Hungarian Academy of Sciences,
Budapest, P.O.Box 67,
H-1450, Hungary;
tel:(36-1)-2100-812
fax: (36-1)-2100-812 or (36-1)-1334-582
e-mail:
A.I.Gulyas, R.Miles, A.Sik, K.Toth, N.Tamamaki and T.F.Freund (1993) Hippocampal pyramidal cells excite inhibitory neurons via a single release site Nature 366:683-687
Transmission often fails (Fig. 1a) at excitatory connections between hippocampal neurons. Failures of synaptic transmission may reflect a low probability of transmitter release, or few release sites. We performed two types of experiments using light and electron microscopy to define the number of transmitter release sites at an excitatory connection: 1) Excitatory interactions between hippocampal pyramidal and inhibitory cells were characterised physiologically and their morphological substrate revealed after injecting biocytin into both cells. 2) We examined contacts between a single filled pyramidal cell axon and inhibitory cells stained immunocytochemically to reveal their characteristic calcium binding protein, parvalbumin.
Simultaneous intracellular records were made from CA3 pyramidal cells and inhibitory cells located close to stratum pyramidale. Inhibitory cells were identified by a distinctive firing pattern, by showing that they inhibited nearby pyramidal cells, and from their morphology revealed after biocytin injection. Excitatory interactions were detected in about 10% of the recorded pairs (Fig. 1a). EPSPs had a mean amplitude of 0.2-1.5 mV and a rise time from 10-90% of 1.2-2.8 ms (n=18). Averaging selected responses to presynaptic action potentials suggested that in 16 of 18 connections some presynaptic action potentials elicited no postsynaptic responses.
Filled cells with complete and reliable light and electron microscopy were recovered from three slices. Inhibitory cells were of different types; one had an axon arborizing in strata lacunosum- moleculare and oriens (Fig. 1a), whereas the other two were basket cells innervating predominantly stratum pyramidale. Pyramidal cell axons were intensely stained, and always terminated in a cut end on the slice surface in CA3 or CA1. In two slices single pyramidal cells were filled. In the third case two pyramidal cells were recovered, although electrical stimulation appeared to generate single action potentials.
Each axon established a single contact on a second order inhibitory cell dendrite at 80-250 æm from the soma (arrow in Fig. 1b). Electron microscopy confirmed that each contact was a conventional synapse (Fig. 2b). Presynaptic elements contained round, clear vesicles, and serial sectioning revealed a single synaptic active zone of diameter 0.4-0.8 æm.
Quantal amplitude (q) and probability of transmitter release (p) were evaluated from EPSP amplitude distributions. For the interaction shown in figure 1, q was estimated as 0.65 ñ 0.25 mV and p was 0.37 with N=1 corresponding to the single anatomically defined site. Fig. 2d shows the resulting fit of the EPSP amplitude distribution. Statistical analysis suggested 14 of the other 17 interactions involved a single release site including 2 cases with confirmatory anatomy (Fig. 3). In three cases, slightly better fits were obtained with N=2 (1) and N=3 (2).
Mean quantal amplitude was estimated as 0.79 ñ 0.20 mV (n=18). We estimate that the coefficient of variation of quantal amplitude was 21-53% (mean = 35 ñ 9%; n=18). This compares with values of 40-50% measured for miniature EPSCs in cortical and hippocampal pyramidal cells when variance from electrotonic sources was minimised. The mean value derived for probability of transmitter release was 0.75 ñ 0.19. In 4 cases p was higher than 0.9. Similar high values for p have been inferred for some excitatory terminals on spinal motorneurons3. However, our estimate may be high since experiments were made in high Ca, which enhances transmitter release, and since larger, more frequently evoked EPSPs were more readily seen.
The hypothesis that pyramidal cells make single synaptic contacts with inhibitory cells was tested further by filling single CA3 pyramidal cells with neurobiotin in vivo. Parvalbumin-containing inhibitory cells (basket and axo-axonic cells) were visualized immunocytochemically in the same sections from the entire hippocampus (Fig. 4). This pyramidal cell axon was followed through 45 sections (60 æm thick) ramifying in the CA3 and CA1 regions and extending 2.7 mm along the longitudinal axis. Figure 4 plots locations, from 9 representative consecutive sections, where axon varicosities contacted parvalbumin- positive inhibitory cell dendrites (134) or soma (1). For 93% of potential target neurones (125 out of 134) a single contact was formed by the filled axon. In another study 85% of potential pyramidal cell contacts with parvalbumin-containing neurons involved a single bouton. Electron microscopy showed that 81% of these potential contacts from light microscopy were indeed synapses with one active zone. Convergence of multiple pyramidal cell axon collaterals onto different dendrites of one parvalbumin-positive cell was not observed.
These data show that pyramidal cells form single axonal contacts with postsynaptic inhibitory neurones. This arrangement differs from other connections such as climbing fibre synapses onto cerebellar Purkinje cells, where there are 200-300 transmitter release sites. In the hippocampus, morphological studies suggest that inhibitory cells form 5-30 synaptic contacts with a single pyramidal cell, implying different strategies for inhibitory and excitatory transmission.
EPSPs initiated in interneurones by single pyramidal cell action potentials are mediated via a single morphological synapse. Thus, transmission failures and release probabilities reflect release from a single site and variation between successive postsynaptic events is variation occurring at one synaptic complex. It is thought that only one vesicle is released from a single presynaptic site, although this has not been directly verified. If so, transmission at this synapse would be mono-quantal as suggested for connections made by CA3 pyramidal cells on CA1 pyramidal cells. Mono-quantal transmission would constrain mechanisms of synaptic plasticity. If at a functional connection there is only one morphological transmitter release site, then latent release sites can not be uncovered. Presynaptic plasticity must then be limited to changes in release probability, which should saturate, or to changes in the amount of transmitter released from one site.
Our results also provide numerical values for connectivity needed to model neuronal network computation. Since over a thousand excitatory synapses terminate on a single inhibitory cell, our data suggest that more than 1000 pyramidal cells converge onto one inhibitory neuron. Approximately 2% of CA3 pyramidal cell targets are immunoreactive for parvalbumin. Thus, a pyramidal cell possessing 10- 50 thousand boutons may excite 200-1000 PV-containing inhibitory cells. EPSPs initiated by single pyramidal cells may cause inhibitory cell firing. This excitatory drive presumably contributes to the high firing frequency of many hippocampal interneurons. Thus, the activation of a single CA3 pyramidal cell may, due to high divergence in excitatory and inhibitory connections, di-synaptically inhibit thousands of pyramidal cells in the CA3 and CA1 regions. This circuitry will operate in some, but not all, behavioural states to exert a strong, widespread inhibitory control of pyramidal cell firing.
Created by Attila Gulyas at
Institute of Experimental Medicine,
Hungarian Academy of Sciences,
Budapest, P.O.Box 67,
H-1450, Hungary;
tel:(36-1)-2100-812
fax: (36-1)-2100-812 or (36-1)-1334-582
e-mail: