Postsynaptic targets of somatostatin-immunoreactive interneurons in the rat hippocampus

POSTSYNAPTIC TARGETS OF SOMATOSTATIN-IMMUNOREACTIVE INTERNEURONS IN THE RAT

HIPPOCAMPUS

I. KATONA, L. ACSAuDY and T. F. FREUND*Institute of Experimental Medicine, Hungarian Academy of Sciences, P.O. Box 67, Budapest H-1450,

Hungary

Abstract––Two characteristic interneuron types in the hippocampus, the so-called hilar perforantpath-associated cells in the dentate gyrus and stratum oriens/lacunosum-moleculare neurons in the CA3and CA1 regions, were suggested to be involved in feedback circuits. In the present study, interneuronsidentical to these cell populations were visualized by somatostatin-immunostaining, then reconstructed,and processed for double-immunostaining and electron microscopy to establish their postsynaptic targetselectivity. A combination of somatostatin-immunostaining with immunostaining for GABA or otherinterneuron markers revealed a quasi-random termination pattern. The vast majority of postsynaptictargets were GABA-negative dendritic shafts and spines of principal cells (76%), whereas other targetelements contained GABA (8%). All of the examined neurochemically defined interneuron types(parvalbumin-, calretinin-, vasoactive intestinal polypeptide-, cholecystokinin-, substance P receptor-immunoreactive neurons) received innervation from somatostatin-positive boutons.

Recent anatomical and electrophysiological data showed that the main excitatory inputs ofsomatostatin-positive interneurons originate from local principal cells. The present data revealed amassive GABAergic innervation of distal dendrites of local principal cells by these feedback drivenneurons, which are proposed to control the efficacy and plasticity of entorhinal synaptic input as afunction of local principal cell activity and synchrony. ? 1998 IBRO. Published by Elsevier Science Ltd.

Key words: inhibition, non-pyramidal cells, neuropeptides, feedback inhibition, GABA, theta.

Interneurons in the cerebral cortex are known to playa crucial role in controlling the activity of largeensembles of principal cells. Recent electrophysio-logical and anatomical studies shed light on thespecific function of various interneuron types in theregulation of population behaviour of principal cellsat different nodes in the hippocampal network (forreview see Ref. 22).

Intracellular labelling of interneurons with biocy-tin in vitro and in vivo15,16,26,59,61 and immunostain-ing for selected neurochemical markers or theircombinations are the most powerful tools to investi-gate the precise connectivity, and neurochemical fea-tures of an interneuron type. The former allows adirect correlation of electrophysiological propertieswith the axonal and dendritic arborization pattern of

a given cell, whereas the latter can visualize largepopulations of a specific cell type and allows corre-lation of neurochemical characteristics with certainmorphological features. By using these methods,many distinct interneuron types were classified in thehippocampus. It was shown that perisomatic inhibi-tory neurons (i.e. basket and axoaxonic cells) containeither the calcium-binding protein parvalbumin35 orthe neuropeptides cholecystokinin (CCK)51 andvasoactive intestinal polypeptide (VIP).1 Inhibitoryneurons that predominantly target the dendrites ofprincipal cells (i.e. dendritic inhibitory cells) includetwo major types that differ in their layer of termin-ation and afferent input. Bistratified cells16 innervatepyramidal cell dendrites in strata radiatum and oriens(in conjunction with Schaffer collaterals) and a largeproportion of them contains the calcium-bindingprotein calbindin.59,61 The other characteristic den-dritic inhibitory neuron type has a dendritic treerestricted to those layers where local collaterals ofprincipal cells of the given hippocampal subfieldarborize, whereas it’s axon terminates on distal den-drites of principal cells in conjunction with entorhinalafferents. These interneurons are the so-called oriens/lacunosum-moleculare (O-LM) cells in the CA147,61

and CA3 regions26 of the hippocampus and the hilarperforant path-associated (HIPP) cells in the hilus ofthe dentate gyrus.15,32,59

*To whom correspondence should be addressed.Abbreviations: ABC, avidin–biotin–horseradish peroxidase

complex; ACPD, 1S,3R-aminocyclopentane dicarboxylicacid; BSA, bovine serum albumin; CCK, cholecystokinin;DAB, 3,3*-diaminobenzidine–4HCl; EPSP, excitatorypostsynaptic potential; HIPP, hilar perforant path-associated; IPSP, inhibitory postsynaptic potential; LTP,long-term potentiation; M2, type 2 muscarinic receptor;mGluR1á, type 1á metabotropic glutamate receptor;NGS, normal goat serum; NPY, neuropeptide Y; O-LM,strata oriens/lacunosum-moleculare; PB, phosphatebuffer; SPR, substance P receptor; TBS, Tris-bufferedsaline; VIP, vasoactive intestinal polypeptide.

Pergamon

Neuroscience Vol. 88, No. 1, pp. 37–55, 1999Copyright ? 1998 IBRO. Published by Elsevier Science Ltd

Printed in Great Britain. All rights reserved0306–4522/99 $19.00+0.00PII: S0306-4522(98)00302-9

37

Fig. 1.

38 I. Katona et al.

In a recent study, Maccaferri and McBain44

showed that induction of long-term depression atSchaffer collateral–CA1 pyramidal cell synapses canincrease the probability of long-term potentiation(LTP) generation at entorhinal afferent-CA1 pyrami-dal cell distal dendritic synapses. The network mech-anism of this peculiar interaction was suggested to befeedback inhibition of distal dendrites by O-LM cells.This is supported by available anatomy demonstrat-ing that the vast majority of excitatory input toO-LM cells derives from local CA1 pyramidal cells.14

The finding that O-LM cells can be discharged bySchaffer collateral stimulation only if it evokes apopulation spike in CA1, and excitatory postsynapticpotentials recorded from O-LM cells always occurafter this population spike, also supports the exclu-sive feedback drive for this cell type.44 O-LM cells, inturn, terminate in stratum lacunosum-moleculare,but whether their major targets are pyramidal cells orinterneurons, also having extensive dendritic arborsin this layer,1,40 is still unknown. Similar circuitryfeatures hold true for HIPP cells in the dentate gyrus.Their dendrites are limited to the hilus, where localcollaterals of granule cell axons (mossy fibres) ter-minate, and have no access to feedforward excitatoryinput, which terminate in stratum moleculare. Incontrast, their axons arborize in the outer two-thirdsof the molecular layer in conjunction with entorhinalafferents, similar to O-LM cells in the Ammon’shorn.31,32,59 Here again, whether the major targetsare granule cell dendrites is still an open question.One may argue that the postsynaptic elements ofO-LM cells in stratum lacunosum-moleculare have tobe distal dendrites of pyramidal cells (or of granulecells in case of HIPP cells), because these are farthe most abundant available targets in these layers.However, this reasoning has already proved to bewrong in the case of calretinin- and VIP-positiveaxon arbors at the stratum oriens-alveus border,where similarly pyramidal dendrites are the mostabundant elements in the neuropil, nevertheless thisGABAergic projection terminates selectively on

GABAergic targets in this layer, and ignores pyrami-dal cell dendrites.2,24 A direct analysis of postsynap-tic elements by immunocytochemical double-stainingtechniques could provide the answer. For such astudy a neurochemical marker is required, whichselectively visualizes the axon terminals of O-LM andHIPP cells.

Interneurons immunoreactive for the neuropeptidesomatostatin have been reported in the hippocampalformation in a number of species, for example inrat,21,36,50,56 mouse,21 rabbit,39 guinea-pig,21 mon-key55 and human.8,20 Somatostatin-immunoreactivestructures have the same laminar and regional distri-bution as O-LM and HIPP cells: i.e. the perikaryaand dendritic trees are mainly located in the hilus ofthe dentate gyrus and stratum oriens of the CA1and CA3 subfields and dense somatostatin-immunoreactive axon terminal fields are present inthe outer molecular layer of the dentate gyrus and inthe stratum lacunosum-moleculare of CA1 and CA3,where entorhinal inputs terminate.

In the present study, using a new antiserumraised against prosomatostatin,41 we aimed to pro-vide further evidence that SOM-positive cells in-deed correspond to O-LM and HIPP cells. Double-immunostaining was employed to investigate thepostsynaptic targets in the dentate molecular layerand in stratum lacunosum-moleculare of theAmmon’s horn according to their GABA content orimmunoreactivity for other interneuron markers.

EXPERIMENTAL PROCEDURES

Perfusion and preparation of tissue sections

Eight male Wistar rats (300–350 g, two-months-old;Charles River, Budapest, Hungary) were deeply anaesthe-tized by Equithesin (chlornembutal, 0.3 ml/100 g), and per-fused through the heart first with saline followed by aphosphate-buffered (PB, 0.1 M) fixative containing 4%paraformaldehyde, 0.15% picric acid and 0.05% glutaralde-hyde in the case of series A (n=5) for single somatostatin-immunostaining and for somatostatin–parvalbumin,somatostatin–CCK, somatostatin–calretinin, somatostatin–VIP, somatostatin–subtance P receptor (SPR) double

Fig. 1. (A) Low power light micrograph showing a specific laminar distribution of somatostatin-immunostaining in CA1 and in the dentate gyrus. Cell bodies and dendrites (arrows) are confined tostratum oriens and the hilus, whereas the main axon terminal fields are located in stratum lacunosum-moleculare and in the outer part of stratum moleculare. Arrowheads indicate a typical main axonemerging from stratum oriens crossing stratum pyramidale and stratum radiatum towards stratumlacunosum-moleculare. (B) At higher magnification a dense network of varicose axon collaterals is visiblein stratum lacunosum-moleculare. Arrowheads point to an axon entering from stratum radiatum tostratum lacunosum-moleculare. (C) Main axons crossing the hippocampal fissure (i.e. the border ofstratum moleculare and stratum lacunosum-moleculare) were often observed (arrowheads). (D) Highpower light micrograph of a somatostatin-immunoreactive neuron located in stratum oriens of CA1. Notethe axon emerging from a proximal dendrite, crossing stratum pyramidale and bifurcating in stratumradiatum. The two main branches were partially reconstructed as shown in Fig. 2 (Cell 1). They reachedstratum lacunosum-moleculare where they extensively arborized, suggesting that they are conventionalO-LM neurons. (E–F) Somatic and dendritic spines were often observed on somatostatin-immunoreactiveneurons. Some of these spines are depicted by arrows. (G) The axon collaterals of Cell 4 bore severaldrumstick-like boutons. Two of them are indicated by arrows. ais, axon initial segment; h., hilus; s.g.,stratum granulosum; s.m., stratum moleculare; s.l-m., stratum lacunosum-moleculare; s.o., stratum oriens;s.p., stratum pyramidale; s.r., stratum radiatum. Scale bars: (A)=100 µm; (B)=30 µm; (C)=80 µm;

(D)=15 µm; (E–F)=10 µm; (G)=5 µm.

Somatostatin-immunoreactive interneurons in hippocampus 39

Fig. 2.

40 I. Katona et al.

immunostaining. In series B (n=3 animals) the fixativecontained 4% paraformaldehyde, 0.15% picric acid and 1%glutaraldehyde dissolved in 0.1 M PB (pH 7.4) for pre-embedding somatostatin-immunostaining combined withpostembedding immunogold staining for GABA. Brainswere removed from the skull, blocks of the hippocampusand overlying neocortex were dissected and coronal sectionsof 60 µm thickness were cut on a Vibratome. After extensivewashes the sections were cryoprotected in 30% sucrose in0.1 M PB overnight, and freeze-thawed in an aluminium foilboat over liquid nitrogen to enhance the penetration ofantisera without destroying the ultrastructure of sections.The sections prepared for light microscopic reconstructionof somatostatin-immunoreactive neurons were treated with0.5% Triton X-100 diluted in 0.05 M Tris-buffered saline(TBS) also containing 5% bovine serum albumin (BSA).

Pre-embedding immunocytochemistry

Following extensive washes and treatment with 1% so-dium borohydride for 30 min (only for animals perfusedwith fixative B), the sections were incubated first in 5% BSAand then in solutions of the following antisera: for singlesomatostatin-immunostaining rabbit anti-somatostatin,diluted 1:20,000,41 was used as primary antiserum(48 h). The second layer was biotinylated anti-rabbit IgG(Vector, 2 h, 1:400) followed by avidin–biotin–horseradishperoxidase complex (Elite ABC, Vector, 1.5 h, 1:400).The immunoperoxidase reaction was developed using3,3*-diaminobenzidine (DAB; Sigma) intensified withammonium nickel–sulphate (DAB–Ni) as a chromogen(black reaction product). The sections were treated with 1%osmium tetroxide in 0.1 M PB for 1 h, dehydrated inethanol and propylene oxide, and embedded in Durcupan(ACM, Fluka). During dehydration the sections weretreated with 1% uranyl acetate in 70% ethanol for 45 min.

Reconstruction of somatostatin-immunoreactive neurons andtheir axonal and dendritic processes

Selected somatostatin-immunoreactive cells were drawnwith the aid of a camera lucida from 60-µm-thick serialVibratome sections. The dendrites and axons of each cellswere reconstructed using #100 oil immersion objective andwere followed as long as it was possible to distinguish themfrom other immunoreactive processes. On the surface ofsections, capillaries near the cut ends of dendrites and axonswere also drawn, to serve as landmarks. In the adjacentsections the same dendrites and axons were identified bymatching the capillaries.

Pre-embedding double immunostaining

After the first immunostaining for somatostatin, thesections were blocked by the ABC blocking kit, thenincubated in solutions of one of the following antisera:rabbit anti-parvalbumin (1:1500);11 rabbit anti-calretinin(1:5000);55 rabbit anti-VIP (1:10,000);23 rabbit anti-CCK

(1:3000);23 and rabbit anti-substance P receptor (SPR)(1:300)58 for 48 h. This was followed by biotinylated anti-rabbit IgG (Vector, 2 h, 1:400) and Elite ABC (Vector,1.5 h, 1:400). This second immunoperoxidase reaction wasdeveloped with DAB as a chromogen (brown reaction endproduct). The sections were treated with 1% osmium tetrox-ide also containing 7% glucose to preserve colour differencebetween DAB–Ni and DAB end products, dehydrated andembedded in Durcupan as above. The specificity of theprimary antisera have been tested by the laboratories oforigin (see references above). In double-stained sectionseach antisera gave the same staining pattern as if appliedin single staining. Although the primary antisera in bothcycles were raised in rabbit, the end product of the firstimmunoperoxidase reaction masked the immune complexso that the antisera of the second cycle could not bind to thefirst. Replacing the primary antisera with normal rabbitserum resulted in the lack of specific immunostaining, onlya faint nuclear background staining was present on thesurface of the sections.

The immunoreactive profiles at the light microscopic levelwere identified by their colour difference (i.e. somatostatin-positive profiles were black, the second markers werebrown). Some of the somatostatin-immunoreactive boutonsin close contact with dendrites and somata of the neuronsimmunostained by the second marker were photographedand re-embedded for further ultrathin sectioning. At theelectron microscopic level, the same profiles were identifiedand examined whether they form synapses.

Postembedding immunogold staining for GABA

Areas with dense somatostatin-immunoreactive axonalstaining from animals perfused with fixative B were selectedboth in the ventral and dorsal part of stratum lacunosum-moleculare of the CA1 and CA3 subfields and stratummoleculare of the dentate gyrus and were re-embedded.Ultrathin sections were cut on a Reichert ultramicrotomeand adjacent sections were mounted on copper and nickelgrids. Postembedding GABA immunostaining was carriedout on the nickel grids according to the protocol of Somogyiand Hodgson,66 using a well-characterized antiserumagainst GABA.33 Incubations were performed on dropletsof solutions in humid Petri dishes. Briefly: 2% periodic acid(H5IO6) for 10 min; washed in distilled water; 2% sodiummetaperiodate (NaIO4) for 10 min; washed in distilledwater; three times 2 min in TBS (pH 7.4); 30 min in 1%ovalbumin dissolved in TBS; three times in TBS containingnormal goat serum (NGS); 1.5 h in rabbit anti-GABAantiserum (1:3000 in NGS/TBS); two times 10 min TBS;10 min in Tris buffer containing 1% BSA, 0.05% Tween 20;2 h goat anti-rabbit IgG-coated colloidal gold (1 nm or15 nm, Amersham, 1:20 in NGS/TBS) in the same solutionas before; two times 5 min washed in distilled water; 30 minsaturated uranyl acetate; washed in distilled water; stainedwith lead citrate; washed in distilled water. In the caseof 1 nm gold particles, additional postfixation with 1%

Fig. 2. Camera lucida drawings of somatostatin-immunoreactive neurons of the CA1 subfield, recon-structed from 10 consecutive sections (60 µm each). (A) The ‘‘Classic’’ type of somatostatin-positiveneurons showing typical morphology of O-LM cells. Namely, the somata and dendrites are confined tostratum oriens, while the main axons arborize predominantly in stratum lacunosum-moleculare. Note thatthe axons in each case emerge from proximal dendrites, which is a typical feature of O-LM cells. Cell 1 isalso shown in Fig. 1D. (B) A type of somatostatin-positive neuron showing striking differences from thosein A. The soma is located in the alveus, whereas some dendrites extend to stratum pyramidale. The axonemerges from the cell body and bifurcates several times close to the axon hillock. One of the main axonsdisappears in the alveus (arrowheads), and the others arborize profusely in stratum pyramidale. Anothermain axon crosses the hippocampal fissure, but it was impossible to follow further due to the dense axonalmeshwork in stratum moleculare (see Fig. 1A–C.). The axon carried several drumstick-like boutons (seeFig. 1G), some of them are indicated by arrows. It should be noted that none of the axon collaterals couldbe followed to their natural end, they were either cut on the section surface, or lost in dense terminal

networks. Thus, each reconstruction is very partial.


glutaraldehyde was performed, and silver intensification(Amersham) was carried out for 4 min to increase the size ofthe label. There was no difference in the specificity andsensitivity between the two methods.

Somatostatin-immunoreactive boutons were searched forin sections on copper grids, and when they formed asynapse, the GABA content of the postsynaptic element wasexamined in the adjacent, GABA-immunostained sectionson nickel grids. Profiles containing at least five times higherdensity of gold particles than neighbouring asymmetrical(presumed glutamatergic) synaptic terminals in two to threeserial sections were considered as GABA-positive. Profilescontaining the same or lower numbers of gold particles asasymmetrical terminals (background level) were consideredGABA-negative, whereas postsynaptic elements withGABA-immunoreactivity between these two levels wereregarded as unidentified. The electron micrographs weretaken on a Hitachi 7100 electron microscope.

The studies were conducted in accordance with the prin-ciples and procedures outlined in the NIH Guide for theCare and Use of Laboratory Animals.

RESULTS

General pattern of somatostatin-immunostaining in thehippocampus

Somatostatin-immunoreactive neurons showed thesame basic morphological features as observed inseveral previous studies in the rat21,34,36,38,50,64 withsome additional morphological details due to theenhanced sensitivity of the present antiserum. Briefly,cell bodies were located in stratum oriens of the CA1,stratum oriens and stratum lucidum of the CA3regions and in the hilus of the dentate gyrus (Fig.1A). Occasionally, some cells were also found instratum radiatum of the CA1 and CA3 subfields. Thevast majority of dendritic processes were confined tothe same layers as the cell bodies. At higher magni-fication, spines were observed on both the dendritesand perikarya (Fig. 1E–F). This is consistent with theobservation that dendrites of somatostatin-positivecells visualized by immunostaining for metabotropicglutamate receptor 1á subunit (mGluR 1á)13 andSPR4 are covered by long thin spines. The moststriking immunoreactive structure was the densemeshwork of axons in stratum lacunosum-moleculare of the CA1 and CA3 subfields and in theouter two thirds of stratum moleculare in the dentategyrus (Fig. 1A–B). Most of the main axons crossedthe cell body and proximal dendritic layers with onlyoccasional side-branches and turns, nevertheless rov-ing axons were seen in all layers throughout thehippocampus. These axon trunks often carried smallround varicosities resembling ‘‘en passant’’ boutons,but in some cases drumstick-like terminals were alsoobserved mainly in the hilus, stratum oriens andstratum radiatum (Fig. 1G). Main axons crossing thehippocampal fissure were also found (Fig. 1C).

Reconstruction of somatostatin-immunoreactiveneurons

CA1 subfield. Despite strong axonal immunostain-ing, it was hard to reconstruct somatostatin-positive

neurons due to the rare staining of proximal mainaxon segments. Thick axons were followed back fromstratum lacunosum-moleculare to stratum oriensin many cases, however only four of them couldbe unequivocally connected to its cell body. Threecells showed typical morphology of O-LM cells(Fig. 2A).43,47,54,61 The somata and dendrites werelocated in stratum oriens. The axons emerged from aproximal dendrite (Figs 1D, 2A), as also seen inintracellularly filled O-LM cells,29,47 crossed stratumpyramidale and radiatum, and began to arborize withmuch thinner collaterals in stratum lacunosum-moleculare. Because of the dense meshwork of axoncollaterals in this layer, it was impossible to followone reliably for a considerable distance. Therefore,we reconstructed only some of the primary andsecondary branches of the arborization in stratumlacunosum-moleculare, which clearly indicated thatindeed, somatostatin-containing O-LM neuronsproject to stratum lacunosum-moleculare in the CA1subfield (Fig. 2A). The main axons also gave somecollaterals in stratum oriens and stratum pyramidale.The axons of the reconstructed O-LM cells werestudded by ‘‘en passant’’ boutons, whereasdrumstick-like terminals were not seen on them.

The fourth cell showed some striking differencesfrom the conventional O-LM type cells (Fig. 2B).First, one of the main axons was faintly labelled anddisappeared in the alveus. Second, it had a moreprofuse axonal arborization both in stratum oriensand stratum radiatum, whereas the main axon randirectly towards the hippocampal fissure, although itwas impossible to follow through due to the verydense somatostatin-positive axon staining. More-over, the axons bore several drumstick-like boutons(Fig. 1G) in addition to the ‘‘en passants’’ varicosi-ties. Other two main axons with similar features havealso been reconstructed, but their cell bodies couldnot be recovered. These main axons were followedfrom the stratum oriens-alveus border to the hilusacross the hippocampal fissure. They arborized ex-tensively both in the hilus and in stratum radiatum ofCA1, but also gave several axon collaterals in allother layers. It is important to note that, in additionto the ‘‘en passant’’ varicosities, these axons borelarge numbers of drumstick-like boutons mainly inthe hilus and in stratum oriens. In spite of the lack ofrecovered cell bodies, these striking morphologicalsimilarities with the fourth cell and conspicuousdifferences from the three former cells strongly sug-gest that these latter axons and Cell 4 belong to asecond type of somatostatin-immunoreactiveneurons in the CA1 subfield of the hippocampus.

CA3 subfield. The stratum lacunosum-moleculareof the CA3 subfield also showed dense somatostatin-immunoreactive axonal immunostaining. Theseaxons probably originated from the somatostatin-positive somata of stratum oriens.26 In stratumradiatum, many roving axons bearing drumstick-like

42 I. Katona et al.

boutons could be observed. They were often found tocross the border of the CA1 and CA3 subfields.Many somatostatin-immunoreactive cell bodies werelocated in stratum lucidum and their axon initialsegments were only occasionally immunostained.However, these axons became usually faintly labelledbeyond the axon initial segment. Some of them couldbe followed through stratum pyramidale, but most ofthem disappeared already in stratum lucidum.

Dentate gyrus. Three somatostatin-immunoreactiveneurons were partially reconstructed in the dentategyrus, all of them showed typical morphologicalcharacteristics of HIPP cells7,15,32,59 (Fig. 3). Thesomata and the dendritic trees were located in thehilus, whereas the main axons crossed stratum granu-losum and arborized in the outer two-thirds ofstratum moleculare. They frequently carried ‘‘enpassant’’ varicosities, while drumstick-like boutonswere never found on these axons. They could be onlypartially reconstructed, because the axons had beenlost in the dense axonal meshwork in stratummoleculare. The axons emerged from the cell body intwo cases (Fig. 3, Cells 2–3) and from a proximaldendrite (Fig. 3, Cell 1) in one case. This latter cellalso had an axon collateral remaining in the hilus.Several somatostatin-immunoreactive cells had amain axon that crossed stratum granulosum intostratum moleculare, coursed parallel with stratumgranulosum, then returned back into the hilus.

Postsynaptic targets of somatostatin-immunoreactiveboutons are mainly, but not exclusively GABA-negative

In the present study we used postembeddingGABA-immunostaining66 to explore the GABAergicor non-GABAergic nature of postsynaptic targets ofsomatostatin-immunoreactive boutons. Altogether101 boutons were examined in GABA-immuno-stained sections at the electron microscopic level instratum lacunosum-moleculare of the CA1 and CA3subfields and in stratum moleculare of the dentategyrus (Table 1). All of them gave single symmetricsynapses. The vast majority (76%) of the postsynaptictargets were clearly GABA-negative (Figs 4, 5),while some (8%) of the profiles were GABA-positive(Fig. 6, for criteria see Experimental Procedures).The remaining elements (16%) were regarded asunidentified due to equivocal staining of the targets,or relatively high background staining. There wereno striking differences in the ratio of GABA-positive/negative targets among the three examined regions(Table 1). Regardless of the targets, somatostatin-immunoreactive boutons were always stronglypositive for GABA.

The postsynaptic elements were mainly dendriticshafts (72%, Figs 4B, 5A), but numerous boutonscontacted spines (27%, Fig. 5D). In one casea synapse was found on a GABA-positive somain the CA1 subfield. The precise localizationof somatostatin/GABA-positive synapses was

Fig. 3. Camera lucida drawings of somatostatin-immunoreactive neurons in the hilus of the dentate gyrus.This somatostatin-containing cell type was most frequently encountered in the hilus and showed manycharacteristics of HIPP cells. The somata and dendrites are confined to the hilus, while the axons crossingstratum granulosum arborize extensively in stratum moleculare. Two of the axons emerge from proximaldendrites (Cells 1 and 2), while the third one originates from the cell body (Cell 3). The main axon of

Cell 1 had a collateral penetrating the hilus.


Fig. 4.

44 I. Katona et al.

remarkable as they usually contacted their targetadjacent to an asymmetric, probably excitatorysynapse (Fig. 4A–B). The average bouton diameter(n=101) was 0.80&0.30 µm (major axis) and0.40&0.12 µm (minor axis). There were no differ-ences among the subfields.

Different interneuron types innervated bysomatostatin-immunoreactive boutons

Nearly 8% of the postsynaptic targets ofsomatostatin-immunoreactive boutons were clearlyGABAergic (Fig. 6). This ratio roughly correspondsto the estimated occurrence of GABAergic andnon-GABAergic elements in the hippocampus.5,71

However, recent studies described novel interneurontypes, which selectively innervate other interneurons,rather than principal cells.2,24,28 Interestingly, oneof the VIP-positive cell types has a dense dendriticarbor confined to stratum lacunosum-moleculare,and other VIP-immunoreactive cells also have exten-sive branches in this layer. Therefore, we aimed todetermine whether the somatostatin-immunoreactiveboutons terminate randomly on all types of inter-neurons in stratum lacunosum-moleculare or theyshow some degree of selectivity for specific subpopu-lations. In stratum lacunosum-moleculare andstratum moleculare nearly all of the somata anddendrites (over a hundred examined for eachcell type) were contacted by somatostatin-immunoreactive boutons, belonging to all examinedinterneuron types visualized by parvalbumin-,cholecystokinin-, SPR-, calretinin- and VIP-immunostaining in stratum lacunosum-moleculare

and stratum moleculare. To confirm our lightmicroscopic observation that somatostatin-immunoreactive boutons (black, Ni–DAB) contactinterneuron dendrites and somata (brown, DAB),some selected boutons were examined at the electronmicroscopic level as well. Five somatostatin-immunoreactive boutons were examined onparvalbumin-immunoreactive dendrites (Fig. 7) andsix boutons each on calretinin- and VIP-positivedendrites and somata (Fig. 8). All of the 17 boutonsformed symmetric synapses with their interneurontargets. Interestingly, in one case a somatostatin-immunoreactive bouton gave two synapses next toeach other, one was found on a VIP-positive somaand another on a VIP-positive proximal dendriteprobably originating from the same soma (Fig.8D–F).

DISCUSSION

On the basis of morphological characteristics,somatostatin-immunoreactive interneurons could beseparated into three distinct subtypes, (i) HIPP cellsof the dentate gyrus, (ii) conventional O-LM cells instratum oriens of the CA1 and CA3 regions and (iii)neurons with dendrites similar to O-LM cells, butwith a characteristically different axon. The first twosubtypes have dense axonal projections to stratumlacunosum-moleculare (O-LM) or stratum molecu-lare (HIPP), where entorhinal afferents terminate.The postsynaptic targets of somatostatin-immunoreactive boutons in these layers showed aquasi-random termination pattern, as most of theirtargets were GABA-negative principal cell dendrites,but all examined GABAergic interneuron types werealso innervated.

Interneurons (strata oriens/lacunosum-moleculare/hilar perforant path-associated) involved in feedbackcircuits contain somatostatin

Two of the most characteristic interneuron types,the O-LM and HIPP cells are wired with a uniquespecificity to participate in feedback inhibition ofprincipal cells. The special features indicating thisrole are the localization of dendritic tree in the samelayer where recurrent axon collaterals of local prin-cipal cells arborize, and the overlap of their axonalarbor with the distal dendritic tree of local principalcells. Direct anatomical evidence for the recur-rent activation of O-LM and HIPP cells has been

Fig. 4. Postsynaptic targets of somatostatin-immunoreactive boutons in stratum moleculare of the dentategyrus were examined by electron microscopy of serial ultrathin sections and postembedding immunogoldstaining for GABA of a random sample of somatostatin-positive boutons. The somatostatin-containingboutons, always positive for GABA (b1 in A; b2 in B), form symmetrical synaptic contacts (arrows) withGABA-negative dendritic shafts (d), as indicated by the lack of immunogold particles in the postsynaptictargets (C and E are serial sections of b1, D and F are of b2). Note that somatostatin-containing boutonsare often localized adjacent to asymmetrical synapses (open arrows), which probably originate from theentorhinal cortex. The stars label other GABA-negative profiles, whereas SOM-negative GABA-positive

boutons are indicated by asterisks. Scale bars=0.4 µm.

Table 1. Postsynaptic target selection of somatostatin-immunoreactive axon terminals

Dentate gyrus CA3 CA1 Totaln=33 n=34 n=34 n=101

GABA-negative 79% 76% 74% 76%GABA-positive 9% 6% 9% 8%Unidentified 12% 18% 17% 16%Dendritic shafts 61% 76% 79% 72%Dendritic spines 39% 24% 18% 27%Cell body 0% 0% 3% 1%

The postsynaptic targets of 101 randomly foundsomatostatin-immunoreactive boutons were identifiedelectron microscopically using postembedding immuno-gold staining for GABA.


provided by Blasco-Ibanez and Freund14 and Acsadyet al.,3 respectively, whereas local principal cells astheir major targets were identified in the presentstudy.

Several recent intracellular labelling studies ident-ified the HIPP and O-LM types of interneurons inall regions of the hippocampal formation both

in vitro26,32,47 and in vivo.15,59,61 The morphologicalcharacteristics of these cells strongly resemble thefirst two types of reconstructed somatostatin-immunoreactive neurons with somata in the hilusand in stratum oriens of CA1, and with axons instratum moleculare and stratum lacunosum-moleculare, respectively. In addition to the present

Fig. 5. Somatostatin-immunoreactive boutons (b1 in A; b2 in D) in stratum lacunosum-moleculare of theCA3 subfield form symmetric synaptic contacts (arrows) with a GABA-negative dendritic shaft (d in Aand B) and with a GABA-negative spine (s in C and D). Other GABA-negative profiles are labelled with

stars, while GABA-positive profiles are labelled with asterisks. Scale bars=0.4 µm.

46 I. Katona et al.

direct demonstration, some earlier indirect evidencealso suggests that this type of somatostatin-immunoreactive neurons in the hippocampal forma-tion can be identified as the O-LM or HIPP cells.Although the neurochemical marker content of theseintracellularly labelled neurons is largely unknown,in one case a HIPP cell was shown to containneuropeptide Y (NPY).22,60 Since NPY and somato-statin co-localize in many hilar interneurons,37 itindicates that HIPP cells express somatostatin aswell. Electrophysiological data showing that themGluR agonist 1S,3R-aminocyclopentane dicarbo-xylic acid (ACPD) generates large inward currentsand current oscillations in identified O-LM neurons47

presumably by activating mGluR1á receptors, whichdensely cover somatostatin-immunoreactive neuronsin stratum oriens,13 provide further indirect evidencethat O-LM cells and somatostatin-containing cells inCA1 are identical.

Interneurons resembling back-projection neurons alsocontain somatostatin

Most hippocampal interneuron types have an ax-onal and dendritic arbor limited to only one and thesame subfield. However, in a recent study a specificinterneuron type, which projects from stratum oriensof the CA1 subfield to the other hippocampal anddentate subfields was described.62 The present studyrevealed several features, which suggest that a smallproportion of somatostatin-immunoreactive neuronscan be classified as back-projection neurons: (i)somatostatin-immunostaining labelled several axonswhich crossed the hippocampal fissure or the borderof the CA3 and CA1 subfields; (ii) these axonsfrequently bore small drumstick-like appendages,which is a typical feature of the axon of theback-projection cell; (iii) several somatostatin-immunoreactive somata were localized in the alveus.

Fig. 6. In some cases, postsynaptic targets (d) of somatostatin-immunoreactive boutons (b1 in A and C; b2

in B) proved to be GABA-positive. These photographs were taken from stratum moleculare of the dentategyrus. The larger size of gold particles is due to silver intensification of 1 nm gold particles used in someof the experiments (see Experimental Procedures). The arrows point to symmetrical synapses formed bysomatostatin-immunoreactive boutons on GABA-positive dendritic shafts. The asterisk shows anotherGABA-positive profile in C, while GABA-negative profiles are labelled with stars. Scale bars=0.4 µm.


Fig. 7. (A) Parvalbumin-immunoreactive dendrites (d1 and d2) were frequently seen to be contacted bynumerous somatostatin-positive boutons (b1 and b2) at the light microscopic level. (B) A random samplewas cut for correlated electron microscopy, two of which are shown here (b1 and b2 in A and B). (C–D)All of the examined boutons formed symmetrical synapses (arrows) with their parvalbumin-immunoreactive targets as seen on these high power electron micrographs of b1 and b2. Open arrow showsan asymmetrical synapse located close to the somatostatin-positive bouton in the same manner as in the

case of GABA-negative postsynaptic targets. Scale bars: (A)=7.5 µm; (B)=2 µm; (C–D)=0.2 µm.

Fig. 8. (A–C) Correlated light (A) and electron micrographs (B–C) of a calretinin-immunoreactive neuron(S) receiving multiple synaptic contacts from somatostatin-immunoreactive boutons (b1 and b2). The thinarrow in B shows the invaginated nucleus of the calretinin-positive neuron, typical of interneurons (alsoshown in F in the case of a VIP-positive neuron). (C) High power electron micrograph of somatostatin-positive boutons forming symmetrical synapses with their calretinin-immunoreactive target (SCR) asindicated by thick arrows. (D–F) somatostatin-immunoreactive boutons also heavily innervate VIP-positive cells. Bouton b1 was found to contact (thick arrows) two VIP-positive profiles, a cell body (SVIP)and a proximal dendrite (d) in the same section. Scale bars: (A, D)=7.5 µm; (B)=2 µm; (E)=1 µm;

(C, F)=0.2 µm.

48 I. Katona et al.

Fig. 8.


Fig. 9.

50 I. Katona et al.

One of these neurons could be partially recon-structed, and the laminar and regional distribution ofthese axon arborization fragments resembled that ofback-projection neurons. Some of the main axonsextensively arborized in stratum oriens and stratumpyramidale, while one main axon disappeared in thealveus after myelination. It was recently shown thatimmunostaining for muscarinic type 2 receptor (M2)labels horizontal cells at the oriens/alveus border withaxons, which occasionally cross the hippocampalfissure.30 Co-localization of M2 receptors and soma-tostatin was found in a small number of interneuronsin stratum oriens of the CA1 subfield,30 which furthersuggests the existence of a somatostatin-positiveback-projection cell population. Moreover, in thisstudy some neurons expressing M2 receptor wereretrogradely labelled from the medial septum. Sinceall nonpyramidal hippocamposeptal projectionneurons contain somatostatin (Katona I., Acsady L.,Freund T.F., unpublished observation), we suggestthat these back-projection neurons, in addition totheir extensive arborization in the entire hippocam-pus, may project to the medial septum.

The postsynaptic target distribution of somatostatin-containing strata oriens/lacunosum-moleculare andhilar perforant path-associated cells

Precise knowledge of connectivity, convergenceand divergence properties is necessary for any pre-dictions to be made about the functional rolesof interneurons (or of any other cell types). There-fore we determined the postsynaptic targets ofsomatostatin-immunoreactive neurons in thoselayers, where they have the most dense axonal arbor-ization. The intrinsic origin of these fibres wasdemonstrated by our cell reconstructions and also byearlier data demonstrating the lack of extrinsicsomatostatin-containing projections to the hippo-campus.12,57 Previous studies assumed that targetsof these axons are the principal cells and showedthat somatostatin-immunoreactive neurons contact

dendritic spines and sometimes shafts both instratum moleculare and in stratum lacunosum-moleculare.10,42,49 However, recent results revealedthat in addition to the principal cell dendrites, severalinterneuron types may possess spines.4,13,25 More-over, the cells of origin of postsynaptic dendriticshafts can rarely be identified by electron microscopyalone. On the other hand, our present results revealedthat dendritic shafts occurred more frequently amongthe targets than spines, which correlates well withearlier data on postsynaptic targets of intracellularlylabeled HIPP neurons31 and O-LM cells26 and re-cently described perforant path-associated neuronslocated in stratum radiatum.69 To provide a directanswer, we used postembedding immunogold label-ling for GABA in sections containing somatostatin-immunoreactive boutons, which allowed in mostcases to identify the postsynaptic elements. Themajority of the postsynaptic targets were clearlyGABA-negative and the somatostatin-positive syn-apses were usually located adjacent to asymmetricsynapses. This suggests that the major mode oftermination of somatostatin-containing interneuronsis on distal dendrites of principal cells in closeassociation with entorhinal excitatory afferents.However, many GABA-positive profiles were alsoamong the targets, but not more than what could beexpected for a quasi-random termination pattern.Thus, the relatively higher incidence of shaft contactsmade by somatostatin-positive boutons cannot beexplained by a preference for interneuron tar-gets. Recent anatomical data have shown thatparvalbumin-immunoreactive neurons are innervatedby HIPP cells.59 Here we demonstrate morphologi-cally that, in addition to principal and basket cells,basically all interneuron types examined may beamong the targets. Electrophysiological data are alsoavailable to support an O-LM cell input to stratumlacunosum-moleculare interneurons, but the types ofcells were not identified: synchronized spontaneousGABAergic inputs that could be modulated byACPD were shared among lacunosum-moleculare

Fig. 9. The wiring and hypothesized activity pattern of the pyramidal cell–O-LM cell network at differentphases of local intracellular theta activity (insets in the top left corners): A, near the peak, and B, at morehyperpolarized phases preceding the peak. Active neurons appear as dark symbols (blue: pyramidal cells;red: O-LM interneuron), whereas they are labelled by light shades of the same colour when inactive (in B).(A) The largest proportion of pyramidal cells are active near the positive peak of intracellular thetaactivity (although even then, only about 0.1% of them fire17), thus, this is the phase (time windowindicated in insert) when O-LM interneurons have the highest probability of being activated during thetaactivity as a result of convergent input from a large number of local pyramidal cell collaterals. It iscalculated that, after pyramidal cell discharge, back-propagating action potentials would arrive to stratumlacunosum-moleculare dendrites at about the same time as the disynaptic IPSP mediated by the O-LMcells (22.5–3 ms48,67). Thus, activity-dependent associative LTP of entorhinal synaptic potentials will beblocked by feedback inhibition even if the EPSPs coincide with the back-propagating action potential. (B)Place cells fire ahead of the population53,63 at more hyperpolarized phases of local theta. Within this timewindow only a small number of pyramidal cells fire synchronously, which is less likely to trigger an actionpotential in the O-LM cell.6 Thus, if perforant path synaptic potentials will coincide with activity in theplace cell, potentiation of these synapses will be allowed by the lack of inhibition in stratum lacunosum-moleculare. By this mechanism feedback inhibition may be able to limit synaptic plasticity both in spaceand time. The same connectivity and activity pattern is proposed for the granule cell-HIPP cell network

as well.


interneurons and pyramidal cells in the CA1region.9

Functional implications

The afferent connectivity of the O-LM and HIPPcells provides predictions as to under what conditionsthese cell types are likely to be active. Both anatomi-cal14 and physiological44,59,61 data support the con-clusion that these cells in CA1 stratum oriens and inthe hilus of the dentate gyrus are driven exclusively ina feedback manner by cortical afferents, and Schaffercollateral or perforant path stimulation dischargesthem only if the stimulus is strong enough to evoke apopulation spike in CA1 or in the dentate gyrus,respectively. Thus, apparently a relatively largenumber of CA1 pyramidal cells or granule cells haveto fire together to trigger repetitive action potentialdischarge in an O-LM cell or HIPP cell. The questionarises, under what physiological conditions is thismost likely to happen? During theta activity, princi-pal cells have the largest probability of dischargearound the positive peak of intracellular thetawaves.72 Within this time window (Fig. 9A) feedbackinterneurons (i.e. O-LM and HIPP cells) will beactivated as a result of large convergence of evendistantly located principal cells68 and may preventactivity-dependent plasticity of entorhinal synapsesin stratum lacunosum-moleculare and stratummoleculare.

Back-propagating action potentials of pyramidalcells were suggested to underlie the Hebbian type ofsynaptic plasticity.45,46 On the basis of conductionvelocity of backpropagating action potentials in CA1pyramidal cells67 and the latency of disynapticinhibition48 it may be calculated that the back-propagating action potential would arrive to thedistal dendrites of CA1 pyramidal cells in approxi-mately the same time as the inhibitory postsynapticpotential (IPSP) evoked by the same action potentialvia activating an O-LM cell (2.5–3 ms after thepyramidal cell action potential in vitro), giving anopportunity to feedback inhibition to veto LTP.44

Other forms of postsynaptic LTP independent ofback-propagating action potentials will also be at-tenuated by feedback inhibition in this layer. Whenpyramidal cell firing carries some specific informa-tion, e.g., about space,52 then action potentials occurearlier than the positive peak, i.e. when pyramidalcells of the subfield are in more hyperpolarizedphases. As the animal crosses the place field of aparticular neuron, the action potentials of this placecell were shown to occur at progressively earlierphases of theta,53,63 as if its excitatory drive becamestronger, overriding greater degrees of perisomaticinhibition. We propose that firing of single place cellsalone, occurring earlier then that of the population, isunlikely to discharge the O-LM cells6,27 or HIPPcells, thus, the probability of recruiting feedbackinhibition in this phase is much lower (see Fig. 9B).

The lack of dendritic inhibition in strata lacunosum-moleculare or moleculare at this time-point wouldallow potentiation of those entorhinal synapseswhich terminate on,44 and are active synchronouslywith, this particular place cell (i.e. when the animal islocated in that place field). This way entorhinalinputs carrying place field information will becomestronger and stronger on a select population of cells,as the animal crosses repeatedly the environment.The present hypothesis suggests that the dynamicformation of place fields may begin with a quasi-random selection of cells that accidentally fire earlierthan the population in a particular place field. Thisway they escape feedback inhibition, which allowsthe potentiation of their simultaneously activeafferent synapses carrying spatial information. Thesepotentiated synapses will then be responsible forphase precession of the cell’s firing upon subsequententries of the same place field, which may lead tofurther enhancement of the excitatory drive andphase precession of the given place cell. This hypoth-esis correlates well with the notion that place fields(maps) will become more refined in time as theanimal explores a new environment.70

O-LM cells are also likely to be activated6 duringhippocampal sharp waves, when large populations ofpyramidal cells fire synchronous bursts,18 and inturn, block LTP of perforant path synapses.44 Thiswould ensure that, during sharp waves, only Schaffercollateral synapses will be potentiated.

The diagrams in Fig. 9 are clearly oversimplified,they focus on the connectivity of this particularinterneuron type and its potential role during thetaactivity. In future studies of entorhinal–hippocampalinformation transfer the functional roles of feed-forward interneuron types activated by entorhinalstimulation and terminating in strata lacunosum-moleculare, radiatum, or pyramidale,29,40,69 whichapparently have a powerful inhibitory effect,65 shouldalso be investigated.

CONCLUSION

The present data demonstrate that somatostatin-containing GABAergic interneurons driven in a feed-back manner terminate predominantly on the mostdistal dendritic segments of pyramidal cells in con-junction with entorhinal afferents (i.e. they are O-LMand HIPP cells). We propose that, if the role of thetaoscillation is to separate signal transmission frombackground firing in time,19 then these feedbackdendritic inhibitory cells may help confine timeperiods when entorhinal synapses on pyramidal cellsmay be effective. In addition, this feedback circuitmay allow associative plasticity only for cells thatcarry specific information (e.g., for place cells of thegiven place field), and limit it to the time windowwhen specific signal transmission takes place (i.e.during ‘‘phase precession’’). During hippocampalsharp waves, these interneurons may attenuate the

52 I. Katona et al.

efficacy of direct entorhinal afferents, ensuring thedominance of Schaffer collateral inputs in drivingCA1 pyramidal cells.44

Acknowledgements—We are grateful to Dr T. J. Gorcs forantisera against somatostatin, vasoactive intestinal polypep-tide and cholecystokinin, to Dr K. G. Baimbridge and toDr J. H. Rogers for antisera against parvalbumin andcalretinin, respectively, and to Dr P. Somogyi for antiseraagainst GABA. The valuable discussions with Drs A. M.

Thomson, Gy. Buzsaki, H. Markram and R. Miles concern-ing the activation of O-LM cells, and the time-course ofback-propagating action potentials and disynaptic IPSPsare highly appreciated. The authors wish to thank forpreparation of the colour figure Ms Au . L. Bodor. Theexcellent technical assistance of Mrs E. Borok, Mrs A. ZoldiSzabone and Mr Gy. Goda is also acknowledged. This workwas supported by the Howard Hughes Medical Institute,the Swiss National Science Foundation, NIMH (MH54671)and OTKA (T16942) Hungary.

REFERENCES

1. Acsady L., Arabadzisz D. and Freund T. F. (1996) Correlated morphological and neurochemical features identifydifferent subsets of vasoactive intestinal polypeptide-immunoreactive interneurons in rat hippocampus. Neuroscience73, 299–315.

2. Acsady L., Gorcs T. J. and Freund T. F. (1996) Different populations of vasoactive intestinal polypeptide-immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus.Neuroscience 73, 317–334.

3. Acsady L., Kamondi A., Sik A., Freund T. F. and Buzsaki G. (1998) GABAergic cells are the major postsynaptictargets of mossy fibers in the rat hippocampus. J. Neurosci. 18, 3386–3403.

4. Acsady L., Katona I., Gulyas A. I., Shigemoto R. and Freund T. F. (1997) Immunostaining for substance P receptorlabels GABAergic cells with distinct termination patterns in the hippocampus. J. comp. Neurol. 378, 320–336.

5. Aika Y., Ren J. Q., Kosaka K. and Kosaka T. (1994) Quantitative analysis of GABA-like-immunoreactive andparvalbumin-containing neurons in the CA1 region of the rat hippocampus using a stereological method, the disector.Expl Brain Res. 99, 267–276.

6. Ali A. B. and Thomson A. M. (1998) Facilitating pyramid to oriens-alveus interneurone inputs: dual intracellularrecordings in slices of rat hippocampus. J. Physiol., Lond. 507, 185–199.

7. Amaral D. G. (1978) A Golgi study of cell types in the hilar region of the hippocampus in the rat. J. comp. Neurol. 182,851–914.

8. Amaral D. G., Insausti R. and Campbell M. J. (1988) Distribution of somatostatin immunoreactivity in the humandentate gyrus. J. Neurosci. 8, 3306–3316.

9. Atzori M. (1996) Pyramidal cells and stratum lacunosum-moleculare interneurons in the CA1 hippocampal regionshare a GABAergic spontaneous input. Hippocampus 6, 72–78.

10. Aylward R. L. and Totterdell S. (1993) Neurons in the ventral subiculum, amygdala and entorhinal cortex whichproject to the nucleus accumbens: their input from somatostatin-immunoreactive boutons. J. chem. Neuroanat. 6,31–42.

11. Baimbridge K. G. and Miller J. J. (1982) Immunohistochemical localization of calcium-binding protein in thecerebellum, hippocampal formation and olfactory bulb of the rat. Brain Res. 245, 223–229.

12. Bakst I., Avendano C., Morrison J. H. and Amaral D. G. (1986) An experimental analysis of the origins ofsomatostatin-like immunoreactivity in the dentate gyrus of the rat. J. Neurosci. 6, 1452–1462.

13. Baude A., Nusser Z., Roberts J. D., Mulvihill E., McIlhinney R. A. and Somogyi P. (1993) The metabotropicglutamate receptor (mGluR1 alpha) is concentrated at perisynaptic membrane of neuronal subpopulations as detectedby immunogold reaction. Neuron 11, 771–787.

14. Blasco-Ibanez J. M. and Freund T. F. (1995) Synaptic input of horizontal interneurons in stratum oriens of thehippocampal CA1 subfield: structural basis of feed-back activation. Eur. J. Neurosci. 7, 2170–2180.

15. Buckmaster P. S. and Schwartzkroin P. A. (1995) Interneurons and inhibition in the dentate gyrus of the rat in vivo.J. Neurosci. 15, 774–789.

16. Buhl E. H., Halasy K. and Somogyi P. (1994) Diverse sources of hippocampal unitary inhibitory postsynapticpotentials and the number of synaptic release sites. Nature 368, 823–828.

17. Buzsaki G. and Eidelberg E. (1983) Phase relations of hippocampal projection cells and interneurons to theta activityin the urethane anesthetized rat. Brain Res. 266, 334–347.

18. Buzsaki G. (1986) Hippocampal sharp waves: their origin and significance. Brain Res. 398, 242–252.19. Buzsaki G. and Chrobak J. J. (1995) Temporal structure in spatially organized neuronal ensembles: a role for

interneuronal networks. Curr. Opin. Neurobiol. 5, 504–510.20. 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 inAlzheimer-type dementia. J. comp. Neurol. 260, 201–223.

21. Feldman S. C., Dreyfus C. F. and Lichtenstein E. S. (1982) Somatostatin neurons in the rodent hippocampus: anin vitro and in vivo immunocytochemical study. Neurosci. Lett. 33, 29–34.

22. Freund T. F. and Buzsaki G. (1996) Interneurons of the hippocampus. Hippocampus 6, 347–470.23. Gulyas A. I., Gorcs T. J. and Freund T. F. (1990) Innervation of different peptide-containing neurons in the

hippocampus by GABAergic septal afferents. Neuroscience 37, 31–44.24. Gulyas A. I., Hajos N. and Freund T. F. (1996) Interneurons containing calretinin are specialized to control other

interneurons in the rat hippocampus. J. Neurosci. 16, 3397–3411.25. Gulyas A. I., Miettinen R., Jacobowitz D. M. and Freund T. F. (1992) 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.

26. Gulyas A. I., Miles R., Hajos N. and Freund T. F. (1993) Precision and variability in postsynaptic target selection ofinhibitory cells in the hippocampal CA3 region. Eur. J. Neurosci. 5, 1729–1751.

27. Gulyas A. I., Miles R., Sık A., Toth K., Tamamaki N. and Freund T. F. (1993) Hippocampal pyramidal cells exciteinhibitory neurons through a single release site. Nature 366, 683–687.


28. Hajos N., Acsady L. and Freund T. F. (1996) Target selectivity and neurochemical characteristics of VIP-immunoreactive interneurons in the rat dentate gyrus. Eur. J. Neurosci. 8, 1415–1431.

29. Hajos N. and Mody I. (1997) Synaptic communication among hippocampal interneurons: properties of spontaneousIPSCs in morphologically identified cells. J. Neurosci. 17, 8427–8442.

30. Hajos N., Papp Cs. E., Acsady L., Levey A. I. and Freund T. F. (1998) Distinct interneuron types express M2muscarinic receptor immunoreactivity on their dendrites or axon terminals in the hippocampus. Neuroscience 82,355–376.

31. Halasy K. and Somogyi P. (1993) Subdivisions in the multiple GABAergic innervation of granule cells in the dentategyrus of the rat hippocampus. Eur. J. Neurosci. 5, 411–429.

32. Han Z. S., Buhl E. H., Lorinczi Z. and Somogyi P. (1993) A high degree of spatial selectivity in the axonal anddendritic domains of physiologically identified local-circuit neurons in the dentate gyrus of the rat hippocampus.Eur. J. Neurosci. 5, 395–410.

33. Hodgson A. J., Penke B., Erdei A., Chubb I. W. and Somogyi P. (1985) Antisera to gamma-aminobutyric acid. I.Production and characterization using a new model system. J. Histochem. Cytochem. 33, 229–239.

34. Johansson O., Hokfelt T. and Elde R. P. (1984) Immunohistochemical distribution of somatostatin-like immuno-reactivity in the central nervous system of the adult rat. Neuroscience 13, 265–339.

35. Katsumaru H., Kosaka T., Heizmann C. W. and Hama K. (1988) Immunocytochemical study of GABAergic neuronscontaining the calcium-binding protein parvalbumin in the rat hippocampus. Expl Brain Res. 72, 347–362.

36. Kohler C. and Chan-Palay V. (1982) Somatostatin-like immunoreactive neurons in the hippocampus: an immuno-cytochemical study in the rat. Neurosci. Lett. 34, 259–264.

37. Kohler C., Eriksson L. G., Davies S. and Chan-Palay V. (1987) Co-localization of neuropeptide tyrosine andsomatostatin immunoreactivity in neurons of individual subfields of the rat hippocampal region. Neurosci. Lett. 78,1–6.

38. Kosaka T., Wu J. Y. and Benoit R. (1988) GABAergic neurons containing somatostatin-like immunoreactivity in therat hippocampus and dentate gyrus. Expl Brain Res. 71, 388–398.

39. Kunkel D. D. and Schwartzkroin P. A. (1988) Ultrastructural characterization and GAD co-localization ofsomatostatin-like immunoreactive neurons in CA1 of rabbit hippocampus. Synapse 2, 371–381.

40. Lacaille J. C. and Schwartzkroin P. A. (1988) Stratum lacunosum-moleculare interneurons of hippocampal CA1region. I. Intracellular response characteristics, synaptic responses, and morphology. J. Neurosci. 8, 1400–1410.

41. Lantos T. A., Gorcs T. J. and Palkovits M. (1995) Immunohistochemical mapping of neuropeptides in thepremamillary region of the hypothalamus in rats. Brain. Res. Rev. 20, 209–249.

42. Leranth C., Malcolm A. J. and Frotscher M. (1990) Afferent and efferent synaptic connections of somatostatin-immunoreactive neurons in the rat fascia dentata. J. comp. Neurol. 295, 111–122.

43. Lorente de No (1934) Studies on the structure of the cerebral cortex—II. Continuation of the study of the ammonicsystem. J. psychol. Neurol. 46, 113–177.

44. Maccaferri G. and McBain C. J. (1995) Passive propagation of LTD to stratum oriens-alveus inhibitory neuronsmodulates the temporoammonic input to the hippocampal CA1 region. Neuron 15, 137–145.

45. Magee J. C. and Johnston D. (1997) A synaptically controlled, associative signal for Hebbian plasticity inhippocampal neurons [see comments]. Science 275, 209–213.

46. Markram H., Lubke J., Frotscher M. and Sakmann B. (1997) Regulation of synaptic efficacy by coincidence ofpostsynaptic APs and EPSPs [see comments]. Science 275, 213–215.

47. McBain C. J., DiChiara T. J. and Kauer J. A. (1994) Activation of metabotropic glutamate receptors differentiallyaffects two classes of hippocampal interneurons and potentiates excitatory synaptic transmission. J. Neurosci. 14,4433–4445.

48. Miles R. (1990) Synaptic excitation of inhibitory cells by single CA3 hippocampal pyramidal cells of the guinea-pigin vitro. J. Physiol., Lond. 428, 61–77.

49. Milner T. A. and Bacon C. E. (1989) Ultrastructural localization of somatostatin-like immunoreactivity in the ratdentate gyrus. J. comp. Neurol. 290, 544–560.

50. Morrison J. H., Benoit R., Magistretti P. J., Ling N. and Bloom F. E. (1982) Immunohistochemical distribution ofpro-somatostatin-related peptides in hippocampus. Neurosci. Lett. 34, 137–142.

51. Nunzi M. G., Gorio A., Milan F., Freund T. F., Somogyi P. and Smith A. D. (1985) Cholecystokinin-immunoreactivecells form symmetrical synaptic contacts with pyramidal and nonpyramidal neurons in the hippocampus. J. comp.Neurol. 237, 485–505.

52. O’Keefe J. and Dostrovsky J. (1971) The hippocampus as a spatial map: preliminary evidence from unit activity infreely moving rats. Brain Res. 34, 171–175.

53. O’Keefe J. and Recce M. L. (1993) Phase relationship between hippocampal place units and the EEG theta rhythm.Hippocampus 3, 317–330.

54. Cajal R. y. (1989) Estructura del asta de Ammon y fascia dentata. Ann. Soc. Esp. Hist. Nat. 22 1893.55. Rogers J. H. (1989) Immunoreactivity for calretinin and other calcium-binding proteins in cerebellum. Neuroscience

31, 711–721.56. Schmechel D. E., Vickrey B. G., Fitzpatrick D. and Elde R. P. (1984) GABAergic neurons of mammalian cerebral

cortex: widespread subclass defined by somatostatin content. Neurosci. Lett. 47, 227–232.57. Senut M. C., Menetrey D. and Lamour Y. (1989) Cholinergic and peptidergic projections from the medial septum and

the nucleus of the diagonal band of Broca to dorsal hippocampus, cingulate cortex and olfactory bulb: a combinedwheatgerm agglutinin–apohorseradish peroxidase–gold immunohistochemical study. Neuroscience 30, 385–403.

58. Shigemoto R., Nakaya Y., Nomura S., Ogawa-Meguro R., Ohishi H., Kaneko T., Nakanishi S. and Mizuno N. (1993)Immunocytochemical localization of rat substance P receptor in the striatum. Neurosci. Lett. 153, 157–160.

59. Sık A., Penttonen M. and Buzsaki G. (1997) Interneurons in the hippocampal dentate gyrus: an in vivo intracellularstudy. Eur. J. Neurosci. 9, 573–588.

60. Sık A., Penttonen M., Ylinen A. and Buzsaki G. (1995) Intracellularly labeled hippocampal hilar interneurons in vivo.Soc. Neurosci. Abstr. 21, 930.

61. Sık A., Penttonen M., Ylinen A. and Buzsaki G. (1995) Hippocampal CA1 interneurons: an in vivo intracellularlabeling study. J. Neurosci. 15, 6651–6665.

54 I. Katona et al.

62. Sık A., Ylinen A., Penttonen M. and Buzsaki G. (1994) Inhibitory CA1–CA3–hilar region feedback in thehippocampus. Science 265, 1722–1724.

63. Skaggs W. E., McNaughton B. L., Wilson M. A. and Barnes C. A. (1996) Theta phase precession in hippocampalneuronal populations and the compression of temporal sequences. Hippocampus 6, 149–172.

64. Sloviter R. S. and Nilaver G. (1987) Immunocytochemical localization of GABA-, cholecystokinin-, vasoactiveintestinal polypeptide-, and somatostatin-like immunoreactivity in the area dentata and hippocampus of the rat.J. comp. Neurol. 256, 42–60.

65. Soltesz I. (1995) Brief history of cortico-hippocampal time with a special reference to the direct entorhinal input toCA1. Hippocampus 5, 120–124.

66. Somogyi P. and Hodgson A. J. (1985) Antisera to gamma-aminobutyric acid. III. Demonstration of GABA inGolgi-impregnated neurons and in conventional electron microscopic sections of cat striate cortex. J. Histochem.Cytochem. 33, 249–257.

67. Spruston N., Schiller Y., Stuart G. and Sakmann B. (1995) Activity-dependent action potential invasion and calciuminflux into hippocampal CA1 dendrites. Science 268, 297–300.

68. Tamamaki N. and Nojyo Y. (1990) Disposition of the slab-like modules formed by axon branches originating fromsingle CA1 pyramidal neurons in the rat hippocampus. J. comp. Neurol. 291, 509–519.

69. Vida I., Halasy K., Szinyei Cs., Somogyi P. and Buhl E. H. (1998) Unitary IPSPs evoked by interneurons at thestratum radiatum-stratum lacunosum-moleculare border in the CA1 area of the rat hippocampus in vitro. J. Physiol.,Lond. 506, 755–773.

70. Wilson M. A. and McNaughton B. L. (1993) Dynamics of the hippocampal ensemble code for space. Science 261,1055–1058.

71. Woodson W., Nitecka L. and Ben-Ari Y. (1989) Organization of the GABAergic system in the rat hippocampalformation: a quantitative immunocytochemical study. J. comp. Neurol. 280, 254–271.

72. Ylinen A., Soltesz I., Bragin A., Penttonen M., Sık A. and Buzsaki G. (1995) Intracellular correlates of hippocampaltheta rhythm in identified pyramidal cells, granule cells, and basket cells. Hippocampus 5, 78–90.

(Accepted 21 May 1998)


Postsynaptic targets of somatostatin-immunoreactive interneurons in the rat hippocampus

Documents