Sex-specific exploration accounts for differences in valence learning in male and female mice
- Integrated Program in Neuroscience, McGill University, Canada
- Department of Psychology, McGill University, Canada
- Ludmer Centre for Neuroinformatics and Mental Health, Canada
Figures
Figure 1 with 2 supplements
Standard indicators of valence learning point to poor task performance.
(A) Overview of the experimental design and apparatus. (B) In females, head entries in the first 10 s of the cue are significantly higher to the CSR compared to the CS− at the end of training in both mixed (left) and appetitive-only (right) paradigms. On recall day (R), discrimination in head entries only remains significant in the mixed paradigm. (C) Male mice are performing more head entries to the CSR compared to the CS− only at the end of the appetitive-only paradigm, which returns to non-significance on recall day. Supplementary file 1a contains full statistical analyses for both sexes, including comparison to CSS. (D) In females, percent time spent freezing is increased in response to CSS compared to CS− in the mixed-valence paradigm, but not in the aversive only paradigm throughout training. On recall day, females have significantly higher freezing levels to the CSS than the CS− in both paradigms. (E) Male mice have increased CSS freezing in the mixed and aversive-only paradigm as compared to CS−, and discrimination remains significant in both paradigms on recall day. Supplementary file 1b contains full statistical analyses for both sexes, including comparison to CSR. *padj < 0.05. Errorbars indicate SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8; Aversive: nMale = 8, nFemale = 8.
Figure 1—figure supplement 1
Port-based readouts of reward learning.
(A) Time in port, (B) licks, and (C) latency to enter the food port during the cue in the mixed (left) and appetitive-only paradigms (right) in females (top) and males (bottom). Stars indicate significance between the CSR and CS−, with full statistical results, including comparisons to the CSS, reported in Supplementary file 1a. Similar to head entries (Figure 1), other port-related metrics do not show evidence of robust associative reward learning in either the mixed or the appetitive-only paradigm in either sex. *padj < 0.05. Errorbars indicate SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8.
Figure 1—figure supplement 2
% Freezing quantified using different motion threshold and freezing duration parameters in ezTrack.
Across all motion thresholds, better discrimination between CS− and CSS is observed in the mixed paradigm compared to the aversive-only paradigm for females, while males discriminate similarly in both the mixed and aversive-only paradigm. With increasing freezing duration, cue discrimination diminishes for both sexes in the mixed paradigm compared to the aversive-only paradigm, suggesting shorter freezing bouts in the mixed protocol. *padj < 0.05. Errorbars indicte SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Aversive: nMale = 8, nFemale = 8.
Figure 2 with 4 supplements
Cue identity can be reliably predicted from data-driven analysis approaches.
(A) Overview of partial least squares discriminant analysis (PLS-DA) approach. (B) Predictions of cue type on recall day using two-way classifiers in individual female (filled points) and male (empty points) mice in the mixed training dataset, mixed test dataset, and appetitive- and aversive-only datasets. Point colors indicate true cue type, with predicted cue type indicated by position on x-axis. (C) Percent accuracy of cue type prediction in females (left) and males (right). All accuracy values surpass chance prediction (50%). (D) Regression weights for individual predictors for each two-way classifier. Significant predictors (p < 0.05) are filled with the color of the cue type an increased value is predictive of, non-significant predictors are gray. Full statistical results are reported in Supplementary file 1d. (E) Distributions of variables predicting cue types. Port distance is decreased and port orientation is increased during the CSR, pausing is increased during the CSS, and exploration (sum of all rearing and locomoting syllables) is highest during the CS−. Errorbars indicate 95% confidence interval for jackknife estimate. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8; Aversive: nMale = 8, nFemale = 8.
Figure 2—figure supplement 1
Trajectory plots of individual syllables.
Videos were used for full evaluation and labeling of individual syllables, in combination with the shown trajectory plots. Sample videos for each syllable are available with all other data at the link provided in the data availability section.
Figure 2—figure supplement 2
Prediction accuracy and component (latent variable) selection.
(A) The cue period was split into 5- or 10-s intervals prior to predicting cue type identity. (B) Accuracy of cue type predictions in training and test datasets of three-way classifiers fitted using all data (left column), female-only data (middle column), and male data (right column) in different using data from different cue intervals. (C) Same as B, but pairwise classifiers were trained to compare each cue type to the other. Best performance that generalizes to the test dataset was observed in the cue mid (5 s) period using data from both sexes (values are enclosed in black boxes). (D) For each pairwise classifier, the optimal number of components (latent variables) is 1.
Figure 2—figure supplement 3
% Freezing correlation with % Pausing in single- or mixed-valence test datasets.
Both metrics are strongly correlated across all cue types, suggesting that pausing is largely capturing freezing behavior.
Figure 2—figure supplement 4
Raw values on recall day of all variables included in the partial least squares discriminant analysis (PLS-DA) model not shown in Figure 2.
This includes location-based metrics (A), attending (B–F), grooming (G), licking (H), rearing (I–M), locomoting (O–R), turning (R–V), and jumping (W–Y). Mixed train and test datasets are combined for this visualization. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8; Aversive: nMale = 8, nFemale = 8.
Figure 3 with 1 supplement
Performance trajectories are mediated by underlying sex differences in exploration.
(A) Cue type prediction throughout training in the mixed paradigm (top) and single-valence paradigms (bottom) throughout training in each two-way classifier. Male and female predictions largely follow each other in the mixed paradigm, whereas sex-specific learning trajectories can be observed in both the appetitive and aversive-only paradigm. (B) Cue type prediction on recall day for each paradigm in females (top) and males (bottom) shows cue discrimination between all cue types in both sexes (contrast CS−–CSR: appetitive female: t-ratio360 = –4.635, padj < 0.001; mixed female: t-ratio360 = –7.92, padj < 0.001; appetitive male: t-ratio360 = –0.547, padj < 0.05; mixed male: t-ratio360 = –3.629, padj < 0.001; contrast CS−–CSS: aversive female: t-ratio360 = –2.165, padj < 0.05; mixed female: t-ratio360 = –15.1, padj < 0.001; aversive male: t-ratio360 = –4.544, padj < 0.001; mixed male: t-ratio360 = –11.746, padj < 0.001; contrast CSR–CSS: mixed female: t-ratio360 = –7.18, padj < 0.001; mixed male: t-ratio360 = –8.117, padj < 0.001). The full statistical model is reported in Supplementary file 1e. (C) Percent time spent exploring in the pre-cue period during habituation, first day of training (day 1), end of training (day 13), and on recall test (R). Males and females differ in their levels of exploration on habituation (contrast female–male: t-ratio30 = –4.281, p < 0.001), and throughout training in the single-valence paradigms (contrast female–male: appetitive: t-ratio85.9 = –2.559, p < 0.05; aversive: t-ratio88.1 = –1.784, p = 0.078) but not the mixed-valence paradigm (contrast female–male: t-ratio94.3 = –0.373, p = 0.71). The full statistical model is reported in Supplementary file 1f. (D) Cue-relevant behaviors port distance (CSR predictive, left) and pausing (CSS predictive, right) negatively correlate with % exploration during cue presentation. *padj < 0.05, **padj < 0.01, ***padj < 0.001. Errorbars and ribbons indicate SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8; Aversive: nMale = 8, nFemale = 8.
Figure 3—figure supplement 1
CSR predictive behaviors in females (top) and males (bottom).
(A) Mice decrease distance to food port in both mixed and appetitive paradigms, but females do so earlier than males. (B) Similarly, port orientation increases in both paradigms, with discrimination being expressed earlier in females than in males. (C) % exploration decreases during the CSR robustly by the end of training in females but not in males. Full statistical results are reported in Supplementary file 1i. *padj < 0.05. Errorbars indicate SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8.
Figure 4
Repeated shock exposure promotes widespread pausing, limiting exploratory behavior in female mice.
(A) Pre-cue exploration and pausing behaviors show significant differences between mixed-valence and aversive-only paradigms throughout training and recall in females (top; contrasts explore mixed vs aversive: all days: t-ratio77.5 = –6.201, padj < 0.001; day 1: t-ratio195 = –3.548, padj < 0.001; day 13: t-ratio191 = –5.219, padj < 0.001; day R: t-ratio186 = –4.81, padj < 0.001; contrasts pause mixed vs aversive: all days: t-ratio77 = 3.338, padj < 0.01; day 1: t-ratio213 = 1.025, padj = 0.306; day 13: t-ratio212 = 2.7, padj < 0.01; day R: t-ratio210 = 2.886, padj < 0.01). In males (bottom), time spent exploring is comparable across paradigms, and time spent pausing is only modestly increased in the aversive-only paradigm during training (contrasts exploring mixed vs aversive: all days: t-ratio73.6 = –2.159, padj < 0.05; day 1: t-ratio185 = –2.419, padj < 0.05; day 13: t-ratio187 = –2.051, padj < 0.05; day R: t-ratio185 = –0.288, padj = 0.774; contrasts pausing mixed vs aversive: all days: t-ratio72 = 1.288, padj = 0.202; day 1: t-ratio210 = 0.367, padj = 0.714; day 13: t-ratio210 = 0.277, padj = 0.782; day R: t-ratio210 = 1.911, padj = 0.057). The full statistical model is reported in Supplementary file 1g. (B) Correlations between CS− and CSS pre-cue and cue exploration (left) and pausing (right) behaviors show that CS− behaviors correlate significantly with pre-cue behaviors, whereas CSS behaviors do not. (C) Percent time spent pausing increases over the course of the session from the first two cues (early; E) to the last two cues (late; L) in the pre-cue period on the first day of training in both paradigms and sexes (contrasts day 1 early vs late: females mixed: t-ratio360.7 = –3.381, padj < 0.001; females aversive: t-ratio360.7 = –6.097, padj < 0.001; males mixed: t-ratio360.7 = –5.249, padj < 0.001; males aversive: t-ratio360.7 = –4.424, padj < 0.001). On day 13 (resolved in bottom plot), both male and female mice show small but significant increases in pausing at the end of the session while, in the aversive-only paradigm, females show a large increase in pausing at the end of the session absent in males (contrasts day 13 early vs late: females mixed: t-ratio360.7 = –2.802, padj < 0.01; females aversive: t-ratio360.7 = –4.023, padj < 0.001; males mixed: t-ratio360.7 = –4.613, padj < 0.001; males aversive: t-ratio360.7 = –1.229, padj = 0.22). On recall day (R), pausing levels are only significantly elevated at the end of the session in males in the mixed paradigm (contrasts day R early vs late: females mixed: t-ratio360.7 = 0.113, padj = 0.91; females aversive: t-ratio360.7 = 0.921, padj = 0.353; males mixed: t-ratio360.7 = –2.355, padj < 0.05; males aversive: t-ratio360.7 = 0.832, padj = 0.406). (D) Similarly, percent time spent pausing to the CS− increases throughout the session on day 13 in females in the aversive-only paradigm, with a more modest increase in males in the mixed-valence paradigm (contrasts day 13 early vs late: females mixed: t-ratio356.5 = –0.925, padj = 0.355; females aversive: t-ratio356.5 = –2.694, padj < 0.01; males mixed: t-ratio360.7 = –3.332, padj < 0.001; males aversive: t-ratio360.7 = –0.492, padj = 0.623). (E) On day 13 levels of pausing are stable from early to late session, except for a modest increase in males in the mixed paradigm (contrasts day 13 early vs late: females mixed: t-ratio356.7 = –0.96, padj = 0.338; females aversive: t-ratio356.7 = 0.335, padj = 0.738; males mixed: t-ratio357.9 = –3.164, padj < 0.01; males aversive: t-ratio356.7 = 0.347, padj = 0.729). Significance levels are shown for day 13 comparisons only (bottom plots), with full statistical results reported in Supplementary file 1h. #padj < 0.1, *padj < 0.05, **padj < 0.01, ***padj < 0.001. Errorbars indicate SE. Sample sizes: Mixed: nMale = 37, nFemale = 37; Appetitive: nMale = 8, nFemale = 8; Aversive: nMale = 8, nFemale = 8.
Additional files
-
Supplementary file 1
Supplementary tables for statistics and syllable description.
(A) Linear mixed-effects regression (LMER) on food-port-related metrics. (B) LMER on freezing with different thresholds. (C) Syllable identities and description. (D) Jackknifing statistics for regression coefficients for PLS models. (E) LMER comparing predicted values from PLS model on recall day to assess cue discrimination. (F) LMER on Pre-Cue exploring including habituation day. (G) LMER on Pre-Cue pausing and exploring. (H) LMER on pausing in early and late cue by day, sex and paradigm with contrasts for pre-cue, CS−, and CSS. (I) LMER on reward predictive behaviors.
- https://cdn.elifesciences.org/articles/108498/elife-108498-supp1-v2.xlsx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/108498/elife-108498-mdarchecklist1-v2.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Sex-specific exploration accounts for differences in valence learning in male and female mice
eLife 14:e108498.
https://doi.org/10.7554/eLife.108498
