Sparse network models reproduce experimentally observed spike timing jitter during inspiratory population rhythms in the pre-Bötzinger complex

Neurons within the pre-Bötzinger complex of the ventral respiratory group (VRG) in the medulla generate rhythmic network activity which has been shown to be essential for normal breathing in vivo. Furthermore, many physiological responses of this network are preserved in transverse medullary brain slice preparations, where bursts of population discharge are typically recorded with large, low-impedance electrodes [1]. Recent analysis of multiunit extracellular recordings from this in vitro preparation has shown that the onset timing of spikes from the individual neurons that make up this rhythm is quite variable on a burst-by-burst basis. A conductance-based model [2] of this network with all-to-all connectivity fails to reproduce this variability, while a model with sparse (but proportionally stronger) synaptic coupling generates firing patterns with a higher fidelity to the experimental results (see Figure 1). Specifically, the mean variance of burst-triggered spike timing was 0.0218 ± 0.0184 (SD) for in vitro recordings, 0.0347 ± 0.0267 for sparsely connected network models, and 0.0004 ± 0.0023 for fully connected models. Continuing simulations will explore further relationships between parameters of network topology and spike timing variability.

Figure 1

In vitro and modeling results for networks with varying connectivity. Schematic of transverse slice preparation of the ventral lateral medulla (A) and connectivity profiles for fully (B) and sparsely connected (C) network models. Example spike time raster of individually identified neurons during normal population rhythms (D) show noticeable variability in burst onset timing. This variability is reproduced in sparsely connected network models (F) but not in fully connected networks (E). For simulation rasters a random selection of 20 of 300 simulated cells is shown.