Fig. 4: Power-law adaptation enables temporal whitening of hippocampal spike trains.

A Reaction of the vesicle cycle when stimulated with a spike train with 1/f power spectrum and an average rate of 1 Hz (Color code as in Fig. 1B, C). B The response of the vesicle cycle model shows a flatter spectrum when compared to the original 1/f spectrum of the input. In comparison, the model with a single vesicle recovery timescale only attenuates high-frequency fluctuations. Power spectra are plotted as normalized densities. The full and single timescale models have comparable average power, with average release rates of 0.26 Hz and 0.29 Hz, respectively. C Power-law adaptation dynamically adjusts the releases per spike to the input rate. Here, we demonstrate this behavior in the strong input regime (~4 Hz), which allows for clear visualization. During high activity, releases per spike are selectively reduced, mainly mediated by slow endocytosis. During low activity, they are enhanced when compared to a single timescale model with similar average release rate. The single timescale model closely follows the linear prediction from input spikes (thin red line). D This can be seen more clearly when computing the average number of vesicles released per spike depending on the input spike rate. C, D Rates were estimated by convolving the spike and release trains with kernels (see Methods). E Reaction of the vesicle cycle when stimulated with a spike train recorded from hippocampus CA1⁶². F Hippocampal spike trains show an approximate f^−0.5 power spectrum over the relevant range. Here the full model nearly flattens the spectrum. Spectra for individual spike trains (transparent lines) behave similarly to their mean, showing that the power law is not a result of averaging spectra of spike trains with different single timescales⁶⁸.