Parallel reconstruction of the excitatory and inhibitory inputs received by single neurons reveals the synaptic basis of recurrent spiking

Abstract

Self-sustained recurrent activity in cortical networks is thought to be important for multiple crucial processes, including circuit development and homeostasis. However, the precise relationship between synaptic input patterns and spiking output of individual neurons remains unresolved during spontaneous network activity. Here, using whole-network high-density microelectrode array (HD-MEA) recordings and patch clamping, we developed a novel experimental approach and analytical tools that provide a comprehensive long-term input-output characterization of individual neurons in cortical cell cultures. We found that, during in vivo-like network activity with excitation(E)-inhibition(I) balance, postsynaptic spiking coincided with the maxima of rapid, network state-dependent fluctuations in the input E/I ratio. Our approach also uncovered the underlying circuit architecture and we identified a few key inhibitory inputs — often from special hub neurons — that were instrumental in mediating these E/I ratio changes. Balanced network theory predicts dynamical regimes governed by input fluctuation and featuring a fast neuronal responsiveness. Our findings — obtained in self-organized neuronal cultures — suggest that the emergence of these favorable regimes and associated network architectures is an inherent property of all cortical networks.

Ensemble learning and ground-truth validation of synaptic connectivity inferred from spike trains

“…The a priori defined structure underlying the LIF network served as the first ground truth to compare established and new connectivity inference methods providing a score ( s ) and a weight ( w ) for each connection (v-vi). Moreover, connectivity-inference performance was also assessed on ground truth data obtained from parallel HD-MEA/patch-clamp recordings [58]. B Schematic depicting the architecture of the ensemble artificial neural network (eANN).…”

“…Next, we applied the developed inference pipeline to experimental data, obtained through parallel HD-MEA network and whole-cell patch-clamp recording in in vitro neuronal cultures (dataset 1; for details on the data see [58]; Fig. 4; n=3 patched cells; culture age: DIV16-18; see Sec S3E for details).…”

“…These data allowed probing the effect of varying network dynamics and recording lengths on network reconstruction performance. In addition, we assessed the performance of connectivity-inference algorithms on in vitro data, where connectivity labels, where inferred from parallel whole-cell patch-clamp and high-density microelectrode array (HD-MEA) recordings obtained from primary neuronal cultures [57, 58]. Finally, we introduce an ensemble artificial neuronal network (eANN) and probe whether ensemble learning techniques [59] can improve today’s methodology to infer synaptic connectivity.…”

RT-Sort: an action potential propagation-based algorithm for real time spike detection and sorting with millisecond latencies

“…Spiking activity of the patched neuron was recorded in cell-attached mode or whole-cell current clamp and was also detected by the MEA. The patch-MEA dataset from (Bartram et al 2023) was used, as well as an additional dataset, generated largely according to the same protocol using only the whole-cell current-clamp mode and internal solution for recording. The experimental protocols involving animal tissue harvesting were approved by the veterinary office of the Canton Basel-Stadt according to Swiss federal laws on animal welfare and were carried out in accordance with the approved guidelines.…”

“…The experimental protocols involving animal tissue harvesting were approved by the veterinary office of the Canton Basel-Stadt according to Swiss federal laws on animal welfare and were carried out in accordance with the approved guidelines. Cell culturing protocols are identical to those described in (Bartram et al 2023). Both datasets contained 2 individually patched neurons.…”