Major Differences in Inhibitory Synaptic Transmission onto Two Neocortical Interneuron Subclasses

Different properties of evoked GABAergic neurotransmission on FS and LTS interneurons

We performed patch-clamp whole-cell recordings from visually identified interneurons in layer V of rat sensorimotor neocortex. Under infrared videomicroscopy, interneurons were distinguished from pyramidal cells by their appearance as round, bipolar, or multipolar cells with small somata and the lack of a thick apical dendrite oriented toward the pia (Bacci et al., 2002, 2003). The firing behavior of neurons was characterized in current clamp using 600 msec depolarizing DC injections of different amplitudes and, on the basis of their membrane and firing properties, they could be classified into FS and LTS interneurons. Analysis of FS and LTS cells revealed large differences in single action potential half-width and afterhyperpolarization amplitude (Table 1). In addition, FS cells were characterized by much higher firing frequencies, more prominent spike frequency adaptation, and lower input resistance than LTS interneurons (Figs. 1C,D, inset, 2A,B; Table 1). FS cells were also characterized by the lack of rebound spikes after hyperpolarizing current step injections (Figs. 1C, 2A) and often displayed GABAergic autaptic responses after action potentials (data not shown) (but see Bacci et al., 2003). LTS interneurons were characterized by a rebound spike (Fig. 1D) or burst of action potentials (Fig. 2A) after injection of hyperpolarizing current steps (Figs. 1D, 2A). Finally, action potentials of both FS and LTS interneurons were brief compared with those of pyramidal neurons and had more pronounced and complex spike afterhyperpolarizations (data not shown). Several neurons were filled with biocytin (n = 12 and 7, for FS and LTS cells, respectively; see Materials and Methods) and reconstructed to assess their somato-dendritic structures and axonal projections (Fig. 1A,B). FS interneurons typically had a dense axonal plexus at the level of their cell body in layer V with some horizontal projections (Fig. 1A), whereas the axonal projections of LTS interneurons ascended to more superficial layers and had a less pronounced perisomatic distribution (Fig. 1B). These cell type-specific morphological features were observed in all biocytin-filled interneurons. The physiological and morphological criteria reported here have been used in several previous studies to distinguish FS and LTS interneurons from each other and from pyramidal neurons (Kawaguchi and Kubota, 1993, 1997, 1998; Cauli et al., 1997; Xiang et al., 1998; Beierlein et al., 2003; Xiang and Prince, 2003).

IPSCs evoked by extracellular stimuli (Fig. 2B,C) were isolated in the presence of ionotropic glutamate receptor blockers DNQX (10 μm) and dl-APV (100 μm) and were selectively and reversibly abolished by the GABA_A receptor blocker gabazine (10 μm; data not shown). eIPSCs in FS interneurons decayed much faster than in LTS cells (Fig. 2B-D). On average, the half-widths of the synaptic responses were 7.45 ± 0.7 and 18.46 ± 2.4 msec in FS and LTS cells, respectively (n = 19 FS; n = 16 LTS; p < 0.00005; independent t test) (Fig. 2D). The decay phase of eIPSCs, fitted with a double exponential and weighted time constants (τ_d,w; see Materials and Methods), measured 17.6 ± 1.9 and 36.0 ± 6.3 msec in FS and LTS cells, respectively (n = 19 FS; n = 16 LTS; p < 0.004; independent t test) (Fig. 2D). Higher extracellular stimulus intensities were required to evoke reliable suprathreshold eIPSCs in LTS interneurons, and responses were typically of lower amplitude than those in FS cells (note the different current calibration bars in Fig. 2B). These differences were not further analyzed. The data above indicate that GABAergic neurotransmission has distinctly different properties in these two interneuronal subtypes.

Different properties of spontaneous GABAergic transmission in FS and LTS interneurons

The above differences in the kinetics of evoked IPSCs in physiologically identified FS and LTS interneurons were also present when sIPSCs were recorded and analyzed. In our experimental conditions, sIPSCs represent a robust estimate of quantal release at individual GABAergic synapses. Indeed, we found no difference between sIPSCs and miniature IPSCs (mIPSCs) (the latter was recorded in the presence of 1 μm TTX) in terms of frequency, amplitude, and half-width in LTS interneurons (data not shown; n = 4; p > 0.1 in all cases). TTX application did not change amplitudes or half-width of sIPSCs in FS interneurons (data not shown; n = 5; p > 0.06). However, we found a significant, although modest (<25%), decrease of IPSC frequency in the presence of TTX in FS cells (data not shown; n = 5; p < 0.05). In LTS cells, the lack of a decrease in IPSC frequency and amplitude in the presence of TTX indicates that sIPSCs are essentially all mIPSCs and thus reflect release from individual sites. In FS cells, the lack of effect of TTX in IPSC half-width and amplitude indicates that the vast majority of spontaneous events recorded in the absence of TTX also reflects IPSC decay at individual synaptic boutons. Figure 3, A and B, shows examples of sIPSCs recorded in both cell types. Gabazine-sensitive (Fig. 3C,D) sIPSCs were strikingly more frequent, had a much larger amplitude, and decayed much more quickly in FS versus LTS cells (Figs. 3A,B,E, 4; Table 2). Although the properties of sIPSCs in FS interneurons were relatively homogenous, those in LTS cells showed a more pronounced variability, especially in decay time constant (coefficient of variation of τ_d,w, 0.21 and 0.53 in FS and LTS cells, respectively), suggesting a possible heterogeneity among LTS cells. Overall, kinetics of IPSCs in LTS interneurons was clearly slower than in FS cells (e.g., the fastest IPSCs in the population of LTS interneurons were longer in duration than the slowest IPSC in FS cells) (τ_d,w range, 5.3-10.6 in FS and 14-53.5 in LTS interneurons). To rule out the possibility that these differences in IPSC kinetics might derive from differences in synaptic location between the two interneuron subtypes, we analyzed sIPSC rise times, which are a sensitive measure of electrotonic distance (Hajos and Mody, 1997). In subsets of FS and LTS interneurons, we recorded sIPSCs under conditions that allow an accurate measure of rise-times [i.e., with less low-pass filtering (10 KHz vs normal 5 KHz) and sampling at 20 KHz]. Rise times were not different between FS and LTS interneurons (Fig. 4C; Table 2). These data indicate that the major differences in sIPSC waveform and frequency cannot be attributed to a different degree of electrotonic filtering of the synaptic response in LTS compared with FS cells.

The GABA_A receptor β-subunit selective agonist loreclezole differentially affects sIPSC kinetics in FS and LTS interneurons

Differences in IPSC kinetics might derive from differences in GABA_A receptor subunit composition (Macdonald and Olsen, 1994; Cherubini and Conti, 2001). We therefore tested the action of the GABA_A receptor β-subunit selective agonist loreclezole (10 μm) on sIPSCs recorded from FS and LTS interneurons. At this concentration, loreclezole selectively augments responses in GABA_A receptors that contain β2 or β3 subunits (Wafford et al., 1994; Wingrove et al., 1994; Korpi et al., 2002). Perfusion with loreclezole resulted in a robust modulation of sIPSCs recorded in FS interneurons (Fig. 5A-C, E). In particular, it produced a several-fold increase of both sIPSC decay time constant and total charge with no significant effect on sIPSC amplitudes (Fig. 5D; Table 3). Loreclezole perfusion increased sIPSC frequency (Fig. 5D; Table 3). Our local perfusion system (see Materials and Methods) allowed for relatively rapid and reversible drug effects (Fig. 5E).

Loreclezole had a less prominent effect on LTS than FS neurons (Fig. 6A-C; Table 3). It increased LTS sIPSC decay time constant and charge but did not affect sIPSC amplitudes or frequencies (Fig. 6D; Table 3). Although loreclezole had a significant effect on both FS and LTS IPSC waveforms, the degree of the modulation was different in the two cell types. Indeed, loreclezole induced a change of sIPSC charge transfer of 315 ± 47 versus 71 ± 21% above control in FS cells and LTS interneurons, respectively (p < 0.002; independent t test; see below), suggesting that the latter cell type has a lower expression of β2-3 synaptic GABA_A receptor subunits.

Differential effects of zolpidem on sIPSCs in FS and LTS interneurons

To test whether LTS and FS cells also differed in their GABA_A receptor α-subunit composition, we recorded sIPSCs and applied 200 nm zolpidem. At this concentration, zolpidem is known to specifically bind the benzodiazepine site of α1-containing and, less efficiently, α2- and α3-containing GABA_A receptors (Korpi et al., 2002; Mohler et al., 2002). In FS interneurons, zolpidem generated a consistent increase in sIPSC decay time constant (Fig. 7A-C; Table 3). Moreover, the drug significantly increased sIPSC amplitudes and charge, but it had no effect on sIPSC frequency (Fig. 7D; Table 3). As with loreclezole (Fig. 5E), the effects of zolpidem were rapid and at least partially reversible (data not shown).

As in FS cells, zolpidem did not alter the frequency of sIPSCs in LTS interneurons (Fig. 8A,D; Table 3). However, in contrast to its actions on FS interneurons, it did not significantly affect sIPSC decay time constants, charge, or amplitudes in LTS interneurons (Fig. 8B-D; Table 3). Although the overall effect in LTS cell population was not statistically significant, zolpidem did increase IPSC decay in some cells, suggesting heterogeneous expression of α1-3 containing GABA_A receptors among LTS interneurons. Overall, the effect of zolpidem on FS sIPSC charge was much greater in FS than LTS cells (117 ± 23 vs 30 ± 22% above control predrug period for FS vs LTS; n = 7 and 6, respectively; p < 0.01; independent t test) (Fig. 8E).

Zolpidem requires GABA_A receptors containing α1 subunits to exert its effects on FS interneurons

In layer V of the neocortex, α1 and α3 subunits of the GABA_A receptor are highly expressed, in contrast to the α2 subunit that is almost undetectable but is expressed in more superficial layers (Wisden et al., 1992; Pirker et al., 2000), suggesting that the zolpidem-mediated effect on FS cell IPSCs could be at GABA_A receptors containing either the α1 or α3 subunit. To distinguish between these possibilities, we tested the effect of zolpidem on sIPSCs in FS cells in control mice and in mice whose GABA_A receptor α1 subunit had been mutated at the benzodiazepine-binding site. The mutation is a substitution of a histidine residue into an arginine [α1(H101R), hereafter referred to as α1 knock-in], which makes the benzodiazepine site of the α1 subunit insensitive to most benzodiazepines and imidazolines, including zolpidem (Rudolph et al., 1999). These mice represent a powerful tool for determining whether zolpidem effects on FS interneuron sIPSCs depend on GABA_A receptors containing α1 or α3 subunits. The dramatic differences in sIPSC kinetics of FS and LTS interneurons in rats described above were also present in either wild-type or α1 knock-in mice (data not shown). In control (wild-type) mice, zolpidem generated a consistent large effect on IPSC half-width, τ_d,w, and charge (Fig. 9C,E; Table 4). Zolpidem did not affect sIPSC amplitude or frequency in either wild-type or knock-in mice (Fig. 9F; Table 4). In contrast, in FS interneurons of α1 knock-in mice, zolpidem application resulted in a modest increase on both half-width, τ_d,w, and charge of sIPSCs (Fig. 9E; Table 4) but had no effect on sIPSC amplitude or frequency (Fig. 9F; Table 4). Although the zolpidem effect was significant in both α1 knock-in and wild-type mice, it was much smaller in FS cells from knock-in animals (sIPSC charge was 164 ± 12 vs 31 ± 12% above control pre-drug period for wild-type vs α1 knock-in mice; n = 4 and 9, respectively; p < 0.0001; independent t test), suggesting that either zolpidem is highly selective for α1 subunits at these concentrations, or that the expression of α3 subunit in FS cells is modest.

Major differences in GABAergic synaptic transmission onto two general classes of cortical interneurons.

We recorded inhibitory synaptic responses from two general classes of layer V neocortical inhibitory cells: fast-spiking and low-threshold spiking interneurons. Although many laboratories use different terminology to describe a variety of non-FS cells (Cauli et al., 1997; Kawaguchi and Kubota, 1997, 1998; Wang et al., 2002), layer V cortical nonpyramidal cells were clearly separated into two distinct groups in this study, FS and LTS interneurons (Fig. 1, Table 1), with membrane and firing properties similar to those described in FS and LTS interneurons in cortical layer IV (Beierlein et al., 2003).

We found that IPSCs are profoundly different in FS and LTS cells in terms of frequency, amplitude, and duration. The frequency of sIPSCs is much higher in FS than LTS cells. Given that interneurons are known to synapse onto cells of the same subclass (Tamas et al., 1998; Gibson et al., 1999; Beierlein et al., 2000), the differences in sIPSC frequency would be expected if FS cells were predominately innervated by other FS cells because they have a higher release probability than LTS interneurons (Gupta et al., 2000; Xiang et al., 2002; Beierlein et al., 2003). The higher IPSC frequency in FS cells might also be attributable to a larger number of GABAergic synaptic contacts on FS cells.

In addition to these sIPSC frequency differences, both sIPSCs and eIPSCs had much faster decay kinetics and larger amplitudes in FS than LTS cells. IPSC decay kinetics was slow and variable among LTS cells, whereas in FS interneurons the kinetics was fast and homogeneous. Several mechanisms may be responsible for the large difference in IPSC kinetic properties in the two interneuron subtypes, including differences in synaptic location and thus different electrotonic distance in the two neuronal populations, lifetime of neurotransmitter in the synaptic cleft (Thompson and Gahwiler, 1992) post-translational modification of GABA_A receptors such as endogenous phosphorylation (Moss and Smart, 1996), and GABA_A receptor subunit composition.

In the hippocampus, it has been reported that different interneuronal classes show distinct IPSC decays and rise times (Hajos and Mody, 1997). Various interneuron classes generate IPSCs with different rise times in pyramidal neurons of both the hippocampus (Banks et al., 1998; Maccaferri et al., 2000) and neocortex (Xiang et al., 2002), indicating differential location of synaptic contacts onto pyramidal cells. Our experiments suggest that the distinct IPSC waveforms in neocortical FS versus LTS interneurons cannot simply be attributed to different synaptic locations, because rise times were fast and comparable in both interneuron subtypes. These cell-specific differences in IPSC waveform and frequency have also been observed in adult rats (>30 d old; our unpublished observations), ruling out the possibility that this dichotomy in IPSC properties could arise from different maturational stages in the two cell types.

The difference in the IPSC decay time constant between FS and LTS interneurons is unlikely to be entirely mediated by a differential phosphorilation of GABA_A receptors. Indeed, GABA_A receptor phosphorylation is known to produce a change in IPSC decay (Jones and Westbrook, 1997; Nusser et al., 1999; Poisbeau et al., 1999; Brandon et al., 2000), but the reported level of IPSC modulation (25-60%) is unlikely to account for the magnitude of the difference of IPSCs waveforms between FS and LTS cells (approximately threefold difference in τ_d,w between FS and LTS interneurons). Similarly, a different lifetime of GABA in the synaptic cleft in FS and LTS cells could, in principle, explain different IPSC kinetics, but it is unlikely that it might generate the large differences between FS and LTS IPSCs (Hill et al., 1998; Nusser et al., 2001; Overstreet and Westbrook, 2003). Accordingly, it has been shown that mIPSCs (which, in our experimental condition, make up the largest part of sIPSCs) are unaffected by the blockade of GABA transporters (Overstreet and Westbrook, 2003). For this latter reason, and because rise times are similar in FS and LTS cells, differential neurotransmitter spillover is also unlikely.

Do differences in IPSCs on FS and LTS interneurons depend on different GABA_A receptor subunit composition?

Fast inhibitory synaptic transmission between interneurons occurs through GABA_A receptors that are pentameric chloride-permeable channels composed of combinations of multiple subunit subtypes (Vicini, 1999; Cherubini and Conti, 2001; Rudolph et al., 2001). In mammals, GABA_A receptor subunits can be divided into several structurally related subfamilies (α1-6, β1-3, γ1-3, δ, ϵ, θ, and ρ1-3) (Vicini, 1999; Cherubini and Conti, 2001; Rudolph et al., 2001). The most common stoichiometry of GABA_A receptors is 2α2βγ (Farrar et al., 1999; Cherubini and Conti, 2001), and different subunit combinations result in profound heterogeneity in the electrophysiological properties of the receptor (Macdonald and Olsen, 1994). Thus, different GABA_A receptor subunit assemblies will likely generate different IPSC kinetics. Indeed, GABA_A-mediated synaptic events change their kinetics during development (Hollrigel and Soltesz, 1997; Dunning et al., 1999; Huntsman and Huguenard, 2000; Okada et al., 2000; Vicini et al., 2001; Banks et al., 2002), and this is correlated with a rearrangement of GABA_A receptor subunits (Dunning et al., 1999; Okada et al., 2000; Vicini et al., 2001). Here, we report that drugs that selectively activate different GABA_A receptor α and β subunit subtypes differentially affect sIPSCs recorded in FS and LTS interneurons. Loreclezole, which selectively enhances the β2-3-containing GABA_A receptors and is less potent for those containing β1 (Wafford et al., 1994; Wingrove et al., 1994; Korpi et al., 2002), dramatically increased sIPSC decay in FS interneurons with a much smaller, although significant, effect on LTS sIPSCs. These data indicate that LTS cells may have a lower expression of functional β2-3-containing GABA_A receptors, suggesting that β1 expression may be an important factor in shaping IPSC decays in LTS interneurons. Indeed, the β1 subunit of the GABA_A receptor is highly expressed in layer V of the rat neocortex (Pirker et al., 2000).

FS and LTS interneurons also showed a differential and more pronounced difference in α-subunits. We used the benzodiazepine-like drug zolpidem, which at the concentration used here (200 nm) enhances IPSCs generated primarily by α1-containing and, less potently, α2- and α3-containing GABA_A receptors (Mohler et al., 2002). Zolpidem significantly increased IPSC decays in FS cells but did not have an effect on IPSCs recorded in LTS cells, indicating that, overall, LTS cells have very low expression of α1-3-containing GABA_A receptors. However, a trend toward prolongation of IPSC decays was present in some neurons, suggesting that LTS cells are heterogeneous in their expression of α subunits. The differential effect by zolpidem in FS and LTS cells is in agreement with previously reported results of different zolpidem sensitivities in morphologically identified interneurons in the hippocampus (Thomson et al., 2000).

Both α1 and α3 subunits are highly expressed in layer V of the neocortex, whereas α2 is almost absent (Pirker et al., 2000). Because both α1 and α3 are zolpidem-sensitive (Mohler et al., 2002), we investigated the specific roles of α1 subunits in shaping FS IPSCs using α1(H101R) knock-in mice, in which the α1 subunit of the GABA_A receptors was made insensitive to agonists acting at the benzodiazepine site (Rudolph et al., 1999). The zolpidem effect on FS cell IPSCs was very much reduced in the knock-in compared with wild-type mice. The small residual effect seen in the mutant mice perhaps reflects the expression of the α3 subunit and strongly suggests that the α1 subtype is the major α subunit expressed in FS interneurons. This is in good agreement with previous reports indicating that the α1 subunit plays a key role in rapid activation and deactivation of GABA_A receptors (Lavoie et al., 1997) and is responsible for accelerating the time course of miniature IPSCs during development in the hippocampus (Goldstein et al., 2002) and cerebellum (Vicini et al., 2001).

Which α subunit is present in LTS cells? According to immunocytochemical studies, the α4 subunit is poorly expressed in neocortical layer V, whereas α5 is relatively highly expressed (Pirker et al., 2000). Interestingly, α5-containing GABA_A receptors have a very low affinity for zolpidem (Mohler et al., 2002), suggesting that LTS cells may express this subunit subtype. Note that the finding of when zolpidem concentration was increased to nonspecific levels (1-2 μm), it strongly prolonged IPSCs in both cell types (our unpublished observation).

Interestingly, loreclezole, but not zolpidem, produced an increase of sIPSC frequency in FS interneurons. This frequency change might be because of a loreclezole-dependent increase in postsynaptic GABA_A receptor sensitivity leading to the unmasking of small events that would have been buried in the noise. If this were the case, why were frequencies not increased by zolpidem? The larger increase of IPSC charge by loreclezole compared with zolpidem (∼4 × vs ∼2 ×) (Fig. 8E) would be more effective in disclosing the activity of remote synapses, given the efficient recovery of synaptic charge (but not amplitude) by somatic voltage clamp (Hausser and Roth, 1997). Consistent with the unmasking of remote events, we found that mean sIPSC rise times were increased more with loreclezole than zolpidem (our unpublished observation). However, we cannot exclude a possible GABAergic presynaptic effect (Zhang and Jackson, 1995).

Conclusion

Experimental and theoretical evidence indicates that GABAergic interneurons in the hippocampus and neocortex form synaptically connected networks that are crucial for generation of several kinds of oscillations associated with specific behaviors (McBain and Fishan, 2001). Rapidly decaying IPSCs between hippocampal interneurons are crucial for network oscillation coherence and frequency control (Bartos et al., 2001, 2002). Different GABAergic neurotransmission kinetics in distinct interneuronal networks might be responsible for entraining cortical oscillations at different frequencies.

Finally, recent data indicate that specific GABA_A receptors may result in distinct functions in terms of behavioral and clinical perspectives (Rudolph et al., 2001; Mohler et al., 2002). The different pharmacological sensitivities of the two neocortical interneuronal subclasses provide possible avenues for selective pharmacological manipulation of cortical inhibitory circuits with important functional effects relevant to both normal and pathophysiological activities.