Criticality-related dynamics accompany long-term functional decline in primary cortical MEA cultures

Abstract

Primary cortical neuron cultures are widely used to study network development and decline, but their long-term electrophysiological trajectories are poorly characterized. We used multi-electrode arrays to longitudinally monitor Wistar rat cortical cultures for up to ~730 days in vitro (DIV), combining activity measures with survival analysis and criticality metrics. Wells shared an early developmental phase, with mean firing rate peaking at 12-14 Hz around DIV 100-150, before diverging into two phenotypes: early-death wells, showing progressive loss of burst organization, longer inter-burst intervals, shorter bursts, and steep declines in active-electrode recruitment; and long-survival wells, which sustained organized bursting and measurable activity well beyond DIV 400, in some cases past DIV 700. Cox proportional-hazards modeling with FDR correction across thirteen candidate features identified burst duration, inter-burst interval, synchrony, mean firing rate, branching ratio, and Kappa index as significant predictors of time to network failure. Criticality metrics offered the most mechanistic insight: in early-death wells, branching ratio declined toward zero and Kappa index rose toward its upper bound as networks approached silencing. Beyond their average levels, the rate of change of branching ratio and Kappa index over recordings was the strongest correlate of survival time among all criticality metrics. Neuronal avalanche size and duration exponents and their scaling relation, remained consistent with scale-free dynamics in both phenotypes while networks were active, though avalanche occurrence rate declined progressively in early-death wells without reaching significance as an independent predictor. These results show that long-term network survival is not determined by firing rate alone, but by preserved burst organization, sustained electrode recruitment and near-critical propagation dynamics. This work offers one of the longest electrophysiological characterizations of rodent cortical cultures to date and identifies criticality-based metrics as promising tools for predicting network decline and evaluating interventions to preserve long-term neuronal function.