Regression analysis of the neuronal activity and behavior

We made linear regression analysis to ask whether the firing rate in each bin of time was correlated with parameters of the task. For this purpose, we first aligned the spikes to each epoch (first, second, penultimate, and last lever press for each sequence) and took 5 s before and 5 s after each epoch. Spike trains were transformed to instantaneous firing rate as described above. We calculated the spike frequency using a 200-ms time window with 10-ms steps. We separated the trials and sorted them by the number of lever presses in the sequences (2,4,6, up to 16, grouped by every increment), for the sequence and transition duration (sorted by duration in descending manner, seven categories), for latency (grouped every one second, starting with 0.5 s.). Then we made a regression analysis with permutation test in two manners. For the first, we used a bin of time of 200 ms with a sliding window of 10 ms and ask whether the variable of interest was correlated with the firing rate (Fig. 3A; Extended Data Fig. 3-1B). For the second regressions analysis, we used specific windows of time: (1) mean firing rate 1 s before the start of the sequence; (2) mean of firing rate during the sequence; (3, 4, and 5) average of firing rate during S1, firing rate during transition or firing rate during S2 (Fig. 3B). To resolve statistically whether the regression’s p value was significant, we ran 1000 permutations and divided the sum of times that the p value > p value initial between the number of permutations. Only regressions with β coefficient different from zero (p < 0.05 and R > 0.6) were accepted.

Both M2 and M1 contain units encoding the temporality of the self-paced sequences of actions. A, top panels, Examples of two units recorded in M2. In each example, each line is a self-paced sequence depicting the neuronal activity (in z score, black to yellow) and lever press (white points subsequence 1, gray points subsequence 2). The color bar to the right of each plot shows the grouped categories plotted in the middle and bottom panels. Middle row panels, Mean firing rate from each category presented in the upper row. Bottom row, Regression fits from the time bin depicted in gray in the middle row panels. B, top row panels, Significant regression analysis per unit; columns i, between the firing rate (FR) 1 s before the start of the sequences (Seq) versus the duration of the sequences; columns ii, FR during the sequences versus duration of sequences; columns iii, FR in the subsequence 1 (S1) versus the duration of S1; columns vi, FR during the transition versus the transition time; and columns v, FR during the subsequence 2 (S2) versus the S2 duration. Bottom panels, The proportion of units that presented significant regression (R² > 0.6 and p < 0.05). The dashed lines depict comparisons with χ² test. Corrections for multiple comparison was considered (see Materials and Methods). Extended Data Figure 3-1 shows the recorded units in M2 or M1 encoding the temporality of the forced sequences of actions. Extended Data Figure 3-2 shows the confirmation of the M2-M1 projections into the LS and linear regressions between the activity and the temporal parameters of the execution of sequences from the photo-identified M2 or M1 cortico-striatal neurons. n.s. = p > 0.05.

Both M2 and M1 contain units encoding the temporality of the forced sequences of actions. A, top row panels, Significant regression analysis per unit columns i, between the firing rate (FR) 1 s before the start of the sequences (Seq) versus the duration of the sequences, columns ii: FR during the sequences versus duration of sequences, columns iii: FR in the subsequence 1 (S1) versus the duration of S1, columns vi: FR during the transition versus the transition time, and columns v: FR during the subsequence 2 (S2) versus the S2 duration. Bottom panels, The proportion of units that presented significant regression (R² > 0.6 and p < 0.05). The dashed lines depict comparisons with χ² test. Corrections for multiple comparison was considered (see Materials and Methods). B, Percentage of recruited units (presenting significant regression, bins of 200-ms sliding window of 10 ms regarding the tittles in each plot). Bars show the average of the recruited units 1 s before and after the start of the sequences except for the transitions, which is aligned to the last press of the subsequence 1. Download Figure 3-1, TIF file.

Confirmation of the M2-M1 cells and their projections into the LS and linear regressions between the activity and the temporal parameters of the execution of sequences from the photo-identified M2 or M1 cortico-striatal neurons. A, Scheme of retrograde tracing of M2 and M1 neurons (left) labeled by retrobead injection into the LS (right). B, left panels, Anteroposterior coronal photomicrographs showing retrobead labeled cells in M2/M1 (left) from the retrobead injections into the LS. Scale bar: 500 μm. Right panels, Diagrams presenting summed data from four animals that were injected with retrobeads into the LS in binary color code. C, Scheme of premotor cortex anterograde tracing using AAV-Retro-Cre-mCherry injection into the LS (injection1) and AAV-DIO-eYFP into the premotor cortex (M2; injection2). D, Left photomicrographs depict an example from one animal of the experiment described in C. Prelimbic cortex (PL), cingulate cortex (Cg1). Scale bar: 500 μm. Right top panels, Zoom-in of the LS and the medial striatum (MS) showing the premotor cortico-striatal projections labeled in green. Scale bar: 50 μm. The bars show the quantification of the M2→LS fibers reaching the LS at AP. 0.3 from bregma (coordinate where most of the fiber tips of the optogenetic inhibition were found). E, Diagram of the injection site to express ChR2 into M1 or M2 cortex and their projections, the electrode array (M2 or M1 cortex), and the optical fiber implantation (LS). F, Raster-plot example of a cortico-striatal PID unit and the perievent time histogram aligned at the start of the first pulse of a train of blue light (10 Hz; 2 mW, 473 nm). Right panels, Mean waveforms, principal component analysis, and latency to response to stimulation during behavior (black) and photo-identification protocol (blue). G, top row panels, Significant regression analysis per unit; columns i, between the firing rate (FR) 1 s before the start of the sequences versus the sequences duration; columns ii, FR during the sequences versus sequences duration; columns iii, FR in the subsequence 1 (S1) versus the duration of S1; columns vi, FR during the transition versus the transition time; and columns v, FR during the subsequence 2 versus the subsequence 2 duration. Bottom row panels, Proportion of units that presented significant regression (R² > 0.6 and p < 0.05) during forced (red) or self-paced sequences (blue). Despite tendencies, no significant differences were found in the comparisons within or between cortical regions, χ² test. Download Figure 3-2, TIF file.