A dedicated hypothalamic oxytocin circuit controls aversive social learning

Abstract

To survive in a complex social group, one needs to know who to approach and, more importantly, who to avoid. In mice, a single defeat causes the losing mouse to stay away from the winner for weeks¹. Here through a series of functional manipulation and recording experiments, 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. Before defeat, aVMHvl^OXTR cells minimally respond to aggressor cues. During defeat, aVMHvl^OXTR cells are highly activated and, with the help of an exclusive oxytocin supply from the SOR, potentiate their responses to aggressor cues. After defeat, strong aggressor-induced aVMHvl^OXTR cell activation drives the animal to avoid the aggressor and minimizes future defeat. Our study uncovers a neural process that supports rapid social learning caused by defeat and highlights the importance of the brain oxytocin system in social plasticity.

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.

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.

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.

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.

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.

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.

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.