a; Sik et al., submitted) have also progressed recently. For example, McBain and Dingledine (1993) discriminated two major classes of CA3 stratum radiatum interneurons on the basis of their current-voltage relations to kainate application and the kinetic properties of spontaneous mEPSCs. They hypothesized that the two groups may have different GluR subunit compositions in their AMPA receptors. In the dentate gyrus, Scharfman (1995a) described heterogeneous physiological properties of interneurons located in the granule cell layer and subgranular layer. Pyramidal- shaped interneurons in the granule cell layer have significantly longer action potentials (1.2 ms) and stronger spike frequency adaptation than those of the hilar group. Cells with similar dendritic morphology often have different axonal arbors. Dentate gyrus interneurons, recorded in vivo in the gerbil, can be classified into “fast- spiking and “slow-spiking” groups (Buckmaster and Schwartzkroin, 1995b). Fast-spiking neurons displayed a low threshold to perforant path stimulation, whereas slow-spiking interneurons responded with predominantly inhibitory potentials. Neurons in the fast-spiking group were not labeled in that study, but one of the slow-spiking interneurons had morphological features similar to NPY and/or SOM cells and spiny CR-immunoreactive cells.

This short summary of the passive and evoked properties of interneurons suggests that some of their morphological characteristics may be predicted from their physiological properties. To date, however, most correlative studies lack rigorous criteria at either the physiological or anatomical level. Although it is not expected that all morphological classes of interneurons have distinct functional features, it may well be that a given physiological property is associated with a unique set of morphological features, and a blend of these features may explain the often subtle differences in the physiological domain.

The dendrites, cell bodies, and the axon initial segment of every principal cell in cortical structures are innervated by inhibitory

TAB

LE 2

.

Den

tate

Gyt

-usl

Lam

inar

A

ffer

ent i

nput

N

euro

chem

ical

mar

kers

di

strib

utio

n of

In

trahi

ppoc

ampa

l R

ecep

tors

C

ell t

ype

Den

drite

s A

xon

Targ

ets

(exc

itato

ry)

Extra

hipp

ocam

pal

CB

Ps/N

OS

Neu

rope

ptid

es

etc.

Ph

ysio

logi

cal f

eatu

res

Axo

-axo

nic

all l

ayer

s s.g

., h

g.is

C

/A, m

ossy

ec

, ms

PV

-

SPR

2,

spik

e do

uble

ts

Bas

ket 1

al

l lay

ers

sg.,

h

r.s, p

d C

/A, m

ossy

ec

, ms

PV

-

SPR

2,

rela

tivel

y w

ide

AP,

acc

omm

odat

ion

Gal

, WA

Gal

, VV

A

Bas

ket 2

al

l lay

ers

s.g.,

h r.s

, pd

C/A

, mos

sy

ec, m

s -

CC

K, V

IP( 2

)

SPR

-

HIP

P h,

(C

A3c

) s.

m. (0)

r.dd

m

ossy

m

s, m

r C

R (

2)

SOM

, NPY

m

Glu

R1,

na

rrow

AP,

lar

ge A

HP,

slo

w s

piki

ng,

HIC

AP

all l

ayer

s s.

m. (

i)

r.pd,

s, dd

C

/A,

mos

sy

ec, m

s, m

r(?)

N

OS

(?)

CC

K (

?)

SPR

(?)

gam

ma

rhyt

mic

ity, a

ctiv

ated

dur

ing

MO

PP

s.m

. (0

) s.

m. (

0)

r.dd

C

/A

SPR

gam

ma

rhyt

hmic

ity

dent

ate

EEG

spi

kes

ec, m

s, m

r(?)

CB

(?)

-

-

Spin

y C

R

h, (

CA

3c)

s.m

. (0

) r.

dd(?

) m

ossy

m

s CR

SO

M, N

PY (

?)

mG

luR

l (?

) -

IS-1

al

l lay

ers

all l

ayer

s i.s

, i.p

d, i.

dd

C/A

, mos

sy

ec,m

s,rn

r C

R

VIP

(2

)

SPR

(2

)

-

IS-2

al

l lay

ers

h.

i.s, i

.pd,

i.dd

C

/A, m

ossy

ec

, ms

CR

VIP

-

(sm

ooth

CR

)

(VIP

)

'Tab

les

24

prov

ide

a lis

t of

mor

phol

ogic

al, n

euro

chem

ical

, and

phy

siol

ogic

al fe

atur

es o

f id

entif

ied

inte

meu

rons

. Tho

se ty

pes

at th

e bo

ttom

of

the

tabl

es, i

ndic

ated

in

bold

face

type

, are

kno

wn

only

from

imm

unoc

ytoc

hem

ical

stud

ies;

ther

efor

e, p

hysi

olog

ical

and

det

aile

d an

atom

ical

dat

a ar

e no

t ava

ilabl

e fo

r the

se ty

pes.

AH

P, sp

ike

afte

rhyp

erpo

lari

zatio

n; A

P, a

ctio

n po

tent

ial;

C/A

, com

- m

issu

ral-

asso

ciat

iona

l pat

h; d

d, d

ista

l den

drite

s; e

c, e

ntor

hina

l cor

tex;

Gal

, GA

BA

-A re

cept

or a

lpha

1 s

ubun

it; g

.is, g

ranu

le c

ell a

xon

initi

al s

egm

ent;

h., h

ilus;

IH, h

yper

pola

riza

tion-

activ

ated

cur

- re

nt; i

.dd,

int

erne

uron

dis

tal d

endr

ite;

i.pd,

int

erne

uron

pro

xim

al d

endr

ite; i

s, in

tern

euro

n so

ma;

15-

1, t

ype

l in

tern

euro

n-se

lect

ive

cells

(CR

); IS

-2, t

ype

2 in

tern

euro

n-se

lect

ive

cells

(VIP

); 15

-3,

type

3 in

tern

euro

n-se

lect

ive

cells

(VIP

); lc

, loc

al c

olla

tera

ls of

pyr

amid

al c

ells

; mr,

med

ian

raph

e; m

s, m

edia

l sep

tum

; pd,

pro

xim

al d

endr

ite; p

is, p

yram

idal

cel

l axo

n in

itial

seg

men

t; r.,

ran

dom

(i

,e., c

onta

cts b

oth

gran

ule

cells

and

inte

rneu

rons

in p

ropo

rtio

n of

thei

r oc

curr

ence

); s.

, som

a; s

.g.,

stra

tum

gra

nulo

sum

; s.

m. (

i), in

ner

stra

tum

rno

lecu

lare

; s.m

. (o)

, out

er s

trat

um m

olec

ular

e; s.

l.,

stra

tum

luci

dum

; s.1

-m.,

stra

tum

lac

unos

um-m

olec

ular

e; s.

o., s

trat

um o

rien

s; s.

P., s

trat

um p

yram

idal

e; s

.r., s

trat

um r

adia

tum

; V

VA

, Vic

in vi

llosa

agg

lutin

in.

20nl

y in

the

dent

ate

gyru

s 3O

nly

on a

xon

term

inal

s.

TAB

LE 3

.

CA

1 R

egio

n

Lam

inar

A

ffer

ent i

nput

N

euro

chem

ical

mar

kers

di

stri

butio

n of

In

trah

ippo

cam

pal

Rec

epto

rs

Cel

l typ

e D

endr

ites

Axo

n T

arge

ts

(exc

itato

ry)

Ext

rahi

ppoc

ampa

l C

BPs

/NO

S N

euro

pept

ides

et

c.

Phys

iolo

gica

l fea

ture

s

Axo

-axo

nic

all l

ayer

s s.

P., S

.O.

pis

C

/A,

lc.

ec, m

s PV

-

d2

,

Gal

, sp

ike

doub

lets

(?),

acco

mm

odat

ion

VV

A

Bas

ket 1

al

l lay

ers

s.P.,

s.o.,

r.s,p

d C

/A, I

c. ec

, ms

PV

-

m22

, Gal

, re

lativ

ely

wid

e A

P, a

ccom

mod

atio

n,

(s.r.

) W

A

low

-am

plitu

de A

HP,

thet

a, g

amm

a,

200-

Hz

rhyt

hms,

A/C

act

ivat

ion

-

CC

K, V

IP(?

) SP

R,

-

Bas

ket 2

al

l lay

ers

s.P.,

s.0,

r.s

,pd

C/A

, Ic

. (4

, ms

0-L

M

s.o.

, ah

. s.1

-m.,

r.dd

Ic

. m

s, m

r C

B( t)

SOM

, m

Glu

R1,

la

rge

IH, l

ow-t

hres

hold

Ca2

+ sp

ikes

, (s

.r.)

rnP(

?)

(s.o

., s.

r.)

NPY

(+)

SPR

( ??)

in

trin

sic

osci

llatio

n at

thet

a fr

eque

ncy,

la

ck o

f C

/A a

ctiv

atio

n, d

isch

arge

by

CA

1 py

ram

idal

cel

ls

Bis

tratif

ied

s.o.

, s.r.

, s.

r., s.

o.,

r.pd,

s,dd

C

/A,

lc.

ms,

mr

CB, N

OS(

?) N

PY( +

?)

5-H

T-3

(?)

C/A

act

ivat

ion

Hor

izon

tal

s.o.

, alv

. s.

r., s

.P.,

r.pd,

s,dd

lc

., C

/A(?

) m

s(?)

, mr(

?)

CB

(?)

SOM

(?),

mW

, la

rge

AH

P, p

rom

inen

t la

te

trila

min

ar

S.O

. N

PY(?

) m

Glu

Rl(

?)

depo

lari

zing

pot

entia

l to

aff

eren

t

s.p.

(S

.P.)

activ

atio

n R

adia

l s.

o., s

.r.,

s.r.,

s.P

., r.p

d,s,

dd

C/A

, Ic

. m

s, m

r (e

c)

CB

(?)

NPY

(?),

SPR

(?)

-

trila

min

ar

s.P.

, s.1

-m. S

.O.

CC

K(?

) B

ack-

s.

o., a

h.

s.r.,

l S.O

.,]

r.pd

,dd

lc.,

C/A

(?)

ms(

?), m

r(?)

N

OS(

?)

SOM

(?),

m2(

?),

smal

l IH,

C/A

inp

uts

hype

rpol

ariz

e pr

ojec

tion

s.p.

, h.l

NPY

(?)

mG

luR

l(?)

(sm

ooth

CR

) S

.O.

i.dd

Gal

( ?)

IS-2

al

l lay

ers

s.r.,

s.P

., i.s

,i.pd

, C

/A,

lc.

ec, m

s C

R( 5)

VIP

G

al(+

) 3s

low

firi

ng, w

ide

AP,

wid

espr

ead

(VIP

) S.

O.

i.dd

inhi

bito

ry in

puts

, low

-thr

esho

ld C

a2+

IS-1

al

l lay

ers

s.r.,

s.P.

, i.s

,i.pd

, C

/A,

lc.

ms,

mr,

(ec)

CR

V

IP( +

) SP

R( ?

), -

spik

es, o

scill

atio

n at

thet

a fr

eque

ncy

IS-3

al

l lay

ers

s.o.

, alv

. i.s

,i.pd

, C

/A, I

c. ec

, ms

CR

VIP

(?

) (V

IP)

i.dd

'Als

o in

oth

er s

ubfi

elds

; i.s

. in

the

CA

3 re

gion

and

in th

e de

ntat

e gy

rus.

20

nly

on a

xon

term

inal

s.

3Cha

ract

eris

tics o

f LM

cel

ls, w

hich

may

incl

ude

IS-2

VIP

cel

ls.

TAB

LE 4

.

CA

3 R

egio

n

Lam

inar

A

ffer

ent i

nput

N

euro

chem

ical

mar

kers

di

stri

butio

n of

In

trah

ippo

cam

pal

Rec

epto

rs

Cel

l typ

e D

endr

ites

Axo

n T

arge

ts

(exc

itato

ry)

Ext

rahi

ppoc

ampa

l C

BPs

/NO

S N

euro

pept

ides

et

c.

Phys

iolo

gica

l fea

ture

s

Axo

-axo

nic

Bas

ket 1

Bas

ket 2

0-L

M

Bis

tratif

ied

Rad

ial

trila

min

ar

Spin

y C

R

IS-1

(s

moo

th C

R)

IS-2

(V

IP)

all l

ayer

s al

l lay

ers

all l

ayer

s

s.o.

, s.r.

,

s.o.

, s.r

.,

all l

ayer

s

S.P.

, s.1.

S.I.,

s.p.

s.1.

all l

ayer

s

all l

ayer

s

S.P.

, S.O

.

S.P.

, S.O

.,

(s.r.

) S.

P., s

.0,

(s.r.

) s.1

-m.,

(s.o

., s.

r.)

s.r.

, s.o

.,

s.r.,

s.P

., 6.

P.)

S.O

.

s.L

m,,l

s.r.

, s.

m. (

0)

s.r.,

s.P.

,

s.o.

, alv

., s.

P., s

.r.

S.O

.

p.is

r.s

, pd

r.s, p

d

r.dd

r.pd

, s, d

d

r.pd

,s,d

d

r.dd

(?)

is, i

.pd,

i.dd

i.s, i

.pd,

i.d

d

C/A

, m

ossy

C

/A,

mos

sy

C/A

, mos

sy

C/A

, mos

sy

C/A

, mos

sy

C/A

, mos

sy

mos

sy

C/A

, m

ossy

C/A

, m

ossy

ec, m

s ec

, ms

ec, m

s

ms,

mr

ms,

mr

ms,

mr,

(ec)

ms

ms,

mr,

(ec)

ec, m

s

PV

PV

CB

( 2)

CB

(?)

CB, N

OS(

?)

CB

( 5)

CR

CR

CR

( ?)

-

-

CC

K, V

IP( 2)

SOM

, NPY

NPY

( ? ?)

NPY

(?),

CC

K(?

)

SOM

, NPY

(?)

VIP

( ?)

VIP

m22

, VV

A

m22

, Gal

V

VA

G

al, m

22(?

), SP

R(?

) m

Glu

R1

5-H

T-3

(?),

Glu

R2(

B)(

+)

SPR

(?),

Glu

R2(

B)

SPR

, m

Glu

Rl(

?)

SPR

( t ),

Gal

( ?) -

-

may

be

disc

harg

ed b

y a

sing

le

pyra

mid

al a

xon

term

inal

-

may

be

disc

harg

ed b

y a

sing

le

pyra

mid

al a

xon

term

inal

-

larg

e A

HP,

thet

a, g

amm

a,

rhyt

hmic

ity a

ctiv

ated

du

ring

den

tate

EEG

spi

kes

-

'Als

o in

oth

er s

ubfi

elds

; is.

in th

e C

A3

regi

on a

nd in

the

dent

ate

gyru

s.

20nl

y on

axon

term

inal

s.


A INTERNEURON OIA

Pl

I

2 -64

D

B PYRAMIDAL CELL n

c INTERNEURON O/A

-58

30 ms

-58

FIGURE 32. Action potentials and afterhyperpolarizations of interneurons at the CAI stratum oriens-radiatum border (A&) and pyramidal cells (B). A: A1 and A2. Averaged action potential at different time scales. Broken lime indicates resting membrane potential. Note lilige amplitude spike afterhyperpolarization (arrow). B: B1 and B2. Average action potential at fast and slow time scales. The spike afterhyperpolarization is small amplitude (arrow), followed by spike depolarization (double arrow) and a medium duration afterhyperpo-

interneurons. The terminals of interneurons release the inhibitory substance GABA and may also release peptidcs that colocalize with GABA in many types of interneurons (Section IV.3). The complex issue of GABA-mediated transmission has recently been reviewed (Mody et al., 1994), so only the major aspects will be briefly summarized here.

VIII.l. GABAergic Inhibition, Shunting, and Excitation of Principal Cells by Interneurons

Stimulation of afferent fibers elicits biphasic IPSPs in principal cells. The early phase o f this event is due to activation of GABAA receptors. Besides the early IPSP-inhibitory postsynaptic current (IPSC), activation of GABAA receptors also increases the membrane conductance and shunts excitatory currents (Staley and Mody, 1992). The late phase is mediated by Kt ion flux through channels linked by G-proteins to GABAs receptors. Numerous questions await answers regarding the modes of in-

larization (triple arrow). C: Depolarizing pulse evoked a train of action potentials (Cl) and a long afterhyperpolarization (arrow in C2). D: Responses to long depolarizing current pulses of increasing intensity (0.125-1.0 nA). Note progressive shortening of the intervals between action potentials. Reprinted from Lacaille J-C and Williams S (1990) "Membrane properties of interneurons in stratum oriens- alveus of the CAI region of the rat hippocampus in vitro." Neuroscience 36:349-359 by permission of Elsevier Science Ltd.

terneuron-mediated inhibition of principal cells and among interneurons themselves. Are separate groups of inhibitory cells rc- sponsible for activating postsynaptic GABAA and CABAS receptors (Alger and Nicoll, 1982a; Segal, 1990a; Muller and Misgeld 1991; Williams and Lacaille, 1990; Samulack and Lacaille, 1993; Williams et al., 1993)? If so, why are spontaneous GABAB-mediated synaptic events not visible (Staley and Mody, 1992)? What is the spatial relationship between the two receptor families? An early hypothesis suggested that dendritic inhibition is mediated by GABAH receptors and perhaps by a separate group of interneurons (Alger and Nicoll, 1982a; Segal, 1990a; Williams and Lacaille, 1992). In support of this hypothesis, IPSPs with similar kinetics to the GABAB-receptor-mediated responses are produced by activation of interneurons in stratum lacunosum-moleculare (Lacaille and Schwartzkroin, 1988b). However, the slow rise time o f somatic IPSPs in postsynaptic pyramidal cells may simply reflect elcctrotonic filtering by the dendrites (Soltisz and Mody, 1994). Furthermore, other experiments that have involved phar-


A extra (CAI int)

'a 5 mV P

in extra (CAI rad)

PP b 'u ' O m V 10 ms I - \

D

FIGURE 33. Feed-forward inhibition of CAI pyramidal cells by the perforant path input. A: Low threshold and monosynatic (6-8 ms) activation of an interneuron in the CAI pyramidal layer (putative basket or chandelier cell) by single-pulse stimulation of the perforant path (PP). Top three traces indicate stimulation with increasing intensity. Asterisk indicates volume-conducted population spike from the dentate gyrus. Bottom trace indicates frequency potentiation (4 Hz, 2 s). Triangle indicates CAI population spike. B: Inhibition of a CAI pyramidal cell. Simultaneous intrasomatic recording (intra) and extracellular field response in the stratum radiatum (extra). Black square indicates trisynaptically evoked Schaffer

collateral response. Arrow indicates early inhibition of the pyramidal cell at a latency comparable to the discharge of the putative bas- kedchandelier cell in A. C: Schematic interpretation of the physiological observations. The entorhinal afferents (PP) excite both pyramidal neurons (p) and chandelier (c) and/or basket (b) cells. The discharging interneurons in turn induce a strong perisomatic inhibition in the pyramidal cell. D: Degenerating perforant path terminals form synapses (arrows) on the distal dendrite of a PV-immunoreactive (putative basket or chandelier) interneuron in stratum lacunosum-moleculare. Data from Bum& and Eidelberg (I 982), Bum& et al. (1995), and Leranth et al. (1996).


macological manipulations show that dendritic inhibition brought about by intracellular activation of the presynaptic interneuron is mediated by GABAA receptors (Buhl et al., I994a, 1995). Is cooperation of several interneurons necessary for activating of GABAB receptors? Are presynaptic GABAB receptors present on terminals of all types of interneurons (Lambert and Wilson, 1993)? Is the ultrastructure of synapses responsible for GABAB IPSPs similar to the synaptic contacts involved in GABAA inhibition? Answers to these questions may come from paired recordings from presynaptic interneurons and postsynaptic principal cells. These experiments may demonstrate not only whether separate groups are involved in these two forms of inhibition but also reveal the type(s) of neurons involved and the number and location of synapses they form with their targets.

Basket cells, chandelier cells, and bistratified cells contact their principal cell targets with multiple (3-12) synaptic contacts (Buhl er al., 1994a,b, 1995; Miles et al., 1996; Halasy et al., 1996), and each cell may innervate between 1,000 and 2,500 pyramidal cells (Li et al., 1992; Gulyas et al., 1993a,b; Buhl et a]., 1994a; Sik et al., 1995). Single action potentials in presynaptic basket cells or chandelier cells evoke large (up to 2 mV) and fast IPSPs in target pyramidal or granule cells in less than 2 ms after the peak of the presynaptic action potential (Miles 1991; Buhl et al., 1994a,b, 1995). Pharmacological experiments with GABA receptor blockers demonstrate that GABA release from presynaptic terminals activate GABAA receptors located at the somata and axon initial segment of the postsynaptic principal cells. Synaptic responses evoked by the chandelier cells appear faster when recordings in pyramidal cells are made in the soma.

In the dentate gyrus, granule cell somatic membrane patches may have at least two single channel conductances (14 and 23 pS) i n response to GABA application (Edwards et a[., 1990). Because the number of release sites on the soma, axon initial segment, and dendrites are similar for these interneuron types, the differences in the response kinetics may arise from the subunit composition of the GABAA channels (Nusser et al., 1995). Furthermore, the magnitude and decay of IPSPs elicited by the same presynaptic basket cell in different pyramidal cells show substantial variability (Buhl et al., 1994a; but see Miles, 1991), although it is not clear whether such variability is due to the number of synapses established by the presynaptic interneuron or other factors.

In contrast to basket and chandelier cells, neurons in stratum lacunosum-moleculare are unidirectionally connected to principal cells because pyramidal cells do not innervate this interneuron class. In paired recordings, burst activity in LM interneurons produce slow rising and slowly decaying (80-90 ms and 90-100 ins, respectively), small amplitude IPSPs in intrasomatically (0.9 mV), and intradendritically (0.7 mV) recorded CA1 pyramidal cells, without any significant change in input resistance (Lacaille and Schwartzkroin, 1988b). Fast activity in the presynaptic interneurons also prevents the discharge of the pyramidal cell brought about by intracellular depolarization. IPSPs are not evident in pyramidal cells, however, when the presynaptic interneu- son is made to fire a single action potential. These findings suggest that interneurons in stratum lacunosum-moleculare require ,tronger synaptic activation for an efficient inhibition of pyramidal cells as opposed to basket cells, chandelier interneurons, and

bistratified cells, or they may have other functions than just action potential regulation (Section XIII.4).

Local application of GABA in the cell body layer in vitro typically produces hyperpolarization in pyramidal cells. However, when GABA is applied to the dendrites, depolarization may be observed. Strong electrical stimulation in the dendritic layers also evokes fast depolarization in pyramidal cells. Both the hyperpolarizing and depolarizing fast responses are mediated by GABAA receptors because they are abolished by GABAA receptor blockers and are enhanced by barbiturates (Alger and Nicoll, 1979, 1982b; Andersen et al., 1980; Perreault and Avoli 1989; Thalmann 1988). These experiments favor the hypotheses that (1) the hyperpolarizing and depolarizing responses are mediated by different receptor subtypes having different ion selectivity and/or (2) the distribution of chloride in somatic and dendritic regions is different (Alger and Nicoll, 1982b). Other observations, however, argue against such a suggestion (Gaiarsa et al., 1995). First, dendritic application of a low concentration of GABA during intrasomatic or intradendritic recording in pyramidal cells produces monophasic hyperpolarizing responses (Wong and Watkins, 1982; Newberry and Nicoll, 1985). Second, GABAA receptors are present on the axon initial segment, soma, and proximal and distal dendrites of principal cells (Houser et al., 1988; Gao and Fritschy, 1994; Nusser et al., 1995). Third, current source density analysis of population IPSPs in the presence of excitatory amino acid receptor antagonists shows that depolarizing IPSPs are evoked only when the amount of released GABA is potentiated by 4-aminopyridine (Lambert et d. , 1991). Fourth, paired recordings from interneuron-pyramidal cell pairs reveal hyperpolarizing GABAA IPSPs even when the interneuron innervates only the dendrites of the postsynaptic pyramidal cell. The rise time is slower and IPSP duration is longer when evoked by the interneurons innervating pyramidal cell dendrites as opposed to basket cells (Buhl et al., 1994a; Miles et al., 1996). Because the number of boutons established by presynaptic bistratified cells (n = 6) is similar to that of the basket cells, the slower kinetics of the IPSPs evoked by bistratified interneurons may reflect electrotonic attenuation due to the large distance between the dendritic synapses and the recording site. Another interpretation of the different kinetics of the interneuron-mediated responses is that they activate different subtypes of GABAA receptors at the soma and dendrites (Nusser et al., 1995). The fast GABAA-mediated current enters at or near the soma, decays rapidly (3-8 ms), is blocked by furosemide, and rapidly curtails the excitatory response. The slower GABAA-mediated current enters the dendrites, decays slowly (30-70 ms), and, importantly, is not blocked by furosemide (Pearce 1993). Subunit composition of GABAA receptors may also change as a function of development (Gaiarsa et al., 1995).

Recent experiments reveal that spontaneous inhibitory events (miniature IPSCs), resulting from action potential-independent release of GABA, arise mostly around the perisomatic region of principal cells (Solt6sz et al., 1995; Miles et a]., 1996), i.e., through fast GABAA receptors. An important practical implication of this finding is that experimental comparison of tonic inhibition due to spontaneous release of GABA from inhibitory terminals around the soma and evoked synaptic inhibition may be


assessing functions of overlapping but different interneuronal populations.

A possible conclusion from these experiments is that when only a small amount of GABA is released by the presynaptic terminals, the postsynaptic membrane of the pyramidal cell is hyperpolarized. In contrast, larger amounts of synaptically released or locally applied GABA may result in postsynaptic depolarization (depolarizing IPSP). These dose-dependent differences may be explained by the hypothesis that, at a low concentration of the transmitter, the GABA effect is primarily mediated by C1- ions. During intense GABAA receptor activation, however, the electrochemical gradient for CI- is diminished and the charges are carried mostly by the efflux of bicarbonate anions (Grover et al., 1993; Staley et al., 1995). Therefore, the hyperpolarizing or depolarizing nature of GABA may depend on the amount of transmitter released. The routinely used 95% 0 2 - 5 % COZ perfusion in slice experiments may facilitate the bicarbonate gradient and produce a more positive reversal potential for GABA. When bicarbonate is reduced by Hepes buffer and the slices are perfused by 100% 0 2 , GABA agonists or tetanic stimulation produces only hyperpolarizing responses (Staley et al., 1995). Because the interstitial concentration of CO2 in vivo varies from virtually zero to a relatively high value, depending on the energy consumption of the locally active neurons, the overall activity of surrounding neurons may affect the hyperpolarizing or depolarizing nature of GABA.

VIII.2. Possible Physiologic Effects of Colocalized Peptides

Although several subtypes of interneurons synthesize various peptides (Section IV.3), their physiologic role in the operations of the hippocampal formation has remained elusive. One major difficulty is that all experiments to date test the pharmacological effects of peptides on neuronal excitability, passive membrane properties, and/or transmitter release by pharmacological means rather than by physiological activation of interneurons. Furthermore, bath application or even pressure ejection of the peptide may act on multiple sites, some of which may exert an opposite effect on the measured experimental variable. Another typical interpretational problem is a mismatch between the axon terminals of the peptide-containing interneurons and the intrahippocampal distribution of receptors for the same type of peptide.

WII.2.a. SOM

Early studies on the effects of SOM reported both excitation and inhibition of pyramidal cells, depending on the dose and the site of application (Dodd and Kelly, 1978; Scharfman and Schwartzkroin, 1988). Perisomatic application of SOM consistently depolarizes pyramidal cells. The physiological relevance of such action is questionable, however, because SOM terminals are absent in the pyramidal cell layer, and it is, therefore, unlikely that interneuron-released peptide would diffuse from release sites at the distal dendrites to the soma. Extracellular studies on evoked field responses failed to observe any effect of SOM application to the apical dendritic layers (H.L. Haas, personal communication). At least part of the intrasomatically observed inhibitory effect is

thought to be mediated by interneurons because physiologically characterized interneurons in the CA1 strata oriens-alveus and pyramidale are powerfully excited by local application of SOM (Scharfman and Schwartzkroin, 1988). Subsequent studies have confirmed the potent inhibitory effect of SOM on GABA-mediated synaptic potentials. However, SOM has little effect on responses to GABA when GABA is exogenously applied (Scharfman and Schwartzkroin, 1989). These findings can be interpreted by assuming that SOM acts at a presynaptic location and depresses GABA release from presynaptic terminals of interneurons. In another study, SOM-induced hyperpolarization of CA1 pyramidal neurons was not blocked by picrotoxin or phaclofen. Although the peptide suppresses both GABAA and GABAB-receptor- mediated IPSPs, it has no significant effect on the EPSI’s. Hyperpolarization of pyramidal cells caused by SOM, however, is reduced in the presence of baclofen, thereby supporting a presynaptic interpretation of SOM action (Xie and Sastry, 1992).

VIII.2.b. NPY In many respects, the effects of NPY appear very similar to

that of SOM. Bath application of the peptide suppresses glutamatergic excitation of CA1 pyramidal cells but has no effect on the passive or active properties of the presynaptic CA3 cells (Colmers et al., 1987, 1988; Haas et al., 1987). It has been suggested that NPY also potentiates NMDA-mediated excitatory responses in CA3 pyramidal cells by an action at a sigma or phen- cyclidine binding site. The peptides NPY (which is active at Yl-Y3 receptors), [Leu31, Pro341NPY (a selective Y1 agonist), and NPYl3-36 (which mimics the effects of NPY in Y2 models) all enhance NMDA-induced activation of the pyramidal neurons in a dose-dependent manner but do not alter the activation of the same neurons by quisqualate (Monnet et al., 1992; de Montigny et al., 1993). A patch-clamp study, however, failed to see these effects in the same hippocampal region (Colmers, 1992).

Although the spatial overlap between excitatory inputs and NPY release sites is greatest in the outer molecular layer of the dentate gyrus, no effect of the peptide has been observed on synaptic inputs to the granule cells or on their passive electrical properties (Klapstein and Colmers, 1993). Recent experiments, however, have demonstrated that NPY suppresses depolarization-induced increases in intracellular Ca2+ concentrations in granule cells by inhibiting an N-type calcium current (McQuiston et al., 1996). An interesting implication of this finding is that N-type calcium channels are also involved in the release of dynorphines from the dendrites of granule cells (Simmons et al., 1995), and therefore NPY may be involved in the regulation dynorphin release.

VIII.2.c. CCK octapeptide (CCK8) In rat hippocampal slices, CCK and the selective CCK-B ag-

onist increases the basal release of endogenous glutamate and aspartate but not that of GABA. The effect is likely mediated by CCK-B receptors because a selective CCK-B receptor antagonist, L-365,260, completely reverses these responses (Migaud et al., 1994). Other experiments have suggested that CCK also affects the transient potassium current in hippocampal neurons by changing the voltage dependence of the inactivation and activa-


tion of the current (Saint and Buckett, 1991). Depression of synaptic excitability has also been demonstrated in an in vitro epilepsy model. CCK applied alone in nanomolar concentrations had no effect on spontaneous activity but did block the effect of kainic acid on synaptic transmission without affecting kainic-acid- induced bursting (Aitken et al., 1991).

VIII.2. d . V I P The first attempts at investigating the physiological effects of

VIP on hippocampal function failed to demonstrate any changes in pyramidal cell activity (Warnick and Pellmar, 1986). A later study, however, found that submicromolar concentrations of VIP, added to the perfusion medium, enhanced the excitability of CA3 and CA1 pyramidal cells, mostly by blocking the Ca2+ and cyclic AMP-dependent potassium current (IAHP) and spike frequency accommodation (Haas and Gahwiler, 1992). These effects persist in tetrodotoxin-containing medium, suggesting a direct effect on pyramidal cells. GABAergic IPSPs are also enhanced, likely reflecting excitation of some interneurons. The calcium regulatory effects of VIP has also been investigated in cultured rat hippocampal neurons by using the calcium-sensitive fluorescent dye fura-2. VIP increases [Ca2+]i but only at micromolar concentrations (Tatsuno et al., 1992).

VIII.2.e. SP Although both hippocampopetal afferents and hippocampal

interneurons may release SP (Section IV.3.0, the only physiological study in the hippocampal formation to date has concluded that it has no effect on pyramidal cells (Dodd and Kelly, 1981).

VIII.2.f. Peptide effects: Questions Several peptides exert a relatively selective action on the release

of GABA from the presynaptic terminals of interneurons. Assuming that peptides are not released without GABA, the peptide will most effectively regulate the release of GABA from the terminals of its parental interneuron on the basis of spatial proximity. Such an effect may be viewed as a variant of autoregulation. Because some interneurons with dendritic targets may not possess presynaptic GABAB receptors (Lambert and Wilson, 1993), an interesting possibility is that autoregulation of GABA release in those neurons may be carried out by a coreleased peptide. It is noteworthy that the axon collaterals of most peptide- containing interneurons target the dendritic domains of principal cells (Section IV.3). It is also possible that high-frequency firing or certain discharge patterns are a prerequisite for peptide release; therefore, presynaptic regulation of GABA release may be discharge pattern dependent.

Yet another potential role that may be assigned to interneuron-released peptides is that they act on the local blood vessels and thereby regulate blood supply to the active neurons. Obviously, such a putative function could not be evaluated in brain slice preparations. Nevertheless, interneurons are perfectly poised for such a function because they represent mean local neuronal activity with high accuracy. The activity of irregularly discharging principal cells is integrated by their common interneu-

rons and converted to changes in firing rate. O n the output side, axon terminals of interneurons are often found in the proximity of blood vessels, and several peptides have been associated with local regulation of metabolic activity and blood flow (Magistretti, 1990).

In future research on peptidergic function, it is essential that experiments more directly assess the physiological effects of peptides. A potentially effective and convincing approach in the as- sessment of peptide function is to record from identified interneuron-principal cells pairs. The physiological effect of peptide release may then be specifically tested with the aid of specific peptide blockers when such compounds are available.

All intrahippocampal and hippocampopetal afferents that terminate on principal cells also innervate interneurons (Buzsdki, 1984). However, very little is known about the way different interneuronal types respond to neurotransmitters and neuromodulators. Neurotransmitter actions on interneurons may be split into two major categories: fast synaptic excitation and inhibition and slow synaptic actions mediated by G proteins. Two neurotransmitters, glutamate and GABA, appear to exert both fast and slow actions via different sets of receptors. To date, most of the available information is indirectly inferred from studying the evoked or spontaneous GABAergic responses in principal cells. Although such indirect information is important, it typically cannot iden- t i k the interneuron types responsible for the action, nor can it reveal whether the different types of actions are mediated by the same interneuron or by distinct classes. Direct recordings from identified interneurons and the selective activation of their presynaptic inputs are needed to dismiss such ambiguities.

When considering the effects of transmitter-neuromodulator action, it is important to recognize the typical dual actions of several neurotransmitters on both the somadendritic surface and presynaptic terminal. In many cases, the apparent “paradoxical” and opposing dose-dependent effects of locally or bath-applied neurotransmitters and modulators may be explained by their distinct actions on interneuronal firing and inhibitory presynaptic terminals. GABA may be released into the synaptic cleft “spontaneously” (i.e., without a preceding action potential in the presynaptic terminal) or be triggered by the action potential-induced depolarization of the presynaptic bouton. Because the size of the action potential-independent IPSPs is much smaller, they are typically referred to as spontaneous “miniature” potentials (Fatt and Katz, 1952). Both the miniature IPSPs and action potential- evoked IPSPs are attenuated by GABA blockers but tetrodotoxin, which blocks sodium-dependent action potentials, can distinguish the two forms because it has no impact on the miniature IPSPs. Given this scenario, neuroactive compounds and clinically used drugs may modulate inhibition of principal cells by affecting both spontaneous and evoked release of GABA (Mody et al., 1994). To date, numerous studies performed with tissue cultures and brain slices have demonstrated that bath or local applications of

426 FRELIND AND BUZSAKI

neurotransmitters and drugs exert a measurable effect on synaptic release without, or in concert with, modulation of presynaptic K+ and Ca2+ currents. An obvious question in this context, however, is whether and how a given neurotransmitter or neuromodulator affects GABA release in the intact brain because the transmitter-modulator release sites are often spatially distant and the physiologically “relevant” concentrations are difficult to predict or measure. Because in vivo preparations often do not have the necessary analytical resolution, methods should include combined slice preparations in which afferent neurons can be selectively activated (T6th et al., submitted).

IX.l. Principal Cell-Interneuron Activation by Excitatory Amino Acids

Nearly all interneurons are innervated by the axon collaterals of principal cells of the hippocampal formation or parahip- pocampal structures. Dual recordings from connected principal cell-interneuton pairs have provided quantitative details on synaptic interactions. In response to intracellular stimulation of CAI pyramidal cells of the guinea pig, short latency unitary EPSPs (2 mV) are produced in basket cells (Knowles and Schwartzkroin, 1981) and interneurons located at the border of stratum oriens and the alveus (Lacaille et al., 1987). Afferent stimulation usually evokes powerful excitatory responses consisting of several action potentials.

In dual intracellular recordings from pairs of CA3 pyramidal cells, discharge of one cell often induces an IPSP in the other cell. Because pyramidal-pyramidal cell synaptic connections are asymmetric, the IPSP must be mediated by an intervening interneuronal synapse. Such disynaptic IPSPs have a short latency (3-5 ms) and a failure rate of less than 0.5 ms (Miles, 1991). Despite such a high fidelity of synaptic transmission, physiological measurements often implicate a single release site (Gulyis et al., 1993b; Arancio et al., 1994). Electron microscopy confirms that each concact is a conventional synapse with round vesicles and a single synaptic active zone of diameter 0.4-0.8 p m (Gulyis et al., 1993b). The mean estimated quantal amplitude is 0.8 mV, with a very high probability of transmitter release (0.360.75). Miniature synaptic events, recorded when presynaptic activity is suppressed, correspond to transmitter release from single presynaptic terminals. At connections wich only one release site, suppression of activity-dependent release should reduce EPSC frequency but should not change its amplitude distribution. This hypothesis was tested in whole-cell recordings from CA3 inhibitory cells (Arancio et al., 1994). Addition of tetrodotoxin and cobalt to suppress voltage-dependent inward currents carried by Nat and Ca2+ reduces EPSC frequency but does not affect EPSC amplitude distributions in most cells. In addition, similar distributions are observed for EPSCs initiated by weak focal stimuli intended to activate a single presynaptic cell and EPSCs recorded later in the presence of tetrodotoxin and cobalt. The single-channel glutamate-activated conductance has yet to be measured in hippocampal interneurons, but it is several times higher in neocortical interneurons than in pyramidal cells (27 pS vs. 9 pS; Hestrin, 1993).

The hypothesis that pyramidal cells make single contacts with

inhibitory cells has been tested fbrther by filling single CA3 pyramidal cells in vivo and visualizing their target PV-immunoreactive cells (Sik et al., 1993). In most cases, a single contact is observed between the pyramidal cell and the target interneurons (84.70/6), but double and triple contacts are also found (15.3%). The majority of the synapses are on the dendrites (84.1%) and less frequently on the cell bodies. Of the more than 15,000 boutons formed by the presynaptic pyramidal cell, only 2.1% contact PV-immunoreactive neurons (approximately 300 basket-chandelier cells).

Transmission between hilar mossy cells and interneurons has also been studied with paired-cell recordings. Action potentials in the presynaptic mossy cells consistently evoke EPSPs, and action potentials are often triggered by the depolarizations. This high fidelity of synaptic transmission is in contrast with the low probability of unitary EPSPs in mossy cell-granule cell pairs (Scharfman, 1995b).

Electrically excitable dendrites in interneurons? The anatomical evidence for a single release site between pyramidal cells and basket cells is difficult to reconcile with the faithful and rapid transmission of action potentials across this synaptic junction (Gulyis et al., 1993b; Arancio et al., 1994; Miles et al., 1996). This paradox may be explained, however, if the interneurori dendrites are electrically excitable (Spencer and Kandel, 1961; Llinas and Nicholson, 1971; Traub and Llinas, 1977; Poolos and Kocsis, 1990; Traub, 1995).

Several observations support the possibility of dendritic fast spike initiation in interneurons. First, the amplitude variability of extracellularly recorded interneurons sometimes exceeds SO%, an observation compatible with multiple site generation of action potentials throughout the dendritic arbor. Occasionally, a bimodal distribution of action potentials is observed (Fig. 34). Second, hyperpolarization of hilar interneurons by somatically placed electrodes reduces the amplitude but not the number of population burst-induced action potentials (Michelson and Wong, 199 1). Third, action potentials equally often emanate from the decaying and rising phases of spontaneous EPSPs without a fixed threshold in CA1 interneurons (Fig. 34; Sik et al., 1995), indicating a dissociation of the locally measured membrane potential and the

FIGURE 34. Possible dendritic generation of fast spikes in interneurons. A: Spontaneous activity of a CAI trilaminar neuron in vivo (urethane anesthesia). The boxed area is shown at a higher resolution. Open arrows indicate spike initiation. The spikes are not necessarily triggered at most depolarized levels of the membrane. The dissociation between local depolarizing potentials and the variability of the spike threshold suggest that fast spikes were initiated at location(s) other than the recording site. B: Four-site recording (tetrode) from a stratum oriens-interneuron in the CAI region in the awake rat. Small (a, arrowheads) and large (b) spikes are present. C: Spikes a and b at a faster scale. Note the double positive- going spikes in b. D: Spatial clustering of spike parameters revealed two possible units (a and b). E: However, cross correlation between spikes a and b revealed a reliable refractory period, suggesting that both spikes were emitted by the same neuron. An alternative possibility is that spikes a and b were generated by interneurons coupled by gap junctions. From K. Moore, Z. Nidasdy, and G. BuzsAki (unpublished findings).


A intracellular

B extracellular

a b E a versus b

I ... . I

-6.4 0 ms 6.4


occurrence of action potentials. Fifth, the critical site of the sodium spike generation in pyramidal cells, granule cells, and mossy cells is the axon initial segment (Spruston et al., 1995; Magee et al., 1995b), which is under the control of densely packed GABAergic terminals of chandelier cells (Somogyi et al., 1983a,b, 1985a; Li et al., 1992; Han et al., 1993; Halasy and Somogyi, 1993b). In contrast, the axon initial segment of interneurons is free of synapses (Halasy and Somogyi, 1993b; Ribak et al., 1993; Buzsiki and Sik, unpublished observations), suggesting that the axon initial segment of interneurons may not be as critical as in principal cells. Sixth, a recent computer model of the synaptic transmission between pyramidal cells and interneurons failed to evoke a short-latency action potentials in the interneuron having branching passive dendrites. However, dendritic EPSPs regularly triggered dendritic action potentials when the density of sodium channels in the dendrites was about half of the somatic region (Traub, 1995; Traub and Miles, 1995). Although these observation are compatible with the idea of active dendritic spike generation, determination of the sodium channel distribution along the somadendritic surface and simultaneous patch-clamp recordings are necessary for providing more direct evidence.

Activation of interneurons by principal cells is mediated by excitatory amino acids via AMPA, NMDA, and metabotropic receptors. These actions will be discussed below.

1X.l.a. Glutamate: AMPA receptors As discussed above, the composition of GluR receptors is dis-

tinctly different in interneurons and principal cells, and it is expected that unique combinations of receptor subunits will be discovered among the different interneuron subgroups (Jonas et al., 1994; Jonas and Burnashev, 1995; Geiger et al., 1995; Koh et al., 1995; Leranth et al., 1996; Section IV.6.a). The larger and faster evoked EPSPs observed in CA3 interneurons vs. those in pyramidal cells (Miles, 199 1) may be conservatively explained by less efficient cable attenuation in interneurons (Thurbon et al., 1994). Likewise, the rise time of sEPSCs in hilar interneurons is much faster (0.5 vs. 1 ms) than in mossy cells, and these interneurons are characterized by fast decay time constants (2-4 vs. 9-10 ms; Livsey and Vicini, 1992). Similar differences in decay time constants are obtained when neocortical interneurons are compared with pyramidal cells (2.5 vs. 4.6 ms; Hestrin, 1993).

Analysis of single-channel currents in outside-out patches from hilar interneurons have revealed larger single-channel currents in interneurons than in mossy cells. Desensitization of AMPA receptors is considerably faster in hilar interneurons (3-5 ms) than in putative mossy neurons, granule cells, or CA3 pyramidal cells (6-15 ms; Livsey et al., 1993; Geiger et al., 1995). Similar differences hold for comparisons of neocortical interneurons and pyramidal cells (3 vs. 12 ms; Hestrin, 1993). Calcium permeability of AMPA-gated channels depends on the presence of the GIuR2 subunit, and channels lacking GluR2 subunits in oocytes allow large influxes of Ca2+ in an order equivalent to NMDA ionophores (Hollmann et al., 1991; Seeburg, 1993). In the study by Geiger et al. (1995), Ca2+ permeability of visually identified hilar interneurons and basket cells was severalfold higher (PcA/PN~ =

0.69 - 1.59) than in mossy neurons, granule cells, or CA3 pyramidal cells (P,-JPN~ < 0.1). Furthermore, a recent patch-clamp single-cell polymerase-chain-reaction study in rat neocortex found that calcium permeability is higher in most nonpyramidal cells than in pyramidal neurons (Jonas et al., 1994). In addition to the low density of GluR2/3 subunits, interneurons have a relative abundance of GluR4 subunits (Leranth et al., 1996) and a higher proportion of the flip module at the R/G site of the subunits compared with principal cells (Geiger et al., 1995). The expression of Ca2' permeable AMPA receptors in interneurons may indicate the presence of a pathway for synaptically mediated Ca2+ entry (Koh et al., 1995). Ca2+ entering through AMPA receptors may activate Ca2+-mediated K+-conductance and contribute to the termination of the EPSC and action potential. In addition, the entering Ca2.+ could inactivate NMDA receptors (Medina et al., 1994) or may contribute to plastic changes of the synapse. Interestingly, Livsey and Vicini (Livsey et al., 1992, 1993) failed to observe significant Ca2+ permeability of AMPA receptors in putative interneurons. In stratum radiatum of the CA3 region, McBain and Dingledine (1993) identified two types of interneurons, the majority with lin- ear current-voltage properties and only a small portion of interneurons possessed inwardly rectifying current-voltage relationships (Fig. 35). The combination of physiological measurements, intracellular filling, and double labeling of interneurons with AMPA receptor subunits will be necessary to elucidate the importance of different receptor types in the different interneuronal classes and the functional importance of such expected differences.

IX.1.b. Glutamate: NMDA receptors

Direct evidence for the existence of NMDA receptors on hippocampal interneurons is available as a result of voltage-clamp analysis of NMDA and non-NMDA currents (Fig. 35; Sah et al.,

FIGURE 35. GluR-mediated effects on interneurons. A: Current-voltage relationship of AMPA receptor and spontaneous mIPCSs in a type 1 interneuron of CA3 stratum radiatum. Ai: Kainate activation of the AMPA receptors show a modest outward rectification. Kainate currents reversed at approximately 0 mV. Aii: Spontaneous mEPSCs recorded in the presence of bicuculline and tetrodotoxin. At a holding potential of -70 mV, mEPSCs are dom- inated by an AMPA-receptor-mediated component. At a holding potential of +50 mV, a two-component EPSC was present. Part of the response was blocked by D-2-amino-5-phosphonopentanoic acid, demonstrating that at positive potentials both AMPA and NMDA receptors are activated in the interneuron. B: Type 2 interneurons of CA3 stratum radiatum possess an inwardly rectifying I-V relationship in response to kainate (Bi). Bii: Averaged spontaneous EPSCs in a type 2 interneuron show only NMDA-receptor-mediated responses at positive potentials. D-APV completely blocked the slow kinetic response at +50 mV, but mEPSCs recorded at -70 mV were not altered. These observations suggest that AMPA receptors possess inwardly rectifying properties at positive potentials. C: 0-LM interneurons in the CA1 region are modulated by presynaptic mGluRs. Ci: Averaged evoked EPSCs are markedly potentiated in the presence of a metabotropic receptor agonist (ACPD). Cii: In addition, ACPD application may induce a large inward current with prolonged oscillatory episodes. Such responses were not observed in CAI basket cells. This figure was kindly prepared by Chris McBain.


Ai 6oo r I Aii

-600

-900 -80 -40 0 40

Membrane potential (mV)

Bi 600

300

r

Ci

-300 -

-600 -

-900 -

-1500

-80 -40 0 40 Membrane potential (mV)

Cii

' 1 0 0 , u M A C P D

Bii

+50mV Control b

I 50pM ACPD


1990; McBain and Dingledine, 1992; Perouansky and Yaari, 1993; Koh et al., 1995). The time course of the non-NMDA component of EPSCs is similar to that recorded in pyramidal cells. The magnitude of NMDA currents on average is somewhat smaller in interneurons, but the individual variability is high. The smaller size of the NMDA current may be important because it would explain why topical application of NMDA in hippocampal cultures evokes a smaller calcium influx in GAD-positive neurons than in pyramidal cells (Segal and Greenberger, 1992). Nevertheless, these findings demonstrate that interneurons also possess NMDA receptors, and despite the different anatomical structure of the synapses formed by the Schaffer collaterals on the two cell populations (spine versus shaft), their physiological and receptor-pharmacological properties are similar.

Owing to their slower kinetics, NMDA channels have a higher probability for synaptic summation than do the fast AMPA- receptor-mediated events, a property that may be particularly important for the maintenance of population synchrony. Inter- neurons may be activated by NMDA receptors under baseline conditions because the magnitude of recurrently evoked IPSPs in CA1 pyramidal cells is attenuated by NMDA blockade (Grunze et a]., 1996).

IX.1.c. Glutamate: Metabotropic receptors Synaptically released glutamate can lead to a short-term exci-

tation of hippocampal interneurons by activation of mGluRs. Application of the mGluR agonist, trans- 1 -amino-cyclopentane- 1,3-dicarboxylic acid (tACPD), reduces spike afterhyperpolarizations and induces slow rhythmic membrane oscillations in unidentified CA3 interneurons (Miles and Poncer, 1993). These interneurons in turn initiate large and rapidly decaying IPSCs in target pyramidal cells (Poncer et al., 1995). The mGluR-mediated activation may be specific or differentially effective on a subset of inhibitory neurons. Identified 0 -LM cells in the CA1 region are strongly excited by tACPD, responding with a large inward current and slow oscillation (Fig. 35), whereas basket cells respond with only a modest inward current (McBain et al., 1994). These physiological findings have been interpreted in light of the hypothesis that 0 -LM (SOM) cells, which specifically express mGluR1a receptors (Baude et al., 1993), are responsible for the tACPD-induced barrage of IPSP in pyramidal cells (McBain et al., 1994). A caveat in this hypothesized circuitry is that 0-LM cells, which have axon terminals on the distal apical dendrites of pyramidal neurons, may not be efficient enough for the fast and large IPSPs produced in the cell body of pyramidal cells in response to tACPD (Miles and Poncer, 1993). The fast effect may be mediated by the few stratum oriens collaterals of the 0-LM cells (Sik et al., 1995) or by other interneuron types that express either mGluRla or other mGluRs (Bradley et al., 1996).

In addition to their soma-dendritic domains, axon terminals of interneurons also express presynaptic mGluRs. Higher concentration of tACPD reduces the frequency of miniature IPSCs recorded from CA3 pyramidal cells in the presence of tetrodotoxin and the absence of Ca2+. In addition to the differences in effective doses of tACPD, kinetic and pharmacological differences sug-

gest that two species of mGluRs are involved in the effects on excitability and GABA release. tACPD excites inhibitory cells with a delay of 2-5 s, whereas the reduction of miniature IPSC frequency has a latency of about 100 s. Quisqualate, which has a much higher affinity for mGluRs 1 and 5 than for mGluKs 2/4, excites inhibitory interneurons but does not affect miniature IPSCs in pyramidal cells. Pharmacological activation of mGluR2/3 receptors reduces miniature IPSC frequency but does not affect spontaneous IPSCs in pyramidal cells. These data suggest that mGluRs 1 and 5 are responsible for the excitation of inhibitory cells, whereas the presynaptic action of glutamate is mediated by mGluR2/3 (Poncer et al., 1995).

It is not clear, however, how glutamate activates presynaptic mGluR receptors because they are spatially distant from glutamate release sites. The distance between the release site of glutamate to postsynaptic mGluRla receptor is about 50 nm (Baude et al., 1993); therefore, it may be activated with repetitive stimulation of excitatory afferents. However, the shortest distance between glutamatergic terminals and GABAergic terminals on dendritic spines is an order of magnitude larger (Halasy and Somogyi, 1993a). Thus, the physiological conditions under which presynaptic mGluRs of interneurons are activated are less clear (Poncer et al., 1994).

IX.2. Interneuron-Interneuron Interactions via GABA

Although it has been known for some time that interneurons innervate not only principal cells but also each other, exploration of the physiological interactions among the different interneuron classes and the implications of such interactions in network operations has just begun. The recent discovery that some hippocampal interneurons innervate only other interneurons but not principal cells (Section 111.4; Acsidy et al., 1996a,b; Gulyis et al., 1996) especially calls for detailed physiological investigations of the interactions among interneuron classes. Misgeld and Frotscher (1986) were among the first to demonstrate spontaneous and evoked IPSPs in interneurons of the hilar region and in the stratum lacunosum-moleculare of the CA3 area. These IPSPs were blocked by bicuculline. In dual recordings of interneuron pairs, the burst discharge of one interneuron in CAI stratum oriens-alveus induced an IPSP in its putative basket cell target (Lacaille et al., 1987). Burst activity in interneurons located in stratum lacunosum-moleculare (LM cells) also evoke ISPSs in interneurons of the pyramidal layer. These IPSPs are comparable to those seen in pyramidal cells, having a mean peak amplitude of 0.7 mV, peak latency of 60 ms, and decay time of 70 ms. In the opposite direction, activity of putative basket-chandelier cells does not produce any change in the membrane potential of LM interneurons (Lacaille and Schwartzkroin, 1988b). However, following pharmacological bockade of the ionotropic glutamate channels, monosynaptic IPSCs are evoked in LM cells by electrical stimulation of several distant regions, including strata radiatum, oriens, lacunosum-moleculare, and even the dentate molecular layer (Khazipov et al., 1995b). These in vitro observations suggest that LM interneurons are part of an extensive inhibitory network.

In response to afferent stimulation, the initial action potential

ZNTERNEURONS OF THE HIPPOCAMPUS 431

burst in basket cells and LM interneurons is followed by early (GABAA) and late (GABAB) IPSPs, which is similar in amplitude to that observed in pyramidal cells (Misgeld et al., 1989; Lacaille, 1991; Khazipov et al., 1995b; Sik et al., 1995). In identified basket cells in vivo, the intracellular membrane oscillation, associated with extracellular theta activity, shows a complete phase reversal between -70 mV and -80 mV, which is in the range of the chloride equilibrium potential. These observations indicate that rhythmic IPSPs in basket cells derive from other hippocampal or extrahippocampal inhibitory cells and are mediated by GABAA receptors (Ylinen et al., 199513). Quantitative data on interneuron-interneuron hnetics or connectivity are scarce. In one study, a biocytin-filled CA1 basket cell made synaptic contacts with at least 60 other PV-immunoreactive cells, in addition to innervating more than 2,000 pyramidal cells (Sik et al., 1995). This ratio is similar to the ratio of PV-immunoreactive cells to pyramidal cells (Aika et al., 1994), indicating no preferential innervation of interneurons or pyramidal cells. In the dentate gyrus, a single NPY-immunoreactive (HIPP) cell was estimated to innervate 4 0 0 4 5 0 other PV-immunoreactive interneurons with an average of three putative contacts per cell (Sik et al., submitted).

The first experiments to address the functional aspects of interneuron-interneuron interactions were done in the hilar region by using a pharmacological blockade of fast excitatory amino acid neurotransmission (Michelson and Wong, 199 1,1994). Inter- neurons were activated by the convulsant compound 4-aminopyridine or by elevation of extracellular potassium concentrations. Population bursts occurred in a group of hilar neurons coinci- dent with large amplitude IPSPs in pyramidal cells and granule cells. In the presence of the GABAA receptor-blocker picrotoxin, bursts and the underlying depolarizing synaptic potentials were completely suppressed in some, but not all, interneurons. Synchronized IPSPs in principal cells continued to occur, but they were less complex in appearance. These findings were explained by the suggestion that GABA is an excitatory transmitter on interneurons (Michelson and Wong, 1991). An alternative hypothesis is that, during 4-aminopyridine-induced bursts, GABA release is intense, and therefore bicarbonate will carry most of the charges through GABAA receptors and depolarize interneurons (Staley et al., 1995). A caveat in this explanation is that principal cells were hyperpolarized rather than depolarized during the interneuron bursts. Thus, at least some differences likely exist between principal cells and interneurons in this respect.

In interneurons that continue to fire bursts after GABAA receptor blockade, intracellular hyperpolarization attenuates the amplitude of action potentials. Intracellular injection of the fluorescent dye Lucifer Yellow into these cells often results in the labeling of two to four cells (dye coupling). It has been argued that a subpopulation of GABAergic neurons can become synchronized by a mechanism that involves gap junctions (Michelson and Wong, 1994). The dye coupling results, however, should be taken with caution as evidence for electronic coupling. Although some interneurons are indeed interconnected by gap junctions and zona adherentia (Katsumaru et al., 1988a; Gulyis et al., 1996), intra- '

cellular labeling of interneurons in the intact brain never results in dye coupling (Buckmaster and Schwartzkroin, 1995a,b; Sik et

al., 1995, submitted). However, even pyramidal cells are dye coupled in the epileptic hippocampus both in vitro and in vivo (Perez- Velazquez et al., 1994; Penttonen et al., 1995), although the existence of gap junctions among principal cells has yet to be demonstrated. The possible physiological role of interneuronal interactions will be considered further in Section XIV.

IX.3. Acetylcholine Cholinergic activation also exerts a dual effect on the GABA

interneuronal system. Although acetylcholine can directly modulate intrinsic ionic conductances and calcium channels of pyramidal cells and affect presynaptic terminals of principal cells (Kmjevic et al., 1971; Benardo and Prince, 1982; Cole and Nicoll, 1984; Madison et al., 1987; Gahwiler and Brown, 1987), similar effects on interneurons make cholinergic regulation particularly complex. As discussed earlier, interneurons in the stratum oriens-alveus region possess particularly dense cholinergic receptors (Section IV.6). Muscarinic agonists or direct stimulation of cholinergic fibers induce a prominent and sustained increase in the frequency and amplitude of spontaneous IPSCs in principal cells, even when fast excitatory neurotransmission is blocked pharmacologically (Pitler and Alger, 1992; Behrends and Ten Bruggencate, 1993). At least part of the effect is related to excitation of interneurons by modulation of Kt conductances because cholinergic agonists induce a fast muscarinic excitation of interneurons in CA1 alveus-oriens and the pyramidal layer (Recce and Schwartzkroin, 1991). In these same preparations, however, muscarinic receptor activation decreased the frequency of miniature IPSCs, suggesting a cholinergic suppression of GABA release at inhibitory terminals. This latter observation is supported by recent morphological evidence showing high density of muscarinic-2 receptors on the terminals of interneurons that innervate the cell bodies and axon initial segment of pyramidal cells (Section IV.6.d). It remains to be determined whether suppression of GABA release is affected at all terminals of the interneuron or occurs selectively on GABAergic terminals contacting the principal cells.

In urethane-anesthetized rats, iontophoreretic application of atropine on putative interneurons reduced their firing rates during theta rhythm, but did not substantially change the discharge rate of successively recorded pyramidal neurons (Stewart et al., 1992). Firing rate changes of putative interneurons have also been reported in aged rats having possible cholinergic deficit (Mizumori et al., 1992).

Given the laminar distribution of the cholinergic terminals and the somadendritic segregation of the terminals of the various interneuronal classes, it is unlikely that acetylcholine in the intact brain equally affects all interneuron types. Such discriminative action of acetylcholine is also evident in its presynaptic action because the terminals of basket cells and chandelier cells have an especially high density of muscarinic-2 receptors (Section IV.6.d). These principles are, of course, also valid for the evaluation of other neurotransmitters acting on principal cells, interneurons, and presynaptic terminals, and only direct recordings from interneurons will reveal whether and how the different interneuronal classes are involved in each action of the neurotransmitter.


IX.4. Norepinephrine In vivo studies have suggested that the overall effect of norep-

inephrine release in the hippocampus is to suppress pyramidal cell discharge (Segal and Bloom, 1974). The suppressive action of norepinephrine may be mediated disynaptically by direct excitation of interneurons. Local application of norepinephrine suppresses the activity of pyramidal cells but excites putative interneurons in vivo. The suppression of pyramidal cells is dominantly mediated by a 1 receptors. However, a 2 and P-adrenergic agonists excite both pyramidal cells and putative interneurons. These findings raise the possibility that a large part of the norepinephrine effect is mediated by interneurons and that the differential responses arise from the activation of distinct populations of noradrenergic receptors (Pang and Rose, 1987). These suggestions are supported by recent in vitro experiments. The norepinephrine-induced excitation of biocytin-filled CAI interneurons is mimicked by an a1-adrenoceptor agonist (phenylephrine), persists in the presence of an cr2-adrenoceptor agonist (atipamezole), and can be blocked by a selective al-adrenoceptor antagonist (Bergles et al., 1996). A P-adrenoceptor-dependent depolarization also occurs in those CA1 interneurons that display time-dependent inward rectification. The axon arborizations of the filled cells indicate that the majority of the recorded cells are basket cells. The P-adrenoceptor agonist isoproterenol also increases the spontaneous discharge rate of putative hilar interneurons and the frequency of GABAA IPSPs in granule cells (Bijak and Misgeld, 1995). Interestingly, in the presence of AMPA and NMDA blockers, norepinephrine reduces the discharge rate of putative hilar interneurons, indicating a complex interaction between interneurons and granule cells or a possible differential effect of norepinephrine on different types of interneurons. Overall, the norepinephrine-induced increase in the firing frequency of inhibitory interneurons and the resulting increase in both the frequency and amplitude of IPSPs in their target principal cells may account for the decrease in spontaneous activity of pyramidal neurons following activation of the locus coeruleus in vivo.

IX.5. Serotonin Actions of serotonin are mediated by second-messenger-linked

5-HT1, 5HT2, and 5-HT4 receptors and membrane-ion- channel-linked 5-HT3 receptors (Bobker and Williams, 1990). Serotonin treatment leads to an enhancement of evoked population spikes both in vivo (Klancnik et al., 1989; Richter-Levin and Segal, 1990) and in vitro (Ropert 1988), yet it hyperpolarizes pyramidal cells (Andrade and Nicoll, 1987; Colino and Halliwell, 1987) and granule cells (Ghadimi et al. 1994). The hyperpolarizing effects of serotonin may be direct on principal cells or may be mediated by excitation of interneurons. It has been inferred from studies on principal cells that a subset of interneurons may be selectively conveying the serotonin-mediated suppression of pyramidal cell activity. Serotonin increases both the frequency and amplitude of GABA-A receptor-mediated unitary IPSPs in pyramidal cells. These effects persist when fast neurotransmission is blocked pharmacologically but is reduced by bicuculline and by 5-HT3 antagonists (Ropert and Guy, 1991). Whole-cell voltage-

clamp recording from interneurons in the CA1 strata radiatum and lacunosum-rnoleculare revealed fast and large inward currents that were attenuated by a 5-HT3 antagonist. Importantly, the effects of serotonin on interneurons is not blocked by bicucullin or pharmacological blockade of fast excitatory neurotransmission, suggesting a direct action of the neurotransmitter (Kauer and McMahon, 1995).

Similar effects have been observed in the dentate gyrus. Local application of serotonin elicits and evokes a transient “burst” of IPSPs that are blocked by tetrodotoxin or bicuculline and the 5- HT3 receptor antagonist dolasetron (Piguet and Galvan, 1994). Focal application of serotonin or a selective 5-HT3 receptor agonist on visually identified “basket cells” in the granule cell layer induced a train of action potentials superimposed on a baseline membrane depolarization. Under voltage-clamp conditions, serotonin evoked an inward current that was accompanied by a multitude of small inward currents of short duration (<I00 ms) caused by serotoninergic excitation of nearby GABAergic presynaptic neurons innervating the recorded principal cell. The serotonin-induced effects in presumed basket cells are blocked by spc- cific blockers of the 5-HT3 receptor subtype. Kawa (1994) concluded that the excitatory action of serotonin occurred exclusively on GABAergic interneurons.

Intraperitoneal injection of a 5-HT3 antagonist, ondasetron, in the awake rat facilitates the induction of long-term potentiation in the CA1 region (Staubli and Xu, 1995). The results support the hypothesis that the drug removes the excitatory effects of serotonin on a subset of hippocampal interneurons.

Other experiments have suggested that 5-HT1A receptors may also be involved in the interneuron-mediated actions of serotonin on pyramidal cells (Segal, 1990b; Ghadimi et al., 1994; Schmitz et al., 1995). Evoked fast and slow IPSPs in CAI pyramidal cells are reduced by the 5-HTlA-receptor agonist 8-OH-DPAT, and this effect is blocked by the 5-HTlA-receptor antagonist NAN- 190. These drugs have no effect, however, on GABA-mediated currents evoked by GABA application to the dendritic or somatic layers, suggesting the involvement of interneurons. Minimal concentrations of serotonin appear to reduce selectively the late, GABA-B receptor-mediated component of evoked IPSPs in both the CA1 (Segal, 1990b) and CA3 regions (Oleskevich and Lacaille, 1992) and in the dentate gyrus (Ghadimi et al., 1994). Importantly, putative interneurons impaled in the pyramidal layer were hyperpolarized and the evoked EPSPs were substantially reduced by both serotonin and 8-OH-DPAT (Schmitz et al., 1995). These results were interpreted by assuming that 5-HTIA receptors directly inhibit interneurons or reduce glutamate release from the presynaptic axon terminals onto interneurons.

Overall, these physiological data suggest that at least part of the action of serotonin on principal cells is mediated by GABAergic interneurons in two different ways: serotonin (1) excites, via 5-HT3 receptors, a subset of hippocampal interneurons, thereby producing GABA-A receptor-mediated IPSPs and (2) may reduce GABA release by activating presynaptic 5-HTlA receptors of hitherto unidentified interneurons. Anatomical evidence for a specific serotonergic innervation of different interneuron types and for the selective interneuronal expression of


5-HT3 receptors is also available to support some of these conclusions (Sections IV.6.c, V.3).

IX.6. Opioids Opiates and the opioid peptide enkephalin may disinhibit hip-

pocampal pyramidal neurons by reducing inhibitory synaptic transmission mediated by GABAergic interneurons (Zieglgiinsberger et al., 1979; Nicoll et al., 1980; Siggins and Zieglgansberger, 1981). Application of met-ENK and opioid agonists decreases the firing rates of putative interneurons in vivo (Zieglgksberger et al., 1979; Raggenbass et al., 1985; Pang and Rose 1989) and hyperpolarizes them in vitro (Madison and Nicoll, 1988). Recent experiments, however, have pointed to the presynaptic effects of opiates (Lambert et al., 1991; Lambert and Wilson, 1993). Both in hippocampal slice cultures and in hippocampal slices, preceptor agonists decreases the frequency of miniature IPSCs without caus- ing a change in their amplitude. At the same time, the amplitude of action potential-dependent GABA release is reduced (Cohen et al. 1992; Reckling, 1993). This reduction suggests a direct action of opioids on the inhibitory presynaptic terminals. Such a presynaptic mechanism may also contribute to the action potential-evoked GABA release by decreasing synaptic inhibition. These presynaptic effects of p opioids do not exclude the possibility that delta or kappa receptors are involved, although iontophoretic and systemic application of a selective kappa-receptor agonist (U- 50488H) has little effect on either granule cells or putative hilar interneurons in vivo (Mayer et al., 1994). Future experiments should reveal the types of interneurons affected and whether different receptor subtypes are involved in the somadendritic and presynaptic effects of opioids.

X.1. Feedback and Feed-Forward Inhibition Inhibitory interneurons in cortical structures provide stability

to the activity of the principal cell populations by feedback and feed-forward inhibition. Groups of interneurons tonically or pha- sically hyperpolarize and/or increase membrane conductance (“shunting”) in the perisomatic and/or dendritic regions of neurons and thereby decrease the efficacy of excitatory afferents in discharging their principal cell targets. Activation of hippocampal interneurons may be brought about by extrahippocampal inputs, by intrahippocampal inputs afferent to interneurons (both feed-forward), or by principal cells of the same hippocampal region (recurrent or feedback). In the feed-forward regulatory system, afferent volleys directly activate the inhibitory neuron (first went) that in turn reduces the probability of firing of the principal cells (second event). In the feedback system, an excitatory input discharges the principal cells, whose excitatory output is fed hack to the inhibitory cell(s) through recurrent axon collaterals (Andersen et al., 1964). The inhibitory interneuron(s) then may discharge and inhibit a group of principal cells, including those that initially activated the interneuron(s). In short, the directions

of firing rate changes of local principal cells and inhibitory interneurons are the same in the feedback systems but opposite in the feed-forward scheme (Buzsiki, 1984).

It has been hypothesized that all extrahippocampal and intrahippocampal pathways that terminate on principal cells also innervate hippocampal interneurons (Buzsiki, 1984). Some interneuron subtypes are innervated exclusively by extrahippocampal afferents (e.g., MOPP cells, LM cells; Section 111.3) and are therefore part of the feed-forward mechanism only (Fig. 36). Most of these interneurons are located in the stratum lacunosum-moleculare and the dentate molecular layer. However, the majority of interneurons in strata pyramidale, granulosum, oriens, and hilar regions are innervated by both “intraregional” and “extrare- gional” and by extrahippocampal inputs and are therefore part of both feedback and feed-forward mechanisms (Buzsiki, 1984; Lacaille et al., 1987). Basket cells are regarded as the “arche type” of feedback (recurrent) interneurons, even though other in- rerneuron types (e.g., trilaminar interneuron, 0-LM cells, HIPP cells) may better fit such requirements (Blasco-Ibanez and Freund, 1995; Sik et al., 1995, submitted). Putative and anatomically identified basket cells are bidirectionally connected to principal cells (Lacaille et al., 1987; Buhl et al., 1994a). Because the time required for the onset of recurrent inhibition in pyramidal cells is shorter than the latency of the action potential-induced depolarizing afterpotential, repetitive spiking can be prevented altogether (Miles et a]., 1996).

Feed-forward inhibition is particularly strong in the hippocampus, although this is often not appreciated when electrical stimulation is used. An obvious problem with artificial stimulation is that recruitment of inhibition and excitation are nonlin- early coupled and the “relevant” physiological patterns are not mimicked properly by stimulation. Low-intensity stimulation as a rule evokes only feed-forward inhibition in all pathways of the hippocampal formation due to the lower discharge threshold of interneurons (Buzsiki, 1984). Additional excitation may or may not override the inhibition, depending on the afferents involved. For example, stimulation of the entorhinal-CA1 path (Bragin and Otmakhov, 1979; Doller and Weight, 1982; Yeckel and Berger, 1990; Colbert and Levy, 1932; Soltksz et al., 1993; Buzsiki et al., 1995), commissural-dentate gyms path (Deadwyler et al., 1975; McNaughton et al., 1978; Buzsiki and Czeh, 1981; Buzsiki and Eidelberg, 1981, 1982), and the dentate associational path (Scharfman, 1394, 1995b) typically evokes hyperpolarization in the target principal cells due to the strong feed-forward inhibition parallel with direct excitation (Fig. 33). Nevertheless, activation of these same pathways by natural stimuli may excite and discharge target cells, probably by suppressing feed-forward activation of the interneurons (Buzsiki et al., 1994, 1995).

X.2. Inhibition, Disinhibition, and Interpretations

When interneurons are serially connected, it is assumed that increased activity of the primary interneuron will lead to increased firing of the target of the secondary interneuron through a process generally referred to as “disinhibition” (Fig. 37). As an example,


CAI pyr

CAI pyr

CAI rad

CAI rad . I

100 0 msec 100

FIGURE 36. Interneuronal activity during sharp wave-related ripples in the CAl region. Averaged field activity reveals a short- lived 200-Hz field oscillation (trace). Top histogram: Auto-correlation of spikes from a putative basket cell in the pyramidal layer. Note increased activity during field ripple and phase modulation of the cell discharge with individual ripple waves. Middle and bottom histograms: Autocorrelations from unidentitied fast fuing interneurons in the stratum radiatum. Interneurons that decrease their discharge frequency or remain unaffected during ripples are very rare.

stimulation of the medial septum enhances population discharges of granule cells in response to perforant path activation (Fig. 38; Alvarez-Leefmans and Gardner-Medwin, 1975). Tetanic stimulation of the medial septal area reduces the frequency of ongoing IPSPs in pyramidal cells, which is associated with an increase of input resistance (Krnjevic et al., 1988). Similarly, stimulation of the septal area in a combined septdhippocampal slice suppresses spontaneous IPSCs in pyramidal neurons (Fig. 38; T6th et al.,

submitted). Because the in vitro experiments were done in the presence of atropine and AMPA and NMDA blockers, it is very likely that septal GARAergic afferents innervating hippocampal interneurons were responsible for the effect (Section V. 1). The suppression of hippocampal interneurons by septal GABAergic cells may be responsible for the enhancement of granule cell discharge (Fantie and Goddard, 1982; Bilkey and Goddard, 1985).

Whereas such stimulation studies will continue to provide important information about connectivity, they often fail to explain firing patterns of populations of neurons. For example, both putative septal GABAergic neurons and hippocampal iqterneurons increase their firing frequency during hippocampal theta activity, and this relationship holds after pharmacological blockade of the septohippocampal cholinergic afferents (Buzsjki et al., 1983; Stewart and Fox, 1989, 1990). As discussed in Section VI.2, CB- immunoreactive interneurons in CA1-CA3 stratum oriens project back to septal GABAergic cells, and such a feed-back loop may substantially modify the firing relationship of neurons during normal physiological operations (Fig. 37). Another type of observation that is not easily compatible with simple schemes of inhibition-disinhibition is that when networks are entrained into oscillations of various frequencies, interneurons and principal cells often discharge with virtually no time lag (Fox and Ranck, 1986; Bragin et al., 1995b; Ylinen et al., 1995a). O n the basis of “Boolean logic” (Fig. 37), one might predict that interneurons should discharge with some delay after the principal cells. In interconnected networks of interneurons, even the direction of frequency changes of individual neurons is hard to predict.

These examples aim to illustrate that, although inhibition and disynaptic disinhibition continue to be useful concepts in the description of physiological effects of interneurons, application of Boolean logic often fails to provide correct predictions in systems where interneurons are interconnected with each other (Fig. 37, Moser, 1996). In networks of interneurons, independent of whether members are connected unidirectionally or mutually, 0s- cillatory activity often emerges (Perkel and Mulloney, 1974; Wang and Rinzel, 1993; Bragin et al., 1995a; Whittington et al., 1995; Wang and Buzsriki, in press; Traub et al., 1996), and the rules that govern the timing of action potentials and the frequency changes of the participating cells can no longer be inferred from the Boolean logic of inhibition and disinhibition (Sections XIV. 1).

X.3. Effectiveness of Inhibition In Vitro and In Vivo

What is the physiological implication of the powerful connectivity between principal cells and interneurons? The in vitro data indicate that the discharge of single interneurons can modify the phase of intrinsic oscillations in its target pyramidal cells and/or prevent their firing altogether (Cobb et al., 1995; Miles et al., 1996). Conversely, a discharging pyramidal cell may induce firing in many of its target interneurons (Gulyk et al., 193313). Because a single pyramidal cell innervates hundreds of interneurons and interneurons in turn innervate 1,000-3,000 pyramidal cells (Li et al., 1992; Buhl et al., 1994a; Sik et al., 1995), a discharging single pyramidal neuron is expected to result in inhibition of tens of thousands of pyramidal cells. Such reasoning may

1NTERNEURONS OF THE HIPPOCAMPUS 435

explain the very low firing rates of principal cells in the slice preparation and in the hippocampus in vivo following removal of its subcortical afferents (Buzsiki et al., 1989). The situation in the intact hippocampus is likely different, however. The effectiveness of feedback activation of interneurons is subject to state-dependent variables. For example, putative and identified basket cells and pyramidal cells in the anesthetized rat can discharge on the opposite phases of theta at least 100 ms apart (Buzsiki and Eidelberg, 1983; Ylinen et al., 1995b; Skaggs et al., 1996). In fact, it is believed that the phase-locked discharge of pyramidal cells during theta is due to their rhythmic inhibition by the interneurons (Fig. 39), with little contribution by the principal cells’ drive to the interneurons (Buzsiki eE al., 1983; Fox and Ranck, 1986; Solt6sz and Deschenes, 1393). Attenuation of the powerful excitatory principal cell-interneuron feedback in vivo may be brought about by subcortical neurotransmitters. Neurotransmit- ters and neuromodulators suppress excitatory synaptic transmission in the hippocampal formation, including acetylcholine (Hounsgaard, 1978; Valentino and Dingledine, 1981; Dutar and Nicoll, 1988), adenosine (Dunwiddie and Haas, 1985; Greene and Haas, 1985; W u and Saggau, 1994), and glutamate via presynaptic metabotropic glutamate receptors (Miles and Poncer, 1993). Furthermore, acetylcholine may also modulate somatic inhibition due to the high density of muscarinergic-2 receptors on axon terminals of basket cells and chandelier cells (Hijos et al., submitted). Understanding the mechanisms of such behaviorally regulated attenuation of recurrent excitation and inhibition (“un- coupling”) by subcortical neurotransmitters and modulators will likely provide clues to their importance in plasticity and other hippocampal function (Hasselmo and Bower, 1993).

FIGURE 37. Inhibition, disinhihition, and problems of interpretation in interneuronal networks. A: The “Boolean” interpretation of inhibition and disinhibition in a simple serial circuitry. Activation of a primary GABAergic interneuron (i) inhibits the secondary interneuron. As a result, the tonic background inhibition onto the principal cell (p) is removed, i.e., the principal cell is disinhibited. B,C: Ambiguity of firing changes in interconnected networks. B: Simple case. Secondary interneurons are unidirectionally connected (il-i2). Excitation of the primary interneuron (i) inhibits secondary interneurons (il, 2) . In turn, the targets of i l become disinhibited. However, the discharge rate changes of interneuron i2 depend on the relative strengths of i-i2 and il-i2 connections. As a result, targets of i2 may be either disinhibited or more inhibited. In the example, dendritic disinhibition may be coupled with increased or decreased somatic inhibition. Such circuitries exist in the hippocampus (see D). C: With increased interneuronal connectivity, firing changes of principal cells are more complex. If the synaptic strengths are different (e.g., arrow), subpopulations of principal cells (pl, p2) may show differential or even opposite changes. D: Similar interneuron types may be reciprocally connected (e.g., basket cells; open arrows), whereas some interneuron types are unidirectionally connected to other interneurons (e.g., NPY-immunoreactive cells in hilus; filled arrows). Both interneuron types may be innervated by a common inhibitory input (e.g., septal GABAergic afferents). Such “master-slave” configurations allow for an immediate synchronization and phase locking of a large number of principal cells (gray tri- angles), but the direction of their firing changes is ambiguous.

Inhibitory circuits may modify the long-term excitability of principal cells in several ways. It has been long recognized that inhibition plays a critical role in electrically induced synaptic plasticity because bath application of GABAA antagonists enhances postsynaptic depolarization brought about by the afferent tetanus and consequently facilitate the induction of long-term potentiation (Bliss and Lomo, 1973; Wigstrom and Gustafsson, 1985;

A

1 1 1 excitation inhibition disinhibition

B L

1 ?

C

?

? ?


C lm mv

200 6 st imulation

T

E

P cell

100 ms

septum dentate gyrus


FIGURE 38. A,B: Septohippocampal sections (80 pm) cut on a Vibratome at a combined parasagittal and oblique angle after bend- ing the fimbria-fornix (@. This sectioning plane allows the septum, hippocampus, and the connecting fibers to be obtained in one plane. A group of choliergic neurons is present in the medial septum (MS) in A; in an adjacent section (B), the same region contains numerous PV-positive (GABAergic) neurons. The framed area is shown at higher magnification in the insert, where many PV-positive fibers (arrows) crossing from the septum to the hippocampus are clearly visible. The fimbria (fi), fornix (fk), the anterior commissure (AC), and subfields of the hippocampus (CAI, CA3, DG) are also indicated. C,D: Electrophysiological studies of GABAergic septal effects on hippocampal inhibition. Acute slices of 400 fim were cut in the same way as in A and B and maintained in vitro. The presence of CNQX, APV, and atropine in the bath ensured that all ionotropic glutamatergic and muscarinic cholinergic transmission was blocked. C: Spontaneous IPSPs recorded with C1-filled electrodes in the hippocampus were suppressed during repetitive stimulation (20 Hz, 8 V, 50-100 p pulses) of the medial septum. The histogram accumulates data from several traces of 12 cells. D: Stimulation in stratum radiatum of the hippocampus (H) evoked fast and slow IPSPs both in an interneuron (I cell) and in a pyramidal cell (P cell) recorded in the CA3 region. In contrast, septal stimulation (S) evoked an IPSP only in the interneuron, whereas the pyramidal cell membrane responded with a slow, small depolarization to the same stimulation. Data from Toth et al., (submitted). E: Septal facilitation of population spikes evoked by entorhinal stimulation (left) and simultaneously recorded single unit responses (right) in the dentate gyrus. Single pulses delivered to the septum (s) failed to evoke detectable field responses (top traces). Near-simultaneous stimulation of the septum and the perforant path input (PP) resulted in substantial faciliation of the evoked population spike (left bottom trace). Righ~ The intensity of the PP stimulation was decreased so that the single putative granule cell was discharged less than half of the time. When the same PP stimulation was preceded by a single-pulse septal stimulation, the neuron was reliably discharged and at an earlier latency (CzLh and Buzsiki, 1979, unpublished findings). Bottom: Interpretation of septal facilitation from Bilkey DK and Goddard GV (1985) Medial septal facilitation of hippocampal granule cell activity is mediated by inhibition of inhibitory interneurons. Brain Res 361:99-106 by permission of Verlag. GABAergic medial septal neurons are assumed to inhibit interneurons (i) and in turn disinhibit granule cells (g) in the dentate gyrus. Scale bar = 1 mm.

Abraham et al., 1986, 1787; Bliss and Collingridge, 1993). Conventional wisdom held that long-term alteration of synaptic transmission took place only at excitatory synapses, and inhibitory interneurons at best only modulated such changes. This assumption is based on the theoretical expectation that modification of synaptic strength among principal cells is the basis of specific associative memory and on the generally accepted view that both long-term potentiation (LTP) and long-term depression (LTD) depend on activity-dependent Ca2’ increase in the postsynaptic neurons (Bliss and Collingridge, 1993; Lisman and Harris, 1993; Singer, 1975). However, recent experiments have suggested that GABAergic synapses on principal cells and interneurons may also undergo long-term modifications (cf. Marty and Llano, 1995). The interneuron circuitry may be modified in a number of ways, including (1) presynaptic changes of excitatory terminals on interneurons, (2) modification of the postsynaptic sites on interneurons, (3) excitability changes of interneurons, (4) presynaptic modification of GABA release, and (5) changes in

postsynaptic sensitivity to GABA. It is also important to stress that enhancement or decrease of inhibition in LTP-LTD proto- cols is not necessarily associated with direct modification of the inhibitory circuitry (Maccaferri and McBain, 1995).

XI.1. LTP-LTD of Interneurons

In the first study on LTP of interneurons, some interneurons in both the dentate gyrus and the CAI region showed reliably increased responsivity to perforant path or comrnissural afferent tetanization (Buzsdu and Eidelberg, 1982). The low current intensities used in those studies reduced the possibility that the potentiation effect was conveyed disynapdcally by recurrent excitatory circuits. These in vivo studies suggest that with low current intensity only IPSPs will be potentiated in principal cells and that potentiation of interneurons is an efficient way for altering local network excitability. In these and related experiments (Taube and Schwartzkroin, 1987; Tomasulo and Steward, 1996), the anatomical identity of the recorded interneurons remained unknown.

Recent in vitro studies on LTP of interneurons support the notion that, similar to LTP of principal cells (cf. Bliss and Collingridge, 1993), a possible site of modification is the excitatory synapses on interneurons (Taube and Schwartzkroin, 1987; Stelzer et al., 1994; Ouardouz and Lacaille, 1995). Tetanic stimulation of stratum oriens in conjunction with postsynaptic depolarization increases postsynaptic EPSCs of interneurons in the CA1 stratum oriens. This form of LTP is prevented by bath application of NMDA and metabotropic glutamate receptor antagonists or nitric oxide synthase blockers and by inclusion of a Ca2+ chelator in the patch pipette. The same stimulation, however, failed to induce EPSC potentiation of neurons in stratum lacunosum-moleculare (Ouardouz and Lacaille, 1995). In a similar study with sharp electrodes, however, identical tetanic stimulation of afferents but without postsynaptic depolarization induced LTP in 40 of the 55 interneurons tested. The proportion of cells that underwent LTP was similar among interneurons in strata pyramidale, oriens, and stratum lacunosum-moleculare (Stelzer et al., 1994). The spontaneous firing of interneurons also increased as a result of tetanization (Stelzer et al., 1994; Poncer et al., 1995). Simultaneous recordings ofpyramidal cells and interneurons in the CAI pyramidal layer show that in several pairs interneurons exhibited a lower threshold for LTP than equidistally recorded pyramidal cells (S. Karnup and A. Stelzer, personal communication). The significantly higher success of LTP in studies using sharp electrodes raises the possibility that the application of whole-cell patch pipettes may wash out factors critical to the induction of LTP, although in some studies LTP was absent even when perforated patch recordings were used (C.J. McBain, personal communication). Other possible sources of interlaboratory variability in the success or failure of LTP induction are the stimulus parameters and the stimulated afferents. Stimulation in stratum oriens may activate a large number of afferents to the interneurons in the same region, whereas stratum radiatum stimulation activates only a fraction of the affferents to the basket and bistratified cells. A recent patch-clamp study found robust LTP in “giant” interneurons of the CAI stratum radiatum (Maccaferri and McBain, 1996). Most of the dendrites of these cells are confined to the stratum radiatum in contrast with the “horizontal” and “vertical” cells of stratum


oriens, which failed to support LTP by stratum radiatum stimulation. Obviously, many technical issues must first be clarified before we can confidently state which interneuron types, if any, display reliable synaptic plasticity.

Much less is known about the conditions that may induce LTD of synaptic excitability in hippocampal interneurons When tetanic stimulation is applied to CA1 stratum radiatum, EPSCs in interneurons recorded in the same layer are depressed. This depression is coupled to the enhancement of evoked responses in pyramidal neurons. A second, unstimulated pathway onto the same interneuron is also depressed, suggesting a postsynaptic locus of depression (McMahon and Kauer, 1995). These observations support previous findings in the CA3 region that showed tetanus-induced depression of IPSPs in simultaneously recorded pyramidal cells (Miles and Wong, 1987b). When using the low- frequency stimulation paradigm (1 Hz, 10 min) and whole-cell or perforated-patch techniques, Maccaferri and McBain (1 995) failed to induce LTD in CA1 stratum oriens interneurons, some of which were identified as 0 -LM cells. Although they regularly observed reduction of EPSPs in response to stratum radiatum stimulation, such a change most likely reflected primary changes in the pyramidal cells, and the decreased EPSPs in the interneurons simply passively reflected LTD of the principal cells. In some cases, high-frequency stimulation was also used and failed to induce LTP in 0 -LM cells. Because Schaffer collaterals do not innervate 0-LM cells (Blasco-Ibanez and Freund, 1995; Maccaferri and McBain, 1995; Sik et al., 1995), the lack of LTP in response to stratum radiatum stimulation is not surprising. However, LTD of 0 -LM cells also failed when low-frequency stimulation was applied to the stratum oriens (Maccaferri and McBain, 1995, 1996). It is noteworthy that the prolonged low-frequency activation paradigm is not effective in producing LTD of pyramidal cells in vivo (Thiels et al., 1994). A possible source of the contradiction may be the differences of the inhibitory circuits in vivo and in vitro. Another explanation may be an age difference of the sub- jects used in these studies. Recent in vitro experiments suggest that LTD is very difficult to induce with low-frequency stimulation in slices derived from adult animals, whereas it is quite robust in slices from juvenile rats (T.C. Foster and R.F. Thomson, personnal communications).

The findings that induction of LTP in interneurons follows the same cellular mechanisms as those in principal cells, including NMDA-receptor activation and intracellular calcium elevation, may have important implications for understanding its underlying mechanisms. Dendritic spines have been suggested to play a critical role in the LTP of pyramidal cells (Lisman and Harris, 1993), yet spines are rare in most interneuron types. However, postsynaptic receptors and numerous postsynaptic mechanisms similar to those of principal cells may account for LTP in at least certain subclasses of interneurons.

XI.2. Modulation of GABAergic Inhibition After Afferent Tetanization

One might expect that LTP of interneurons would reduce the overall potentiation effect on principal cells, yet a typical conse-

quence of LTP-inducing tetanization is that, when the slope of the potentiated field EPSP is returned to its pretetanization level by lowering the stimulus strength, the EPSP is then associated with a larger population spike than before. Early field potential studies suggested that this effect, termed “EPSP-spike po tentiation” (Bliss and Lomo, 1973), is a result of greater LTP of excitatory inputs onto principal cells than of inputs onto feed-forward inhibitory interneurons (Abraham et al., 1987; Tomasulo et al., 1991).

Postsynaptic depolarization is critical for the induction of LTP (Malinow and Miller, 1986). A primary mechanism of sustained depolarization during high-frequency afferent activation is the reduction of postsynaptic inhibition. Such disinhibition is believed to result from a decreased release of GABA due to activation of presynaptic GABAB autoreceptors (Davies et al., 1991; Mott and Lewis, 1991; Thompson et al., 1993). In contrast, lower frequency synaptic stimulation (5-10 Hz) is associated with a prolonged membrane hyperpolarization and a decrease in neuronal input resistance in CAI pyramidal cells (Alger et al., 1990). These latter conditions may be critical for the induction of LTD.

Several LTP studies have analyzed changes in synaptic inhibition. Postsynaptic GABA sensitivity, as assessed by measuring the reduction of the population spike in response to focally applied GABA to the CA1 pyramidal cell layer, is unchanged after induction of LTP in the Schaffer collaterals (Scharfman and Sarvey, 1985). Similarly, Haas and Rose (1982; 1984) found no change in either the slope of IPSPs or in the rate of spontaneous IPSP occurrence in CA1 pyramidal cells following development of LTP. In contrast, tetanic stimulation of the alveus resulted in enhancement of IPSPs recorded in pyramidal cells. This inhibitory LTP is abolished by NMDA blockade (Grunze et al., 1996). In the CA3 region, tetanic stimulation of the mossy fibers increased the IPSP amplitude in most cases, whereas the initially observed evoked IPSP is converted to a predominant EPSP after the tetanur in other cases (Misgeld and Nee, 1984). Facilitation of inhibitory transmission can be explained by increased excitability of interneurons. Attenuation of inhibition in other cases suggests, however, that the inhibitory synapse itself underwent modifications. When stronger or repetitive tetanization are used, IPSPs and GABA sensitivity typically decrease after tetanization (Yamamoto and Chujo, 1978; Misgeld and Nee, 1984; Miles and Wong, 1987a,b; Stelzer et al., 1987). Such depression of the inhibitory synapse may have short- and long-term components. A recent study has suggested that these inconsistencies may be explained by the differential changes of inhibition at somatic and dendritic locations (Stelzer et al., 1994). In this study, orthodromically evoked IPSPs exhibited variable changes in somata but a consistent decrease when recorded at dendritic sites associated with a large reduction of GABAA conductance. GABAA currents, evoked by iontophoretic application of a specific agonists, decreased in both dendritic and somatic recordings, even in those cells that exhibited no changes in the orthodromic IPSPs. Amplitudes of spontaneous IPSPs decreased in both somatic and dendritic recordings (Stelzer et al., 1994). Such somatic-dendritic compartmentaliza- tion of orthodromic IPSP changes during LTP may reflect distinctive subcellular modification of synaptic inhibition.


A more dramatic form of inhibitory plasticity was recently described in CA1 pyramidal cells by pairing postsynaptic depolarization with stimulation induced inhibition or local application of GABA (Collin et al., 1995). Both forms of pairing resulted in the transformation of hyperpolarization into depolarization or even discharge of the cell, lasting for up to an hour. The induction of the long-term transformation was not affected by pharmacological blockade of fast glutamatergic transmission but was blocked by bicucullin or the C1- pump-blocker furosemide. Whereas the hyperpolarizing GABA response reversed its polar- ity at - 70 mV, the depolarizing responses reversed between - 35 mV and -45 mV, suggesting that charges were carried by bicarbonate (Grover et al., 1993; Staley et al., 1995). The findings may be interpreted by the hypothesis that strong release of GABA paired with bursting dicharge of the postsynaptic pyramidal cell may result in long-term depression of the chloride pump.

Short-term changes in the inhibitory synapse may also be brought about by strong postsynaptic depolarization and rises in intracellular Ca2+ (Pitler and Alger, 1992). It has been suggested that a diffusible retrograde messenger activates a G-protein-linked pathway in the presynaptic inhibitory terminals (Pitler and Alger, 1994). Besides electical stimulation, long-term depression of IPSPs (> 1 h) have been described in CA1 pyramidal cells after extended application of mGluR agonists (Liu et al., 1993). In this case, the effect was attributed to postsynaptic changes because it was abolished by perfusing the postsynaptic cell with a GTP ana- logue. However, kindling studies in awake animals have suggested that strong and repetitive activation of the hippocampal network can also lead to long-term potentiation of inhibitory synapses. The amplitude but not the frequency of miniature IPSCs in granule cells increased in rats that underwent a kindling procedure (Otis et al., 1994). These observations indicate that use-dependent activation may lead to lasting enhancement of IPSCs.

This summary of the roles of interneurons in synaptic plasticity indicates that the enhanced interneuronal output during LTP may be associated with an altered GABAA receptor function. It remains to be seen whether these “average” effects are due to specific changes at the inputs and outputs of interneurons or they may be explained by a combination of various factors, including specific changes in some interneuronal types but not in others, alterations in the interneuron-interneuron relationships, and long-term changes in cell coupling through gap junctions.

XI.3. Possible Contribution of Interneurons to Synaptic Plasticity In Vivo

As will be discussed in Section XII.1, the majority of hippocampal interneurons increase their firing rates during theta-associated explorative behavior. Interestingly, this reliable correlation may be disrupted when the rat enters a new environment. Wilson and McNaughton (1993) observed that the majority of simultaneously recorded interneurons in the CA1 pyramidal layer became transiently silent when the rat was allowed to visit a new compartment of the testing apparatus. They suggested that suppression of interneuronal activity might facilitate the synaptic modification necessary to encode new spatial information.

Importantly, the activity of all recorded pyramidal cells were also suppressed on entering the new environment; thus, it could not be excluded that the decreased interneuronal activity was a con- sequence of their reduced recurrent activation.

During hippocampal sharp waves (Section XI11.4), population bursts of principal cells coincide with the most rapid firing of the majority of interneurons. Some interneurons in stratum radiatum may transiently discharge up to 800 Hz (Buzsiki et al., 1983), whereas others may not be affected or even decrease their activity (Fig. 36). What might be the advantage of such intense collective activity in both principal cells and interneurons? As discussed above, potentiation of interneurons may be associated with diminished inhibition of principal cells in LTP paradigms, and the large amounts of GABA may lead to depolarization in target principal cells. O n the basis of in vitro evidence, we may speculate that intensely discharging interneurons during hippocampal or entorhinal sharp waves (SPWs) may in fact induce a hyperpolarizing-depolarizing sequence in pyramidal cells. Hyperpolariza- tion followed by depolarization is the appropriate sequence for de-inactivation of low-threshold calcium channels in pyramidal cell dendrites (Magee and Johnston, 1995b). Therefore, intense recurrent excitation via AMPA receptors and GABA-mediated depolarization during SPW bursts may potentiate the activation of NMDA channels and lead to the cooperation of voltage-dependent and NMDA-dependent calcium influx into the dendrites. Testing the reality of this scenario will require anatomical identification of fast-firing interneuron types, voltage-clamp moni- toring of pyramidal cell dendrites during hippocampal SPWs, and paired recordings from the interneuron type(s) in question and pyramidal cells. From this perspective, activation of interneurons with dendritic targets could either prevent or facilitate synaptic plasticity, depending on their cooperative synchrony and firing frequency (Lambert and Grover 1995; Miles et al., 1996).

Different population patterns, as reflected by spontaneous field potentials and rhythms, are present in the hippocampal formation, including theta activity and associated gamma patterns (40-100 Hz oscillation), hippocampal SPWs, and associated high-frequency (200 Hz) oscillation (“ripple”) and dentate spikes (cf. O’Keefe and Nadel, 1978; Buzsiki et al., 1983, 1994; Bland, 1990; Bragin et al., 1995b). Interneurons appear critically involved in the induction and maintenance of network oscillations in the theta, gamma (40-100 Hz), and ultrafast (200 Hz) frequency ranges and regulate the recruitment of principal cells during SPW bursts (Buzsiki et al., 1983, 1992; Fraser and MacVicar, 1991; Soltksz and Deschenes, 1993; Bragin et al., 1995a; Minen et al., 1995a,b).

XII.l. Theta Rhythm

It has been hypothesized that part of the extracellular field associated with theta activity is due to perisomatic hyperpolarizing potentials in pyramidal cells that are mediated by interneurons


(Buzsriki and Eidelberg, 1983; Buzsiki et al., 1983). Phasing of hippocampal interneurons at the theta frequency may derive from (1) rhythmic excitation by the septal cholinergic input, (2) rhythmic excitation coupled with rhythmic inhibition of the same interneurons by the septal GABAergic input, (3) tonic cholinergic and/or glutamatergic activation coupled with rhythmic modula-

tion by the septal GABAergic input, (4) hippocampal interneu. rons with intrinsic slow oscillation, or ( 5 ) depolarization-induced slow oscillation of pyramidal cells (Petsche et al., 1962; Andersen and Eccles, 1962; Buzsiki et al., 1983; Konopacki et al., 1987; Fox, 1989; MacVicar and Tse, 1989; Stewart and Fox. 1990; Fraser and MacVicar, 1991 ; Leung and Yim, 1991; Soltksz and

B

D extracellular A

CA1 pyramidal cell I

FIGURE 39. Rhythmic bursts of basket cells hyperpolarize pyramidal cells during theta. A: Simultaneous recording of extracellular theta activity in the CA1 pyramidal layer (extracellular) and intracellular activity of a basket interneuron. Action potentials are clipped. An epoch with poor extracellular theta was chosen to ac- centuate the very regular intracellular rhythm. Note rhythmic 30-60- Hz bursts of action potentials with every theta cycle. B: Camera lucida drawing of the dendritic tree and axon collaterals of the basket cell in the pyramidal layer. C: Evoked field (upper) and intracellu-

lar responses to commissural (c) stimulation. The basket cell fired earlier than the peak of the population spike (arrow) and emitted three fast spikes. D: Simultaneous recording of extracellular elec- troencephagraphic activity in the CA1 pyramidal layer (extracellular) and intracellular activity of an identified pyramidal cell. Note hyperpolarization of the pyramidal cell membrane at the onset of spontaneous theta activity (arrow). Dotted line, -60-mV “resting” potential. From Ylinen et al. (1995b).


a, 10- n E 3

c .

0

L O i

, I.l

C

0 90 180 270 360 t h e t a p h a s e ( d e g r e e s )

FIGURE 40. Relationship between theta phase and interneuronal activity in the behaving rat. A: Sinusoid wave represents a single theta wave. B: Distribution of maximal-firing phases of interneurons in the CAI region. Most cells were recorded from stratum oriens (data adopted from Buzs& et al., 1983). C: Distribution of maximal-firing phases of interneurons in the CAI region. Most cells were recorded from the pyramidal layer (data adopted from Skaggs et al., 1996). Note bimodal distribution of phase relationship in both studies. Although individual interneurons are strongly modulated and their phase-locking to theta waves is quite constant, the phase relationship may vary among the interneuron groups.

Deschenes, 1993; Lee et a!., 1994). In the intact rat, the combination of the “pacemaker” inputs from the septum (mechanisms 1-3) and the resonant intrinsic and network properties of hippocampal interneurons and principal cells (mechanisms 4 and 5 ) contributes to the prominent entrainment of hippocampal neurons during “theta behaviors” (Buzsiki et al., 1983) and to the production of field theta activity.

Although nearly all interneurons express rhythmic discharges during theta, their relationship to the phase of the theta cycle shows a large variability (Buzsiki et al., 1983; Fox and Ranck, 1986). Identified basket cells in the CAI region discharge on the phase opposite to the maximal activity of the pyramidal cells in the urethane-anesthetized animal (Fig. 39). The basket cells are rhythmically hyperpolarized by either the septa1 GABAergic input (Freund and Antal, 1988) or by other hippocampal interneurons winen et al., 1995b). In turn, they hyperpolarize their target pyramidal cells at the theta frequency (Fig. 39). In the awake rat, interneurons in the CAI oriens and pyramidal layer show a bimodal distribution with respect to the phase of theta (Fig. 40),

and it is logical to assume that these subgroups represent morphologically and functionally separate sets of interneurons, although their anatomical identity has yet to be revealed. The discrepancy of the phase relationship of interneurons to the theta cycle in the anesthetized and drug-free animals (Fox and Ranck, 1986) may be explained by the suggestion that recurrent excitation of interneurons is very sensitive to blockade of NMDA receptors (Grunze et al., 1996). Therefore anesthetics, such as ketamine and urethane, may attenuate principal cell-interneuron feedback.

Experiments on the regulation of theta activity have indicated a complex interaction between interneurons and target principal cells. Rhythmic activation of identified basket or chandelier cells at the theta frequency in vitro instantly phase lock simultaneously recorded postsynaptic CA1 pyramidal cells (Cobb et al., 1995). During the entrainment, the interneurons and pyramidal cells fire on alternate phases, as in the urethane-anesthetized rat (Buzsiki and Eidelberg, 1983; Ylinen et al., 1995b). Subthreshold membrane oscillations, induced by depolarization of the pyramidal cell, are also entrained. Furthermore, pacing of hippocampal interneurons by rhythmic stimulation of the septohippocampal GABAergic pathway in a combined septum-hippocampus slice preparation also induced theta oscillations in pyramidal cells (T6th et al., submitted). These observations indicate that regulation of intrinsic conductances of their targets is as fundamental a function in the repertoire of interneurons a s their membrane polarizing and shunting effects (Cobb et al., 1995). “Resonant” properties of some interneuron types may further enhance the hippocampal network to respond to rhythmic subcortical inputs. 0 -LM cells have been shown to fire very regularly at theta frequency even in the absence of fast synaptic transmission (McBain, 1994). A possible candidate for such intrinsic, pacemaker conductances is the hyperpolarization-activated current, IH (Maccaferri and McBain, in press).

Rhythmic cooperative discharge patterns of interneurons, ev- idenced by field theta waves, show a predictive relationship with behavior. They also display some limited correlation with the animal’s spatial position, suggesting that they receive a dispropor- tionally stronger excitation from “spatially” firing pyramidal cells (O’Keefe and Nadel, 1978) than from other inputs (Barnes et al. 1991). These observations offer the possibility that local interactions between interneurons and principal cells may allow substantial deviation from average population behavior. From this perspective, interneurons are not only responsible for a global control of the principal cells but may also be involved in finely tuned local regulation of their principal cell targets.

XII.2. ”Antitheta” Interneurons

A very small portion (<I%) of interneurons have diametri- cally opposite behavioral correlates as opposed to typical theta- modulated interneurons. These cells, termed “antitheta” or “theta-off’’ cells (Buzsiki et al., 1983; Colorn and Bland, 1987; Mizumori et al., 1990) substantially decrease their firing rate during exploration-associated theta activity. In some of these studies, antitheta cells could clearly be distinguished from pyramidal cells, granule cells, and mossy cells based on their location (stra-


tum radiatum, hilus) and physiological correlates. In the absence of theta waves, antitheta neurons fire very regularly in a "clock- like" manner, typically 15-40 Hz (Fig. 41). Their firing rate becomes especially regular after inactivation of the medial septum, suggesting that subcortical inputs tonically suppress these neu- tons. Alternatively, antitheta cells may be inhibited by other hippocampal interneurons. At least some antitheta cells exhibited spatial selectivity (Mizumori et al., 1990). Although small in number, given their peculiar relationship to network activity, expected distinguishing intrinsic properties, and differential subcortical and hippocampal interneuronal innervation, it would be especially re- warding to reveal the anatomical identity of antitheta interneurons.

XII.3. Gamma Oscillation A second prominent population rhythm of the hippocampal

formation is gamma oscillation (Stumpf, 1965; Buzsiki et al., 1983; Leung, 1992). The gamma rhythm (40-100 Hz in the awake rat) is intrinsic to the hippocampal formation but modulated by the slower theta waves (Fig. 42; Bragin et al., 1995a). A major significance of gamma rhythm is that similar oscillations are simultaneously present in other forebrain areas during behavioral activation (cf. Gray, 1994). Coupling of neocortical and hippocampal oscillations is a candidate mechanism for binding neuronal representations associated with currently perceived and retrieved information (Bragin et al., 1995a; Buzsiki and Chrobak, 1995). Ample evidence supports the importance of interneurons in the generation of gamma oscillation. (1) Extracellular current sources are localized to the cell body layers (Buzsaki et al., 1983;

w a l k

Bragin et al., 1995a). (2) Gamma waves are associated with phase- locked membrane oscillations in pyramidal cells when chloride- filled electrodes are used (SoltPsz and Deschenes, 1993). (3) Putative interneurons fire phase locked to the gamma rhythm in the awake rat (Buzsiki et al., 1983; Bragin et al., 1995a). (4) Identified CA1 basket cells discharge phase locked to both gamma and theta frequencies (Fig. 39; Ylinen et al., 1995b). (5) Several types of hilar interneurons sustain membrane oscillations phase locked to the field gamma waves (Sik et al., submitted). These observations support the hypothesis that gamma oscillations are generated by chloride-dependent GABAA receptor-mediated IPSPs. In contrast with the slow theta rhythm, gamma oscillation is generated by the intrinsic hippocampal network because it sur- vives damage to the septum or removal of all subcortical inputs (Buzsiki et al., 1983, 1987). A possible reason for this is that the gamma pattern can emerge in interneuronal networks isolated from the principal cells (Section XIV. I). However, GABAergic cells in the basal forebrain possess an especially high propensity of voltage-dependent oscillation (Alonso et al., in press), and thesc neurons are reciprocally connected to hippocampal interneurons (T6th et al., 1993). It is therefore expected that in the in vivo brain these GABAergic neuronal networks play a critical role in the maintenance of a clock function in the coordination of principal cell activity (Section XIII.2). In future work, it will be necessary to reveal what interneuron classes possess the intrinsic mechanisms necessary for voltage-dependent generation of gamma oscillations as opposed to those interneurons that are brought into the oscillatory population activity by rhythmic IPSPs and EPSPs.

s t i l l

. . . . . . . I .

FIGURE 41. Interactions among interneurons during theta at- tivity. Simultaneous recordings of three interneurons (1-3) in the hilar region. Interneurons 1 and 2 fired rhythmically during waking- related theta activity, whereas interneuron 3 (antitheta cell) was completely suppressed. Cessation of discharge of interneuron 1 (open arrow) coincided with the discharge of the antitheta cell. Interneuron 2 continued its periodic discharge beyond interneuron 1 (dots be-

low trace 2). When interneurons 1 and 2 stopped discharging (filled arrow), the antitheta cell fired regularly at 18 Hz. Such a reciprocal relationship may be brought about by common inputs to the two cell types acting on different receptors or theta-related interneurons may be innervating antitheta cells. Antitheta cells constitute fewer than 1% of all interneurons (Buzsiiki et ai., 1983; Colom and Bland, 1987; Mizumori et al., 1990).


FIGURE 42. Theta modulation of gamma activity. a: Gamma activity in the hilar region during exploration. Upper trace: Wide band recording. Middle and lower traces: Gamma activity and unit firing (500 Hz-5 kHz), respectively. Note both theta- and gamma- related modulation of the isolated neuron. b: Gamma-triggered average of the wide band activity. c: Cross correlogram of unit firing and peaks of the gamma waves. Note the relationship between av-

eraged field (b) and unit discharges (c). d: Coherence of gamma activity in the longitudinal axis of the hippocampus. Electrodes 1-7 are indicated by black dots. e: Superimposed traces of simultaneously recorded signals from the hilus in the longitudinal axis (electrodes 1-7 in d). Note near-zero phase shift of both theta and gamma in the longitudinal axis.

XII.4. Sharp Waves and Fast Oscillations (200-Hz Ripples)

Intermittent SPWs of 40-120-ms duration are observed during consummatory behaviors, behavioral immobility, and slow wave sleep (Buzsiki et al., 1983; Buzsiki, 1986; 1989; 1996 Suzuki and Smith, 1987). Released from inhibition associated with theta, the highly interconnected CA3 network exhibits population bursts. These bursts produce synchronized depolarization of CA1 pyramidal cells and interneurons and an associated large field response (an SPW). The transient depolarization associated with the CA3 burst sets into motion a short-lived interaction between interneurons and pyramidal cells, the product of which is a fast (200 Hz) field oscillation (ripple) and a phase-related discharge of CA1 pyramidal cells (O’Keefe and Nadel, 1978; Buzsiki et al., 1992). Putative basket cells in the awake rat and anatomically identified basket cells in the anesthetized animal fire at ripple frequency and phase locked to the negative peak of the local ripple waves, i.e., at the same time when pyramidal cells discharge (Fig. 43). Action potentials of other interneurons in stratum oriens are also entrained to the fast oscillations, but they may not necessarily sustain high-frequency discharge at ripple frequency and typically emit action potentials 1-2 ms after the maximal firing of pyramidal cells (Ylinen et al., 1995a). These observations

indicate that the different subgroups of interneurons are recruited to the population oscillation by different mechanisms. Because stratum oriens interneurons are not innervated by basket cells, chandelier cells, or CA3 pyramidal cells, the delay may reflect their recruitment into the population oscillation by the CAI pyramidal cells. Interneuronal communication through gap junctions has been suggested to be critical for providing coherent synchrony of ripples throughout the CAI region and subiculum (Ylinen et al., 1995b; Traub 1995; Chrobak and Buzsiki, 1996).

The majority, but not all, interneurons are activated during SPW-associated population bursts. A subgroup of cells in CA1 stratum radiatum does not change its discharge rate during the population bursts, although the group is phase modulated by the theta rhythm (Fig. 36). Obviously, these interneurons are not innervated by the axon collaterals of either CA1 or CA3 pyramidal cells. Yet another type of cell in CA1 stratum radiatum specifically decreases its discharge frequency during SPW bursts (Urioste et al., 1992). It is tempting to speculate that suppression of activity in this cell type is brought about by the discharging bistratified, trilaminar, and back-projection neurons and VIP- or CR-immunoreactive interneurons (Acsidy et al., 1996a,b; Gulyh et al., 1996) because axon collaterals of these cells course in stratum radiatum and terminate on other interneurons. Another unexplored mechanism for the participation of the interneuronal


a extra (50 Hz-5 kHz)

b extra (50-30L Hz)

intra i ' 1 ' 20 ms -

FIGURE 43. Coherent discharge of interneurons time the occurrence of pyramidal cells action potentials. a: Single trace of wide band recording from the C A I pyramidal layer. Putative basket cell recorded in the awake rat. Note rhythmic discharge of the interneuron phase locked to the field oscillation. b: Intracellular recording from a CAl basket cell during ripple activity. Note phase- locked discharge of the interneuron to the extracellular field. Arrows indicate spike failures and rhythmic membrane oscillation (urethane anesthesia). c: Relationship between high-frequency field oscillation and neuronal spike activity. Simultaneous recording from 40 pyra-

types in SPW bursts is the hippocampal interneuron-medial septum CABAergic loop (Sections IV.2.b, V). Clearly, the precise relationship between population rhythms and interneuronal cell types awaits further experimentation.

The first significant interaction between systems neuroscience and cellular neuroscience has occurred in the field of neuronal oscillations. The recognition that neurons are not simply "integratc- and-fire" devices but are endowed with different intrinsic ligand- and voltage-dependent mechanisms (cf. Llinas, 1988) has begun to exert an impact on hypotheses of network activity. The "pace-

C I

20 ms PYR d

50 ,

0 20 ms 50 -1

30 401 0 20 ms

midal cells and 2 putative interneurons (i at electrodes 4,5). Averaged field event (bottom) and superimposed spike events in the CAI pyramidal layer. d: Cross correlogram between field oscillation and summed pyramidal cell activity (PYR) and interneuronal firing (INT). Continuous line indicates Gabor function curves. Modified from Ylinen A, Sik A, Bragin A, NBdasdy Z, Jand6 G, Szab6 I, Buzsdki G, (1995) "Sharp wave-associated high-frequency oscillation (200 Hz) in the intact hippocampus: Network and intracellular mechanisms." J Neurosci 1530-46 by permission of Journal of Neuroscience.

maker-follower'' schemes are gradually being replaced by models based on interactions between the emerging network properties and resonant behavior of the participating neurons. At the very least, this new effort provides an alternative approach to such very complex issues, such as perception, cognition, movement initiation, and memory (Murthy and Fetz, 1992; Buzsiki et al., 1994; Gray, 1994; Llinas et al., 1994; Singer, 1994). Although the neuronal mechanisms of communication between neocortical and hippocampal representations could not be realistically addressed by any previous physiological theory, coupling via network oscillations is a novel and unexplored possibility.

A hypothesis that was advanced in connection with these issues suggests that networks of inhibitory interneurons impose a coordinated oscillatory "context" for the "content" carried by networks of principal cells. The hypothesis implies that GABAergic neuronal "super networks" may cooperatively entrain large populations of

INTERNEURONS OF THE HZPPOCAMPUS 445

pyramidal cells throughout the forebrain. These oscillating inhibitory nerworks may therefore provide the precise temporal structure necessary for ensembles of neurons to perform specific hnc- tions (Buzsik~ and Chrobak, 1995). In such combined systems, spatially distal principal cells can discharge either irregularly or rhythmically, yet temporal coincidences of their action potentials may be preserved even when they are not interconnected by excitatory collaterals. From this perspective, macroscopic oscillations per se do not carry any specific information. Nevertheless, interneuron-imposed network oscillations are critical for the emergence of spatiotempo- ral synchrony of the principal cell population and thereby essential for the coordination and integration of functions in anatomically distributed structures. In testing this hypothesis, the first step is to examine the rules of the emergence of network oscillation and the neuronal mechanisms that support them.

XIII.l. Oscillations in Interneuronal Nets

The two major issues in neuronal oscillations are the cellular origin of rhythmicity and the mechanisms of large-scale population coherence. Recent modeling studies have indicated that population oscillations beyond the theta frequency range likely cannot be achieved by the fast AMPA-type connections between pyramidal neurons (van Vreeswijk et al., 1995; Wang and Buzsiki, in press). However, isolated interneuronal networks maintain oscillation in the gamma band (Wang and Buzsiki, in press; Whittington et al., 1995; Traub et al., 1996). These combined in vitro, in vivo, and modeling studies indicate that gamma oscillation arises in the hippocampus when interneurons receive a tonic or slowly varying excitation and provide some insights into the rules of oscillations in this frequency range.

Transient gamma field oscillation and fast, rhythmic IPSPs in simultaneously recorded pyramidal cells are observed in the hippocampal slice preparation during blockade of ionotropic glutamate receptors. The oscillation can be prolonged by metabotropic glutamate receptor agonists and blockade of GABAB receptors. Therefore, the critical element for the induction of gamma oscillation appears to be a mutual GABAA receptor-mediated inhibition among interneurons (Whittington et al., 1995). The rich interconnectivity among hippocampal interneuronal classes provides the anatomical support for the observed physiological observations (Misgeld and Frotcher, 1986; Lacaille et al., 1987; Khazipov et al., 1995a; Sik et al., 1995; Acsidy et al., 1996a,b; Gulyk et al., 1996). It has been hypothesized that IPSPs can synchronize neuronal discharges, provided that the synaptic kinetics are sufficiently slow as compared with the oscillation frequency of synaptically uncoupled neurons F a n g and Rinzel, 1993). Subsequent pharmacological and computational experiments have further defined the necessary and sufficient conditions for gamma oscillation (Wang and Buzsiki, in press; Traub et al., 1996), and the results can be summarized as follows. Large-scale network synchronization requires a critical (minimal) network connectedness and sufficiently dense synaptic inter- connections among interneurons. Synchron-ization by GABAA synapses requires (1) that the synaptic current does not decay too rapidly (>5 ms) and (2) that the synaptic potential of the participating interneurons is more negative than the spike afierhyperpo-

larization, so that the synaptic action always produces hyperpolarizations. The frequency of gamma oscillation can be regulated by several factors, including the synaptic decay time constant, the efi- cacy of GABAA synapses (i.e., the amplitude of IPSCs), and the driving currents to the interneurons. In contrast with the monotoni- cally varying frequencies of single interneuron firing rates (0-400 Hz) brought about by these variables, the degree of network synchronization shows a relatively narrow peak in the gamma frequency range (20-70 Hz). In the slice experiment, thiopental was used to prolong the GABAA time constant, the GABAA conductance was varied by bicuculline and diazepam, and the driving current was increased by increasing doses of L-glutamate. In general, there was a good fit between the experiments and the model.

In vivo, field gamma “tails,” associated with phase-locked interneuron discharges, are observed in the absence of principal cell firing between epileptic spikes (Traub et al., 1996). The gamma frequency of the interneuronal network can be “modulated” by slow driving currents, enabling the interneurons to express gamma and theta rhythms at the same time, as occurs in the intact animal (Buzsiki et al., 1983; Bragin et al., 1995a; Ylinen et al., 1995b). Although modeling studies assumed, based on preliminary physiological observations (Whittington et al., 1995; Ylinen et al., 1995b), that basket cells are involved in gamma oscillations, the contribution and importance of other interneuron types, particulary those involved in selective innervation of other interneurons (Section I11.4), have yet to be examined. In addition, in the modeling studies the connectivity among the interneuronal pool is statistically homogeneous and random, whereas in the hippocampus the different interneuronal types are rarely interconnected symmetrically. For example, basket cells and chandelier cells may not innervate other interneuron types, but they themselves are heavily interconnected and are innervated by other interneuron types (Fig. 37). The ubiquitous nature of gamma oscillation in numerous central structures (cf. Gray, 1994) may be explained by the assumption that physiological properties of interneurons and their connectivity and effects on target cells have been preserved during phylogeny.

XIII.2. Physiological Role of Interneuron- Generated Oscillations: Keeping Time

In general, principal cells in the cortex discharge relatively slowly and irregularly, and it has been extensively debated whether the information they carry is embedded in the precise timing of the action potentials or in the firing rate of neurons (cf. Bialek et al., 1991; Softky and Koch, 1993; Shadlen and Newsome, 1994; Hopfield, 1995). Networks of interneurons impose a coordinated subthreshold oscillation on rheir principal cell targets. This is an efficient way ( I ) for keeping principal cells in a “readiness” state by allowing them to respond to relatively weak excitatory inputs and (2) for making principal cells capable of implementing codes that use precise spike times. Thus, a postulated function of interneuronal nets is to bring sparsely firing, spatially distant principal cells into functionally cooperative ensembles (Buzsiki and Chrobak, 1995). The extent of interneuronal nets go beyond the borders of the hippocampal formation. As discussed earlier,


GABAergic neurons in the basal forbrain are reciprocally connected to hippocampal interneurons (Sections V. 1, VI.2). Moreover, these basal forebrain GABAergic neurons also innervate inhibitory neurons of the neocortex (Freund and GulyPs, 1991; Freund and Meskenaite, 1992) and the GABAergic neurons in the reticular nucleus of the thalamus (Asanuma and Porter, 1992). In such widespread interneuron-controlled networks, timing of principal cell action potentials within and outside the hippocampal formation may be determined by a combination of the oscillatory phase (i.e., the level of inhibition) and the strength of the excitatory drive. Stronger excitation produces a larger phase advance of the occurrence of the action potentials. In other words, the information about the integration of excitatory inputs to a given cell is expressed by the timing of the action potentials with respect to average population oscillation (Fig. 44). A formal computation model building on similar ideas has recently been published (Hopfield, 1995).

Phase precession of principal cell discharge relative to the population theta rhythm has been observed experimentally (O’Keefe and Recce, 1993; Skaggs et al., 1996). The spikes of granule cells and CA1 pyramidal cells advance to earlier phases of theta as the rat passes through the cell’s place field (i.e., region of spatially restricted firing), and the magnitude of phase advance is a function of the animal’s position within the place field. Simultaneously

recorded pyramidal neurons with overlapping spatial fields display predictable relative spike sequences. These experimental findings indicate that the momentary position of the animal can be precisely determined by respective occurrence of principal cell action potentials relative to the phase of rhythmic hyperpolarizations provided by the interneuron network (Skaggs et al., 1996). It remains to be seen whether phase precession of principal cells occurs in a relatively continuous or “quantal” manner, as would be expected by the superimposed gamma oscillation on theta waves.

Empirical data on phase coding of neuronal information is rather limited in the mammalian nervous system. However, n u - merous computational models use timing of action potentials with respect to an ongoing collective oscillator as a code for representing analog information (Richmond and Optican, 1990; Bialek et al., 1991; Burgess et al., 1994; Hopfield, 1995; Lisman and Idiart, 1995; Tsodyks et al., 1996). To date, no data are available to allow speculation on whether a phase precession of action potentials similar to that observed in the theta range also occurs within gamma waves or within the individual wavelets of SPW- associated 200-Hz ripples. Demonstration of phase advance at such short time scales would reflect that the timing of events in the mammalian nervous system is as precise as demonstrated in the brains of lower animals (Heiligenberg, 1991). Such new information would also support the hypothesis that oscillatory col-

I I

I

I

threshold _ _ . . . . . . . . . . . . . . . . . . . . . . .

\

\

interneuron

V principal cell

I

FIGURE 44. Oscillation of inhibitory networks provides a clock signal for timing of action potentials of principal cells. The hypothesized “distributed clock” leads to a zero-phase lag of inhibitory sub- roups (gray ticks indicate action potentials of interneurons). As a result, the membrane potential of spatially distant principal cells are rhythmically hyperpolarized in a highly coherent fashion (sinusoid wave indicates rhythmic hyperpolarization of principal cell membrane potential). Tonic or ramplike excitatory inputs may therefore produce rhythmic discharges of principal cells (black ticks correspond to discharging principal cells highlighted in black in the embedded network); the stronger the excitation, the earlier the phase advance in discharge (top four cells). Phasic inputs discharge the principal cells irregularly, but, again, timing is determined by a combination of the oscillatory phase (i.e., the level of inhibition) and the strength of the

excitatory drive (bottom four cells). In this scenario, coherent membrane oscillations, mediated by the interconnected inhibitory network, is a necessary but not sufficient condition for representing information. Information is assumed to be embedded in the temporal relationship of action potentials of principal cells during a given cycle. Oscillations of two or more different frequencies (e.g., theta and gamma) can co-occur and may represent different time scales for different representations. The frequency of the distributed clock is determined by the excitatory (local and subcortical) inputs impinging on the interneurons, the inhibitory inputs from other interneurons, and the intrinsic oscillatory properties of interneurons. Reproduced Buzsiki G, Chrobak JJ (1995) Temporal structure in spatially organized neuronal ensembles: a role for interneuronal networks. Curr Opin Neurobiol5:504-510 by permission of Current Biology Limited.


lective patterns representing theta, gamma, and 200-Hz rhythms reflect “processing time windows” for different categories of neuronal representations. Revelation of the types and drive of interneurons, predominantly involved in each of these oscillations, may provide insights into these challenging issues.

XIII.3. Physiological Role of Interneuron- Generated Oscillations: Homeostasis and Plasticity

As discussed earlier, repetitive activation of interneurons by stimulation very quickly leads to desensitization of GABA synapses. Nevertheless, coordinated and maintained activity of interneurons during naturally occurring oscillatory rhythms do not indicate any evidence of GABA desensitization. What is the source of such a large discrepancy between the experiments and activity in real brains? The simplest explanation is that GABA is always released by the same activated terminals in the experiment, whereas rhythmic hyperpolarization of neurons is provided by an ever-changing set of presynaptic interneurons in natural oscillations. An argument against this hypothesis is that during theta activity, for example, the number of action potentials per theta wave in basket cells does not fluctuate across a large range. Therefore, transmitter action might be maintained without significant sensitization or desensitization in oscillatory networks. Such a homeostatic mechanism may be brought about by a cyclic sequence of voltage-dependent activation of ionic conductances. From this perspective, the postsynaptic action of transmitter may be quite different even at the same membrane potential, depending on whether that potential was preceded by depolarization or hyperpolarization.

However, perturbation of such homeostatic cycles may facilitate synaptic plasticity. The prevailing view is that LTP and LTD are induced by stimulation patterns at different frequencies (cf. Teyler et al., 1994; Singer, 1995). LTD and depotentiation is produced by 1-15 min of low (1-5 Hz) stimulation (Dudek and Bear, 1992; Mulkey and Malenka, 1992). However, it is hard to imagine such maintained behavioral-network patterns for such extended periods in the awake animal. However, recent work has indicated that an increase or decrease in synaptic weights can be achieved in oscillatory systems and phase locking of the stimulus to the population pattern rather than to its frequency, or intensity will determine whether LTP or LTD is induced (Stanton and Sejnowski, 1989; Huerta and Lisman, 1995). These observations support the view that cyclic sequences of voltage- and ligand-dependent conductances underlie the rhythmic membrane fluctua- tions in individual neurons (Llinas, 1988). In turn, they suggest that the sequential changes of conductances, maintained and regulated by interneuron networks, are critical for both neuronal transmission and network plasticity.

XIII.4. Physiological Roles of Interneurons: Separation of Fast Spike Transmission From Plasticity

Different inhibitory cell types target precisely specified areas of pyramidal cell membrane (Han et al., 1993; Gulyis ec al., 1993b;

Buhl et al., 1994a; Sik et al. 1995). Perisomatic connections typically involve multiple, closely spaced contacts, whereas a dendritic axon generally contacts multiple dendritic branches of a target pyramidal cell. Membrane channels controlling pyramidal cell excitability are also expressed in a segregated fashion. Sodium channels have a high density in axon initial segment and soma, whereas calcium channel expression is primarily dendritic (Hamill et al. 1991). Thus, there is a striking cosegregation of specialized interneuron groups and the generations sices of the two kinds of action potentials in pyramidal cells. A hypothesis has been submitted that dendritic inhibitory cells control calcium-dependent electrogenesis, whereas somatic inhibitory cells regulate generation of sodium-spikes (Traub et al., 1994; Miles et al., 1996).

Paired recordings from presynaptic basket cells and postsynaptic CA3 pyramidal cells reveal that the recurrent inhibitory circuitry is typically faster than the peak of the depolarizing afterpotential following a spike. The result of this well-tuned timing mechanism is that afferent excitation typically evokes only a single spike in the pyramidal cell because the fast somatic IPSP (4-10 ms after the spike), brought about by the recurrent interneurons, suppresses further firing (Fox and Ranck, 1981; Buzsiki and Eidelberg, 1982; Miles 1991; Miles et al., 1996). Thus, a main objective of perisomatic inhibition is timing and suppression of sodium spikes generated in the axon initial segment. What then is the function in spike genesis of interneurons with dendritic targets? It is generally assumed that dendritic shunting can locally “filter” or “divide” the efficacy of local excitatory inputs. Such a function may be brought about by cooperative action of several interneurons because, in contrast to basket cells and chandelier cells, interneurons with dendritic targets typically do not innervate the same dendrites with multiple synapses. In addition, such cooperative action may also affect calcium spike genesis (Miles et al., 1996) because this is where calcium spikes are initiated (Wong et al., 1979; Spruston et al., 1995).

Intradendritic recordings from CA1 pyramidal cells in vivo reveal that calcium spikes can be delayed, prevented, or aborted by commissural activation of interneurons, without affecting sodium spike genesis (Fig. 45; Buzsiki et al., 1996). The studies in vitro have used minimal stimulation of inhibitory fibers in the presence of 6-cyano-7-nitroquinoxaline-2,3-dione and 2-amino-5- phosphonovaleric acid to block excitatory transmission. Surgical cuts in the slice assured that perisomatic inhibition was not activated by dendritic layer stimulation. Single pulses delivered to somatic layers delay or suppress repetitive discharge of fast sodium spikes. In contrast, single stimuli in dendritic zones (s. radiatum) suppress the discharge of calcium-dependent spikes initiated by current injection in dendritic recordings (Fig. 46; Miles et al., 1996). In addition, stimulation in the distal dendritic layers reduce the spike-associated [CaZf]i changes in the distal dendrites but have little effect on the changes in the cell body (Tsubokawa and Ross, in press). A possible explanation for these observations is that activity of certain incerneurons that target principal cell dendrites can selectively interfere with dendritic calcium spike genesis while leaving sodium spike generation-propagation unaffected (Traub et al., 1994; Miles et al., 1996). An alternative explanation is based on demonstrations that the dendritic mem-


brane actively supports propagation of action potentials in dendritic compartments, and back propagation of sodiums spike bursts are associated with a large voltage-dependent calcium influx (Jaffe et al., 1992; Stuart and Sakmann, 1994; Magee and Johnston, 19951.3; Spruston et al., 1995). The success or failure of dendritic invasion of a spike series depends on the level of membrane polarization, which is preset by the inhibitory interneurons. It follows that activation of interneurons with dendritic terminals can also prevent the active regeneration of the back-propagating sodium spikes. In the absence of the regenera- tive process, the passively propagating sodium spikes fail to trig- ger dendritic calcium spikes (Bus& et al., 1996).

Dendritic inhibition could also selectively regulate the efficacy of NMDA receptors and prevent plasticity simply due to the dendritic localization of NMDA receptors. However, plastic changes could be largely independent of the regulation of neuronal excitability at the somatic level, as has been demonstrated in the pyriform cortex (Kanter et al., 1996).

An important implication of these experiments is that by selec-

tive activation of certain interneuron classes plastic changes in a given circuitry may be selectively and temporarily removed, yet leaving neuronal transmission by fast sodium spikes intact. In other words, the same network can function in a “standard” transmission mode, or use-dependent activity may bring about lasting changes of synaptic fiinction, depending on the cooperative activity of interneuron classes with dendritic targets. In this context, it is important to re- iterate the differential subcortical innervation of interneuron types with dendritic and perisomatic targets (Section V). Such differential connectivity suggests that the different subcortical neurotransmitters may separately affect spike transmission and synaptic plasticity by way of the intercalated interneurons on which they terminate. In line with this suggestion, extracellular recording studies have demonstrated that associative and heterosyntaptic LTP or LTD between fiber systems terminating on nonoverlapping dendritic segments can be induced only after blockade of GAB&.-mediated inhibition, and GABAergic neurons with dendritic targets have been suggested as responsible for this regulation (Tomasulo et al., 1993; Zhang and Levy, 1993). Future experiments will reveal

Aa

7 b

50 mV

100 ms -

i FIGURE 45. Inhibition-induced changes in the active dendritic propagation of Na+ spikes. Dendritic recording from a CAI pyramidal cell in vivo. Aa: Depolarization of the dendrite by current injection evoked fast spikes and a slow (calcium) spike. Ab: Weak commissural stimulation (arrow)-induced inhibition, paired with dendritic depolarization, abolished the slow spike without much affecting the Na+ spikes. Ac: Repetitive (2 Hz), strong commissurd stimulation

tially decreased (long arrows). B: Hypothesis: dendritic inhibition hyperpolarizes and/or shunts the membrane and attenuates active back- propagation of Na+ spikes from soma to dendrites. i, inhibitory interneuron with dendritic targets. After Bum& G, Penttonen M, Nadady Z, Bragin A (1996) “Pattern and inhibition-dependent invasion of pyramidal cell dendrites by fast spikes in the hippocampus in vivo.” Prco Natl Acad Sci (USA) 939921-3925 by permission of . .

led to repetitive spiking. The amplitude oc the fast spikes is substan- Proceedings from the National Academy of Science.


A FIGURE 46. Pyramidal cell activity is differentially controlled by somatic and dendritic inhibition. A: A perisomatic IPSP can suppress repetitive firing of fast action potentials. Three action potentials were elicited by current injection into a pyramidal cell (top). Field IPSP was evoked by focal stimulation in the somatic region (middle). When the IPSP was initiated just after the first spike, subsequent spikes were suppressed (lower). B: A cut was made in the slice (shown in the diagram) to prevent activation of apical dendritic inhibitory synapses. Dendritic generation of Ca-dependent spikes are more effectively suppressed by dendritic IPSPs than by perisomatic IPSPs. Current injection into a CA3 pyramidal cell dendrite elicited a complex burst consisting of small fast spikes followed by a slower depolarization (top). A dendritic IPSP suppressed the slow Ca-dependent component of the potential in 40 of 53 trials. A perisomatic IPSP, of similar amplitude in the dendritic recording, was effective in only 13 of 51 trials. Five superimposed traces are shown. The slice preparation and electrode placement for selective activation of somatic and apical dendritic inhibitory synapses is shown in the diagram. In A and B, excitatory synapses were blocked with CNQX (20 pM) and APV (100 pM). C-F: Control of repetitive pyramidal cell firing by single perisomatic IPSPs. C: A single inhibitory cell (1) could block repetitive firing of Na spikes in a target pyramidal cell (2). Two action potentials elicited by depolarization of cell 2 (dotted trace). An IPSP initiated during the pyramidal cell depolarizing afterpotential suppressed the second spike. The second spike was suppressed on 73 of 11 5 trials, and in the other trials was delayed by 5-32 ms beyond its mean control latency. D: Light and electron microscopy showed the inhibitory cell (1) made three synaptic contacts on the soma of the pyramidal cell (2) at the sites indicated by black dots. E: The difference in pyramidal cell membrane repolarization after a single spike in the absence (dotted) and the presence (solid line) of an IPSP initiated by cell 1 (arrow) is shown. F: The IPSP initiated by cell 1 in the absence of activity in cell 2 is shown. Reproduced from Miles et d. (1996) “Distinct functional roles for somatic and dendritic inhibition in the hippocampus.” Neuron, 16:815-823 by permission of Cell Press. This figure was kindly prepared by Richard Miles.

jyiR zF+ & h 2 0 mV b~ I 2o mV

I 5 m v - - Soma I I - 10ms 20 ms Stim

10 ms

the anatomical identity of interneurons dominantly involved in the control of sodium spikes and calcium spikes and characterize the conditions necessary for the activation of these cells.

prevalent hypotheses proposed to explain the selective vulnerability of hippocampal neurons is the “excitotoxic hypothesis” of Olney (1978), which postulates that an increase in the release of excitatory amino acids leads to neuronal death by allowing a lethal

Given the widespread involvement of hippocampal interneurons in the control of principal cell networks, interneuronal impairment is expected to exert far-reaching consequences on hippocampal function. The consistent observation that blockade of GABAergic transmission precipitates seizures led to the assumption that loss of inhibitory neurons or impairment of GABAergic transmission may be causally related to epilepsy (Prince, 1978; Ben-Ari et al., 1979; Traub and Wong, 1982; Dichter and Ayala, 1987; McNamara, 1994; During et al., 1995). Many studies have been stimulated by these findings, but the results are rather controversial. The following anatomical and physiological descriptions will be largely limited to phenomena in which the involvement of interneurons is apparent.

Epileptic and/or ischemic challenges may induce long-term changes in the hippocampal circuitry. To date, one of the most

abnormal degree of cation and water influx into the cells, which may be calcium independent, and (2) the delayed “apoptotic-like” cell death (Pollard et al., 1994; Nitatori et al., 1995), which is calcium dependent (for review, see Schmidt-Kastner and Freund, 199 1). Besides calcium dependence, the two mechanisms also differ in reversibility and selectivity: the former affects all neurons that are predisposed to a toxic level of glutamatergic input and have glutamate receptors (not necessarily calcium permeable). Depending on the magnitude of the excitotoxic impact, these cells may recover to a normal ion and water homeostasis, unless the plasma membrane has been ruptured. Thus, there is also differential vulnerability in acute mechanism, but it depends largely on the level of excitotoxic input. In contrast with the acute mechanism, delayed cell death affects only the population of selectively vulnerable neurons, which in the hippocampus are the CA1 pyramidal cells. The delayed mechanism affects primarily the CA1


pyramidal neurons both in ischemia and epilepsy, whereas interneurons appear to be destroyed acutely (Freund and Magloczky, 1993; Hsu and Buzsiki, 1993; Magl6czky and Freund, 1993, 1995; Freund et al., 1990a,b, 1992). Accordingly, the crucial factors determining vulnerability via the acute mechanism-relevant from the point of interneurons-are the types and density of glutamatergic inputs and receptors, whereas the factors determining vulnerability via the delayed mechanism (primarily affecting pyramidal cells) include the calcium buffering capacity and/or the presence of the necessary biochemical machin- ery to activate apoptosis (Pollard et al., 1994; Nitatori et al., 1995). Based on this logic, the calcium-binding protein content of neurons might predict vulnerability, if at all, only in delayed calcium-dependent cell death.

Several studies have attempted to correlate the vulnerability of neurons with the presence or absence of the calcium-binding proteins PV, CB, and CR (Mudrick and Baimbridge, 1989; Nitsch et al., 1989a,b; Sloviter, 1989; Freund et al., 1990a,b, 1992; Baimbridge et al., 1992; Freund and Magloczky, 1993; Johansen et al., 1990). Intracellular injection of the calcium chelator BAPTA effectively protects neurons from stimulation-induced cell death in vitro (Scharfman and Schwartzkroin, 1989). Furthermore, intracellular injection of PV or CB was shown to reduce the rate of rise of calcium and altered the kinetics of its decay in dorsal root ganglion cells (Chard et al., 1993). Additional support for the “calcium-buffer’’ hypothesis derives from experiments showing that a subpopulation of SOM-containing interneurons, but not PV- and CB-immunoreactive interneurons, degenerate in ischemia and status epilepticus (Johansen et al., 1987, 1990; Sloviter 1989). However, studies that systematically examined the relationship between the distribution of calcium- binding proteins and neuronal vulnerability failed to demonstrate such a relationship in either forebrain ischemia (Freund et al., 1990a; Johansen et al., 1990; Freund and Magloczky, 1993; Magloczky and Freund, 1993) or status epilepticus (Freund et al., 1992). PV-immunoreactive neurons in all hippocampal regions are resistant. CA1 interneurons survive ischemic insults (Johansen et al., 1983; Nitsch et al., 1989a,b) and even persist for several months, despite a total loss of the surrounding pyramidal cells (Murdick and Baimbridge, 1989; Hsu et al., 1994). However, neurons in the reticular nucleus of the thalamus containing both PV and CR are especially vulnerable (Freund et al, 1990a). The spiny class of CR-positive cells in the hilus and the CA3 region is damaged shortly after ischemia, whereas the aspiny neurons containing the same calcium-binding protein are fairly resistant (Hsu and Buzsiki, 1993; Freund and Magloczky, 1993). The content of neuropeptides (NPY, CCK, or VIP) also fails to predict susceptibility to an ischemic insult (Grimaldi et al., 1990; Yanagihara et al., 1785). Some SOM-positive cells in the hilar region are vulnerable, whereas others in the CA3-CA1 regions are resistant. The lack of correlation with calcium-binding protein content is expected if interneurons are indeed damaged by an acute ion and water influx, which may be calcium independent. The correlation between calcium-binding protein content and vulnerability or resistence does not hold for the delayed death of pyramidal cells either because those in the CA3 region are far

more resistant (both in ischemia and epilepsy) than are pyramidal cells in CA1, and yet most of the latter cells contain CB, whereas the former do not.

Under extreme excitotoxic conditions, the acute mechanism may induce irreversible damage in both interneurons and principal cells at a consistent order of vulnerability. The sequence in decreasing order of vulnerability for principal cells is mossy cells > CA3c pyramidal cells > CA3a and CA3b pyramidal cells; the sequence for interneurons is spiny CR-containing cells in the hilus and CA3 stratum lucidum > SOM-containing hilar cells (partly overlapping with CR-positive cells) > other types of interneurons with radially oriented dendritic trees. This order of vulnerability confirms the lack of direct relationship not only with calcium- binding protein, neuropeptide, or transmitter content but also with the presence or absence of the GluR2 subunit, which is responsible for calcium permeability of AMPA channels (Hollmann et al., 1991; Geiger et al., 1995). The most vulnerable groups are the mossy cells and the CR and/or SOM-containing hilar neurons. Mossy cells lack calcium-binding proteins and contain the GluR2 subunit (Leranth et al., 1996), whereas the sensitive interneurons lack the GluR2 subunit and contain a calcium-binding protein. There is, however, one striking similarity between these otherwise totally different cell types, i.e., that the majority of their excitatory input derives from mossy fiber collaterals. In contrast, those cells at the end of the list of vulnerability (CA3a, CA3b pyramidal cells and interneurons with radial dendritic trees) have only relatively short segments of their dendritic trees in layers innervated by mossy fibers. Thus, the order of vulnerability shows a remarkable correlation with the density of mossy fiber innervation of both principal cells and interneurons (Freund and Magloczky, 1993; Hsu and BuzsPki, 1993). Dentate granule cells are highly resistant to excitotoxic impacts and may become hy- perexcitable and fire NMDA receptor-mediated bursts (at least in some epilepsy models; Mody and Heinemann, 1987; Otis et al., 1994). Therefore, they may represent the ultimate source of the deadly glutamate release leading to acute, irreversible swelling of cells in the hilus and CA3 stratum lucidum.

Another important message from the anatomical-pathological studies is that, with the exception of the SOM- and/or CR-containing hilar inhibitory cells, interneurons are relatively well preserved in epileptic tissue (Babb et al., 1989; Sloviter 1989). Perhaps their impaired afferent drive, due to principal cell loss, may be responsible for the altered inhibition (Sloviter, 1991; Bekenstein and Lothman, 1993). In contrast with this view, spontaneous IPSCs in granule cells are increased in brain slices obtained from epileptic (kindled) rats, coupled with a reduction of presynaptic inhibition of GABA release (Buhl et al., 1996). I t was hypothesized that the enhanced inhibition collapses during a seizure as a result of Zn2+ release by the mossy fibers.

Paradoxically, despite the critical involvement of impaired inhibition in a number of epilepsy models, direct recordings from inhibitory interneurons during epileptic or epileptiform activity are extremely scarce (Kogure, 1987; Scharfman, 1994; McBain, 1994; Bragin et al., 1995a). In the absence of direct evidence, it has been assumed that the recruited excitation overrides inhibition (Sutula et al., 1988). Inhibition fails either because the

ZNTERNEURONS OF THE HZPPOCAMPUS 451

amount of released GABA decreases, because the target principal cells become less sensitive to GABA (Thompson and Giihwiler, 1989; During et al., 1995), or because GABA may depolarize pyramidal neurons (Staley et al., 1995). Simultaneous recordings from interneuronal populations in vivo suggest that decreased firing of interneurons is yet another explanation for the emergence of population bursts in epileptic afterdischarges (Bragin et al., in press). The failure of a subset of interneurons to maintain a concerted high-frequency spiking during seizures may be brought about by accumulation of activity-dependent extracellular adenosine and the consequent presynaptic suppression of interneuron excitation (Dunwiddie and Haas, 1985; Khazipov et al., 1993, 1995; Katchman et al., 1994; Manzoni et al., 1994) or by their depolarization block due to excessive levels of glutamate and [K+],. Direct recordings from identified interneuronal populations are required to begin to reveal the role of inhibitory interneurons in the induction and maintenance of epileptic discharges.

The effect of anoxia on interneurons in CA1 stratum lacunosum-moleculare has been directly examined (Congar et al., 1995; Khazipov et al., 1995a,b). Anoxia generates anoxic and postanoxic outward currents, which are more prominent in interneurons than in pyramidal cells. The anoxic episode rapidly and reversibly suppresses evoked EPSCs in LM neurons, and this suppression is the major cause of depression of polysynaptic IPSCs in both principal cells and interneurons. EPSCs are more rapidly depressed in interneurons than in simultaneously recorded pyramidal cells. The depression is mediated by presynaptic Al-type adenosine receptors. Interestingly, GABA release and its regulation by presynaptic GABAB receptors appears unaffected by anoxia, suggesting that release of transmitter from GABAergic terminals is not affected by adenosine (Yoon and Rothman, 1991). These observations suggest that LM interneurons in the CA1 region are functionally disconnected from their excitatory glutaminatergic inputs by anoxia and immediately after the anoxic episode (Zhu and Krnjevic, 1994; Congar et al., 1995; Khazipov et d. , 1995a), which is similar to the disconnection of some hilar interneurons during seizures (Bragin et al., in press).

In summary, the circuitry in which the vulnerable interneurons are located appears more important in the induction of cell damage than does intracellular calcium-buffering capacity or peptide content. The resistance of the majority of interneurons may be explained by the degree of functional disconnection from glutamate inputs during epilepsy and the postischemic period.

As pointed out in the Introduction, the present review largely focused on rodents, particularly rats, but one should be aware of the difference among species with regard to various properties of interneurons. For example, considerable differences exist in the nrurochemical characteristics of interneurons between rodent species as close as the rat and the mouse. However, there are basic rules in connectivity, which are similar even between rats and primates. Species difference is an important issue to be addressed

here, not only because of the often false cross-species interpretation and correlation of anatomical and physiological data but also because variability of some features or phylogenetical conservation of others may indicate major differences in the relative importance of various characteristics for basic network operations in the archicortex. Such comparisons make sense only if archicortical structures are compared because the neocortex has a different origin and parallel evolution. Furthermore, such comparative data may also demonstrate whether the function of a cell type or any of its features was already present when archicortex was first formed in various reptilian species or it appeared during later stages of archicortical evolution. The cerebral cortex of the lizard shows remarkable similarities to the mammalian hippocampus and may be considered to be the most ancient archicortical structure in phy- logenesis, which already possesses the cell types, connections, and lamination typical of the mammalian hippocampus. The primate cerebral cortex possesses major quantitative and qualitative changes compared with rodents, whereas the hippocampus as an archicortical area is a more conserved structure. Here we aim to draw attention to these issues by providing pro and contra examples for the cross-species difference and consistency of different cell types and connections. However, it would be very difficult to give a complete account of the innumerable species differences that occur even if considering only interneurons.

XV.1. The Archicortex of Lizards

The dorsal and dorsomedial areas of the lizard cerebral cortex appear to be homologous to the hippocampus and the medial cortex to the dentate gyrus of mammals. The pyramidal cell bodies form a compact layer and have radially running apical and basal dendrites forming the outer (corresponding to strata radiatum and lacunosum-moleculare) and inner (corresponding to stratum oriens) plexiform layers, respectively (Regidor et al., 1974). GABAergic axon terminals of local origin innervate all major membrane domains of the pyramidal cells, the cell body, the axon initial segment, and the dendritic tree (Schwerdtfeger and Lopez- Garcia, 1986; Martinez-Guijarro et al., 1991, 1993). GABAergic boutons in the outer one-third of the outer plexiform layer are remarkably dense, which is similar to stratum lacunosum-molecdare of the mammalian CA1 region. According to Golgi studies, it is possible to distinguish interneurons that give rise to the perisomatic GABAergic terminals and to interneurons with axon arborizing in the dendritic layers (Martinez-Guijarro et al., 1990). Afferents to these interneuron types originate from the ipsilateral medial cortex (Zn-containing projection, which is analogous to the mammalian mossy fiber projection), the contralateral dorsal and dorsomedial cortices, and collaterals of local principal cells (Lohman and van Woerden-Verkley, 1976; Lopez-Garcia and Martinez-Guijarro, 1988; Martinez-Guijarro et al., 1990, 1993; Bernabeu et al., 1994). Interestingly, they are also supplied by subcortical afferents such as the GABAergic septa1 and serotonergic raphe projections, which, similarly to mammals (see Section V), selectively innervate these interneurons (Martinez-Guijarro and Freund, 1992b; Martinez-Guijarro et al., 1994). These basic rules in interneuronal connectivity are remarkably similar to those


found in the mammalian hippocampus. Therefore, they are likely to be phylogenetically ancient features of cortical architecture, which have been highly conserved in the archicortex during evolution.

However, when studying more complex connectivity features and neurochemical properties, numerous differences become apparent. For example, there is no evidence for the existence of interneurons that specifically innervate other interneurons in the lizard cortex as the CR- and some of the VIP-containing cell do in the rat hippocampus (see Section 111.4). In lizards, CR coex- ists with the other two calcium-binding proteins, PV and CB, and is present in GABAergic basket cells innervating the perisomatic region of pyramidal neurons (Martinez-Guijarro and Freund, 1992a). It is interesting to note here that these three calcium-binding proteins, or at least two of them, often occur in the same cells during early cortical development in rodents (Alcantara et al., 1996) and even in monkey (Yan et al., 1995), although in adults they are present in interneuron types with different connectivity and function (see Section IV.2). Neuropeptides, including SOM and NPY, are also present in GABAergic interneurons in the lizard cortex. They show a minor overlap with the interneurons that contain calcium-binding proteins and are located in the inner or outer plexiform layers. They are likely to be involved in dendritic inhibition because they form an axon plexus in the outer one-third of the outer plexiform layer (Divila et al., 1991, 1993; Martinez-Guijarro et al., 1993) in a fashion remarkably similar to the SOM-NPY-containing 0-LM cells in the mammalian hippocampus (Sections III.3.2.a, IV.3.a, IV.3.b). Thus, the segregation of interneuron types responsible for the selective innervation of the perisomatic versus the distal dendritic regions of principal cells appears to have taken place very early in archicortical evolution. This notion further suggests that the differential control of synaptic plasticity and principal cell spike transmission by separate classes of interneurons (Miles et al., 1996) is a fundamental and indispensable component of network operations in the archicortex. The selective innervation of these separate interneuron classes by different subcortical pathways, which may subserve a behavior-dependent modulation of network events, is also a phylogenetically well-preserved organizational principle (Freund, 1992; Martinez-Guijarro and Freund, 1992b; Martinez-Guijarro et al., 1994).

XV.2. Variability Among Rodent Species The most widely used laboratory animal species are the rat,

mouse, guinea pig, and gerbil. As expected, the anatomical wiring of interneuronal circuits is similar among these species. The guinea pig appears to be a relatively distant species from the rat and mouse, a point that is also apparent in its hippocampal structure. The approximate frequency, laminar distribution, afferent inputs, and postsynaptic targets of interneurons show rather small differences. Interneurons in stratum lacunosum-moleculare are more frequent in the guinea pig than in other rodents (K. T6th and T.F. Freund, unpublished observations), but the postsynaptic targets of these neurons are unknown. The majority of cross- species variability is found in the neurochemical characteristics of

certain interneuron types and principal cells. Nevertheless, the general distribution of neurons containing calcium-binding proteins and neuropeptides is similar, but there are interesting examples of differences. Numerous PV-immunoreactive axon terminals arising from entorhinal pyramidal cells form asymmetrical synapses in stratum moleculare of the gerbil hippocampus (Scotti and Nitsch, 1991). PV is not present in perforant path terminals in other rodents or in primates. The spiny CR-positive cells, described in the rat hippocampal formation (Gulyis et al., 1992), are absent both in the hilus and in CA3 stratum lucidum of guinea pigs. Instead, guinea pigs have numerous aspiny CR-positive cells in stratum lacunosum-moleculare, which are few in number in the rat and mouse (K. T6th and T.F. Freund, unpublished observations). In this respect, the guinea pig hippocampal formation is more similar to the primate than to other rodents. This notion is further supported by the presence of SP in hilar neurons and in the supramammillary afferents of the hippocampal formation in primates and guinea pigs, which is absent in rats and mouse (Ljungdahl et al., 1978; Gall and Selawski, 1984; Haglund et al., 1984; Nitsch and Leranth, 1994).

The major differences between rats and mouse are in the neurochemical characteristics of principal cells, whereas GABAergic interneurons appear very similar. For example, CCK is present in mossy fibers in mouse and guinea pig (Gall and Selawski, 1984; Gall et al., 1986) but not in the rat (Handelmann et al., 1981; Kosaka et al., 1985; Sloviter and Nilavet, 1987). In addition to being present in intetneurons, CR is expressed in hilar mossy cells and some CA3c pyramidal cells in the ventral hippocampus of the mouse (Liu et al., 1996), but it is confined to interneurons in the rat and guinea pig (Gulyis et al., 1992). Pyramidal cells in the CA1 region of the guinea pig lack CB, whereas the “superficial cells” in other rodents show a characteristic staining (Section IV.2.b). Additional species differences will probably emerge in future colocalization studies, but to date such data in the mouse, guinea pig, and gerbil are scarce. However, even this sniall col- lection illustrates that anatomical data, especially if based on immunocytochemical localization of different neuropeptides and calcium-binding proteins, should be considered carefully when employed to interpret findings in different species, even in closely related rodents.

XV.3. Relevance to Primate Anatomy The major differences between primates and rodents in hip-

pocampal cytoarchitectonics, principal cell numbers, and morphology have been extensively reviewed elsewhere (Rosene and Van Hoesen, 1987; Seress, 1988, 1992; West, 1990; Frotscher ct al., 1991). The axonal arborization patterns of interneurons are difficult to compare because the most complete picture available about such features of interneurons derives from in vivo intracellular recording and labeling studies in rodents, which are rarely performed in monkeys, even in vitro. Thus, the morphological features of interneuron types in rodents can be compared only with partially visualized examples from Golgi studies in monkey and human. Although the dendritic trees of Golgi impregnated neurons are usually extensive only fractions of the entire axon ar-


bors are revealed. Nevertheless, these details are often sufficient to establish the major connectivity features of the cells, the potential sources of inputs, and postsynaptic target specificity. At this level of analysis, primate and rodent interneuron types are similar. It has been repeatedly emphasized that, although processes of principal cells increase in complexity and variability in the monkey compared to rodents, the basic features of basket and chandelier cells are remarkably constant (Seress and Frotscher, 1991; Somogyi et al., 1983a). An interesting finding is the relatively small number of GABAergic synapses on the cell bodies of granule cells in primates vs. the rat (Seress and Ribak, 1992). This finding, taken together with a smaller proportion of pyramidal- like basket cells in the dentate gyrus (Seress and Pokorny, 1981), led to the assumption that sparse perisomatic inhibition may explain why primate granule cells are prone to epileptiform burst firing (Seress, 1992). This, however, turns out to be a far more complex issue (Section XIV).

Greater variability is found when the distribution of neurochemically identified cell types is considered. For example, immunostaining for PV visualizes a type of basket or chandelier cell in the monkey that lacks “basal” dendrites penetrating the hilus, suggesting that these cells have apparently no access to input from local mossy fiber collaterals (i.e., no feedback drive). Furthermore, a horizontally oriented PV-positive cell type in the outer molecular layer has been reported in the monkey dentate gyrus (Seress et al., 1991). Such neurons have not been seen in the rat (Sloviter, 1989; Ribak et al., 1990). In the CA1 subfield of the monkey hippocampus, some basket cells have very large cell bodies and multipolar dendritic trees (Seress et al., 1991; Pitkanen and Amaral, 1993), whereas the same cell type is much smaller in rodents (Kosaka et al., 1987). This difference was suggested to be associated with the expansion of the pyramidal cell layer, i.e., a larger space has to be supplied by axons of single basket cells (Ribak et al., 1993). PV-positive boutons of unknown origin are abundant in stratum lacunosum-moleculare of the CA2 subfield in primates (Leranth and Ribak, 1991) but are absent in this layer in rats. Other types of PV-positive cells and all CB-containing interneuron types are similar (see Section IV.2a.b), with the exception of CB-positive cells in the dentate molecular layer. These cells have horizontal dendritic trees and are more numerous in the monkey than in the rat and even more so in the human (Seress et al., 1991, 1993a). The distribution of CR-positive neurons shows more pronounced differences. The spiny cell type, which was originally described in the hilus and CA3 stratum lucidum in the rat (Gulyis et. al., 1392), has not been observed in the monkey (Seress et al., 199%). These neurons in the rat are likely to correspond to the long-spined multipolar cells described by Amaral (1978). The question arises in such cases as to whether a cell type is indeed missing from that particular species or even ex- ists but lacks that particular marker. If the particular cell type ex- ists, evidence could be obtained by Golgi impregnation. However, the absence of a cell type is very difficult to prove experimentally. Another striking difference in the distribution of CR is the abundance of CR-containing cells in stratum lacunosum-moleculare and the bordering stratum radiatum in monkey and human (Nitsch and Leranth, 1993; Seress et al., 1993b; Nitsch and Ohm,

1995). Cells in this location are also frequent in the guinea pig but scarce in the rat (see above). Similar differences between rats and primates are apparent in the number of CR-positive cells in the dentate molecular layer. The radially oriented CR-positive cells occur in all layers of the hippocampus in rats, whereas these cells are distributed more toward the distal stratum radiatum and extremely few cells are visible in stratum oriens in primates (Nitsch and Leranth, 1993; Seress et al., 1993b; Nitsch and Ohm, 1995).

The overall distribution of axonal and dendritic arbors of interneurons immunoreactive for NPY and/or SOM is highly conserved throughout phylogeny. A similar population of cells in the hilar polymorphic area is immunoreactive for these peptides in the rat, monkey, and human dentate gyrus and give rise to a dense axon terminal field in the outer two-thirds of the molecular layer (Bakst et al., 1985, 1986; Chan-Palay et al., 1986; Kohler et al., 1986, 1987; Chan-Palay, 1987; Amaral et al., 1988; Nitsch and Leranth, 1991). The same cell population in primates also contains SP (Seress and Leranth, 1996), whereas there is a much smaller degree of colocalization in rats, although SP immunostaining in this species is highly variable and strongly depends on the type of fixative used (Z. Borhegyi, L. Seress, and C. Leranth, personal communication). In the hippocampus, the density of NPY- and SOM-containing fibers is highest in stratum lacunosum-moleculare, and the parent cell bodies are located predominantly in stratum oriens in both primates and rodents. Thus, GABAergic interneurons specialized to innervate the most distal dendrites of principal cells appear to be similar throughout the phylogenetic scale from lizards through rodents up to the monkey and human archicortex, not only in their efferent connections but also in neuropeptide content and in afferent drive (mostly feedback; see Section IV.3.a). However, there are qualitative and quantitative differences in cell types located (and probably also terminating) in the dendritic layers of both the dentate gyrus and the hippocampus. The CB- and CR-containing neurons in these locations are more numerous in primates (see above), suggesting that feed-forward dendritic inhibition, involved in the control of afferent synaptic plasticity, may be more complex in these species.

The rich interconnectivity of interneurons and their hypothesized role in synaptic plasticity provide the possibility that the different interneuronal types may selectively influence the efficacy of afferent inputs. For example, during exploration-associated theta-gamma oscillations, VIP-positive interneurons with dendritic targets may suppress other interneurons innervating the distal apical dendrites of principal cells (0-LM and HIPP cells). As a result, the direct entorhinal-CA1 transmission is facilitated, whereas the efficacy of the intrahippocampal associational connections are suppressed. Obviously, working hypotheses like these will be continuously generated and modified in the process of re- vealing the precise connectivity and physiological properties of in-


terneurons. Clarifying the relation between the chemical content of the neuron and its axonal targets is the first logical step in the classification process. A further level of analysis is required to establish a correlation between electrophysiological properties and morphology/immunocytochemistry. Intracellular recordings from interneurons in vitro and their subsequent visualization are a direct approach for studying the functional properties of interneurons that provides the essential bridge between physiological features and their structural substrate. In turn, intracellular recordings from anatomically identified interneurons in whole animals enables us to relate interneuron activity to network patterns. The differential involvement of the interneuronal types in the various oscillations and intermittent bursts indicates that examina- tion of their associations with population patterns is the correct strategy for comparing their functional and anatomical features. Because some of these population patterns can also be induced in the anesthetized animal, the neurons involved can be labeled and anatomically characterized. The location, firing pattern, and afferent responses of the subtypes can be examined in the behaving animal. Extracellular labeling of the physiologically and behaviorally characterized interneurons in the awake animal is a possibility that should be explored. Understanding the complex interactions between interneurons and principal cells will benefit from computational modeling. Overall, in vivo and in vitro studies will complement each other and provide a precise knowledge about physiological properties, connectivity, and immunocytochemical identification of interneurons. Such knowledge is a prerequisite for understanding of the operational principles of the hippocampal network and may serve as a model for exploring neocortical circuitries.

Caveats and expected difficulties in the classification process should also be ackowledged. Precise biophysical, pharmacological, and network analyses of interneuron-mediated effects depend on a reliable taxonomy of interneuron types. To date, interneuron classes cannot always be unanimously defined even when multiple criteria are used. For example, “typical” basket cells also have their variants innervating the proximal dendrites, and at least two subsets exist that contain different neurochemical markers. However, even within the same subclasses, the dendritic and axonal features may have large variability (Gulyis et al., 1993a; Sik et al., 1995). Such observations raise the intriguing possibility that axon length and even the targets of interneurons may vary not only as part of the ontogenetic development but also in the adult in a use-dependent manner. Such plasticity of interneurons would especially be significant for shaping the topographical representations of information in hippocampal networks. The available evidence along these lines is in support of the uniquely plastic properties of CABAergic axons. When a sciatic nerve is implanted into the thalamus, the overwhelming majority of cells that grow axons into the nerve graft are GABAergic cells of the reticular nucleus (David and Aguayo, 1981). Although neurons of the reticular nucleus normally innervate only thalamic nuclei, their newly generated axon collaterals into the peripheral nerve grow longer than 50 mm. NPY-immunoreactive interneurons of the hilar region also display a remarkable sprouting of their axon collaterals (Deller et al., 1995a). Obviously, the task is to determine how much variability is reasonable in the classification process and to examine

whether such hypothesized plasticity is reflected in the variation of axon length only or whether the somadendritic targets also change. Establishing a physiology-based taxonomy of interneurons and determination of their precise connectivity within and outside the hippocampal formation will reveal new horizons on the road on which Ramon y Cajal began his walk many decades ago.

Acknowledgments

This work was supported by the NIH (NS34994), Human Frontier Science Program Organization, Howard Hughes Medical Institute, OTKA (T 16942), Hungary, and the Whitehall Foundation. We thank Richard Miles and Roger Traub for their constant encouragement, discussions, and friendship. We are grateful to our immediate colleagues, LhzM Acsidy, Attila Gulyis, Norbert Hijos, Isrva‘n Katona, Zoltan Nadasdy, Attila Sik, and Katalin Toth, for their help with suggestions to several versions of this manuscript and with the preparation of figures. We are grateful to Katalin Halasy, Toshio Kosaka, Jean-Claude Lacaille, Csaba Leranth, Chris McBain, Richard Miles, and Peter Somogyi for their contribution of valuable figures to the review. Consultation with several individuals helped shape the final version of the manuscript, including David G. Amaral, Zsolt Borhegyi, Eberhard Buhl, James J. Chrobak, Cynthia Dolorfo, Michael Frotscher, Toshio Kosaka, Brian Leonard, Csaba Leranth, Zsofia Magloczky, Francisco Martinez-Guijarro, Chris McBain, Isrvin Mody, Charles Ribak, Philip A. Schwartzkroin, Liszl6 Seress, IvLn Solttsz, Peter Somogyi, Armin Stelzer, James Tepper, and Xiao-Jing Wang. The excellent technical assistance of Erzsebet Borok, Agnes Miiller, Andrea Zoldi, Ggbor Terstyinszky, and Isrvin Csap6 with the preparation of the il- lustrations and the text is gratefully acknowledged.

Abraham WC, Gustafsson B, Wigstrom H (1986) Single high strength afferent volleys can produce long-term potentiation in the hippocampus in vitro. Neurosci Lett 70:217-222.

Abraham WC, Gustafsson B, Wigstrom H (1987) Long-term potentiation involves enhanced synaptic excitation relative to synaptic inhibition in guinea-pig. J Physiol (Lond) 394:367-380.

Acsddy L, Halasy K, Freund T F (1993) Calretinin is present in nonpyramidal cells of the rat hippocampus-111. Their inputs from the median raphe and medial septa1 nuclei. Neuroscience 52:829-841.

Acsddy L, Papp E, Hdjos N, Freund T F (1994) Subcortical innervation of VIP-containing interneurons in the hippocampus. Eur J Neurosci (Suppl) 7:209.

Acsddy L, Arabadzisz D, Freund T F (19964 Correlated morphological and neurochemical features indentify different subsets of VIP-immunoreactive interneurons in rat hippocampus. Neuroscience 73:

Acsddy L, Gorcs TJ, Freund TF (1996b) Different populations of VIP- immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus. Neuroscience 73:317-334.

Acsidy L, Katona I, Arabadzisz D, Gulyds AI, Shigemoto R, Freund TF (submitted) Substance P receptor labels GABAergic cells with distinct termination pattern in the hippocampus. J Comp Neurol.

299-3 15.

INTERNEURONS OF THE HIPPOCAMPUS 455 ~~

Aika Y, Ren JQ, Kosaka K, Kosaka T (1994) Quantitative analysis of GABA-like-immunoreactive and parvalbumin-containing neurons in the CAI region of the rat hippocampus using a stereological method, the disector. Exp Brain Res 99:267-276.

Aitken PG, Jaffe DB, Nadler JV (1991) Cholecystokinin blocks some effects of kainic acid in CA3 region of hippocampal slices. Peptides 12: 127-129.

Alcantara S, de Lecea L, Del Rio JA, Ferrer I, Soriano E (1996) Transient co-localization of parvalbumin and calbindin D28K in the postnatal cerebral cortex: evidence for a phenotypic shift in developing nonpyramidal neurons. Eur J Neurosci 8: 1329-1339.

Alger B, Nicoll RA (1979) GABA-mediated biphasic inhibitory responses in hippocampus. Nature 28 1 :3 15-3 17.

Alger B, Nicoll RA (1982a) Feed-forward dendritic inhibition in rat hippocampal pyramidal cells studied in vitro. J Physiol (Lond) 328: 105-1 23.

Alger B, Nicoll RA (1982b) Pharmacological evidence for two kinds of GABA receptor on rat hippocampal pyramidal cells studied in vitro. J Physiol (Lond) 328:125-141.

Alger BE, P i t h TA, Williamson A (1990) A prolonged post-tetanic hyperpolarizations in rat hippocampal pyramidal cells in vitro. Brain Res 521:118-124.

Alonso A, Kohler C (1982) Evidence for separate projections of hippocampal pyramidal and non-pyramidal neurons to different parts of the septum in the rat brain. Neurosci Lett 31:209-214.

Alonso A, Khateb A, Fort P, Jones BE, Muhlethaler M (in press) Differential oscillatory properties of cholinergic and non-cholinergic nucleus basalis neurons in guinea big brain slices. J Neurophysiol.

Alvarez-Leefmans FJ, Gardner-Medwin AR (1975) Influences of the septum on the hippocampal dentate area which are unaccomplanied by field potentials. J Physiol (Lond) 249: 14P-16P.

Amaral DG (1978) A Golgi study of cell types in the hilar region of the hippocampus in the rat. J Comp Neurol 182:851-914.

Amaral DG, Witter MP (1 989) The three-dimensional organization of the hippocampal formation: a review of anatomical data. Neurosci- ence 31:571-591.

Amaral DG, Witter MP (1995) Hippocampal formation. In: The rat nervous system, 2nd ed (Paxinos G, ed), pp 443-494. New York: Academic Press.

Amaral DG, Insausti R, Campbell MJ (1988) Distribution of somatostatin immunoreactivity in the human dentate gyrus. J Neurosci 8: 3306-3316.

Amaral DG, Dolorfo C, Alvarez-Roy0 P (1991) Organization of CA1 projections to the subiculum: a PHA-L analysis in the rat. Hippo- campus 4:415-436.

Andersen P, Eccles J C (1962) Inhibitory phasing of neuronal discharge. Nature (Lond) 196:645-647.

Andersen P, Eccles JC, Loyning Y (1963) Recurrent inhibition in the hippocampus with identification of the inhibitory cell and its synapse. Nature 198:540-542.

Andersen P, Eccles JC, Loyning Y (1964) Location of synaptic inhibitory synapses on hippocampal pyramids. J Neurophysiol 27:592-607.

Andersen P, Gross GN, Lomo T, Sveen 0 (1 969) Participation of inhibitory and excitatory interneurons in the control of hippocampal cortical output. MIT Press, Boston. In: The interneuron (Brazier M, ed), pp 415-432.

Andersen P, Dingledine R, Gjerstad L, Langmoen LA, Mosfeldt-Laursen A (1980) Two different responses of hippocampal pyramidal cells to application of gamma-aminobutyric acid. J Physiol (Lond) 305: 279-296.

Andrade R, Nicoll RA (1987) Pharmacologically distinct actions of serotonin on single pyramidal neurons of the rat hippocampus recorded in vitro. J Physiol (Lond) 39499-124.

Arancio 0, Korn H , Gulyas A, Freund TF, Miles R (1994) Excitatory synaptic connections onto rat hippocampal inhibitory cells may involve a single transmitter release site. J Physiol (Lond) 481(2): 395-405.

Arimatsu Y, Naegele JR, Barnstable CJ (1987) Molecular markers of neu-

ronal subpopulations in layers 4, 5, and 6 of cat primary visual cortex. J Neurosci 7:1250-1263.

Asanuma C, Porter LL (1992) Light and electron microscopic evidence for a GABAergic projection from the caudal basal forebrain to the thalamic reticular nucleus in rats. J Comp Neurol 302:159-172.

Ashwood TJ, Lancaster B, Wheal HV (1984) In vivo and in vitro studies on putative interneurones in the rat hippocampus: possible me- diators of feed-forward inhibition. Brain Res 293:279-29 1.

Babb TL, Pretorius JK, Kupfer WR, Crandall PH (1989) Glutamate decarboxylase-immunoreactive neurons are preserved in human epileptic hippocampus. J Neurosci 9:2562-2574.

Baimbridge KG, Miller JJ (1982) Immunohistochemical localization of calcium-binding protein in the cerebellum, hippocampal formation and olfactory bulb of the rat. Brain Res 245:223-229.

Baimbridge KG, Miller JJ, Parkes CO (1982) Calcium-binding protein distribution in the rat brain. Brain Res 239:519-525.

Baimbridge KG, Celio MR, Rogers J H (1992) Calcium-binding proteins in the nervous system. Trends Neurosci 15:303-308.

Bakst I, Morrison JH, Amaral DG (1985) The distribution of somatostatin-like immunoreactivity in the monkey hippocampal formation. J Comp Neurol 236:423-442.

Bakst I, Avendano C, Morrison JH, Amaral DG (1986) An experimental analysis of the origins of somatoscatin-like immunoreactivity in the dentate gyrus of the rat. J Neurosci 6:1452-1462.

Barnes CA, Foster TC, Rao G, McNaughton BL (1991) Specificity of functional changes during normal brain aging. Ann NY Acad Sci 640:80-85.

Battenberg E, Morales M, de Lecea L, Johnson DS, Bloom FE (1994) Immunocytochemical characterization of neurons expressing the 5TH3 receptor. SOC Neurosci Abstr 20: 1 156.

Baude A, Nusser Z, Roberts JD, Mulvihill E, McIlhinney RA, Somogyi P (1993) The metabotropic glutamate receptor (mGluR1 alpha) is concentrated at perisynaptic membrane of neuronal subpopulations as detected by immunogold reaction. Neuron 11:771-787.

Baude A, Nusser Z, Molnar E, McIlhinney RAJ, Somogyi P (1995) High- resolution immunogold localization of ampa type glutamate receptor subunits at synaptic and non-synaptic sites in rat hippocampus. Neuroscience 69: 103 1-1055.

Behrends JC, Ten Bruggencate G (1 993) Cholinergic modulation of synaptic inhibition in the guinea pig hippocampus in virro: excitation of GABAergic interneurons and inhibition of GABA release. J Neurophysiol 69:626-629.

Bekenstein JW, Lothman EW (1993) Dormancy of inhibitory interneurons in a model of temporal lobe epilepsy. Science 25997-100.

Ben-Ari Y (1994) Kainate-induced apoptotic cell death in hippocampal neurons. Neuroscience 63:7-18.

Ben-Ari Y, Krnjevic K, Reinhardt W (1979) Hippocampal seizures and failure of inhibition. Can J Physiol Pharmacol 57: 1462-1466.

Benardo LS, Prince DA (1982a) Cholinergic excitation of mammalian hippocampal pyramidal cells. Brain Res 249:315-33 1.

Benardo LS, Prince DA (1 982b) Ionic mechanisms of cholinergic excitation in mammalian hippocampal pyramidal cells. Brain Res 249:333-344.

Bergles DE, Doze VA, Madison DV, Smith SJ (1996) Excitatory actions of norepinephrine on multiple classes of hippocampal CAI interneurons. J Neurosci 16:572-585.

Bernabeu A, Martinez-Guijarro FJ, Luis de la Iglesia JA, Lopez-Garcia C (1994) An axosomatic and axodendritic multipolar neuron in the lizard cerebral cortex. J Anat 184:567-582.

Berzaghi M, Freund TF, Papp E, Lindholm D, Zirrgiebel U, Castren E, Thoenen H (submitted) Brain-derived neurotropic factor (BDNF) in the rat cerebral cortex: mismatch between site of synthesis and accumulation. J Neurosci.

Bialek W, Rieke F, van Steveninck RR (1991) Reading a neural code. Science 252:1854-1857.

Bijak M, Misgeld U (1995) Adrenergic modulation of hilar neuron activity and granule cell inhibition in the guinea-pig hippocampal slice. Neuroscience 67:541-550.

Bilkey DK, Goddard GV (1985) Medial septa1 facilitation of hip-


pocampal granule cell activity is mediated by inhibition of inhibitory interneurons. Brain Res 361:99-106.

Blackstad TW (1956) Commissural connections of the hippocampal region of the rat, with special reference to their mode of termination. J Comp Neurol 105:417-537.

Blackstad TW, Kjaerheim A (1961) Special axo-dendritic synapses in the hippocampal cortex: electron and light microscopic studies on the layer of mossy fibers. J Comp Neurol 117:113-I 59.

Bland BH (1990) Physiology and pharmacology of hippocampal formation theta rhythms. Prog Neurobiol 26: 1-54.

Blasco-Ibanez JM, Freund T F (1995) Synaptic input of horizontal interneurons in stratum oriens of the hippocampal CAI subfield: structural basis of feed-back activation. Eur J Neutosci 7:2170-2180.

Bliss TVP, Collingridge GL (1993) A synaptic model of memory: long- term potentiation in the hippocampus. Nature 361 :31-39.

Bliss ’rVP, Lomo T (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol (Lond) 232:331-356.

Bobker D H , Williams JT (1990) Ion conductances affected by 5-HT receptor subtypes in mammalian neurons. Trends Neurosci

Bradley SR, Levey AI, Hersch SM, Conn PJ (1996) Immunocyto-chemical localization of group I11 metabotropic glutamate receptors in the hippocampus with subtype-specific antibodies. J Neurosci 16:2044- 2056.

Bragin AG, Otmakhov NA (1979) Comparison of direct influences of the perforant path on hippocampal CA1 and CA3 neurons in vitro [in Russian]. Neurophysiology 11:220-225.

Bragin A, Jand6 G, NadBsdy Z, Hedke J, Wise K, Buzsdki G (1995a) 40-100Hz, Oscillation in the Hippocampus of the behaving rat. J Neurosci 15:47-60.

Bragin A, Jand6 G, Nadisdy Z, van Landeghem M, and Buzsiki G (1995b) Dentate EEG spikes and associated population bursts in the hippocampal hilar region of the rat. J Neurophysiol 73:1691-1705.

Bragin A, Csicsvary J, Penttonen M, Buzsdki G (in press) Epileptic af- terdischarge in the hippocampo-entorhinal system: current source density and unic studies. Neuroscience.

Buckmaster PS, Schwartzkroin PA (1 9954 Interneurons and inhibition in the dentate gyrus of the rat in vivo. J Neurosci 15:774-789.

Buckmasrer PS, Schwarzkroin PA (1 995b) Physiological and morphological heterogeneity of dentate gyrus-hilus interneurons in the gerbil hippocampus in vivo. Eur J Neurosci 7:1393-1402.

Buckmaster PS, Strowbridge BW, Kunkel DD, Schmiege DL, Schwartzkroin PA (1992b) Mossy cell axonal projections to the dentate gyms molecular layer in the rat hippocampal slice. Hippocampus 2:349-362.

Buckmaster PS, Kunkel DD, Robbins RJ, Schwartzkroin PA (1994) Somatostatin immunoreactivity in the hippocampus of mouse, rat, guinea pig and rabbit. Hippocampus 4:167-180.

Buckmaster PS, Wentzel HJ, Kunkel DD, Schwartzkroin PA (1996) Axon arbors and synaptic connections of hippocampal mossy cells in the rat in vivo. J Comp Neurol 366:270-292.

Buhl EH, Halasy K, Somogyi P (1994a) Diverse sources of hippocampal unitary inhibitory postsynaptic potentials and the number of synaptic release sites. Nature 368:823-828.

Buhl EH, Han ZS, Lorinczi Z, Stezhka W, Karnup SV, Somogyi P (1994b) Physiological properties of anatomically identified axo-axonic cells in the rat hippocampus. J Neurophysiol 71:1289-1307.

Buhl EH, Cobb SR, Halasy K, Somogyi P (1995) Properties of unitary IPSPs evoked by anatomically identified basket cells in the rat hippocampus. Eur J Neurosci 7: 1989-2004.

Buhl EH, Otis TS, Mody I (1996) Zinc-induced collapse of augmented inhibition by GABA in a temporal lobe epilepsy model. Science 271 :369-373.

Burgess N, Recce M, O’Keefe J (1994) A model of hippocampal function. Neural Networks 7: 1065-1081.

Burnashev N, Khodorova A, Jonas P, Helm PJ, Wisden W, Monyer H,

10:502-509.

Seeburg PH, Sakmann B (1 992) Calcium-permeable AMPA-kainate receptors in fusiform cerebellar glial cells. Science 256: 1 566-1 570.

Buzsiki G (1984) Feed-forward inhibition in the hippocampal formation. Prog Neurobiol 22:131-153.

Buzsdki G (1986) Hippocampal sharp waves: their origin and significance. Brain Res 398:242-252.

Buzsiki G (1989) A two-stage model of memory trace formation: a role for “noisy” brain states. Neuroscience 31:55 1-570.

Buzsiki G (1 996) The hippocampal-neocortical dialogue. Cereb Cortex 6:s 1-92.

Buzdki G, Chrobak JJ (1995) Temporal structure in spatially organized neuronal ensembles: a role for interneuronal nerworks. Curr Opin Neurobiol 5:504-510.

Buzsiki G, Czeh G (1981) Commissural and perforant path interactions in the rat hippocampus: field potentials and unitary activity. Exp Brain Res 225:234-247.

Buzsiki G, Eidelberg E (1981) Commissural projection to the dentate gyrus of the rat: evidence for feed-forward inhibition. Brain Res 230:346-350.

B u ~ s i k i G, Eidelberg E (1982) Direct afferent excitation and long-term PO- tentiation of hippocampal interneurons. J Neurophysiol 48:597-607.

Buzsiki G, Eidelberg E (1983) Phase relations of hippocampal projection cells and interneurons to theta activity in the urethane anesthetized rat. Brain Res 266:334-338.

Buzsiki G, Leung I,, Vandenvolf C H (1983) Cellular bases of hippocampal EEG in the behaving rat. Brain Res Rev 6:139-171.

Ruzsdki G, Gage FH, Czopf J, Bjorklund A (1987) Restoration of iyth- mic slow activity in the subcortically denervated hippocampus by fe- tal CNS transplants. Brain Res 400:334-347.

Buzsdki C, Ponomareff G, Bayardo, F, Ruiz, R, Gage FH (1989) Neuronal activity in the subcortically denervated hippocampus: a chronic model for epilepsy. Neuroscience 28:527-538.

Buzsdki G, Horvath Z, Urioste R, Hetke J, Wise K (1992) High-frequency network oscillation in the hippocampus. Science 256:

Buzsiki G, Bragin A, Chrobak JJ, Nadasdy Z, Sik A, Hsu M, Minen A (1994) Oscillatory and intermittent synchrony in the hippocampus: relevance to memory trace formation. In: Temporal coding in the brain, pp 145-172. G Buzsiki, R Llinas, W Singer, A Berthoz, Y Christen (eds). Berlin: Springer-Verlag.

Buzsdki G, Ylinen A, Penttonen M, Bragin A, Nadasdy Z, Chrobak JJ (1995) Possible physiological role of the perforant parh-CAl projection. Hippocampus 5:141-146.

Buzsiki G , Penttonen, M, Nadasdy Z, Bragin A (1996) Pattern and inhibition-dependent invasion of pyramidal cell dendrites by fast spikes in the hippocampus in vivo. Proc Natl Acad Sci (USA).

Caffe AR, van Leeuwen PW, Luiten I’G (1987) Vasopressin cells i n the medial amygdala of the rat project to the lateral septum and ventral hippocampus. J Comp Neurol 261:237-252.

Celio MR (1 986) Parvalbumin in most gamma-aminoburyric acid-con- raining neurons of the rat cerebral cortex. Science 231:995-997.

Celio MR (1990) Calbindin D-28k and parvalbumin in the rat nervous system. Neuroscience 35:375-475.

Chan-Palay V (1987) Somatostatin immunoreactive neurons in the human hippocampus and cortex shown by immunogold/silver intensification on vibratome sections: coexistence with neuropeptide Y neurons, and effects in Alzheimer-type dementia. J Comp Neurol 260:20 1-22.3.

Chan-Palay V, Kohler C, Haesler U, Lang W, Yasargil G (1986) Dis- tribution of neurons and axons immunoreactive with antisera against neuropeptide Y in the normal human hippocampus. J Comp Neurol

Chard 1 5 , Bleakman, Christakos S , Fullmer CS, Miller RJ (1993) Calcium buffering properties of calbindin D28K and parvalbumin in rat sensory neurones. J Physiol (Lond) 472: 341-357.

Chavkin C, Shoemaker WJ, McGinty JF, Bayon A, Bloom FE (1985) Characterization of the prodynorphin and proenkephalin neuropeptide systems in rat hippocampus. J Neurosci 5:808-816.

1025-1 027.

248 ~360-375.


Chedotal A, Cozzari C, Faure MP, Hartman BK, Hamel E (1994) Distinct choline acetyltransferase ( C U T ) and vasoactive intestinal polypeptide (VIP) bipolar neurons project to local blood vessels in the rat cerebral cortex. Brain Res 646:181-193.

Chickwendu A, McBain CJ (in press) Two temporally overlapping “delayed-rectifiers” determine the voltag-dependent potassium current phenotype in cultured hippocampal interneurons. J Neurophysiol.

Choi D W (1988) Glutamate neurotoxicity and diseases of the central nervous system. Neuron 1623-634.

Chrobak JJ, Buzsiki G (1996) High-frequency oscillations in the output networks of the hippocampal-entorhinal axis of the freely-behaving rat. J Neurosci 16:3056-3066.

Chronwall BM, DiMaggio DA, Massari VJ, Pickel VM, Ruggiero DA, O’Donohue TL (1985) The anatomy of neuropeptide-Y-containing neurons in rat brain. Neuroscience 15: 1 159-1 18 1.

Claiborne BJ, Amaral DG, Cowan WM (1986) A light and electron microscopic analysis of the mossy fibers of the rat dentate gyrus. J Comp Neurol 246:435-458.

Claiborne BJ, Amaral DG, Cowan WM (1 990) Quantitative, 3-dimensional analysis of granule cell dendrites in the rat dentate gyrus. J Comp Neurol 302:206-219.

Cobb SR, Buhl EH, Halasy K, Paulsen 0, Somogyi P (1995) Synchronization of neuronal activity in hippocampus by individual GABAergic interneurons. Nature 378:75-98.

Cohen GA, Doze VA, Madison DV (1 992) Opioid inhibition of GABA release from presynaptic terminals of rat hippocampal interneurons. Neuron 9:325-335.

Cohen NJ, Eichenbaum H (1993) Memory, amnesia, and the hippocampal system. Cambridge, MA: MIT Press.

Colbert CM, Levy WB (1 992) Electrophysiological and pharmacological characterization of perforant path synapses in CA1: mediation by glutamate receptors. J Neurophysiol 68: 1-8.

Cole AE, Nicoll RA (1984) Characterization of a slow cholinetgic postsynaptic potential recorded in vitro from rat hippocampal pyramidal cells. J Physiol (Lond) 352:173-188.

Colino A, Halliwell JV (1 987) Differential modulation of three separate K-conductances in hippocampal CA1 neurons by serotonin. Nature

Collin C, Devane WA, Dahl D, Lee C-J, Axelrod, Alkon DL (1995) Long-term synaptic transformation of hippocampal CA1 gamm- aminobutyric acid synapses and the effect of anandamide. Proc Natl Acad Sci USA 92:10167-10171.

Colmers WF (1992) Neuropeptide Y does not alter NMDA conductances in CA3 pyramidal neurons: a slice-patch study. Neurosci Lett 138:26 1-274.

Colmers WF, Lukowiak K, Pittman QJ (1987) Presynaptic action of neuropeptide Y in area CAI of the rat hippocampal slice. J Physiol (Lond) 383:285-299.

Colmers WF, Lukowiak K, Pittman QJ (1988) Neuropeptide Y action in the rat hippocampal slice: site and mechanism of presynaptic inhibition. J Neurosci 8:3827-3837.

Colom LV, Bland BH (1987) State-dependent spike train dynamics of hippocampal formation neurons: evidence for theta-on and theta-off cells. Brain Res 422:277-286.

Colonnier M (1966) The structural design of the neocortex. In: Brain and conscious experience (Eccles JC, ed), pp 1-23. New York: S pringer-Verlag.

Colonnier M (1968) Synaptic patterns on different cell types in the different laminae of the cat visual cortex. An electron microscope study. Brain Res 9:268-287.

Congar P, Khazipov R, Ben-Ari Y ( I 995) Direct demonstration of functional disconnection by anoxia of inhibitory interneurons from excitatory inputs in rat hippocampus. J Neurophysiol 73:421-426.

(:onner JM, Muir D, Varon S, Hagg T, Manthorpe M (1992) The localization of nerve growth factor-like immunoreactivity in the adult rat basal forebrain and hippocampal formation. J Comp Neurol 31 9:454-462.

328:73-77.

Cozzari C, Howard J, Hartman B (1990) Analysis of epitopes on choline acetyltransferase ( C U T ) using monoclonal antibodies. SOC Neurosci Abstr 16(1):200.

Danos P, Frotscher M, Freund TF (1991) Non-pyramidal cells in the CA3 region of the rat hippocampus: relationships of fine structure, synaptic input and chemical characteristics. Brain Res 546: 195-202.

David S, Aguayo AJ (1981) Axonal elongation into PNS “bridges” after CNS injury in adult rats. Science 214:931-933.

Davies CH, Starkey SJ, Pozza MF, Collingridge GL (1991) GABAB autoreceptors regulate the induction of LTP. Nature 349:609-611.

Davies S, Kohler C (1985) The substance P innervation of the rat hippocampal region. Anat Embryo1 173:45-52.

Divila JC, de la Calle A, Gutierrez A, Megias M, Andreu MJ, Guirado S (1991) Distribution of neuropeptide Y (NPY) in the cerebral cortex of the lizards Psummodromus ulgirus and I’udarcis hipanicu: co-localization of NPY, somatostatin and GABA. J Comp Neurol 308:397-408.

Divila JC, Megias M, de la Calle A, Guirado S (1993) Subpopulation of GABA neurons containing somatostatin, neuropeptide Y, and parvalbumin in the dorsomedial cortex of the lizard Psummodromus al- girzls. J Cornp Neurol 336161-173.

Divila JC, Megias M, Andreu MJ, Real MA, Guiradi S (1 995) NADPH diaphorase-positive neurons in the lizard hippocampus: a distinct subpopulation of GABAergic interneurons. Hippocampus 5:60-70.

de Montigny C, Monnet FP, Fournier A, Debonnel G (1993) Sigma ligands and neuropeptide Y selectively potentiate the NMDA response in the rat CA3 dorsal hippocampus: in vivo electrophysiological studies. Clin Neuropharmacol (Suppl) 15:145A-l46A.

de Quidt ME, Emson PC (1986) Distribution of neuropeptide Y-like immunoreactivity in the rat central nervous system-11. Immuno- histochemical analysis. Neuroscience 18:545-618.

Deadwyler SA, West JR, Cotman CW, Lynch G (1975) Physiological studies of the reciprocal connections between the hippocampus and entorhinal cortex. Exp Neurol 49:35-57.

DeFelipe J (1993) Neocortical neuronal diversity: chemical heterogeneity revealed by colocalization studies of classic neurotransmitters, neuropeptides, calcium-binding proteins, and cell surface molecules. Cereb Cortex 3:273-289.

Deller T, Leranth C (1990) Synaptic connections of neuropeptide-Y (NPY) immunoreactive neurons in the hilar area of the rat hippocampus. J Comp Neurol 300:433-447.

Deller T, Nitsch R, Frotscher M (1994) Associational and commissural afferents of parvalbumin-immunoreactive neurons in the rat hippocampus: a combined immunocytochemical and PHA-L study. J Comp Neurol 350:612-622.

Deller T, Frotscher M, Nitsch R (1995a) Morphological evidence for the sprouting of inhibitory commissural fibers in response to the lesion of the excitatory entorhinal input to the rat dentate gyrus. J Neurosci 15:6868-6878.

Deller T, Nitsch R, Frotscher M (1995b) Phaseulus vulgaris-leucoagglutinin tracing of commissural fibers to the rat dentate gyrus: evidence for a previously unknown commissural projection to the outer molecular layer. J Comp Neurol 352:55-68.

Dent AJ, Galvin NJ, Stanfield BB, Cowan WM (1983) The mode of termination of the hypothalamic projection to the dentate gyrus: an EM autoradiographic study. Brain Res 258:l-10.

Dichter MA, Ayala GF (1987) Cellular mechanisms of epilepsy: a status report, Science 237:157-163.

Dinerman JL, Dawson TM, Schell MJ, Snowman A, Snyder SH (1994) Endothelial nitric oxide syiithase localized to hippocampal pyramidal cells: implications for synaptic plasticity. Proc Natl Acad Sci USA 9 1 :42 14-42 18.

Dodd J, Kelly JS (1 978) Is somatostatin an excitatory transmitter in the hippocampus? Nature 273:674.

Dodd J, Kelly JS (1981) The actions of cholecystokinin and related peptides on pyramidal neurones of the mammalian hippocampus. Brain Res 338:97-113.

458 FREUND AND %UZSAKI

Doller HJ, Weight FF (1982) Perforant pathway activation of hippocampal CAI stratum pyramidale neurons: electrophysiological evidence for a direct pathway. Brain Res 237:l-13.

Douglas RJ, Martin KAC (1991) Opening the grey box. TINS 14: 286-293.

Drake CT, Mulligan KA, Wimpey TL, Hendrickson A, Chavkin C (1 99 1) Characterization of Vicia villosa agglutinin-labeled GABAer- gic interneurons in the hippocampal formation and in acutely dissociated hippocampus. Brain Res 554:176-185.

Du J, Zhang L, Weiser M, Rudy B, McBain CJ (1996) Developmental expression and functional characterization of the potassium-channel subunit Kv3.1 b in parvalbumin-containing interneurons of the rat hippocampus. J Neurosci 16:506-518.

Dudek SM, Bear MF (1992) Homosynaptic long-term depression in area CA1 of the hippocampus and the effects of NMDA receptor blockade. Proc Natl Acad Sci USA 89:4363-4367.

Dun NJ, Dun SL, Wong RK, Forstermann U (1994) Colocalization of nitric oxide synthase and somatostatin immunoreactivity in rat dentate hilar neurons. Proc Natl Acad Sci USA 91:2955-2959.

Dunwiddie TV, Haas HL (1985) Adenosine increases synaptic facilitation in the in vitro rat hippocampus: evidence for a presynaptic site of action. J Physiol 69:365-77.

During MJ, Ryder KM, Spencer D D (1995) Hippocampal GABA trans- porter function in temporal-lobe epilepsy. Nature 376: 174-177.

Dutar P, Nicoll RA (1 988) Classification of muscarinic responses in hippocampus in terms of receptor subtypes and second-messenger systems: electrophysiological studies in vitro. J Neurosci 8:4212-4224.

Eccles JC (1 964) The physiology of synapses. Berlin: Springer-Verlag. Eckenstein F, Baughman RW (1984) Two types of cholinergic innerva-

tion in cortex, one co-localized with vasoactive intestinal polypeptide. Nature 309:153-155.

Eckenstein F, Thoenen H (1 983) Chlonergic neurons in the rat cerebral cortex demonstrated by immunohistochemical localization of choline acetyltransferase. Neurosci Lett 36:211-215.

Edwards FA, Konnerth A, Sakmann B (1990) Quanta1 analysis of inhibitory synaptic transmission in the dentate gyrus of rat hippocampal slices: a patch-clamp study. J Physiol (Lond) 430:213-249.

Esclapez M, Houser CR (1995) Somatostatin neurons are a subpopulation of GABA neurons in the rat dentate gyrus: evidence from colocalization of pre-prosomatostatin and glutamate decarboxylase messenger RNAs. Neuroscience 64:339-355.

Esclapez M, Chang DK, Houser CR (1996) Subpopulations of GABA neurons in the dentate gyrus express high levels of the alpha1 subunit of the GABA-A receptor. Hippocampus 6:225-238.

Fantie BD, Goddard GV (1982) Septa1 modulation of the population spike in the fascia dentata produced by perforant path stimulation in the rat. Brain Res 252:227-37.

Fatt P, Katz B (1952) Spontaneous subthreshold activity at motor nerve endings. J Physiol (Lond) 117:109-128.

Finsen BR, Tonder N, Augood S, Zimmer J (1992) Somatostatin and neuropeptide Y in organotypic slice cultures of the rat hippocampus: an immunocytochemical and in situ hybridization study. Neurosci- ence 47:105-113.

Fox SE (1989) Membrane potential and impedance changes in hippocampal pyramidal cells during theta rhythm. Exp Brain Res 77:283-294.

Fox SE, Ranck JB J r (1981) Electrophysiological characteristics of hippocampal complex-spike cells and theta cells. Exp Brain Res 4 1:299-3 13.

Fox SE, Ranck JB Jr (1986) Hippocampal theta rhythm and the firing of neurons in walking and urethane anesthetized rats. Exp Brain Res 62495-508.

Fraser DD, MacVicar BA (1 99 1) Low-threshold transient calcium current in rat hippocampal lacunosum-moleculare interneurons: kinetics and modulation by neurotransmitters. J Neurosci 1 1 :2812-2820.

Freedman R, Wetmore C, Stromberg I, Leonard S, Olson L (1993) Alpha-bungarotoxin binding to hippocampal interneurons: immunocytochemical characterization and effects on growth factor expression. J Neurosci 13:1965-1975.

Freund T F (1992) GABAergic septa1 and serotonergic median raphe afferents preferentially innervate inhibitory interneurons in the bip- pocampus and dentate gyrus. Epilepsy Res (Suppl) 7:79-91.

Freund TF (1993) Anterograde PHAL-tracing combined wirh pre- and postemhedding immunocytocheniistry. In: Immunohistochemistry, 2nd ed (Cuello AC ed), pp 329-348. Chichester: John Wiley & Sons.

Freund TF, A n d M (1988) GABA-containing neurons in the septum control inhibitory interneurons in the hippocampus. Nature

Freund TF, Gulyis A1 (1991) GABAergic interneurons containing calbindin D28K or somatostatin are major targets of GABAergic basal forebrain afferents in the rat neocortex. J Comp Neurol314:187-199.

Freund TF, Magl6czky 2 (1993) Early degeneration of calretinin-containing neurons in the rat hippocampus after ischemia. Neuroscience 56:581-596.

Freund TF, Meskenaite V (1992) Gamma-aminobutyric acid-containing basal forebrain neurons innervate inhibitory interneurons in the neocortex. Proc Natl Acad Sci USA 89:738-742.

Freund TF, Somogyi P (1989) Synaptic relationships of Golgi impregnated neurons as identified by electrophisiological or immunocytochemical techniques. In: Neuroanatomical tract-tracing methods 11. Recent progress (Heimer L, Zaborszky L, eds), pp 201-238. New York: Plenum.

Freund TF, Martin KA, Smith AD, Somogyi P (1983) Glutamate decarboxylase-immunoreactive terminals of Golgi-impregnated axo-axonic cells and of presumed basket cells in synaptic contact with pyramitial neurons of the car’s visual cortex. J Comp Neurol 221:263--278.

Freund TF, Buzsiki G, Leon A, Baimbridge KG, Somogyi P (1990a) Relationship of neuronal vulnerability and calcium binding protein irnmunoreactivity in ischemia. Exp Brain Res 83:55-66.

Freund TF, Buzsiki G, Leon A, Somogyi P (1990b) Hippocampal cell death following ischemia: effects of brain temperature and anesthesia. Exp Neurol 108:251-260.

Freund TF, Gulyis AI, Acsidy L, Gorcs T, T6th K (1 990c) Serotonergic control of the hippocampus via local inhibitory interneurons. Proc Natl Acad Sci USA 87:8501-8505.

Freund TF, Ylinen A, Miettinen R, Pitkanen A, Lahtinen H, Bairnbridge KG, Riekkinen PJ (1992) Pattern of neuronal death in the rat hippocampus after status epilepticus. Relationship to calcium binding protein content and ischemic vulnerability. Brain Res Bull 2827-38.

Frotscher M (1985) Mossy fibres form synapses with identified pyramidal basket cells in the CA3 region of the guinea-pig hippocampus: a combined Golgi-electron microscope study. J Neurocytol 14:245- 259.

Frotscher M (1989) Mossy fiber synapses on glutamate decarboxylase- immunoreactive neurons: evidence for feed-forward inhibition in the CA3 region of the hippocampus. Exp Brain Res 75:441-445.

Frotscher M, Leranth C (1983) Cholinergic innervation of the rat hippocampus as revealed by choline acethyltransferase immunocytochemistry:a combined light and electron microscopic study. J Comp Neurol 239:237-246.

Frotscher M, Leranth C (1986) The cholinergic innervation of the rat fascia dentata: identification of target structures on granuie cells by combining choline acetyltransferase irnmunocytochemistry and Golgi impregnation. J Comp Neurol 243:58-70.

Frotscher M, Leranth C (1988) Catecholaminergic innervation of pyramidal and GABAergic nonpyramidal neurons in the rat hippocampus. Double label immunostaining with antibodies againsr tyrosine hydroxylase and glutamate decarboxylase. Histochemistry 88:3 13- 319.

Frotscher M, Zimmer J (1983a) Commissural fibers terminate on nonpyramidal neurons in the guinea pig hippocampus-a combined Golgi/EM degeneration study. Brain Res 265:289-293.

Frotscher M, Zimmer J (1983b) Lesion-induced mossy fibers to the inner molecular layer of the rat fascia dentata: identification of postsynaptic granule cells by the Golgi-EM technique. J Comp Neurol

336: 170-173.

215:299-311.


Frotscher M, Leranth C, Lubbers K, Oertel WH (1984) Commissural afferents innervate glutamate decarboxylase immunoreactive nonpyramidal neurons h the guinea pig hippocampus. Neurosci Lett 46: 137-143.

Frotscher M, Schlander M, Leranth C (1986) Cholinergic neurons in the hippocampus. A combined light- and electron-microscopic immunocytochemical study in the rat. Cell Tissue Res 246:293-301.

Frotscher M, Seress L, Schwerdtfeger WK, Buhl E (1991) The mossy cells of the fascia dentata-a comparative study of their fine structure and synaptic connections in rodents and primates. J Comp Neurol 312:145-163.

Gahwiler B, Brown D (1987) Muscarine affects calcium-currents in rat hippocampal pyramidal cells in vitro. Neurosci Lett 76:301-306.

Gaiarsa JL, McLean H , Congar P, Leinekugel X, Khazipov R, Tseeb V, Ben-Ari Y (1995) Postnatal maturation of gamma-aminobutyric acid A and B-mediated inhibition in the CA3 hippocampal region of the rat. J Neurobiol 26:339-349.

Gall C (1988) Seizures induce dramatic and distinctly different changes in enkephalin, dynorphin, and CCK immunoreactivities in mouse hippocampal mossy fibers. J Neurosci 8:1852-1862.

Gall C, Selawski L (1984) Supramammilary afferents to guinea pig hippocampus contain substance P-like immunoreactivity. Neurosci Lett 5 1: 171-1 76.

Gall C, Brecha N, Karten HJ, Chang KJ (1981) Localization of enkephalin-like immunoreactivity to identified axonal and neuronal populations of the rat hippocampus. J Comp Neurol 198:335-350.

Gall C, Berry LM, Hodgson LA (1986) Cholecystokinin in the mouse hippocampus: localization in the mossy fiber and dentate commissural systems. Exp Brain Res 62:431437.

Gao B, Fritschy JM (1 994) Selective allocation of GAB& receptors containing the a1 subunit to neurochemically distinct subpopulations of rat hippocampal interneurons. Eur J Neurosci 6:837-853.

Garthwaite J (1 99 1) Glutamate, nitric oxide and cell-cell signalling in the nervous system. Trends Neurosci 14:60-67.

Gaykema RP, Luiten PG, Nyakas C, Traber J (1990) Cortical projection patterns of the medial septum-diagonal band complex. 293: 103- 124.

Geiger, JRP, Melcher T, Koh D-S, Sakmann B, Seeburg PH, Jonas P, Monyer H (1995) Relative abundance of subunit mRNAs determines gating and Ca2+ permeability on AMPA receptors in principal neurons and interneurons in rat CNS. Neuron 15:193-204.

Gernstein G, Mandelbrot B (1964) Random walk models for the spike activity of a single neuron. Biophys J 4:417-419.

Ghadimi EM, Jarolimek W, Misgeld U (1994) Effects of serotonin on hilar neurons and granule cell inhibition in the guinea pig hippocampal slice. Brain Res 633:27-32.

Gorcs TJ, Penke B, Boti Z, Katarova Z, Himori J (1993) Immuno histochemical visualization of a metabotropic glutamate receptor. NeuroReport 4:283-286.

Gottlieb DI, Cowan WM (1973) Autoradiographic studies of the commissural and ipsilareral associarion connecrions of rhe hippocampus and dentate gyms of the rat: I. The commissural connections. J Comp Neurol 149:393-420.

Gray CM (1994) Synchronous oscillations in neuronal systems: mechanisms and functions. J Comput Neurosci 1:ll-38.

Gray EG (1959) Axo-somatic and axo-dendritic synapses of the cerebral cortex: an electron microscope study. J Anat 93:420-433.

Greene RW, Haas HL (1985) Adenosine actions on CA1 pyramidal neurones in rat hippocampal slices. J Physiol (Lond) 366:119-127.

Greenwood RS, Godar SE, Reaves TA Jr, Hayward JN (1981) Cholecystokinin in hippocampal pathways. J Comp Neurol203:335- 350.

Grimaldi R, Zoli M, Agnati LF, Ferraguti F, Fuxe K, Toffano G, Zini I (1990) Effects of transient forebrain ischemia on peptidergic neurons and astroglial cells: evidence for recovery of peptide immunoreactivities in the neocortex and striatum but not hippocampal formation. Exp Brain Res 82:123-136.

Grover LM, Lambert NA, Schwartzkroin PA, Teyler TJ (1993) Role of HC03-ion in deporarizing GABAA receptor-mediated responses in pyramidal cells of rat hippocampus. J Neurophysiol 69:1541-1555.

Grunze HCR, Rainnie DG, Hasselmo ME, Barkai E, Hearn EF, McCarley, Greene RW (1 996) NMDA-dependent modulation of CA1 local circuit inhibition. J Neurosci 16:2034-2063.

Gulyis AI, Freund TF (in press) Calbindin-containing interneurons innervate pyramidal cell dendrires in the hippocampus. Hippocampus.

Gulyis AI, Gorcs TJ, Freund T F (1990) Innervation of different peptide-containing neurons in the hippocampus by GABAergic septa1 afferents. Neuroscience 37:3 1 4 4 .

Gulyis AI, Seress L, T6th K, Acsidy L, Antal M, Freund TF (1991a) Septa1 GABAergic neurons innervate inhibitory interneurons in the hippocampus of the macaque monkey. Neuroscience 41 :38 1-390.

Gulyis AI, T6th K, Danos P, Freund TF (1991b) Subpopulations of GABAergic neurons containing parvalbumin, calbindin D28k, and cholecystokinin in the rat hippocampus. J Comp Neurol 3 12:371- 378.

Gulyis AI, Miettinen R, Jacobowitz DM, Freund TF (1 992) Calretinin is present in non-pyramidal cells of the rat hippocampus-I. A new type of neuron specifically associated with the mossy fibre system. Neuroscience 48: 1-27.

Gulyis AI, Miles R, Hijos N, Freund TF (1993a) Precision and variability in postsynaptic target selection of inhibitory cells in the hippocampal CA3 region. Eur J Neurosci 5:1729-1751.

Gulyis AI, Miles R, Sik A, Toth K, Tamamaki N, Freund T F (1 993b) Hippocampal pyramidal cells excite inhibitory neurons through a single release site. Nature 366:683-687.

Gulyis AI, Hijos N, Freund TF (1996) Interneurons containing calretinin are specialized to control other interneurons in the rat hippocampus. J Neurosci 16:3397-3411.

Haas HL, Gahwiler BH (1992) Vasoactive intestinal polypeptide modulates neuronal excitability in hippocampal slices of the rat. Neuro- science 47:273-277.

Haas HL, Hermann A, Greene RW, Chan-Palay V (1 987) Action and location of neuropeptide tyrosine (Y) on hippocampal neurons of the rat in slice preparation. J Comp Neurol 257:208-217.

Haas HL, Rose G (1982) Long-term potentiation of excitatory synaptic transmission in the rat hippocampus: the role of inhibitory processes. J. Physiol (Lond) 329:541-552.

Haas HL, Rose G (1984) The role of inhibitory mechanisms in long- term potentiation. Neurosci Lett 47:303-308.

Haglund L, Swanson LW, Kohler C (1984) The projection of the supramammillary nucleus to the hippocampal formation: an immunohistochemical and anterograde transport study with the lectin PHA-L in the rat. J Comp Neurol 229:171-185.

Hijos N, Acsidy L, Freund TF (1996) Target selectivity and neurochemical characteristics of VIP-immunoreactive interneurons in the rat dentate gyrus. Eur J Neurosci 8:1415-1431.

Hijos N, Cs. Papp E, Acsidy L, Levey AI, Freund TF (submitted) Distinct interneuron types express m2 muscarinic receptor immunoreactivity on their dendrites and/or axons in the rat hippocampus. Neuroscience.

Halasy K, Somogyi P (1993a) Distribution of GABAergic synapses and their targets in the dentate gyrus of rat: a quantitative immunoelec- tron microscopic analysis. J Hirnforsch 34:299-308.

Halasy K, Somogyi P (1993b) Subdivisions in the multiple GABAergic innervation of granule cells in the dentate gyrus of the rat hippocampus. Eur J Neurosci 5:411-429.

Halasy K, Miettinen R, Szabat E, Freund T F (1992) GABAergic interneurons are the major postsynaptic targets of median Raphe afferents in the rat dentate gyrus. Eur J Neurosci 4:144-153.

Halasy K, Buhl EH, Lorinczi Z, Tamas G, Somogyi P (1996) Synaptic target selectivity and input of GABAergic basket and bistratified interneurons in the CA1 area of the rat hippocampus. Hippocampus 6:306-329.

Hamill OP, Huguenard JR, Prince DA (1991) Patch-clamp studies of


voltage-gated currents in identified neurons of the rat cerebral cortex. Cereb Cortex 1:48-61.

Han ZS, Buhl EH, Lorinczi Z, Somogyi P (1993) A high degree of spatial selectivity in the axonal and dendritic domains of physiologically identified local-circuit neurons in the dentate gyrus of the rat hippocampus. Eur J Neurosci 5:395-410.

Handelmann G, Meyer DK, Beinfeld MC, Oertel WH (1981) CCK- containing terminals in the hippocampus are derived from intrinsic neurons: an immunohistochemical and radioimmunological study. Brain Res 224:180-184.

Harris KM, Marshall PE, Landis DM (1985) Ultrastructural study of cholecystokinin-immunoreactive cells and processes in area CA1 of the rat hippocampus. J Comp Neurol 233:147-158.

Hasselmo ME, Bower JM (1993) Acetylcholine and memory. TINS

Heiligenberg WF (1991) Neural nets in electric fish. Cambridge, MA: MIT Press.

Hell JW, Westenbroek RE, Warner C, Ahlijanian MK, Prystay W, Gilbert MM, Snutch TP, Catterall WA (1993) Identification and differential subcellular localization of the neuronal class C and class D L-type calcium channel alpha 1 subunits. J Cell Biol 123:949-962.

Hestrin S (1 993) Different glutamate receptor channels mediate fast excitatory synaptic currents in inhibitory and excitatory cortical neurons. Neuron 11:1083-1091.

Hill JA, Zoli M, Bourgeois J-P, Chngeux J-P (1 993) Immunocytochem- ical localization of a neuronal nicotinic receptor: the alpha2-subunit. J Neurosci 13:1551-1558.

Hofer M, Pagliusi SR, Hohn A, Leibrock J , Barde YA (1990) Regional distribution of brain-derived neurotrophic factor mKNA in the adult mouse brain. EMBO J 92459-2464.

Hokfelt T, Johansson 0, Ljungdahl A, Lundberg JM, Schultzberg M (1980) Peptidergic neurones. Nature 284:5 15-522.

Hollmann M, Heinemann S (1994) Cloned glutamate receptors. Annu Rev Neurosci 17:31-108.

Hollmann M, Hartley M, Heinemann S (1991) Ca2+ permeability of KA-AMPA-gated glutamate receptor channels depends on subunit composition. Science 252235 1-853.

Hope BT, Michael GJ, Knigge KM, Vincent SR (1991) Neuronal NADPH diaphorase is a nitric oxide synthase. Proc Natl Acad Sci USA 88:2811-2814.

Hopfield JJ (1995) Pattern recognition computation using action potential timing for stimulus representation. Nature 376:33-36.

Hounsgaard J (1978) Presynaptic inhibitory action of acetylcholine in area CAI of the hippocampus. Exp Neurol 62:787-797.

Houser CR, Esclapez M (1 994) Localization of mRNAs encoding two forms of glutamic acid decarboxylase in the rat hippocampal formation. Hippocampus 4:530-545.

Houser CR, Crawford GD, Salvaterra PM, Vaughn JE (1985) Immuno- cytochemical localization of choline acetyltransferase in rat cerebral cortex: a study of cholinergic neurons and synapses. J Comp Neurol 234:17-34.

Houser CR, Olsen RW, Richard JG, Mohler H (1988) Immunocyto- chemical localization of benzodiazepine/GABAA receptors in the human hippocanipal formation. J Neurosci 8: 1370-1 383.

Hsu M, Buzsiki G (1993) Vulnerability of mossy fiber targets in the rat hippocampus to forebrain ischemia. J Neurosci 13:3964-3979.

Hsu M, Sik A, Gallyas F, Horvath Z, Buzsaki G (1994) Short-term and long-term changes in the postischemic hippocampus. Ann NY Acad Sci 743:121-140.

Huerta PT, Lisman JE (1 995) Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillation in CA1 in virro. Neuron 15: 1053-1 063.

Iritani S, Fujii M, Satoh K (1989) The distribution of substance P in the cerebral cortex and hippocampal formation: an immunohistochemical study in the monkey and rat. Brain Res Bull 22:295-303.

Isackson PJ, Huntsman MM, Murray KD, Gall CM (1991) BDNF

16~218-222.

mRNA expression is increased in adult rat forebrain after limbic seizures: temporal patterns of induction distinct from NGF. Neuron 6337-948.

Ishizuka N, Weber J, Amaral D G (1990) Organization of inrrahip- pocampal projections originating from CA3 pyramidal cells in the rat. J Comp Neurol 295:580-623.

Ishizuka N, Cowan WM, Amaral DG (1995) A quantitative analysis of the dendritic organization of pyramidal cells in the rat hippocampus. J Comp Neurol 362:17-45.

Jacobowitz DM, Winsky L (1991) Immunocytochemical localization of calretinin in the forebrain of the rat. J Comp Neurol 304:198-218.

Jaffe DB, Johnston D, Lasser-Ross N, Lisrnan JE, Miyakawa H, Ross WN (1992) The spread of Na+ spikes determines the pattern of dendritic Ca + 2 entry into hippocampal neurons. Nature 357:244-246.

Jahnsen H , Llinas RR (1984) Ionic basis for the electroresponsiveness and oscillatory properties of guinea-pig thalamic neurons in vitro. J Physiol (Lond) 349:227-247.

Johansen FF, Jorgensen MB, Diember N H (1983) Resistance of hippocampal CA-1 interneurons to 20 min of transient cerebral ischemia in the rat. Acta Neuropath 61:135-140.

Johansen FF, Zimmer J , Diemer N H (1987) Early loss of somatostatin neurons in dentate hilus after cerebral ischemia in the rat precedes CA-1 pyramidal cell loss. Acta Neuropathol 73:llO-114.

Johansen FF, Tonder N , Zimmer J , Baimbridge KG, Diemer NH (1990) Short-term changes of parvalbumin and calbindin immunoreactivity in the rat hippocampus following cerebral ischemia. Neurosci Lett 120:171-174.

Johansson 0, Hokfelt T, Elde RP (1984) Irnmunohistochernical distribution of somatostatin-like irnmunoreactivity in the central nervous system of the adult rat. Neuroscience 13:265-339.

Jonas I’, Burnashev N (1 995) Molecular mechanisms controlling calcium entry through AMPA-type glutamate recepcor channels. Curr Opin Neurobiol 15:987-990.

Jonas P, Racca C, Sakmann 8, Seeburg PH, Monyer H (1994) Differences in Ca2+ permeability of AMPA-type glutamate receptor channels in neocortical neurons caused by differential GluR-B subunit expression. Neuron 12: 128 1-1 289.

Kaneko T, Shigemoto R, Nakanishi S, Mizuno N (1 994) Morphological and chemical characteristics of subsrance P receptor-immunoreactive neurons in the rat neocortex. Neuroscience 60:199-211.

Kanter ED, Kapur A, Haberly LB (1996) A dcndritic GABAA-mediated IPSP regulates facilitation of NMDA-mediated responses to burst stimulation of afferent fibers in piriform cortex. J Neurosci 16:

Katchman AN, Vicini S, Hershkowitz N (1994) Mechanism of early anoxia-induced suppression of the GABA-A-mediated inhibitory postsynaptic current. J Neurophysiol 71 :1128-1138.

Katona I, Acsidy L, Gulicsi A, Freund TF (1996) Somatostatin-con- raining cells in the rat hippocampus:connectivity and heterogeneity. Eur J Neurosci Suppl 9:176.

Katsumaru H, Kosaka T, Heizman CW, Hama K (1988a) Gap-junctions on GABAergic neurons containing the calcium-binding protein parvalbumin in the rat hippocampus (CA1 regions). Exp Brain Res 72:363-370.

Katsumaru H, Kosaka T , Heizmann CW, Hama K (1988b) Immuno- cytochemical study of GABAergic neurons containing the calcium- binding protein parvalbumin in the rat hippocampus. Exp Brain Res 72:347-362.

Kauer JA, McMahon LL (1975) Hippocampal s. radiatum interneurons express functional serotonin receptors. SOC Neurosci Abstr 2 1 :590.

Kawa K (1994) Distribution and functional properties of 5-HT3 receptors in the rat hippocampal dentatc gyrus: a patch-clamp study. J Neurophysiol 71 :1935-1947.

Kawaguchi Y, Hama K (1987a) Fast-spiking non-pyramidal cells in the hippocampal CA3 region, dentate gyms and subicuium of rats. Brain Res 425:351-355.

307-312.


Kawaguchi Y, Hama K (1987b) Two subtypes of non-pyramidal cells in rat hippocampal formation identified by intracellular recording and HRP injection. Brain Res 411:190-195.

Kawaguchi Y, Hama K (1988) Physiological heterogeneity of nonpyramidal cells in rat hippocampal CA1 region. Exp Brain Res 72:494- 502.

Kawaguchi Y, Katsumaru H, Kosaka T, Heizmann CW, Hama K (1987) Fast spiking cells in rat hippocampus (CA1 region) contain the calcium-binding protein parvalbumin. Brain Res 416:369-374.

Khazipov R, Bregestovski P, Ben-Ari Y (1993) Hippocampal inhibitory interneurons are functionally disconnected from excitatory inputs by anoxia. J Neurophysiol 70:2251-2259.

Khazipov R, Congar P, Ben-Ari Y (1995a) Hippocampal CA1 lacunosum-moleculare i nterneurons: comparison of effects of anoxia on excitatory and inhibitory postsynaptic currents. J Neurophysiol 74: 2 138-21 49.

Khazipov R, Congar P, Ben-Ari Y (1995b) Hippocampal CAI lacunosum-moleculare interneurons: modulation of monosynaptic GABA- ergic IPSCs by presynaptic GABAb receptors. J Neurophysiol 7 4 2 126-2 137.

Kiss J, Buzdki G, Morrow JS, Glantz SB, Leranth C (1996) Entorhinal cortical innervation of parvalbumin-containing neurons (basket and chandelier cells) in the rat Ammon’s horn. Hippocampus 6:239-246.

Klancnik JM, Baimbridge KG, Philips AG (1989) Increased population spike amplitude in the dentate gyrus following systemic administra- tion of 5-hydroxytryptophan or 8-hydroxy-2-(di-n-propylamino)- tetralin. Brain Res 505: 145-148.

Klapstein GJ, Colmers WF (1993) On the sites of presynaptic inhibition by neuropeptide Y in rat hippocampus in vitro. Hippocampus

Kneisler TB, Dingledine R (1995a) Spontaneous and synaptic input from granule cells and perforant path to dentate basket cells in rat hippocampus. Hippocampus 5:151-164.

Kneisler TB, Dingledine R (1995b) Synaptic input from CA3 pyramidal cells to dentate basket cells in rat hippocampus. J Physiol (Lond) 487:125-146.

Knowles WD, Schwartzkroin PA (1981) Axonal ramifications of hippocampal CA1 pyramidal cells. J Neurosci 1:1236-1241.

Kogure S (1987) Simultaneous recordings from two types of hippocampal nonpyramidal cells during electrically induced paroxysmal discharges. Exp Neurol 95:763-767.

Koh D-S, Geiger JRP, Jonas P, Sakmann B (1995) Ca2+-permeable AMPA and NMDA receptor channels in basket cells of rat hippocampal dentate gyrus. J Physiol (Lond) 485:383-402.

Kohler C (1982) Distribution and morphology of vasoactive intestinal polypeptide-like immunoreactive neurons in regio superior of the rat hippocampal formation. Neurosci Lett 33:265-270.

Kohler C (1983) Morphological analysis ofvasoactive inrestinal polypeptide (VIP)-like immunoreactive neurons in the area dentata of the rat brain. J Comp Neurol 221:247-262.

Kohler C, Chan-Palay V (1 982) Somatostatin-like immunoreactive neurons in the hippocampus: an immunohistochemical study in the rat. Neurosci Lett 34:259-264.

Kohler C, Steinbusch HWM (1982) Identificatios of serotonin and non- serotonin-containing neurons of the mid-brain raphe projecting to the entorhinal area and the hippocampal formation. A combined immunohistochemical and fluorescent retrograde tracing study in the rat brain. Neuroscience 7:951-975.

Kohler C, Eriksson L, Davies S, Chan-Palay V (1986) Neuropeptide Y innervation of the hippocampal region in the rat and monkey brain. J Comp Neurol 244:384-400.

Kohler C, Eriksson LG, Davies S, Chan-Palay V (1987) Co-localization of neuropeptide tyrosine and somatostatin immunoreactivity in neurons of individual subfields of the rat hippocampal region. Neurosci Lett 78: 1-6.

Konopacki J, MacIver MB, Bland BH, Roth SH (1987) Theta in hip-

3:103-112.

pocampal slices: relation to synaptic responses of dentate neurons. Brain Res Bull 18:25-27.

Kosaka T (1983a) Axon initial segments of the granule cell in the rat dentate gyrus: synaptic contacts on bundles of axon initial segments. Brain Res 274:129-134.

Kosaka T (1983b) Gap junctions between non-pyramidal cell dendrites in the rat hippocampus (CA1 and CA3 regions). Brain Res 271:157- 161.

Kosaka T (1983~) Neuronal gap junctions in the polymorph layer of the rat dentate gyrus. Brain Res 277347-351.

Kosaka T, Hama K (1985) Gap junctions between non-pyramidal cell dendrites in the rat hippocampus (CA1 and CA3 regions): a combined Golgi-electron microscopy study. J Comp Neurol 231:150-161.

Kosaka T, Heizmann C W (1989) Selective staining of a population of parvalbumin-containing GABAergic neurons in the rat cerebral cortex by lectins with specific affinity for terminal N-acetylgalactosamine. Brain Res 483:158-163.

Kosaka T, Kosaka K, Tateishi K, Hamaoka Y, Yanaihara N, Wu JY, Hama K (1 985) GABAergic neurons containing CCK-8-like and/or VIP-like immunoreactivities in the rat hippocampus and dentate gyrus. J Comp Neurol 239:420-430.

Kosaka T, Katsumaru H, Hama K, Wu JY, Heizmann C W (1987) GABAergic neurons containing the Ca2+-binding protein parvalbumin in the rat hippocampus and dentate gyrus. Brain Res 419:119-130.

Kosaka T, Tauchi M, Dahl JL (1 988a) Cholinergic neurons containing GABA-like and/or glutamic acid decarboxylase-like immunoreactivities in various brain regions of the rat. Exp Brain Res 70:605-617.

Kosaka T, Wu JY, Benoit R (1988b) GABAergic neurons containing somatostatin-like immunoreactivity in the rat hippocampus and dentate gyrus. Exp Brain Res 71:388-398.

Kosofsky BE, Molliver ME (1987) The serotoninergic innervation of cerebral cortex: different classes of axon terminals arise from dorsal and medial raphe nuclei. Synapse 1: 153-168.

Krnjevic K, Schwartz S (1 967) The action of gamma-aminobutyric acid on cortical neurons. Exp Brain Res 3:320-336.

Krnjevic K, Pumain R, Renaud L (1971) The mechanism of excitation by acetylcholine in the cerebral cortex. J Physiol (Lond) 215:247-68.

Krnjevic K, Ropert N, Casullo J (1988) Septohippocampal disinhibition. Brain Res 438:182-192.

Kuffler SW (1960) Excitation and inhibition in single nerve cells. Harvey Lect 54:176.

Kunkel DD, Schwartzkroin PA (1 988) Ultrastructural characterization and GAD co-localization of somatostatin-like immunoreactive neurons in CA1 of rabbit hippocampus. Synapse 2:371-381.

Kunkel DD, Lacaille JC, Schwartzkroin PA (1 988) Ultrastructure of stratum lacunosum-moleculare interneurons of hippocampal CA1 region. Synapse 2:382-394.

Lacaille J-C (1 991) Postsynaptic potentials mediated by excitatory and inhibitory amino acids in interneurons of stratum pyramidale of the CA1 region of rat hippocampus slices in vitro. J Neurophysiol 66:1441-1454.

Lacaille J-C, Schwartzkroin PA (1988a) Stratum lacunosum-moleculare interneurons of hippocampal CAI region. I. Intracellular response characteristics, synaptic responses, and morphology. J Neurosci 8: 1400-14 10.

Lacaille J-C, Schwartzkroin PA (1 988b) Stratum laculosum-moleculare interneurons of hippocampal CA1 region. 11. intrasomatic and intradendritic recordings of local circuit synaptic interactions. J Neurosci 8:1411-1424.

Lacaille J-C, Williams S (1990) Membrane properties of interneurons in stratum oriens-alveus of the CAI region of rat hippocampus in vitro. Neuroscience 36:349-359.

Lacaille J-C, Mueller A, Kunkel DD, Schwartzkroin PA (1987) Local circuit interactions between oriens/alveus interneurons and CA1 pyramidal cells in hippocampal slices: electrophysiology and morphology. J Neurosci 7:1979-1993.


Lambert N, Grover L (1995) The mechanism of biphasic GABA responses. Science 269:928-929.

Lambert NA, Wilson WA (1993) Heterogeneity in presynaptic regulation of GABA release from hippocampal inhibitory neurons. Neuron

Lambert NA, Borroni AM, Grover LM, Teyler TJ (1991) Hyperpolar- izing and depolarizing GABAA receptor-mediated dendric inhibition in area CAI of the rat hippocampus. J Neurophysiol66:1538-1548.

Laurberg S (1979) Commissural and intrinsic connections of the rat hippocampus. J Comp Neurol 184:685-708.

Laurberg S, Sorensen KE (1 98 1) Associational and commissural collaterals of neurons in the hippocampal formation (hilus fasciae dentatae and subfield CA3). Brain Res 212:287-300.

Laurerborn IC, Tran TMD, Isackson PJ, Gall CM (1 993) Nerve growth factor mRNA is expressed by GABAergic neurons in rat hippocampus. NeuroReport 5:273-276.

Lee MG, Chrobak JJ, Sik A, Wiley RG, Buzsriki G (1994) Hippocampal theta activity following selective lesion of the septa1 cholinergic system. Neuroscience 62:1033-1047.

Leranth C, Frotscher M (1 986) Synaptic connections of cholecystokinin- immunoreactive neurons and terminals in the rat fascia dentata: a combined light and electron microscopic study. J Comp Neurol 254:51-64.

Leranth C, Frotscher M (1987) Cholinegic innervation of hippocampal GAD- and somatostatin-immunoreactive commissural neurons. J Comp Neurol 261:33-47.

Leranth C, Ribak CE (1991) Calcium-binding proteins are concentrated in the CA2 field of the monkey hippocampus: a possible key to this region’s resistance to epileptic damage. Exp Brain Res 85: 129-136.

Leranth C, Frotscher M, Tombol T, Palkovits M (1984) Ultrastructure and synaptic connections of vasoactive intestinal polypeptide-like immunoreactive non-pyramidal neurons and axon terminals in the rat hippocampus. Neuroscience 12:53 1-542.

Leranth C, Malcolm AJ, Frotscher M (1 990) Afferent and efferent synaptic connections of somatostatin-immunoreactive neurons in the rat fascia dentata. J Comp Neurol 295:lll-122.

Leranth C, Szeidemann Z, Hsu M, Buzsiki G (1996) AMPA receptors in the rat and primate hippocampus: a possible absence of Glur213 subunits in most interneurons. Neuroscience 70631-652.

Leung L-WS, Yim C-YC (1991) Intrinsic membrane potential oscillations in hippocampal neurons in vitro. Brain Res 553:261-274.

Leung LS (1992) Fast (beta) rhythms in the hippocampus: a review. Hippocampus 293-98.

Levey AI, Wainer BH, Rye DB, Mufson EJ, Mesulam MM (1984) Choline acetyltransferase-immunoreactive neurons intrinsic to rodent cortex and distinction from acetylcholinesterase-positive neurons. Neuroscience 13:34 1-353.

Levey AI, Edmunds SM, Koliatsos V, Wiley RG, Heilman CJ (1995) Expression of ml-m4 muscarinic acetylcholine receptor proteins in rat hippocampus and regulation by cholinergic innervation. J Neurosci 15:4077-4092.

Li X-G, Somogyi P, Tepper JM, Buzsriki G (1992) Axonal and dendritic arborization of an intracellularly labeled chandelier cell in the CA1 region of rat hippocampus. Exp Brain Res 90:519-525.

Li X-G, Somogyi P, Minen A, Buzsiki G (1994) The hippocampal CA3 network: an in vivo intracellular labeling study. J Comp Neurol

Lidov HG, Grzanna R, Molliver ME (1980) The seroronin innervation of the cerebral cortex in the rat-an immunohistochemical analysis. Neuroscience 5:207-227.

Lisman JE, Harris KM (1993) Quanta1 analysis and synapric anatomy- integrating two views of hippocampai plasticity. TINS 16: 141-147.

Lisman JE, Idiart MAP (1995) A mechanism for storing 7 -+ 2 short- term memories in oscillatory subcycles. Science 267:1512-1514.

Liu Y, Fujise N, Kosaka T (1996) Distribution of calretinin immunoreactivity in the mouse dentate gyrus. Exp Brain Res 108:389- 403.

11~1057-1067.

339: 18 1-208.

Liu Y-B, Disterhoft JF, Slater N T (1993) Activation of metabotropic glutamate receptors induces long-term depression of GABAergic inhibition in the hippocampus. J Neurophysiol 69: 100-104.

Livsey CT, Vicini S (1 992) Slower spontaneous excitatory postsynaptic currents in spiny versus aspiny hilar neurons. Neuron 8:745-755.

Livsey CT, Costa E, Vicini S (1 993) Glutamate-activated currents in outside-out patches from spiny versus aspiny hilar neurons of rat hippocampal slices. J Neurosci 135324-5333.

Ljungdahl A, Hokfelt T, Nilsson G (1978) Distribution of substance 1’- lie immunoreactivity in the central nervous system of the rat--1. cell bodies and nerve terminals. Neuroscience 3:861-943.

Llinas RR ( I 988) The intrinsic electrophysiological properties of mammalian neurons: insight into central nervous system function. Science 242: 1654-1664.

Llinas RR, Nicholson C (1971) Electrophysiological properties of dendrites and somata in alligator Purkinje cells. J Neurophysiol 34:532-551.

Llinas R, Ribary U, Joliot M, Wang X-J (1994) Content and context in temporal thalamocortical binding (Buzsaki G, Llinas RR, Singer W, Berthoz A, Christen Y, eds), Temporal coding in the brain, pp 25 1- 272. Berlin: Springer-Verlag.

Lloyd DPC (1946) Facilitation and inhibition of spinal motoneurons. J Neurophysiol 9:42 1.

Lohman AHM, Van Woerden-Verkley I (1976) Further studies on the cortical connections of the Tegu lizard. Brain Res 103:9-28.

Lopes da Silva FH, Witter MP, Boeijinga PH, Lohman AHM (1990) Anatomic organization and physiology of the limbic cortex. Physiol Rev 70:453-5 1 1.

Lopez-Garcia C, Marinez-Guijarro FJ (1 988) Neurons in the medial cortex give rise to Timm-positive boutons in the cerebral cortex of lizards. Brain Res 463:205-217.

Loren I, Emson PC, Fahrenkrug J, Bjorklund A, Alumets J, Hakanson R, Sundler F (1 979) Distribution of vasoactive intestinal polypeptide in the rat and mouse brain. Neuroscience 4:1953-1976.

Lorente de N6 R (1934) Studies on the structure of the cerebral cortex-11. Continuation of rhe study of the ammonic system. J Psycho1 Neurol 46:113-177.

Loy R, Koziell DA, Lindsey JD, Moore RY (1980) Noradrenergic innervation of the adult rat hippocampal formation. J Comp Neurol

Lujan R, Nusser Z, Roberts JDB, Shigemoto R, Somogyi P (1996) Perisynaptic location of metabotropic glutamate receptors mGluRl and mGluR5 on dendrites and dendriric spines in the rat hippocampus. Eur J Neurosci 8: 1488-1 500.

Lytton WW, Sejnowski TJ (1991) Simulations ofcortical pyramidal neurons synchronized by inhibitory interneurons. J Neurophysiol 66: 1059-1079.

Maccaferri G, McBain CJ (1995) Passive propagation of LTD ro stratum oriens-alveus inhibitory neurons modulates the temporoam- monic input to the hippocamapl CA1 region. Neuron 15:137-145.

Maccaferri G, McBain JC (1996) Long-term potentiation in distinct subtypes of hippocampal non-pyramidal neurons. J Neurosci 165334- 5343.

Maccaferri G, McBain CJ (in press) The hyperpolarization-activated current (Ih) contributes to pacemaker activity in rat CA1 hippocampal st. oriens/alveus inrerneurones. J Physiol (Lond) (in press).

MacDermott AB, Mayer ML, Westbrook GL, Smith SJ, Barker JL (1 986) NMDA-receptor activation increases cytoplasmic calcium concentration in cultured spinal cord neurones. Nature 321:5 19-522.

MacVicar BA, Tse FWY (1989) Local neuronal circuitry underlying cholinergic rhyrhmic slow activity in CA3 area of rat hippocampal slices. J Physiol (Lond) 417:197-212.

Madison DV, Nicoll RA (1 988) Enkephalin hyperpolarizes interneu- Tones in the rat hippocampus. J Physiol (Lond) 398:123-130.

Madison DV, Lancaster B, Nicoll RA ( I 987) Voltage clamp analysis of cholinergic action in the hippocampus. J Neurosci 7:733-741.

Magee JC, Johnston D (1995a) Characterization of single voltage-gated

189~699-710.

INTERNEURONS OF THE HIpPocAMpus 463

Na+ and Ca2' channels in apical dendrites of rat CA1 pyramidal neurons. J Physiol 487:67-90.

Magee JC, Johnston D (1995b) Synaptic activation of voltage-gated channels in the dendrites of hippocampal pyramidal neurons. Science 268:30 1-304.

Magistretti PJ (1990) VIP neurons in the cerebral cortex. Trends Physiol Sci 11:250-254.

Magl6czky Z, Freund TF (1993) Selective neuronal death in the contralateral hippocampus following unilateral kainate injections into the CA3 subfield. Neuroscience 56:317-336.

Maglbczky Zs, Freund TF (1995) Delayed cell death in the contralateral hippocampus following kainate injections into the CA3 subfield. Neuroscience 66:847-860.

Maglbczky 2, Acsidy L, Freund T F (1994) Principal cells are the postsynaptic targets of supramammillary afferents in the hippocampus of the rat. Hippocampus 4:322-334.

Malinow R, Miller JP ( I 986) Postsynaptic hyperpolarization during conditioning reversibly blocks induction of long-term potentiation. Nature 320:529-530.

Mmzoni OJ, Manabe T, Nicoll RA (1994) Release of adenosine by activation of NMDA receptors in the hippocampus. Science

Marksteiner J, Wahler R, Bellmann R, Order M, Krause JE, Sperk G (1992) Limbic seizures cause pronounced changes in the expression of neurokinin B in the hippocampus of the rat. Neuroscience

Martin LJ, Blackstone CD, Huganir RL, Price DL (1992) Cellular localization of a metabotropic glutamate receptor in rat brain. Neuron

Martin LJ, Blackstone CD, Levey AI, Huganir RL, Price DL (1993) AMPA glutamate receptor subunits are differentially distributed in rat brain. Neuroscience 53:327-358.

Martinez-Guijarro FJ, Freund TF (1 992a) Distribution of GABAergic interneurons immunoreactive for calretinin, calbindin D28k, and parvalbumin in the cerebral cortex of the lizard Podarcis hispanica. J Comp Neurol 322:449-460.

Martinez-Guijarro FJ, Freund TF (1992b) GABA-immunoreactive basal forebrain afferents innervate GABA-immunoreactive non-pyramidal cells in the cerebral cortex of the lizard Podarcis hispanica. Neuroscience 5 1 :425-437.

Martinez-Guijarro FJ, Desfilis E, Lopez-Garcia C (1 990) Organization of the dorsomedial cortex in the lizard podarcis hispanica. In: The forebrain in non-mammals. New aspects of structure and development (Schwerdtfeger WK, Germroth P, eds), pp 77-92. Berlin: Springer-Verlag.

Martinez-Guijarro FJ, Soriano E, Del Rio JA, Lopez-Garcia C (1991) Parvalbumin-immunoreactive neurons in the cerebral cortex of the lizard Podarcis hispanica. Brain Res 547:339-343.

Martinez-Guijarro FJ, Soriano E, Del Rio JA, Blasco-Ibanez JM, Lopez- Garcia C (1 993) Parvalbumin-containing neurons in the cerebral cortex of the lizard Podarcis hispanica: morphology, ultrastructure, and coexistence with GABA, somatostatin, and neuropeptide Y. J Comp Neurol 336:447-467.

Martinez-Guijarro FJ, Blasco-Ibanez JM, Freund TF (1994) Serotonin- ergic innervation of nonprincipal cells in the cerebral cortex of the lizard podarcis hispanica. J Comp Neurol 343:542-553.

Marty A, Llano I (1995) Modulation of inhibitory synapses in the mammalian brain. Curr Opin Neurobiol 535-341.

Marty S, Berninger B, Carrol P, Thoenen H (1996) GABAergic stimulation regulates the phenotype of hippocampal interneurons through the regulation of brain-derived neurotrophic factor. Neuron 16:

Mayer JH, Steffensen SC, Henriksen SJ (1994) Electrophysiological effects of selective opioid agonisrs on spontaneous and evoked neuronal activity in the dentate gyrus of the hippocampus in vivo. Neurophar- macology 33:963-975.

McBain C, Dingledine R (1992) Dual-component miniature excitatory

265~2098-2 10 1.

49:383-395.

9~259-270.

565-570.

synaptic currents in rat hippocampal CA3 pyramidal neurons. J Neurophysiol 68: 16-27.

McBain CJ ( I 994) Hippocampal inhibitory neuron activity in the ele- vated potassium model of epilepsy. J Neurophysiol 72: 1-1 1 .

McBain CJ, DiChiara TJ, Kauer JA (1994) Activation of metabotropic glutamate receptors differentially affects two classes of hippocampal interneurons and potentiates excitatory synaptic transmission. J Neurosci 14:4433-4445.

McBain CJ, Dingledine R (1993) Heterogeneity of synaptic glutamace receptors on CA3 stratum radiatum interneurones of rat hippocampus. J Physiol 462373-392.

McCormick DA, Pape H-C (1 990) Properties of a hyperpolarization-activated cation current and its role in rhythmic oscillation in thalamic relay neurones. J Physiol (Lond) 431:291-318.

McMahon LL, Kauer JA (1995) Tetanus-induced LTP in hippocampal pyramidal cells is accompanied by synaptic depression in neighbor- ing interneurons. SOC Neurosci Abstr 21: 1102.

McNamara J O (1994) Cellular and molecular basis of epilepsy. J Neurosci 14:34 13-3425.

McNaughton BL, Douglas RM, Goddard GV (1978) Synaptic enhancement in fascia dentata: cooperativity among coative afferents. Brain Res 157:277-293.

McQuiston AR, Petrozzino JJ, Connor JA, Colmers WF (1996) Neuropeptide Y1 receptors inhibit N-type calcium currents and reduce transient calcium increases in rat dentate granule cells. J Neurosci 15: 1422-1429.

Medina I, Fillipova Barbin G, Ben-Ari Y, Bregestovski P (1994) Kainate- induced inactivation of NMDA currents via an elevation of intracellular Ca2+ in hippocampal neurons. J Neurophysiol 72456465 .

Michelson HB, Wong RKS (1991) Excitatory synaptic responses mediated by GABA receptors in the hippocampus. Science 253:1420- 1423.

Michelson HB, Wong RKS (1994) Synchronization of inhibitory neurones in the guinea-pig hippocampus in vitro. J Physiol 4 7 7 3 4 5 .

Miettinen R, Freund TF (1992a) Convergence and segregation of septal and median raphe inputs onto different subsets of hippocampal inhibitory interneurons. Brain Res 594263-272.

Miettinen R, Freund T F (1992b) Neuropeptide-Y-containing interneurons in the hippocampus receive synaptic input from median raphe and GABAergic septal afferents. Neuropeptides 22: 185-1 93.

Miettinen R, Gulyis AI, Baimbridge KG, Jacobowitz DM, Freund TF (1992) Calretinin is present in non-pyramidal cells of the rat hippocampus-11. Co-existence with other calcium binding proteins and GABA. Neuroscience 4829-43.

Migaud M, Roques BP, Durieux C (1994) Effects of cholecystokinin octapeptide and BC 264, a potent and selective CCK-B agonist on aspartate and glutamate release from rat hippocampal slices. Neuropharmacology 33:737-743.

Miles R (1 99 1) Tetanic stimuli induce a short-term enhancement of recurrent inhibition in the CA3 region of guinea-pig hippocampus in vitro. J Physiol (Lond) 443:669-682.

Miles R, Poncer J-C (1993) Metabotropic glutamate receptors mediate a post-tetanic excitation of guinea-pig hippocampal inhibitory neurones. J Physiol (Lond) 463:461-473.

Miles R, Wong RKS (1987a) Inhibitory control of local excitatory circuits in the guinea-pig hippocampus. J Physiol (Lond) 388:611-629.

Miles R, Wong RKS (1 787b) Latent synaptic pathways revealed after tetanic stimulation in the hippocampus. Nature 329:724-726.

Miles R, Tbth K, Gulyis AI, Hijos N, Freund TF (1996) Differences between somatic and dendritic inhibition in the hippocampus. Neuron 16415-823.

Milner TA, Bacon CE (1989a) GABAergic neurons in the rat hippocampal formation: ultrastructure and synaptic relationships with catecholaminergic terminals. J Neurosci 9 3 4 10-3427.

Milner TA, Bacon CE (1989b) Ultrastructural localization of somatostatin-like immunoreactivity in the rat dentate gyrus. J Comp Neurol 290:544-560.

464 FREUND AND SUZSAKl

Milner TA, Veznedaroglu E (1992) Ultrastructural localization of neuropeptide-Y-like immunoreactivity in the rat hippocampal formation. Hippocampus 2: 107-1 26.

Misgeld U, Frotscher M (1 986) Postsynaptic-GABAergic inhibition of non-pyramidal neurons in the guinea-pig hippocampus. Neuroscience

Misgeld U, Nee MR (1984) Long-term potentiation and inhibition in CA3 neurons. Exp Brain Res 9:325-332.

Misgeld U, Muller W, Brunner H (1989) Effects of (-)badofen on inhibitory neurons in the guinea pig hippocampal slice. Eur J Physiol 4 14: 139-1 44.

Mizumori SJY, Barnes CA, McNaughton BL ( I 990) Behavioral correlates of theta-on and theta-off cells recorded from hippocampal formation of mature young and aged rats. Exp Brain Res 80:365-373.

Mizumori SJY, Barnes CA, McNaughton BL (1992) Differential effects

19: 193-206.

- of age on subpopulations of hippocampal theta cells. Neurobiol Aging 13:673-679.

Mody I, Heinemann U (1987) NMDA receptors of dentate gyrus granule cells participate in synaptic transmission following kindling. Nature 326:701-704.

Mody I , De Koninck Y, Otis TS, SoltCsz I (1994) Bridging the cleft at GABA synapses in the brain. TINS 17:517-525.

Molnar E, Baude A, Richmond SA, Pate1 PB, Somogyi P, McIlhinney RA (1 993) Biochemical and immunocytochemical characterization of antipeptide antibodies to a cloned GluRl glutamate receptor subunit: cellular and subcellular distribution in the rat forebrain. Neuroscience 53:307-326.

Monnet FP, Fournier A, Debonnel G, de Montigny C (1992) Neuropeptide Y potentiates selectively the N-methyl-D-aspartate response in the rat CA3 dorsal hippocampus. I. Involvement of an atypical neuropeptide Y receptor. J Pharmacol Exp Ther 263: 1212- 1218.

Morales M, Battenberg E, de Lecea L, Sanna P, Bloom FE (1994) Development and characterization of antibodies against the 5TH3 receptor. SOC Neurosci Abstr 20:1156.

Morilak DA, Garlow SJ, Ciaranello RD (1993) Immunocytochemical localization and description of neurons expressing serotonin2 receptors in the rat brain. Neuroscience 54:701-717.

Morrison JH, Molliver ME, Grzanna R (1979) Noradrenergic innervation of cerebral cortex: widespread effects of local cortical lesions. Science 205:313-316.

Morrison JH, Benoit R, Magistretti PJ, Ling N, Bloom FE (1982) Immunohistochemical distribution of pro-somatostatin-related peptides in hippocampus. Neurosci Lett 34: 137-142.

Moser EI (1996) Altered inhibition of dentate granule cells during spatial learning and an exploration task. J Neurosci 16:1247-1259.

Motr DD, Lewis DV (1991) Facilitation of the induction of long-term potentiation by GABA-B receptors. Science 252:1718-1720.

Moudy AM, Kunkel DD, Schwartzkroin PA (1993) Development of dopamine-beta-hydroxylase-positive fiber innervation of the rat hippocampus. Synapse 15:307-318.

Mudrich LA, Baimbridge KG (1989) Long-term structural changes in the rat hippocampal formation following cerebral ischemia. Brain Res 493: 179-1 84.

Mulkey KM, Malenka RC (1992) Mechanisms underlying induction of homosynaptic long-term depression in area CAI of the hippocampus. Neuron 9:967-975.

Muller W, Misgeld U (199 1) Picrotoxin- and 4-aminopyridine-induced activity in hilar neurons in the guinea pig hippocampal slice. J Neurophysiol 65:141-147.

Murthy VN, Fetz EE (1992) Coherent 25- to 3 5 - H ~ oscillations in the sensorimotor cortex of awake behaving monkeys. Proc Natl Acad Sci USA 89:5670-5674.

Naegele JR, Katz LC (1990) Cell surface molecules containing N-acetylgalactosamine are associated with basket cells and neurogliaform cells in cat visual cortex. J Neurosci 10:540-557.

Naegele JR, Arimatsu Y, Schwartz P, Barnstable CJ (1988) Selective

staining of a subset of GABAergic neurons in cat visual cortex by monoclonal antibody VCI. 1. J Neurosci 8:79-89.

Nafstad PH (1967) An electron microscope study on the termination of the perforant path fibers in the hippocampus and the fascia dentata. Zellforsch Mikrosk Anat 76:532-542.

Nakagawa F, Schulte BA, Wu JY, Spicer SS (1986) GABAergic neurons of rodent brain correspond partially with those staining for glyco- conjugate with terminal N-acetylgalactosamine. J Neurocytol 15:

Nakaya Y, Kaneko T, Shigemoto R, Nakanishi S, Mizuno N (1994) Immunohistochemical localization of substance P receptor in the central nervous system of the adult rat. J Comp Neurol 347:249-274.

Nawa H, Carnahan J, Gall C ( I 995) BDNF protein measured by a novel enzyme immunoassay in normal brain and after seizure: partial dis- agreement with mRNA levels. Eur J Neurosci 7: 1527-1535.

Newberry NR, Nicoll RA (1985) Comparison of the action of haclophen with gamma-aminobutyric acid on rat hippocampal pyramidal cells in vitro. J Physiol (Lond) 360:161-181.

Nicoll RA, Alger BE, Jahr CE (1980) Enkephalin blocks inhibitory pathways in the vertebrate CNS. Nature 287:22-25.

Nitatori T, Sato N, Waguri S, Karasawa Y, Araki H, Shibanai K, Kominami E, Uchiyama Y (1995) Delayed neuronal death in the CA1 pyramidal cell layer of the gerbil hippocampus following transient ischemia is apoptosis. J Neurosci 15:lOOl-1011.

Nitsch R, Leranth C (1991) Neuropeptide Y (NPY)-immunoreactive neurons in the primate fascia dentata; occasional coexistence with calcium-binding proteins: a light and electron microscopic study. J Comp Neurol 309:430-444.

Nitsch R, Leranth C (1993) Calretinin immunoreactivity in the monkey hippocampal formation. 11: Intrinsic GABAergic and hypothala-- mic nonGABAergic systems. An experimental tracing and coexistence study. Neuroscience 55:797-812.

Nitsch R, Leranth C (1994) Substance P-containing hypothalamic af-. ferents to the monkey hippocampus. An immunocytochemical, trac-. ing and coexistence study. Exp Brain Res 101:231-240.

Nitsch R, Leranth C (1996) GABAergic neurons in the rat dentate gyrus are innervated by subcortical calretinin-containing afferents. J Comp Neurol 364:425-438.

Nitsch R, Ohm TG (1995) Calretinin immunoreactive structures in the human hippocampal-formation. J Comp Neurol 360:475-487.

Nitsch C, Goping G, Klatzo I (1989a) Preservation of GABAergic perikarya and boutons after transient ischemia in the gerbil hippocampal CA1 field. Brain Res 495:243-252.

Nitsch C, Scotti A, Sommacal A, Kalt G (1989b) GABAergic hippocampal neurons resistant to ischemia-induced neuronal death contain the Ca2+-binding protein parvalbumin. Neurosci. Lett 105: 263-268.

Nitsch R, Leranth C, Frotscher M (1990a) Most somatosratin-immunoreactive neurons in the rat fascia dentata do not contain the calcium-binding protein parvalbumin. Brain Res 528:327-329.

Nitsch R, Soriano E, Frotscher M (1990b) The parvalbumin-containing nonpyramidal neurons in the rat hippocampus. Anat Embryo1 18 1 :4 13-425.

Nunzi MG, Gorio A, Milan F, Freund TF, Somogyi P, Smith AD (1985) Cholecystokinin-immunoreactive cells form symmetrical synaptic contacts with pyramidal and nonpyramidal neurons in the hippocampus. J Comp Neurol 237:485-505.

Nusser Z, Roberts DB, Baude A, Richards JG, Sieghart W, Somogyi I’ (1995) Immunocytochemical localization of the a1 and b2/3 subunits of the GABAA receptor in relation to specific GABAergic synases in the dentate gyrus. Eur J Neurosci 7:630-646.

Nyakas C, Luiten PGM, Spencer DG, Traber J (1987) Detailed projection patterns of septa1 and diagonal band efferents to the hippocampus in the rat with emphasis on innervation of CAI and dentate gyrus. Brain Res Bull 18:533-545.

O’Keefe J, Nadel L (1978) The hippocampus as a cognitive map. Oxford: Clarendon.

389-396.


O’Keefe J, Recce ML (1993) Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 3:317-330.

Oleskevich S , Lacaille JC (1992) Reduction of GABAB inhibitory post- syaptic potentials by serotonin via presynaptic and postsynaptic mechanisms in CA3 pyramidal cells of rat hippocampus in vitro. Synapse 12: 173-1 88.

Oleskevich S, Descarries L, Lacaille JC (1989) Quantified distribution of the noradrenaline innervation in the hippocampus of adult rat. J Neurosci 9:3803-3815.

Oleskevich S, Descarries L, Watkins KC, Seguela P, Daszuta A (1991) Ultrastructural features of the serotonin innervation in adult rat hippocampus-an immunocpochemical description in single and serial thin sections. Neuroscience 42:777-791.

Olney JW (1978) Neurotoxicity of excitatory amino acids. In: Kainic acid as a tool in neurobiology (McGeer E, Olney JW, McGeer P, eds), pp 201-217. New York: Raven Press.

Otis TS, De Koninck Y, Mody I (1994) Lasting potentiation of inhibition is associated with increased number of g-aminobutyric acid typ A receptors activated during miniature inhibitory postsynaptic currents. Proc Natl Acad Sci USA 91:7698-7702.

Ouardouz M, Lacaille J-C (1 995) Mechanisms of selective long-term potentiation of EPSCs in interneurons of stratum oriens in rat hippocampal slices. J Neurophysiol 73: 8 10-8 19.

Paakkari I, Lindsberg P (1995) Nitric oxide in the central nervous system. Ann Med 27:369-377.

Pang K, Rose GM (1987) Differential effects of norepinephrine on hippocampal complex-spike and theta-neurons. Brain Res 425: 146- 158.

Pang K, Rose GM (1989) Differential effects of methionin5-enkephalin on hippocampal pyramidal cells and interneurons. Neuropharmacol-

Panula P, Pirvola U, Auvinen S, Airaksinen MS (1989) Histamine-immunoreactive nerve fibers in the rat brain. Neuroscience 28:585-610.

Pearce RA (1 993) Physiological evidence for two distinct GABAA ce- sponses in rat hippocampus. Neuron 10: 189-200.

Penttonen M, Bragin A, Sik A, Buzsa’ki G (1995) Dye coupling of neurons in the hippocampus implies a role for gap junctions in epilepsy. SOC Neurosci Abstr 21:1971.

Perez-Velazquez JL, Valiante TA, Carlen PL (1994) Modulation of gap junctional mechanisms during calcium-free induced field burst activity: a possible role for electrotonic coupling in epileptogenesis. J Neurosci 144308-4317.

Perkel DH, Mulloney B (1974) Motor pattern production in reciprocally inhibitory neurons exhibiting postinhibitory rebound. Science 185:181-183.

Perney TM, Marshall J, Martin KA, Hockfield S, Kanmarek LK (1992) Expression of the mRNA for the Kv3.1 potassium channel gene in the adult and developing rat brain. J Neurophysiol 68:756-766.

Perouansky M, Yaari Y (1993) Kinetic properties of NMDA receptor- mediated synaptic currents in rat hippocampal cells versus interneurones. J Physiol (Lond) 465:223-244.

Perreault P, Avoli M (1989) Effects of low concentrations of 4-aminopyridine on CA1 pyramidal cells of the hippocampus. J Neurophysiol 6 1 : 95 3-970.

Petralia RS, Wenthold RJ (1992) Light and electron immunocytochemical localization of AMPA-selective glutamate receptors in the rat brain. J Comp Neurol 318:329-354.

Petsche H, Stumpf C, Gogolak G (1962) The significance of the rabbit’s septum as a relay station between the mid-brain and the hippocampus. The control of hippocampal arousal activity by septum cells. Electroencephalogr Clin Neurophysiol 14:202-2 11.

Piguet P, Galvan M (1 994) Transient and long-lasting actions of 5-HT on rat dentate gyrus neurones in vitro. J Physiology 481:629-639.

Pitkanen A, Amaral DG (1993) Distribution of parvalbumin-immunoreactive cells and fibers in the monkey temporal lobe: the hippocampal formation. J Comp Neurol 331:37-74.

Pitler TA, Alger BE (1992) Postsynaptic spike firing reduces synaptic

O ~ Y 28~1175-1181.

GABAA responses in hippocampal pyramidal cells. J Neurosci 12: 41224132.

Pitler TA, Alger BE (1994) Depolarization-induced suppression of GABA- ergic inhibition in rat hippocampal pyramidal cells: G protein involvement in a presynaptic mechanism. Neuron 13:1447-1455.

Pollard H, Charriaut-Marlangue C, Cantagrel S, Represa A, Robain 0, Moreau J, Ben-Ari Y (1994) Kainate-induced apoptotic cel death in hippocampal neurons. Neuroscience 63:7-18.

Poncer J-C, Shinozaki H, Miles R (1995) Distinct metabotropic glutamate receptors modulate synaptic inhibition in the rat hippocampus. J Physiol (Lond) 485:121-134.

Poncer J-C, Shinozaki H, Miles R (1995) Dual modulation of synaptic inhibition by distinct metabotropic gluatamate receptors in the rat hippocampus. J Physiol (Lond) 485:124-154.

Poolos N, Kocsis J (1 990) Dendritic action potential activated by NMDA receptor-mediated EPSP’s in CAI hippocampal cells. Brain Res 524:342-346.

Prince DA (1978) Neurophysiology of epilepsy. Annu Rev Neurosci 1 :3 59-4 1 5.

Radpour S, Thomson AM (1992) Synaptic enhancement induced by NMDA and Qp receptors and presynaptic activity. Neurosci Lett

Raggenbass M, Wuarin JP, Ghwiler BH, Dreifuss JJ (1985) Opposing effects of oxytocin and a p-receptor agonistic opioid peptide on the same class of non-pyramidal neurons in the rat hippocampus. Brain Res 344:392-396.

Ramon y Cajal S (1 893) Estructura del asfa de Ammon y fascia dentata. Ann SOC Esp Hist Nat 22.

Ramon y Cajal S (1 91 1) Histologie de systeme nerveux de I’Homme et des vertebres tomme 11. Paris: Maloine.

Ranck JB Jr (1973) Studies on single neurons in dorsal hippocampal formation and septum in unrestrained rats. I. Behavioral correlates and firing repertoires. Exp Neurol 42:461-531.

Recce LJ, Schwartzkroin PA (1991) Effects of cholinergic agonists on two non-pyramidal cell types in rat hippocampal slices. Brain Res 566: 1 15-1 26.

Reckling JC (1 993) Effects of met-enkephalin on GABAergic spontaneous miniture IPSPs in organotypic slice cultures of the rat hippocampus. J Neurosci 13:1954-1964.

Regidor J, Martin-Trujillo JM, Lopez-Garcia C, Marin-Guiron F (1 974) Estudio citoarquitectonico de la corteza cerebral de reptiles. 11. Tipologia dendritica y distribucion neuronal en areas corticales de Lacerta galloti. Trab Inst Cajal Inv Biol 66:l-32.

Resibois A, Rogers J H (1992) Calretinin in rat brain-an immunohistochemical study. Neuroscience 46: 101-1 34.

Ribak CE, Peterson GM (1991) Intragranular mossy fibers in rats and gerbils form synapses with the somata and proximal dendrites of basket cells in the dentate gyrus. Hippocampus 1:355-364.

Ribak CE, Seress L (1983) Five types of basket cell in the hippocampal dentate gyrus: a combined Golgi and electron microscopic study. J Neurocytol 12:577-597.

Ribak CE, Vaughn JE, Saito K (1978) Immunocytochemical localization of glutamic acid decarboxylase in neuronal somata following colchicine inhibition of axonal transport. Brain Res 140:315-332.

Ribak CE, Vaughn JE, Barber RP (1981) Immunoqtochemical localization of GABAergic neurons at the electron microscopical level. Histochemistry 13:555-582.

Ribak CE, Seress L, Amaral DG (1985) The development, ultrastructure and synaptic connections of the mossy cells of the dentate gyrus. J Neurocytol 148355857,

Ribak CE, Seress L, Peterson GM, Seroogy KB, Fallon JH, Schmued LC (1986) A GABAergic inhibitory component within the hippocampal commissural pathway. J Neurosci 63432-3498.

Ribak CE, Nitsch R, Seress L (1990) Proportion of parvalbumin-positive basket cells in the GABAergic innervation of pyramidal and granule cells of the rat hippocampal formation. J Comp Neurol300449461.

Ribak CE, Seress L, Leranth C (1993) Electron microscopic immuno-

13811 19-122.


cytochemical study of the distribution of parvalbumin-containing neurons and axon terminals in the primate dentate gyrus and Ammon’s horn. J Comp Neurol 327:298-321.

Richards JG, Schoch P, Haring P, Takacs B, Mohler H (1987) Resolving GABAA/benzodiazepine receptors: cellular and subcellular localization in the CNS with monoclonal antibodies. J Neurosci 7:1866- 1886.

Richmond BJ, Optican LM (1990) Temporal encoding of two-dimensional patterns by single units in primate primary visual cortes: 11. Information trasrnission. J Neurophysiol 64:370-380.

Richter-Levin G, Segal M (1990) Effects of serotonin releasers on dentate granule cell excitability in the rat. Exp Brain Res 82:199-207.

Roberts GW, Woodhams PL, Polak JM, Crow TJ (1984) Distribution of neuropeptides in the limbic system of the rat: the hippocampus. Neuroscience 11 :35-77.

Rocamora N, Pascual M, Acsidy L, de Leces L, Freund TF, Soriano E ( 1 996) Expression of NGF and NT3 mRNA in hippocampal interneurons iiiiiervated by the GABAergic septohippocampal pathway. J Neurosci 16:3991-4004.

Rogers J H (1992) Irnmunohistochemical markers in rat cortex-co-localization of calretinin and calbindin-d28k with neuropeptides and GABA. Brain Res 587:147-157.

Ropert N (1988) Inhibitory action of serotonin in CA1 pyramidal neurons in vitro. Neuroscience 36:631-641.

Ropert N , Guy N (1991) Serotonin facilitates GABAergic transmission in the CA1 region of rat hippocampus in vitro. J Physiol (Lond) 44 1 : 12 1-1 36.

Rose G, Schubert P (1977) Release and transfer of [3H]adenosine de- rivatives in the cholinergic septa1 system. Brain Res 121:353-357.

Rosene DL, Van Horsen GW (1987) The hippocampal formation of the primate brain. A review of some comparative aspects of cytoarchi- tecture and connections. In: Cerebral cortex, vol 6: further aspects of cortical function, including hippocampus (Jones EG, Peters A, eds), pp 345-356. New York: Plenum Press.

Sah P, Hestrin S, Nicoll RA (1990) Properties of excitatory postsynaptic currents recorded in vitro from rat hippocampal interneurones. J Physiol (Lond) 430:605-616.

Saint DA, Buckett KJ ( I 99 1) Modulation of the transient potassium current in rat hippocampal neurons by cholecystokinin. Neuropeptides

Samulack DD, Lacaille JC (1 993) Hyperpolarizing synaptic potentials evoked in CAI pyramidal cells by glutamate stimulation of interneurons from the orienslalveus border of rat hippocampal slices. 11. Sensitivity ro GABA antagonists. Hippocampus 3:345-358.

Schaffer K (1 892) Beitrag zur histologie der Amnionshornformation. Arch Mikrosk Anat 39:611-632.

Scharfman HE (1 994) EPSPs of dentate gyrus granule cells during epileptiform bursts of dentate hilar “mossy” cells and area CA3 pyramidal cells in disinhibited rat hippocampal slices. J Neurosci 14:604 1-6057.

Scharfman H E (1995a) Electophisiological diversity of pyramidal-shaped neurons at the granule cell laydhilus border of the rat dentate gyms recorded in vitro. Hippocampus 5:287-305.

Scharfman HE (1 995b) Electrophysiological evidence that dentate hilar mossy cells are excitatoty and innervate both granule cells and interneurons. J Neurophysiol 74: 179-1 94.

Scharfman HE, Sarvey JM (1 985) Gamma-aminobutyrate sensitivity does not change during long-term potentiation in rat hippocampal slices. Neuroscience 15:695-702.

Scharfman HE, Schwartzkroin PA (1988) Further studies of the effects of somatostatin and related peptides in area CAI of rabbit hippocampus. Cell Mol Neurobiol 8:411-429.

Scharfman HE, Schwartzkroin PA (1989) Selective depression of GABA- mediated IPSPs by somatostatin in area CA1 of rabbit hippocampal slices. Brain Res 493:205-211.

Schmidt-Kastner R, Freund T F (1991) Selective vulnerability of the hippocampus in brain ischemia. Neuroscience 40:599-636.

20: 15 1-1 57.

Schmitz D, Empson RM, Heinemann U (1995) Serotonin reduces inhibition via 5-HTlA receptors in area CAI of rat hippocampal slices in vitro. J Neurosci 15:7217-7225.

Schwartzkroin PA, Kunkel DD (1985) Morphology of identified interneurons in the CA1 regions of guinea pig hippocampus. J Comp Neurol 232:205-218.

Schwartzkroin PA, Mathers LH (1 978) Physiological and morphological identification of a nonpyramidal hippocampal cell type. Brain Res 157: 1-10.

Schwerdtfeger WK, Lopez Garcia C (1986) GABAergic neurons in the cerebral cortex of the brain of a lizard (Podarcis hirpanicu). Neurosci Lett 68:117-121.

Scotti AL, Nitsch C (1991) The perforant path in the seizure sensitive gerbil contains the Ca(2+)-binding protein parvalbumin. Exp Brain Res 85:137-143.

Seay-Lowe SI., Claiborne RJ (1 992) Morphology of intracellularly labeled interneurons in the dentate gyrus ofthe immature rat. J Comp Neurol 32423-36.

Seeburg PH (1993) The TINS/TIPS lecture. The molecular biology of mammalian glutamate receptor channels. TINS 16:359-365.

Segal M (1990a) A subset of local interneurons generate slow inhibitory postsynaptic potentials in hippocampal neurons. Brain Res 5 11: 163-164.

Segal M (1990b) Serotonin attenuates a slow inhibitory postsynaptic potential in rat hippocampal neurons. Neuroscience 36:63 1-641.

Segal M, Bloom FE ( I 974) The action of norepinephrine in the rat hippocampus. 11. Activation of the input pathway. Brain Res 7299-114.

Segal M, Greenberger V (1992) Acidic amino acids evoke a smaller [Ca2+]; in GABAergic than non-GABAergic hippocampal neurons. Neurosci Lett 140:243-246.

Seress L (1 988) Interspecies comparison of the hippocampal formation shows increased emphasis on the regio superior in the Ammon’s horn of the human brain. J Hirnforsch 29:335-340.

Seress L (1992) Morphological variability and developmental aspects of monkey and human granule cells: differences between the rodent and primate dentate gyrus. Epilepsy Res (Suppl) 7:3-28.

Seress L, Frotscher M (1 99 1) Basket cells in the monkey fascia dentata- a Golgi/electron microscopic study. J Neurocytol 20:9 15-928.

Seress L, Leranth C (1 996) Distribution of substance P-immunoreactive neurons and fibers in the monkey hippocampal formation. Neuro- science 71:633-650.

Seress L, Pokorny J (1981) Structure of the granular layer of the rat dentate gyrus: a light microscopic and Golgi study. J Anat 133: 18 1-195.

Seress L, Ribak CE (1983) GABAergic cells in the dentate gyrus appear to be local circuit and projection neurons. Exp Brain Res 50:173-182.

Seress L, Ribak CE (1984) Direct commissural connections to the basket cells of the hippocampal dentate gyrus: anatomical evidence for feed-fonvard inhibition. J Neurocytol 13:2 15-225.

Seress L, Ribak CE (1 985) A combined Golgi-electron microscopic study of non-pyramidal neurons in the CA1 area of the hippocampus. J Neurocytol 1 4 7 17-730.

Seress L, Ribak CE (1990a) Postnatal development of the light and electron microscopic features of basket cells in the hippocampal dentate gyrus of the rat. Anat Embryo1 181:547-565.

Seress L, Ribak CE (1990b) The synaptic connections of basket cell axons in the developing rat hippocampal formation. Exp Brain Res

Seress L, Ribak CE (1992) Ultrastructural features of primate granule cell bodies show important differences from those of rats-axosomatic synapses, somatic spines and infolded nuclei. Brain Res 569: 353-357.

Seress L, Gulyis AI, Freund TF (1991) Parvalbumin- and calbindin D28k-immunoreactive neurons in thc hippocampal formation of the macaque monkey. J Comp Neurol 313:162-177.

Seress L, Gulyis AI, Ferrer I, Tunon T, Soriano E, Freund ‘TF (1993a) Distribution, morphological features, and synaptic connections of

8 1~500-508.


pawalbumin- and calbindin D28k-immunoreactive neurons in the human hippocampal formation. J Comp Neurol 337:208-230.

Seress L, Nitsch R, Leranth C (1993b) Calretinin immunoreactivity in the monkey hippocampal formation-I. Light and electron microscopic characteristics and co-localization with other calcium-binding proteins. Neuroscience 55:775-796.

Shadlen MN, Newsome WT (1994) Noise, neural codes and cortical organization. Curr Opin Neurobiol 4:569-579.

Shinoda K, Tohyama M, Shiotani Y (1987) Hippocampofugal gamma- aminobutyric acid (GABA)-containing neuron system in the rat: a study using a double-labeling method that combines retrograde tracing and immunocytochemistry. Brain Res 409:18 1-1 86.

Siesjo BK, Bengtsson F (1989) Calcium fluxes, calcium antagonists, and calcium-related pathology in brain ischemia, hypoglycemia, and spreading depression: a unifying hypothesis. J Cereb Blood Flow Metab 9:127-140.

Siggins GR, Zieglgansberger W (1 98 1) Morphine and opioid peptides reduce inhibitory synaptic potentials in hippocampal pyramidal cells without altering membrane potential. Proc Natl Acad Sci USA

Sik A, Tamamaki N, Freund TF (1993) Complete axon arborization of a single CA3 pyramidal cell in the rat hippocampus, and its relationship with postsynaptic parvalbumin-containing interneurons. Eur J Neurosci 5:1719-1728.

Sik A, Minen A, Penttonen M, BuzsAki G (1994) Inhibitory CA1-CA3- hilar region feedback in the hippocampus. Science 265: 1722-1724.

Sik A, Penttonen M, Ylinen A, Buzsdki G (1 995) Hippocampal CA1 interneurons: an in vivo intracellular labeling study. J Neurosci 15:665 1-6665.

Sik A, Penttonen M, Buzsdki G (submitted) Interneurons in the dentate gyrus: An in vivo intracellular labeling study. Eur J Neurosci.

Simmons ML, Terman GW, Gibbs SM, Chavkin C (1995) L-type calcium channels mediate dynorphin neuropeptide release from dendrites of but not axons of hippocampal granule cells. Neuron 14: 1265-1 272.

Sims KB, Hoffman DL, Said SI, Zimmerman EA (1980) Vasoactive intestinal polypeptide (VIP) in mouse and rat brain: an immunocyco- chemical study. Brain Res 186:165-183.

Singer W (1994) Time as coding space in neocortical processing: a hypothesis. In: Temporal coding in the brain (Buzsaki G, Llinas RR, Singer W, Berthoz A, Christen Y, eds), pp 5 1-79. Berlin: Springer- Verlag.

Singer W (1995) Development and plasticity of cortical processing ar- chitectures. Science 270:758-763.

Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta phase precession in neuronal populations and the compression of temporal sequences. Hippocampus 6: 149-172.

Sloviter RS (1 989) Calcium-binding protein (calbindin-D28k) and parvalbumin immunocytochemistry: localization in the rat hippocampus with specific reference to the selective vulnerability of hippocampal neurons to seizure activity. J Comp Neurol 280:183-196.

Sloviter RS (1 99 1) Permanently altered hippocampal structure, excitability, and inhibition after experimental status epilepticus in the rat: the “dormant basket cell” hypothesis and its possible revelance to temporal lobe epilepsy. Hippocampus 1 :41-66.

Sloviter RS, Nilaver G (1 987) Immunocyrochemical localization of GABA-, cholecystokinin-, vasoactive intestinal polypeptide-, and somatostatin-like immunoreactivity in the area dentata and hippocampus of the rat. J Comp Neurol 256:42-60.

Snyder SH (1992) Nitric oxide: first in a new class of neurotransmitters. Science 257:494496.

Softky WR (1995) Simple codes vs. efficient codes. Curr Opin Neurobiol 5: 239-247.

Softky WR, Koch C (1993) The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J Neurosci 13:334-360.

78~5235-5239.

SoltCsz I, Deschenes M (1 993) Low- and high-frequency membrane potential oscillations during theta activity in CA1 and CA3 pyramidal neurons of the rat hippocampus under ketamine-xylazine anesthesia. J Neurophysiol 70:97-116.

Solttsz I, Mody I (1994) Patch-clamp recordings reveal powerful GABAergic inhibition in dentate hilar neurons. J Neurosci 14:2365- 2376.

Solttsz I, Smetters DK, Mody I (1995) Tonic inhibition originates from synapses close to the soma. Neuron 14:1273-1283.

Somogyi P (1977) A specific “axo-axonal” interneuron in the visual cortex of the rat. Brain Res 136:345-350.

Somogyi P (1978) The study of Golgi stained cells and of experimental degeneration under the electron microscope: a direct method for the identification in the visual cortex of three successive links in a neuron chain. Neuroscience 3: 167-180.

Somogyi P, Freund TF (1989) Immunocytochemistry and synaptic relationships of physiologically characterized, HRP-filled neurons. In: Neuroanatomical tract-tracing methods, vol 2: recent progress (Heimer L, Zaborszky L, eds), pp 239-264. New York: Plenum.

Somogyi P, Freund TF, Halasz N, Kisvarday ZF (1981) Selectivity of neuronal [3H]GABA accumulation in the visual cortex as revealed by Golgi staining of the labeled neurons. Brain Res 225:431-436.

Somogyi P, Nunzi MG, Gorio A, Smith AD (1 983a) A new type of specific interneuron in the monkey hippocampus forming synapses exclusively with the axon initial segments of pyramidal cells. Brain Res 259:137-142.

Somogyi P, Smith AD, Nunzi MG, Gorio A, Takagi H, Wu JY (1 983b) Glutamate decarboxylase immunoreactivity in the hippocampus of the cat: distribution of immunoreactive synaptic terminals wiht special reference to the axon initial segment of pyramidal neurons. J Neurosci 3: 1450-1468.

Somogyi P, Hodgson AJ, Smith AD, Nunzi MG, Gorio A, Wu JY (1984) Different populations of GABAergic neurons in the visual cortex and hippocampus of cat contain somatostatin- or cholecystokinin-ini- munoreactive material. J Neurosci 4:2590-2603.

Somogyi P, Freund TF, Hodgson AJ, Somogyi J, Beroukas D, Chubb IW (1985a) Identified axo-axonic cells are immunoreactive for GABA in the hippocampus and visual cortex of the cat. Brain Res 332: 143-149.

Somogyi P, Hodgson AJ, Chubb IW, Penke B, Erdei A (1 985b) Antisera to gamma-aminobutyric acid. J Histochem Cytochem 33:240-248.

Soriano E, Frotscher M (1 989) A GABAergic axo-axonic cell in the fascia dentata controls the main excitatory hippocampal pathway. Brain Res 503:170-174.

Soriano E, Frotscher M (1993a) GABAergic innervation of the rat fascia dentata: a novel type of interneuron in the granule cell layer with extensive axonal arborization in the molecular layer. J Comp Neurol 334:385-396.

Soriano E, Frotscher M (1993b) Spiny nonpyramidal neurons in the CA3 region of the rat hippocampus are glutamate-like immunoreactive and receive convergent mossy fiber input. J Comp Neurol 333:435448.

Soriano E, Frotscher M (1994) Mossy cells of the rat fascia dentata are glutamate-immunoreactive. Hippocampus 4:65-70.

Soriano E, Nitsch R, Frotscher M (1990) Axo-axonic chandelier cells in the rat fascia dentata: Golgi-electron microscopy and immunocytochemical studies. J Comp Neurol 293:l-25.

Soriano E, Del Rio JA, Martinez A, Super H (1994) Organization of the embryonic and early postnatal murine hippocampus. I. Immunocytochemical characterization of neuronal populations in the subplate and marginal zone. J Comp Neurol 342:571-595.

Spencer WA, Kandel ER (1961) Electrophysiology of hippocatnpal neurons: IV. Fast prepotentials. J Neurophysiol 24260-271.

Spruston N, Schiller Y, Stuart G, Sakmann B (1995) Activity-dependent action potential invasion and calcium influx into hippocampal CAI dendrites. Science 268:297-300.


Staley KJ, Mody I (1992) Shunting of excitatory input to dentate gyms granule cells by a depolarizing GABAA receptor-mediated postsynaptic conductance. J Neurophysiol 68: 197-212.

Staley KJ, Soldo BL, Proctor WR (1995) Ionic mechanisms of neuronal excitation by inhibitory GABAA receptors. Science 269:977-8 I .

Stanfield BB, Cowan WM (1984) An EM autoradiographic study of the hypothalamo-hippocampal projection. Brain Res 309:299-307.

Stanton PK, Sejnowski TJ (1989) Associative long-term depression in the hippocampus induced by hebbian covariance. Nature 339:2 15- 227.

Staubli U, Xu FB (1995) Effects of 50HT3 receptor antagonism on hippocampal theta rhythm, memory and LTP induction in the freely moving rat. J Neurosci 15:2445-2452.

Stelzer A, Slaw NT, Bruggencate G (1987) Activation of NMDA receptors blocks GABAergic inhibition in an in vitro model of epilepsy. Nature 326:698-701.

Stelzer A, Simon G, Kovacs G, Rai A (1994) Synaptic disinhibition during maintenance of long-term potentiation in the CAI hippocampal subfield. Proc Natl Acad Sci USA 91:3058-3062.

Stevens CF, Wang Y (1995) Facilitation and depression at single central synapses. Neuron 14795-802.

Stewart M, Fox SE (1989) Two populations of rhythmically bursting neurons in rat medial septum are revealed by atropine. J Neurophysiol 61 :982-993.

Stewart M, Fox SE (1 990) Do septal neurons pace the hippocampal theta rhythm? TINS 13:163-168.

Stewart M, Luo Y, Fox SE (1992) Effects of atropine on hippocampal theta cells and complex-spike cells. Brain Kes 591 :122-128.

Stichel CC, Singer W (1985) Organization and morphological characteristics of choline acetyltransferase-containing fibers in the visual thalamus and striate cortex of the cat. Neurosci Lett 53:155-160.

Storm-Mathisen J, Leknes AK, Bore AT, Vaaland JL, Edminson P, Haug FMS, Ottersen OP (1983) First visualization of glutamate and GABA in neurones by immunocynochemistry. Nature 301 :5 17-520.

Stuart G, Sakmann B (1994) Active propagation of somatic action potentials into neocortical pyramidal cell dendrites. Science 367:69-72.

Stumpf C (1965) Drug action on the electrical activity of the hippocampus. Int Rev Neurobiol 8:77-138.

Sutula T, He XX, Cavazos J, Scott G (1988) Synaptic reorganization in the hippocampus induced by abnormal functional activity. Science 239:1147-1150.

Suzuki SS, Smith GK (1987) Spontaneous EEG spikes in the normal hippocampus: 111. Relations to evoked potentials. Electroencephalogr Clin Neurophysiol 69:541-549.

Swanson LW (1983) Neuropeptides-new vistas on synaptic transmission. Trends Neurosci 6:3-19.

Swanson I,W, Sawchenko PE, Cowan WM (1980) Evidence that the commissural, associational and septal projections of the regio inferior of the hippocampus arise from the same neurons. Brain Res 197:

Szentigothai J (1 962) O n the synaptology of the cerebral cortex. In: Structure and function of the nervous system (Sarkissov SA, ed), pp 6-14. Moscow: Medgiz.

Szentigothai J (1965a) The synapse of short local neurons in the cerebral cortex. In: Modern trends in neuromorphology (Szentigothai J, ed), pp 251-276. Budapest: Akademiai Kiado.

Szentigothai J (1965b) The use of degeneration methodsin the investigation of short neuronal connections. In: Degeneration patterns in the nervous system (Singer M, Schade JP, eds), pp 1-32. Amsterdam: Elsevier.

Szentigothai J, Arbib MA (1974) Conceptual models of neural organization. Neurosci Res Prog Bull 12:307-510.

Tamamaki N, Nojyo Y (1990) Disposition of the slab-like modules formed by axon branches originating from single CAI pyramidal neurons in the rat hippocampus. J Comp Neurol 291:509-519.

Tamamaki N, Nojyo Y (1 993) Projection of the entorhinal layer-I1 neu-

207-212.

rons in the rat as revealed by intracellular pressure-injection of neu-. robiotin. Hippocampus 3:471-480.

Tarnamaki N, Nojyo Y (1995) Preservation of topography in the connections between the subiculum, field CAI, and the entorhinal cortex in rats. J Comp Neurol 353:379-390.

Tamamaki N, Waltanabe K, Nojyo Y (1984) A whole image of the hippocampal pyramidal neuron revealed by intracellular pressure-injection of horseradish peroxidase. Brain Res 307:336-340.

Tatsuno I, Yada T, Vigh S, Hidaka H, Arimura A (1992) Pituitary adeny- late cyclase activating polypeptide and vasoactive intestinal peptide increase cytosolic free calcium concentration in cultured rat hippocampal neurons. Endocrinology 131 :73-8 1.

Taube JS, Schwartzkroin PA (1987) Intracellular recording from hippocampal CA1 interneurons before and after development of long- term potentiation. Brain Res 419:32-38.

Tecott LH, Maricq AV, Julius D (1993) Nervous system distribution of the serotonin 5-HT3 receptor mRNA. Proc Natl Acad Sci IJSA 90: 1430-1434.

Teyler TJ, Cavus I, Coussens C, DiScenna P, Grover L, Lee YI’, Little Z (1 994) Multideterminant role of calcium in hippocampal synaptic plasticity. Hippocampus 4:623-634.

Thalmann RH ( I 988) Blockade of a late inhibitory postsynaptic potential in hippocampal CA3 neurons in vitro reveals a late depolarizing potential that is augmented by pentobarbital. Neurosci L.ett 95:

Thiels E, Barrionuevo G, Berger TW (1994) Excitatory stimulation during postsynaptic inhibition induces long-term depression in hippocampus in vivo. J Neurophysiol 72:3009-3016.

Thompson SM, Capogna M, Scanziani M (1993) Presynaptic inhibition in the hippocampus. ’Trends Neurosci 16:222-227.

Thompson SM, Gahwiler BL (1989) Activity-dependent disinhibition. 111. Desensitization and GABAb receptor-mediated presynaptic inhibition in the hippocampus in vitro. J Neurophysiol 61:524-533.

Thurbon D, Field A, Redman S (1994) Electrotonic profiles of interneurons in stratum pyramidale of the CAI region of rat hippocampus. J Neurophysiol 71: 1948-1 958.

Tomasulo RA, Steward 0 (1996) Homosynaptic and heterosynaptic changes in driving of dentate gyrus interneurons after brief tetanic stimulation in vivo. Hippocampus 6:62-71.

Tomasulo RA, Levy WB, Steward 0 (1991) LTP-associated EPSP/spike dissociation in the dentate gyrus: GABAergic and non-GABAergic components. Brain Res 561:27-34.

Tomasulo RA, Kamirez JJ, Steward 0 (1993) Synaptic inhibition regulates associative interactions between afferents during the induction of long-term potentiation and depression. Proc Natl Acad Sci USA

Tork I (1990) Anatomy of the serotonergic system. Ann NY Acad Sci

T6th K, Freund TF (1 992) Calbindin D28k-containing nonpyramidal cells in the rat hippocampus: their immunoreactivity for GABA and projection to the medial septum. Neuroscience 49:793-805.

T6th K, Borhegyi Z, Freund TF (1993) Postsynaptic targets of GABAergic hippocampal neurons in the medial septum-diagonal band of broca complex. J Neurosci 13:3712-3724.

Toth K, Freund TF, Miles R (submitted) Disinhibition of rat hippocampal pyramidal cells by GABAergic afferents from the septum. J Physiol.

Traub RD (1995) Model of synchronized population bursts in electrically coupled interneurons containing active dendritic conductances. J Comput Neurosci 2:283-289.

Traub RD, Llinas R (1977) The spatial distribution of ionic conductances in normal and axotomiwd motorneurons. Neuroscience 2:829-849.

Traub RD, Miles R (1995) Pyramidal to inhibitory spike transduction explicable by active dendritic conductances in inhibitory cells. J Comput Neurosci 2:291-298.

155-1 60.

90:11578-11582.

600:9-34.


Traub RD, Wong RKS (1982) Cellular mechanism of neuronal synchronization in epilepsy. Science 216:745-747.

Traub RD, Jefferys JGR, Miles R, Whittington MA, T6th K (1994) A branching dendritic model of a rodent CA3 pyramidal neurone. J Physiol (Lond) 481:79-95.

Traub RD, Whittington MA, Colling SB, Buzsiki G, Jefferys JGR (1996) Analysis of gamma rhythms in the rat hippocampus in vitro and in vivo. J Physiol (Lond) 493:471-484.

I’sodyks MV, Skaggs WE, Sejnowski TJ, McNaughton BL (1996) Population dynamics and theta rhythm phase precession of hippocampal place cell firing: a spiking neuron model. Hippocampus.

l’subokawa H, Ross WN (in press) IPSPs modulate spike backpropaga- tion and associated [Ca2+], changes in the dendrites of hippocampal CA1 pyramidal neurons. J Neurophysiol

Turner DA, Li X-G, Pyapali GK, Ylinen A, Buzsiki G (1995) Morpho- metric and electrical properties of reconstructed hippocampal CA3 neurons recorded in vivo. J Comp Neurol 365:580-594.

Urioste R, Chrobak JJ, Buzsriki G (1992) Physiological classification of hippocampal feedback and feedfonvard interneurons. SOC Neurosci Abstr 18:320.

Valentino RJ, Dingledine R (198 1) Presynaptic inhibitory effect of acetylcholine in the hippocampus. J Neurosci 1:784-792.

Valtschanoff JG, Weinberg RJ, Kharazia VN, Nakane M, Schmidt H (1993) Neurons in rat hippocampus that synthesize nitric oxide. J Comp Neurol 331:lll-121.

van Groen T, Wyss JM (1990) Extrinsic projections from area CA1 of the rat hippocampus: olfactory, cortical, subcortical, and bilateral hippocampal formation projections. J Comp Neurol 302:5 15-528.

van Vreeswijk C, Abbott LF, Ermentraut GB (1995) When inhibition, not excitation synchronizes neuronal firing. J Comput Neurosci 1:313-322.

Vandenvolf C (1989) A general role for serotonin in the control of behavior: studies with intracerebral5,7-dihydroxytryptamine. Brain Res 504: 192-1 98.

Verney C, Baulac M, Berger B, Alvarez C, Vigny A, Helle KB (1985) Morphological evidence for a dopaminergic terminal field in the hippocampal formation of young and adult rat. Neuroscience 14:1039- 1052.

Vertes RP (1992) PHA-L analysis of projections from the supramammillary nucleus in the rat. J Comp Neurol 326:595-622.

Vincent SR, Kimura H (1992) Histochemical mapping of nitric oxide synthase in the rat brain. Neuroscience 46:755-784.

Wang X-J, Rinzel J (1993) Spindle rhythmicity in the reticularis thalamic nucleus: synchronization among mutually inhibitory neurons. Neuroscience 53:899-904.

Wang X-J, Buzsaki G (in press) Gamma oscillation by synaptic inhibition in a hippocampal interneuronal network model. J Neurosci

Warnick JE, Pellmar TC (1986) Resistance of afterhyperpolarizations in hippocampal pyramidal cells to prostaglandins and vasoactive intestinal peptide (VIP). Neurosci Lett 70: 295-299.

Weiser M, Vega-Saenz de Miera E, Kentros C, Moreno H, Franzen L, Hillman D , Baker H and Rudy B (1994) Differential expression of Shaw-related channels in the rat central nervous system. J Neurosci 14:949-972.

West MJ (1990) Stereological studies of the hippocampus:a comparison of the hippocampal subdivisions of diverse species including hedge- hogs, laboratory rodents,wild mice and men. In: Progress in brain research (Storm-Mathisen J, Zimmer J, Ottersen OP, eds) 83:13-36. New York: Elsevier Science.

Whittington MA, Traub RD, Jefferys JG (1995) Synchronized oscillation in interneuron networks driven by metabotropic glutamate receptor activation. Nature 373:612-615.

Wigstrom H, Gustafsson B (1985) Facilitation of hippocampal long-lasting potentiation by GABA antagonists: Acta Physiol Scand 125: 159-72.

6:271-280.

Wilcox BJ, Unnerstall JR (1990) Identification of a subpopulation of neuropeptide Y-containing locus coeruleus neurons that project to the entorhinal cortex. Synapse 6:284-291.

Williams S, Lacaille J-C (1 990) Bicuculline- and phaclofen-resisant hyperpolarizations evoked by glutamate applications to stratum lacunosum-moleculare in CAI pyramidal cells of the rat hippocampus in vitro. Eur J Neurosci 2:993-1003.

Williams S, Vachon P, Lacaille J-C (1993) Monosynaptic GABA-mediated inhibitory postsynaptic potentials in CA1 pyramidal cells of hy- perexcitable hippocampal slices from kainic acid-treated rats. Neuroscience 52:54 1-5 54.

Williams S, Samulack DD, Beaulieu C, Lacaille JC (1994) Membrane properties and synaptic responses of interneurons located near the stratum lacunosum-moleculare/radiatum border of area CAI in whole-cell recordings from rat hippocampal slices. J Neurophysiol 71 :22 17-2235.

Wilson MA, McNaughton BL (1993) Dynamics of the hippocampal ensemble code for space. Science 261:1055-1058.

Wong RKS, Watkins DJ (1982) Cellular factors influencing GABAA response in hippocampal pyramidal cells. J Neurophysiol48:938-954.

Wong RKS, Prince DA, Basbaum A1 (1979) Intradendritic recordings from hippocampal neurons. Proc Natl Acad Sci USA 76:986-990.

Woodson W, Nitecka L, Ben-Ari Y (1989) Organization of the GABAergic system in the rat hippocampal formation: a quantitative immunocytochemical study. J. Comp Neurol 280:254-271.

Woodward MP, Young WW, Bloodgood RA (1985) Detection of monoclonal antibodies specific for carbohydrate epitopes using periodate oxidation. J Immunol Methods 78: 143-1 53.

Wouterlood FG, Saldana E, Witter MP (1990) Projection from the nucleus reuniens thalami to the hippocampal region: light and electron microscopic tracing study in the rat with the anterograde tracer Pbaseohs vulgaris-leucoagglutinin. J Comp Neurol 296: 179-203.

Wu L-G, Saggau P (1994) Pharmacological identification of two types of presynaptic voltage-dependent calcium channels at CA3-CA1 synapses of the hippocampus. J Neurosci 14:5613-5622.

Xie Z, Sastry BR (1992) Actions of somatostatin on GABA-ergic synaptic transmission in the CA1 area of the hippocampus. Brain Res 59 1 :239-247.

Yamamoto C, Chujo T (1978) Long-term potentiation in thin hippocampal sections studied by intracellular and extracellular recordings. Exp Neurol 58:242-250.

Yamamoto C, Marshall P, Hemmendinger LM, Boyer AB, Caviness VS Jr. (1988) Distribution of glucuronic acid-sulfate-containing glycoproteins in the central nervous system of the adult mouse. Neurosci Res 5:273-298.

Yan YH, Van Brederode JF, Hendrickson AE (1995) Transient co-localization of calretinin, parvalbumin and calbindin-D28k in developing visual cortex of monkey. J Neurocytol 24:825-837.

Yanagihara T, Yoshimine T, Morimoro K, Yamamoto K, Homburger HA (1985) Immunohistochemical investigation of cerebral ischemia in gerbils. J Neuropathol Exp Neurol 44:204-215.

Yeckel MF, Berger TW (1990) Feedfonvard excitation of the hippocampus by afferents from the entorhinal cortex: redefinition of the role of the trisynaptic pathway. Proc Natl Acad Sci USA 87:5832- 5836.

Minen A, Bragin A, Nadasdy Z, Janda G, Szabo I, Sik A, Buzsiki G (1 9953 Sharp-wave-associated high-frequency oscillation (200 Hz) in the intact hippocampus: network and intracellular mechanisms. J Neurosci 14:30-46.

Minen A, Solttsz I, Bragin A, Penttonen M, Sik A, Buzsiki G (1995b) Intracellular correlates of hippocampal theta rhythm in identified pyramidal cells, granule cells and basket cells. Hippocampus 5:78-90.

Yoon KW, Rothman SM (1991) Adenosine inhibits excitatory by not inhibitory synaptic transmission in the hippocampus. J Neurosci 11:2217-2235.

Zhang L, McBain CJ (1 995a) Potassium conductances underlying ac-


tion potential repolarization and afterhyperpolarization in rat CA1 hippocampal interneurons. J Physiol (Lond) 488:661-672.

Zhang L, McBain CJ (1995b) Voltage-gated potassium currents in stratum oriens-alveus inhibitory neurones of the rat CA1 hippocampus. J Physiol (Lond) 488:647-660.

Zhang DX, Levy WE3 (1993) Bicucullin permits the induction of long- term depression by heterosynaptic, translaminar conditioning in the hippocampal dentate gyrus. Brain Res 613:309-312.

Zhu PJ, Krnjevic K (1994) Endogenous adenosine deaminase does not modulate synaptic transmission in rat hippocampal slices under nor- moxic or hypoxic conditions. Neuroscience 63:489-97.

Zieglgansberger W, French ED, Bloom FE (1979) Opioid peptides may

excite hippocampal pyramidal neurons by inhibiting adjacent inhibitory interneurons. Science 205:415417.

Zimmer J, Laurberg S, Sunde N (1983) Neurobiology of the hippocampus. In: Neuroanatomical aspects of normal and transplanted hippocampal tissue (Seifert W, ed), pp 39-64. New York: Academic Press.

Zimprich F, Zezula J, Sieghart W, Lassmann H (1991) Immunohis- to-chemical localization of the alpha 1, alpha 2 and alpha 3 subunit of the GABA-A receptor in the rat brain. Neurosci Lett 127:125-128.

Zipp F, Nitsch R, Soriano E, Frotscher M (1989) Entorhinal fibers form synaptic contacts on pawalbumin-immunoreactive neurons in the rat fascia dentata. Brain Res 495:161-166.

freund-buzsaki1994

Documents

Transcript of freund-buzsaki1994