Sex-specific processing of social cues in the medial amygdala

Abstract

Animal-animal recognition within, and across species, is essential for predator avoidance and social interactions. Despite its essential role in orchestrating responses to animal cues, basic principles of information processing by the vomeronasal system are still unknown. The medial amygdala (MeA) occupies a central position in the vomeronasal pathway, upstream of hypothalamic centers dedicated to defensive and social responses. We have characterized sensory responses in the mouse MeA and uncovered emergent properties that shed new light onto the transformation of vomeronasal information into sex- and species-specific responses. In particular, we show that the MeA displays a degree of stimulus selectivity and a striking sexually dimorphic sensory representation that are not observed in the upstream relay of the accessory olfactory bulb (AOB). Furthermore, our results demonstrate that the development of sexually dimorphic circuits in the MeA requires steroid signaling near the time of puberty to organize the functional representation of sensory stimuli.DOI: http://dx.doi.org/10.7554/eLife.02743.001.

Keywords: behavior; medial amygdala; pheromones; sensory representation; sexual dimorphism; vomeronasal system.

Figure 1.. Experimental system for recording MeA sensory responses.

(A) Vomeronasal and olfactory structures are shown in yellow and blue, respectively. Multichannel electrophysiological probes are stereotaxically positioned in the MeA to continuously record neural responses to sensory stimulation. VNO stimulus presentation (orange arrow) is achieved by placing nonvolatile stimuli in the nostril followed by electrical stimulation of the sympathetic nerve trunk (SNT) to activate the VNO pump and permit access of stimuli into the VNO. VNO stimuli are washed out through the NPD. MOE stimulation is achieved by controlling airflow of volatile stimuli into the nostril (blue arrow), which access the MOE, and are eliminated through a tracheotomy. (B) Diagram illustrating a coronal section through the posterior MeA with red dots indicating the expected dorsal–ventral distribution of recording sites. A single fluorescent electrode tract accurately targeted to MeA is shown in the inset. (C) Timecourse of VNO and MOE stimulation trials (top). Electrophysiological signals recorded from a single MeA electrode during four successive trials reveal a well-isolated unit responding only to female stimuli following electrical stimulation of SNT (stimulation artifacts are evident at 20 s). (D) Percentage of MeA units responding to VNO vs MOE stimulation with chance rates indicated by a horizontal dashed line. (E) Sagittal section of whole brain showing DAPI staining and the site of MeA FluoroGold iontophoresis. (F) Dense retrograde labeling of AOB projection neurons. (G) Fraction of AOB (99.8%) and MOB (0.2%) neurons that are retrogradely labeled by FluoroGold iontophoresis in the MeA. DOI: http://dx.doi.org/10.7554/eLife.02743.003

Figure 1—figure supplement 1.. Clustering and analysis of multichannel electrophysiological recordings.

(A) Spike waveforms of four MeA units as recorded on two channels (channels on which signals from these units were not detected are not shown for simplicity). (B) Projections of spike waveforms on the first Principal components for each of the two channels shown in A. Even in this partial representation (projections on additional Principal Components are not shown) four clusters are clearly evident (same color scheme as in A). Initial clusters were generated by the KlustaKwik program and then adjusted manually. (C) Histograms showing the distribution of interspike intervals within each cluster. Units a and d were classified as ‘multi-units’ as they did not show a clear refractory period. Units b and c showed negligible refractory period violations (<4 ms) and were accordingly classified as ‘single units’. (D) The average spike waveform for all well-isolated MeA units. The lighter lines indicate STD. (E) Distribution of the spike maximum to spike minimum ratio for MeA units. (F) Distribution of spike widths (distance between the minimum and maximum) for all well-isolated MeA units. (G) Scatterplot showing the relationship between spike width and the min/max ratio. (H) Spike waveforms for four AOB units (as recorded on two channels). (I) Principal component projections for the four clusters depicted in H. (J) ISI histograms for the units shown in H and I. Unit b was classified as ‘multi-unit’ due to significant violations of the refractory period. The three other units (a, c, d) recorded by these channels were classified as single units. (K) Average spike waveform for all well-isolated AOB units, lighter lines indicate STD. (L) Distribution of spike maximum to spike minimum ratio for AOB units. (M) Distribution of spike widths for all well-isolated AOB units. (N) Scatterplot showing the relationship between spike width and the min/max ratio for AOB units. DOI: http://dx.doi.org/10.7554/eLife.02743.004

Figure 1—figure supplement 2.. Baseline electrophysiological characteristics of MeA responses.

(A) Firing rates (spikes/second) for the 20 s prior to VNO stimulation are plotted against the firing rate during the 40 s after stimulus application. Each point indicates an isolated single unit with the most significant (p<0.01, Nonparametric ANOVA) stimulus induced response plotted for the post stimulus firing rate. Nearly all units lie above the line of unity (black diagonal) indicating that VNO stimulation positively drives MeA activity. (B) Single MeA units were grouped based on the pre stimulus firing rate, and the average response selectivity is plotted for units within each of these groups. A clear correlation was observed with less active neurons responding more specifically (black line; R = −0.31: p<0.0001). DOI: http://dx.doi.org/10.7554/eLife.02743.005

Figure 1—figure supplement 3.. MOE-driven responses in the MOB and PLCO.

To ensure that the MOE stimulation used in our anesthetized preparation was capable of driving responses in olfactory areas, we recorded from the MOB and PLCO. The top three rows are single units recorded from the MOB and the bottom row is a unit recorded from olfactory amygdala. Each column indicates the presentation of a different stimulus. For each panel, raster plots indicating the timing of individual action potentials elicited by multiple presentations of the same stimulus (shaded region). Histograms of the mean response and standard error for the same data are shown below each raster plot. Time zero indicates alignment to the start of stimulus presentation. All units responded significantly to at least one stimulus (p<0.01; Nonparametric ANOVA). Gray boxes indicate that a particular stimulus was not presented during that unit's recording session. DOI: http://dx.doi.org/10.7554/eLife.02743.006

Figure 2.. MeA sensory responses to VNO stimuli.

(Left) Each row shows the responses elicited in a single MeA unit by four different VNO stimuli, with each stimulus presented 5 times. The order of stimulus presentation was randomized during the experiment, but is shown grouped by stimulus for clarity. (Right) Histograms showing the mean response and standard error (shaded region) for each unit. Responses were aligned to the onset of stimulus presentation. All significant responses (p<0.01; Nonparametric ANOVA) are indicated by an asterisk in the top right corner of the average histogram plots. Response magnitudes for each unit were normalized to the maximum response for all stimuli. Colored bars (top and bottom) indicate the 40 s epoch following stimulus presentation that was considered for all analyses. DOI: http://dx.doi.org/10.7554/eLife.02743.007

Figure 2—figure supplement 1.. MeA units responsive to different stimulus categories are spatially segregated.

(A) A single electrophysiological probe was positioned in the MeA of an adult male mouse (left) and 13 MeA units were simultaneously recorded on 8 evenly spaced recording sites (right: rows). Dashed lines indicate the recording sites/depth. The average responses to four stimuli (columns) are shown for each unit (asterisks indicate significant responses; bold asterisks indicate unit classification as ‘predator’ or ‘conspecific’). (B) The dorsoventral distance between pairs of neurons with different stimulus classifications (either ‘conspecific’ or ‘predator’) is shown for all simultaneously recorded pairs. Positive values indicate the conspecific unit was dorsal to the predator unit. Red, blue, and shaded purple curves represent data obtained from female, male, and all animals respectively with each curve skewed towards conspecific units being located dorsally (p<0.00001 for all distributions: signrank test; horizontal dashed line indicates the mean for all animals: 160, 95% CI = 123 to 189 µm). (C) The dorsoventral distance is plotted against the difference in stimulus specificity for all unit pairs. Regression analysis indicates a weak relationship between the dorsoventral distance and the difference in stimulus specificity for unit pairs (R² = 0.033, F = 8.6, p=0.003; black line; light gray points omitted due to high leverage). Therefore, the dorsoventral topography is largely a result of quantitative biases in the numbers of ‘conspecific’ and ‘predator’ neurons located dorsal vs ventral, while only 3% of the variance is explained by a smooth transition from ‘predator’ selective neurons (ventral) to ‘conspecific’ selective neurons (dorsal). DOI: http://dx.doi.org/10.7554/eLife.02743.008

Figure 2—figure supplement 2.. AOB sensory responses to VNO stimuli.

Each row (A-D) shows data obtained from a single well-isolated AOB unit, and each column indicates the stimulus presented (female, male, predator, Ringer's). For each panel, raster plots (top, shaded region) show the timing of spikes that occurred during each of 5–7 trials. Histograms of the mean spiking response are shown below each raster plot with standard error (shaded color region). (E) Sagittal section through the olfactory bulb. DAPI staining is shown in blue and the electrodes were coated with DiI (red) prior to insertion. DOI: http://dx.doi.org/10.7554/eLife.02743.009

Figure 3.. Decreased frequency and increased specificity of sensory responses in the MeA compared to the AOB.

(A) The percentage of single AOB units (dashed curves) and MeA units (solid curves) exhibiting statistically significant responses to male (blue), female (red), and predator (green) vomeronasal stimuli as the threshold for inclusion was varied from p<0 to p<0.2 (abscissa). The diagonal gray line indicates the predicted false positive rate. (B) Distribution of response selectivity (‘Materials and methods’) showing a shift towards higher specificity in MeA (solid line) as compared to the AOB (dashed line). (C) Selectivity of sensory responses for units recorded in the adult AOB (197 units from male and female animals). (D) Selectivity of sensory responses for units recorded in the adult MeA (274 units from male and female animals). Each point represents the response profile of an individual unit, with at least one significant response, to male, predator, and/or female stimuli. Points located near a vertex (more frequent in the MeA) represent units that respond most strongly for the stimulus indicated at that vertex whereas points at the center (more frequent in the AOB) represent units that respond similarly to all stimuli. Insets (C and D) show correlation between responses for each pair of stimuli. DOI: http://dx.doi.org/10.7554/eLife.02743.010

Figure 3—figure supplement 1.. Leverage analysis for stimulus response correlations.

(Top row) Pseudocolor plots of correlation values with all data points included. (Middle row) Scatter plot of stimulus 1 (abscissa) against stimulus 2 (ordinate) for all comparisons. High leverage points (lev >2*Rn/N; where Rn = # regression parameters and N = the size of the data set) are shown in red. (Bottom row) Pseudocolor plots of correlation values with high leverage data points omitted. Each column represents a different data set (adult AOB, adult MeA, juvenile MeA). The similarity between correlations with or without high leverage points included (A to G; B to H; C to I) indicates that the correlation is not driven by only a few neurons, but rather is a property of the population. DOI: http://dx.doi.org/10.7554/eLife.02743.011

Figure 3—figure supplement 2.. Principal components analysis of MeA categorization data.

(A top) Principal component scores for principal components 1–5. (A bottom) For each matrix: rows indicate the normalized response of a single MeA unit to ten different stimuli drawn from three behaviorally relevant categories male (balbC M, C57 M, CBA M), female (balbC F, C57 F, CBA F), predator (fox, rat, bobcat), and a Ringer's control. Columns indicate the stimulus presented, and vertical white lines indicate category boundaries. Each matrix shows the same data set sorted by the five largest principle components. Together, these components account for over 90% of the variance in sensory responses observed in our recordings. (B) Pairwise correlation of MeA responses, for all responding units, elicited by VNO stimulation. Positive correlations between stimulus-induced responses are indicated by red and anti-correlations are indicated by blue. Within-category correlations are generally positive and between-category correlations are generally negative. (C) Multidimensional scaling analysis shows that stimuli from a single category populate a region of the response space that is non-overlapping with stimuli drawn from a different category. Lines connecting individual data points indicate ethologically defined stimulus categories. DOI: http://dx.doi.org/10.7554/eLife.02743.012

Figure 4.

Sexual dimorphism of adult MeA responses. (A) Responses of AOB neurons to vomeronasal stimuli in adult male (210 units) and female (64 units) mice. (B) Responses of MeA neurons to vomeronasal stimuli in adult male (106 units) and female (91 units) mice. (C) Responses of MeA neurons to vomeronasal stimuli in juvenile male (37 units) and female (50 units) mice. Units shown in panels A–C are the same data shown in Figure 3C,D but classified according to the sex of the animal recorded. Blue circles indicate units recorded from male mice and red squares indicate data recorded from female mice. (D–F) Sex-specificity (‘Materials and methods’) histograms are shown for all units recorded from male (blue) and female (red) animals in the adult AOB (D), adult MeA (E) and juvenile MeA (F). Horizontal lines (above) indicate the mean and 95% confidence interval (bootstrap CI) of the mean for each distribution. Data collected from males vs females were only different in the adult MeA (AOB: p=0.26 adult MeA: p<0.00001; juvenile MeA: p=0.18; permutation tests). DOI: http://dx.doi.org/10.7554/eLife.02743.013

Figure 4—figure supplement 1.. Sexual dimorphism in the dominance of predator versus conspecific responses.

A predator/conspecific index was calculated as: (R_pred-R_conspecific)/(R_pred + R_conspecific), where RP is the response to predator stimuli and RC is the stronger of the responses to male or female stimuli. Histograms show predator/conspecific index for units recorded from male (blue) and female (red) animals in the adult AOB (A), adult MeA (B), and juvenile MeA (C). Horizontal lines above histograms indicate the bootstrapped 95% confidence interval for the mean. Only data from the MeA of adult males and females exhibited a clear difference (AOB: p=0.35; adult MeA: p<0.001; juvenile MeA: p=0.42; permutation tests). DOI: http://dx.doi.org/10.7554/eLife.02743.014

Figure 5.. Importance of aromatase signaling for the development of sexually dimorphic MeA responses.

(A) Population summary of MeA responses to male, female, and predator stimuli recorded from adult male ArKO mice (light blue circles) or heterozygous male littermates (dark blue diamonds). All plotted units responded significantly to at least one stimulus. (B) Sex-specificity histograms for units recorded from ArKO males (blue fill), heterozygous male littermates (dark blue; no fill), and wild-type juvenile males (light blue; no fill). Horizontal lines indicate the mean and 95% confidence interval of the mean of each distribution. (C) Matrices of correlation for the population responses between pairs of sensory stimuli for heterozygous males (top) and ArKO males (bottom). DOI: http://dx.doi.org/10.7554/eLife.02743.015

Figure 6.. Estrogen influences the development of sexually dimorphic MeA responses.

(A) Population summary of MeA responses to male, female, and predator stimuli recorded from adult untreated female mice (red squares) or estrogen treated adult females (dark red triangles). (B) Sex-specificity histograms for units recorded from estrogen-treated adult females (dark red fill), adult untreated females (red; no fill), and untreated adult males (blue; no fill) for comparison. Estrogen treatment reduced responses in adult female mice to male stimuli (leftward) and increased the frequency of responses to same-sex stimuli (rightward). Horizontal lines indicate the bootstrapped 95% confidence interval for the mean of each distribution. (C) Comparison of the dorsal-ventral locations of male vs predator responsive units (blue) and female vs predator responsive units (red; see Figure 2—figure supplement 1) in the MeA of estrogen treated females. Both male responsive and female responsive units are dorsal to predator responsive units. DOI: http://dx.doi.org/10.7554/eLife.02743.016