Olfactory responses in the cortico-hippocampal pathway.

(A) Schematic representation of the experimental setup. Infrared thermography was used to measure respiration. (B) Illustration of recorded brain areas AON, aPCx, LEC, CA1, and SUB. (C) Examples of neurons responding to either of two different familiar odors in each of the five recorded areas. Green shading indicates the time window of odor delivery. (D) Fraction of recorded neurons per area responding to at least one odor. Odor responsiveness per area (not significant, p>0.3, Kruskal Wallis test). Error bars: ± standard deviation (SD) between experiments. (E) Raster plot during presentation of odors for two example neurons in the aPCx (top) and the SUB (bottom). Histogram of the ISI for those neurons and fitting value of gamma function (right). (F) Distribution of alpha parameters (p<0.01, Kruskal-Wallis). AON and aPCx significantly different from LEC, CA1, and SUB (**p<0.01, Mann-Whitney). (G) Responses of all recorded neurons (n=1414) to familiar odors. Yellow indicates an increase from baseline; blue indicates a decrease. Each row represents one neuron. (H) Responses of all recorded neurons (n=1414) to novel odors. Yellow indicates an increase from baseline; blue indicates a decrease. Each row represents one neuron.

Experience modulates odor responses across the cortico-hippocampal pathway.

(A) Example raster plots of one neuron in each brain area during the presentation of either a novel (orange) or a familiar (blue) odor. The analysis window is highlighted in green in the raster plots. The histograms represents the spike/trials per 50ms bins of time. (B) Cumulative inhalation counts for all experiments in the time window around odorant delivery (time point 0). All subsequent analyses were performed in a 400ms analysis window (highlighted in green), when breathing was indifferent between novel and familiar conditions (ns: not significant, *p<0.05, **p<0.01, paired t-test). (C) Difference in the proportion of neurons significantly excited (red) or inhibited (blue) by novel relative to familiar odors. The difference in the proportion of unresponsive neurons between novel and familiar conditions is shown in grey. (D) Sankey diagram showing the direction and magnitude of changes in the proportion of neurons significantly excited (red) or inhibited (blue) by familiar (left) and novel (right) odors. The direction and magnitude of changes in the proportion of unresponsive neurons is shown in grey. The width of the band represents the number of neurons in each category. Neurons with mixed response patterns were excluded from this analysis. (E) Absolute difference in firing rate between novel and familiar odors for excited (red) and inhibited (blue) neurons (**p<0.01, *** p<0.001, Mann-Whitney U test). (F) Fraction of neurons with significant responses to a given number of odorants (1-10). (G) Fraction of neurons with significant responses to odors of either familiar, novel, or both stimulus categories. (H) Decoding accuracy of odor identities calculated for an increasing sample of neurons. (I) Decoding accuracy of experience calculated for an increasing sample of neurons. To ensure accuracy can be compared between regions, we subsampled in each region the largest common number of neurons recorded (185 neurons). (J) Decoding accuracy of odor identities (p<0.01, ANOVA). The AON significantly differed from all other regions (**p<0.01, Student’s t-test). Error bars: 95% CI of the mean. (K) Decoding accuracy of experience (black, Experience, error bars: SD) and decoding accuracy using cross-condition generalization performance (grey, CCGP, error bars: SD between excluded odors). The decoding accuracy was significantly different between regions (p<0.01, ANOVA). The AON had a significantly higher decoding accuracy than any other region (p<0.01, Welch t-test). The decoding accuracy was not significantly different between CCPG and Experience in AON and aPCx (ns), but significant in the other regions (***p<0.001, Wilcoxon test).

A subpopulation of neurons in the aPCx is tuned to familiar stimuli.

(A) Accuracy of decoding the chemical identity of familiar and novel odors (*p<0.05, **p<0.01, paired t-test). Error bars: 95% CI of the mean. (B) Mutual information (MI) between odor identities and neuronal responses (**p<0.01, paired t-test). Error bars: 95% CI of the mean. (C) Top: Simplified classification scheme in which two neurons contribute differently to decoding stimulus identity. The weight vector is orthogonal to the hyperplane separating the data. Bottom: Distribution of weights of all neurons. Neuron 1 (n1, blue) has a large weight, neuron 2 (n2, green) has a low weight in encoding stimulus identity. (D) Distribution of weights (absolute values) of the SVMs decoding the identity of either novel or familiar odorants. Significant differences were observed in aPCx, CA1, and LEC (ns: not significant, *p<0.05, **p<0.01, Wilcoxon rank- sum test). (E) MI of outlier subpopulations (ns: not significant, **p<0.01, Welch’s t-test). Error bars: 95% CI of the mean.

The representation of identity and experience become independent from each other along the cortico- hippocampal pathway.

(A) Pearson correlation coefficients between SVM weights for decoding familiar and novel odor identity (*p<0.01, **p<0.01). Error bars: Standard error of the mean (SEM). (B) Schematic representation of the methodology to calculate error rate inside and outside the experience category (novel vs. familiar) with the example of a novel odor A. (C) Proportion of errors per region. Solid and dashed lines represent the proportion of errors inside and outside the same experience category, respectively. (ns: not significant, **p<0.01, Wilcoxon rank-sum test). Error bars: SD. (D) Position of odors in the neural space for the 10 trials of four familiar odors (blue) and six novels (orange). The first two components of the PCA were used. Percentage explained by the first two dimensions: AON 31%, aPCx 15%, LEC 12%, CA1 15%, and SUB 28%. (E) Normalized Euclidean distance of principal components between novel and familiar odors divided by the average distance between all trials, pooled irrespective of experience (p<0.01, Kruskal- Wallis). Post-hoc pairwise comparison revealed significant differences between all regions (*p<0.05, **p<0.01, Mann- Whitney). Error bars: 95% CI of the mean. (F) Probability of responding to one or more familiar odors if a neuron responded to at least one novel odor (p<0.001, Kruskal-Wallis). AON and aPCx significantly different from LEC, CA1, and SUB (p<0.01, Mann-Whitney). Error bars: 95% CI of the mean. (G) Standard deviation of the firing rate in response to olfactory stimuli (p<0.001, Kruskal Wallis test). Response variability progressively decreases from AON to SUB (ns: not significant, *p<0.01, **p<0.01, Mann-Whitney). Error bars: 95% CI of the mean.

Proposed mechanism for the transformation of olfactory representations.

Illustration showing the responses of 5 neurons in AON (top) and SUB (bottom) to either of 3 novel (orange) or familiar (blue) odorants. Novel stimuli evoked larger responses than familiar stimuli in the AON, whereas in hippocampal areas, novelty was reflected by the number of responsive neurons. Moreover, the stimulus selectivity decreases from AON to SUB (narrowing of response curves).

Probes and coordinates.