Supplementary Figure 5: VTA^DA→BA axon responses during training and following completion of training.

a, Left: all single-trial responses (rows) from 8 mice during Pavlovian training of RC paired with Ensure delivery (“During training” refers to sessions during the first 3 days of training before introducing aversive trials), sorted by onset of first lick following RC onset (blue ticks). Green ticks: time of Ensure delivery. Responses were mostly evident during Ensure consumption. This was especially pronounced when Ensure delivery was unexpected, as seen in rows with longer latencies to first lick. Right: behavioral performance of mice during training (n = 8 sessions, one session/mouse) used for recordings in panels c-e, below. Percentage of reward cue trials with licking during the cue (‘hit rate’) and percentage of false alarms (incorrect licking after neutral cues or during blank trials) were both high early in learning, indicative of poor discrimination between cues. Mean ± s.e.m. b, Left: all single-trial responses (rows) from 10 mice following completion of training, sorted by onset of first lick following RC onset (blue ticks). Green ticks: time of Ensure delivery. Responses were locked to cue onset, not to motor response onset. Right: behavioral performance of trained mice (n = 10 sessions, one session/mouse) used for recordings in Fig. 2d, e and in panels c-e, below. Left: percentage of correct responses across reward cue trials (‘hit rate’). Right: percentage of false alarms (incorrect licking after neutral or aversive cues) across all neutral or aversive trials. Mean ± s.e.m. c, Mean response timecourses of VTA^DA→BA axons during training and following completion of training on task. Response timecourse of Ensure delivery related activity was obtained by subtracting a monoexponential fit of the cue response. The window used for analysis of Ensure delivery responses is indicated by a blue bar. Error bars: s.e.m. across mice. Z: Z-score. d, Comparison of cue responses and Ensure delivery responses during training vs. following training. Mean ± s.e.m. *** p < 0.0001, two-sided Wilcoxon rank-sum. e, The outcome response bias index shifted from positive to negative following training, indicating a shift in response from the time of Ensure delivery to the time of cue onset. Mean ± s.e.m. *** p < 0.0001, two-sided Wilcoxon rank-sum. f, Population response, averaged across those AC-Av trials early in training that involved licking during the behavioral response window (“false alarm”, thereby eliciting quinine delivery) or across trials without licking (“correct reject”, thereby eliciting no outcome). A monoexponential fit was subtracted to isolate the response magnitude to quinine delivery or no outcome. The window used for analysis of outcome responses is indicated by a blue bar. Mean ± s.e.m., n = 10 mice. Z: Z-score. g, Comparison of AC-Av cue and outcome responses for false alarm trials vs. correct reject trials. Mean ± s.e.m., n = 10 mice, * p = 0.04, two-sided Wilcoxon sign-rank. h, Population response, averaged across those NC trials early in training that involved licking during the behavioral response window (“false alarm”) or across trials without licking (“correct reject”). In both cases, no outcome occurs following cue offset. A monoexponential fit was subtracted to isolate the response magnitude during the outcome window. The window used for analysis of outcome responses is indicated by a blue bar. Mean ± s.e.m., n = 10 mice. Z: Z-score. i, Comparison of NC cue and outcome responses during false alarm trials vs. correct reject trials. Mean ± s.e.m., n = 10 mice.