Experimental C57BL/6J female and male mice aged 8 weeks were group housed with five mice of the same sex in each cage (n = 10). All cages were maintained on a 12-h light/dark cycle (lights on 7:00 am), with food and water ad libitum. All surgical and behavioral experiments were performed during light hours and began when mice were aged 10 weeks. Aggressive retired male breeder CD1 mice from Charles River Laboratories were used for all social defeat experiments and nonaggressive retired male breeder CD1 mice were used for social interaction testing. All experiments were conducted in compliance with National Institutes of Health Guidelines for Care and Use of Experimental Animals and approved by the Institutional Animal Care and Use Committees at Columbia University Irving Medical Center and the New York State Psychiatric Institute.

Mice were injected with the helper virus AAV8-CMV-TVA-mCherry-2A-oG (0.5–0.9 μL, Salk Institute California, titer of 1.28 × 10⁸ TU/mL) into the BLA bilaterally (from bregma: AP −1.03, ML ±3.34, DV −4.89, angle 0°) under a combined isoflurane anesthetic (2–5%, SomnoSuite, Kent Scientific) and meloxicam analgesic (5 mg/kg) protocol (Figure 1A). Mice were allowed to recover and fully express the virally delivered mCherry-tagged TVA and G-protein complexes for 3 weeks. GFP-tagged glycoprotein-deleted pseudorabies virus, AAV8-RVdG-eGFP (0.5–0.9 μL, Salk Institute California, titer of 2.72 × 10⁸ TU/mL), was then injected into the BLA of each hemisphere (from bregma: AP −1.03 mm, ML ±4.43 mm, DV −5.09 mm, angle 11°) (Figure 1B) and mice were given a week prior to behavioral testing to allow for full recovery and expression of the GFP-tagged glycoprotein-deleted pseudorabies virus. Post-hoc histological confirmation of BLA-centered targeting was performed on all brains (Figure 2C) and only those meeting viral localization were included in further analysis.

One hour before beginning ASDS, urine was collected from nonaggressive CD1 mice and thoroughly applied (10–25 μL) to the urogenital region of all female C57 mice as previously described (Harris et al., 2018) to increase the propensity of CD1 aggression upon female mice (Mugford and Nowell, 1970; Chamero et al., 2007). Each male C57 was placed into the home cage of a CD1 aggressor for 2 min. This was repeated three times sequentially with three different CD1 aggressors, with no rest periods. Experimental female mice were maintained in group housing in the experimental room during the CD1-on-male C57 aggression and were then each individually exposed to acute stress in an identical paradigm as to which the males were exposed (Figure 1C). As previously reported (Grossman et al., 2022), all mice were singly housed for the remainder of the hour and tested for SI 60 min following the onset of ASDS. For this experiment, no control C57 mice were included, as the focus of the study was individual variability in behavioral response.

Precisely 1 h after the onset of the first defeat in ASDS, C57BL/6J mice were placed into an open-field arena (45 cm × 45 cm) with a removable wire-mesh enclosure secured in clear Plexiglas (10 cm W × 6 cm D × 30 cm H) placed against the middle of one of the inner walls of the arena. Mice were allowed to habituate to the arena for 2.5 min. After 2.5 min, a novel non-aggressive CD1 “target” mouse was placed in the wire-mesh enclosure. The C57BL/6J mouse was then allowed to explore for another 2.5 min (Figure 1D). Open-field arenas and wire-mesh enclosures were wiped with a cleaning solution between tests. All tests were conducted during the lights on period (between 13:00 and 19:00) under infrared light conditions. A video-tracking system (Ethovision XT 15, Noldus Information Technology) was used to record the movements of the C57 mouse during the “target absent” and “target present” periods. The total time spent in the “interaction zone,” defined as a 24 cm × 8 cm corridor surrounding the wire-mesh enclosure, and corner zones, 9 cm × 9 cm in the two corners opposite from the interaction zone, was recorded. The SI ratio was calculated by the absolute time spent in the interaction zone (blue, Figure 1D) with the target present divided by the absolute time spent in the interaction zone with the target absent. The corner ratio was calculated analogously, as the absolute time spent in the corner zone (red, Figure 1D) with the target present divided by the absolute time spent in the corner zone with the target absent (Grossman et al., 2022), and was then inverted for enhanced comparability with SI scores (i.e., lower scores predict social avoidance, and higher scores predict social approach).

Mice were anesthetized with isoflurane, checked by toe pinch, and sacrificed by transcardial perfusion with 4% paraformaldehyde immediately after SI testing. Each perfusion started one-hour post-ASDS to support literature findings of peak cFos protein levels occurring between 60 and 90 min after cellular activation (Zinck et al., 1993; Skar et al., 1994; Li et al., 1997). The 60-min mark was chosen to ensure the capture of activated neurons was most time-locked to the first exposure to the social stressor. Perfusions were performed at a flow rate of 9 mL/min for 10 min, followed by decapitation and submandibular cuts for increased fixation. Brains were post-fixed for 72 h, deskulled, and sectioned at 100 μm using a vibratome (speed 8.5, frequency 9; Leica Biosystems), at which point the left hemisphere was marked with a “nick” for hemisphere-specific analysis of injection accuracy. Sectioned brains were stored in 0.9 M PBS with 0.5% sodium azide until immunohistochemical analysis was performed. Sections were serially stained with rabbit anti-cFos (9F6) primary monoclonal antibody (Cell Signaling, catalog number: 2250S, lot number: 12, 1:2,000 in 0.3% PBST + 5% Normal Goat Serum + 1% BSA + 0.2% Cold Water Fish Skin Gelatin block solution) and Alexa Fluor-conjugated AffiniPure goat anti-rabbit secondary antibody (Alexa Fluor-647, Sigma, code number: 111-605-033, lot number: 157045, 1:500 in 0.3% PBS-T). All sections were incubated in 4′,6-diamidino-2-phenylindole (DAPI) for histological identification.

A Nikon Eclipse Ti2-E Motorized inverted microscope and the Element Acquisition Software were used for all imaging. Initially, all coronal sections along the wide brain were scanned for long-range projections, but upon identification of only local projections, sequential amygdala sections were imaged for analysis. For histological confirmation of injection sites, low resolution (4×) 16-bit images of sections spanning the entire anterior-posterior extent of the amygdala were inspected. Triple or quadruple channel z-stacks (stacks of 2 μm thickness spanning a total of 20 μm) were then acquired to image GFP, mCherry, CY5, and sometimes DAPI signal and the EDF function of the Element Software applied to collapse the z-stack. For projection analysis of cFos, GFP-tagged Rabies, and mCherry-tagged TVA-G cells (Figure 2A), higher resolution (10×, and 40×) 16-bit images were acquired at the amygdala (Figure 2B). Further analysis was done on two selected sections for each animal, one corresponding to the anterior part of the amygdala complex (AP −1.2 to −1.5 from bregma) and one to the more posterior part (AP −1.8 to −2.1 from bregma). Due to the strength of Cy5, the laser power was tuned as low as possible to detect cFos⁺ cells, to minimize leakage from the mCherry fluorescent reporter. Cross comparison was also performed on mCherry-tagged cells dorsal to the amygdala that were infected along the injection line, but that were expected to be cFos⁻ following social stress and it was confirmed that these cells were not detected in the Cy5 channel. Notably, these cells did not receive GFP-Rabies virus, due to the multi-angle injection approach, and were thus not confounding starter cell populations, allowing for the use of this mild off-target infection as a confirmation of Cy5 filter specificity. The nd2 files acquired were imported in Fiji (Schindelin et al., 2012) and modified for counting and reconstruction analysis. Each split channel was smoothed and a seven pixels BG subtraction applied. Before proceeding to counting, non-rigid free-form registration of the selected 100 μm sections was performed on the GFP channel with the open-source WholeBrain Software (Fürth et al., 2018) in RStudio. Briefly, each image was first matched to a corresponding plate of the Mouse Allen Brain Reference Atlas CCF v3 (Wang et al., 2020). The contour of the brain section was segmented out using the autofluorescence of the down-sampled section itself, and a set of correspondence points that align the selected reference atlas plate with the tissue-section was automatically generated by the “thin-plate splines algorithm.” Landmarks such as ventricles were used to manually adjust, add, or remove correspondent points when necessary. The two sets of reference points (atlas and tissue-section reference points) constitute a 2D final mapping of each brain region along the antero-posterior axis. The extracted warped contours of the amygdala complex were exported from RStudio and overlapped to the original 16-bit images in ImageJ (Schneider et al., 2012) with a custom-made macro (Mahadevia et al., 2021). GFP Rabies⁺ cells were counted after thresholding the resulting images (analyze particle plugin). CY5 cFos⁺ cells were counted after applying a white top hat to the images (MorphoLibJ plugin). For colocalization analysis of starter cells (GFP Rabies⁺ mCherry TVA-G⁺) and stress-activated projection cells (GFP Rabies⁺ CY5 cFos⁺), the image calculator was implemented using the AND function. Method validation was performed with the coloc function. The schematic of projections and starter cells for the left and right hemispheres (Figures 2C,E) were reconstructed in ImageJ (Schneider et al., 2012) by overlapping and realigning in a stack the segmented images for the channel of interest over the representative plate from the Paxinos Mouse Brain Atlas 4th edition (Paxinos and Franklin, 2012) in Adobe Illustrator V for final color modifications. All statistical analyses including correlations and FDR corrections were performed in excel using the built-in statistical functions.

Following adequate recovery time from stereotactic viral injections (Figures 1A,B), mice were exposed to our previously developed ASDS paradigm (Grossman et al., 2022), which consists of three consecutive two-minute bouts of social stress (Figure 1C) and was here extended to include female mice. After 1 h of rest, the experimental mice underwent an SI test as a behavioral readout of individual variability in behavioral response to stress, measured as time spent in the interaction zone, near a novel “target” mouse of the same breed as the aggressor, and the time spent in the corner zones (Figure 1D). Given the explicit readout of individual variability in behavioral response to stress, all mice underwent the social stress paradigm and unstressed controls were not included. ASDS elicited a spectrum of behavioral responses ranging from social avoidance to social approach in both female (n = 5) and male (n = 5) mice (Figures 1E–H). Female and male mice showed no statistically significant difference in time spent in the interaction zone (repeated measures ANOVA: F vs. M, p = 0.135; phase, p = 0.0863; interaction, p = 0.522) (Figure 1E), corresponding to SI ratio (t-test: p = 0.6000) (Figure 1G), or time spent in the corner zone (repeated measures ANOVA: F vs. M, p = 0.198; phase, p = 0.0456; interaction, p = 0.899) (Figure 1F), corresponding to inverse corner ratios (t-test: p = 0.4566) (Figure 1H). These metrics provide a justification for concatenating female and male data for a total sample size of 10 animals that could be used for analyses exploring neural correlates of individual variability in behavioral stress responses.

Post-hoc histological confirmation of the injection site was performed on all brains and overlaid with DAPI for regional identification of the amygdala. All 10 animals demonstrated injection endpoints of both the mCherry-tagged TVA and G-protein containing viral complex and the GFP-tagged pseudotyped G-deleted rabies virus (Figure 2C) within the bilateral amygdala complex. After thorough optimization of immunohistochemical conditions, consistent cFos fluorescence was observed in all 10 animal samples. Since all animals met viral localization criteria satisfactorily (Grossman et al., 2022), no subjects were excluded from analysis. Two amygdala sections from each brain were imaged in detail, one from the more anterior portion of the amygdala complex (AP −1.2 to −1.5 from bregma) and one from the more posterior end of the amygdala complex (AP −1.8 to −2.1 from bregma) (example in Figure 2B). Since there were no identifiable differences in hemispheric projections, projections from both hemispheres were analyzed in all animals. Starter cells (GFP⁺mCherry⁺) were identified across the amygdala complex as well as surrounding cortical regions in a total of 19 subregions, and starter cell location and total number was not predictive of the behavioral phenotype (Figure 2D). Eight of the 13 amygdala subregions also contained projection cells (GFP⁺mCherry⁻), so further analyses were restricted to these eight subregions. By controlling for the individual variability in the infection rate using total starter cell number per animal (GFP⁺mCherry⁺), structural (not responsive to stress, GFP⁺cFos⁻) and, functional (stress-activate, GFP⁺cFos⁺) projection cell counts, as well as alternately activated cFos expression (GFP-cFos+) compared to behavioral response were compared across these eight amygdala-associated subregions (Figures 2E–G) No correlations between structural or functional projection cell number and SI scores in any of the eight amygdala subregions reached FDR-corrected significance threshold, though we propose this methodological approach for future investigation.