A dedicated hypothalamic oxytocin circuit controls aversive social learning

Summary

To survive in a complex social group, one needs to know who to approach but, more importantly, who to avoid. A single defeat causes the losing mouse to stay away from the winner for weeks.¹ Here, we identify oxytocin neurons in the retrochiasmatic supraoptic nucleus (SOR^OXT) and oxytocin receptor-expressing cells in the anterior subdivision of the ventromedial hypothalamus, ventrolateral part (aVMHvl^OXTR) as a key circuit motif for defeat-induced social avoidance through a series of functional manipulation and recording experiments. Before defeat, aVMHvl^OXTR cells respond minimally to aggressor cues. During defeat, aVMHvl^OXTR cells are highly activated and, with the help of a private oxytocin supply from the SOR, potentiate their responses to aggressor cues. After the defeat, the strong aggressor-induced aVMHvl^OXTR cell activation drives the animal to avoid the aggressor and minimizes future defeat. Our study uncovers the neural process supporting the rapid social learning caused by defeat and highlights the importance of the brain oxytocin system in social plasticity.

Introduction

Fighting is a major means to compete for limited resources in the wild. After the fighting ends, typically by the retreat of the loser, the traumatic defeat experience is clearly remembered. The loser continuously avoids close interaction with the winner and readily flight away when confronted. In male mice, a single 10-minute defeat can induce avoidance of the winner for 15 days¹. The defeat-induced avoidance is observed across species, including humans². In the US, a quarter of teenagers reportedly experienced bullying and showed increased social isolation and school avoidance³.

The neural mechanisms underlying the fast and long-lasting behavior changes induced by defeat remain incompletely understood. Early studies focused on conditioned defeat⁴ in male hamsters and concluded that defeat and non-social aversive experiences, e.g., foot shock, utilize the same brain circuit, including the prefrontal cortex, basolateral amygdala, and hippocampus, for associative fear learning^5–7. Recently, several studies have suggested a potential role for the VMHvl in social defense and avoidance^8–11. VMHvl is a part of the social behavior network and is highly activated by conspecific cues^12,13. Sakurai et al. captured VMHvl cells activated during defeat and found that their re-activation elicited fear responses towards a benign conspecific⁸. Conversely, inactivating the VMHvl and its surrounding area reduces social avoidance of the aggressor one day after defeat⁹. Our previous study further revealed functional heterogeneity within the VMHvl: anterior VMHvl (aVMHvl) is preferentially activated during defeat, while the posterior VMHvl (pVMHvl) is most activated during attack¹⁰. Optogenetically activating aVMHvl cells elicits freezing, upright postures, and avoidance of a conspecific, while activating the pVMHvl cells elicits approach, social investigation, and attack^10,13. These studies support the role of the aVMHvl in social avoidance and fear. However, whether the aVMHvl mediates defeat-induced behavior changes and, if so, how remains unknown. Here, we investigated this question using a series of recording, functional, and molecular tools and found that aVMHvl cells expressing oxytocin receptor (VMHvl^OXTR) undergo dramatic changes during defeat with the help of a private source of oxytocin to mediate defeat-induced social avoidance and fear.

Results

One day defeat induces avoidance and fear

We used the social interaction (SI) test¹⁴ to characterize the behavior changes induced by defeat. During SI test, a C57BL/6 (C57) male or female mouse freely interacted with a cupped aggressive Swiss Webster (SW) male or lactating female mouse for 10 min (ED Fig. 1a-b). We performed the SI tests one day before and after a 10-min resident-intruder (RI) test during which the C57 test mouse was introduced into the home cage (HC) of the aggressor, the same as the one in SI tests, for 10 min (ED Fig. 1c). During the RI test, SW aggressors attacked and defeated the C57 intruders quickly and repeatedly (ED Fig. 1d-e). After several bouts of defeat, C57 intruders spent more time immobile (body center velocity < 1 pixel/frame) in the corner (ED Fig. 1f-g). During the pre-defeat SI test, C57 males repeatedly approached and investigated the cupped aggressor and spent approximately half the time around the cup (ED Fig. 1h-m). After defeat, the animal spent most of the time staying at the end far from the aggressor and less time investigating the aggressor due to combined reduced approach frequency and shorter investigation duration per visit (ED Fig. 1h-o). Additionally, when the C57 was far from the aggressor, it significantly reduced its movement velocity, i.e., freezing more (ED Fig. 1p-q). Thus, a single 10-min defeat is sufficient to induce both social avoidance—measured by the reduced interaction time with the SW, and social fear—measured by the reduced movement velocity when the aggressor is far away. The defeat-induced behavior change was qualitatively similar between males and females, although the extent of avoidance was lower in females than in males (ED Fig. 1k-q).

To address whether the avoidance is specific to the same SW aggressor, we examined the behavior towards an SW aggressor different from one that defeated the test mice (ED Fig. 1r). The defeated animal also reduced time around the cupped novel aggressor and decreased movement velocity when away from the cup. However, the decrease level was lower than when encountering the same SW aggressor (ED Fig. 1s-x). To understand whether the defeat-induced avoidance is generalizable to mice with genetic backgrounds different from the aggressor, we compared the behaviors of the test C57 towards the same SW aggressors, Balb/C (BC) non-aggressors and unfamiliar C57 mice using a multi-animal social interaction (MSI) test (ED Fig. 2a-b). One day after defeat, the C57 test males spent significantly less time investigating or around the cupped SW aggressor and approached the aggressor fewer times (ED Fig. 2c, e-g). In contrast, the interaction between the defeated C57 and an unfamiliar C57 or a previously encountered BC remained unchanged (ED Fig. 2c, e-g). We observed qualitatively similar results in C57 female mice during the MSI test (ED Fig. 2d, h-j). Thus, 10-min defeat induced avoidance of winner-like conspecifics.

Winner cues drive loser aVMHvl^OXTR cells

We previously showed higher c-Fos expression in the aVMHvl after defeat than winning¹⁰. Nasanbuyan et al. reported that defeat-induced c-Fos overlaps with OXTR at the VMHvl¹⁵. We thus examined defeat-induced c-Fos in OXTR^Cre:Ai6 (OXTR^ZsGreen) mice and found that defeat induced more OXTR and c-Fos double-positive cells (OXTR+Fos+) in the aVMHvl (Bregma: −1.34 mm to −1.50 mm) than attack while these two behaviors induced a similar number of OXTR+Fos+ cells in the pVMHvl (Bregma: −1.66 mm to −1.82 mm) (ED Fig. 3a-b). Compared to estrogen receptor alpha (Esr1), a gene marker for aggression-related VMHvl cells^16,17, OXTR is expressed more abundantly in the aVMHvl (ED Fig. 3c-d). Approximately 10% of aVMHvl^OXTR cells express Esr1, whereas the overlap increases to 30% in the pVMHvl (ED Fig. 3e-f).

Fiber photometry recording of GCaMP6f expressing aVMHvl^OXTR cells in male mice (OXTR^GCaMP6) further revealed low cell activity during investigating or attacking a non-aggressive BC male intruder despite a large introduction response (ED Fig. 4a-c, f-j). In contrast, when the test mouse fought and was defeated by an aggressive SW mouse, in either the test mice or the SW aggressors’ cages, aVMHvl^OXTR cells strongly increased activity (ED Fig. 4d-e, h-j). Overall, aVMHvl^OXTR cells responded strongly during defeat but not during winning or social investigation (ED Fig. 4h-j). Female aVMHvl^OXTR cells were also highly excited during defeat by a lactating SW female while showing no activity change during non-agonistic interactions with naïve C57 females (ED Fig. 5a-j).

After defeat, as expected, OXTR^GCaMP6 male mice decreased interaction with the cupped aggressor during the SI test (ED Fig. 4k-o). Strikingly, aVMHvl^OXTR cells showed a drastic increase in response to the cupped aggressor (ED Fig. 4p-r). Furthermore, the avoidance level and the post-defeat response increase were significantly and positively correlated (ED Fig. 4s). During the post-defeat SI test, the test animal often quickly retreated from the cupped aggressor after investigation. At the retreat onset, the Ca²⁺ signal reached the maximum and gradually decreased during the retreat (ED Fig. 4t-v). Contrastingly, when the animal stayed immobile far from the aggressor, Ca²⁺ activity remained low (ED Fig. 4w-y). Similar response patterns were observed in female mice (ED Fig. 5k-v). These results suggest that aVMHvl^OXTR cells may drive social avoidance but likely not immobility.

This defeat-induced response increase was winner-specific as responses to previously encountered BC or unfamiliar C57 male mice did not change in the post-defeat MSI test (Fig. 1a-h, Supplementary Video 1). Again, there was a significant correlation between post-defeat avoidance levels and changes in cell responses (Fig. 1i). Similarly, female mice strongly avoided the lactating SW females while increasing interaction time with the C57 females in the post-defeat MSI test (ED Fig. 6a-e). The response towards the SW mothers, but not C57 females, increased after defeat, and the response increase and social investigation time decrease were significantly correlated (ED Fig. 6f-i).

Fig. 1. Male aVMHvl^OXTR cells increase responses to aggressors after defeat.

a. Schematics and representative histology. Scale bar: 200 μm. The brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline and behavior test illustrations.

c. A recording mouse’s body center distribution in pre- and post-defeat MSI tests.

d. Time spent around each cup during pre- and post-defeat MSI tests.

e-f. Representative GCaMP6f traces during pre- (e) and post-defeat (f) MSI tests.

g. PETHs of GCaMP6f signals aligned to different cup investigations during pre- and post-defeat MSI tests.

h. Average GCaMP6f signal during cup investigation in the pre- and post-defeat MSI tests.

i. Z-scored GCaMP6f signal change (post-defeat – pre-defeat) and investigation time change of all stimuli are significantly correlated.

j. Slice recording timeline.

k. Recording site and a recorded aVMHvl OXTR^ZsGreen cell. IR-DIC: Infrared differential interference contrast. Right shows the enlarged boxed area. Scale bars: 500 μm (left) and 50 μm (right).

l. Representative sEPSCs from various groups.

m-n. Amplitude (m) and frequency (n) of sEPSCs from various groups. Cells are from 3–4 mice/group.

o-q. sIPSC results, following plotting conventions in l-n.

Shades and error bars represent ± SEM. Circles and lines in d, h, and i represent individual animals. Circles in m, n, p and q represent individual cells. Numbers on the plots indicate the number of animals (d, g-i) or cells (m, n, p, q). Kruskal-Wallis test with Dunn’s multiple comparisons (m, n, p); One-way ANOVA with Tukey’s multiple comparisons (q); Two-way repeated-measure ANOVA with Sidak’s multiple comparisons (d, h); Pearson cross-correlation (i). All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Our in vivo recording result is also consistent with the c-Fos expression pattern. In male mice that the aggressor defeated for two days, aVMHvl^OXTR cells, but not pVMHvl^OXTR cells, expressed more c-Fos after interaction with the cupped aggressor for 10 minutes compared to mice that interacted with the cupped aggressor for two days (ED Fig. 7).

To understand the physiological and synaptic changes underlying the in vivo response change, we performed patch-clamp recordings of aVMHvl^OXTR cells in brain slices from OXTR^Cre:Ai6 male mice that experienced a 10-min defeat (D), non-agonistic social interaction (SI), or no intruder (SH) one day before the recording (Fig. 1j-k). We found that the magnitude of spontaneous excitatory postsynaptic current (sEPSC) of aVMHvl^OXTR cells in the D males was significantly higher than that in SH males, while SI animals showed the opposite change (Fig. 1l-m). The sEPSC frequency did not differ across groups (Fig. 1n). Spontaneous inhibitory postsynaptic currents (sIPSCs) also increased slightly but significantly in magnitude, but not frequency, in defeated animals, compared to other groups (Fig. 1o-q). Parameters reflecting intrinsic cell properties, including the current-frequency (I-F) curve, resting membrane potential, rheobase, and input resistance, did not differ significantly between D and SH males (ED Fig. 8). However, cells in D males appeared more excitable than those in SI animals due to their opposite trends of changes (ED Fig. 8b). These data suggest that potentiation of the excitatory synapses onto the aVMHvl^OXTR cells is likely the main contributor of increased in vivo cell responses to the aggressor after defeat.

aVMHvl^OXTR response is socially specific

We next asked whether aVMHvl^OXTR cells are activated only by aversive social cues or aversive cues in general. First, we recorded aVMHvl^OXTR cell Ca²⁺ activity during the investigation of 1% 2-methyl-2-thiazoline (2MT), an analog of predator odor component 2,4,5-trimethyl-3-thiazoline (TMT) that is highly aversive to mice¹⁸ (ED Fig. 9a-j). For comparison, we defeated the recording male mice one day before the test and recorded the response to the cupped aggressor in the same session. While Ca²⁺ signal consistently and robustly increased during aggressor investigation, we observed no activity increase during 2MT investigation although 2MT elicited strong avoidance and fear-like behaviors at least comparable to those induced by the aggressor (ED Fig. 9h-j).

Next, we employed an olfactory fear conditioning paradigm to pair a neutral odor (Pentyl acetate or (R)-(+)-limonene) with 1 mA foot shock (ED Fig. 9k-l). Control odor was delivered with no shock. One day after training, the animals showed effective aversive learning as they reduced movement velocity when the paired odor but not the unpaired odor was delivered (ED Fig. 9m-n). Throughout the post-conditioning odor test, aVMHvl^OXTR cells showed little Ca²⁺ activity fluctuation, and the average GCaMP6 signal did not differ among pre-odor, shock-unpaired, and -paired odor delivery periods (ED Fig. 9o-p).

As it was difficult to determine the exact moment when the odor reached the test animal, we also recorded aVMHvl^OXTR cell responses to shock-unpaired and -paired odors, as well as 2MT, delivered directly to the animal on a Q-tip using a head-fixed preparation (ED Fig. 10a). We again found no change or suppressed aVMHvl^OXTR cell Ca²⁺ activity during all non-social aversive odor presentations (ED Fig. 10b-d). In particular, 2MT caused a decreased cell activity for at least 30 sec beyond the odor delivery (ED Fig. 10b3, c3). In contrast, SW aggressor urine significantly increased aVMHvl^OXTR cell activity in one-day defeated test mice (ED Fig. 10b2, c2). These results suggest that the aVMHvl^OXTR cells are activated by aversive social, but not non-social, olfactory cues.

Avoidance expression requires aVMHvl^OXTR

What is the increased aVMHvl^OXTR cell response after defeat good for? To address this question, we optogenetically activated aVMHvl^OXTR cells (OXTR^ChR2) in naïve animals during SI tests (Fig. 2a-b). Control animals expressed GFP in aVMHvl^OXTR cells (OXTR^GFP). Upon light delivery, OXTR^ChR2, but not OXTR^GFP animals, strongly avoided the cupped animal as indicated by the significantly increased distance from the cup, reduced investigation time, and approach frequency (Fig. 2c-e, Supplementary Video 2). aVMHvl^OXTR cell activation also elicited social fear as indicated by significantly decreased movement velocity (Fig. 2f). The stimulation-induced avoidance and fear are not aggressive mouse-specific as similar behaviors were induced when the cupped animal was a non-aggressive BC male (ED Fig. 11a-f). In the absence of any target, the stimulated animals showed interleaved freeze and flight as indicated by increased time spent staying immobile (velocity < 1 pixel/frame) and moving fast (velocity > 20 pixel/frame) (ED Fig. 11g-k). In the real-time place preference (RTPP) test, the animal spent significantly less time in the light-paired chamber, indicating the aversive nature of aVMHvl^OXTR cell activation (ED Fig. 11l-n). These results suggest that the increased aVMHvl^OXTR response to the winner after defeat is functionally relevant as high activity of the cells drives social avoidance and fear and evokes a negative emotional state.

Fig. 2. aVMHvl^OXTR cells bi-directionally modulate social avoidance.

a. Virus schematics and representative histology. Scale bar: 200 μm.

b. Timeline and behavior assay illustration.

c-f. Average distance to the cupped aggressor (c), percentage of time spent investigating the aggressor (d), approach frequency (e), and median body center velocity (f) during ON (blue shades) and OFF periods in GFP and ChR2 groups. Statistical results are for ChR2 group. All p> 0.05 for GFP group.

g. Schematics and representative histology. Scale bar: 200 μm.

h. Timeline and light delivery protocol.

i. Distribution of the distance between the test animal’s body center and cupped aggressor during pre- and post-defeat SI tests for Gt-OFF (i1) and Gt-ON (i2) groups.

j-l. Change index of the investigation time percentage (j), around cup time percentage (k), and the median velocity when far from the aggressor (l) during SI tests for various groups. Change index: (P_after – P_before) / (P_after + P_before). P_before and P_after: behavior performance during pre- and post-defeat SI tests, respectively.

m-r. Behavior changes in post-defeat SI tests after chemogenetic inhibition of aVMHvl^OXTR cells. Plots follow conventions in (g-l). Scale bar in (m): 200 μm.

Shades and error bars represent ± SEM. Circles represent individual animals. Numbers on the plots indicate the number of animals. Two-way repeated-measure ANOVA with Sidak’s multiple comparisons (c-f), Kruskal-Wallis test with Dunn’s multiple comparisons (k, q), and One-way ANOVA with Tukey’s multiple comparisons (j, l, p, r). All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001 and ***p<0.0001. Illustration in (a, g, m) is based on a brain atlas from https://atlas.brain-map.org/. See Supplementary Table 1 for detailed statistics.

To understand whether the increased activity of aVMHvl^OXTR cells after defeat is necessary for the behavior change, we optogenetically inhibited male aVMHvl^OXTR cells using stGtACR¹⁹ (OXTR^Gt-ON) during the post-defeat SI test (Fig. 2g-h). Three control groups of animals were also tested, including OXTR^Gt-OFF (expressing stGtACR2 and receiving no light), OXTR^mCh-ON (expressing mCherry and receiving light), and OXTR^mCh-OFF (expressing mCherry and receiving no light) (Fig. 2g). Compared to control animals, OXTR^Gt-ON males spent more time surrounding and investigating the cupped aggressor, suggesting a necessary role of aVMHvl^OXTR cells in expressing post-defeat social avoidance (Fig. 2i-k). However, the effect of aVMHvl^OXTR optogenetic inhibition on post-defeat social fear could not be determined as light delivery itself affected immobility. Although control OXTR^mCh-OFF and OXTR^Gt-OFF animals showed decreased movement velocity during the post-defeat SI test, both control OXTR^mCh-ON and test OXTR^Gt-ON animals did not (Fig. 2l).

As a complementary strategy to inhibit aVMHvl^OXTR cells without light, we chemogenetically inhibited the aVMHvl^OXTR cells (OXTR^Gi-A21) using hM4Di²⁰ during the post-defeat SI test (Fig. 2m-n). Two control groups were also tested, including OXTR^Gi-Sal (expressing hM4Di and injected with saline) and OXTR^mCherry-A21 (expressing mCherry and injected with Agonist 21, an hM4Di specific ligand²¹) (Fig. 2m). Compared with OXTR^Gi-Sal and OXTR^mCh-A21 groups, OXTR^Gi-A21 animals spent more time surrounding and investigating the cupped aggressor during the post-defeat SI test (Fig. 2o-q). However, OXTR^Gi-A21 animals showed reduced movement velocity comparable to that of control animals, suggesting that the manipulation did not reduce social fear (Fig. 2r). Thus, the activity of aVMHvl^OXTR cells is necessary for social avoidance but not social fear expression.

Social avoidance learning requires OXTR

Is the aVMHvl oxytocin-OXTR signaling essential for defeat-induced social avoidance and fear learning and/or expression? To address this question, we knocked out OXTR in the aVMHvl by bilaterally injecting Cre-GFP virus into OXTR^flox/flox male mice (OXTR^aVMHvl-KO) (Fig. 3a). Control animals were injected with GFP virus (OXTR^aVMHvl-GFP) (Fig. 3a). To confirm the successful OXTR knockout, we injected Cre-GFP and GFP viruses each into one side of aVMHvl in OXTR^flox/flox male mice and performed in vitro patch clamp recording of GFP positive cells (ED Fig. 12a). While 7/16 (44%) of GFP+ cells in GFP expressing side were depolarized by TGOT, a highly specific OXTR agonist, no GFP+ cell was responsive to TGOT in Cre-GFP expressing side, suggesting that OXTR has been effectively knocked out by Cre-GFP (ED Fig. 12b-c).

Fig. 3. OXTRs in aVMHvl are essential for defeat-induced social avoidance learning.

a. Schematics and representative histology. Scale bar: 200 μm.

b. Experimental timeline and behavior assay.

c. Total defeat duration of the OXTR knockout and control mice in RI tests.

d. Body center distributions of representative OXTR KO and control mice during pre- and post-defeat SI tests.

e-g. Change index of aggressor investigation time percentage (e), around cup time percentage (f), and median velocity when far from the aggressor (g) during SI tests for various groups.

h. Cannula implantation schematics.

i. Representative histology. Scale bar: 200 μm.

j-k. Timelines for antagonizing OXTR during defeat (j) or post-defeat SI test (k).

l. Total defeat time of saline (Sal)- and OXTR antagonist (OXTRA)-injected animals during RI tests.

m-r. Change index of aggressor investigation time percentage (m, p), around cup time percentage (n, q), and median velocity when far from the aggressor (o, r) during SI tests of animals injected before RI (m-o) or post-defeat SI tests (p-r).

Error bars represent ± SEM. Circles represent individual animals. Numbers on the plots indicate the number of animals. Mann-Whitney test (c, o, q), and unpaired t-test (e-g, l, m-n, p, r). All statistical tests are two-tailed. **p<0.01, ***p<0.001. Illustration in (a, h) is based on https://atlas.brain-map.org/. See Supplementary Table 1 for detailed statistics.

We then asked whether OXTR knockout at the aVMHvl affects defeat-induced social avoidance and fear (Fig. 3b). During RI tests with SW aggressors, OXTR^aVMHvl-KO and OXTR^aVMHvl-GFP male mice were defeated for a similar amount of time (Fig. 3c). In the post-defeat SI test, OXTR^aVMHvl-KO males spent significantly more time surrounding and investigating the cupped aggressor than OXTR^aVMHvl-GFP males, suggesting an essential role of aVMHvl OXTR in defeat-induced social avoidance (Fig. 3d-f). In contrast, both test and control groups showed decreased velocity after defeat when far away from the cupped aggressor, seemingly suggesting a less critical role of the aVMHvl OXTR in social fear (Fig. 3g). However, we noticed that the absolute movement velocity of OXTR^aVMHvl-KO males was higher than that of OXTR^aVMHvl-GFP males during pre-defeat SI tests, suggesting tonic changes in the aVMHvl^OXTR cell output after OXTR KO, which may lead to circuit compensation (Supplementary Note 1).

To avoid the potential circuit compensation that may mask the endogenous OXTR functions and address whether OXTR signaling is required for acquiring or expressing defeat-induced social avoidance, we injected L-368,899 hydrochloride (100 μM, 250 nL/side), a potent OXTR antagonist (OXTRA), into the aVMHvl of wildtype male mice either 20 min before the RI test (aVMHvl^OXTRA-RI) or 20 min before the post-defeat SI test (aVMHvl^OXTRA-SI) (Fig. 3h-k). Control males were injected with saline (aVMHvl^Sal-RI and aVMHvl^Sal-SI). OXTR antagonist- and saline-injected animals were defeated for a similar amount of time during the RI tests (Fig. 3l). During post-defeat SI test, aVMHvl^OXTRA-RI male mice spent more time surrounding and investigating the cupped aggressor and showed less reduction in movement velocity when far away from the aggressor in comparison to aVMHvl^Sal-RI mice (Fig. 3m-o). In contrast, injecting the OXTR antagonist before the post-defeat SI test did not affect social avoidance or fear (Fig. 3p-r). These results suggest that OXTR signaling at the aVMHvl is necessary for social avoidance and fear learning during defeat but not their expression during post-defeat social encounters.

OXT facilitates synaptic potentiation

Next, we aimed to identify the oxytocin source for aVMHvl^OXTR cells. Given that OXTR signaling in the aVMHvl is required during defeat for social avoidance learning, we reasoned that the relevant oxytocin release should occur during defeat. We found that defeat-induced c-Fos and oxytocin double-positive cells were present in the paraventricular hypothalamic nucleus (PVN), supraoptic nucleus (SON), and retrochiasmatic supraoptic nucleus (SOR)²², a small region caudal to SON (and sometimes considered as a subdivision of SON^23,24) (ED Fig. 13a-b). SOR is particularly interesting as it contained the highest percentage (~50%) of oxytocin cells expressing defeat-induced c-Fos (ED Fig. 13c). Anatomically, SOR is located right next to the aVMHvl^OXTR cells (Fig. 4a), making it well positioned to provide oxytocin to aVMHvl cells through somatodendritic release^25,26.

Fig. 4. SOR is the primary source of OXT for aVMHvl^OXTR cells.

a. OXT (red) and OXTR (green) mRNA expression in male SOR and aVMHvl. Scale bars: 200 μm.

b. Slice recordings of aVMHvl cell responses to PVN or SOR OXT inputs.

c-d. ChR2 expression in PVN^OXT (c) and SOR^OXT (d) cells and their axons surrounding the aVMH. The aVMH image in (c) was digitally enhanced. Scale bars: 200 μm.

e. Strategy to examine aVMHvl cell responses to OXT input.

f. Representative recording traces of aVMHvl cells (f1) responsive to SOR^OXT stimulation (SOR Light+); (f2) unresponsive to SOR^OXT stimulation but depolarized by TGOT (SOR Light-TGOT+); and (f3) unresponsive to SOR^OXT activation and TGOT (SOR Light-TGOT-).

g-h. RMP change of aVMHvl cells after SOR^OXT (g) or PVN^OXT (h) light activation and TGOT.

i. Distribution of response types. Cells in g-i are from 4 male mice/group.

j. Slice recording schematics.

k. ChrimsonR expressing PA cells (k1) and their axons in the aVMHvl (k2). Scale bars: 200 μm.

l. Recording and light delivery protocols to examine OXT’s role in synaptic potentiation.

m. aVMHvl^OXTR cell input resistance before and after 10 min TGOT.

n, q. Slopes of light-evoked EPSP (oEPSP) of aVMHvl^OXTR cells before and after TGOT and PA-VMHvl stimulation. Open and closed circles indicate oEPSPs with and without action potential (AP), respectively.

o, p, r, s. Normalized oEPSP slope (o, r) and light-evoked firing probability (p, s) at different recording periods. Cells are from 5 (o, p) and 4 (r, s) male mice.

Error bars represent ± SEM. Lines in gj-h, o, p, r, and s represent individual cells. Numbers on the plots indicate the number of cells (g-i, m, o-p, r-s). One-way RM ANOVA with Tukey’s multiple comparisons (r), Chi-square test (i), and Friedman test with Dunn’s multiple comparisons (o, p, s), and paired t-test (m). All statistical tests are two-tailed. *p<0.05, **p<0.01, and ***p<0.001. Illustration in (j) is based on a brain atlas from https://atlas.brain-map.org/. See Supplementary Table 1 for detailed statistics.

To understand the influence of oxytocin input on aVMHvl cell activity, we virally expressed ChR2 in SOR^OXT or PVN^OXT cells using OXT^Cre male mice and performed current clamp recording of aVMHvl cells on brain slices while delivering light pulses (1 ms, 20 Hz) for 5 min to activate SOR^OXT or PVN^OXT input (Fig. 4b-d). Upon activation of SOR^OXT input, 11/22 aVMHvl cells showed > 4 mV increase in resting membrane potential, consistent with the reported effect of oxytocin on VMHvl cells²⁷, and we considered those cells as putatively OXTR+ (Fig. 4e, f1, g, and i). In comparison, only 2/24 aVMHvl cells were depolarized by PVN^OXT terminal activation (Fig. 4h-i). For aVMHvl cells that were not depolarized by PVN^OXT or SOR^OXT activation, we applied TGOT to functionally determine OXTR expression (Fig. 4e). We found that 3/11 of SOR^OXT activation unresponsive cells were depolarized by TGOT while 11/22 of PVN^OXT stimulation unresponsive cells did so (Fig. 4e, f2-f3, g-i). Altogether, SOR^OXT and PVN^OXT inputs influenced 79% (11/14 cells) and 15% (2/13 cells) of aVMHvl^OXTR cells, respectively. The SOR^OXT impact on aVMHvl^OXTR cells is OXTR-dependent. Pre-application of OXTR antagonist prevented SOR^OXT activation-induced depolarization in all recorded aVMHvl cells (ED Fig. 12d-e).

The vast majority of oxytocin cells express vesicular glutamate transporter 2 (Vglut2), suggesting that oxytocin cells may co-release glutamate (ED Fig. 12f-i). However, we did not observe optogenetically evoked EPSC (oEPSC) with light delivery to activate SOR^OXT or PVN^OXT inputs (ED Fig. 12j-m). These results suggest that SOR^OXT cells are the primary oxytocin source for aVMHvl cells, and they likely do not form glutamatergic synapses with aVMHvl cells.

We next asked whether oxytocin-OXTR signaling at the aVMHvl could facilitate synaptic potentiation as observed in defeated animals (Fig. 1m). To control the excitatory input to aVMHvl^OXTR cells, we virally expressed ChrimsonR²⁸ in the posterior amygdala (PA) cells and performed current clamp recording of aVMHvl^OXTR cells using OXTR^Cre:Ai6 male mice (Fig. 4j-k). PA is the primary extrahypothalamic glutamatergic input to the VMHvl and evokes strong monosynaptic EPSC from VMHvl cells^29–31. For each recorded aVMHvl^OXTR cell, we probed its postsynaptic responses to PA inputs with 1 ms, 0.1 Hz, 605 nm light pulses for 5 min (Fig. 4l). Then, we delivered 20 Hz, 5 ms, 25 sec light pulses for 3 times to mimic the strong PA input that could occur naturally during intruder encounter²⁹(Fig. 4l). After the light train, the postsynaptic response of aVMHvl^OXTR cells to the PA input only increased slightly (Fig. 4n-o). Strikingly, when the light pulses were delivered in the presence of TGOT, the increase in light-evoked excitatory postsynaptic potential (oEPSP) was significantly larger (Fig. 4n-o). Consequently, aVMHvl^OXTR cells were more likely to fire action potentials upon 1-ms PA terminal stimulation (Fig. 4p).

TGOT application increased aVMHvl^OXTR cell input resistance, raising the possibility that TGOT alone could increase oEPSP due to alternation in synaptic integration (Fig. 4m)³². We thus probed oEPSP after 10 min TGOT perfusion and found no significant change in the oEPSP magnitude or spiking probability (Fig. 4l, q-r). We then repeatedly delivered light pulse trains to activate PA terminals and observed consistent increases in oEPSP amplitude and firing probability (Fig. 4q-s). These synaptic changes were maintained for at least 10 minutes after TGOT wash-off (Fig. 4q-s). Thus, simultaneous OXTR activation and excitatory synaptic inputs to aVMHvl cells are required for synaptic potentiation, possibly through a postsynaptic voltage-dependent mechanism³³, although the contribution of inhibitory synapses could not be excluded.

Noxious stimuli activate SOR^OXT cells

To understand the in vivo response patterns of SOR^OXT cells, we recorded their Ca²⁺ activity using fiber photometry in male OXT^Cre mice during pre- and post-defeat MSI tests and RI tests (Fig. 5a-b, ED Fig. 14a-b). SOR^OXT cells did not increase activity during the initial intruder encounter, social investigation, or attack (Fig. 5c-j). However, SOR^OXT cells were highly activated during each episode of fight and defeat, regardless of the territorial environment (Fig. 5d-e, i-j). After defeat, SOR^OXT cells showed no increase in response to SW aggressor or any other conspecifics, which is qualitatively different from the post-defeat response patterns of aVMHvl^OXTR cells (ED Fig. 14).

Fig. 5. SOR^OXT cells are activated by noxious stimuli.

a. Virus schematics and representative histology. Scale bar: 200 μm. Illustration is based on a brain atlas from https://atlas.brain-map.org/.

b. Three RI test conditions.

c-e. Representative Z-scored GCaMP6f traces during RI tests with a BC male intruder (c), a SW male intruder (d), or resident SW (e). c2, d2, and e2 show the enlarged boxed areas.

f. PETHs of GCaMP6f signal aligned to initial opponent encounters. Only sessions without defeat or attack during the first 10 sec are included.

g. Peak GCaMP6f response within the first 10 sec of RI tests.

h, i. PETHs of GCaMP6f signals aligned to close investigation (h) and agonistic interactions (i).

j. Average Z-scored responses of SOR^OXT cells during various social behaviors.

k. Schematics of head-fixed fiber photometry recording of SOR^OXT cells.

l. Representative raw GCaMP6f trace of SOR^OXT cells during various stimulus presentations.

m. PETHs of Z-scored GCaMP6f signals aligned to the onset of various stimulus presentations. Gray and black dashed lines indicate the onset and offset of stimulus presentation.

n. Average Z-scored ΔF/F during various stimulus presentations.

Shades and error bars represent ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. Unpaired t-test (g), Two-way repeated measure ANOVA with Sidak’s multiple comparisons (j), and One-way repeated measure ANOVA with Tukey’s multiple comparisons (n). All statistical tests are two-tailed. *p<0.05, **p<0.01, and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Given the specific response of SOR^OXT cells during strenuous fight and defeat, we hypothesized that SOR^OXT cells are activated by sensory inputs associated with being attacked, i.e., pain. To test this hypothesis, we recorded Ca²⁺ activity of SOR^OXT cells in head-fixed animals while presenting aggressor urine, gentle touch on the back, and pinch on the back and tail (Fig. 5k-l). SOR^OXT cells showed robust and consistent activity increase during tail pinch and back pinch but no response during gentle touch or urine presentation (Fig. 5l-n). Female SOR^OXT cells responded similarly to those in males (ED Fig. 15). In contrast, aVMHvl^OXTR cells did not show consistent activity change during the delivery of the noxious somatosensory stimuli (ED Fig. 10e-g). These results suggest that SOR^OXT cells are activated specifically during defeat, likely due to the behavior-associated noxious somatosensory inputs.

SOR^OXT boosts social avoidance learning

To understand the functional importance of SOR^OXT cells in defeat-induced social avoidance and fear learning, we virally expressed diphtheria toxin receptor (DTR)³⁴ to ablate SOR^OXT cells using OXT^Cre male mice (OXT^DTR) (Fig. 6a). After DT injection, oxytocin staining in the SOR disappeared in DTR- but not GFP-expressing animals (OXT^GFP) whereas vasopressin expression in OXT^DTR mice was intact (Fig. 6a-b). During the RI test, OXT^DTR and OXT^GFP males were defeated for a similar amount of time by SW aggressors (Fig. 6c-d). However, during the post-defeat SI test, OXT^DTR mice spent more time around and investigating the cupped aggressor than OXT^GFP males (Fig. 6e-f). OXT^DTR also showed slightly less reduction in movement velocity when away from the aggressor than OXT^GFP animals (Fig. 6g). These results support a functional role of SOR^OXT cells in defeat-induced social avoidance and social fear.

Fig. 6. SOR^OXT cells are essential for social avoidance learning.

a. Virus schematics. SOR OXT expression was detectable in a GFP but not in a DTR mouse. Vasopressin (AVP) expression remained in a DTR mouse. Scale bars: 100 μm.

b. Number of SOR OXT+ and AVP+ cells in GFP and DTR mice.

c. Experimental timeline. DT: diphtheria toxin.

d. Total defeated time of GFP and DTR animals during the RI test.

e-g. Change index of aggressor investigation time (%) (e), around cup time (%) (f), and median movement velocity when far from the aggressor (g) during SI tests of GFP and DTR mice.

h. Virus schematics and histology. Scale bar: 500 μm.

i. Experimental timeline.

j-m. Behavioral results of Gt and control animals during defeat (j) and post-defeat SI tests (k-m).

n. Virus schematics and histology. Scale bar: 200 μm.

o. Experimental timeline and illustration.

p. Total defeat time during the RI tests.

q-s. Change index of aggressor investigation time (%) (q), around the cup time (%) (r), and median movement velocity when far from the aggressor (s) during SI tests.

t. A model of SOR^OXT-aVMHvl^OXTR oxytocin signaling-dependent social avoidance learning after defeat.

Error bars represent ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. One-way ANOVA with Tukey’s multiple comparisons (b), unpaired t-test (d, f-g, j-m), Mann-Whitney test (e), and Two-way repeated measure ANOVA with Sidak’s multiple comparisons (p-s). All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. Illustration in (a, h, n) is based on a brain atlas from https://atlas.brain-map.org/. See Supplementary Table 1 for detailed statistics.

Like aVMHvl OXTR KO mice, mice with ablated SOR^OXT cells showed increased movement velocity compared to control animals, even during pre-defeat SI tests (Supplementary Note 1). To eliminate permanent ablation-induced chronic behavior changes and potential circuit compensation, we optogenetically inhibited the SOR^OXT cells during RI tests using stGtACR2 (OXT^Gt) while control animals were injected with GFP virus (OXT^GFP) (Fig. 6h-i). During defeat, OXT^Gt and OXT^GFP male mice were defeated for a similar duration (Fig. 6j). During the post-defeat SI test, OXT^Gt mice spent significantly less time investigating and around the cupped aggressor (Fig. 6k-l). Additionally, while OXT^GFP mice decreased movement velocity when far from the aggressor after defeat, OXT^Gt mice showed no change in mobility in comparison to pre-defeat level (Fig. 6m). These results further support a critical role of the SOR^OXT cells in defeat-induced social avoidance and fear learning.

Lastly, we asked whether activating SOR^OXT cells could facilitate defeat-induced social avoidance. We virally expressed ChR2 or GFP in SOR^OXT cells using OXT^Cre male mice (OXT^ChR2 and OXT^GFP) and subjected the animals to subthreshold defeat (total defeat time: ~ 5 sec) (Fig. 6n-p), which was insufficient to induce social avoidance (Fig. 6q-s). Light delivery during the brief RI test to OXT^ChR2 animals but not OXT^GFP males caused a significant decrease in the time spent around and investigating the cupped aggressor during the post-defeat SI test (Fig. 6q-r). However, social fear, measured as movement velocity, did not differ across groups (Fig. 6s). Thus, enhancing the activity of SOR^OXT cells is sufficient to facilitate social avoidance learning after a mildly negative social experience. This effect is not due to the valence change caused by SOR^OXT cell activation, as OXT^ChR2 animals did not avoid or prefer the light-paired chamber in the RTPP test (ED Fig. 11o-p).

Discussion

To survive in a complex social group, it is important to learn to stay away from superior competitors. Indeed, a 10-minute defeat is sufficient to induce multi-week avoidance of the winner¹. Our study provides new mechanistic insight into the neural process supporting this rapid behavior change. Before defeat, aVMHvl^OXTR cells show minimum response to aggressor cues and animals do not avoid the aggressor (Fig. 6t). During defeat, pain, likely caused by biting from the aggressor, evokes strong activation of oxytocin neurons in the SOR and presumably releases OXT, which then binds to OXTR of aVMHvl cells and facilitates the long-term potentiation of synapses that carry the aggressor information. After defeat, when the animal reencounters the aggressor, due to the strengthened input that carries aggressor cues, aVMHvl^OXTR cells are now strongly activated, which in turn drives social avoidance to ensure the animals stay away from potentially disadvantageous conflicts (Fig. 6t) (More in Supplementary Note 1).

Our results reveal distinct roles of aVMHvl^OXTR cells in social avoidance and fear. While the cells are indispensable for expressing the former, it is unnecessary for the latter. Inhibiting aVMHvl^OXTR cells reduced defeat-induced social avoidance but did not impair social fear. However, OXT-OXTR signaling during defeat is essential for the emergence of both. Blocking either SOR^OXT cells or aVMHvl OXTR during defeat, but not post-defeat SI test, reduced social avoidance and fear. Based on the specific role of aVMHvl^OXTR cells in social fear learning but not expression, we propose the aVMHvl as an input region to “teach” the social fear circuit (Supplementary Note 1). During defeat, aVMHvl input to the social fear circuit is essential in inducing the circuit changes. However, once the changes are complete, the social fear circuit can operate independently of the aVMHvl input (Supplementary Note 1).

Our results affirmed a critical role of OXT in social behavior plasticity³⁵ and expanded the list of regions through which oxytocin could modulate negative social responses^36–40. The findings that SOR^OXT cells are exclusively activated by painful stimuli and serve as a private source for aVMHvl^OXTR cells during defeat raises the possibility that there are distinct OXT subsystems dedicated to promoting social learning during positive and negative social encounters. Indeed, in contrast to the SOR^OXT cell responses, PVN^OXT cells are activated by positive social experiences, such as gentle social touch, non-antagonistic social interaction, and maternal shepherding^41–44. Future studies identifying such OXT subsystems will be essential for harvesting the therapeutic potential of oxytocin in disorders with social deficits⁴⁵.

Methods

Mice

All procedures were approved by the NYULMC Institutional Animal Care and Use Committee (IACUC) in compliance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals. Mice were housed under a 12-hour light-dark cycle (dark cycle; 10 a.m. to 10 p.m. or 6:30 p.m. to 6:30 a.m.), with food and water available ad libitum. Room temperature was maintained between 20–22 °C and humidity between 30–70%, with a daily average of approximately 45%. Oxtr^Cre (Strain#: 031303)⁴⁶, OXT^Cre (Strain#: 024234)⁴⁶, Vglut2^Cre (Strain#: 0169963)⁴⁷, and OXTR^flox mice (Strain#: 008471)⁴⁸ were purchased from Jackson Laboratory. Ai6 (Strain#: 007906)⁴⁹ mice were from Jackson Laboratory and crossed with Oxtr^Cre and Vglut2^Cre mice. Test mice were between 8 to 24 weeks at the time of behavior testing or recording. Stimulus animals in the RI test were BALB/c male (> 9 weeks), and C57BL/6N male and female mice (> 8 weeks) originally purchased from Charles River and then bred in-house. Swiss Webster male and female mice (> 11 weeks) were purchased from Taconic, Charles River, or bred in-house. All mice were group-housed until adulthood. After surgery with fiber or cannula implantation, all test mice were single-housed. Animals were randomly assigned to control and test groups.

Viruses

The following AAVs were used in this study, with injection titers as indicated. AAV2 CAG-Flex-GCaMP6f-WPRE-SV40 (1.8 × 10¹² vg/ml, UPenn, #V5747S) for fiber photometry was purchased from UPenn vector core. For the functional manipulation, the following AAVs were used. Optogenetic activation: AAV2 Ef1a-DIO-hChR2(H134R)-EYFP (4.2 × 10¹² vg/ml, UNC, #AV4378); Chemogenetic inactivation: AAV2 hSyn-DIO-hM4Di-mCherry (1.5 × 10¹³ vg/ml, Addgene, #44362-AAV2); Optogenetic inactivation: AAV1 hSyn-SIO-stGtACR2-FusionRed (1.9 × 10¹³ vg/ml, Addgene, #105677-AAV1); OXTR knockout: AAV1 CMV-HleGFP-Cre (1.1 × 10¹³ vg/ml, Addgene, #105545-AAV1) or AAV2 hSyn-GFP (3.4 × 10¹² vg/ml, UNC, #4876D); Ablation of oxytocin cells: AAV8 hSyn-DIO-DTR (9.1 × 10¹³ vg/ml, Boston Children’s Hospital), AAV2 hSyn-DIO-GFP (4.0 × 10¹² vg/ml, UNC, #4530C), or AAV2 hSyn-DIO-mCherry (1.8 × 10¹³ vg/ml, Addgene, #50459-AAV2). For slice electrophysiology, the following AAVs were used: AAV2 Ef1a-DIO-hChR2(H134R)-EYFP (4.2 × 10¹² vg/ml, UNC, #AV4378) and AAV9 hSyn-ChrimsonR-tdTomato (2.6 × 10¹³ vg/ml, Addgene, #59171-AAV9).

Drugs

For chemogenetic inhibition, 1 mg/kg Agonist21²¹ (Tocris, #5548) in saline was administered intraperitoneally. To block OXTRs in the aVMHvl, 250 nL/side 100 μM L-368,899 hydrochloride (Tocris, #2641) in saline was injected through implanted cannulae bilaterally. To ablate SOR^OXT cells, 50 μg/kg diphtheria toxin in saline (Sigma-Aldrich, #D0564) was administered intraperitoneally on two consecutive days (7 days before the pre-defeat SI test). A mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) in saline was administered intraperitoneally before perfusion.

Stereotaxic surgery

Mice (8–12 weeks old) were anesthetized with 1–1.5% isoflurane and placed in a stereotaxic apparatus (Kopf Instruments Model 1900). Viruses were delivered into the targeted brain regions through glass capillaries using a nanoinjector (World Precision Instruments, Nanoliter 2010) at a speed of 20 nL/min. 100–120 nL of AAV was injected into each targeted brain region. Stereotaxic injection coordinates were based on the Paxinos and Franklin mouse brain atlas⁵⁰. For fiber photometry and optogenetic manipulation, polished optical fiber (440- or 230-μm diameter, Thorlabs) was implanted 150 or 200 μm above the virus injection site either immediately after virus injection or 2–3 weeks later. During the same surgery as optic fiber implantation, a 3D-printed head-fixation ring⁵¹ was cemented on the skull (C&B Metabond dental cement, Parkell) to allow head-fixation during fiber attach and detach, drug injection through cannula and head-fixed fiber photometry recording. Mice were single-housed after optic fiber implantation. Histology was obtained from all test animals, and only animals with correct virus expression and optic fiber placement were included in the final analysis.

Common for all behavior tests

Mouse behaviors in all experiments were recorded from both the side and top of the cage using two synchronized cameras (Basler, acA640–100 gm and 120 gm) in a semi-dark room with infrared illumination. Video acquisition was achieved using StreamPix 5 (Noprix) at 25 frames/sec. Manual behavioral annotations were performed on a frame-by-frame base using custom software written in MATLAB (https://pdollar.github.io/toolbox/)¹³. Most videos were annotated by people who did not perform behavior assay blindly and then checked and modified by an experimenter who was not blind to the animal’s group assignment. A subset of videos were only annotated blindly. There is high consistency (around > 90%) between annotations performed blindly and those not. During annotation, the neural responses were unknown to the experimenter. Custom DeepLabCut⁵²-based models were constructed to track animals’ body center, head center, and nose points in top-view videos. Movement velocity was calculated as the distance of an animal’s body center between two adjacent frames.

Resident-intruder test

For inter-male RI tests with the goal to defeat the test animal, we introduced the test animal into the home cage of a sexually experienced, aggressive, and single-housed SW male mouse or introduced the SW male aggressor into the home cage of the test mouse (see individual experiments for details). For females, we introduced the test female into the home cage of a lactating female of SW or mixed background. The RI test typically lasted for 10 min, although for several wildtype male mice in Extended Data Fig. 1, the RI test was terminated after 5 min as the aggressor was highly aggressive and attacked the test animal continuously. Across experiments, each test mouse was defeated for approximately 20 sec during the RI test. “Defeat” was annotated when the aggressor’s front-end contacts the test animal’s back, presumably to bite, as the aggressor attacks the test mouse. “Fight” was annotated if the test animal successfully pushes or bites the aggressor when being attacked. For RI tests with non-aggressors, we introduced a Balb/C (BC) non-aggressive group-housed male mouse into the home cage of a male test mouse and a C57 naïve female mouse into the home cage of a test female mouse. “Attack” was annotated when the test mouse initiated a suite of fast actions towards the non-aggressive intruder, including lunges, bites, and tumbles. To analyze the behaviors, in addition to manual annotation, we also tracked the animals and calculated the percentage of time the test animal spent immobile (body center velocity < 1 pixel/frame) during each minute of RI tests.

Social interaction test

For the SI test, the stimulus animals were always the same as those used in RI tests except for Extended Data Figure 2. In the aVMHvl^OXTR optogenetic activation experiment, group-housed non-aggressive Balb/C males were also used as stimulus animals. For females, on the day before the first SI test, the test animal was habituated to the empty metal wire cup (diameter of the cup bottom: 7.5 cm; height: 10.5 cm) in a clean cage for 10–15 min. There was no habituation for SI test in males. During the SI test, the stimulus animal was placed under a cup at one end of a clean cage with bedding, and then the test animal was introduced into the cage and freely interacted with the cupped animal for 5 or 10 min. For each experiment, the SI test was performed one day before and one day after the RI test. For each experiment, the same set of aggressors was used for test and control groups to reduce the variability in defeat.

To analyze behaviors during the SI test, we calculated the distance of the test animal’s body center to the cup center for each frame and constructed the histograms showing the distribution of the test animal-cupped aggressor distance. The percentage of time an animal spent around the cup was calculated as the percentage of frames when the test animal’s body center-to-cup distance was within 15 cm (< 250 pixels), which is approximately half of the cage length. The investigation behavior was annotated manually as the time period when the test animal’s nose point was in close proximity to the cup. When the test animal stayed far from the cupped aggressor (distance > 300 pixels) and when its body center and head center velocity were < 1 pixel/frame for >0.5s, the animal was considered as “immobile”. We then calculated the change index (CI) for each parameter (P), including time around aggressor (%), investigation time (%), and median velocity when far from the aggressor as (P_after – P_before) / (P_after + P_before). P_before: the value of the parameter before defeat; P_after: the value after defeat. CI ranges from −1 to 1. Positive CI values indicate increases after defeat, while negative CIs indicate decreases after defeat. Values close to −1 or 1 indicate large changes.

Multi-animal social interaction test

For the MSI test, the test arena (L × W × H: 22”× 18”× 16”) contained four wire meshed cups, one in each corner. On the habituation days (2 days), test animals freely explored the arena for around 20 min without any cups. On the test day, one cup was left empty, and the other three cups were each with a stimulus animal. The test animals were allowed to explore the arena freely for 10 min. MSI test was performed one the day before and one day after RI tests. For each male MSI test, we also introduced an aggressive SW male (16–40 weeks), group-housed C57 male (16–24 weeks) and a BC non-aggressive group-housed male (14–24 weeks), each under a cup, as stimulus animals. The same BC was encountered during the RI test, during which the test animal either attacked or investigated it but was never defeated by it. The C57 stimulus males were unfamiliar and only encountered during the MSI tests. For each female MSI test, the SW mother aggressor (16–32 weeks), a single-housed SW virgin female (10–24 weeks), and a single-housed C57 virgin female (10–24 weeks), each under a cup, were introduced as stimulus animals. The same C57 virgin mice were encountered during the RI test (only investigation). The SW virgin mice were unfamiliar and only encountered during the MSI tests.

To analyze the behaviors, we calculated the distance between the animal’s head center and each cup’s center. When the distance was < 3× radius of the cup (r_cup), the test animal was considered to be around the cupped animal. We also calculated the distance between animal’s nose point to the center of each cup. When the distance was < 1.5× r_cup, the test animal was considered investigating the cupped animal. We then calculated the percentage of time each test animal spent on investigating and around the cup during before- and after-defeat MSI tests. We also calculated the number of approaches to each stimulus during the MSI tests. When the animal moved from a far distance, and its nose reached the distance to the cup center shorter than the cup diameter, we considered it an approaching event.

Odor-shock paired conditioning

The odor delivery chamber was custom-made by opening one odor delivery port (diameter: 1/4”) and one vacuum port (diameter: 1/4”) onto the opposite short walls of an acrylic chamber (L × W × H: 14 1/2” x 7 1/2” x 11 7/8”). The odors (10% Pentyl acetate (Sigma-Aldrich, #109584) and (R)-(+)-limonene (Sigma-Aldrich, #183164)) were delivered at a rate of 2L/min through Tygon tubing (MFLX06422–07 and MFLX07407–75, Cole-Parmer) inserted directly into the odor delivery port. Each odor was delivered for 3 sec, 8 times, with an inter-trial interval of 60 sec. There was a 10 min no-odor period between Pentyl acetate and limonene presentations. A second tube was inserted into the vacuum port and connected to the building vacuum system to move air constantly. The odor delivery chamber was free of bedding or any objects.

The foot shock – odor conditioning occurred in a foot-shock chamber (ENV-307W-CT, Med Associates) that was modified for odor delivery by adding the custom odor delivery and vacuum ports. During the conditioning session, we first delivered one odor (10% Pentyl acetate or limonene) at 2 L/min, 8 times for 3 sec each, with 60 sec intervals without any shock. Then, 10 min after completing the first odor delivery, we delivered a second odor (2 L/min, 8 times for 3 sec each within 60 sec intervals). The foot shock (1 mA, 1 sec) was delivered between 2 and 3 sec of each odor delivery trial. The odor (10% Pentyl acetate or limonene) paired with the shock was randomly selected for each animal. We examined the behavior and neural responses of OXTR^GCaMP male mice in the odor delivery chamber one day before and one day after the shock-odor conditioning.

Optogenetic activation

To activate aVMHvl^OXTR cells, we injected 120 nL Cre-dependent ChR2 (control: GFP) expressing virus unilaterally into the aVMHvl (Bregma coordinates: AP, −1.455 mm; ML, −0.65 mm; DV, −5.70 mm) of OXTR^Cre mice and placed a 230-μm multimode optic fiber (Thorlabs, FT200EMT) 200 μm above injection site. On the test day, the implanted fiber was connected to a matching patch cord using a plastic sleeve (Thorlabs, ADAL1) to allow light delivery (Shanghai Dream Lasers). During the test, a non-aggressive BC male mouse or an aggressive SW male mouse was placed under in a metal wire cup located at one end of a clean cage, and 20 ms, 20 Hz, 1.5–2 mW light was delivered for 60 sec, followed by 0 mW sham light for 60 sec, and then repeated once. In a separate group of animals, we stimulated aVMHvl^OXTR cells without any target in a clean cage using the same stimulation protocol (20 ms, 20 Hz, 1.5–2 mW light for 60 sec, followed by 0 mW sham light for 60 sec with one light-sham repeat).

To activate SOR^OXT cells, we injected 120 nL Cre-dependent ChR2 (control: GFP) expressing virus bilaterally into the SOR (Bregma coordinates: AP, −1.36 mm; ML, ±0.90 mm; DV, −5.69 mm) of OXT^Cre male mice and placed two 230-μm multimode optic fibers (Thorlabs, FT200EMT) 200 μm above the injection sites. The behavior test started three weeks after the virus injection. Each ChR2 and GFP animal underwent two rounds of 3-day SI-RI-SI tests with at least 3 weeks in between. No light was delivered during SI tests. During the RI test, a SW male aggressor was introduced into the home cage of the test mice until the aggressor defeated the test moue for 3–4 times with a total defeat duration of approximately 5 sec. For one round of the SI-RI-SI test, the light (3.5–4 mW, 20 ms, 20 Hz) was delivered during the entire RI test. For the other round of the SI-RI-SI test, the test animals received no light during the RI tests. The order of light delivery during RI tests was counterbalanced across animals.

The real-time place preference (RTPP) test was performed to investigate the valence of aVMHvl^OXTR and SOR^OXT activation. The test area contains two equal size chambers (13 cm (L) x 16 cm (W) x 25 cm (H) per chamber) made with transparent acrylic boards. During the test, the animal was allowed to freely move in the chamber for 15 minutes without any stimulation, and then the light (3.5–4 mW, 20 ms, 20 Hz) was manually delivered through the implanted optic fiber whenever the animal entered a pre-assigned chamber for 10 minutes. The body center of the animal was tracked using custom tracking software (https://pdollar.github.io/toolbox/)¹³ and used to calculate the time spent in each chamber.

Optogenetic and chemogenetic inhibition

To inactivate aVMHvl^OXTR cells, we injected 120 nL Cre-dependent stGtACR2 (control: mCherry) virus bilaterally into the aVMHvl (Bregma coordinates: AP, −1.455 mm; ML, ±0.65 mm; DV, −5.70 mm) of OXTR^Cre mice and placed two 230-μm multimode optic fibers (Thorlabs, FT200EMT) 200 μm above the injection sites. Three weeks after the virus injection, all test animals went through 3-day SI-RI-SI tests. For each test animal, the aggressor was the same SW male mouse throughout the tests. No light was delivered during the first SI test or RI test. During the post-defeat SIT test, one group of stGtACR2 and mCherry animals received light (473nm, 3.5–4 mW, 20 ms, 20 Hz) for 5 min, and the other group received no light.

To chemogenetically inhibit aVMHvl^OXTR cells, we injected 100–110 nL Cre-dependent hM4Di (control: mCherry) virus bilaterally into the aVMHvl (Bregma coordinates: AP, −1.455 mm; ML, ±0.65 mm; DV, −5.70 mm) of OXTR^Cre mice. Three weeks later, animals underwent 3-day SI-RI-SI tests with SW male aggressors. No drug was injected during the pre-defeat SI or RI tests. One hour before the post-defeat SI test, test animals were injected with 250 uL of saline or agonist 21 solutions (1 mg/kg, Tocris, #5548) intraperitoneally.

To optogenetically inactivate SOR^OXT cells, we injected 120 nL Cre-dependent stGtACR2 (control: mCherry or GFP) virus bilaterally into the SOR (Bregma coordinates: AP, −1.36 mm; ML, ±0.90 mm; DV, −5.69 mm) of OXT^Cre mice and placed two 230-μm multimode optic fibers (Thorlabs, FT200EMT) 200 μm above the injection sites. Three weeks after the virus injection, all test animals underwent 3-day SI-RI-SI tests (all 10 min). For each test animal, the aggressor was the same SW male mouse throughout the tests. No light was delivered during the SI tests. During RIT test, all animals received light (473nm, 3.5–4 mW, 20 ms, 20 Hz) for 10 min continuously.

Knockout of aVMHvl OXTRs

To knock out OXTR in the aVMHvl, we bilaterally injected 100 nL AAV expressing GFP-Cre (control: GFP) into the aVMHvl (Bregma coordinates: AP, −1.46 mm; ML, ±0.65 mm; DV, −5.70 mm) of OXTR^flox/flox male mice. 3–4 weeks after the virus injection, all test animals went through the 3-day SI-RI-SI tests. During RI tests, SW aggressors were introduced into the home cage of the test animals for 10 min.

OXTR antagonist application

To block aVMHvl OXTR, we implanted bilateral cannulae (PlasticsOne, center-to-center distance: 1.5 mm) 0.7 mm above the aVMHvl (Bregma coordinates: AP, −1.45 mm; ML, ±0.75 mm; DV, −5.00 mm) of wild type C57BL/6 male mice. One week after the surgery, all animals went through the 3-day SI-RI-SI tests. To block aVMHvl OXTR during defeat, we injected 250 nL/side 100 μM L-368,899 hydrochloride (Tocris, #2641) into the aVMHvl through the cannula using a syringe (Hamilton, #65457–02) 20 min before RI test when the animal was head-fixed on a running wheel. To block OXTR during the post-defeat SI test, the same drug was injected 20 min before the SI test. Control animals were injected with saline and went through the same behavior tests. During the waiting time after the drug injection, the animal was returned to its home cage. Before sacrificing the animals, we injected 250 nL 10 ng/mL DiI (ThermoFisher) to mark the injection site.

Ablation of OXT cells in the SOR

To ablate SOR^OXT cells, we injected 120 nL AAV expressing Cre-dependent DTR (control: GFP) into the SOR (Bregma coordinates: AP, −1.36 mm; ML, ±0.90 mm; DV, −5.69 mm) bilaterally using OXT^Cre male mice. Three weeks later, we intraperitoneally injected each animal with 50 μg/kg diphtheria toxin (DT, Sigma-Aldrich, #D0564) per day for two consecutive days. One week after the 1^st DT injection, all animals went through the 3-day SI-RI-SI tests.

Fiber photometry recording

We injected 120 nL AAV expressing Cre-dependent GCaMP6f into the SOR or aVMHvl unilaterally of 10–12 weeks old OXT^Cre and OXTR^Cre male and female mice. The following Bregma coordinates were used. Male SOR: AP, −1.36 mm; ML, −0.90 mm; DV, −5.69 mm from the top of the skull; female SOR: AP, −1.355 mm; ML, −0.88 mm; DV, −5.68 mm from the skull surface. Male aVMHvl: AP, −1.46 mm; ML, −0.65 mm; DV, −5.7 mm from the skull surface. Female aVMHvl: AP, −1.455 mm, ML, −0.645 mm; DV, −5.72 mm from the skull surface. Recording started at least 3 weeks after the virus injection.

Prior to fiber photometry recording, a ferrule sleeve (ADAL1–5, Thorlabs) was used to connect a matching patch cord to the implanted optic fiber when the animal was head fixed. For recordings, a 390-Hz sinusoidal 470-nm blue LED light (35 mW; LED light (M470F1, Thorlabs) driven by a LED driver (LEDD1B, Thorlabs) was bandpass-filtered (passing band: 472 ± 15 nm, Semrock, FF02–472/30–25) and delivered to the brain in to excite GCaMP6f. The emission light then passed through the same optic fiber, a bandpass filter (passing band: 534 ± 25 nm, Semrock, FF01–535/50), detected by a Femtowatt Silicon Photoreceiver (Newport, #2151) and recorded using RZ5 real-time processor (Tucker-Davis Technologies). The envelope of 390-Hz signals from the photoreceiver was extracted in real-time using a custom-written program (Tucker-Davis Technologies) as the readout of GCaMP6f intensity. Top- and side-view behavior videos were simultaneously recorded (Basler, acA640–100 gm and 120 gm) and acquired using StreamPix 5 (Noprix) at 25 frames/sec. Timestamps of video frames were used to align GCaMP6f signal and behaviors videos. For the head-fixed recording of SOR^OXT and aVMHvl^OXTR cells, aggressor urine was collected from SW male aggressors or lactating SW female mice on the same day of recording. Urine was pooled from multiple aggressors, including the mouse that defeated the test animal during the RI test. We then added 100 μL of urine to a Q-tip using a pipette and presented it to the recording animal manually for 10 sec with 50 sec in between. Male urine was presented to male test mice, and female urine was presented to female animals. The gentle touch was performed onto the back of test animals with a large fluffy cotton ball (5 swipes for neck to tail base for each trial). Urine exposure and gentle touch were performed 6 times for each animal. Back and tail pinches were applied with a pair of fine tweezers (FST, #91100–12) at a force that did not cause visible skin damage. 12 pinches were applied, each lasting for approximately 3 sec, with 50–60 sec in between. Additionally, we recorded aVMHvl^OXTR cell responses to unpaired and shock paired-odor in head-fixed animals one day after odor-pairing. 100 μL of 10% pentyl acetate or limonene was added to a Q-tip and manually presented to the recording animals 6 times, each for 10 sec with 50 sec in between. Unpaired odor and paired odor were delivered sequentially with 5 min intervals in between.

We also compared the aVMHvl^OXTR cell responses to 2-methyl-2-thiazoline (2MT) and aggressor urine in free-moving and head-fixed animals. On the day before the recording, the test animal was defeated by a SW male aggressor for 10 min in the SW home cage. On the recording day, the test animal went through a 10-min SI test with the cupped SW aggressor in a clean cage, and then 5 min later, was presented with 100 μL 1% 2MT in PBS on a filter paper placed at one end of the clean cage for 10 min. After recording in free-moving animals, we also recorded the cell response to saline, 2MT, and aggressor urine in head-fixed animals. During recording, 100 μL of saline, urine from the same SW aggressor, and 1% 2MT were presented to the animals on Q-tips, each for 6 times, 10 sec per trial with 50 sec between trials. There was a 5-min interval between presentations of different stimuli.

To analyze the free-moving recording data, MATLAB function ‘‘msbackadj’’ with a moving window of 20% of the total recording duration was first applied to obtain the instantaneous baseline signal. The instantaneous ΔF/F was calculated as (F_raw-F_baseline)/F_baseline. The Z-scored ΔF/F of the entire recording session was calculated as (ΔF/F – mean(ΔF/F))/std(ΔF/F). The peri-event histogram (PETH) of Z-scored ΔF/F aligned to a given behavior was constructed for each animal and then averaged across animals. The response during a specific behavior for each animal was calculated by averaging the Z-scored ΔF/F during all periods when the behavior occurred.

To analyze the recording data in head-fixed animals, we constructed the PETHs of the raw fluorescence signal and then calculated the averaged PETH for each animal and used the −20 to 0 sec recording trace before the stimulus onset as the baseline to calculate the mean and standard deviation for Z-scoring the whole PETH. The mean Ca²⁺ signal of each stimulus for each animal was calculated by averaging the Z-scored PETH values from 0 sec to the average duration of the stimulus presentation (approximately 10 sec). We did not perform baseline correction for head-fixed recording as we noticed sustained signal suppression during the 2MT presentation, making the low-pass-based baseline correction inaccurate.

Slice electrophysiology

To prepare brain slices for patch clamp recording, mice were anesthetized with isoflurane, and brains were quickly removed and then immersed in ice-cold cutting solution for 1–2 min (in mM: 110 choline chloride, 25 NaHCO3, 2.5 KCl, 7 MgCl2, 0.5 CaCl2, 1.25 NaH2PO4, 25 glucose, 11.6 ascorbic acid and 3.1 pyruvic acid). The 275 μm aVMHvl coronal sections were cut using a Leica VT1200s vibratome, collected in oxygenated (95% O2 and 5% CO2) and pre-heated (32–34 °C) artificial cerebrospinal fluid (ACSF) solution (in mM: 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 1 MgCl2, 2 CaCl2 and 11 glucose) and incubated for 30 min. The sections were then transferred to room temperature and continuously oxygenated until use.

Current and voltage whole cell patch clamp recordings were performed with micropipettes filled with intracellular solution containing (in mM: 145 K-gluconate, 2 MgCl2, 2 Na2ATP, 10 HEPES, 0.2 EGTA (286 mOsm, pH 7.2) or 135 CsMeSO3, 10 HEPES, 1 EGTA, 3.3 QX-314 (chloride salt), 4 Mg-ATP, 0.3 Na-GTP and 8 sodium phosphocreatine (pH 7.3 adjusted with CsOH)). Signals were recorded with MultiClamp 700B amplifier (Molecular Devices) and Clampex 11.0 software (Axon Instruments), digitized at 20 kHz with Digidata 1550B (Axon Instruments). After recording, data were analyzed using Clampfit (Molecular Devices) or MATLAB (Mathworks).

To characterize the physiological and synaptic properties of aVMHvl^OXTR cells, we identified ZsGreen positive cells in the aVMHvl on slices from OXTR^ZsGreen mice using an Olympus 40 × water-immersion objective with a GFP filter. For investigating intrinsic excitability, cells were recorded in current-clamp mode, and the number of action potentials was counted over 500-ms current steps. The current steps consisted of 30 sweeps from −20 pA to 270 pA at 10 pA per step. sEPSCs and sIPSCs were recorded in the voltage-clamp mode. The membrane voltage was held at −70 mV for sEPSC recordings and at 0 mV for sIPSC recordings.

To investigate the efficacy of OXTR knockout, 120 nL AAV1-Cre-GFP and 120 nL AAV2-GFP were injected into the left (KO) and right (Control) sides of the brain, respectively. 3–4 weeks later, GFP-positive cells in the aVMHvl from both KO and control sides were recorded in current clamp mode. All cells were recorded for 3–5 min after break-in until RMP was stable and then perfused with TGOT (250 nM) for 10 min. Cells that increased RMP for > 4 mV after 10 min TGOT perfusion were considered OXTR positive.

In PVN^OXT and SOR^OXT optogenetic activation experiment, we injected 120–140 nL AAV2-Ef1a-DIO-ChR2-EYFP into either PVN or SOR of OXT^Cre male mice. After 4 weeks of virus incubation, aVMHvl cells were recorded in current-clamp mode. After the cell membrane potential was stabilized, we delivered 1 ms, 20 Hz blue light pulses (pE-300white; CoolLED) for 5 min to activate PVN^OXT or SOR^OXT processes. If the cell did not show a significant increase in RMP (> 4 mV) after light delivery, we then perfused TGOT (250 nM) in bath for 10 min to functionally determine the OXTR expression. The cells were separated into three categories based on their response to the light activation and TGOT perfusion: Light+, Light-TGOT+, and Light-TOGT-. Light+ and Light-TGOT+ cells were considered as putative OXTR-expressing cells. Light-TGOT- cells were classified as OXTR-. To confirm that SOR^OXT optogenetic activation induces RMP changes of aVMHvl cells through activation of OXTR, we pre-incubated brain sections in ACSF with 1 uM L-368,899 hydrochloride (Tocris, #2641) and performed the same SOR^OXT stimulation protocol.

To examine the effect of OXT on synaptic transmission, we injected 100 nL AAV9 hSyn-ChrimsonR-tdTomato into the PA of OXTR^ZsGreen male mice. After 4 weeks of virus incubation, we obtained brain slices and performed current clamp recording of aVMHvl^OXTR cells. aVMHvl^OXTR cells are determined based on their ZsGreen expression. Before recording, the expression of ChrimsonR-tdTomato in PA was examined using an Olympus 40× water-immersion objective with a TXRED filter. The recording only proceeded if ChrimsonR-tdTomato was correctly and robustly expressed in the PA. To probe the excitatory synaptic responses of recorded aVMHvl^OXTR cells, we injected a small positive or negative current to keep the cell membrane potential around −70 mV and delivered a 1 ms 605-nm full-field light pulse every 10 sec (0.1 Hz) (pE-300white; CoolLED). After 5 min probing, we then delivered 3 trains of 20 Hz, 5 ms light pulses with 25 sec per train and 5 sec in between. During light pulse train delivery, the cells were not injected with any positive or negative current. After light train delivery, we again injected negative or positive current to maintain the membrane potential around −70 mV and probed the light-evoked EPSPs for 5 or 10 min, and then bath perfused 250 nM TGOT. After 10 min of TGOT perfusion, we delivered a second set of light pulse trains and probed the light-evoked EPSPs for 5 or 10 min. In a second protocol, we probed the light-evoked EPSPs before and after perfusing TGOT (250 nM) for 10 min, and then delivered 3 trains of 25 s, 20 Hz, 5 ms light pulses, with 5 sec between trains. After light stimulation, light-evoked EPSPs were probed for 5 min, and then TGOT was washed out with ACSF while the light-evoked EPSPs were continuously probed for 10 min.

Immunohistochemistry

For c-Fos and OXT staining, animals were deeply anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) and transcardially perfused with 10 ml of PBS, followed by 10 ml of 4% paraformaldehyde in PBS. After perfusion, brains were harvested, soaked in 30% sucrose in PBS for 24 hours at 4 °C, and then embedded with OCT compound (Fisher Healthcare). 40 μm thick coronal brain sections were cut using a cryostat (Leica). Brain sections were washed with PBST (0.3% Triton X-100 in PBS, 10 min), blocked in 5% normal donkey serum (NDS, Jackson Immuno Research) in PBST for 30 min at room temperature (RT), and then incubated with primary antibodies in 5% NDS in PBST overnight at RT (about 18 hours). Sections were then washed with PBST (3×10 min), incubated with secondary antibodies in 5% NDS in PBST for 4 hours at RT, washed with PBST (2×10 min) and DAPI-mixed (1:10000, Thermo Scientific) PBS solution (1×20 min). Slides were coverslipped using a mounting medium (Fluoromount, Diagnostic BioSystems) after drying.

The primary antibodies used were: rabbit anti-Oxytocin (1:5000, Immunostar, #20068, Lot #1607001), guinea pig anti-c-Fos (1:2000, Synaptic Systems, 226–005, Lot #2–10, 2–13), rabbit anti-Vasopressin (1:5000, Immunostar, #20069, Lot #1004001), rabbit anti-Esr1 (1:2000, Invitrogen, PA1–309, lot#: YA352477), and anti-GFP (1:2000, abcam, ab13970, lot#GR3190550–2). The secondary antibodies used were: Cy3-AffiniPure donkey anti-rabbit IgG (1:500, Jackson Immuno Research, 711–165-152, lot#124528), Cy5-AffiniPure donkey anti-rabbit IgG (1:250, Jackson Immuno Research, 711–175-152, lot#150312), Alexa Fluor 488-conjugated goat anti-guinea pig IgG (1:500, Invitrogen. #A11073, lot#2160428), or Alexa Fluor 488-conjugated donkey anti-chicken IgY(IgG) (1:500, Jackson Immuno Research. 703–545-155, lot#116967). The 10× or 20× fluorescent images were acquired to determine the overall expression pattern in each brain region by Olympus VS120 Automated Slide Scanner and its specific software OlyVIA. The 20× fluorescent confocal images were acquired by Zeiss LSM 800 and its specific software (Zeiss, ZEN 2.3 system) for cell counting.

In situ hybridization

To prepare the sections for in situ hybridization (ISH), 10–12-weeks-old C57BL/6 male mice were anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) and transcardially perfused with 10 ml of DEPC treated PBS (DEPC-PBS), followed by 10 ml of 4% paraformaldehyde in DEPC-PBS (PFA, from paraformaldehyde 32% solution, Electron Microscopy Sciences). After perfusion, brains were harvested, soaked in 30% of sucrose in DEPC-PBS for 24 hours at 4 °C and then embedded with OCT compound (Fisher Healthcare). 30 μm thick coronal brain sections were cut using a cryostat (model #CM3050S, Leica). The sections were placed on MAS-coated glass slides (MAS-03, Matsunami) and stored at −80 °C before use.

To synthesize the cDNA for the OXT and OXTR probes, their original templates were from mouse brain cDNA (cDNA-mmu-01, Biosettia). cDNA was amplified by PCR methods using the following oligo-DNA primers, and the products were purified with micro spin columns (MACHEREY-NAGEL, 74060910). Each reverse primer also possesses T3 sequence for transcription.

OXT1-forward: TGGCTTACTGGCTCTGACCT

OXT1-reverse: AATTAACCCTCACTAAAGGGAGGAAGCGCGCTAAAGGTAT

OXTR1-forward: GGCGGTCCTGTGTCTCATAC

OXTR1-reverse: AATTAACCCTCACTAAAGGGCTCCACATCTGCACGAAGAA

OXTR2-forward: TTCATCATTGCCATGCTCTT

OXTR2-reverse: AATTAACCCTCACTAAAGGGGGGTGGCTCTCATTTCCTTT

OXTR3-forward: GCTGGAGATAGGAGGCAGTG

OXTR3-reverse: AATTAACCCTCACTAAAGGGGCTGTGTCACTCACCAGACG

The OXT probes are approximately 400 bp in length, and the OXTR probes are approximately 750 bp to 1000 bp. ISH probes were prepared by in vitro transcription with DIG RNA Labeling Mix (Roche Applied Science, #11277073910) or Fluorescein RNA Labeling mix (Roche Applied Science, #11685619910) and T3 polymerase (Roche Applied Science, #11031163001). OXT probe was labeled with Flu, and OXTR probes were labeled with DIG.

Brain sections, including the VMHvl, underwent ISH at 56 °C overnight. After a series of post-hybridization washing and blocking, Flu-positive cells were visualized with anti-FITC antibody (PerkinElmer, #NEF710001EA, 1:200 in blocking buffer) followed by TSA biotin amplification reagent (PerkinElmer, #NEF749A001KT, 1:100 in 1 × plus amplification diluent) and streptavidin Alexa488 (Invitrogen, #S11223, 1:250 in blocking buffer). DIG-positive cells were visualized with anti-DIG antibody (Roche Applied Science, #11207733910, 1:250 in blocking buffer) and TSA Cy3 amplification regent (PerkinElmer, #NEL744001KT, 1:100 in 1 × plus amplification diluent). Sections were counterstained with 4’,6-diamino-2-phenylindole dihydrochloride (DAPI, 1:10000 in PBS, Thermo Scientific) and mounted with a cover glass using Fluoromount (Diagnostic BioSystems, #K024). The 20× fluorescent images were acquired using a slide scanner (Olympus VS120). 20× fluorescent confocal images were acquired using Zeiss LSM 800 (Zeiss, ZEN 2.3 system).

Quantification and statistical analysis

No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications^29,53–56. All experiments were conducted using 1 to 2 cohorts of animals. The results were reproducible across cohorts and combined for final analysis. Statistical analyses were performed using MATLAB (version 2019b, 2021b, or 2023b, Mathworks) and Prism9 and Prism10 (GraphPad Software, RRID: SCR_002798). All statistical analyses were two-tailed. Parametric tests, including one sample t-test, paired t-test, unpaired t-test, and One-way ANOVA, were used if distributions passed Kolmogorov–Smirnov (for sample size ≥ 5) or Shapiro-Wilk tests (for sample size < 5) for normality or else nonparametric tests, including one sample Wilcoxon test, Wilcoxon matched-pairs signed rank test, Mann-Whitney test, and Kruskal-Wallis test were used. For comparisons across multiple groups and variables, Two-way ANOVA was used without formally testing the normality of data distribution. Following Two-way ANOVA, differences between groups were assessed using Sidak’s multiple comparison test or Tukey’s multiple comparisons test based on the Prism recommendation. When more than two one-sample t-tests were performed, the p values were adjusted using Holm-Šídák correction. *p< 0.05; **p<0.01; ***p<0.001; ****p<0.0001. If not indicated, p>0.05. Error bars represent ± SEM. For detailed statistical results, including exact p values, F values, t values, degree of freedom, and cohort number, see Supplementary Table 1.

Extended Data

Extended Data Fig. 1. One-day 10-min social defeat is sufficient to induce social avoidance of winner-like conspecifics.

a. Experimental timeline.

b. Cartoon illustration of the social interaction (SI) test (top) and a video frame overlaid with DLC tracking results (bottom). Red dot: body center; green dot: head center; yellow dot: nose point; blue dot: cup center.

c. The resident-intruder (RI) test illustration and procedure.

d. The latency to first defeat in male and female mice.

e. The total defeat duration during the 10-min RI test in male and female mice.

f. Video frames from SI tests overlaid with the movement trajectories of the SW mother aggressor (maroon) and C57 test female (pink) during the 1^st (top) and 4^th (bottom) minute of the test.

g. The percentage of time the male (g1) and female (g2) test mice and the aggressors spent on staying stationary (velocity < 1 pixel/frame) over the course of the RI tests.

h. Heatmaps showing the body center location of a representative female mouse during pre-defeat and post-defeat SI tests.

i-j. Representative traces showing the distance between the test animal body center to the cup center (i) and the movement velocity of the test animal (j) in pre- and post-defeat SI tests. Color shades indicate manually annotated behaviors.

k. Distribution of the distance between the test animal’s body center and cup center during the pre-defeat (gray) and post-defeat (color) SI tests for males (k1) and females (k2). Shades shown in gray represent the distance range considered as “around cup”.

l. The percentage of total time the male (l1) and female (l2) test mice spent around the aggressor cup (distance < 250 pixels) during SI tests.

m. The percentage of total time the male (m1) and female (m2) test mice spent investigating the aggressor cup during SI tests.

n. The average duration of each investigation episode of the male (n1) and female (n2) test mice during SI tests.

o. The cup approach frequency of the male (o1) and female (o2) test mice during SI tests.

p. Accumulative plots showing the distribution of movement velocity when the male (p1) and female (p2) test mice are far away (distance > 300 pixels) from the cupped aggressor during the pre-defeat (gray) and post-defeat (color) SI tests.

q. The median velocity when the male (q1) and female (q2) test mice are far away (distance > 300 pixels) from the cupped aggressor during SI tests.

r. Experimental design to test whether social avoidance after defeat is specific to the same aggressor. The aggressor in the RI and SI tests is the same in the AAA paradigm. Different SW aggressors are used for RI and SI tests in the ABA paradigm.

s. Distribution of the distance between the test animal’s body center and cupped aggressor during the pre- and post-defeat SI tests for AAA (s1) and ABA (s2) paradigms.

t. Accumulative plots showing the distribution of movement velocity when the test mice are far away (distance > 300 pixels) from the cupped aggressor during the pre-defeat and post-defeat SI tests in AAA (t1) and ABA (t2) paradigms.

u. The total defeat time during the RI tests in AAA and ABA paradigms is comparable.

v. The change index of time spent around the cupped aggressor during the SI tests in AAA and ABA paradigms. The change index is defined as (P_post-P_pre)/(P_post+P_pre). P_pre and P_post are the behavior parameter values during the pre-defeat and post-defeat SI tests, respectively.

w. The change index of investigation time of the cupped aggressor during the SI tests in AAA and ABA paradigms.

x. The change index of the median movement velocity when the test animal is far away from the aggressor in the SI tests in AAA and ABA paradigms.

Shades and error bars represent ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. Mann-Whitney test (d, v, w), unpaired t-test (e, u, x), two-way repeated measure ANOVA with Sidak’s multiple comparisons test (g), Wilcoxon matched-pairs signed rank test (l1, m1, o1, l2, m2, and n2), and paired t-test (n1, q1, o2, and q2). All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 2. Defeated animals do not avoid conspecifics with genetic backgrounds different from the aggressor.

a. Experimental timeline.

b. The left shows a cartoon illustration and a snapshot of the multi-animal social interaction (MSI) test. The right shows the stimulus animals used for male and female MSI tests.

c-d. Heatmaps showing the body center location of a representative male (c) and a female (d) mouse in pre- and post-defeat MSI tests.

e-f. Total time male test mice spent investigating (e) and around (f) each cupped animal during pre- and post-defeat MSI tests.

g. The number of approaches towards each cup during pre- and post-defeat MSI tests.

h-j. Data from female mice. Plots follow the convention of e-g.

Error bars represent ± SEM. Lines represent individual animals. Numbers on the plots indicate the number of animals. (e-j) Two-way repeated measure ANOVA with Sidak’s multiple comparisons test. All statistical tests are two-tailed. **p<0.01, and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 3. The relationship between OXTR and defeat-induced c-Fos and Esr1 in the VMHvl.

a. Representative histological images showing c-Fos (red) and OXTR (OXTR^ZsGreen, green) expression in the aVMHvl (Bregma: −1.50 mm) and pVMHvl (Bregma: −1.82 mm) in OXTR^Cre × Ai6 (OXTR^ZsGreen) male mice after attack or social defeat. Insets showing enlarged views of the boxed areas. Dashed lines mark the boundary of the aVMH. Scale bars: 50 μm.

b. The number of c-Fos and OXTR double-positive cells after attack and defeat in the aVMHvl (Bregma: −1.34 to −1.50 mm) and pVMHvl (Bregma: −1.66 to −1.82 mm).

c. Representative histological images showing OXTR (green) and Esr1 (red) expression in the aVMHvl (Bregma: −1.50 mm) and pVMHvl (Bregma: −1.82 mm) in OXTR^Cre × Ai6 (OXTR^ZsGreen) male mice. Insets showing enlarged views of the boxed areas. Dashed lines mark the boundary of the VMH. Scale bars: 100 μm.

d. Number of OXTR and Esr1 positive cells in the aVMHvl and pVMHvl.

e. The percentage of Esr1 and OXTR double-positive cells in OXTR positive cells in the aVMHvl and pVMHvl

f. The percentage of Esr1 and OXTR double-positive cells in Esr1 positive cells in the aVMHvl and pVMHvl.

Error bars: ± SEM. (b, d, e and f) n = three 40-μm sections were counted per region per animal, 3 animals per group. Circles in b and lines in d-f represent individual animals. (b and d) Two-way repeated measure ANOVA with Sidak’s multiple comparisons test. (e) Paired t-test. (f) Mann-Whitney test. All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 4. aVMHvl^OXTR cells increase responses to the aggressor after defeat in male mice.

a. Schematics of virus injection and a representative histology image. Dashed line marks the aVMH. Scale bar: 200 μm. Brain illustration in is based on a reference atlas from https://atlas.brain-map.org/.

b. Cartoon illustration of the RI test.

c-e. Representative Z-scored GCaMP6f traces during RI tests with a BC male intruder (c), a SW male intruder (d), or resident SW (e). c2, d2 and e2 show the enlarged boxed areas.

f. Post-event histograms (PETHs) of GCaMP6f signal aligned to initial opponent encounters, only including sessions without defeat or attack during the first 10 sec.

g. Peak GCaMP6f response within the first 10 sec of RI tests.

h-i. PETHs of GCaMP6f signals aligned to close investigation (h) and agonistic interactions (i).

j. Averaged Z-scored responses during various social behaviors.

k. Experimental timeline.

l. Heatmaps showing the body center location of a representative test male during pre- and post-defeat SI tests.

m. Distribution of the distance between the test animal’s body center and cup center during pre- and post-defeat SI tests.

n. The percentage of total time test mice spent around the aggressor cup during pre- and post-defeat SI tests.

o. The percentage of total time the test mice spent investigating the cupped aggressor during pre- and post-defeat SI tests.

p. Representative Z-scored GCaMP6f traces from a recording male mouse during pre-defeat (p1) and post-defeat (p2) SI tests. Shades mark investigation events.

q. PETHs of Z-scored GCaMP6f signals aligned to the onset of investigating aggressor during pre- and post-defeat SI tests.

r. Average Z-scored ΔF/F of aVMHvl^OXTR cells during investigating aggressor in pre- and post-defeat SI tests.

s. Scatter plot showing the correlation between investigation time change index and change in Z-scored GCaMP response to the aggressor during post-defeat SI tests from the pre-defeat level.

t. Representative Z-scored GCaMP trace (black) overlaid with the velocity trace (orange) during the post-defeat SI test. Blue indicates the period when the test animal quickly retreated from the aggressor.

u. PETHs of Z-scored GCaMP signal (black) and velocity (orange) aligned to the retreat onset.

v. The retreat onset GCaMP signal is significantly higher than the retreat offset signal.

w. Representative Z-scored GCaMP trace (black) overlaid with the velocity trace (orange) during the post-defeat SI test. Gray indicates when the test animal stayed immobile and far from the aggressor.

x. PETHs of Z-scored GCaMP signal (black) and velocity (orange) aligned to immobility onset. Immobility trials are defined as velocity < 1 pixel/frame lasting for > 0.5 sec.

y. The mean GCaMP signal during immobility.

Shades and error bars represent ± SEM. Lines and circles represent individual animals. Numbers on the plots indicate the number of animals. One-way ANOVA with Tukey’s multiple comparisons (g), Two-way repeated-measure ANOVA with Sidak’s multiple comparisons (j), Wilcoxon matched-pairs signed rank test (n and o), paired t-test (r, v), one-sample t-test (y), and Pearson cross-correlation (s). All statistical tests are two-tailed. *p<0.05. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 5. Female aVMHvl^OXTR cells increase responses to the aggressor after defeat.

a. Schematics of virus injection and a representative histology image. Dashed lines mark the boundary of the aVMH. Scale bar: 200 μm. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline.

c. Cartoon illustration of the SI test.

d-e. Representative Z-scored GCaMP6f traces from a recording female mouse during RI tests with a non-aggressive naïve C57 (d) and an aggressive lactating SW (e) female mouse. d2 and e2 show the zoomed-in view of the boxed area in d1 and e1, respectively.

f. PETHs of Z-scored GCaMP6f signals aligned to C57 intruder introduction and introduction of the test mouse to the SW lactating female’s cage. Only sessions with no defeat during the first 10 sec are included.

g. The peak GCaMP6f response within the first 10 sec of intruder/resident encounter. Only sessions with no defeat or attack during the first 10 sec are included.

h-i. PETHs of Z-scored GCaMP6f signals aligned to the onset of investigating C57 intruders (h), and investigating and being defeated by SW residents (i).

j. Average Z-scored ΔF/F of aVMHvl^OXTR cells during various social behaviors in the RI tests.

k. Heatmaps showing the body center location of a representative test female during pre-defeat and post-defeat SI tests.

l. Distribution of the distance between the test animal’s body center and cup center during pre- and post-defeat SI tests.

m. The percentage of the total time the test mice spent around the aggressor cup during pre- and post-defeat SI tests.

n. The percentage of the total time the test mice spent investigating the aggressor cup during pre- and post-defeat SI tests.

o. Representative Z-scored GCaMP traces from a recording female during pre-defeat (o1) and post-defeat (o2) SI tests. Shades represent investigation events.

p. PETHs of Z-scored GCaMP signals aligned to the onset of investigating the cupped aggressor during pre- and post-defeat SI tests.

q. Average Z-scored ΔF/F of aVMHvl^OXTR cells during investigating aggressor in pre- and post-defeat SI tests.

r. Representative Z-scored GCaMP trace (black) overlaid with the velocity trace (orange) during the post-defeat SI test. Gray indicates the periods when the test animal stayed immobile and far from the aggressor, and blue indicates a fast retreat event.

s. PETHs of Z-scored GCaMP signal (black) and velocity (orange) aligned to the retreat onset.

t. The GCaMP signal at the retreat onset is significantly higher than the signal at the retreat offset.

u. PETHs of Z-scored GCaMP signal (black) and velocity (orange) aligned to immobility onset. Immobility trials are defined as velocity < 1 pixel/frame lasting > 0.5 sec.

v. The mean GCaMP signal during immobility.

Shades and error bars represent ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. Kruskal-Wallis test with Dunn’s multiple comparisons test (j), Wilcoxon matched-pairs signed rank test (n), unpaired t-test (g), paired t-test (m, q, t), and one-sample t-test (v). All statistical tests are two-tailed. *p<0.05, **p<0.01, and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 6. aVMHvl^OXTR cells increase response to the aggressor after defeat in female mice.

a. Virus injection site and a representative histology image. Scale bar: 200 μm. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline and cartoon illustration of the behavior assay.

c-e. Summary plots showing the investigation time (c), time around each cup (d), and number of approach (e) during pre- and post-defeat MSI tests. E: Empty; Cv: C57 virgin female; Sv: unfamiliar SW virgin female; Sm: SW lactating female aggressor

f. Representative raw traces showing the Z-scored GCaMP6 signal in the pre- (f1) and post-defeat (f2) MSI tests. Shades represent investigation episodes. Empty cup investigation episodes are not marked.

g. PETHs aligned to the investigation onset of different stimuli in pre- and post-defeat MSI tests.

h. The mean Z-scored GCaMP6 signal during investigation of different targets in pre- and post-defeat MSI tests. h1 and h2 are the same data shown in different arrangements.

i. Scatter plots showing the correlation between change index in investigation time and change in Z-scored GCaMP responses to various social targets after defeat from the pre-defeat level.

Error bars and shades: ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. (c, d, e, h) Two-way repeated measure ANOVA with Sidak’s multiple comparisons test. (i) Pearson cross-correlation. All statistical tests are two-tailed. *p< 0.05, **p<0.01, and ***p<0.001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 7. Defeat experience enhances aggressor cue-induced c-Fos in aVMHvl^OXTR cells during subsequent encounters.

a. Experimental design. CCC: SI (Cup)-SI (Cup)-SI (Cup) (top); DDC: RI (Defeat)-RI(Defeat)-SI (Cup) (bottom).

b and c. Representative images showing the expression of OXTR (OXTR^ZsGreen, green) and c-Fos (red) in the aVMHvl (b) and pVMHvl (c) in animals experienced CCC or DDC. Insets showing enlarged views of the boxed areas in the aVMHvl and pVMHvl. Dashed lines mark the boundary of the VMH. Scale bars: 50 μm.

d. The percentage of CCC and DDC-induced c-Fos cells that express OXTR in the aVMHvl and pVMHvl. Numbers indicate the number of animals. Error bars: ± SEM. Circles indicate individual animals. Two-way repeated measure ANOVA with Sidak’s multiple comparisons test. All statistical tests are two-tailed. *p<0.05. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 8. No change in excitability of aVMHvl^OXTR cells one day after defeat.

a. Representative recording traces of aVMHvl^OXTR cells under specific current steps, ranging from 50 pA to 250 pA, from single-housed (SH), defeated (D) and socially interacted (SI) male mice.

b. Frequency-current (F-I) curve of aVMHvl^OXTR cells in SH, D, and SI groups. Two-way repeated measure ANOVA with Sidak’s multiple comparison test. *p<0.05 for D vs. SI comparisons. If not indicated, p> 0.05.

c. Resting Membrane Potential (RMP) of aVMHvl^OXTR cells in SH, D, and SI groups.

d. Rheobase of aVMHvl^OXTR cells in SH, D, and SI groups.

e. Input resistance of aVMHvl^OXTR cells in SH, D, and SI groups.

Error bars in b-e represent ± SEM. Circles in c-e represent individual recording cells. Numbers on the plots indicate the number of cells. Cells are from 3 male mice per group. One-way ANOVA with Tukey’s multiple comparisons test (c) and Kruskal-Wallis test with Dunn’s multiple comparisons test (d-e). All statistical tests are two-tailed. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 9. aVMHvl^OXTR cells do not respond to non-social aversive odors.

a. Virus injection schematics and a representative histology image. Scale bar: 200 μm. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline for testing 2MT responses and cartoon illustration of the behavioral assay.

c. Distribution of the distance between the test animal’s body center and cupped aggressor (gray) or 2MT (orange) during the test. The test animals were defeated one day before the recording.

d. The percentage of the total time the test mice spent around the cupped aggressor or 2MT (distance < 250 pixels) during the test. Only animals that showed clear avoidance of the aggressor (< 20% of total time investigating the aggressor cup) were included in the analysis.

e. The percentage of the total time the test mice spent investigating the aggressor cup or 2MT.

f. Accumulative plots showing the distribution of movement velocity when the test mice are far away (distance > 300 pixels) from the cupped aggressor or 2MT.

g. The median velocity of the test mice when far from the cupped aggressor or 2MT.

h. A representative trace showing continuous Z-scored GCaMP signal during aggressor (magenta) and 2MT (orange) encounters. Shade represents investigation episodes. Dashed lines indicate the aggressor cup and 2MT introduction.

i. PETHs of Z-scored GCaMP signals aligned to the onset of investigating the aggressor and 2MT.

j. Average Z-scored ΔF/F of aVMHvl^OXTR cells during aggressor and 2MT investigation.

k. Virus injection schematics. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

l. Experimental timeline for the shock-odor conditioning and testing. US-paired odor is always delivered after unpaired odor presentation.

m. A representative trace showing the movement velocity of an animal when exposed to shock-paired and unpaired odors one day after shock-odor conditioning. Bars indicate pre-odor and odor delivery periods for calculation in (n).

n. A summary of median velocity before and during odor delivery in the post-conditioning test.

o. A representative Z-scored GCaMP recording trace of an animal when exposed to shock-paired and unpaired odors one day after shock-odor conditioning. Bars indicate pre-odor and odor delivery periods for calculation in (p).

p. A summary of mean GCaMP response (Z-scored ΔF/F) before and during odor delivery in the post-conditioning test.

Error bars and shades: ± SEM. Numbers on the plots indicate the number of animals. Lines represent individual animals. Wilcoxon matched-pairs signed rank test (d); Paired t-test (e, g and j); One-way ANOVA with repeated measures followed by Tukey’s multiple comparisons test (n and p). All statistical tests are two-tailed. *p< 0.05. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 10. aVMHvl^OXTR cells do not respond to non-social aversive cues or noxious somatosensory stimuli in head-fixed animals.

a. (left) Virus injection location and the schematics of head-fixed fiber photometry recording of aVMHvl^OXTR cells. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/. (right) Experimental timelines. The responses to shock-paired and unpaired odors are recorded in one session one day after the shock-odor conditioning. The responses to saline, 2MT, and aggressor urine are recorded in a separate session one day after defeat.

b. Representative raw GCaMP trace of aVMHvl^OXTR cells during delivery of saline (b1), aggressor urine (b2), 2MT (b3), shock-unpaired odor (b4) and shocked-paired odor (b5). All stimuli are presented on Q-tips placed approximately 1 cm in front of the mouse nose for 10 sec.

c. PETHs of Z-scored GCaMP signals aligned to the onset of various odor presentations. Red dotted horizontal lines indicate Z = 0. Gray and black vertical dashed lines indicate the onset and average offset of stimulus presentation.

d. Average Z-scored ΔF/F during various stimulus presentations. Circles represent individual animals. One-way ANOVA with repeated measures followed by Tukey’s multiple comparisons test. *p<0.05, and **p<0.01.

e. Representative raw GCaMP trace of aVMHvl^OXTR cells during gentle touch (e1), back pinch (e2), back poke (e3), and tail pinch (e4). All stimuli were manually delivered.

f. PETHs of Z-scored GCaMP signals aligned to the onset of various somatosensory stimuli. Gray and black vertical dashed lines indicate the onset and average offset of stimulus presentation.

g. Average Z-scored ΔF/F during various stimulus presentations. Lines represent individual animals. One-way ANOVA with repeated measures followed by Tukey’s multiple comparisons test.

Error bars and shades: ± SEM. Numbers on the plots indicate the number of animals. All statistical tests are two-tailed. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 11. Behavior changes induced by optogenetic activation of aVMHvl^OXTR and SOR^OXT cells.

a. Experimental timeline, light delivery protocol, and cartoon illustration of the behavioral assay.

b. Video frames from SI tests overlaid with movement trajectories of a GFP (gray) and a ChR2 animal (green) during interleaved light-on (ON) and light-off (OFF) trials.

c-f. Average distance to the cupped male (c), percentage of time spent investigating the cupped BC male (d), frequency of approaching the cup (e), and the median movement velocity (f) during light-on (blue) and light-off periods in GFP (black) and ChR2 (green) groups. Statistical results were between ON and OFF periods in ChR2 animals. All p>0.05 for GFP animals.

g. Experimental timeline, stimulation protocol, and cartoon illustration of the behavioral test.

h. Representative traces showing the movement velocity of an OXTR^GFP animal (h1) and an OXTR^ChR2 animal (h2) during the light-on (blue shade) and light-off period.

i. Plots showing the distribution of movement velocity during the light-on and light-off periods of GFP (i1) and ChR2 (i2) animals.

j. The percentage of time the animals spent immobile (velocity < 1 pixel/frame).

k. The percentage of time the animals spent flight (velocity > 20 pixels/frame).

l. Experimental timeline and schematic illustration of the real-time place preference test.

m. Heatmaps showing the body center distribution of representative OXTR^GFP and OXTR^ChR2 animals during the 10-min RTPP tests.

n. The percentage of time the animals spent in the light-paired chamber. Circles represent individual animals.

o. Heatmaps showing the body center distribution of representative OXT^GFP and OXT^ChR2 animals during the 10-min RTPP tests.

p. The percentage of time the animals spent in the light-paired chamber.

Shades and error bars represent ± SEM. Lines and circles represent individual animals. Numbers indicate the number of animals. Two-way RM ANOVA with Sidak’s multiple comparisons test (c-f, j-k) and unpaired t-test (n, p). All statistical tests are two-tailed. *p<0.05, **p<0.01, and ***p<0.001. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 12. SOR^OXT affects aVMHvl cell activity by activating OXTR, not glutamatergic synaptic transmission.

a. Strategy to evaluate OXTR knockout efficiency. </P/> b. Representative recording traces of GFP cells from knockout (KO) and control (Ctrl) sides under TGOT perfusion (red bar).

c. Number of aVMHvl cells (from 4 animals) depolarized (> 4 mV) by TGOT in KO and control sides. Chi-square’s test. **p< 0.01.

d. Representative traces showing the membrane potential changes of two aVMHvl cells, one recorded in ACSF (d1) and the other in the presence of 1 μM L-368, 889, a highly specific OXTR antagonist (d2).

e. The number of cells depolarized (ΔRMP > 4 mV) by the SOR^OXT optogenetic stimulation (20 Hz, 1 ms, 5 min) and not. n = 23 (without OXTRA) cells from 5 animals and 23 (with OXTRA) cells from 4 animals. Chi-square’s test. ****p< 0.0001.

f-h. Histology images showing oxytocin (OXT, red) immunostaining and Vglut2 (green) expression in the PVN (f), SON (g), and SOR (h) from Vglut2^Cre × Ai6 male mice. Scale bars: 50 μm.

i. The percentage of OXT-positive cells that express Vglut2 in the PVN, SON, and SOR. Circles represent individual animals. Error bars: ± SEM. n = 3 male mice.

j. Slice recording schematics.

k-l. Representative voltage clamp recording traces from aVMHvl^OXTR cells when a 1 ms light pulse (blue vertical bar) was delivered to activated PVN^OXT (k) or SOR^OXT (l) input.

m. None of the aVMHvl^OXTR cells showed light-evoked EPSC during PVN^OXT (0/14 cells) or SOR^OXT optogenetic activation (0/12 cells).

See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 13. Overlap between OXT and defeat-induced c-Fos.

a. Defeat-induced c-Fos (red) and oxytocin (OXT, green). Scale bars: 50 μm.

b-c. Number of c-Fos+OXT+ cells (b) and percentage of c-Fos+ cells in OXT+ cells (c) in different regions. Every other brain section was counted.

Error bars represent ± SEM. Lines represent individual animals. One-way RM ANOVA with Tukey’s multiple comparisons (b-c). All statistical tests are two-tailed. **p<0.01. See Supplementary Table 1 for detailed statistics.

Extended Data Fig. 14. SOR^OXT cells do not increase responses to aggressors after defeat in male mice.

a. Schematics of virus injection and a representative histological image for fiber photometry recording of SOR^OXT cells in male mice. The dashed line marks SOR. Scale bar: 200 μm. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline.

c. Heatmaps showing the body center location of a recording mouse in MSI tests before and after defeat. E: empty; BC: familiar non-aggressive Balb/C male; C57: unfamiliar C57 male; and SW: SW aggressor.

d. Time spent around each cup during MSI tests before and after defeat.

e-f. Representative Z-scored GCaMP6f traces from a male recording mouse during pre-defeat (e) and post-defeat (f) MSI tests. Shades represent investigation periods of different cupped stimulus animals. Periods investigating the empty cup are not marked.

g. PETHs of Z-scored GCaMP6f signals aligned to the onset of investigation of different cupped stimuli. Gray: pre-defeat; Color: post-defeat.

h. Average Z-scored ΔF/F of SOR^OXT cells during the investigation of various cupped stimuli in the pre-defeat and post-defeat MSI tests.

Shades and error bars represent ± SEM. Lines represent individual animals. Numbers on the plots indicate the number of animals. Two-way repeated measure ANOVA with Sidak’s multiple comparisons (d, h). All statistical tests were two-tailed. See Extended Data Table 1 for detailed statistics.

Extended Data Fig. 15. SOR^OXT cells in female mice are activated by noxious stimuli.

a. Schematics of virus injection and a representative histological image for fiber photometry recording of SOR^OXT cells in female mice. The dashed line marks SOR. Scale bar: 200 μm. Brain illustration is based on a reference atlas from https://atlas.brain-map.org/.

b. Experimental timeline.

c-d. Representative Z-scored GCaMP6f traces of SOR^OXT cells from an animal that encountered a naïve C57BL/6 female intruder (c) or a SW lactating female mouse in the SW cage (d). c2 and d2 show enlarged views of boxed areas in c1 and d1, respectively.

e. PETH of Z-scored GCaMP6f signal aligned to C57 female intruder introduction. As all lactating mothers attacked the test mouse within 10 sec, the introduction response cannot be isolated.

f. The peak GCaMP6f response within the first 10 sec after C57 intruder introduction.

g-h. PETHs of Z-scored GCaMP6f signals aligned to close investigation (CI) of C57BL/6 female intruders (g), investigating and being defeated by SW mothers (h).

i. Average Z-scored ΔF/F of SOR^OXT cells during various social behaviors.

j. Heatmaps showing the body center location of a recording mouse in MSI tests before and after defeat. E: empty; Cv: familiar C57BL/6 virgin female; Sv: unfamiliar virgin SW female; and Sm: SW mother.

k. Time spent around each cup during MSI tests before and after defeat.

l. Representative Z-scored GCaMP6f traces from a female recording mouse during pre-defeat (l1) and post-defeat (l2) MSI tests. Shades represent investigation periods of different cupped stimulus animals. Periods investigating the empty cup are not marked.

m. PETHs of Z-scored GCaMP6f signals aligned to the onset of investigation of different cupped stimuli. Gray: pre-defeat; Color: post-defeat.

n. Average Z-scored ΔF/F of SOR^OXT cells during the investigation of various cupped stimuli in the pre-defeat and post-defeat MSI tests.

o. Schematics of head-fixed fiber photometry recording of SOR^OXT cells and presented stimuli.

p. Representative raw GCaMP6f trace of SOR^OXT cells during delivery of aggressor urine on a Q-tip, gentle touch, back pinch, back poke, and tail pinch.

q. PETHs of Z-scored GCaMP6f signals aligned to the onset of aggressor urine presentation (q1), gentle touch (q2), back pinch (q3), back poke (q4), and tail pinch (q5). Gray and black dashed lines indicate the onset and average duration of stimulus delivery, respectively.

r. Average Z-scored ΔF/F during various stimulus delivery.

Shades and error bars represent ± SEM. Circles and lines represent individual animals. Numbers on the plots indicate the number of animals. Kruskal-Wallis test with Dunn’s multiple comparisons test (i); Two-way repeated measure ANOVA with Sidak’s multiple comparisons test (k, n); and One-way repeated measure ANOVA with Tukey’s multiple comparisons test (r). All statistical tests are two-tailed. *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. See Supplementary Table 1 for detailed statistics.

Data Availability

Raw values associated with each figure panel can be found in the source data files. Fiber photometry recording data, behavior annotations, tracking, and raw representative histology images can be downloaded from 10.5281/zenodo.8417540. Behavior videos and additional histology images are available from the corresponding authors upon reasonable request. They are not deposited to the public database due to their large sizes and size limitation of online depositories. Illustrations of the coronal brain sections were based on the Allen Mouse Brain Atlas https://atlas.brain-map.org/.