~ Pergamon

PII: S0306-4522(00)00496-6
Neuroscience Vol. 102, No. 3, pp. 527-540, 2001
© 2001 IBRO. Published by Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0306-4522/01 $20.00+0.00
www.elsevier.com/locate/neuroscience

TOTAL NUMBER AND DISTRIBUTION OF INHIBITORY AND EXCITATORY

SYNAPSES ON HIPPOCAMPAL CA1 PYRAMIDAL CELLS

M. MEG[AS, ZS. EMRI, T. F. FREUND and A. i. GULY,~S*

Institute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, P.O. Box, H-1450, Hungary

Abstract--The integrative properties of neurons depend strongly on the number, proportions and distribution of excitatory
and inhibitory synaptic inputs they receive. In this study the three-dimensional geometry of dendritic trees and the density of
symmetrical and asymmetrical synapses on different cellular compartments of rat hippocampal CA I area pyramidal cells
was measured to calculate the total number and distribution of excitatory and inhibitory inputs on a single cell.
A single pyramidal cell has -12,000 txm dendrites and receives around 30,000 excitatory and 1700 inhibitory inputs, of
which 40% are concentrated in the perisomatic region and 20% on dendrites in the stratum lacunosum-moleculare. The pre-
and post-synaptic features suggest that CA1 pyramidal cell dendrites are heterogeneous. Strata radiatum and oriens dendrites
are similar and differ from stratum lacunosum-moleculare dendrites. Proximal apical and basal strata radiatum and oriens
dendrites are spine-free or sparsely spiny. Distal strata radiatum and oriens dendrites (forming 68.5% of the pyramidal cells'
dendritic tree) are densely spiny; their excitatory inputs terminate exclusively on dendritic spines, while inhibitory inputs
target only dendritic shafts. The proportion of inhibitory inputs on distal spiny strata radiatum and oriens dendrites is low
(-3%). In contrast, proximal dendritic segments receive mostly (70 100%) inhibitory inputs. Only inhibitory inputs
innervate the somata (77 103 per cell) and axon initial segments. Dendrites in the stratum lacunosum-moleculare possess
moderate to small amounts of spines. Excitatory synapses on stratum lacunosum-moleculare dendrites are larger than the
synapses in other layers, are frequently perforated (-40%) and can be located on dendritic shafts. Inhibitory inputs, whose
percentage is relatively high (-14-17%), also terminate on dendritic spines.
Our results indicate that: (i) the highly convergent excitation arriving onto the distal dendrites of pyramidal cells is
primarily controlled by proximally located inhibition; (ii) the organization of excitatory and inhibitory inputs in layers
receiving Schaffer collateral input (radiatulrdoriens) versus perforant path input (lacunosum-moleculare) is significantly
different. © 2001 IBRO. Published by Elsevier Science Ltd. All rights reserved.

Key words: synaptic convergence, dendrite geometry, serial reconstruction, 3D, electron microscopy, database for modeling.

The principal cells of the hippocampal CA1 area, the
pyramidal cells, integrate information arriving from
several sources. Schaffer collaterals from the CA3 area
form synapses on dendrites located in strata radiatum and
oriens; input from the entorhinal cortex and different
subcortical structures (e.g. nucleus reuniens thalami,
amygdala) innervate the distal apical dendrites in stratum
lacunosum-moleculare, whereas recurrent collaterals
from CA1 pyramidal cells innervate basal dendrites. 3'36
Several functionally different interneuron populations
control pyramidal cell activity via axon terminals that
target precise domains of the postsynaptic cells. 7'14'17
Inhibitory cells terminating in the perisomatic region
control the generation of Na + spikes and thus the output
of the neurons. Inhibition in the dendritic region, in
*Corresponding author. Tel: +36-1-2100-819/246; fax: +36-1-313-
9498.
E-mail address: gulyas@koki.hu (A. I. Guly~is).
Abbreviations: AMPA, c~-amino-3-hydroxy-5-methyl-4-isoxazole-
propionate; BDA, biotinylated dextran amine; CA, comu ammonis;
CB, calbindin D28k; CCK, cholecystokinin; CR, calretinin; DAB,
3Y-diaminobenzidine; PB, phosphate buffer; PV, parvalbumin;
SLM, stratum lacunosum-moleculare dendrites; SRO, strata radia-
turn and oriens dendrites; TBS, Tris-buffered saline; VIP, vaso-
active intestinal polypeptide.
527
contrast, controls the generation of dendritic Ca 2+ spikes
and is associated with synaptic plasticity in dendrites. 2v
Although several studies have estimated the total
n u m b e r of inputs converging onto a pyramidal cell by
counting spines.,5'23'3°'33 important aspects of pyramidal
cell innervation may have gone unnoticed at the light
microscopic level. Measurements of spine density did
not reveal the detailed organization of excitatory inputs
on different cell domains. The number, distribution and
ratio of inhibitory synapses converging onto pyramidal
cells has not been studied either. The arrangement of
excitatory and inhibitory inputs relative to each other,
their ratio and the distribution of the inputs on the
cells' surface are important parameters influencing inte-
grative properties and cellular activity patterns. This
information may shed light on how pyramidal cells trans-
form signals arriving from different sources.
We recently described the number, ratio and distribu-
tion of excitatory and inhibitory inputs on three subsets
of hippocampal interneurons in the CA1 area, j5 and
found significant differences as well as c o m m o n prin-
ciples in the organization of synaptic inputs among
these cell populations. Interneurons and principal cells
show rather different average activity levels ~° and
firing patterns. ~ Mapping the distribution of terminals
528 M. Megfas et al.

on pyramidal cells provides structural parameters that are
n e c e s s a r y to understand h o w interactions b e t w e e n exci-
tatory and inhibitory inputs c o m b i n e to shape cellular
activity.
To a n s w e r these questions, the c o m p l e t e dendritic
trees o f C A 1 p y r a m i d a l cells in the adult rat h i p p o c a m p u s
w e r e r e c o n s t r u c t e d in 3D at the light m i c r o s c o p i c level,
f o l l o w e d by serial electron m i c r o s c o p i c reconstruction o f
excitatory and inhibitory synaptic inputs c o n v e r g i n g onto
different dendrites, somata and axon initial segments.
The density, distribution and ratio o f inhibitory and exci-
tatory inputs were measured, and the total n u m b e r o f
c o n v e r g i n g inputs was calculated.
Light microscopic sampling
The BDA injection sites contained large amounts of labeled
neurons hindering the identification of individual dendrites (Fig.
1B). Therefore we selected neurons for reconstruction at a
distance from the injection sites, where individually visualized
cells could also be found. We selected only those cells where all
the dendrites could be clearly followed until their natural end.
The dendritic segments were classified into different subclasses in
each layer, based on their order, spine density and distance from
the soma. The division between subclasses was evident due to
the organization principles of pyramidal cell dendrites. Cells
(n = 20) were reconstructed from six to 13 sections using camera
lucida. The subclasses of dendrites and their location were
recorded. The 3D structure of the neurons and data on the length
of dendritic subclasses were obtained after the digitalization of
the reconstructed cells using the program ARBOR developed by
S. Pomah~izy and modified by A. I. Guly~s. t5'35

EXPERIMENTAL PROCEDURES Sampling for electron microscopy. Four to six segments of

each dendritic subclass were selected, re-embedded and serially
Seven adult (300 g) male Wistar rats (Charles River, Budapest)
sectioned for electron microscopy (samples derive from five
were used in these experiments. The animals were anesthetized
animals, each dendritic subclass was sampled in at least two
with Equithesin (chlornembutal, 0.3 ml per 100g of body
animals, at least two samples per animal). Before and after the
weight). Biotinylated dextran amine (BDA, Molecular Probes;
serial ultrathin sectioning, dendrites were drawn using a camera
10% in phosphate buffer, PB) was injected into the CAI region
lucida to determine the length of the sectioned segment. To mini-
at the following coordinates: Bregma: - 3 . 2 mm; lateral: 2 mm
mize the error in the measurements we cut long series of ultrathin
(left and right); - 2 . 4 m m from pia mater; and Bregma:
sections containing on average 25-/xm-long segments of den-
- 4 . 3 mm; lateral: 3 mm (left and right); - 2 . 4 from pia mater.
drites. The synapses were classified as symmetrical or asym-
The BDA was iontophoretically applied by pulsed positive
metrical from the postsynaptic density and the type of vesicles. L3
current for 5 min, 5 IxA, 7 s on/7 s off via a glass capillary.
Spine densities shown in Table 2 derive from the electron micro-
After two days survival, the animals were anesthetized with
scopic measurements. We noticed differences in the size of
Equithesin (chlornembutal, 0.3 ml per 100 g of body weight) and
synapses in different layers. We made a semi-quantitative
perfused transcardially with 50ml of physiological saline
measurement: in the case of 30 synapses from each dendritic
followed by 300 ml of fixative containing 3% paraformaldehyde,
subclass we counted the number of sections the synapses occu-
1% glutaraldehyde and 15% saturated picric acid in 0.l M PB
pied to gain a distribution of synapse size. To count the number
(pH 7.4). All experiments were approved by the animal care
and types of synapses on somata and axon initial segments, long
committee of the Institute. Every effort was made to minimize
serial sections (300-400 sections, 55 nm thickness) were cut
animal numbers and suffering during the experiments.
from the CA1 pyramidal cell layer. Unlabeled pyramidal cell
The brains were removed from the skull and postfixed (1 h).
bodies (n = 5, from two animals) were completely reconstructed.
Then 60-/xm-thick sections were cut on a Vibratome. The
Axon initial segments of labeled cells with a long straight passage
sections were kept in sequential order to allow serial reconstruc-
were also reconstructed (n = 4, from three animals).
tion of the dendritic arbors. To calculate possible shrinkage
In case of the input density measurements of this paper, stan-
during the experimental procedure, several sections were placed
dard errors are not given, since, due to the rather laborious nature
on a slide and their profiles and cut capillaries were drawn using a
of the sampling, the sample size for individual dendrite
camera lucida. To increase penetration of reagents, sections were
subclasses from each layer was small (three to five segments).
cryoprotected by immersion in 25% sucrose and 10% glycerol in
0.1 M PB for 30 rain, and freeze-thawed three times over liquid
nitrogen. Sections were treated with 1% NaBH4 for 30 min. BDA lmmunogold detection of GABA. The immunostaining proce-
was then visualized by incubation in Elite ABC (1:200, Vector, dure followed those described by Somogyi and Hodgson, 32 with
2 h), and developed using 0.03% DAB (3,3r-diaminobenzidine- small modifications, ~5 using a well-characterized antiserum
4HCI, Sigma) and hydrogen peroxide (0.01%, 5 min). Sections against GABA. 21 The steps were carried out on droplets of
were treated with 1% OsO4 in 0.1 M PB for 40 rain, dehydrated Millipore-filtered solutions in humid Petri dishes, as follows:
and embedded in Durcupan (ACM Fluka). The shrinkage was 2% periodic acid (H5IO6) for l0 min; wash by dipping in several
estimated by checking the change in the size of the sections changes of double-distilled water; 2% sodium metaperiodate
drawn before the immunostaining. (NaIO4) for 10 min; wash as before; three times 2 min in TBS

Fig. 1. CAI pyramidal cells visualized by extracellular BDA injection. After BDA injection in CA1, a tightly packed group of
stained pyramidal ceils was observed at the injection site (long arrow in B). At distances of 1 or 2 ram, some ceils were labeled
retrogradely (short arrow in B) and individual cells could be distinguished and reconstructed completely. (A) Apical dendrite. The
spine density increases from the proximal to the distal part of the dendrite. In the proximal part (P: radiatum/Thick/proximal) no
spines are visible; they start to appear at around 100 ~m from the soma, and increase progressively in number through the middle
segment (m, radiatum/Thick/medial) until they reach the distal third (d, radiatum/Thick/distal) of stratum radiatum where their
density is maximal. Open arrows label subclass boundaries. (C) Thin spiny dendrites (short arrows, radiatum/thin) branch from the
apical dendrite in stratum radiatum. These dendrites are homogeneous in appearance and could be clearly distinguished from labeled
intemeuron dendrites (long arrow). (E) In the stratum oriens, proximal dendrites bore no spines (curved arrow, oriens/proximal).
Spine number increased gradually with the distance from the soma, reaching and maintaining a maximal density (arrows) on the
most distal portions (oriens/distal). Axon initial segments were also labeled (arrowheads). (F) The density of spines decreased
considerably at the stratum radiatum (r, long arrow)/stratum lacunosum-moleculare (lm, short arrow) border. (G) Spines in the
stratum lacunosum-moleculare are larger and less frequent, their number decreases from thick (large arrow, l-m/Thick) to medium
dendrites (short arrow, l-m/medium). (H) The most distal dendrites in the stratum lacunosum-moleculare are very thin and long (l-m/
thin). They bear very few spines (large arrows) and show varicosities (short arrows). Scale bars = 4 ixm (A), 250 ixm (B), 20/xm (D),
10 ~m (C, E-H).
 Synaptic convergence onto CA I pyramidal cells
(pH 7.4); 30 rain in 1% ovalbumin dissolved in TBS; three times
10 min in TBS containing 1% normal goat serum; l - 2 h in a
rabbit anti-GABA antiserum (Code No. 9, diluted 1:1000 in
normal goat serum/TBS); two times 10rain TBS; 10rain in
0.05 M Tris buffer (pH 7.4) containing 1% bovine serum albumin
and 0.5% Tween 20; goat anti-rabbit IgG-coated colloidal gold
(12 nm, Jackson) for 2 h (diluted 1:20 in the same buffer); two
times 5 min wash in double-distilled water; saturated uranyl
Fig. 1.
529
acetate for 30 min; wash in four changes of double-distilled
water; staining with lead citrate; wash in distilled water. Pro-
files showing a density of colloidal gold particles at least five
times above background level, in two to three adjacent sec-
tions, were considered GABA-immunoreactive. Axon terminals
forming asymmetrical synapses (presumed glutamatergic) were
used to establish background density. Replacement of the GABA
antiserum with normal rabbit serum in the postembedding
530 M. Megfas et al.

immunogold reaction resulted in a loss of specific staining; i.e. no
signs of colloidal gold accumulation could be detected over any
profiles.
RESULTS
Light microscopy
Biotinylated dextran amine (BDA) injected into the
CA1 area labeled a dense group of cells in the center
of the injection site, as well as cells several hundred
micrometers away from the injection site, probably by
retrograde transport. These solitary cells (Fig. 1D)
showed very dense precipitation of DAB so that den-
drites could be followed and morphological features
described in detail (Fig. I C). The labeling visualized
the fine dendritic spines and the dendrites could be
followed until their natural ends terminated abruptly,
without fading within the sections. We used these soli-
tary cells for the reconstructions.
During the experimental procedures the shrinkage of
the slices was checked. The dimensions of the sections
before and after the BDA visualization and embedding
remained the same, so no corrections were applied for
shrinkage. Note also that for the calculation of the total
number of synapses, no correction was necessary, since
light and electron microscopic data were obtained from
the same material.
Dendritic features and types. Although several studies
and databases have quantified the dendritic length of
CA1 pyramidal cells, 5,23,3°,33 we needed to classify the
dendrites and to measure the length of each subclass
separately in order to calculate the total synaptic conver-
gence. We reconstructed 20 pyramidal cells from the
CAI region of the dorsal hippocampus (the detailed
geometry of the cells together with additional data, not
presented in the paper, can be downloaded from: http://
www.koki.hu/-gulyas/calcells). As shown in Figs 1 and
2, pyramidal cell dendritic trees are rather heterogeneous.
The dendrites were classified into nine subclasses
according to their location, diameter, and the density of
spines on them. Dendritic subclasses and their character-
istic diameters are listed in Fig. 2 and Table 1.
Pyramidal cells have a basal dendritic tree consisting
of three to five (3.55 + 0.82) primary dendrites and an
apical dendritic tree which arises in the form of a thick
apical dendrite ascending until stratum lacunosum-
moleculare and emitting secondary branches in stratum
radiatum. Two dendritic subclasses were distinguished in
the basilar dendritic tree: (i) thick dendrites, close to the
soma with few or no spines in upper stratum oriens (oriens/
proximal); and beyond that (ii) thinner distal dendrites
densely covered with spines (oriens/distal). The spine
density progressively increased from spine free segments
(up to 3 0 - 5 0 ~m from the soma) until the first branch-
point, where they reached their maximum density (Figs
1E, 2).
The thick apical trunk of the pyramidal cells emits
many secondary thinner dendrites in stratum radiatum,
and upon reaching stratum lacunosum-moleculare bifur-
cates several times to form thinner branches of different
appearance. The bifurcating apical dendrites turn, and
run parallel with the hippocampal fissure spanning a
considerable distance. The diameter (lateral spread) of
the apical dendritic tree in stratum lacunosum-moleculare
is thus often larger (350-750 I~m, 462.4 + 132.5 p~m)
than the diameter of the dendritic arbor in stratum oriens
(170-320 txm, 250.5 +-54.5 txm) and stratum radiatum
(200-350 p~m, 255.6 _+ 54.2 Ixm).
Thick apical dendrites were divided into three
subclasses, based on spine density: (i) a proximal seg-
ment of approximately 100 I~m with no spines (radiatum/
Thick/proximal); (ii) a medial segment ( - 1 5 0 txm) with
a low spine density (radiatum/Thick/medial); and (iii) a
thick distal segment ( - 2 0 0 Ixm) with very high density
of spines (radiatum/Thick/distal, Figs 1A and 2). Thinner
branches, originating from the apical dendrites in stratum
radiatum formed the fourth group (radiatum/thin). These
dendrites represented the majority of dendrites in this
layer and were similar in appearance to oriens/distal
dendrites.
The spine density decreased on thick (bifurcated)
apical dendrites entering stratum lacunosum-moleculare
(Fig. IF). Here we distinguished three dendrite sub-
classes: (i) the prolongation of the thick apical dendrite
(l-m/Thick, - 1 . 1 i~m) that were covered by fewer but
larger spines than in stratum radiatum (Fig. 1G, long
arrow); (ii) after further bifurcations the segments carried
fewer spines of smaller size (1-m/medium, Fig. 1G, short
arrow); and finally (iii) thin distal dendrites (Fig. I H)
representing the majority of the dendrites in stratum
lacunosum-moleculare (1-m/thin). These dendrites pos-
sessed very few spines and also showed varicosities
and segments without spines.
Dendritic lengths. Following the light microscopic
reconstruction, the total dendritic length as well as the

Fig. 2. Dendritic structure of a CA1 pyramidal cell filled with BDA. The drawing illustrates the subclassesof dendrites distinguished
and sampled in the study (illustrated by light micrographs in Fig. 1). Two types of dendrites were classified in stratum oriens.
Proximal basal dendrites carried no spines or were sparsely spinous (oriens/proximal).Spine density increased until the first branch
point (-50 p~m),where the second subclass of dendrites (oriens/distal)began; these carried large numbers of spines throughout their
extent. In stratum radiatum, four subclasses of dendrites were distinguished.The thick apical dendritic trunk was divided into three
segments: a proximal part with no spines (radiatum/thick/proximal),a medial sparsely spiny part (radiatum/Thicldmedial)and a
densely spiny distal part (radiatum/Thick/distal).The fourth type correspondedto the thin side branches (radiatum/thin)arising from
the thick apical dendrite. In stratum lacunosum-moleculare,three subclassesof dendrites were identifiedon the basis of diameter and
spine density: dendrites which were a continuation of the radiatum thick dendrites possessed fewer spines than the radiatum
dendrites, and were relatively thick (1-m/Thick). These dendrites became thinner and sparsely spinous (l-m/medium). In the
more distal apical parts, the dendrites were nearly spine-free, and often became beaded and very thin (l-m/thin). For every dendritic
subclass the density of asymmetrical, symmetrical (left and middle numbers, respectively, number/l ~m) and the proportion of
symmetrical synapses (right number) are shown. Scale bar= 100 ~m.
 ~,.jjy '~ 1-m/thin
~ 10.4 0i0917%1
l-m/medium
1-m/Thick

/ radiatum
radiatum/thin \ Thick distal
16.9 ~1
3.5 o.15 3% I r

radiatum
Thick medial
12.3 0+5 18%I

radiatum
Thick proximal
0.03 1!7 98%I
oriens/proximal
0.64 0i61 48%

oriens/distal
"-j[3.4 0.1 3%1

density of ratio of
excitatory inhibito~ inhibito
input input : input

Fig. 2.
532 M. Megias et aI.

Table 1. Length and proportion of different dendritic subclasses

A 160001
Length (~m) Contribution Dendrite
140004 to total diameter (Ixm)
1ooot .t .T,
I0O0O] • •
Ori/dist 3815 + 1023 33.0% 0.3; 0.25-0.45
8°°°4
600041I I
I I
Ori/prox 382.8 + 129.1 3.3% 0.7; 0.50-0.9
Rad/T/prox 114.2 ± 19.4 1.0% 2.1; 1.8-2.5
Rad/T/med 117.4 + 47.6 1.0% 2.0; 1.6-2.2
Rad/T/dist 311.3 ± 137.5 2.7% 1.2; 1.0-1.5
Rad/t 4095 + 895 35.5% 0.5; 0.45-0.55
L-M/T 283.0 ± 174.8 2.5% 1.1; 0.8-1.2
L-M/M 610.2 ± 216.6 5.3% 0.6; 0.3-0.8
L-M/t 1818 ± 631 15.7% 0.2; 0.15-0.4
Total 11549 - 2010
B Ori
Rad
4198 - 1056
4638 + 1022
36.3%
40.2%
L-M 2712 + 873 23.5%

o
4000t
3o001 •
i
•
Proximal
Distal
Apical
Basal
497.0 ± 140.9
11052 ± 1992
7350 + 1494
4198 + 1056
4.3%
95.7%
63.6%
36.4%
T, Thick; M, medium; t, thin dendrites, respectively. The most char-
'E 1000~ m ~ ~ L ~ L ~ acteristic diameter value is followed by the range of diameters
within a dendrite subclass. Rad, stratum radiatum; Ori, stratum
oriens; L-M, stratum lacunosum-moleculare; prox, med, dist repre-
• g g sent proximal, medial and distal parts, respectively.

5D) were associated with asymmetrical synaptic special-

Fig. 3. Dendritic length of CAI pyramidal cells. (A) Total dendritic
length, length by layers, by distance from the soma and length of
apical versus basal dendritic tree. (B) Length of each dendritic
subclass examined. Note that the majority of dendrites belong to
the densely spiny thin oriens and radiatum dendrites.
izations. We found a large difference in synaptic organi-
zation between stratum lacunosum-moleculare dendrites
and the relatively similar strata oriens and radiatum
dendrites (Figs 4, 5). Therefore, the two sets of dendrites
will be described separately.

length of dendrites in each layer were calculated (Fig.

3B, Table 1). The total dendritic length was
11.5-+ 2.0 mm and it was distributed unimodally. The
contribution of the strata radiatum and oriens dendrites
to the total dendritic length was similar (4.63 - 1.0 mm
and 4.2 + 1.0 mm, respectively), whereas the stratum
lacunosum-moleculare dendrites represented approxi-
mately 2.7 -+ 0.8 mm. Thin dendrites, densely covered
by spines in strata oriens and radiatum, represented more
than 60% of the total dendritic length. The contribution
of 1-m/thin dendrites is also significant. The apical
branches form about two thirds of the total dendritic tree.
Electron microscopy
The number and ratio of excitatory and inhibitory
inputs terminating on each subclass of dendrite was
calculated from electron microscopic observations of
serial sections. 'Excitatory' and 'inhibitory' synapses
were identified in two ways. In a smaller set of samples,
postembedding G A B A immunostaining was made on
serial sections of all subclasses of dendrites. Inhibitory
terminals were identified on the basis of accumulated
immunogold particles. On a larger set of samples the
symmetrical or asymmetrical nature of the synapses
was identified. The two methods gave similar results.
All symmetrical synaptic contacts were formed by
GABA-positive (Figs 4D, E, 5F, 7E, F) terminals,
whereas the GABA-negative terminals (Figs 4D, E,
Strata radiatum and oriens dendrites. In agreement
with previous studies m the asymmetric synapses termin-
ated on spines, regardless of the subclass of dendrite
(Figs 4, 9B). Every spine received only one synaptic
contact that was always asymmetrical. The morpho-
logical heterogeneity of spines was similar to that
reported in previous studies. 5'2° No correlation was
apparent between the dendrite subclasses and the spine
size or shape. The size of the synapses also varied (Fig.
4C). By counting the number of sections spanned by a
single synapse we compared the size of asymmetrical
synapses on different dendritic subclasses. Analysis of
30 synapses for each dendritic subclass resulted in simi-
lar, unimodal distributions, with extents ranging from
two sections (approximately 20%) to six sections
(approximately 15%). Synapses most often occupied
three to four sections (approximately 30%).
The density of spines and thus the density of asym-
metrical synapses varied among the different dendrite
subclasses. On oriens/proximal and radiatum/Thick/
proximal dendrites, there were no asymmetrical contacts.
In contrast, on radiatum/Thick/distal dendrites, the site
of the highest spine (synapse) density, there were 6.98
asymmetrical synapses (spine) per micrometer. The oriens/
distal and radiatum/thin dendrites were extensively
covered by a rather similar density of spines (3.08 and
3.52 synapses/txm, respectively; Figs 2, 6A and Table 3).
Perforated (or partitioned) asymmetrical synapses
 Synaptic convergence onto CA1 pyramidal cells 533

Fig. 4. Synaptic inputs on dendrites in strata oriens and radiatum. (A) Thick apical dendrite (radiatum/Thick/distal) located in distal
radiatum. The BDA staining reliably visualized all spines. (B) and (C) Consecutive sections from the field shown by a rectangle in A.
Asymmetrical synapses were always found on spines (curved arrows), whereas symmetrical synapses were located on the shafts
(white arrowhead). Sometimes symmetrical synapses formed another synaptic contact with additional dendrites (double arrowhead).
The asymmetrical synapses showed large differences in size and contacted spines of different types and sizes. The larger spines
spanned five to seven sections (curved arrow on the left), whereas small spines can be seen only in two sections (curved arrow on the
right). (D) and (E) In both strata oriens (D) and radiatum (E), all the terminals tested for GABA immunoreactivity were negative on
spines (curved arrow) and positive on the shafts (arrowhead). Scale bars = I ~xm (A), 0.25 p~m (B-E).

were observed on spines of different dendritic subclasses.
Because such structures have been implicated in synaptic
plasticity,12 we recorded their occurrence. W e found no
difference in the ratio of perforated synapses on spines of
different dendritic segments in these layers. They
appeared to distribute randomly along the different
spiny dendrites and represented 10.7% of the asym-
metrical synapses.
Symmetrical synapses on strata radiatum and oriens
dendrites were always located on dendritic shafts (Figs
4 B - D , 7B, D, 9C). Their density and distribution
c o m p l e m e n t e d the distribution of the spines and thus of
asymmetrical synapses. They were most abundant on
spineless proximal dendrites (oriens/proximal and radia-
tum/Thick/proximal), where less asymmetrical synapses
were found. The number of symmetrical synapses per
micrometer increased towards the soma, i.e. the radia-
turn/Thick/proximal and oriens/proximal dendrites
showed 1.72 and 0.61 symmetrical synapses per micro-
meter, respectively, whereas the radiatum/thin and
oriens/distal showed 0.11 symmetrical synapses per
micrometer (Table 2). Close to the soma, synapses
were nearly exclusively inhibitory (98.1% on radiatum/
Thick/proximal and 48.6% on oriens/proximal dendrites),
534 M. Megfas et al.

Fig. 5. Pyramidal cell dendrites in stratum lacunosum-moleculare. The synaptic organization of stratum lacunosum-moleculare
dendrites differed from dendrites in other layers. (A), (B) Spines originating from dendritic shafts (S) in stratum lacunosum-
moleculare were larger than in strata radiatum and oriens. Presynaptic boutons were also larger (curved arrow) and often formed
perforated synapses (straight arrow). Symmetrical synapses making synaptic contact on spines could also be seen (arrowhead).
These inhibitory terminals frequently contacted other postsynaptic targets (double arrowhead). (C), (D) Medium dendrites (l-m/
medium) possessed few spines (black curved arrow in D) and often received asymmetrical synapses from GABA-negative terminals
onto their shafts (white curved arrow). Asterisks (in D) show GABA-positive profiles. (E) Perforated synapses were observed both
on spines (straight arrow) and dendritic shafts (S, curved arrow). The spine shown was innervated by a terminal that formed
symmetrical synapses on it (arrowhead) and also on a dendritic shaft (double arrowhead). (F) Symmetrical synapses (arrowhead)
in stratum lacunosum-moleculare were formed by GABA-positive terminals. Scale bars = 1 txm (C), 0.25 Ixm (A, B, D-F).

whereas their ratio was only 3 . 3 - 2 . 9 % on the thin spiny
dendrites in strata oriens and radiatum. Symmetrical
synapses were in general larger than asymmetrical
synapses, and were present in five to 14 sections.
S t r a t u m l a c u n o s u m - m o l e c u l a r e . As apical dendrites
entered stratum lacunosum-moleculare, spine density
dropped to 1.72 spines per micrometer. This density was
decreased to 0.60 spines per micrometer in 1-m/medium
dendrites, and to 0.37 spines per micrometer in the l-m/
thin dendrites (Fig. 1 F - H ; Table 2). In contrast to the strata
radiatum and oriens, asymmetrical synaptic contacts could
also be observed on dendritic shafts (Fig. 5 C - E ) , mainly
on the l-m/medium and l-m/thin dendrites, so in this region
the density of asymmetrical synapses was higher than spine
density (Fig. 6; Table 2). In general the size of asym-
metrical synapses was larger than in the strata radiatum
and oriens. Asymmetrical synaptic contacts frequently
spanned five to six sections (30%), while small synaptic
contacts ranging from one to two sections represented only
10-15% of all asymmetrical synapses. On the 1-m/thin
dendritic segments, where spines were infrequently
observed, asymmetrical synapses were located on dendritic
varicosities. The ratio of asymmetrical synapses showing
perforated synaptic junctions was also higher than in strata
oriens and radiatum. The perforated synapses were most
 Synaptic convergence onto CA1 pyramidal cells
A 8
II °GA.A+
m .GABA
2 VIP terminals, it suggests that the majority of the somatic
/ + ,-oOO -,+ -.
1°°111!i1
B and 23 synaptic contacts, respectively (Fig. 7G), giving
90%~
so~
7o~
60%J
so%1
40%-t
10%J
'~ 0%:
Fig. 6. Density and ratio of excitatory and inhibitory synapses on
different subclasses of dendrite. (A) Density of asymmetrical and
symmetrical synapses expressed as number per ixm. (B) Ratio of
asymmetrical and symmetrical synapses on different dendritic
subclasses. Note that on dendrites giving the majority of total dendri-
tic length (ori/dist and rad/t) the GABA-negative inputs are much
more frequent than GABA-positive inputs.
abundant on 1-m/Thick (34, 37%) and 1-m/medium (47,
89%) dendrites. They represented 28.7% of the total asym-
metrical synapses in the stratum lacunosum-moleculare
(Figs 5A, B, E, 8 and Table 2).
The ratio of symmetrical synapses was higher (15%,
Figs 6, 8C and Table 2) on all subclasses of stratum
lacunosum-moleculare dendrites than in the strata oriens
and radiatum dendrites, excluding the proximal region.
The number of symmetrical synapses decreased with
dendritic thickness from 0.28 _+ 0.19 synapses per micro-
meter on 1-m/Thick dendrites to 0.09 -+ 0.06 synapses per
micrometer on 1-m/thin dendrites. Symmetrical synapses
were located mainly on shafts (Fig. 5F), but around 10%
of them contacted spine heads (Fig. 5A, B, D).
Somatic input. The number of synapses on the pyra-
midal cell bodies was obtained by serial reconstruction of
complete somata (n = 5). All synaptic inputs converging
onto the somata were symmetrical (Fig. 7A, C) and
GABA-positive (Fig. 7E). On average, 91.6_+ 12 syn-
apses per cell body were found. With soma surfaces of
465.4 -+ 50.2 sq. micrometer, the density of synapses is
thus 19.91 synapse/100 sq. micrometer. Synapses were
distributed all over the soma surface without evident
clustering, although areas in close contact with other
somata were not innervated. CA1 pyramidal cells
receive somatic inhibitory input from two distinct sets
of basket cells: l those containing parvalbumin and
535
those immunopositive for CCK/VIP. The axon terminals
of the two types of cells can be differentiated in the
electron microscope, because only CCK/VIP terminals
contain dense-core vesicles. In our sample, 18% of the
presynaptic terminals contained at least one dense-core
vesicle when examined in the serial sections. Although
this might be a minimum estimate for the ratio of CCK/
input derives from parvalbumin-containing basket cells.
Axon initial segments. Four axon initial segments were
serially sectioned (36-, 32-, 24- and 50-~xm-long
segments were reconstructed) and received 25, 27, 22
0.68 synapses per micrometer. However, the distribu-
tions were not homogeneous; portions with different
coverage of synaptic terminals could be observed. One
of the initial segments was followed until the first col-
lateral branchpoint ( - 5 0 Ixm from the soma), and syn-
aptic contacts were still apparent at this distance. Since
axons were not followed until the beginning of myelina-
tion, these values may under-estimate the total number of
axo-axonic synapses per pyramidal cell, and therefore
were not included when the total number of synapses
per cell was calculated.
Total number of synapses
The total number of synaptic inputs on CA1 pyramidal
cells was calculated from the length of different dendrite
subclasses and the associated values for synapse density.
We estimate that a typical CAI pyramidal cell receives
32,351 _+ 5486 synaptic contacts on its dendritic tree
(Fig. 8A, Table 3), of which 94.7% are asymmetrical
and only 5.3% are symmetrical (Fig. 8B, Table 3). The
majority of the excitatory input is concentrated in stratum
radiatum (55.1%, Table 3) and to a smaller extent in
oriens (39.1%), because a considerable fraction of the
total dendritic tree is formed by the densely spiny, thin
dendrites in these layers (Fig. 3B). The total number of
spines (30,382 -+ 5214) is less than the total number of
GABA-negative synapses (30,637 _+ 5259) since, in stra-
tuna lacunosum-moleculare, GABA-negative synapses
terminate on both dendritic shafts and spines. As with
the organization of the dendritic tree, a larger part of
the synapses is located on the apical dendritic tree
(60.9%), than on the basal dendritic tree (39.1%). For a
typical pyramidal cell, perforated synapses form only
10.2% of the total, and the majority of these contacts
are on different subclasses of the stratum lacunosum-
moleculare dendrites (7.4-48.0%, depending on sub-
class; Fig. 9A). Unconventionally located asymmetrical
synapses on stratum lacunosum-moleculare dendritic
shafts (10.7-19.2%) account for only 0.8% of the total
synapses (Fig. 9B).
As shown in Fig. 8C, inhibitory synapses are concen-
trated on the soma and proximal dendrites (40.5% of all
symmetrical synapses). GABAergic synapses terminat-
ing on pyramidal cell soma and axon initial segments,
the only inputs to these domains, form 7.2% of all inhi-
bitory synapses. Inhibitory inputs predominate (98.1% of
536 M. Meg/as et al.

Fig. 7. The somatic and proximal dendritic region of a CA1 pyramidal cell is shown in (A). Open arrows identify segments enlarged
in B-D. Pyramidal cells received exclusively symmetrical synapses in their perisomatic region; they were found on the soma (C) and
on the proximal apical (B) and basal dendrites (D). All symmetrical synapses studied with postembedding GABA immunostaining
were positive (arrowheads) both on the somata (E) and the proximal apical dendrites (F). (G) The axon initial segments received only
symmetrical synapses (arrowhead). Scale bars = 2.5 Ixm (A), 0.25 Ixm (B-G).

synapses are symmetrical) on the radiatum/Thick/proximal
dendrites, and represent around half of the synapses
(48.6%) on oriens/proximal dendrites. These proximal
dendrites, which form only 6.7% of the total dendritic
tree, carry 33.4% of the total symmetrical synapses
converging onto a pyramidal cell. As shown above, the
ratio of symmetrical synapses on different subclasses
of stratum lacunosum-moleculare dendrites is higher
(14.1-17.0%) than on other parts of the dendritic tree.
Another unique feature of symmetrical synapses in stratum
lacunosum-moleculare is that they sometimes target
dendritic spines. These contacts form only 0.08% of
the total synapses (Fig. 9C).
DISCUSSION
The main findings of the present study are: (1) Pyra-
midal cell dendrites are heterogeneous in diameter, spine
density and in the number of asymmetrical and sym-
metrical synaptic inputs. Strata radiatum and oriens
dendrites (SRO) are similar to each other both at the
light and electron microscopic levels and are different
from stratum lacunosum-moleculare dendrites (SLM)
with respect to dendritic shaft diameter, branching
pattern and spine density. (2) The ratio of symmetrical
versus asymmetrical synapses on different parts of the
CA1 pyramidal dendritic tree is highly variable but
consistent with dendrite type. (3) The majority of the
excitatory input arrives onto spiny SRO dendrites.
Here, asymmetrical synapses terminate exclusively on
dendritic spines, while symmetrical synapses contact
dendritic shafts at much lower density. (4) Spine-free or
sparsely spiny proximal dendritic shafts receive only
symmetrical synapses, similar to somata and axon initial
segments, thus inhibition is concentrated in the perisomatic
region. (5) Asymmetrical synapses in stratum lacunosum-
moleculare are larger than in other layers, they are
frequently perforated and also contact dendritic shafts. In
this layer, symmetrical synapses, whose percentage is rela-
tively high, also terminate on dendritic spines.
Pyramidal cells had a total dendritic length of approxi-
mately 12 mm, consistent with previous results. 2,5,28,31
Earlier estimates of the total synaptic inputs were based
on the observation that nearly all the asymmetrical
synapses terminate on spines of hippocampal pyramidal
cells, j9 Therefore, excitatory inputs have been largely
quantified from light microscopic measurements of
spine density. This method may underestimate the
number of spines, since smaller spines, and those located
under and above the dendritic shaft, may not be detected
in the light microscope. This might explain the reported
lower spine density values (0.7-1.9 spines per micro-
meter; see Refs 2, 4) compared to the present measure-
ments at the electron microscopic level (3.0-3.5 spines
per micrometer on spiny SRO dendrites). Our calculations
gave a density similar to that reported by Bannister and
 Synaptic convergence onto CA I pyramidal cells 537

--+
,~<

&

,,< --: c~ :5

E
+
~.~
~D ¢).

~ _ ~ ~
I
c ~
<D

~ d - ~ - ~
"U
m +~ +~ +~ +~ +~ +~ +~ +~ +~ +~ +1 +1 +1
.&
(o
"T.

o
& 4-1 +1 +1 +1
Y. m Z
._=
E

~5

,o dd d~dd~d +1 +1 +1 +1

e [-
+ ~ c~CD

-a +l +1 +l +1 +l +[ +1 +1 +l +1 +1 +1 +1

[--
©

,q.~. o.
[--
_~R~~~-,
+1 +l +1 +1 +1 +1 +1 +i +1 +1 +1 +1 +1

E
c,,i

~C~ ,
+1 +i +1 H +1 +l +i +1 +1 +i +1 H +l '-1
g ~1 ~-P'-

E
c~

e"h ,,~-
H +1 +1 H +l +1 +1 +1 +l +1 -H ÷l +1 O

t~,l P- ("q ~

E E

~o~,,'.~ '~-
538 M. Megias et al.

A A 60- perforated
40000 ==total s y n a p s o s
35000,
DGABA+ 40
30000 FI s p i n e s
25000 30
20000 20
15000'
10000
5000'
0 , ~ , , ~-'~--~1 ,

2
B ~~ o ,
B C
÷ ~25 ] GABA- on shaft on spine

o, . . . . - 0,m,, ,, m

2 Fig. 9. Ratio of special types of synapses on subclasses of dendrites.

C100% (A) Proportion of perforated synapses. (B) GABA-negativeterminals
90%j ............ on dendritic shaft. (C) GABA-positive terminals on spines. Note the
large difference between stratum lacunosum-moleculare dendrites
70% ii!~,~iiiil 'i~::' and the rest of the dendritic tree.

distinct synaptic organization is associated with different

subtypes of pyramidal cell dendrites. This suggest that, in
ili;iii;iiiiii functional terms, the dendritic tree of pyramidal cells
10%' ~ii,i~i......
l iiii+ii,!ii;ii!: ii;iii~!i should be divided into SRO dendrites and SLM dendrites
O%
rather than into basal and apical trees. The distinct char-
acteristics suggest that integration of synaptic inputs
arriving from the CA3 area via the Schaffer collaterals
Fig. 8. Total number of different types of synapses converging onto a in SRO differs from those arriving from entorhinal cortex
CA1 pyramidal cell. (A) Number of synapses and spines in different and different subcortical structures (e.g. nucleus reuniens
layers. (B) Ratio of GABA-negative and GABA-positiveterminals in
each layer. (C) Distribution of synapse types among different domains thalami, amygdala) onto SLM dendrites.3
of a pyramidal cell. Note that, while a significant portion of the A strong current sink has been shown in SLM 25
GABA-positive terminals is concentrated on the soma and on the following perforant path stimulation and during theta
proximal dendrites, excitatory synapses are devoid in these areas. activity, 6 suggesting that entorhinal input is powerful in
spite of its distal location. This is supported by the
Larkman, 5 with correction for hidden spines. Spine present morphological data showing larger asymmetrical
density measurements may also underestimate total synapses in SLM, and a horizontal trajectory of dendrites
asymmetrical synaptic input because, as we also showed, in this layer allowing multiple contacts from similarly-
some of them arrive onto shafts of SLM dendrites. Inhi- oriented entorhinal fibers. It should also be noted that
bitory synapses were quantified only in the present study. larger synapses are associated with more A M P A recep-
Besides the glutamatergic and GABAergic local and tors, 29 and that perforated (partitioned) synapses (having
extrinsic axons, axons of different transmitter content a high incidence in this layer) are associated with more
ascend from subcortical areas into the hippocampus. effective synaptic transmission following long-term
However, the axon terminals of subcortical projections potentiation. ~2 The arrangement of the inhibitory inputs
using monoamines or acetylcholine 34 only occasionally nicely matches the distribution of excitatory inputs. The
form classical synaptic specializations, and therefore stronger excitation is coupled with an elevated inhibitory
their contribution to the total synaptic input of pyramidal synapse ratio, providing a strict inhibitory control in
cells is minimal. SLM. That might reflect the requirement of a more
precise transmission of information from diverse after-
ents representing the sensory environment (thalamus,
Differences in inputs between strata radiatum/oriens entorhinal cortex, and amygdala) for the formation of
versus stratum lacunosum-moleculare dendrites receptive fields (place cells). 24
Our light and electron microscopic data show that In addition to differences in synaptic inputs to SLM
 Synaptic convergence onto CA1 pyramidal cells
and SRO, we also observed a variation in the size of
asymmetrical synapses contacting distinct dendritic
subclasses. Rather small synapses spanning only two
consecutive sections formed one fifth of the synaptic
inputs on SRO dendrites. Similar proportions of small
synapses were supposed to be 'silent' synapses 22 in the
Schaffer-collateral/CA1 pyramidal cell input system by
demonstrating that they lack AMPA receptors. 2~ This
may suggests that around one fifth of the excitatory
synapses converging onto a pyramidal cell from the
Schaffer-collateral system may be 'silent'.
h~hibitory input is concentrated in the perisomatic region
The thick proximal dendrites represent the gateway tbr
synaptic potentials to reach the soma. Very few, if any,
excitatory synapses are found on these dendrites, and the
density of inhibitory contacts is the highest here. The
arrangement of synaptic inputs on these dendrites thus
resembles that on the somata and axon initial segments,
which receive exclusively inhibitory inputs. Thus, the
thick, proximal apical dendrite is essentially equivalent
to the cell body in terms of synaptic organization. Conse-
quently, a large part of the cell's inhibitory terminals
converges onto the perisomatic region (40%, Fig. 8C).
This suggests that as dendritic inhibition is sparse, the
perisomatically clustered inhibition--which controls the
output of the neurons27--does not perform a selective
control over individual dendritic excitatory synaptic
events, but rather exercises a non-selective veto function.
There is, however, specificity associated with different
dendritic inhibitory cells. Their axons terminate in speci-
fic layers and thus they control different sets of input
pathways. 7'16-1~ However, as shown by Miles et al., 26
the sparse dendritic inhibition can only be effective if
large populations of dendritic inhibitory cells are
synchronously activated.
Comparison of interneurons and pyramidal cells
In a recent paper, ~ we described the organization of
synaptic inputs on three different types of CA1 area inter-
neurons (parvalbumin-, PV, calbindin D28k-, CB, and
calretinin-, CR, containing cells) in much the same way
as in this study. There are clear differences and simi-
larities in the organization of the inputs compared to
pyramidal cells.
The differences are that pyramidal cells receive several
times more (two to 10 times) synaptic inputs (-32,000)
than interneurons (--3000-16,000, consistent with the
fact that their total dendritic length is longer than for
interneurons), and the ratio of inhibitory terminals on
them is lower ( - 3 % for pyramidal cells, - 4 - 3 0 % for
interneurons). An important feature shared with inter-
neurons is that the inhibitory terminals are concentrated
in the perisomatic region. However, whereas inhibitory
cells receive both excitatory and inhibitory somatic
REFERENCES
Acsady L., Gores 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.
539
inputs, pyramidal cell somata receive only inhibitory
input, s The innervation of the axon initial segment is
also different. Interneurons receive only a few (7-10)
axo-axonic synapses onto the first few micrometers of
their axon initial segment, while the examined pyramidal
cells received over 25 inhibitory terminals distributed
along the sampled length (up to 50 txm) of the axon
initial segment. These features may at least partly explain
that, despite the large number of converging excitatory
inputs, the average activity level in pyramidal cells is
significantly lower than in interneurons. 9
We detected differences in synaptic inputs to distinct
types of interneurons. CB and CR cells received rela-
tively fewer inputs (3800 and 2200, respectively) with
a high ratio of inhibitory inputs (32 and 23%, respec-
tively). In contrast, PV cells were contacted by a higher
number of terminals (16,300) and the ratio of inhibitory
inputs was rather low (6.4%). Thus, from the three inter-
neuron subclasses, PV cells share the most common
features with pyramidal cells: the density of excitatory
synapses is comparably high and the ratio of inhibitory
synapses is comparably low on the dendrites of the two
cell populations. Another important similarity is that the
ratio of inhibitory inputs is higher on the SLM dendrites
of both cell types, suggesting that their input from the
entorhinal cortex and other subcortical structures is under
a strong inhibitory control. Since these cells receive
inputs from similar sources, the excitatory and inhibitory
innervation of PV cells and pyramidal cells is rather
similar.
In conclusion, the present quantitative analysis
demonstrated a very high convergence of excitatory
inputs onto CA1 area pyramidal cells. Excitatory inputs,
terminating primarily on distal spiny dendrites, are
balanced by inhibitory inputs that are largely concen-
trated in the perisomatic region. This separation of exci-
tatory and inhibitory inputs suggests that dendritic
excitatory events are vetoed by the potent perisomatic
inhibition, in'espective of their origin. The relatively
high ratio of inhibitory terminals in SLM suggests that
entorhinal cortical input is under a stronger inhibitory
control than Schaffer collateral input. The similarity in
the synaptic inputs of pyramidal cells and PV-containing
interneurons--including basket and axo-axonic cells--
suggests that the dynamics of perisomatic inhibition
controlling the pyramidal cell's output are closely
coupled to the activation of the pyramidal cell itself.
Acknowledgements--We are grateful to Drs L. Acsfidy, K. Kaila
and R. Miles for helpful discussions and comments on the
manuscript, and to Mrs E. Bor6k, Mr G. Goda and Mrs E.
Oswald for excellent technical assistance. This work was
supported by the Howard Hughes Medical Institute; the
McDonnell Foundation; NIH (MH 54671); FPI grants,
Ministerio de Educi6n y Ciencia, Spain; and OTKA (Hungarian
Scientific Research Found, T23261), Hungary.
540 M. Megfas et al.

2. Amaral D. G., Ishizuka N. and Claiborne B. (1990) Neurons, numbers and the hippocampal network. Prog. Brain Res. 83, 1-11.
3. Amaral D. G. and Witter M. P. (1989) The three-dimensional organization of the hippocampal formation: a review of anatomical data.
Neuroscience 31, 571-591.
4. Andersen P. and Soleng A. F. (1998) Long-term potentiation and spatial training are both associated with the generation of new excitatory
synapses. Brain Res. Rev. 26, 353-359.
5. Bannister N. J. and Larkman A. U. (1995) Dendritic morphology of CAI pyramidal neurones from the rat hippocampus: II. Spine distributions.
J. comp. Neurol. 360, 161-171.
6. Bragin A., Jando G., Nadasdy Z., Hetke J., Wise K. and Buzsaki G. (1995) Gamma (40-100 Hz) oscillation in the hippocampus of the behaving
rat. J. Neurosci. 15, 47-60.
7. Buhl E. H., Halasy K. and Somogyi P. (1994) Diverse sources of hippocampal unitary inhibitory postsynaptic potentials and the number of
synaptic release sites. Nature 368, 823-828.
8. Colonnier M. (1968) Synaptic patterns on different cell types in the different laminae of the cat visual cortex. An electron microscope study.
Brain Res. 9, 268-287.
9. Csicsvari J., Hirase H., Czurko A. and Buzsaki G. (1998) Reliability and state dependence of pyramidal cell-interneuron synapses in the
hippocampus: an ensemble approach in the behaving rat. Neuron 21, 179-189.
10. Csicsvari J., Hirase H., Czurko A., Mamiya A. and Buzsaki G. (1999) Oscillatory coupling of hippocampal pyramidal cells and interneurons in
the behaving Rat. J. Neurosci. 19, 274-287.
11. Freund T. F. and Buzsaki G. (1996) Interneurons of the hippocampus. Hippocampus 6, 345-470.
12. Geinisman Y., Detoledomorrell L., Morrell F., Heller R. E., Rossi M. and Parshall R. F. (1993) Structural synaptic correlate of long-term
potentiation--formation of axospinous synapses with multiple, completely partitioned transmission zones. Hippocampus 3, 435-446.
13. Gray E. G. (1959) Axo-somatic and axo-dendritic synapses of the cerebral cortex: an electron microscope study. J. Anat. 93, 420-433.
14. Gulyas A. I., H~jos N. and Freund T. F. (1996) Interneurons containing calretinin are specialized to control other interneurons in the rat
hippocampus. J. Neurosci. 16, 3397-3411.
15. Gulyas A. I., Megias M., Emri Z. and Freund T. F. (1999) Total number and ratio of excitatory and inhibitory synapses converging onto single
interneurons of different types in the CA 1 area of the rat hippocampus. J. Neurosci. 19, 10,082-10,097.
16. Gulyas A. I., Miles R., Hfijos N. and Freund T. F. (1993) Precision and variability in postsynaptic target selection of inhibitory cells in the
hippocampal CA3 region. Eur. J. Neurosci. 5, 1729-1751.
17. Halasy K., Buhl E. H., Lorinczi Z., Tamas G. and Somogyi P. (1996) Synaptic target selectivity and input of GABAergic basket and bistratified
interneurons in the CAI area of the rat bippocampus. Hippocampus 6, 306-329.
18. Han Z. S., Buhl E. H., Lorinczi Z. and Somogyi P. (1993) A high degree of spatial selectivity in the axonal and dendritic domains of
physiologically identified local-circuit neurons in the dentate gyrus of the rat hippocampus. Eur. J. Neurosci. 5, 395-410.
19. Harris K. and Kater S. B. (1994) Dendritic spines: cellular specializations imparting both stability and flexibility to synaptic function. Ann. Rev.
Neurosci. 17, 341-371.
20. Harris K. M., Jensen F. E. and Tsao B. (1992) Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CAI) at
postnatal day 15 and adult ages: implications for the maturation of synaptic physiology and long-term potentiation. J. Neurosci. 12, 2685-2705.
21. Hodgson A. J., Penke B., Erdei A., Chubb I. V. and Somogyi P. (1985) Antisera to -/-aminobutyric acid. I. Production and characterization using
a new model system. J. Histochem. Cytochem. 33, 229-239.
22. Isaac J. T. R., Nicoll R. A. and Malenka R. C. (1995) Evidence for silent synapses: implications for the expression of LTP. Neuron 15,427-434.
23. Isbizuka N., Cowan W. M. and Amaral D. G. (1995) A quantitative analysis of the dendritic organization of pyramidal cells in the rat
hippocampus. J. comp. Neurol. 362, 17-45.
24. Katona I., Acsady L. and Freund T. F. (1999) Postsynaptic targets of somatostatin-immunoreactive interneurons in the rat hippocampus.
Neuroscience 88, 37-55.
25. Leung L. S., Roth L. and Canning K. J. (1995) Entorhinal inputs to hippocampal CAI and dentate gyrus in the rat: a current-source-density
study. J. Neurophysiol. 73, 2392-2403.
26. Miles R., Toth K., Gulyas A. I., H~ijos N. and Freund T. F. (1994) Functional differences between dendritic and perisomatic hippocampal
inhibition. Soc. Neurosci. Abstr. 20, 301-304.
27. Miles R., Toth K., Gulyas A. I., Hfijos N. and Freund T. F. (1996) Differences between somatic and dendritic inhibition in the hippocampus.
Neuron 16, 815-823.
28. Moser M.-B., Trommald M., Egeland T. and Andersen P. (1997) Spatial training in a complex enviroment and isolation alter the spine
distribution differently in rat CAI pyramidal cells. J. comp. Neurol. 380, 373-381.
29. Nusser Z., Lujan R., Laube G., Roberts J. D., Molnar E. and Somogyi P. (1998) Cell type and pathway dependence of synaptic AMPA receptor
number and variability in the hippocampus. Neuron 21, 545-559.
30. Pyapali G. K., Sik A., Penttonen M., Buzsaki G. and Turner D. A. (1998) Dendritic properties of hippocampal CA 1 pyramidal neurons in the rat:
intracellular staining in vivo and in vitro. J. comp. Neurol. 391, 335-352.
31. Pyapali G. K. and Turner D. (1996) Increased dendritic extent in hippocampal CA1 neurons from aged F344 rats. Neurobiol. Aging 17,
601-611.
32. Somogyi P. and Hodgson A. J. (1985) Antisera to 3,-aminobutyric acid. III. Demonstration of GABA in Golgi-impregnated neurons and in
conventional electron microscopic sections of cat striate cortex. J. Histochem. Cytochem. 33, 249-257.
33. Trommald M., Jensen V. and Andersen P. (1995) Analysis of dendritic spines in rat CA I pyramidal cells intracellularly filled with a fluorescent
dye. J. comp. Neurol. 353, 260-274.
34. Umbriaco D., Garcia S., Beaulieu C. and Descarries L. (1995) Relational features of acetylcholine, noradrenaline, serotonin and GABA axon
terminals in the stratum radiatum of adult rat hippocampus (CAI). Hippocampus 5, 605-620.
35. Wolf E., Birinyi A. and Pomahazi S. (1995) A fast 3-dimensional neuronal tree reconstruction system that uses cubic polynomials to estimate
dendritic curvature. J. Neurosci. Meth. 63, 137-145.
36. Wouterlood F. G., Saldana E. and Witter M. P. (1990) Projection from the nucleus reuniens thalami to the hippocampal region: light and
electron microscopic tracing study in the rat with the anterograde tracer Phaseolus vulgaris-leucoagglutinin. J. comp. Neurol. 296, 179-203.

(Accepted 18 October 2000)