Extended Data Fig. 3: Interneuron firing properties remain overall stable across days and over training.

a. Example PV cell recorded for 5 consecutive imaging sessions, plotted as in Fig. 3. Dashed lines: significant field over a specific odor (line covers corresponding trials) or through both odors (Day 4). b. Progression of firing peak time bins across days for PV (left) and SST cells (right) recorded across multiple sessions. Gray: Non-odor-specific fields. Yellow, green: odor A- or odor B-specific fields, respectively. Open circles: nonsignificant peaks. Lines connect a cell’s progression. c. Progression of odor selectivity index across days, displayed as in b (sessions on x axis). d. Odor-decoding accuracy of SVM decoders trained on odor-specific cells, non-odor-specific cells, or no-field cells, during odor-presentation, in naïve (top) and trained sessions (bottom), plotted as in Fig. 2n. e–g. Progression of average firing rate (e), theta power (f) and phase locking strength (mean vector length; g) across all trials, as in c (field-type not displayed). Right: Mean ± SE for last naïve vs first trained session and for every X vs X + 1 trained session. * P = 0.0068, paired-sample two-sided t-test (distributions of PV and SST cells were pooled for this comparison; n = 11. P > 0.05 for all other comparisons). h. From top: Evolution of mean locomotion per session, mean firing rate per cell over all trials, theta power and theta modulation of spiking (vector length) per cell across PV-Cre mice and PV cells (left) or SST-Cre mice and SST cells (right), plotted as in Fig. 3i, j. For PV: n = 31 vs 73 recordings, 31 vs 76 cells; from top: P = 9.44 × 10⁻¹⁰, 0.218, 0.0146, 0.0152; two-sided WT. For SST: n = 35 vs 49 recordings, 36 vs 51 cells; from top: P = 1.23 × 10⁻¹¹, 0.0123, 0.466, 0.543 for SST.