Tentori, E.; Kastellakis, G.; Maschietto, M.; Leparulo, A.; Poirazi, P.; Mazzucato, L.; Allegra, M.; Vassanelli, S.

Dissociated neuronal cultures on High-Density Multi-Electrode Arrays (HD-MEAs) provide a controllable in vitro platform to systematically investigate the formation, development, and functioning of neural circuits, as well as their response to a wide array of electrical and pharmacological manipulations. Yet, this potential cannot be fully exploited without a a reliable characterization of the intricate web of causal interconnections within the culture. In this work, we present a protocol to map the network-wide impact of focal stimulation by sequentially stimulating individual sites and recording the evoked responses across the culture. We demonstrate that evoked responses are consistent and replicable across stimulation trials, allowing us to map causal links in terms of the strength of stimulus-response effects, termed Interventional Connectivity (IC). However, this time-consuming protocol may be inconvenient for experimental settings that demand rapid recordings or no perturbation, such as analyses of spontaneous plasticity. We thus investigate whether IC can be predicted from the network`s ongoing activity by comparing it to Effective Connectivity (EC) estimates derived from relatively short (10 minutes) recordings of spontaneous, unperturbed activity. We show that EC is strongly informative of IC, regardless of the specific metric used for EC inference (Transfer Entropy, Signed Cross-Correlation, or Cross-Covariance), proving that spontaneous activity can reliably reveal key aspects of the culture`s causal architecture. Our findings open the way to robust network analyses of in vitro neuronal systems, and the design of effective, targeted stimulation experiments exploiting the culture`s strongest connections or most highly connected sites.