Research Article

Spike Transmission and Synchrony Detection in Networks of GABAergic Interneurons

See allHide authors and affiliations

Science  22 Jun 2001:
Vol. 292, Issue 5525, pp. 2295-2299
DOI: 10.1126/science.1061395


The temporal pattern and relative timing of action potentials among neocortical neurons may carry important information. However, how cortical circuits detect or generate coherent activity remains unclear. Using paired recordings in rat neocortical slices, we found that the firing of fast-spiking cells can reflect the spiking pattern of single-axon pyramidal inputs. Moreover, this property allowed groups of fast-spiking cells interconnected by electrical and γ-aminobutyric acid (GABA)–releasing (GABAergic) synapses to detect the relative timing of their excitatory inputs. These results indicate that networks of fast-spiking cells may play a role in the detection and promotion of synchronous activity within the neocortex.

In vivo studies indicate that coherent firing among cortical neurons may be related to sensory stimulation and behavioral states (1–4). However, how postsynaptic cells read out the pattern of activity of their presynaptic axons and how spike synchrony among input axons may be detected by the cortical network remain poorly understood (3, 5).

Studies in intact animals as well as theoretical work suggest that inhibitory interneurons may coordinate neuronal activity in cortical networks (6–11). Fast-spiking (FS) cells are a prominent subtype of GABA-releasing (GABAergic) interneurons (12), exerting powerful inhibitory control of both excitatory and inhibitory cortical cells (13–17). The properties of excitatory synapses at FS cells (13, 18), as well as their voltage-dependent conductances (19–21), suggest that these cells may be particularly sensitive to the timing of their inputs, as has been shown for hippocampal interneurons (22,23). Moreover, cortical networks of FS cells are interconnected by electrical synapses (14, 15,17, 24), and these electrical connections can promote synchronous spiking (14, 15). We investigated how the presynaptic pattern of firing can be transmitted through single-axon pyramidal-to-FS connections and how the degree of synchrony among excitatory inputs may be detected by groups of FS cells.

Precise timing of spike transmission. Pairs of pyramidal and FS cells were recorded in rat neocortical slices (25). Monosynaptic pyramidal–FS cell connections were detected by the generation of short-latency unitary excitatory postsynaptic potentials (EPSPs) in response to individual pyramidal spikes (mean latency, 0.63 ± 0.05 ms; range, 0.4 to 1.0 ms; n = 12 pairs) (26). To reproduce the ongoing synaptic activity that occurs in vivo (27), we injected into the postsynaptic FS cells fluctuating current waveforms that changed from trial to trial, resulting in an irregular firing of ∼5 to 50 Hz (Fig. 1A, middle panels) (28). The impact of a presynaptic spike on the postsynaptic firing (that is, spike transmission) was studied by producing a timed action potential in the pyramidal cell with every trial (Fig. 1A, upper panel). Spike transmission was examined by constructing a peristimulus time histogram (PSTH) of the postsynaptic cell. This revealed a sharp increase in the firing frequency of the FS cell after the presynaptic spike (Fig. 1A, lower panel, and Fig. 1B). We defined the temporal precision of spike transmission as the width (10 to 90 %) of the PSTH peak (29). The average temporal precision in seven pairs was 1.5 ± 0.3 ms (Fig. 1D), and the latency between the peaks of the presynaptic spike and of the PSTH was 1.7 ± 0.1 ms (n = 7 pairs). We also defined the spike transmission gain (STG) as the area of the PSTH peak. STG represents the average number of spikes added in the postsynaptic cell by a single presynaptic action potential. In seven pairs, the STG ranged from 0.046 to 0.711 (Fig. 1C). We found an approximately linear relationship between the STG and the amplitude of the unitary EPSP [r = 0.98 (Fig. 1C)]. Thus, under our experimental conditions, the STG per 1-mV EPSP was ∼0.04. High temporal precision was not limited to large synaptic inputs. Small inputs were unreliable, but the postsynaptic spikes they produced were precisely timed (Fig. 1D).

Figure 1

Properties of spike transmission at pyramidal-to-FS cell connections. (A) Irregular spike activity [postsynaptic membrane potential (V m)] was generated in a postsynaptic FS cell by injecting pulses of random signals (post I) (28). Two examples of these injections are shown in the middle panels. A single spike in the presynaptic pyramidal cell (top) produced a sharp peak in the FS cell PSTH (bottom) (n = 560 trials, bin width = 1 ms). (B) (Top) Superposition of 75 consecutive postsynaptic recordings at an expanded time scale. (Bottom) Temporal relationship between the peak of the PSTH and the presynaptic spike (superimposed dashed line, bin width = 250 μs ). The width of the PSTH was 0.9 ms (10 to 90 %), and the latency between both peaks was 1.4 ms. Data are from the same pair shown in (A). (C) Linear relationship between the amplitude of the single-axon EPSP and the STG (r = 0.98). (D) Lack of correlation (r = –0.09) between the EPSP amplitude (amp) and the temporal precision of spike transmission. Solid symbols in (C) and (D) represent data obtained from seven different pairs at basal presynaptic stimulation frequencies (0.33 to 1 Hz). Open symbols illustrate additional data obtained from the same pairs by means of paired-pulse protocols. Data from a given pair are plotted with symbols of the same shape.

We next examined the temporal relationship between the PSTH peak and the waveforms of the corresponding single-axon excitatory postsynaptic current (EPSC) and EPSP (Fig. 2, A and B) (30). The duration of the PSTH peak was much shorter than that of the unitary EPSP (Fig. 2B), as previously shown in spinal motoneurons (31). The PSTH peak occurred during the rise time or peak of the EPSP and, on average, preceded the peak of the EPSP by 0.37 ± 0.2 ms (n = 7 pairs). The high temporal precision of spike transmission in these connections could allow the firing of postsynaptic cells to reflect the temporal pattern of presynaptic spikes. We thus generated in the presynaptic cell two action potentials separated by a brief interval, and found that the postsynaptic PSTH showed two independent peaks [n = 3 pairs; interval, 4 to 7 ms (Fig. 2C)]. The second PSTH peak was smaller in amplitude, reflecting paired-pulse depression, and was in agreement with a linear relationship between EPSP amplitude and STG (Fig. 1C). Thus, spike transmission at pyramidal-to-FS cell connections can transfer high-frequency signals and preserve the pattern of the presynaptic inputs.

Figure 2

Pyramidal-to-FS cell connections can preserve the pattern of presynaptic inputs with millisecond precision. (A) Comparison of the time course of a single-axon EPSC (top) (holding potential = –70 mV, average of 50 traces), the corresponding single-axon EPSP recorded near resting conditions (middle) (V m = –79 mV, average of 100 traces), and the resulting PSTH (bottom) (n = 560 trials, bin width = 1 ms). These three traces have been scaled, superimposed, and expanded in (B), where the value of 0 on the time scale corresponds to the presynaptic spike peak. The polarity of the unitary EPSC (dotted line) has been reversed (binning of the PSTH is 250 μs). (C) Two presynaptic action potentials separated by an interval of 4 ms (top) produced two independent peaks in the postsynaptic PSTH (bottom) (n = 685 trials, bin width = 1 ms). Middle traces show corresponding single-axon EPSCs (holding potential = –70 mV, average of 50 traces) and EPSPs. Data were obtained from the same connection shown in (A).

EPSC kinetics and temporal precision. The rate of membrane depolarization may affect the efficacy of spike-generating mechanisms (32, 33), and this could contribute to the discrepancy between the shape of the EPSP and the PSTH peak. We thus simulated synaptic potentials (sim-EPSPs) of variable kinetics by injecting depolarizing current waveforms into the FS cell (34). Two sim-EPSPs of similar amplitudes and decay time courses but different rise times (10 to 90% = 1 ms versus 8 ms) produced different effects on the cell spiking activity [n = 4 cells (Fig. 3A)]. Because fast-decaying single-axon EPSCs are correlated with fast-rising EPSPs (35) (n = 6 pairs; r = 0.88), we next examined the relationship between the EPSC decay time constant and the temporal precision of spike transmission (Fig. 3B). The temporal precision was correlated with the decay time constants of both synaptic EPSCs and sim-EPSCs [r = 0.96 (Fig. 3B)] (36). These results indicate that the kinetics of the postsynaptic currents can affect the temporal precision of spike transmission.

Figure 3

The kinetics of the synaptic currents are important for synaptic integration of spike transmission. (A) PSTH (n = 1800 trials, bin width = 1 ms) obtained from an FS cell in response to the coinjection of synthetic current waveforms and two simulated EPSPs of different rise times (8 ms versus 1 ms) but similar decay time courses (upper trace). (B) Relationship between the decay time constant of single-axon EPSCs or sim-EPSCs and the width of the PSTH peak (10 to 90%, r = 0.96). The upper traces illustrate the different PSTH peaks obtained by two sim-EPSCs (dashed lines) with similar rise times but different decay time constants of 2 ms (left) and 5 ms (right). The corresponding sim-EPSPs are shown superimposed (continuous lines). Left vertical calibration bar: 13 pA, 0.35 mV, and 8.5 Hz. Right vertical calibration bar: 9 pA, 0.35 mV, and 6 Hz.

Depolarization induces EPSP acceleration and boosting. At excitatory synapses among pyramidal neurons, depolarization of the postsynaptic cell slows down the EPSP time course (23,37–39). FS cells, however, express voltage-dependent conductances distinct from those of principal neurons (19–21). Moreover, at hippocampal interneurons, the EPSPs time course is relatively insensitive to membrane depolarization (23). Thus, we compared the time course of pyramidal-to-FS–cell unitary EPSPs recorded at resting and near-threshold potentials (Fig. 4A) (40). The decay time course of FS single-axon EPSPs at depolarized potentials (–48 ± 1 mV) was faster than that of the synaptic inputs recorded at resting potentials [–73 ± 2 mV,n = 5 pairs (Fig. 4A)]. The weighted decay time constant of EPSPs at the resting potential was 9.7 ± 1.1 ms as compared to 3.0 ± 1.4 ms at depolarized membrane potential [n = 5 pairs, paired t test,P < 0.02 (Fig. 4B)] (41). Using sim-EPSPs, we also examined whether the shape of the events could affect their voltage-dependent modification. We compared fast- (∼1 ms) and slow-rising (∼5 to 10 ms) sim-EPSPs and found that fast-rising events, similar to single-axon EPSPs, were selectively modified at near-threshold potentials. Their amplitude was boosted by 54% (P < 0.001, n = 6 experiments) and their decay accelerated by 37% (P < 0.05,n = 6 experiments). Slow-rising events did not show significant changes [P > 0.05 (Fig. 4C)]. Tetrodotoxin (0.5 μM) prevented the boosting and reduced the acceleration of fast-rising events, indicating the involvement of voltage-dependent Na+ channels in these modifications [P < 0.01, n = 5 (42)]. The depolarization-induced acceleration of the EPSP time course may have a profound effect on temporal summation (Fig. 4D). At resting potential, a 200-Hz train of three sim-EPSPs exhibited summation, whereas at depolarized potential, the acceleration prevented the summation. These findings suggest that the temporal properties of spike transmission at pyramidal-to-FS cell connections reflect the fast kinetics of their postsynaptic receptors (43,44) as well as voltage-dependent conductances (19–21, 23, 45).

Figure 4

Pyramidal-to-FS cell EPSPs decay faster near threshold potentials. (A) Superimposition of the time course of a pyramidal-to-FS cell unitary EPSP recorded at restingV m (thin trace, –77 mV) and a depolarized V m (thick trace, –51 mV). The traces (average of 150 trials) are shown scaled in the right panel. (B) Faster decay time course of unitary EPSP at depolarized versus resting V m (n = 5 pairs,P < 0.02). (C) Superimposition of sim-EPSPs with fast (1.1 ms, left) and slow (6.4 ms, right) rise times recorded at resting (thin trace) and near-threshold potentials (thick trace) in a FS cell. Each trace is the average of 60 traces. (D) Three sim-EPSPs (10 to 90%, rise time = 1.1 ms, decay τ = 6.3 ms, 200 Hz) recorded from a FS cell at resting (V m = –70 mV) and depolarized (V m = –33 mV) potentials. Each trace is the average of 40 trials.

Synchrony detection in interneuron networks. The temporal precision of spike transmission that we describe here may allow groups of FS cells interconnected by electrical and GABAergic synapses (14, 15, 17) to be sensitive to synchronous excitatory inputs. When a pair of FS cells is coupled only electrically (Fig. 5A, bottom), a presynaptic spike generated a biphasic response consisting of a brief depolarization [reflecting the spike itself (open arrow)] and a slower hyperpolarizing component [reflecting the presynaptic AHP (solid arrow)] in the postsynaptic cell. If the cells are connected by GABAergic synapses, a hyperpolarizing inhibitory postsynaptic potential (IPSP) is produced (Fig. 5A, top) (46). When both types of connections coexist, nearly synchronous inputs into different FS cells will be aided by the mutual depolarizing phase of the electrical response, and synchronous spiking will be promoted (14, 15). On the other hand, nonsynchronous inputs will be relatively inefficient, because FS cells receiving late inputs would be influenced by the hyperpolarizing response, mediated by both GABAergic and electrical connections, from neurons receiving early inputs. We recorded from pairs of FS cells connected by both electrical and GABAergic synapses (Fig. 5B). We created sim-EPSPs by injecting EPSC-like waveforms in both cells, and depolarized the cells so that individual sim-EPSPs would generate a spike in 50 to 70 % of the trials. Next, we tested whether the relative timing of two independent inputs could affect the spiking of the FS cells. When the time interval between the sim-EPSPs at each cell was 0 to 1 ms, both FS cells fired with higher probability within a narrow time window (Fig. 5C). Under these conditions, the probability of firing increased by 33 ± 10% (n = 3 pairs, P < 0.05). In contrast, when the two inputs were separated by an interval of 5 ms, the firing of the cell receiving the delayed input was reduced by 59 ± 20% [P < 0.05 (Fig. 5D)].

Figure 5

Precise spike transmission of excitatory inputs allows pairs of FS cells connected by both GABAergic and electrical synapses to detect synchronous activity. (A) (Upper traces) Example of an isolated single-axon GABAergic IPSP (calibration bars: vertical, 50 mV and 1 mV; horizontal, 5 ms). (Lower traces) Postsynaptic response (post-V m = –37 mV) produced by a spike in the presynaptic neuron (pre-V m = –40 mV) in a pair of FS cells connected exclusively by electrical synapses (DC coupling coefficient, 10.1 %; calibration bar, 30 and 0.5 mV). (B) Recordings from a pair of FS cells connected by both electrical and GABAergic synapses (left, schematic drawing). A presynaptic spike produces a biphasic response in the postsynaptic cell (pre-V m = −38 mV, post-V m = −35 mV). (C) Two sim-EPSPs separated by 1 ms (schematically represented in the top traces) produced the synchronous firing of both FS cells (PSTH bin = 0.5 ms). (D) If two sim-EPSPs are separated by 5 ms, the firing of the postsynaptic cell is dramatically reduced. Δt represents the interval between sim-EPSPs.

FS cells receive strong thalamocortical inputs (15) as well as corticocortical and local excitatory inputs (13, 18,47–49) and form networks with other FS cells interconnected by both GABAergic and electrical synapses (14, 15,17, 24). Here we provide evidence that the temporal characteristics of spike transmission and the properties of electrical and GABAergic synapses could endow networks of FS cells with sensitivity to the synchrony of their excitatory inputs. These properties would allow a group of FS cells to fire in synchrony when receiving coincident excitatory inputs. Given that FS cells make inhibitory contacts with both pyramidal and nonpyramidal neurons (13–17, 48, 50–52), the simultaneous firing of several FS cells could then coordinate the activity of their postsynaptic cells (7). Thus, we propose that local networks of FS cells could play a role in the readout and transmission of information that may be encoded by the synchronous activity of excitatory neurons.

  • * To whom correspondence should be addressed. E-mail: galarreta{at}


View Abstract

Navigate This Article