Extended Data Fig. 8: Early sensory and late motor-related components of current-source density and cell-type spiking profiles in visual cortex.

(a) The current source density (colormap, CSD) and event-related potential (black traces, ERP) for auditory and visual stimulus changes in the same example session (MST mouse). (b) Histogram of z-scored videoME (0–500 ms post-change) across visual and auditory trials of all sessions with LFP recordings in V1 (all cohorts). To separate the contribution of motor activity to the LFP, all trials were split into ‘still’ and ‘moving’ trials based on the amount of motor activity. ‘Still’ trials had z-scored videoME between −0.5 and 0.5 and ‘moving’ trials a z-scored videoME larger than 1. (c) For each session a CSD map was constructed using either still or moving trials given the same visual stimuli. Average across n = 46 sessions (NE: 12 sessions; UST: 7; MST: 27). Visual stimuli evoked a consistent and characteristic current source density (CSD) profile with an early sink in L4 and subsequent sink-source pairs in L2/3 and L5/6, in line with earlier reports^33,36,87. (d) The difference between the Vstill and Vmoving maps in c, which we interpret as mostly related to motor differences. Note how most of the motor-related CSD power is expressed after 200 ms in L2-5 and predominantly in superficial and middle layers. (e) Same as c, but for auditory trials. Note how the early sinks and sources in deep layers of the auditory CSD map in the example session of a are only partially reflected in the average. (f) Difference map of the Astill and Amoving maps in e. Note how the movement-associated CSD pattern resembles that of visual trials (d), but is generated somewhat earlier in time. (g) Absolute ERP response (in μV) averaged across cortical depth for selected trial categories. The tick marks and text denote the first time bin the LFP response is different from baseline (−500 to 0 ms) during auditory or visual trials irrespective of motor activity (p < 0.05, Wilcoxon signed rank test, Bonferroni correction). These latencies closely match spiking onset latencies (Fig. 3b). The LFP response for auditory trials can be seen to diverge between still and moving trials around 100 ms after stimulus onset and was significantly different after 243.4 ms (bootstrap test, n = 1000 resamples, p < 0.05) and after 324.4 ms for visual trials (p < 0.05) suggestive of late motor-related signals. Line and shading are mean ± SEM. (h) Laminar organization of onset latencies of visual and auditory responses in V1 (spiking data, not LFP). Top histogram shows the distribution of onset latencies of all significantly auditory responsive neurons (red) and visually responsive neurons (blue). Significance and onset latency were assessed using a binning-free algorithm, ZETA⁷⁷. Spiking onset was significantly earlier for auditory versus visual stimuli (55.3 ms (31.4–108.5 ms) versus 80.3 ms (61.5–98.5 ms); median and interquartile range; F(1,411) = 5.37, p = 0.0209), similar to our earlier population-averaged approach (Fig. 3b). Bottom panel shows each neuron’s onset latency as a function of its recorded depth and cell type. If neurons are bimodally responsive they appear twice. Symbols are scaled by response magnitude. Putative pyramidal cells (broad-spiking) and putative parvalbumin expressing cells (narrow-spiking) were classified based on their waveform. L1 is mostly empty because almost no cells were recorded in that layer. Visually driven cells first began to fire significantly in the middle and superficial layers and later in deeper layers, consistent with the canonical sensory processing scheme^93,94. Auditory-evoked firing started at similar latencies across layers, with many auditory responsive neurons in deep layers. Cortical depth was significantly correlated to spiking onset latency during visual trials (r = 0.71, p = 0.015, Pearson correlation), but not auditory trials (r = 0.13, p = 0.696). *p < 0.05, **p < 0.01.