Experimental design and examples of single unit recordings.

A. Icons describing the different subject and stimulus donors. Mouse colors denote the strain, while background colors denote the status (estrus, non-estrus for females, naive, dominant, castrated for males). B. Schematic off the recording setup. A multisite electrode probe is advanced to the external cell layer of the AOB. Stimuli (yellow drop) are applied to the nostril and, after a delay, the sympathetic nerve trunk (SNT) is stimulated. Between stimulus presentations, the nasal cavity and VNO are flushed with Ringer’s solution. VNO: Vomeronasal organ. C. Timeline of an individual trial. After a 20 second delay, the SNT is activated to induce VNO activation. Following another 40 seconds, the flushing procedure is initiated. An inter-trial interval (ITI) of 18 seconds is applied between consecutive stimulus presentations. During the experiment, stimuli are delivered repeatedly in blocks. Within each block, each of the individual stimuli are presented in a random order. D. Examples of single unit responses to all stimuli. For each unit, responses to each of 11 stimuli are shown. Left panels show the mean firing rate (PSTH, peri-stimulus time histogram), while right panels shown corresponding raster displays. The red vertical line indicates SNT stimulation, except for the unit on the right, where it indicates stimulus application to the nostril. In this specific example, responses began immediately after stimulus application, prior to SNT stimulation.

Basic response characteristics.

A. Colormap of responses. Each neuron (n = 546) is represented by one row, and each stimulus by one column, as indicated at the bottom. Responses are normalized between -1 and 1. In this representation (and some analyses), non-significant responses are assigned a value of 0. B. Classical multidimensional scaling of population level responses using the data shown in panel A and a correlation distance metric. Here, the first two dimensions account for 44.6% of the variability. The mean error due to the reduced 2-D representation is 0.29, whereas the mean of the original distances is 0.73. Note the square indicating a distance of 0.1 for comparison. C. Distribution of the number of significant responses across all neurons. D. Percentage of neurons responding to each of the stimuli, separately for rate increases (upper bars, red), and decreases (lower bars, green). E. Mean response magnitude, across all recorded neurons to each of the stimuli. Green lines indicate the standard error of the mean. See Table 1 for statistical comparison of stimulus magnitudes.

Comparison of stimulus response magnitudes across stimulus pairs (related to

Fig. 2E). The tests were conducted using a non-parametric ANOVA (kruskalwallis), and the multcompare MATLAB function with the Tukey Kramer multiple comparisons correction (top), or without correction (bottom).

Analysis of response patterns.

A. Graphical representation of response pattern definitions. To fulfill a pattern, each of the stimuli indicated in red should elicit a stronger response than each of the stimuli shown in blue. The left and center panels show the basic response patterns and their complementary patterns, respectively. The panels on the right provide two examples of adjusted response patterns. For example, a dominant/strain pattern requires fulfilment of three conditions, so that within each strain, responses will be strongest to the dominant stimulus. Within a given condition, hatched squares indicate stimuli that are irrelevant for the pattern in question. The complementary adjusted responses patterns are not shown. B. All response patterns across neurons. Note that a neuron may fulfill more than one pattern and thus may appear in more than one row. The right panel provides an expanded view of some response patterns. The adjusted categories are not shown in this representation. C. Frequency analysis of each of the basic and complementary response patterns. The response pattern name is indicated next to each panel. Within each panel, the blue histogram shows the shuffled distribution of pattern frequencies (n = 100,000 shuffles). The vertical blue lines show the mean value of the shuffled distribution. Red vertical lines indicate the actual number of observed patterns. Green shaded panels indicate significant cases (i.e., the estimated probability to obtain the observed number of cases by chance is less than 5%). In most cases, the probability is considerably lower (as indicated by the numbers within each panel). D. Like C for adjusted patterns.

Comparison of population level representations in estrus and non-estrus females.

A. Responses of individual neurons. Same data as in Fig. 2A, divided by reproductive states. B. Pairwise population responses distances in non-estrus (left) and estrus (right) using the correlation distance measure. Each pixel represents the distance between two columns within a given matrix from panel A. By definition, these matrices are symmetric around the diagonal. C. Same as B, using the Euclidean distance measure. D. Correlation of population level pairwise distances across the two reproductive states, using the correlation distance measure. E. Same as in D using the Euclidean distances. In both D and E, the correlation coefficient (CC) and the probability to obtain it by chance (p-value) are indicated within each panel.

Comparison of responses in estrus and non-estrus females.

A. Percentage of responding neurons to each of the stimuli in non-estrus (gray) and estrus (red) neurons. p-values correspond to the binominal exact test. B. Mean response magnitude to each stimulus under the two states. p-values correspond to non-parametric ANOVA. In both A and B, the Bonferroni adjusted p-value given 11 comparisons, at the 0.05 level, is 0.0045. The single significant difference is indicated in bold. In A and B, n = 305 non-estrus, and n = 241 estrus. C. Correlation between mean pairwise preference indices under the two reproductive states. The correlation coefficient and the probability to obtain it by chance are indicated. D. Comparison of pairwise preference indices under the two states. Each pair of bars corresponds to one comparison. The stimuli are indicated by the icons and text. For example, the first pair of bars on the left corresponds to pairwise comparison of dominant and naive ICR male urine. In both estrus states, there is a stronger response to the dominant stimulus as indicated by the downward pointing bars. However, the difference is not significant. E. Comparison of pairwise preference indices between stimulus pairs comprising male and female stimuli. In D and E, significance is determined using a shuffling test (n = 100,000). Bonferroni adjusted p-values at the 0.05 level are 0.0056 and 0.01 for D, and E, respectively. Significant differences among the states are indicated in bold. Note that in most cases, significant differences correspond to cases where responses during estrus are stronger to the less virile male stimulus.

Response selectivity under the two reproductive states.

A. Lifetime sparseness distributions for non-estrus (left, n = 305, mean: 0.29 median: 0.27), and estrus (right, n = 241, mean: 0.33, median: 0.31), neurons. Non-parametric ANOVA p-value: 0.0078, reflecting higher selectivity during estrus. B. Triangle plots showing relative response magnitude to stimuli from each of the three male states, separated according to strains. Neurons recorded in non-estrus and estrus females are indicated in gray and red, respectively. C. Distributions of the male state selectivity index (calculated from the data shown in B) under the two reproductive states, calculated for each strain separately. p-values corresponding to non-parametric analysis of variance (Kruskal Wallis test): (ICR: KW p: 0.0017, BC: KW p: 0.24, C57: KW p: 0.08). Higher selectivity during estrus is observed for the stimuli from ICR male mice (p = 0.0017). Mean selectivity scores: ICR: estrus: 0.52, n = 161, non-estrus: 0.42, n = 181. BC: estrus: 0.47, n = 153, non-estrus: 0.43, n = 181. C57: estrus: 0.37, n = 139, non-estrus: 0.41, n = 206.

Detailed comparison of selectivity for pairs of stimuli between estrus and non-estrus neurons (Related to Fig. 6).

p-values are calculated using a non-parametric ANOVA (kruskalwallis). Significant differences after adjustment for 23 comparisons (0.05/23 = 0.0022) are shown in red. Significant differences without adjustment (at the 0.05) are shown in blue. For each stimulus pair, values for the state with the higher selectivity are shown in bold font. The number of neurons in each comparison is indicated in parenthesis in the corresponding selectivity columns.

Comparison of response frequency under the two reproductive states.

A. Frequencies of basic patterns (including complementary patterns). B. Frequencies of adjusted patterns (including complementary patterns). In both panels, frequencies in estrus and non-estrus females are shown above and below the horizontal axis, respectively. Significantly represented categories (using a shuffling test as described in Fig. 3) are indicated in green. P-values for differences for each pattern under the two states, using the binomial exact test, are listed. The Bonferroni adjusted p-values at the 0.05 level are 0.0031 (0.05/16) and 0.0042 (0.05/12), for A and B, respectively, and thus none of the differences are significant. Relaxing the correction, three categories fulfil the 0.05 criterion (indicated in bold). None of them correspond to increased representation of dominant male stimuli during estrus. Note that in some cases, a category is overrepresented only during one of the states, but this is not necessarily reflected by a significant difference among the two states (e.g., ∼female).

Schematic illustration of sampling of trait selective molecules.

A. Molecular profiles associated with three individuals sharing a common trait. Trait selective molecules are shown in red and all others (unique molecules) are shown in green, within a 30 x 30 grid representing all (900) molecules in the world. The three individuals are indicated on the three columns on the left. The rightmost column shows their union. Rows correspond to different proportions of individual specific and trait selective molecules. B. Different sampling scenarios by AOB neurons. The three rows correspond to the scenarios shown in A. The left column is the union of all molecules under each scenario, as shown in A. The three other columns represent sampling schemes that differ in the proportion of trait selective vs. other molecules sampled, moving from trait avoiding, to balanced, to trait selective sampling. The two values above each column indicate the probability of sampling unique (left), or common (trait-selective) molecules (right). We speculate that sampling properties are matched to component frequencies, thereby balancing robustness of trait detection with redundancy. We note that receptive fields of AOB-MCs are actually determined by VSNs and by their interactions with AOB-MCs.