In 7 developing rat cortical cultures, PCA power-law and PC-locality changes did not come from firing rate or electrode count; distance coupling explained PC locality but not spectral steepening

Nakamuta et al. report that in developing rat cortical cultures on high-density electrode arrays, the PCA variance spectrum becomes a power law, PCs become spatially local, and neuronal avalanches appear. They explain this with heavy-tailed synaptic coupling, but they did not control for firing rate or for the number of active electrodes. We used their public data (DANDI 002020, 7 cultures, 113 of 129 daily sessions) and their methods, and added controls. The published changes reproduced. The spectral exponent increased from 0.54-0.98 to 1.02-1.33, and the participation ratio fell in 7 of 7 cultures. The PC locality slope became steeper in 7 of 7, and the branching ratio increased from about 0.6 to about 0.9 in 7 of 7. Firing rate and electrode count had no consistent day trend. The day trends remained with 200 matched electrodes and after partial control for rate and electrode count. We reduced a mature session to the rate and electrode count of each other day: this gave no day trend in the exponent, decay rate, participation ratio or locality. A surrogate with the same rates and burst timing but no pair structure gave flatter spectra (exponent 0.70 vs 1.11) and no PC locality. Thus pair structure is necessary. A model in which coupling depends only on inter-electrode distance gave the real PC locality (slope -0.35 vs -0.29, p = 0.59) and the fall in dimensionality. It gave a lower exponent (0.82 vs 1.11) and about one third of its developmental increase. When we added per-electrode coupling gains, the model gave about four fifths of the increase. But the gains became less heavy-tailed with maturation (Gini fell in 7 of 7 cultures). Sessions with heavier gain tails had flatter spectra beyond day. Spectrum shape and avalanche exponent did not co-vary beyond their shared day trend. Shot-noise correction showed that activity becomes lower-dimensional, not higher-dimensional. Limitations: unsorted channel-level detections, unknown days in vitro, 7 cultures, and functional (not anatomical) coupling proxies. Measured connectivity is necessary to test the heavy-tailed-synapse mechanism.

neuronal culture · HD-MEA · development · PCA variance spectrum · power law · dimensionality · neuronal avalanches · branching ratio · self-organization · firing rate confound · shot noise · spatial coupling · local Moran's I · structure-function · replication · DANDI 002020