Dendritic release of glutamate suppresses synaptic inhibition of pyramidal neurons in rat neocortex

Abstract

1. Dual whole-cell recordings were made in layer 2/3 of the rat neocortex in synaptically connected pyramidal cells and fast-spiking non-accommodating (FSN) interneurons. In 75 % of cell pairs (n = 80), the cells formed reciprocal synaptic connections.

2. Trains of backpropagating action potentials in pyramidal cells induced Ca²⁺ transients in dendrites followed by inhibition of unitary IPSPs. IPSP depression was prevented by loading pyramidal cells with 5 mm BAPTA or EGTA.

3. IPSP depression was mimicked by the metabotropic glutamate receptor (mGluR) agonist ACPD and was prevented by a mixture of the mGluR antagonists CPCCOEt and EGLU.

4. IPSP depression was prevented by loading pyramidal cells with the antagonists of vesicular exocytosis botulinum toxin D (light chain) and GDP-β-S.

5. It is concluded that Ca²⁺-dependent release of a retrograde messenger, most probably glutamate, from pyramidal cell dendrites suppresses the inhibition of pyramidal neurons via activation of mGluRs located in FSN interneuron nerve terminals.

The activity of neocortical pyramidal neurons is controlled by an amazing variety of inhibitory interneurons (Kawaguchi, 1995; Cauli et al. 1997; Gupta et al. 2000). Pyramidal cells and interneurons frequently form reciprocal synaptic connections creating elementary microcircuits. One question that arises is whether a local feedback mechanism regulating synaptic efficacy exists in these connections. Synaptic feedback provided by a retrograde messenger was suggested for inhibitory connections in cerebellar Purkinje cells (Llano et al. 1991) and CA1 pyramidal cells (Pitler & Alger, 1992). In both cell types, glutamate was proposed as a candidate for the retrograde messenger, since depolarization-induced suppression of inhibition (DSI) was mediated through activation of presynaptic mGluRs (Glitsch et al. 1996; Morishita et al. 1998).

Recently we showed (Zilberter et al. 1999) that in excitatory synapses between pyramidal cells and bitufted interneurons (Reyes et al. 1998) in layer 2/3 of the neocortex, backpropagating action potentials (APs) in interneuron dendrites evoked a transient Ca²⁺ influx leading to the release of GABA. GABA, as a retrograde messenger, activated presynaptic GABA_B receptors resulting in inhibition of glutamate exocytosis from pyramidal cell axon terminals. The results of the present study show that regulation of synaptic efficacy by the release of a retrograde messenger from subsynaptic dendrites also exists in inhibitory connections between pyramidal cells and another type of interneuron (FSN neuron) and thus may represent a general property of synapses between pyramidal cells and interneurons in the neocortex.

METHODS

Brain slices (300 μm thick) were prepared from the somatomotor cortex of 14- to 16-day-old Wistar rats as described previously (Markram et al. 1997). Rats were anaesthetized with halothane and decapitated in accordance with national guidelines. Simultaneous dual whole-cell voltage and current recordings were made in pyramidal cells synaptically connected to FSN interneurons. FSN neurons and pyramidal cells in layer 2/3 were identified by infrared differential interference contrast (IR-DIC) video microscopy and subsequent measuring of the neuron firing properties (see Fig. 1C). The extracellular solution contained (mm): 125 NaCl, 2.5 KCl, 25 glucose, 25 NaCO₃, 1.25 Na₂PO₄, 2 CaCl₂ and 1 MgCl₂. All experiments were performed at 32°C in oxygenated extracellular solution. The pipette solution contained (mm): 100 potassium gluconate, 20 KCl, 4 Mg-ATP, 10 sodium phosphocreatine, 0.3 GTP and 10 Hepes (pH 7.3, 310 mosmol l⁻¹).

Figure 1. Reciprocally connected pyramidal and FSN neurons.

A, IR-DIC image of a pyramidal cell (left) and a FSN interneuron during dual whole-cell recordings. Scale bar, 65 μm. The scheme on the right illustrates the reciprocal connection. B, camera lucida reconstruction of reciprocally connected pyramidal and FSN cells. Somata and dendrites are blue and red, axons are black and green for the FSN and pyramidal cells, respectively. Synaptic contacts are marked by asterisks of the same colour as the corresponding axons. C, firing patterns of the pyramidal (Pyr) and FSN cells recorded during current injection in a soma. The resting potentials of the cells are indicated below the traces. D, examples of IPSPs and EPSPs recorded in reciprocally connected pyramidal and FSN cells during paired-pulse stimulation (100 ms interpulse interval). E, current-voltage relationship of the inhibitory transmission between FSN and pyramidal cells (n = 3).

Data are given as means ± s.d. Statistical significance was analysed by Student’s paired t test.

Five neuron pairs were morphologically reconstructed with the aid of a computerized camera lucida system. Neurons were filled during experiments with 2 mg ml⁻¹ neurobiotin.

The conditioning protocol for raising the dendritic Ca²⁺ concentration was the same as that described previously (Zilberter et al. 1999). Briefly, a train of 10 APs (unless otherwise noted) at 50 Hz in the pyramidal neuron caused a transient increase in the dendritic Ca²⁺ concentration (Helmchen et al. 1996). One or two successive IPSPs in the pyramidal neuron were evoked by stimulating the FSN interneuron 250 ms after the postsynaptic burst of APs. This pattern of sequential pre- and postsynaptic stimulation was repeated every 7 s.

The paired-pulse ratio (PPR) was calculated as IPSP2/IPSP1, where IPSP1 and IPSP2 were mean IPSP amplitudes in response to the first and second FSN cell APs, respectively. The mean amplitude of unitary IPSPs was measured from 50–100 sweeps.

In Ca²⁺-imaging experiments, neurons were filled with fura-2 via pipettes containing 250 μm of the dye added to the pipette solution. A monochromatic light source was used for fluorescence excitation (T.I.L.L. Phototonics, Planegg, Germany). A back-illuminated frame transfer CCD camera (Princeton Instruments, Trenton, USA) was used to acquire 356 nm/380 nm fluorescence ratio images from up to eight regions of interests simultaneously at a frequency of 100 Hz.

RESULTS

Unitary synaptic connections between FSN interneurons and pyramidal cells in layer 2/3 of the rat neocortex were studied using dual patch-clamp recordings. Figure 1A shows an image of a synaptically connected interneuron and pyramidal cell made under IR-DIC microscopy. The FSN and pyramidal neurons frequently formed reciprocal connections (in 75 % of synaptically connected cell pairs measured, n = 80). Figure 1B shows a camera lucida reconstruction of the reciprocally connected FSN and pyramidal neurons. The FSN neurons were non-pyramidal cells of remarkably similar morphological type resembling ‘neurons with axons forming arcades’ (Peters & Saint Marie, 1984) or type 2/3 neurons (Jones, 1975). However, from a combination of the morphological and electrophysiological properties, descriptions of analogous cells could not be found elsewhere. FSN neurons had oval cell bodies. Sparsely spinous dendrites radiated from the soma (over 800–1000 μm), usually being more numerous towards layer IV. The axon originated from the soma or dendrite, and extended above the cell body where it formed a rich plexus coextensive with basal dendrites and/or proximal branches of the pyramidal cell apical dendrites. Some branches arched downwards and typically one or more long branches descended parallel to the pyramidal cell axon, branching, as did the latter, in layer V. These axons formed between two and eight presumptive synaptic contacts with basal dendrites or branches of the apical dendrites of pyramidal neurons.

Interneurons had a resting potential of −70.1 ± 2.4 mV (mean ± s.d., n = 23) and unlike pyramidal cells displayed a high frequency non-accommodating firing pattern (Fig. 1C). Both unitary EPSPs and IPSPs recorded in the FSN and pyramidal neurons, respectively, showed prominent paired-pulse depression (PPD) at a 100 ms interval (10 Hz) between APs (Fig. 1D). Inhibitory connections were characterized by a high transmitter release probability resulting, in most connections, in the absence of synaptic failures in response to a single AP.

Current-voltage relationships of inhibitory connections showed a tendency for outward rectification (Fig. 1E). The IPSC reversal potential was about −40 mV, close to the chloride equilibrium potential (-48 mV), indicating a GABAergic origin of the inhibitory connections. Moreover, synaptic transmission was completely blocked by a specific GABA_A receptor antagonist, bicuculline (50 μm; not shown).

A conditioning train of backpropagating APs (10 APs at 50 Hz) in pyramidal cell dendrites resulted in a significant decrease in IPSP amplitude measured 250 ms after the train (Fig. 2A). The mean IPSP amplitudes are indicated by horizontal lines and are also shown in the top panel of Fig. 2A. IPSP depression could be reproduced in the same experiment, though to a smaller extent in this case, by repeated conditioning. IPSPs during application of the first conditioning protocol were always used for data analysis. Figure 2B shows mean IPSPs measured in 16 cell pairs in control, during conditioning and after conditioning. Synaptic depression during conditioning was observed in all these cell pairs. In each experiment, IPSPs were normalized to the mean IPSP amplitude in control. Then, IPSPs in 16 cell pairs recorded at equivalent times during the experimental protocol were averaged (open circles). Filled circles indicate the mean IPSPs for each minute of the experiments. These data suggest that synaptic depression develops during the first minute of conditioning (probably after the first conditioning train application), and that synaptic transmission recovers quickly after conditioning termination. Furthermore, a clear IPSP potentiation was observed right after the termination of conditioning, indicating that the potentiation process developed in parallel with the synaptic depression. In 16 cell pairs, the mean IPSP decreased during conditioning to 68 ± 18 % of control (P < 0.007). In eight cell pairs in these experiments, the long-lasting potentiation of IPSPs (to 152 ± 42 % of control, P < 0.002) following the conditioning protocol was observed, and could be recorded until termination of the experiment (in some cases, for more than 60 min after conditioning).

Figure 2. IPSP depression induced by a conditioning train of 10 backpropagating APs.

A, IPSP amplitudes during one experiment. The mean IPSP amplitudes during the indicated times are shown by horizontal lines. ○, amplitude of IPSPs in the absence of conditioning; •, amplitude of IPSPs during the conditioning protocol. These IPSPs are also shown in the top panel. B, IPSPs in control, IPSP depression during conditioning and IPSP potentiation after the termination of conditioning measured in 16 cell pairs. IPSPs were normalized in each experiment to the mean IPSP amplitude in control. Then, IPSPs in 16 cell pairs recorded at equivalent times during the experimental protocol were averaged (○). •, mean IPSPs within each minute. C, repeated application of 100 μm ACPD during the experiment. The mean IPSPs in the top panel were normalized to the amplitude of first IPSP during the paired-pulse stimulation. •, IPSP amplitude in the presence of ACPD.

To test the contribution of the presynaptic site to IPSP depression, in 11 of these experiments the PPR was measured, and was found to be 72 ± 12 % in control, 87 ± 15 % during conditioning (P < 0.006) and 70 ± 7 % after the termination of conditioning. This suggests that the decrease in the probability of GABA release in FSN neuron nerve terminals underlies synaptic depression (Wilcox & Dichter, 1994; Jensen et al. 1999). On the other hand, in eight of these experiments where potentiation of IPSPs developed, the PPR did not change significantly (76 ± 10 and 71 ± 7 % in control and during IPSP potentiation, respectively, P < 0.2). However, the potentiation of inhibitory transmission was not studied in detail in the present work.

In an additional series of experiments, IPSCs were measured under voltage-clamp conditions. The conditioning train of APs was replaced by one depolarizing 500 ms pulse from the cell resting potential to 0 mV. In six cell pairs, the mean IPSC was reduced to 70 ± 8 % of control during conditioning (P < 0.01).

The possibility that the IPSP depression may be induced by dendritic release of glutamate, leading to activation of mGluRs in the GABAergic axon terminals of FSN neurons was then explored. ACPD (100 μm), an agonist of group I and II mGluRs, evoked strong and reversible block of IPSPs (Fig. 2C). In seven cell pairs, the mean IPSP amplitude was reduced to 28.7 ± 8.2 % of control (P < 0.005). Meanwhile, ACPD induced a significant decrease in PPD. The mean IPSPs recorded during control and ACPD application and scaled to the amplitude of the first IPSP in a paired-pulse protocol are shown in the top panel of Fig. 2C. ACPD converted IPSP depression to facilitation. The PPR was 73 ± 7 % in control, 110 ± 19 % during ACPD application (P < 0.005), and 74 ± 13 % after washout (n = 7). Moreover, synaptic failures appeared under the action of ACPD (see Fig. 2C). These results imply that ACPD activates presynaptically localized mGluRs followed by a reduction in GABA release probability.

Application of 7 μm DCG-IV, a specific agonist of group II mGluRs, also inhibited IPSPs, though to a lesser extent than ACPD (to 47.6 ± 6.3 % of control, P < 0.002, n = 5). The PPR was 64 ± 8 % in control, 80 ± 22 % during DCG-IV application, and 67 ± 9 % after washout. ACPD applied in the presence of 100 μm EGLU, a specific antagonist of group II mGluRs, still inhibited IPSPs (to 29 ± 16 %, n = 3). This suggests that both group I and group II mGluRs may be present in the FSN cell axon terminals.

To test whether activation of mGluRs is involved in the IPSP depression, a mixture of specific antagonists of group I and II mGluRs, 50 μm CPCCOEt and 100 μm EGLU, respectively, was used. In five experiments, application of these antagonists completely prevented the IPSP depression: the mean IPSP amplitude was 1.73 ± 1.6 mV in control and 1.86 ± 1.6 mV during conditioning (P < 0.3; see also Fig. 3C). This implied that activation of mGluRs did underlie synaptic depression and also suggested the release of a retrograde messenger, most probably glutamate, from the pyramidal cell dendrites.

Figure 3. Ca²⁺ dependence and antagonists of IPSP depression.

A, dependence of IPSP depression on dendritic Ca²⁺ influx. The integrals of dendritic Ca²⁺ transients during the stimulation protocol are plotted on the horizontal axis. Ca²⁺ transients were induced by 4 AP trains, 10 AP trains, and under voltage-clamp (VC) conditions by 50 ms depolarizing pulses from the cell resting potential to 0 mV. The spatial distribution of Ca²⁺ transients in pyramidal cell dendrites was highly inhomogeneous: measured in 37 regions of 8 pyramidal cells, the amplitude of Ca²⁺ transients evoked by a single AP varied from 14 to 323 nm. This explains the large deviations of Ca²⁺ signal integrals shown in the figure. Examples of Ca²⁺ transients in pyramidal cell dendrites measured in the same region and induced by different stimulation patterns are shown on the right. B, mean IPSPs in 9 cell pairs in control and during conditioning in the presence of 5 mm BAPTA (n = 4) or EGTA (n = 5). Here and in C-E, IPSPs were normalized as described in the legend to Fig. 2B. C, mean IPSPs in 5 cell pairs in control and during conditioning in the presence of 50 μm CPCCOEt and 100 μm EGLU. D, mean IPSPs in 9 cell pairs in control and during conditioning in the presence of 400 nm BoTx-D. E, mean IPSPs in 3 cell pairs in control and during conditioning in the presence of 0.6 mm GDP-β-S. F, effect of blocking Na⁺ channels with 2.5 mm QX-314 loaded into the pyramidal cell in reciprocally connected pyramidal and FSN cells. The left panel shows IPSC amplitudes in control (○) and during conditioning (•) in one experiment. In the right panel, upper traces represent currents measured in the pyramidal cell during application of a depolarizing pulse from the resting potential (-83.4 mV) to 0 mV (note a large inward Na⁺ current in the absence of QX-314). Capacitative transient currents are truncated. Lower traces show the corresponding voltages measured in the FSN neuron: the EPSP in control disappeared in the presence of QX-314 (mean of 5 consecutive sweeps).

Indeed, synaptic depression was prevented by 5 mm BAPTA (n = 4) or EGTA (n = 5) loaded into the pyramidal cells (Fig. 3B). The mean IPSP amplitude, averaged from all these experiments, was 1.36 ± 0.65 mV in control and 1.55 ± 0.56 mV during conditioning (P < 0.2). Importantly, 5 mm EGTA inhibited glutamate release from pyramidal cell axon terminals by only 50 % on average (A. Rozov & N. Burnashev, personal communication) though it completely prevented IPSP depression. The dependence of IPSP depression on the rise in dendritic Ca²⁺ concentration is shown in Fig. 3A. The integrals of dendritic Ca²⁺ transients during the stimulation protocol are plotted on the horizontal axis. Ca²⁺ transients were induced by trains of four APs, 10 APs, and under voltage-clamp conditions by 500 ms depolarizing pulses from the cell resting potential to 0 mV. The data indicate that the process of IPSP depression was already saturated at the dendritic Ca²⁺ elevation induced by the 10 AP conditioning trains. Meanwhile, in 37 dendritic regions, the mean amplitude of the Ca²⁺ transients evoked by the 10 AP trains was 198 ± 252 nm, and for those evoked by depolarizing pulses under voltage-clamp conditions it was 234 ± 315 nm. This suggests a high sensitivity of the process underlying synaptic depression to the dendritic Ca²⁺ concentration.

The prominent dependence of synaptic depression on the dendritic Ca²⁺ concentration suggested, as in our previous work (Zilberter et al. 1999), that dendritic release of glutamate may proceed by vesicular exocytosis. To verify this assumption, a specific antagonist of vesicular exocytosis, botulinum toxin D light chain (BoTx-D), was applied to the neurons. In nine experiments, 400 nm BoTx-D prevented IPSP depression during conditioning (Fig. 3D). Most of these experiments were performed in reciprocally connected cells and, to ensure diffusion of BoTx-D to the dendritic sites of synaptic connections, the conditioning AP train was applied only when complete block of EPSPs in FSN neurons had been detected. This waiting period was between 20 and 50 min. The mean IPSP amplitudes were 5 ± 2.1 mV in control and 4.9 ± 2 mV during conditioning (P < 0.3). Meanwhile, the mean amplitude of the Ca²⁺ transients measured in 15 dendritic regions of three neurons increased to 182 ± 56 % of control after loading cells with BoTx-D.

Additional experiments (n = 3) were performed with 0.6 mm GDP-β-S loaded into pyramidal cells. GDP-β-S is a potent inhibitor of presynaptic transmitter release (Hess et al. 1993; Zilberter et al. 1999). In all these experiments, the cells were reciprocally connected. In two of them, GDP-β-S completely blocked EPSPs measured in FSN cells, and in one case EPSPs were reduced to 15 % of control. GDP-β-S prevented IPSP depression (Fig. 3E): the mean IPSP amplitudes were 1.5 ± 0.15 mV in control, and 1.73 ± 0.27 mV during conditioning.

Finally, the possibility that glutamate was released from the pyramidal cell axon collaterals innervating the FSN nerve in the vicinity of the active synaptic contacts under investigation had to be ruled out. To prevent active voltage propagation in the pyramidal cell axon, an antagonist of Na⁺ channels, QX-314 (2.5 mm), was loaded into pyramidal cells and IPSCs were measured under voltage-clamp conditions (n = 5). In three of these experiments, the neurons were reciprocally connected and EPSPs measured in FSN cells disappeared soon after loading of QX-314 into pyramidal cells, confirming the block of voltage propagation (Fig. 3F). Meanwhile, QX-314 did not affect IPSC depression during conditioning: the mean IPSC was reduced to 70 ± 9 % (P < 0.04) of control. Thus, glutamate was most probably released from the pyramidal cell dendrites.

DISCUSSION

The results presented here suggest that synaptic inhibition of pyramidal cells may be effectively adjusted to their output activity by the dendritic release of glutamate evoked by backpropagating APs. The properties of this release fit well with those of vesicular exocytosis (Augustine & Neher, 1992; Huang & Neher, 1996; Maletic-Savatic & Malinow, 1998). Interestingly, backpropagating APs in FSN neurons presumably evoke EPSP depression by dendritic GABA release (author’s unpublished results), similar to that described by Zilberter et al. (1999). In both pyramidal and FSN cells, metabotropic receptors located in the ‘alien’ axon terminals (GABA_B receptors in glutamatergic terminals of pyramidal cells and mGluRs in GABAergic terminals of interneurons) are activated by the retrograde messengers released from subsynaptic dendrites, resulting in a decrease of transmitter release from axon terminals.

The EPSP depression in pyramidal-to-bitufted cell connections described previously (Zilberter et al. 1999) and IPSP depression in pyramidal-to-FSN cell synapses are similar in some respects to DSI in cerebellar Purkinje cells (Llano et al. 1991) and hippocampal pyramidal cells (Pitler & Alger, 1992): all these processes are triggered by an increase in the dendritic Ca²⁺ concentration resulting in the release of a retrograde messenger. However, available results of studies on hippocampal pyramidal cells suggest a significant difference in the mechanism of dendritic release. It was reported that DSI was not affected by GDP-β-S loaded into pyramidal cells (Pitler & Alger, 1994) or by removal of GTP from the intracellular solution (Lenz & Alger, 1999). In the experiments in this study, GDP-β-S completely prevented dendritic transmitter release. GTP-binding proteins are necessary for several steps in vesicle docking and trafficking during the exocytotic cycle (for review, see Augustine et al. 1999). This explains why GDP-β-S is an effective inhibitor of vesicular exocytosis (Hess et al. 1993; Zilberter et al. 1999), indicating that this process might underlie the dendritic release of transmitters in the present experiments. However, here GDP-β-S was not considered to be a specific antagonist of vesicular exocytosis. Hence BoTx-D, which is a specific antagonist of vesicular exocytosis (Xu et al. 1998), was also used, and was shown to prevent dendritic release of a retrograde messenger.

A simplified scheme of the possible feedback interactions in a microcircuit consisting of a reciprocally connected pyramidal cell (P) and interneuron (I) is shown in Fig. 4. In the absence of retrograde signalling, simultaneous activity of the two cells creates a strong negative feedback for pyramidal cell excitation: release of glutamate from the pyramidal cell axon terminals increases the excitability of the interneuron, therefore enhancing pyramidal cell inhibition. Interestingly, in both cells, suppression of synaptic efficacy by retrograde transmitter release leads to the same result: the maintenance of excitability of the pyramidal cell. Depression of the pyramid-to-interneuron connections decreases interneuron excitation and, consequently, pyramidal cell inhibition. Depression of the interneuron-to-pyramid connections directly maintains pyramidal cell excitability. Thus, the balance between the impact of the axonal and dendritic parts of transmitter release on pyramidal cell excitation presumably modulates the temporal pattern of pyramidal cell output activity. There is the intriguing possibility that in abnormal conditions it may contribute to epileptic activity in the cerebral cortex.

Figure 4. Schematic representation of the reciprocally connected pyramidal cell and interneuron.

Abbreviations: APs, axonal action potentials; BAPs, dendritic backpropagating APs.