Interneurons are critical in regulating the excitability of principal cells in neuronal circuits, thereby modulating the output of neuronal networks. We investigated synaptically evoked inhibitory responses in CA3 pyramidal cells mediated by metabotropic glutamate receptors (mGluRs) expressed somatodendritically by interneurons. Although pharmacological activation of mGluRs in interneurons has been shown to enhance their excitability, the inability to record mGluR-mediated synaptic responses has precluded detailed characterization of mGluR function in hippocampal interneurons. We found that a single extracellular pulse in CA3 stratum pyramidale was sufficient to induce disynaptic inhibitory responses mediated by postsynaptic mGluRs of the interneurons in CA3 pyramidal cells of hippocampal slice cultures. The disynaptic inhibitory response followed a short-latency monosynaptic inhibitory response, and was observed at stimulus intensities evoking half-maximal monosynaptic IPSCs. Synergistic activation of mGluR1 and mGluR5 was required to induce the full inhibitory response. When recordings were obtained from interneurons in CA3 stratum radiatum or stratum oriens, a single extracellular stimulus induced a slow inward cationic current with a time course corresponding to the slow inhibitory response measured in pyramidal cells. DCG IV, a group II mGluR agonist, which specifically blocks synaptic transmission through mossy fibres, had no effect on mGluR-mediated synaptic responses in interneurons, suggesting that feed-forward inhibition via mossy fibres is not involved. Thus, postsynaptic mGluR1 and mGluR5 in hippocampal interneurons cooperatively mediate slow feedback inhibition of CA3 pyramidal cells. This mechanism may allow interneurons to monitor activity levels from populations of neighbouring principal cells to adapt inhibitory tone to the state of the network.
Neuronal networks in the brain integrate a diversity of incoming information, which is then processed into a meaningful output. The operation of neuronal networks is suggested to depend on the precise circuit connections and dynamic interactions between inhibitory interneurons and excitatory principal cells (Freund & Buzsáki, 1996). Two types of inhibitory circuits can be distinguished giving rise to feed-forward or feedback inhibition. Feed-forward inhibitory conductances shunt excitatory signals onto dendrites with a synaptic delay thereby restricting the effects of depolarizing inputs to target cells (Alger & Nicoll, 1982; Buzsáki & Eidelberg, 1982). Feed-forward inhibition efficiently modulates the kinetics of EPSPs, especially their rising phase (Karnup & Stelzer, 1999), and has recently been shown to set the temporal window for the integration of excitatory inputs in CA1 pyramidal cells (Pouille & Scanziani, 2001). Feedback inhibition regulates the spread of neuronal excitation within a network as a result of polysynaptic recurrent excitatory inputs onto interneurons (Kandel et al. 1961; Andersen et al. 1963; Miles, 1990). In addition, feedback inhibition, with its synchronous refractory period, facilitates synchronization of principal cell discharge and plays a critical role in the generation of hippocampal oscillations (Buzsáki et al. 1992; Fisahn et al. 1998).
Pharmacological activation of somatodendritic metabotropic glutamate receptors (mGluRs) induces slow responses such as depolarization or inward currents in interneurons (Miles & Poncer, 1993; McBain et al. 1994; Poncer et al. 1995; Llano & Marty, 1995; Strata et al. 1995; Yanovsky et al. 1997; van Hooft et al. 2000). However, because postsynaptic mGluRs exhibit an extrasynaptic localization in interneurons (Luján et al. 1996), synaptic signalling by mGluRs in interneurons may normally be rather restricted. To assess the functional role of postsynaptic mGluRs directly, it is therefore necessary to record synaptic currents mediated by mGluRs in interneurons, which has hitherto not been possible. Here, we have taken advantage of the intact connectivity between neurons in hippocampal slice cultures to record for the first time mGluR-mediated inhibitory synaptic responses in hippocampal CA3 pyramidal cells in response to single pulses at intensities inducing submaximal ionotropic inhibitory synaptic responses. In our experiments we have addressed the following questions: (1) does disynaptic mGluR-dependent inhibition of CA3 pyramidal cells involve a feed-forward or a feedback circuit, (2) can synaptic activation of somatodendritic mGluRs in interneurons induce sufficient depolarization to fire action potentials, (3) which ionic conductance underlies the postsynaptic current mediated by mGluRs in interneurons, and (4) which mGluR subtypes are responsible for this synaptic response.
Preparation of slice cultures
Slice cultures were prepared from 6- to 7-day-old Wistar rat pups killed by decapitation following a protocol approved by the Veterinary Department of the Canton of Zurich, and maintained as previously described (Gähwiler et al. 1998). In brief, 400 μm thick hippocampal slices were attached to glass coverslips, placed in sealed test tubes with serum-containing medium, and maintained in a roller-drum incubator at 36 °C for 14-28 days.
Cultures were superfused with an external solution containing 148.8 mm Na+, 2.7 mm K+, 149.2 mmCl−, 2.8 mm Ca2+, 2.0 mm Mg2+, 11.6 mm HCO3−, 0.4 mm H2PO4−, 5.6 mm d-glucose and 10 mg l−1 Phenol Red (pH 7.4). All experiments were performed at a temperature of 28 °C. Synaptic currents were evoked with a monopolar metal electrode using single pulses (100 μs, 0.5-60 μA). The stimulation intensity was adjusted to obtain a submaximal peak response at 1 min intervals. Recordings were obtained from CA3 pyramidal cells, and stratum radiatum and stratum oriens interneurons (Axopatch 200B amplifier; Axon Instruments, Union City, CA, USA) with patch pipettes (2-5 MΩ). For the CA3 pyramidal cells, recording pipettes were filled with: 130 mm caesium methanesulfonate, 10 mm Hepes, 10 mm EGTA, 5 mm QX-314-Cl, 2 mm Mg-ATP, 5 mm creatine phosphate (CrP) and 0.4 mm GTP (pH 7.25). For the interneurons, recording pipettes were filled with: 140 mm caesium acetate, 10 mm Hepes, 0.1 mm EGTA, 1 mm QX-314-Cl, 2 mm Mg-ATP, 5 mm CrP and 0.4 mm GTP (pH 7.25). The actual membrane potentials were corrected for the liquid junction potential for each internal solution. Series resistance (typically between 5 and 15 MΩ) was regularly monitored, and if a change of more than 10 % occurred, cells were excluded.
Drugs and chemicals
6-Cyano-7-nitroquinoxaline-2,3(1H,4H)-dione (CNQX, 40 μm), 3-((R)-2-carboxypiperazin-4yl)-propyl-1-phosphonic acid (CPP, 40 μm) and CGP 62349 (5 μm) were always present in the bathing fluid. Adenosine, CNQX, (2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl)glycine (DCG IV), (S)-3,5-dihydroxyphenylglycine (DHPG), (S)-α-methyl-4-carboxyphenylglycine (MCPG), 2-methyl-6-(phenylethynyl)-pyridine (MPEP) and (+)-2-methyl-4-carboxyphenylglycine (LY 367385) were purchased from Tocris Cookson (Bristol, UK); ATP, EGTA and (-)-bicuculline methochloride from Sigma/Fluka (Buchs, Switzerland); tetrodotoxin (TTX) from Latoxan (Valence, France); gabazine (SR 95531) from Sanofi-Synthelabo (Paris, France); QX-314 from Alomone Labs (Jerusalem, Israel); and CPP and CGP 62349 were kindly provided by Novartis AG (Basel, Switzerland).
Data acquisition and analysis
Signals were filtered at 2 kHz and digitally recorded on a computer using Clampex 7 software (Axon Instruments). To quantify the slow bursts of IPSCs evoked by electrical stimulation (Fig. 2 and Fig. 3), the area under the burst waveform was integrated as charge transfer. The interpolated reversal potential (Vrev) of the responses was calculated by fitting data points of the current-voltage (I-V) relationship with third-order polynomials. Numerical data in the text are expressed as means ±s.e.m. Student's t test was used to compare values when appropriate. P < 0.05 was considered significant.
Evoked slow IPSC in CA3 pyramidal cells
We evoked synaptic responses in CA3 pyramidal cells voltage clamped at 0 mV by electrical stimulation in the CA3 pyramidal cell layer in the presence of the ionotropic glutamate receptor antagonists CNQX (40 μm) and CPP (40 μm), and the metabotropic GABAB receptor antagonist CGP 62349 (5 μm) (Fig. 1A and B). A single pulse of 100 μs duration at a intensity of 5 μA induced only a fast outward current. With increasing stimulation intensity a slow outward current, which appeared to consist of a burst of IPSCs, followed the initial fast outward current (latencies: fast outward current, 3.3 ± 0.2 ms; slow outward current, 342.7 ± 20.7 ms; n= 16). A plot of the stimulus-response relationships shows that the slow outward current appeared before the fast outward current reached a plateau and increased in amplitude with increasing stimulation intensity (n= 8; Fig. 1B). The fast and the slow outward currents are synaptic responses, as both were blocked by TTX (0.5 μm; n= 4, data not shown). The I-V relationships for both the fast and the slow outward current were determined by clamping the CA3 pyramidal cells at holding potentials (Vh) from -100 to +20 mV (Fig. 1C). Although the I-V relationships show different slopes, both responses reversed close to the equilibrium potential for Cl− calculated as -88.3 mV (fast IPSC, -92.4 ± 1.6 mV; slow IPSC, -91.3 ± 1.9 mV; n= 5), demonstrating that the fast and the slow currents are both mediated by a membrane conductance specific for Cl−.
The origin of the slow IPSC
In the hippocampus, fast inhibitory synaptic transmission is mediated exclusively by GABAA receptors (Mori et al. 2002). Bath application of gabazine (10 μm), a specific GABAA receptor antagonist, blocked completely both the fast IPSC and the slow IPSC in CA3 pyramidal cells (n= 7; data not shown), indicating that the slower response is also mediated by GABAA receptors. Adenosine (100 μm), which inhibits glutamate release from pyramidal cells (Thompson et al. 1992), selectively blocked the slow response (fast IPSC, by 4.9 ± 1.7 %; slow IPSC, by 97.5 ± 0.5 %: n= 4; Fig. 2A and B). These results indicate that the circuit underlying the slow IPSC includes both GABAergic and glutamatergic synapses. Since fast glutamatergic transmission was blocked under our experimental conditions, mGluRs are likely to mediate the glutamatergic component for the slow IPSC response. Indeed, MCPG (1 mm), a broad-spectrum mGluR antagonist, blocked the slow IPSC (by 96.4 ± 1.7 %: n= 4) but not the fast IPSC (Fig. 2A and B).
Taken together, these findings are consistent with a circuit in which the fast IPSC is evoked by direct stimulation of interneurons and the slow IPSC is evoked through a disynaptic pathway involving the synaptic activation of mGluRs situated on interneurons, which in turn release GABA onto targeted CA3 pyramidal cells (Fig. 2C).
Group I mGluRs mediate the slow IPSC
The synaptically evoked slow IPSC could be mimicked by bath application of DHPG (10 μm), a group I mGluR agonist, onto CA3 pyramidal cells held at 0 mV (n= 4; Fig. 3A). In the presence of TTX (0.5 μm), DHPG failed to induce the outward current (n= 3; data not shown), indicating that activation of group I mGluRs elicits action potential firing in interneurons in order to generate the slow IPSC in postsynaptic CA3 pyramidal cells. Similarly, gabazine (10 μm) blocked the outward current in response to DHPG (n= 3; data not shown).
To identify which group I mGluR subtype expressed in interneurons mediates the slow IPSC, experiments were performed utilizing antagonists selective for either mGluR1 or mGluR5. LY 367385 (1-300 μm), a specific mGluR1 antagonist, reversibly inhibited the slow IPSC (IC50, 27.9 μm; Hill coefficient (nH), 1.3; n= 5) without affecting the fast IPSC (Fig. 3B). At 300 μm, LY 367385 appeared to also inhibit the fast IPSC (by 7.8 ± 4.7 %, P= 0.19, n= 5). MPEP (0.01-3 μm), a specific mGluR5 antagonist, irreversibly blocked the slow IPSC (IC50, 0.21 μm; nH, 1.5; n= 8), with no effect on the fast IPSC (Fig. 3C). The residual slow outward current, which persisted in the presence of MPEP (3 μm), could be completely blocked by LY 367385 (100 μm; n= 4; Fig. 3D).
Group I mGluRs mediate a slow synaptic excitatory current in interneurons
The same stimulation protocol (single pulses: 100 μs, 0.5-60 μA) evoked a slow inward current (-30.9 ± 2.9 pA, n= 41) in interneurons of the CA3 area clamped near the resting potential (-80 mV) (Fig. 4). The time course of the slow inward current corresponds to the slow IPSC recorded in CA3 pyramidal cells (Fig. 1). For these experiments, the GABAA receptor antagonist bicuculline (100 μm) was added to the standard cocktail of antagonists (CNQX, CPP and CGP 62349). No significant difference in the incidence of the slow EPSCs recorded in interneurons from the stratum radiatum or the stratum oriens was observed (stratum radiatum interneurons: 57.4 %, n= 47; stratum oriens interneurons: 46.7 %, n= 30). The stratum oriens interneurons exhibited a larger EPSC than the stratum radiatum interneurons (stratum radiatum interneurons: -26.5 ± 2.8 pA, n= 27; stratum oriens interneurons: -39.3 ± 6.0 pA, n= 14: P < 0.05). The slow EPSC is mediated by mGluRs as it was blocked by MCPG (1 mm; n= 4; Fig 4A). Furthermore, LY 367385 as well as MPEP reduced the slow EPSC by more than 50 % (LY 367385 (100 μm): 74.6 ± 12.1 %, n= 4; MPEP (1 μm): 68.6 ± 8.3 %, n= 7) and coapplication of LY 367385 (100 μm) and MPEP (1 μm) abolished the slow EPSC (n= 4), indicating that activation of both mGluR1 and mGluR5 is involved in inducing the slow synaptic response. The small rapid inward current that often preceded the slow EPSC (Fig. 4A, B and D) was resistant to CNQX, CPP, CGP 62349, bicuculline and MCPG, and may reflect the activation of P2X ionotropic ATP receptors (Mori et al. 2001). The slow EPSC at Vh values between -80 and +20 mV and a representative I-V relationship are illustrated in Fig 4B. Vrev was close to 0 mV (-10.2 ± 2.7 mV, n= 8) with a caesium-based internal patch solution, indicating that activation of a non-selective cationic conductance is likely to underlie the slow EPSC.
To determine whether this synaptically evoked inward current in interneurons is sufficient to account for the slow IPSP recorded in CA3 pyramidal cells, we obtained current-clamp recordings from interneurons. Under these conditions, evoked slow excitatory postsynaptic potentials reliably triggered action potential firing in the interneurons, and these responses were again blocked by MCPG (1 mm; n= 5; Fig. 4C).
Source of the glutamate activating mGluRs
Organotypic hippocampal slice cultures lack excitatory inputs to the CA3 area from the entorhinal cortex and the contralateral hippocampus (Gähwiler et al. 1998). Therefore, with our stimulation paradigm, in which we place the extracellular stimulating electrode in the CA3 pyramidal cell layer, glutamate reaching interneurons may be released through depolarization of both recurrent collaterals from CA3 pyramidal cells to induce feedback inhibition, and mossy fibres from dentate granule cells to induce feed-forward inhibition. To distinguish between these two inputs, we selectively inhibited mossy fibre transmission by activating group II mGluRs, which are expressed on the mossy fibre but not the associational fibre terminals (Shigemoto et al. 1997; Maccaferri et al. 1998). Bath application of the specific group II mGluR agonist DCG IV (5 μm) had no effect on the slow EPSC in the interneurons (n= 5; Fig. 4D), suggesting that the recurrent collaterals of CA3 associational fibres represent the main source of glutamate for the synaptic activation of mGluR1s and mGluR5s in CA3 interneurons.
In this study, we have characterized the mechanisms underlying slow synaptic inhibition in the hippocampal CA3 area. Our approach, allowing us to reproducibly evoke mGluR-mediated synaptic responses, greatly facilitates the characterization of these receptors in hippocampal interneurons. We find that a single extracellular stimulus evokes a synchronous burst of IPSCs in CA3 pyramidal cells by depolarizing interneurons via recurrent collaterals originating from the pyramidal cells. This form of slow feedback inhibition involves the synergistic activation of mGluR1 and mGluR5, which depolarizes interneurons by increasing a non-selective cationic membrane conductance.
Slow feedback inhibition mediated by mGluRs
The specific group II mGluR agonist DCG IV did not modify the slow EPSC evoked in interneurons (Fig. 4D), indicating that the mGluR-dependent synaptic inhibition of pyramidal cells is mediated by recurrent collaterals from CA3 associational fibres rather than by mossy fibres, and thus involves feedback rather than feed-forward inhibition. Feedback inhibition is critical for hippocampal function. Inhibitory circuits involving feedback activation of interneurons are essential in the generation of coherent network oscillations (Jefferys et al. 1996), while disruption of negative feedback loops disinhibits recurrent excitation between CA3 pyramidal cells, leading to epileptic bursting (Miles & Wong, 1987).
Interaction between mGluR1 and mGluR5
Both LY 367385, a specific inhibitor of mGluR1 (IC50= 27.9 μm; Fig. 3B), and MPEP, a specific inhibitor of mGluR5 (IC50= 0.21 μm; Fig. 3C), inhibit the slow mGluR-dependent response by more than 50 % (Figs 3B and C and 4A). Furthermore, both antagonists had to be applied to block responses completely (Fig. 3D and Fig. 4A). These results suggest that cooperativity occurs between mGluR1 and mGluR5, which are extrasynaptically localized also in interneurons (Luján et al. 1996), or between downstream elements of their respective signalling systems. Immunohistochemical data have shown that pyramidal cells usually make only a single synaptic contact with interneurons in vivo (Gulyás et al. 1993). However, a train of action potentials induced in a presynaptic CA3 pyramidal cell failed to evoke an mGluR-mediated response in a synaptically coupled interneuron (n= 7; M. Mori & U. Gerber, unpublished observations). The slow synaptic current in interneurons could be evoked with a single extracellular stimulus that simultaneously activates a population of pyramidal cells. Cooperativity may arise at the intracellular level, through the accumulation of G-proteins or second messengers activated by concomitant stimulation of several synaptic inputs onto a given interneuron. The relatively slow time course of the mGluR-mediated response (≈2 s) may reflect the prolonged elevation of intracellular signalling molecules, which in turn would promote pooling in order to activate common ion channel effectors.
In conclusion, the mGluR-mediated burst of IPSCs evoked in CA3 pyramidal cells represents a much longer-lasting form of inhibition than typical GABAA receptor-mediated fast inhibition. This longer window of response is expected to promote the temporal summation of responses impinging on the interneurons from multiple CA3 recurrent inputs, thereby allowing interneurons to filter high frequency components of neuronal discharge in CA3 pyramidal cells originating from the same excitatory recurrent inputs. In fact, the high connectivity between neighbouring CA3 pyramidal cells (Li et al. 1994; Debanne et al. 1995) may be critical in producing the synchronous activity that appears to be necessary for inducing mGluR-mediated feedback inhibition.
We thank Beat H. Gähwiler for his generous support and for providing us with slice cultures. We thank H. Blum, S. Giger, H. Kasper, A. Nussbaumer, L. Rietschin and R. Schöb for excellent technical assistance, and B. H. Gähwiler, Y. Fischer and M. Scanziani for helpful discussions and critical reading of the manuscript. This work was funded by the Swiss National Science Foundation, the NCCR on Neural Plasticity and Repair, Hartmann Müller Foundation, and the Ministry of Education, Science, and Culture of Japan.