Median raphe serotonergic innervation of medial septum/diagonal band of Broca (MSDB) parvalbumin-containing neurons: Possible involvement of the MSDB in the desynchronization of the

Median Raphe Serotonergic Innervationof Medial Septum/Diagonal Band of Broca

(MSDB) Parvalbumin-ContainingNeurons: Possible Involvement

of the MSDB in the Desynchronizationof the Hippocampal EEG

CSABA LERANTH1 AND ROBERT P. VERTES2

1Department of Obstetrics and Gynecology and Section of Neurobiology, Yale University,School of Medicine, New Haven, Connecticut 06520

2Center for Complex Systems, Florida Atlantic University, Boca Raton, Florida 33431

ABSTRACTActivation of median raphe serotonergic neurons results in the desynchronization of

hippocampal electroencephalographic (EEG) activity. This could be a direct effect, becauseserotonin (5-HT) fibers terminate on a specific population of hippocampal interneurons. Onthe other hand, it could be an indirect action through the medial septum/diagonal band ofBroca (MSDB) pacemaker cells, because, in addition to previously described inhibitory effects,excitatory actions of 5-HT have been demonstrated on MSDB g-aminobutyric acid (GABA)-containing neurons through 5-HT2A receptors. Electron microscopic double immunostainingfor Phaseolus vulgaris-leucoagglutinin (PHA-L) injected into the median raphe (MR) andparvalbumin, choline acetyltransferase, or calretinin as well as double immunostaining for5-HT and parvalbumin, and colocalization for parvalbumin and 5-HT2A receptors were done inrats. The results demonstrated that: 1) MR axons form perisomatic and peridendritic basketsand asymmetric synaptic contacts on MSDB parvalbumin neurons; 2) these fibers do notterminate on septal cholinergic and calretinin neurons; 3) 5-HT fibers form synapses identicalto those formed by PHA-L-immunolabeled axons with parvalbumin neurons; and 4) MSDBparvalbumin cells contain 5-HT2A receptors. These observations indicate that 5-HT has a dualaction on the activity of hippocampal principal cells: 1) an inhibition of the input sector byactivation of hippocampal GABA neurons that terminate exclusively on apical dendrites ofpyramidal cells, and 2) a disinhibition of the output sector of principal neurons. MSDBparvalbumin-containing GABAergic neurons specifically innervate hippocampal basket and chan-delier cells. Thus, 5-HT-elicited activation of MSDB GABAergic neurons will result in a powerfulinhibition of these GABAneurons. J. Comp. Neurol. 410:586–598, 1999. r 1999 Wiley-Liss, Inc.

Indexing terms: Phaseolus vulgaris-leucoagglutinin; double immunostaining; acetylcholine;

g-aminobutyric acid; disinhibition; serotonin 2A receptor

It is well established that the medial septum/diagonalband of Broca (MSDB) serves a critical role in the control ofthe hippocampal electroencephalogram (EEG), i.e., hippo-campal synchronization (theta rhythm) and desynchroni-zation. The MSDB contains a population of g-aminobutyricacidergic (GABAergic) and cholinergic cells that fire in burstsin synchrony with the theta rhythm to pace this rhythm(pacemaker cells). The disruption of the rhythmic dischargeof the pacemaker cells disrupts theta or desynchronizes thehippocampal EEG (for review, see Vertes and Kocsis, 1997).

An extensive body of evidence indicates that the seroto-nin (5-HT)-containing median raphe nucleus (MR) is in-

Grant sponsor: National Institutes of Health; Grant numbers: NS26068and NS35883.

*Correspondence to: Csaba Leranth, M.D., Ph.D., Department of Obstet-rics and Gynecology, Yale University, School of Medicine, 333 Cedar Street,FMB 328, New Haven, CT 06520-8063. E-mail: [email protected]

Received 11 May 1998; Revised 1 February 1999; Accepted 5 March 1999

THE JOURNAL OF COMPARATIVE NEUROLOGY 410:586–598 (1999)

r 1999 WILEY-LISS, INC.

volved directly in the desynchronization of the hippocam-pal EEG. It has been shown that MR stimulationdesynchronizes the hippocampal EEG (Macadar et al.,1974; Assaf and Miller, 1978; Vertes, 1981). MR lesionsproduce an ongoing theta rhythm independent of behavior(Maru et al., 1979), and injections of various pharmacologi-cal agents into the MR in urethane-anesthetized rats(Vertes et al., 1994; Kinney et al., 1994, 1995) that inhibitthe activity of serotonergic MR neurons or intravenousinjections of 5-HT1A receptor agonists in behaving cats(Marrosu et al., 1996) generate theta rhythm (i.e., blockdesynchronization).

The MR is the source of pronounced projections to boththe septum and the hippocampus (Moore and Halaris,1975; Azmitia and Segal, 1978; Vertes and Martin, 1988)and can exert desynchronizing effects on the hippocampalEEG through one or both routes. MR fibers projecting tothe hippocampus predominantly target calbindin- andcalretinin (CR)-containing interneurons of the hippocam-pus (Freund et al., 1990; Halasy et al., 1992). Evidencesuggests that desynchronizing actions of the MR on thehippocampal EEG are mediated primarily through theMSDB. Assaf and Miller (1978) demonstrated that MRstimulation disrupts the rhythmic discharge of septalpacemaker cells and desynchronizes the hippocampal EEG,and Kinney et al. (1996) showed that the suppression ofMR 5-HT cells rhythmically activates MSDB pacemakercells and generates theta rhythm.

In further support of a 5-HT influence on the MSDB indesynchronization of the hippocampal EEG are the impor-tant findings of Alreja and colleagues (Alreja, 1996; Liuand Alreja, 1997). Recently, they demonstrated that 5-HTexcites putative GABA-containing MSDB cells primarilythrough 5-HT2A receptors, and these GABAergic MSDBcells, in turn, inhibit subsets of GABAergic/cholinergicpacemaker neurons of the MSDB. These findings suggestan MR 5-HT activation of GABAergic MSDB neurons andsubsequent suppression of septal GABAergic/cholinergicpacemaker cells in the desynchronization of the hippocam-pal EEG.

Although MR fibers distribute densely to the MSDB, toour knowledge, only a few preliminary reports have exam-ined MR-septal projections at the ultrastructural level(Honda and Semba, 1993; Milner and Veznedaroglu, 1993).The primary aim of the present study was to examine themode of termination of MR fibers on identified cell types ofthe MSDB to further elucidate the circuitry controlling thehippocampal EEG.

MATERIALS AND METHODS

Animals

Seventeen adult Sprague-Dawley male and female rats(body weight, 300–320 g) were used in this study. Twelveanimals were devoted to the anterograde tracer experi-ments, three rats to the 5-HT and parvalbumin (PA)double-immunostaining experiments, and two to the colo-calization study.Animals were kept under standard labora-tory conditions, with tap water and regular rat chow adlibitum, in a 12 hour light, 12 hour dark cycle. All of theexperiments performed were in accordance with NationalInstitutes of Health Guidelines and were approved by theYale Animal Care and Use Committee, and all efforts weremade to minimize animal suffering.

Anterograde labeling of MR-septal afferents

Phaseolus vulgaris-leucoagglutinin (PHA-L; 2.5% in sa-line; Vector Laboratories, Burlingame, CA) was appliediontophoretically (5 µA positive current; 2 seconds on, 2seconds off for 15–30 minutes) through a micropipette(25–35 µm inner diameter) into the MR (lateral, 0 mm;ventral, 8.0–8.5 mm; and anteroposterior, 27.3–8.3 mmbehind Bregma). Animals were killed 9–11 days after thePHA-L injection.

Tissue preparation

Animals were killed under ether anesthesia by transcar-dial perfusion of 50 ml heparinized saline followed by afixative containing 0.1% glutaraldehyde, 4% paraformalde-hyde, and 15% picric acid in 0.1 M phosphate buffer (PB),pH 7.35. Tissue blocks were dissected out and postfixed for2 hours in the same but glutaraldehyde-free fixative.Then, they were rinsed in several changes of ice-cold PB,and 60 µm Vibratome (Lancer, St. Louis, MO) sectionscontaining the mesencephalon and septal complex werecut. Subsequently, the sections were treated with 1%sodium borohydride in PB for 10–30 minutes to eliminateunbound aldehydes from the tissue (Kosaka et al., 1986).

Double immunostaining for PHA-L and eitherPA, choline acetyltransferase, or CR

Sections for light microscopy were incubated first in 10%normal goat serum containing 1% Triton X-100 in PB for30 minutes. Sections for electron microscopy were trans-ferred into vials containing 0.5 ml 10% sucrose (in PB) andrapidly frozen by immersing the vial in liquid nitrogen,then thawed to room temperature, and washed in PB.Thereafter, sections for both light and electron microscopywere incubated in a mixture of primary antisera directedagainst PHA-L (biotinylated goat anti-PHA-L; 1:200, Vec-tor Laboratories) and either PA (mouse anti-PA; 1:5,000;Sigma, St. Louis, MO), choline acetyltransferase (ChAT;rat anti-ChAT; 1:10, Boehringer Mannheim, Indianapolis,IN), or CR (rabbit anti-CR; 1:6,000; Rogers, 1987) for 48hours at 4°C. Sections were then placed in a mixturecontaining avidin biotinylated-horseradish peroxidase com-plex (ABC; 1:200; Vector Laboratories) and either goatanti-mouse immunoglobulin G (IgG) for PA, rabbit anti-ratIgG for ChAT, or goat anti-rabbit IgG for CR (1:50; DAKO,Carpinteria, CA) for 3 hours at 20°C. The first immunoper-oxidase (for PHA-L) was developed by using a dark-blue toblack, nickel-intensified diaminobenzidine (Ni-DAB) reac-tion (15 mg DAB, 0.12 mg glucose oxidase, 12 mg ammo-nium chloride, 12 mg nickel-ammonium sulfate, and 60 mgb-D-glucose in 30 ml PB for 10–15 minutes). Then, thesections were further incubated in either mouse (for PA)rat (for ChAT), or rabbit (for CR) peroxidase-antiperoxi-dase complex (PAP; 1:100; DAKO) for 2 hours at roomtemperature, and the tissue-bound peroxidase was visual-ized by using a brown DAB reaction (15 mg DAB, 165 µl0.3% H2O2 in 30 ml PB). Sections were washed thoroughlyin PB between each incubation step. All antibody dilutionswere made in PB containing 0.1% sodium azide. Sectionsfor light microscopy were mounted on gelatin-coated slides,dehydrated, and mounted in Permount. For electron mi-croscopy, putative synaptic contacts between PHA-L-immunoreactive axons and immunolabeled septal neuronsfirst were photographed on wet-mounted sections (Fig. 3a).Then, from each animal, seven to nine Vibratome sectionscontaining putative synaptic contacts were osmicated (1%

5-HT INNERVATION OF MEDIAL SEPTAL PARVALBUMIN NEURONS 587

OsO4 in PB; 30 minutes), dehydrated through increasingethanol concentrations (the 70% ethanol contained 1%uranyl acetate; 30 minutes), and embedded in Durcupan(ACM Fluka, Buchs, Switzerland). Ribbons of ultrathinsections cut from the putative synaptic contacts wereexamined in a Philips (Mahwah, NJ) CM-10 electronmicroscope.

Double immunostaining for 5-HT and PA

Vibratome sections for light microscopy were Tritontreated, whereas sections for electron microscopy werefreeze-thaw treated (see above). After several rinses in PB,sections were incubated (48 hours at 4°C) in a mixture ofrabbit anti-5-HT (1:3,000; Incstar, Stillwater, MN) andmouse anti-PA (1:5,000; Sigma). Sections were furtherincubated (2 hours at room temperature) in a mixture ofbiotinylated goat anti-rabbit IgG (1:250; Vector Laborato-ries) and goat anti-mouse IgG (1:50; DAKO). This step wasfollowed by incubation (2 hours at room temperature) inABC Elite (1:200; Vector Laboratories) and a Ni-DABreaction to visualize 5-HT fibers. Thereafter, sections wereincubated in mouse PAP (1:100; overnight at 4°C; DAKO),which was followed by a DAB reaction. Between eachincubation step, sections were rinsed three times for 15minutes each in PB. Vibratome sections for light andelectron microscopy (seven to nine from each animal) weretreated further as described above.

Colocalization of 5-HT2A receptorsand PA or ChAT

The very specific ‘‘mirror’’ colocalization technique ofKosaka et al. (1985) was used. Consecutive, 50-µm Vibra-tome sections of the septal complex were placed in alter-nate wells of a tissue culture plate. Light-microscopicsingle immunostaining for 5-HT2A and PA or 5-HT2A and

ChAT was performed on the adjacent sections. The pri-mary and secondary antisera used for 5-HT2A were mouseanti-5-HT2A (Jakab and Goldman-Rakic, 1998; clone G186–1117; Pharmingen, San Diego, CA; 1:1,000 in PB overnightat room temperature) and biotinylated horse anti-mouseIgG (1:250 in PB; 2 hours at room temperature; VectorLaboratories), respectively. Immunostaining for PA andChAT was carried out as described above. Visualization oftissue-bound peroxidase was accomplished using a DABreaction. Consecutive sections were mounted on slides(two sections per slide, one immunostained for 5-HT2A andone for PA or ChAT) such that the posterior side of the firstsection and the anterior side of the second section wereface up. Light-microscopic examination was then made on12 pairs of sections (taken from each animal), and videophotomicrographs were taken from consecutive sections ofthe same identified perikarya to determine colocalizationof the two substances.

Control experiments

All of the antisera and antibodies used in this study arewidely used and well characterized. Therefore, only thespecificity of the double immunostainings was tested.Double immunostainings were performed as describedabove; however, one or both of the primary antisera wereomitted or replaced with corresponding normal sera. Inthese experiments, no or single immunostaining could beobserved.

The placement of the PHA-L injections also was con-trolled. Alternate Vibratome sections of the mesencepha-lon were single immunostained for PHA-L or 5-HT. Onlythose septal sections from animals in which the epicenterof the PHA-L injection was exactly in the MR (Fig. 1) wereused.

Fig. 1. a,b: Low-power light photomicrographs show a typicalPhaseolus vulgaris-leucoagglutinin (PHA-L) injection (b) and seroto-nin (5-HT)-immunoreactive neurons (a) on consecutive Vibratome

sections immunostained for PHA-L and 5-HT, respectively. Note thatthe epicenter of the PHA-L injection is in the median raphe (Mr;arrowheads). Aq, aqueduct. Scale bar 5 1,000 µm.

588 C. LERANTH AND R.P. VERTES

Videophotography

Color light-microscopic images, as well as the black-and-white images used in the colocalization studies werecaptured with a CCD video camera connected to a Matrox(Chicago, IL) digitizing board of a computer and stored asTIF files. By using the Adobe PhotoShop 4 program (AdobeSystems, Mountain View, CA), the contrast and sharpnessof the images were increased by 5–10%. The same programwas used to prepare the composite pictures (Fig. 2) and togenerate the labeling.

RESULTS

MR fibers innervate only septalPA-containing neurons

Dark-blue to black, Ni-DAB-labeled, PHA-L-containingfibers and boutons were seen throughout the septal com-plex. Light microscopic analysis of serial Vibratome sec-tions throughout the injection site and the septum re-vealed that the anterogradely labeled fibers reach theseptum through the medial forebrain bundle. The densestpopulation of these profiles could be observed in the lateralaspects of the MSDB. However, less densely packed,PHA-L-immunoreactive fibers and boutons were found ina homogeneous distribution in the lateral septal area, andPHA-L-labeled fibers were detected in the midsagittal partof the medial septum. The location of the brown DAB-labeled, PA-, ChAT-, and CR-immunoreactive neuronscorresponded to earlier descriptions (see, e.g., Kiss et al.,1997).

The immunostaining for ChAT and CR was excellent.Fine dendrites of these neurons were quite visible, andboth ChAT- and CR-immunopositive cells were sur-rounded by a PHA-L-immunoreactive network. However,no putative synaptic contacts between PHA-L-labeledaxons and these neurons were found. This observation,however, does not exclude the possibility that very fine,unlabeled dendritic processes of these cells are contactedby MR afferents. On the other hand, large populations ofPA-containing neurons located in the medial septum aswell as in the vertical and horizontal diagonal band weresurrounded by PHA-L-labeled, basket-like structures, andtheir somata and dendrites were putative synaptic targetsof PHA-L-labeled axon terminals. PHA-L-containing bou-tons contacted both the somata and dendritic processes ofthese cells (Fig. 2). The majority of these putative synapticcontacts (64.6% 6 8 of 1,950 contacts counted on 65 sec-tions) were observed on dendrites.

It is important to note that both the location and thedensity of PHA-L-immunoreactive fibers, as well as ofMSDB PA neurons that were contacted by these antero-gradely labeled MR fibers, were different in each experi-ment. However, if PHA-L-labeled baskets were seen, thenthey always were associated with PA neurons. It appearedthat the location of the anterogradely labeled fibers wasdetermined by the placement of the PHA-L microinjec-tions, with their anteroposterior position being an impor-tant factor. Thus, because none of the PHA-L injectionsresulted in a complete labeling of all of the MR neurons, ameaningful statistical analysis could not be performedwith regard to the percentage of MSDB PA neurons thatwere contacted by MR fibers.

Electron microscopic analysis of serial sections cut fromthese putative synaptic contacts (a total of 109) showed

that the relatively large (1.0–1.4 µm) PHA-L-labeled bou-tons form exclusively asymmetric synaptic contacts withboth the somata and dendrites (Fig. 3) of PA-immunoreac-tive neurons. Furthermore, the diameter of the synapticmembrane specializations was short (0.2–0.3 µm), and, onoccasion, 14 of 109 boutons examined, even after surveyingcomplete sets of serial sections, the synaptic contact couldnot be found. These nonsynaptic contacts always wereassociated with the somata of PA-containing neurons.

5-HT innervation of MSDB PA neurons

The previous experiment demonstrated that MR neu-rons project to the MSDB and terminate on PA-containingneurons. Because cells other than 5-HT also project fromthe raphe to the septum (Semba et al., 1988), a compara-tive analysis between the innervation of MSDB PA neu-rons by anterogradely PHA-L-labeled and by 5-HT-containing fibers was necessary. PHA-L fibers originatingin the MR were found to innervate only the PA-containingcells of the MSDB. Therefore, double immunostaining foronly 5-HT and PA was performed. Under the light micro-scope, the distribution pattern of 5-HT fibers in the septumwas very similar to that of PHA-L-labeled axons. However,the density of 5-HT-containing fibers was much higherthan that of anterogradely labeled axons. 5-HT-immunore-active fibers, similar to the PHA-L fibers, seemed to formbaskets exclusively around PA-containing somata anddendrites, and 5-HT-containing boutons established puta-tive axosomatic (Fig. 4a) and, more frequently, axoden-dritic synaptic contacts with these neurons. It appearedthat all of the PA-immunopositive cells located on thesurface of the Vibratome sections were contacted by 5-HT-containing boutons. Under the electron microscope, themajority of 5-HT-immunoreactive profiles were myelin-ated axons. Furthermore, similar to the synaptic contactsof the PHA-L-labeled boutons, the large 5-HT-immunoposi-tive axon terminals were found to form exclusively asym-metric synaptic contacts, and the length of the synapticmembrane specializations was short as well (Figs. 4, 5).Very frequently, although complete sets of serial sections ofputative synaptic contacts (a total of 27 putative contactsexamined) between 5-HT-immunoreactive boutons and thesomata of PA-containing neurons were analyzed, no synap-tic membrane specializations could be observed (Fig. 4b).In contrast, all of the putative axodendritic synapsesexamined (a total of 62) could be verified under theelectron microscope (Figs. 4d, 5a–c).

MSDB PA neurons contain 5-HT2A receptors

5-HT2A receptor-containing neurons were found through-out the MSDB. The majority of these cells were located inthe lateral part of the medial septum, the vertical limb ofthe diagonal band, and the dorsal portion of the horizontallimb of the diagonal band (Fig. 6a). It appeared that all ofthe PA-containing neurons (Fig. 6b) contained 5-HT2Areceptors (Fig. 6). In contrast, when the same cells werestained in different sections for 5-HT2A receptors andChAT, none of the 5-HT2A receptor-immunoreactive neu-rons had corresponding ChAT-positive counterparts in theadjacent section.

DISCUSSION

The major findings of this study were that 1) axonsoriginating in the MR form perisomatic and peridendritic


Fig. 2. a–e: Light photomicrographs demonstrate the result ofdouble immunostaining for PHA-L and parvalbumin (PA) in themedial septum/diagonal band of Broca (MSDB). Varicose PHA-L-

immunoreactive fibers (arrowheads) form putative synaptic contactswith the somata (a,b,d) and with initial (a,e) and proximal (c) dendriticsegments of PA-containing neurons. D, dendrite. Scale bars 5 10 µm.


Fig. 3. a–d: Light (a) and electron photomicrographs (b–d) showthe results of correlated light and electron microscopic double immuno-staining for PHA-L and PA. Low-power (b) and high-power magnifica-tions (c,d) of axodendritic putative synaptic contacts of PHA-L-labeledboutons shown on the light photomicrographs (arrows) are seen on the

electron photomicrographs. Serial sections (c,d) of one of these PHA-L-containing axon terminals (long arrow in a and b) demonstrate thatthese contacts represent asymmetric synaptic membrane specializa-tions (arrowheads). Scale bars 5 10 µm in a; 1 µm in b–d.

Fig. 4. a–e: Light (a) and electron photomicrographs (b–f) show theresults of correlated light and electron microscopic double immuno-staining for 5-HT and PA (photomicrograph in the lower right cornerbelongs to b and shows a part of neuron N1). Electron photomicro-graphs demonstrate asymmetric synaptic contacts (arrowheads) be-tween 5-HT-immunoreactive axon terminals (numbered 1–5 in a) andPA-immunoreactive neurons (N1–N3 in a). Bouton 1 contacts a

PA-containing dendrite (b,d); the others (boutons 3–5) form axoso-matic synaptic contacts. Although a complete set of serial sections ofbouton 2 was analyzed, no synaptic membrane specialization wasfound between this 5-HT-immunoreactive bouton and PA-immunoreac-tive cell N1. The photomicrograph in d is a high-power magnificationof the axodendritic synaptic contact of 5-HT bouton 1. Scale bars 5 10µm in a; 1 µm in b–f.

Fig. 5. a–c: Electron photomicrographs demonstrate three of the very frequently observed asymmetricaxodendritic synaptic contacts (arrowheads) between heavily nickel-intensified diaminobenzidine-labeled5-HT axon terminals and PA-containing dendrites (D). Scale bars 5 1 µm.


Fig. 6. Light micrographs show adjacent sections of the same neurons immunoreactive for 5-HT2Areceptors (a) and PA (b) in the horizontal limb of the diagonal band. Note that all of the 5-HT2A-containingcells (a; 1–12) located on the surface of the Vibratome section are immunoreactive for PA (b; 1–12).Asterisks label identical capillaries. Scale bar 5 20 µm.


baskets and asymmetric synaptic contacts on MSDB PA-immunoreactive neurons; 2) MR fibers may not terminateon septal cholinergic and CR-containing neurons; 3) 5-HT-immunoreactive fibers form synapses identical to thoseformed by PHA-L-labeled axons with PA-immunoreactiveMSDB neurons; and 4) MSDB PA-immunoreactive cellscolocalize 5-HT2A receptors. There was a marked overlapin the distribution patterns of PHA-L-labeled elementsand 5-HT-immunoreactive terminals in the MSDB, indicat-ing that they originate from the same source. In effect, thissuggests that serotonergic fibers that innervate the MSDBoriginate virtually entirely from the MR and that themajority of MR fibers projecting to the MSDB are seroton-ergic. This is further evidenced by the present resultsshowing that the synaptic profiles for the two sets ofimmunolabeled terminals (5-HT- and PHA-L-labeled) werevery similar; that is, with large boutons, asymmetricsynaptic contacts, and short membrane specializations(about 0.2–0.3 µm in diameter). The discrepancy betweenthe density of MSDB PA-containing neurons that areinnervated by 5-HT-immunoreactive axons and those thatare targeted by anterogradely labeled axons can be ex-plained methodologically. Thus, as mentioned above (seeResults), although all of the PA- and 5-HT-containingelements can be visualized on the surface of the Vibratomesections, not all of the MR neurons can be labeled by asingle PHA-L microinjection.

The large PHA-L- and 5-HT-immunoreactive boutonsdescribed here correspond closely to large varicositiespreviously shown to characterize 5-HT MR fibers in theforebrain, including the septum, and they are distinctlydifferent from those of the dorsal raphe nucleus (DR). Forinstance, Molliver and associates (Kosofsky and Molliver,1987) and, subsequently, several others (Mulligan andTork, 1988; Hornung et al., 1990; Voigt and de Lima, 1991)demonstrated that the forebrain contains two major typesof 5-HT fibers: 1) thick fibers with large, spherical varicosi-ties, shown to originate from the MR; and 2) thin fiberswith small, oval or fusiform varicosities, shown to origi-nate from the DR. Accordingly, the terminals shown hereare of the MR type and not the DR type. Consistent withthis, the DR projects to the lateral septum and, at best,only sparsely to the MSDB (see Fig. 7; Vertes, 1991).

In some respects, the present observations contradict anearlier report (Milner and Veznedaroglu, 1993) on the5-HT innervation of the septohippocampal neurons. Inthat work, the authors reported that the average diameterof 5-HT-immunoreactive boutons is smaller (0.3–1.2 µm)than that (1.0–1.4 µm) observed in this study. Further-more, and more importantly, we observed exclusivelyasymmetric synaptic contacts between both the PHA-L-and 5-HT-immunoreactive boutons and the PA cells. Incontrast, they reported that 5-HT-containing axon termi-nals formed both symmetric and asymmetric membranespecializations with septohippocampal neurons. However,they found that the majority (70%) of those contacts wereasymmetric. In our study, the type of synaptic contactsalways was determined by use of a goniometer, which mayhelp to explain this discrepancy. The common observationof the two studies is the frequent appositions of serotoner-gic terminals to cell bodies without synaptic membranespecializations. These may represent a nonsynaptic, para-crine release of 5-HT.

Physiological relevance

The present demonstration that serotonergic MR fibersform asymmetric synapses with PA (GABAergic) cells ofthe MSDB suggests that they have an excitatory influenceon these GABAergic cells (Eccles, 1964). Consistent withthis possibility, Alreja and colleagues, in correspondingreports in septal slices (Alreja, 1996; Liu and Alreja, 1997),showed that 5-HT selectively excited putative septohippo-

Fig. 7. Schematic representation of the proposed direct and indi-rect route of action of 5-HT on the hippocampus. MR 5-HT neuronsthat send direct fibers to the hippocampus innervate a subpopulationof hippocampal GABAergic interneurons, which terminate exclusivelyon the apical dendrites of principal cells (Freund et al., 1990).Stimulation of these GABAergic neurons will effectively block theinput sector of principal cells. On the other hand, MR 5-HT cellsterminating in the MSDB (or in both the MSDB and the hippocampus)activate septohippocampal GABAergic neurons, which innervate pre-dominantly hippocampal basket and chandelier cells (Freund andAntal, 1988) and also a very limited number of calbindin-containingGABA neurons that terminate on the apical dendrites of principal cells(Miettenen and Freund, 1992). Thus, stimulation of septohippocampalGABAergic neurons results mainly in a disinhibition of the outputsector of hippocampal principal neurons and, through their local axoncollaterals (Leranth and Frotscher, 1989), depresses the activity ofseptohippocampal cholinergic cells and unidentified MSDB GABAcells (not shown). ACH, acetylcholine.


campal GABAergic neurons of the MSDB. In addition,they showed that the excitatory actions of 5-HT on GABAer-gic MSDB cells primarily involved 5-HT2A receptors. PAcells that receive serotonergic input from the MR, asshown here, express the 5-HT2A receptor, indicating that5-HT MR neurons excite PA-containing GABAergic cells ofthe MSDB. One question remains, however: Do all of theMSDB PA-containing GABAergic cells project to the hippo-campus? Although there is no direct evidence available,our unpublished observations indicate that they all do.Three weeks after fimbria fornix transection, similar to theMSDB cholinergic neurons (Neumann et al., 1992; Peter-son, et al., 1992, 1993), which survive but do not expressChAT, no PA-immunoreactive neurons can be detected inthe MSDB. Thus, axotomy seems to target all of the MSDBPA cells.

This system of connections would appear to underlie thewell-documented desynchronizing effects of the MR on thehippocampal EEG (Vertes and Kocsis, 1997). MR stimula-tion desynchronizes the hippocampal EEG (Macadar et al.,1974; Assaf and Miller, 1978; Vertes, 1981), and MR lesionsor pharmacological suppression of 5-HT MR cells gener-ates theta (i.e., blocks desynchronization; Maru et al.,1979; Kinney et al., 1994, 1995; Vertes et al., 1994).Furthermore, MR cells discharge in a manner consistentwith a direct role in the desynchronization of the hippocam-pal EEG; that is, they fire at the highest rates duringstates of hippocampal desynchronization (quiet wakingand slow-wave sleep) and are virtually silent during statesin which theta is present in the hippocampus [explorationof waking and rapid-eye-movement (REM) sleep]. Regard-ing the latter, Marrosu et al. (1996) recently reported that5-HT MR neurons virtually ceased firing in the presence oftheta during waking and REM sleep in behaving rats.They stated: ‘‘In addition to displaying the characteristicsuppression of activity during REM sleep, these neuronsalso exhibited a marked decrease or suppression in activ-ity, during a variety of waking behaviors, which have beenassociated with increased hippocampal theta activity. Thesebehaviors include exploratory movements about the record-ing chamber, orientation to strong or novel stimuli (e.g.,door opening or presentation of an inaccessible rat), pos-tural shifts, and walking/running on a treadmill.’’

Evidence suggests that the desynchronizing actions ofthe MR on the hippocampal EEG involve primarily adisruption of the rhythmic activity of MSDB cells that pacetheta (pacemaker cells). For instance, it has been shownthat MR stimulation disrupts the rhythmic discharge ofthe pacemaker cells (Assaf and Miller, 1978), and pharma-cological suppression of serotonergic MR cells with 5-HTagonists rhythmically activates these cells (Kinney et al.,1996). In the study discussed previously by Alreja (1996),which showed that 5-HT activates GABAergic MSDB cells,it was reported further that these GABAergic cells, inturn, inhibit subsets of cholinergic and unidentified GABA-ergic neurons of the MSDB. The observation that all of theanterogradely labeled and 5-HT-immunopositive basketswere associated with PA neurons indicates that the afore-mentioned GABA cells are septohippocampal PA-contain-ing neurons and that their axon collaterals terminate oncholinergic cells. Taken together, these findings suggestthat serotonergic MR cells activate PA-containing GABA-ergic cells of the MSDB, which, in turn, inhibit the activity(or disrupt the rhythmic firing) of cholinergic pacemakercells to desynchronize the hippocampal EEG. This view is

supported by the observation that MSDB cholinergic cellsare innervated massively by GABAergic terminals (Ler-anth and Frotscher, 1989) that originate in the MSDB(lateral septal area GABA neurons do not project to theMSDB; Leranth et al., 1992).

In addition to the circuitry described above, other sys-tems, in parallel with the MR-septohippocampal pathway,may be involved in the control of the hippocampal EEG. Ithas been demonstrated that 5-HT MR neurons projectdirectly to the hippocampus and terminate predominantlyon hippocampal interneurons (Freund et al., 1990; Halasyet al., 1992). Also, as discussed, MSDB PA (GABAergic)neurons that can be activated by 5-HT send fibers to thehippocampus (Liu and Alreja, 1997) and terminate primar-ily on hippocampal interneurons (Freund and Antal, 1988;Gulyas et al., 1990). These observations, as illustrated inFigure 7, suggest that the 5-HT-containing MR neuronsmay have a dual influence on the activity of the principalcells of the hippocampus, possibly in the control of thehippocampal EEG, in other words, an inhibition of theinput sector through a direct MR activation of a specificpopulation of hippocampal interneurons that terminateexclusively on apical dendrites of pyramidal cells and adisinhibition of the output sector through an MR activa-tion of MSDB PA-containing GABAergic neurons thatspecifically innervate hippocampal basket and chandeliercells.

To fully understand the mechanism(s) responsible forthe desynchronization of hippocampal EEG after MRstimulation, several other factors should be elucidated.Most importantly, it is not known whether the same MRcells activate both MSDB GABAergic septohippocampalcells and hippocampal interneurons. Thus, if differentpopulations of MR neurons terminate in the MSDB andhippocampus, then the two suggested effects, inhibition ofthe input sector and disinhibition of the output sector ofthe principal cells, may not be synchronized. This seems bethe case, because, in a preliminary study, we examined thepossible collateral projections from the MR to the MSDBand the hippocampus. In ten rats to date, we found thatonly 7–8% of MR cells project to both the MSDB and thehippocampus.

5-HT and memory

A growing body of evidence indicates that the thetarhythm serves an important role in mnemonic functions ofthe hippocampus (for review, Vertes and Kocsis, 1997). Forinstance, it has been shown that 1) long term potentiation(LTP) in the hippocampus is elicited optimally with pat-terns of stimulation mimicking the theta rhythm (Larsonand Lynch, 1986; Larson et al., 1986; Rose and Dunwiddie,1986; Staubli and Lynch, 1987; Diamond et al., 1988;Greenstein et al., 1988; Leung et al., 1992); 2) stimulationdelivered in the presence, but not in the absence, of thetapotentiates population responses in the hippocampus (Pav-lides et al., 1988; Bramham and Srebro, 1989; Huerta andLisman, 1993); and 3) discrete medial septal lesions thatabolish theta in the hippocampus produce severe learning/memory deficits in rats (Winson, 1978; Mizumori et al.,1990; M’Harzi and Jarrard, 1992).

In line with these data, a 5-HT MR-mediated disruptionof theta (hippocampal desynchronization) may disruptLTP and/or memory or, alternatively, a suppression of theMR and, hence, the elicitation of theta, may enhancememory. Consistent with this, it has been shown that


serotonergic agents block LTP (Corradetti et al. 1992;Staubli and Otaky, 1994), and 5-HT antagonists (mainly5-HT3 antagonists) enhance LTP and/or memory (Staubliand Xu, 1995; Buhot, 1997; Vertes and Kocsis, 1997).

ACKNOWLEDGMENTS

The authors thank M. Shanabrough and N. Todorova fortheir excellent technical help and Dr. J.H. Rogers for thecalretinin antiserum. This work was supported by grantsNS26068 (to C.L.) and NS35883 (to R.P.V.) from theNational Institutes of Health.

LITERATURE CITED

Alreja M. 1996. Excitatory actions of serotonin on GABAergic neurons ofthe medial septum and diagonal band of Broca. Synapse 22:15–27.

Assaf SY, Miller JJ. 1978. The role of the raphe serotonin system in thecontrol of septal unit activity and hippocampal desynchronization.Neuroscience 3:539–550.

Azmitia EC, Segal M. 1978. An autoradiographic analysis of the differentialascending projections of the dorsal and median raphe nuclei in the rat. JComp Neurol 179:641–668.

Bramham CR, Srebro B. 1989. Synaptic plasticity in the hippocampus ismodulated by behavioral state. Brain Res 493:74–86.

Buhot M-C. 1997. Serotonin receptors in cognitive behaviors Curr OpinNeurobiol 7:243–254.

Corradetti R, Ballerini L, Pugliese AM, Pepeu G. 1992. Serotonin blocks thelong-term potentiation induced by primed burst stimulation in the CA1region of rat hippocampal slices. Neuroscience 46:511–518.

Diamond DM, Dunwiddie TV, Rose GM. 1988. Characteristics of hippocam-pal primed burst potentiation in vitro and in the awake rat. J Neurosci8:4079–4088.

Eccles JC. 1964. The physiology of synapses. Heidelberg: Springer.Freund TF, Antal M. 1988. GABA-containing neurons in the septum control

inhibitory interneurons in the hippocampus. Nature 336:170–173.Freund TF, Gulyas AI, Acsadi L, Gorcs T, Toth K. 1990. Serotonergic control

of the hippocampus via local inhibitory interneurons. Proc Natl AcadSci USA 87:8501–8505.

Greenstein YJ, Pavlides C, Winson J. 1988. Long-term potentiation ispreferentially induced at theta periodicity. Brain Res 438:331–334.

Gulyas AI, Gorcs TJ, Freund TF. 1990. Innervation of different peptide-containing neurons in the hippocampus by GABAergic septal afferents.Neuroscience 37:31–44.

Halasy K, Miettinen R, Szabat E, Freund TF. 1992. GABAergic interneu-rons are the major postsynaptic targets of median raphe afferents in therat dentate gyrus. Eur J Neurosci 4:144–153.

Honda T, Semba K. 1993. Serotonergic synaptic input to GABAergicseptohippocampal neurons in the rat: a pre-embedding triple-labelelectron microscopic study. Soc Neurosci Abstr 19:1442.

Hornung J-P, Fritschy J-M, Tork I. 1990. Distribution of two morphologi-cally distinct subsets of serotonergic axons in the cerebral cortex of themarmoset. J Comp Neurol 297:165–181.

Huerta PT, Lisman JE. 1993. Heightened synaptic plasticity of hippocam-pal CA1 neurons during a cholinergically induced rhythmic state.Nature 364:723–725.

Jakab RL, Goldman-Rakic PS. 1998. 5-Hydroxytryptamine2A serotoninreceptors in primate cortex: possible site action of hallucinogens inpyramidal cell apical dendrites. Proc Natl Acad Sci USA 95:735–740.

Kinney GG, Kocsis B, Vertes RP. 1994. Injections of excitatory amino acidantagonists into the median raphe nucleus produce hippocampal thetarhythm in the urethane anesthetized rat. Brain Res 654:96–104.

Kinney GG, Kocsis B, Vertes RP. 1995. Injections of muscimol into themedian raphe nucleus produce hippocampal theta rhythm in the rat.Psychopharmacology 120:244–248.

Kinney GG, Kocsis B, Vertes RP. 1996. Medial septal unit firing characteris-tics following injections of 8-OH-DPAT into the median raphe nucleus.Brain Res 708:116–122.

Kiss J, Magloczky, Z, Somogyi J, Freund TF. 1997. Distribution of calretinin-containing neurons relative to other neurochemically identified celltypes in the medial septum of the rat. Neuroscience 78:399–410.

Kosaka T, Kosaka K, Tateishi K, Hamaoka Y, Yanaihara N, Wu J-Y, Hama

K. 1985. GABAergic neurons containing CCK-8-like and/or VIP-likeimmunoreactivities in the rat hippocampus and dentate gyrus. J CompNeurol 239:420–430.

Kosaka T, Nagatsu H, Wu JY, Hama K. 1986. Use of high concentrations ofglutaraldehyde for immunocytochemistry of transmitter-synthesizingenzymes in the central nervous system. Neuroscience 18:975–990.

Kosofsy BE, Molliver ME. 1987. The serotonergic innervation of cerebralcortex: different classes of axon terminals arise from the dorsal andmedian raphe nuclei. Synapse 1:153–168.

Larson J, Lynch G. 1986. Induction of synaptic potentiation in hippocam-pus by patterned stimulation involves two events. Science 232:985–988.

Larson J, Wong D, Lynch G. 1986. Patterned stimulation at theta frequencyis optimal for the induction of long-term potentiation. Brain Res368:347–350.

Leranth C, Frotscher M. 1989. The organization of the septal region in therat brain: cholinergic-GABAergic interconnections and the terminationof hippocampo-septal fibers. J Comp Neurol 289:304–314.

Leranth C, Deller T, Buzsaki G. 1992. Intraseptal connections redefined:lack of lateral septum to medial septum path. Brain Res 583:1–11.

Leung LS, Shen B, Kaibara T. 1992. Long-term potentiation induced bypatterned stimulation of the commissural pathway to hippocampal CA1region in freely moving rats. Neuroscience 48:63–74.

Liu W, Alreja M. 1997. Atypical antipsychotics block the excitatory effects ofserotonin in septohippocampal neurons in the rat. Neuroscience 79:369–382.

Macadar AW, Chalupa LM, Lindsley DB. 1974. Differentiation of brain stemloci which affect hippocampal and neocortical electrical activity. ExpNeurol 43:499–514.

Marrosu F, Fornal CA, Metzler CW, Jacobs BL. 1996. 5-HT1A agonistsinduce hippocampal theta activity in freely moving rats: role of presyn-aptic 5-HT1A receptors. Brain Res 739:192–200.

Maru E, Takahashi LK, Iwahara S. 1979. Effects of median raphe nucleuslesions on hippocampal EEG in the freely moving rat. Brain Res163:223–234.

M’Harzi M, Jarrard LE. 1992. Effects of medial and lateral septal lesions onacquisition of a place and cue radial maze task. Behav Brain Res49:159–165.

Miettinen R, Freund TF. 1992. Convergence and segregation of septal andmedian raphe inputs onto different subsets of hippocampal inhibitoryinterneurons. Brain Res 594:263–272.

Milner TA, Veznedaroglu E. 1993. Serotonin-containing terminals synapseon septohippocampal neurons in the rat. J Neurosci Res 36:260–271.

Mizumori SJY, Perez GM, Alvarado MC, Barnes CA, McNaughton BL.1990. Reversible inactivation of the medial septum differentially affectstwo forms of learning in rats. Brain Res 528:12–20.

Moore RY, Halaris AE. 1975. Hippocampal innervation by serotoninneurons of the midbrain raphe in the rat. J Comp Neurol 164:171–184.

Mulligan KA, Tork I. 1988. Serotonergic innervation of the cat cerebralcortex. J Comp Neurol 270:86–110.

Neumann T, Peterson GM, Frotscher M. 1992. Fine structure of ratseptohippocampal neurons: II. A time course analysis following axotomy.J Comp Neurol 325:219–242.

Pavlides C, Greenstein YJ, Grudman M, Winson J. 1988. Long-termpotentiation in the dentate gyrus is induced preferentially on thepositive phase of theta-rhythm. Brain Res 439:383–387.

Peterson GM, Neumann T, Frotscher M. 1992. Identified septohippocampalneurons survive axotomy: a fine-structural analysis in the rat. NeurosciLett 138:81–85.

Peterson GM, Neumann T, Frotscher M. 1993. Light- and electro-microscopic studies of identified septohippocampal neurons survivingaxotomy. Ann NY Acad Sci 678:291–298.

Rogers, JH. 1987. Calretinin: a gene for novel calcium-binding proteinexpressed principally in neurons. J Cell Biol 105:1343–1353.

Rose GM, Dunwiddie TV. 1986. Induction of hippocampal long-termpotentiation using physiologically patterned stimulation. Neurosci Lett69:244–248.

Semba K, Reiner PB, McGeer, Fibiger HC. 1988. Brainstem afferents to themagnocellular basal forebrain studied by axonal transport, immunohis-tochemistry, and electrophysiology in the rat. J Comp Neurol 267:433–453.

Staubli U, Lynch G. 1987. Stable hippocampal long-term potentiationelicited by ‘theta’ pattern stimulation. Brain Res 435:227–234.

Staubli U, Otaky N. 1994. Serotonin controls the magnitude of LTP inducedby theta bursts via an action on NMDA-receptor-mediated responses.Brain Res 643:10–16.


Staubli U, Xu F. 1995. Effects of 5-HT3 receptor antagonism on hippocampaltheta rhythm, memory, and LTP induction in the freely moving rat. JNeurosci 15:2445–2452.

Vertes RP. 1981. An analysis of ascending brain stem systems involved inhippocampal synchronization and desynchronization. J Neurophysiol46:1140–1159.

Vertes RP. 1991. A PHA-L analysis of ascending projections of the dorsalraphe nucleus in the rat. J Comp Neurol 313:643–668.

Vertes RP, Kocsis B. 1997. Brainstem-diencephalo-septohippocampal sys-tems controlling the theta rhythm of the hippocampus. Neuroscience81:893–926.

Vertes RP, Martin GF. 1988. An autoradiographic analysis of ascendingprojections from the pontine and mesencephalic reticular formation andthe median raphe nucleus in the rat. J Comp Neurol 275:511–541.

Vertes RP, Kinney GG, Kocsis B, Fortin WJ. 1994. Pharmacologicalsuppression of the median raphe nucleus with serotonin1A agonists,8-OH-DPAT and buspirone, produces hippocampal theta rhythm in therat. Neuroscience 60:441–451.

Voigt T, de Lima AD. 1991. Serotonergic innervation of the ferret cerebralcortex. I. Adult pattern. J Comp Neurol 314:403–414.

Winson J. 1978. Loss of hippocampal theta rhythm results in spatialmemory deficit in the rat. Science 201:160–163.


Documents

Median raphe serotonergic innervation of medial septum/diagonal band of Broca (MSDB) parvalbumin-containing neurons: Possible involvement of the MSDB in the desynchronization of the