Millisecond timescale synchrony among hippocampal neurons

Abstract

Inhibitory neurons in cortical circuits play critical roles in composing spike timing and oscillatory patterns in neuronal activity. These roles in turn require coherent activation of interneurons at different timescales. To investigate how the local circuitry provides for these activities, we applied resampled cross-correlation analyses to large-scale recordings of neuronal populations in the cornu ammonis 1 (CA1) and CA3 regions of the hippocampus of freely moving rats. Significant counts in the cross-correlation of cell pairs, relative to jittered surrogate spike-trains, allowed us to identify the effective couplings between neurons in CA1 and CA3 hippocampal regions on the timescale of milliseconds. In addition to putative excitatory and inhibitory monosynaptic connections, we uncovered prominent millisecond timescale synchrony between cell pairs, observed as peaks in the central 0 ms bin of cross-correlograms. This millisecond timescale synchrony appeared to be independent of network state, excitatory input, and γ oscillations. Moreover, it was frequently observed between cells of differing putative interneuronal type, arguing against gap junctions as the sole underlying source. Our observations corroborate recent in vitro findings suggesting that inhibition alone is sufficient to synchronize interneurons at such fast timescales. Moreover, we show that this synchronous spiking may cause stronger inhibition and rebound spiking in target neurons, pointing toward a potential function for millisecond synchrony of interneurons in shaping and affecting timing in pyramidal populations within and downstream from the circuit.

Keywords: fast oscillations; gap junctions; hippocampus; interneurons; networks; synchrony.

Figure 1.

Number of recording sessions per interneuron. Where possible, we tracked interneurons across recording sessions based on spike waveform (for specific examples, see Fig. 7). Tracking was performed very conservatively; most cells could only be reliably attributed to a single session.

Figure 2.

Hippocampal microcircuits in the freely moving rat. Bottom, Network of pyramidal cells (triangles) and interneurons (circles) connected through excitatory (cyan), inhibitory (black), and millisecond synchronous connections (red) in a recording session with an 8 shank × 8 electrode (inset; ∼20 μm between electrodes, and 200 μm between shanks; x-axis, shank number) in CA3 and 4 shank × 8 electrodes in CA1. Only connected cells are shown, and distances between cells are not to scale. Letters indicate individual cells. Top, CCGs for pairs comprised of a reference cell (first letter) and target cell (second letter). Connections were assessed based on 95% global bands using 5 ms jitter for inhibition, and 1 ms jitter for excitation and millisecond synchronicity. However, for simplicity, significance bands in all (top) 15 panels are illustrated with 5 ms jitter. Dashed green line indicates global significance bands; dashed red line indicates local significance bands; blue represents the mean. A sample excitatory connection between pyramidal cell (k) and interneuron (h) is depicted in panel kh. The pair in panel gh exhibit both an inhibitory connection, from cell (g) onto cell (h), as well as millisecond synchronicity, evidenced by a significant peak in the 0 ms time bin. Additional examples of millisecond synchronous pairs of interneurons are shown in panels ah–nq. An example of a nonsignificant pair (aj) is also shown. For this pair, the CCG peak did not reach significance using 1 ms jitter, although it did with 5 ms jitter as shown (see Materials and Methods). Millisecond synchronicity was also observed between pyramidal cells and interneurons both in CA3 (examples bh and dh) and in CA1 (example or). Many such pairs also exhibited an inhibitory connection. In a few rare instances, millisecond synchronicity was observed in pyramidal–pyramidal pairs (fi). JBSI, reflecting the height of the observed central bin relative to jittered spike train, is provided on the top of each panel. Colors of interneuron cell bodies correspond to putative interneuron types defined in Figure 8.

Figure 3.

JBSIs for synchronous interneuron pairs in CA3 and CA1. Higher millisecond synchrony is observed among interneuron pairs in CA3 (bottom, median = 0.0083) than in CA1 (top, median = 0.0078), with more CA3 pairs found in the right tail of the skewed distribution (JBSI > 0.02). The same data are shown in both linear (left) and logarithmic (right) scale (Buzsáki and Mizuseki, 2014).

Figure 4.

Millisecond synchrony was not dependent on the network state. Significant 0 ms peaks were observed even when spikes were partitioned according to network state, during non-θ LIA, θ, and ripples alone (sample pairs ah and gh from Fig. 2). Local significance bands (dashed red line) from 5 ms jitter are shown, along with the jittered mean (dashed blue line). Similar results were observed in other pairs.

Figure 5.

γ independence of millisecond synchrony. Significant bin counts were observed at 0 ms for synchronous pairs when the CCGs were recalculated for subsets of spikes, according to the phase of the local γ oscillation recorded from the same shank as the reference neuron. Pairs ah and gh from Figure 2 are illustrated. Four phase quadrants correspond to the rising, peak, descending, and trough phases of γ. Although γ modulation is evident in the CCGs, significant peaks were observed in the 0 ms bin for all phase quadrants. Similar results were observed in other pairs.

Figure 6.

High-resolution cross-correlations between millisecond synchronous pairs. CCGs were calculated with time bins set at the sampling resolution of the recording system (here 0.03 ms). This generally resulted in less obvious peaks between −0.5 ms and 0.5 ms, as 0 ms bin counts were distributed among more (finer) bins (e.g., ah, gh), in which some very small peaks could occasionally be seen (e.g., jm, and hm). However, in a few cases, a very well-defined temporal relationship could be resolved in the spike times of interneuron pairs (e.g., eh, ch), including apparent inhibitory influence surrounding a prominent peak slightly offset from 0. In other pairs (e.g., aj, jl), wider yet unmistakable peaks were apparent, without the visible troughs. In some pairs, a small and narrow central peak was observable, bordered by negative dips (e.g., nq and op).

Figure 7.

Interneuron firing responses during sharp-wave ripples were generally consistent across days. Cells g and m from Figure 2 are shown across multiple recording days and sessions. Multiple sessions from different recording days are depicted. Top panel for each session, Probability of firing averaged across all ripples detected during that session. Time was normalized into 32 bins, with 0 corresponding to the peak power in the ripple frequency band (130–230 Hz), and ripple onset and offset corresponding to −1 and 1, respectively. Bottom panel for each session, Average waveform of spikes attributed to the interneuron on each of 8 recording electrodes.

Figure 8.

Putative classification of interneurons based on firing response during sharp-wave ripples. Interneurons were sorted into nine classes (shown in different colors) based on visual inspection of firing during sharp-wave ripples averaged across all connectable sessions (Fig. 7). Although some of the sorting was inevitably subjective, classification approximately corresponded to k-means clusters obtained from principal component features from the time-normalized firing responses to ripples from all interneurons recorded in sessions with at least 20 ripples. a, Top, Manual clusters with corresponding colors to b, with nearby classes chosen to be close in color space. Bottom, Automated clusters based on top five principal components (arbitrary coloring) along first and second principal components. b, The ripple firing response for all cells assigned to each class of nine classes were averaged. Normalized time −1 indicates the ripple onset, 0 indicates the ripple peak, and 1 indicates the ripple offset. Overall, some cells firing robustly during ripples (blue Group 1–3), whereas others showed transient and varying responses (green Group 4–6) or decreased in firing (red Group 7–8) or failed to show a discernible response (9). The colors used for interneurons in Figure 2 were based on the same color/category relationships. c, Histograms for these cell types display the distribution of the total number of cells in all recordings with at least 20 ripples that were assigned to each class group (black), those that were excited by a simultaneously recorded pyramidal unit (brown), those that displayed millisecond synchrony (red), those that displayed millisecond synchrony with cells from Classes 1–3 (orange), those that displayed millisecond synchrony with cells from Classes 4–9 (dark yellow), those that inhibited at least one other simultaneously recorded cell (light yellow), and those that were inhibited by a simultaneously recorded cell (white). These color codes are unique to this panel and not used elsewhere. Synchronous pairs were observed in all nine cell types (data not shown), and all cell types formed synchronous pairs with cells in the blue group (Types 1–3).

Figure 9.

θ and γ phase preferences of different cell types. Scatterplots were composed based on preferred firing phase of neurons during CA3 γ and CA1 θ phase for cells in both regions. Some general group differences could be observed in the clustering of phase preferences of different neuronal types; for example, Type 4–6 (green) and Type 7–9 (orange) interneurons in CA1 (a) showed different θ and γ phase preferences, and Type 1–3 (blue) and Type 7–9 (orange) classes in CA3 (b) showed different θ phase preferences. Color code is the same as in Figure 8b.

Figure 10.

Millisecond timescale synchrony between pyramidal cells and interneurons oscillates during sharp-wave ripples. The CCG of these pairs (Fig. 2, pair bh in CA3 and pair or in CA1) using spikes firing during ripples alone indicates they did not fire antiphasically, as might be expected in interneuron–pyramidal pairs, but instead fired with apparent oscillatory coupling at ∼200 Hz, corresponding to the ripple oscillation. Data pooled from significant 0 ms pyramidal–interneuron pairs across all sessions shows millisecond synchrony during θ (a) and during ripples (b), with notable secondary peaks at 5 ms and −6 ms. Similarly, pooled data from significant 0 ms interneuron–interneuron pairs in CA3 (c) or in CA1 (d) across all sessions show millisecond synchrony during ripples but without visible secondary peaks.

Figure 11.

Sample interneuron population raster in a freely moving rat. Shown here is a 500 ms raster plot from the 21 interneurons depicted in Figure 2, along with the CA1 pyramidal layer local field potential on the top. Red dots indicate instances of synchronized firing between two or more cells. Millisecond synchronous firing appeared to be largely stochastic and naturally varying with firing rates, with different interneurons synchronizing at different time points.

Figure 12.

Effect of millisecond synchrony on pyramidal populations. All interneuron spikes across all sessions were labeled according to whether they occurred within 0.5 ms of spikes of interneurons on other electrodes. For synchronous spiking of CA1-CA1 cell pairs, (a) CCGs were calculated between synchronous interneuron spikes and all CA3 pyramidal cell spikes (top left) and all CA1 pyramidal cell spikes (top right). y-axes indicate total spike counts (×1000) from all combined pyramidal cell spikes. These can be compared with similar CCGs calculated between an equal number subset of randomly selected nonsynchronous interneuron spikes and pyramidal cell firing in CA3 (bottom left) and CA1 (bottom right). These comparisons demonstrate greater immediate inhibition in CA1 along with a subsequent (postinhibitory) increase in excitation after millisecond synchronous activity. b, For synchronous interneuron spiking in CA3-CA3 cell pairs, CCGs were calculated between synchronous interneuron spikes and all CA3 pyramidal cell spikes (top left) and all CA1 pyramidal cell spikes (top right). These can be compared with similar CCGs calculated between an equal number of randomly selected nonsynchronous interneuron spikes and pyramidal cell firing in CA3 (bottom left) and CA1 (bottom right). These comparisons demonstrate that CA3 pyramidal cell spiking was more likely to precede, and less likely to follow, synchronous spiking in CA3 interneurons. Furthermore, a stronger subsequent rise in pyramidal cell activity can be seen in CA1 after these events.