Submillisecond firing synchrony between different subtypes of cortical interneurons connected chemically but not electrically

Abstract

Synchronous firing is commonly observed in the brain, but its underlying mechanisms and neurobiological meaning remain debated. Most commonly, synchrony is attributed either to electrical coupling by gap junctions or to shared excitatory inputs. In the cerebral cortex and hippocampus, fast-spiking (FS) or somatostatin-containing (SOM) inhibitory interneurons are electrically coupled to same-type neighbors, and each subtype-specific network tends to fire in synchrony. Electrical coupling across subtypes is weak or absent, but SOM-FS and FS-FS pairs are often connected by inhibitory synapses. Theoretical studies suggest that purely inhibitory coupling can also promote synchrony; however, this has not been confirmed experimentally. We recorded from 74 pairs of electrically noncoupled layer 4 interneurons in mouse somatosensory cortex in vitro, and found that tonically depolarized FS-FS and SOM-FS pairs connected by unidirectional or bidirectional inhibitory synapses often fired within 1 ms of each other. Using a novel, jitter-based measure of synchrony, we found that synchrony correlated with inhibitory coupling strength. Importantly, synchrony was resistant to ionotropic glutamate receptors antagonists but was strongly reduced when GABA(A) receptors were blocked, confirming that in our experimental system IPSPs were both necessary and sufficient for synchrony. Submillisecond firing lags emerged in a computer simulation of pairs of spiking neurons, in which the only assumed interaction between neurons was by inhibitory synapses. We conclude that cortical interneurons are capable of synchronizing both within and across subtypes, and that submillisecond coordination of firing can arise by mutual synaptic inhibition alone, with neither shared inputs nor electrical coupling.

Figure 1.

Firing patterns and spike parameters of FS and SOM interneurons. A, B, Voltage responses (top traces) of representative FS and SOM cells to hyperpolarizing and depolarizing intracellular current steps (bottom traces). Note the pronounced difference in firing frequency adaptation. C, Superimposed, color-coded spike waveforms of representative FS and SOM interneurons (the SOM spike is larger and slower). D, Scatterplot of adaptation ratio and spike width for all cells. Note the clear segregation of the SOM and FS populations, and the intermixing of FS interneurons from X94 and G42 mice. Note also that the non-GFP-expressing SOM interneuron was not different from the others. The dotted lines indicate the cutoff values (adaptation ratio of 0.47; spike width of 0.28 ms) that best separate the two populations.

Figure 2.

Synaptic connectivity between FS and SOM interneurons. Representative records from unidirectionally (top) and bidirectionally (bottom) connected FS–FS (left) and SOM–FS (right) pairs of layer 4 interneurons; simultaneous records from each pair are shifted vertically and color-coded red or blue. In each panel, the bottom traces show the action potentials fired by the presynaptic neuron and the top traces show the averaged postsynaptic response, recorded at a holding potential of −50 mV. Capacitive artifacts, which were more noticeable on channel 2 of the amplifier (blue traces), are blanked. The diagrams illustrate the recording configuration.

Figure 3.

Submillisecond firing lags between FS and SOM interneurons. Representative records from unidirectionally (top) and bidirectionally (bottom) connected FS–FS (left) and SOM–FS (right) pairs of layer 4 interneurons; simultaneous records from each pair are superimposed and color-coded red or blue. In each panel, the top traces show 250-ms-long segments of firing induced by intracellular current injection into both cells, with spike pairs occurring at submillisecond lags indicated by asterisks. A 50 ms segment is shown at an expanded timescale in the bottom traces. The arrows point to examples of IPSPs.

Figure 4.

Submillisecond synchrony occurred at different firing frequencies. Records from bidirectionally connected FS–FS (left) and unidirectionally connected FS–SOM (right) pairs of layer 4 interneurons; simultaneous records from each pair are superimposed and color-coded red or blue. In each pair, the phasically firing neuron was kept at a constant firing rate, whereas the tonically firing neuron was depolarized to three different firing rates. In all cases, submillisecond spike lags (asterisks above spikes) were observed. The arrows point to examples of IPSPs.

Figure 5.

Cross-correlation histograms reveal excess of submillisecond spike lags. Raw cross-correlograms of spike lags in FS–FS (left) and SOM–FS (right) pairs, averaged from all unconnected (top), unidirectionally connected (middle), and bidirectionally connected (bottom) pairs. In one-way connections, the presynaptic cell was used as the reference cell; the center right correlogram is an average of both SOM→FS and FS→SOM pairs. Bin width is 2 ms; the central bin (indicating lags of ±1 ms) is filled black. Error bars represent SEM.

Figure 6.

Average Z scores and JSSI values for different patterns of connectivity. Error bars are SEM. Note that nonconnected pairs had average JSSI values <0.10 and Z scores that were only marginally significant; synaptically connected pairs had average JSSI values ∼0.20 and highly significant Z scores. The horizontal dotted lines indicate a Z score of 2 (equivalent to p = 0.05) and a Z score of 3.3 (equivalent to p = 0.001).

Figure 7.

The role of IPSPs and EPSPs in promoting submillisecond firing coordination. A, C, Scatterplots of JSSI values versus the ICS for all SOM–FS (A) and FS–FS (C) pairs. Symbols representing bidirectionally connected pairs are filled gray. The linear regression lines apply to all data points and have r² values of 0.50 (A) and 0.35 (C). B, D, The effect of blocking fast glutamate receptors (APV/CNQX), GABA_A receptors (GBZ), or both (APV/CNQX/GBZ) on the JSSI, for nine SOM–FS pairs (B) and seven FS–FS pairs (D). Symbols representing the same pair are connected by lines. Note that blocking ionotropic glutamate receptors had little effect on synchrony, whereas blocking GABA_A receptor profoundly reduced it.

Figure 8.

Submillisecond synchrony in simulated pairs of neurons connected by unidirectional (left) and bidirectional (right) inhibitory synapses. A, B, Representative simulated IPSPs (top traces) and presynaptic action potentials (bottom) (compare with Fig. 2). C, D, Representative simulated paired spike trains (compare with Fig. 3). E, F, Cross-correlograms averaged from ∼10 simulated pairs each (compare with Fig. 5). G, H, Scatterplots of JSSI values versus ICS (compare with Fig. 7A,C).