Serotonin and neurotensin inputs in the vCA1 dictate opposing social valence

Abstract

The ability to evaluate valence of a social agent based on social experience is essential for an animal’s survival in its social group¹. Although hippocampal circuits have been implicated in distinguishing novel and familiar conspecifics^2–7, it remains unclear how social valence is constructed on the basis of social history and what mechanisms underlie the heightened valence versatility in dynamic relationships. Here we demonstrate that the ventral (v)CA1 integrates serotonin (5-HT) inputs from the dorsal raphe and neurotensin inputs from the paraventricular nucleus of the thalamus (PVT) to determine positive or negative valence of conspecific representations. Specifically, during an appetitive social interaction 5-HT is released into the vCA1 and disinhibits pyramidal neurons through 5-HT1_B receptors, whereas neurotensin is released during an aversive social interaction and potentiates vCA1 neurons directly through NTR1s. Optogenetic silencing of dorsal raphe 5-HT and PVT neurotensin inputs into the vCA1 impairs positive and negative social valence, respectively, and excitation flexibly switches valence assignment. These results show how aversive and rewarding social experiences are linked to conspecific identity through converging dorsal raphe 5-HT and PVT neurotensin signals in the vCA1 that instruct opposing valence, and represent a synaptic switch for flexible social valence computation.

Aversive and rewarding social experiences are linked to conspecific identity through converging dorsal raphe 5-HT and paraventricular nucleus of the thalamus neurotensin signals in the vCA1 that instruct opposing valence, representing a synaptic switch for flexible social valence computation.

Main

Many mammals live in social groups and navigating a social environment requires the ability to flexibly update valence of a social agent, unlike the predetermined valence towards predator odour, foot shock or food reward^1,8–10. Owing to the dynamic nature of social relationships, the temporal scales and plasticity may differ from learned valence of inanimate objects and contexts^11–14. Indeed, it is poorly understood how social history influences social valence and how valence information can be flexibly updated. In a changing environment, how does the brain integrate different and even conflicting information to calculate new behavioural decisions? We reasoned that distinct neuromodulatory inputs from converging circuit motifs may represent a potential neural plasticity mechanism, which would allow flexible behavioural switch based on social history. To interrogate this hypothesis, we established new behavioural models to determine the causal relationship between neuromodulator release and social valence.

Unlike innate valence, assessment of social valence has been challenging in rodents, as it is difficult to dissociate conditioning to a non-social stimulus (for example, foot shock pairing) from valence of the social interaction itself^15,16. To address how social valence is regulated we developed new social cognitive models, which enabled us to assess valence association with a conspecific. We presented strong stimuli consisting of a potential mate or an aggressor before a three-chamber social memory assay to elicit rewarding and aversive social experiences (Fig. 1a,b). First, we evaluated whether C57BL/6J test mice show memory of aggressor identity using a standard social memory assay without valenced interactions, as subjects only encounter the aggressive mice under a cup. Indeed, C57BL/6J subjects can distinguish novel and familiar aggressive mice, as evidenced by their preference for the chamber containing the novel mouse (Extended Data Fig. 1a). It is important to dissociate innate valence, such as natural response to predators, from social valence, which can be flexibly assigned based on social history¹. Thus, we examined whether subjects showed any avoidance behaviour towards aggressive mice under a cup and found that without previous direct interaction, test mice showed no preference (Extended Data Fig. 1b). To assess negative social valence assignment, the subject mouse sequentially interacts with a same-sex aggressor or a non-aggressor (male CD-1 or female CFW) for 5 min separated by 5-min intervals (Fig. 1a). Similar to the resident–intruder test^17,18, the interaction with the aggressor is in its home cage whereas the interaction with the non-aggressor is in a new cage. After a 5-min interval, mice were tested for social valence assignment to the aggressor in a three-chamber box, where both aggressor and non-aggressor were confined under a cup. With previous direct aggressor and non-aggressor interaction, both male and female subjects preferred the chamber with the non-aggressor in the subsequent social valence test (Fig. 1a and Extended Data Fig. 1c), thereby demonstrating negative social valence association with the conspecific based on the previously aversive interaction.

Fig. 1. Inhibition of the dCA2 and vCA1 disrupts social memory and valence.

a, Left, schematic of the negative social valence (sv) test. Right, duration subjects spent in chamber containing the non-aggressor with a previous more neutral (ntrl) encounter or in chamber containing the aggressor with a previous more negative (neg) encounter (t₁₇ = 4.091, P = 0.0008, n = 18). b, Left, schematic of positive social valence test. Right, duration subjects spent in chamber containing the opposite-sex partner with a previous more neutral (ntrl) encounter or in chamber containing the potential with a previous more positive (pos) encounter (t₂₅ = 5.98, P = 0.000003, n = 26). c,d, Schematic and representative image of injection site. Social memory test: duration in chamber with familiar mouse (fm) or novel mouse (nm) and discrimination (d) scores (c, t₁₆ = 3.841, P = 0.0014, n = 17; d, t₁₁ = 3.885, P = 0.0025, n = 12). Negative sv, duration in either chamber and d scores (c, t₁₆ = 4.224, P = 0.0006, n = 17; d, t₁₁ = 3.552, P = 0.0045, n = 12). Positive sv, duration in either chamber and d scores (c, t₁₇ = 3.817, P = 0.0014, n = 18; d, t₁₁ = 2.817, P = 0.0168, n = 12). e, Schematic of TRAP2;Ai14 experiment. f, Quantification of tdTomato (tdT)-positive cells in brain regions (ntrl, neg, n = 7; pos, n = 6): medial septum (MS, F_2,17 = 0.3618, P = 0.7016), nucleus of the diagonal band (NDB, P = 0.058), the anterodorsal (AD), anteroventral thalamus (AV) (F_2,16 = 0.8131, P = 0.461, neg n = 6), anterior region of the PVT (aPVT, F_2,17 = 30.08, P = 0.000003), supramammillary nucleus (SUM, F_2,17 = 0.2254, P = 0.8006), dCA2 (F_2,17 = 0.6871, P = 0.5165), vCA1 (P = 0.121), nucleus reuniens (RE) (F_2,17 = 18.36, P = 0.00006), DR (F_2,17 = 18.86, P = 0.00005) and median raphe (MR) (F_2,17 = 15.05, P = 0.0002). Statistical tests: two-tailed paired Student’s t-test (a–d); one-way ANOVA with Tukey’s post hoc test, Kruskal–Wallis test was used for NDB and vCA1 (f). NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Data are mean ± s.e.m. Scale bars, 200 μm (c,d). HC, home cage. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Extended Data Fig. 1. Without prior interaction mice do not show preference in the 3-chamber test.

a, Left, schematic of social memory test. Right, duration C57BL/6J test mice spent in chamber with familiar mouse (fm) or novel mouse (nm), which are both aggressive CD-1s under a cup with which the subjects had no prior direct interaction (t₉ = 2.343, P = 0.0438, n = 10). b, Schematic of experiment and duration C57BL/6J test mice spent in chamber containing an aggressive and non-aggressive CD-1 under a cup without prior direct interaction (t₇ = 0.6644, P = 0.5277, n = 8). c, Left, same data as Fig. 1a, but divided into males and females. Right, discrimination (d) scores (t₁₆ = 0.2009, P = 0.8433, males n = 10, females n = 8). d, Schematic of experiment and duration C57BL/6J male mice spent in chamber containing C57BL/6J female in pro/estrus cycles and not in those cycles without prior direct interaction (t₉ = 0.1644, P = 0.8731, n = 10). e, Left, same data as Fig. 1b, but divided into males and females. Right, d scores (t₂₄ = 0.5349, P = 0.5976, n = 13). Statistical tests: duration (a – e): two-tailed paired Student’s t-test. d scores (c, e): two-tailed unpaired Student’s t-test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m.

Source data

To test positive social valence association, C57BL/6J subjects undergo sequential 5-min direct interactions with a potential mate and a ‘neutral’ partner followed by a 5-min three-chamber test, each separated by 5-min intervals (Fig. 1b). The neutral interaction is less appetitive, but not completely neutral. Male subjects interacted with an age-matched female conspecific in the female partner’s home cage while the experimenter held the partner three times for 3 s to simulate the mating posture¹⁹ and allowed the male to sniff the rear of the female. Without previous direct interaction, male subjects do not show any preference for females in proestrus or oestrus cycles over other cycles in the three-chamber assay (Extended Data Fig. 1d). Since female subjects may not experience the same level of reward during a male partner interaction, we constructed a barrier small enough that the female could escape to another compartment and thus control the timing of the interaction, which has been reported to be rewarding^20,21. The neutral (less positive) experiences are without mating posture or barrier in a new cage. After a direct neutral and potential mate interaction, the male and female subjects spent more time in the chamber with the potential mate in the subsequent social valence test (Fig. 1b and Extended Data Fig. 1e), thereby demonstrating positive social valence association with the conspecific based on the previously appetitive interaction.

Social valence requires the vCA1

The dorsal (d)CA2 and vCA1 are critical for distinguishing novel and familiar conspecifics^2–6,22, but it remains unclear whether they participate in valence association of social representations. To investigate the role of the dCA2 and vCA1 in social valence, we expressed an inhibitory designer receptor exclusively activated by designer drugs (DREADD) hM4Di fused to mCherry (mCh) in the dCA2 of Amigo2-Cre mice, a dCA2-specific Cre-driver line², or vCA1 of C57BL/6J mice and administered the DREADD activator clozapine-N-oxide (CNO) or saline before three different social memory and valence tests. In the social memory test^2,3,7, subjects encounter a novel and a familiar age-matched same-sex conspecific, which are relatively neutral interactions. After saline administration as a within animal control, subjects with hM4Di expression in the dCA2 (Fig. 1c) and vCA1 (Fig. 1d) showed normal social memory as well as negative and positive social valence assignment, as they spent more time in the chamber with the novel mouse, the non-aggressor and the potential mate, respectively. However, the same cohorts of hM4Di-expressing mice receiving CNO showed impaired social memory and valence (Fig. 1c,d). The discrimination scores^3,7, which allowed us to directly compare saline and CNO conditions, confirmed the disruption of both social memory and valence in mice in which dCA2 and vCA1 neuronal activity was inhibited (Fig. 1c,d). As an extra control, we expressed mCh alone in the dCA2 and the vCA1, which did not affect social memory nor valence after saline or CNO administration (Extended Data Fig. 2a,b). Baseline sociability, assessed by the initial phase of the neutral social memory test, was not affected by hM4Di-mediated silencing of the dCA2 and vCA1, nor by expression of mCh in the control cohort (Extended Data Fig. 2c,d). To test whether the positive social valence test is primarily driven by reproductive related behaviours, we assessed the time subjects spent in the chamber with a male or female mouse under a cup. Like saline administration, CNO administration in the hM4Di vCA1 cohort did not decrease the preference of subjects for the chamber with females over males (Extended Data Fig. 2e). Further, the number of times female subjects escaped through the barrier from male partners did not significantly differ after hM4Di-mediated vCA1 inhibition, suggesting that the positive social valence test does not affect innate sexual behaviours (Extended Data Fig. 2f).

Extended Data Fig. 2. Inhibition of the vCA1 does not influence sociability and sexual behaviours.

a, b, social memory: duration subjects spent in chamber with familiar mouse (fm) or novel mouse (nm) and discrimination (d) scores (a: F_3,34 = 6.693, P = 0.0011, hM4Di n = 17, mCh n = 10; b: F_3,35 = 6.257, P = 0.0016, hM4Di n = 12, mCh n = 14). Negative social valence (sv): duration in chamber containing the non-aggressor with a previous ntrl encounter or in chamber containing the aggressor with a previous neg encounter and d scores (a: F_3,34 = 8.018, P = 0.0004, hM4Di n = 17, mCh n = 10; b: F_3,34 = 5.437, P = 0.0037, hM4Di n = 12, mCh n = 13). Positive sv: duration in chamber containing the partner with a previous ntrl encounter or in chamber containing the potential mate with a previous pos encounter and d scores (a: F_3,35 = 4.622, P = 0.0079, hM4Di n = 17, mCh n = 10; b: F_3,34 = 6.909, P = 0.0009, hM4Di n = 12, mCh n = 13). c, d, duration in chamber with a novel object (no) or nm and d scores (c: F_3,34 = 0.8793, P = 0.4615, hM4Di n = 17, mCh n = 10; d: F_3,37 = 3.258, P = 0.0323, hM4Di n = 13, mCh n = 14). e, Left, Schematic of experiment and duration subjects spent in chamber containing a conspecific male or female. Right, d scores (t₉ = 0.3229, P = 0.7541, n = 10). f, Schematic of experiment and the number of times female test mice escaped from males into the separate compartment of the cage within 5 min (t₄ = 0.3203, P = 0.7648, n = 5). Statistical tests, duration (a – f) and d scores (e, f): two-tailed paired Student’s t-test. d scores (a – d): two-way ANOVA with Šidák’s post-hoc test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m.

Source data

Although our results show that information encoded in the dCA2 and vCA1 is important for both positive and negative social valence association, it is possible that an impairment of conspecific identity memory also affects valence evaluation. To determine whether the dCA2 and vCA1 are indeed involved in social valence encoding, independent assessment of valence and identity is necessary. We postulated that different hippocampal inputs may send information mediating positive and negative association and thus be dissociable when manipulated separately. To identify regions that form monosynaptic connections with the dCA2 and vCA1, we used monosynaptic rabies tracing, which yielded labelled cells in the medial septum, the nucleus of the diagonal band, the anterodorsal, anteroventral thalamus, the supramammillary nucleus and the median raphe for both starter regions (Extended Data Fig. 3a–c). Direct connection to the vCA1 was only found in the nucleus of the reuniens, the anterior region of the paraventricular thalamic nuclei paraventricular nucleus of the thalamus (aPVT) and dorsal raphe (DR) (Extended Data Fig. 3b). To uncover dCA2 and vCA1-projecting regions, involved in valence induction, we exposed TRAP2;Ai14 mice to a neutral social partner, a potential mate and an aggressor, and quantified the number of tdTomato-positive cells in the dCA2, vCA1 and all extra-hippocampal input regions (Fig. 1e). Increased numbers of tdTomato-positive cells were detected in the aPVT, reuniens and median raphe only after an aversive interaction and in the DR only after a rewarding interaction (Fig. 1f).

Extended Data Fig. 3. Monosynaptic rabies tracing reveals multiple vCA1 and dCA2 input regions.

a, Schematic of experimental set-up. b, c, Representative images (1 of 3 mice) of injection sites in the (b) vCA1 and (c) dCA2 and respective presynaptic labelling with GFP in the Medial septum (MS), nucleus of the diagonal band (NDB), the anterodorsal (AD), anteroventral (AV) and paraventricular (PVT) thamalamic nuclei, supramammillary nucleus (SUM), nucleus reuniens (RE), dorsal (DR) and median raphe (MR). Scale bar, 200 μm. n = 3. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Distinct inputs mediate opposing valence

To address whether distinct inputs can confer opposing social valence, we used a viral intersectional chemogenetic approach to achieve projection-specific manipulations. As the DR primarily projects to the ventral hippocampus²³ (Extended Data Fig. 3), we injected canine adenovirus type 2 expressing Cre (CAV2-Cre) into the vCA1 and Cre-dependent inhibitory DREADD AAV-DIO-hM4Di-mCh into the DR (Fig. 2a). In two control cohorts, mice expressing hM4Di with saline administration or expressing mCh with saline or CNO administration showed normal social memory (Fig. 2b) as well as negative (Fig. 2c) and positive social valence (Fig. 2d). On the contrary, hM4Di mice that received CNO showed impaired positive social valence, while social memory and negative social valence were unperturbed (Fig. 2b–d). Consistent with manipulations of the vCA1, hM4Di-mediated silencing of vCA1-projecting DR neurons did not alter baseline sociability (Extended Data Fig. 4a). Because the vCA1 is implicated in non-social valence association^24–28, which could influence social valence, we tested whether inhibiting the DR to vCA1 projection induced conditioned place preference (CPP) or aversion. hM4Di-mediated silencing of the vCA1-projecting DR cells had no effect on non-social reward and aversion as test mice spent an equal amount of time on the surfaces paired with CNO or saline (Extended Data Fig. 4b). Neither sociability nor CPP were significantly affected by CNO administration in the mCh control cohort (Extended Data Fig. 4a,b).

Fig. 2. Bi-directional social valence regulation by DR and PVT inputs into the vCA1.

a, Schematic of injection and representative image of AAV-DIO-hM4Di-mCh expression in the DR. b, Left, duration in chamber with familiar mouse (fm) or novel mouse (nm). Right, discrimination (d) scores (F_3,36 = 0.04147, P = 0.9886, n = 13). c, Left, duration in chamber containing non-aggressor with previous neutral encounter or aggressor with previous negative encounter. Right, d scores (F_3,36 = 0.3357, P = 0.7996, n = 13). d, Left, duration in chamber containing partner with previous ntrl encounter or potential mate with previous pos encounter. Right, d scores (F_3,36 = 3.992, P = 0.0149, n = 13). e, Schematic of injection and representative image of AAV-DIO-hM4Di expression in the aPVT. f–h, Left, duration that subjects spent in chamber containing fm or nm and d scores (F_3,26 = 0.2621, P = 0.852, hM4Di n = 11; mCh n = 9) (f), non-aggressor with previous ntrl encounter or aggressor with previous neg encounter and d scores (F_3,25 = 3.556, P = 0.0286, hM4Di n = 10; mCh n = 9) (g) and a partner with previous ntrl encounter or potential mate with previous positive encounter and d scores (hM4Di F_3,26 = 0.7687, P = 0.522, n = 11; mCh n = 9) (h). Statistical tests: duration, two-tailed paired Student’s t-test. d score, two-way ANOVA with Šidák’s post hoc test. Wilcoxin rank test was only used for mCh saline (d) and hM4Di saline duration (h). Data are mean ± s.e.m. Scale bars, 200 μm (a,e). Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Extended Data Fig. 4. Inhibition of MR to dCA2 projection perturbs social memory and inhibition of RE to vCA1 projection has no effect on social memory and valence.

a, Duration that subjects spent in chamber with a novel object (no) or nm and discrimination (d) scores (F_3,36 = 0.617, P = 0.6083, n = 13). b, Duration spent on previously CNO or saline paired surfaces (F_3,34 = 1.948, P = 0.1404, hM4Di: n = 13; mCh: n = 12). c, Duration that subjects spent in chamber with a nm or familiar mouse (fm) and d-scores (t₁₄ = 4.034, P = 0.0012, n = 15). d, Duration that subjects spent in chamber with a novel object (no) or nm and d scores (t₁₄ = 1.346, P = 0.1996, n = 15). e – g, Left, Duration that subjects spent in chamber with (e) fm or nm, (f) non-aggressor with previous ntrl encounter or aggressor with previous neg encounter and d scores, (g) partner with previous ntrl encounter or potential mate with previous pos encounter and d scores (e: t₁₂ = 0.2963, P = 0.772, n = 13; f: t₁₂ = 0.2874, P = 0.7787, saline n = 14, CNO n = 13; g: t₁₃ = 0.9073, P = 0.3807, n = 14). h, Duration that subjects spent in chamber with a no or nm and d scores (t₁₂ = 0.4988, P = 0.627, saline n = 14, CNO n = 13). i, Duration that subjects spent in chamber with a no or nm and d scores (F_3,26 = 2.552, P = 0.0774, hM4Di: n = 11; mCh: n = 9). j, Duration spent on previously CNO or saline paired surfaces (F_3,29 = 0.8432, P = 0.4814, hM4Di: n = 14; mCh: n = 9). k, l, Left, Sample traces of DR (k) and PVT (l) neuron spiking in response to CNO. Right, quantification of spiking (k: uninfected control t₅ = 1.581, P = 0.1747, infected t₅ = 4.583, P = 0.0059, n = 6; l: uninfected control P = 0.75, n = 6, infected P = 0.0156, n = 7). Scale bars, x-axis = 100 ms, y-axis = 10 mV. Statistical tests, duration (a, c – i) and d scores (d – h, k): two-tailed paired Student’s t-test. d scores (a, b, i, j): two-way ANOVA with Šidák’s post-hoc test. l, Wilcoxon rank test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m.

Source data

To identify inputs mediating negative social valence, we used the same viral intersectional strategy to chemogenetically manipulate projections of the median raphe to the dCA2, reuniens to the vCA1 and aPVT to the vCA1 based on connectivity found in our tracing experiment. Inhibition of the median raphe to dCA2 projection impaired social memory but not sociability (Extended Data Fig. 4c,d), whereas inactivation of the reuniens to vCA1 projection had no effect on social memory, social valence nor sociability (Extended Data Fig. 4e–h). hM4Di-mediated silencing of the aPVT to vCA1 projection (Fig. 2e) specifically ablated negative social valence without affecting social memory and positive social valence (Fig. 2f–h). Saline administration in the same hM4Di cohort or CNO administration in the mCh control cohort did not influence social memory nor valence (Fig. 2f–h). No manipulations of vCA1-projecting aPVT neurons in hM4Di or mCh-expressing cohorts affected baseline sociability and CPP (Extended Data Fig. 4i,j). Similarly to previously reported studies^29,30, bath application of CNO significantly reduced the excitability of hM4Di-expressing DR and PVT cells compared to non-expressing control cells (Extended Data Fig. 4k,l). Together, these results indicate that positive and negative social valence information converge in the vCA1 through DR and PVT inputs.

Social valence requires 5-HT and NT

We questioned how vCA1 neurons dissociate opposing valence information. To explore the possibility that distinct neuromodulators facilitate memory valence encoding in the same brain region, we assessed differential neuromodulatory signals from the DR and PVT. As the DR and PVT are major release sites for 5-HT (refs. ^23,31) and neurotensin (NT)^9,32, respectively, we tested whether DR 5-HT and aPVT NT neurons send projections to the vCA1. Consistent with previous reports^9,23,33, expressing Cre-dependent GFP in the DR of Sert-Cre mice (Extended Data Fig. 5a) and Cre-dependent mCh in the aPVT of NT-Cre mice (Extended Data Fig. 5b) revealed 5-HT and NT axon innervation in the vCA1. To assess the involvement of DR 5-HT neurons in positive social valence induction, we expressed Cre-dependent NpHR3.0eYFP or enhanced yellow fluorescent protein (eYFP) alone in the DR and implanted optical fibres adjacent to the vCA1 (Fig. 3a). Optogenetic inhibition of DR 5-HT inputs in the vCA1 did not alter social memory (Extended Data Fig. 5c) and negative social valence (Fig. 3b) but only blocked positive social valence assignment (Fig. 3c). The same manipulation did not affect baseline sociability (Extended Data Fig. 5d) nor induce real-time place preference (RTPP) (Extended Data Fig. 5e), suggesting no change in innate reward and aversion. Moreover, the NpHR3.0 cohort without light stimulation (light off) or the eYFP cohort with or without light stimulation (light on or off) showed no significant changes in social memory, social valence, sociability and RTPP (Fig. 3b,c and Extended Data Fig. 5c–e).

Extended Data Fig. 5. Inhibition of DR 5-HT and PVT NT inputs into the vCA1 do not influence social memory, sociability, and real time place preference (RTPP).

a, Schematic of experiment and representative images of injection site (1 of 3 mice) in the DR of Sert-Cre animals and axon innervation in the vCA1, n = 3. b, Schematic of experiment and representative images of injection site (1 of 3 mice) in the aPVT of NT-Cre animals and axon innervation in the vCA1, n = 3. c, Left, duration mice spent in chamber with familiar mouse (fm) or novel mouse (nm). Right, discrimination (d) scores (P = 0.1833, NpHR3.0: n = 13; eYFP: on n = 15, on n = 16). d, Duration that subjects spent in chamber with a novel object (no) or nm and d) scores (P = 0.4517, NpHR3.0: n = 13; eYFP: on n = 14). e, Average time spent in light paired or unpaired chambers in initial and reversal trials (F_3,22 = 1.602, P = 0.2174, NpHR3.0: n = 13; eYFP: n = 6). f, Left, Duration that subjects spent in chamber with fm or nm encounter and d scores (F_3,34 = 0.8096, P = 0.4974, NpHR3.0: n = 14; eYFP: off n = 12, on n = 11). g, Duration that subjects spent in chamber with a novel object (no) or nm and d scores (F_3,34 = 0.1939, P = 0.8998, NpHR3.0: n = 14; eYFP: off n = 12, on n = 11). h, Average time spent in light paired or unpaired chambers in initial and reversal trials (F_3,28 = 1.649, P = 0.2005, NpHR3.0: n = 10; eYFP: n = 11). i, Time courses of average RCaMP2 transient z scores event-locked to direct interactions and quantification of peak z score during interaction (Top: t₂ = 30.31, P = 0.021, n = 2; Bottom: t₂ = 24.95, P = 0.0016, n = 3). j, Left, time courses of average EGFP-CAAX transient z scores event-locked to direct interactions. Right, quantification of peak z score during interaction (F_2,9 = 0.2688, P = 0.7702, n = 4). k, Left, representative images of AAV-sgRNA-EGFP expression in the DR with Tph2 IHC. Top right, schematic of virus injection. Bottom right, quantification of Tph2 fluorescence intensity (P = 0.0286, n = 4). l, Schematic of qPCR experiment. Fold change of Tph2 mRNA level in DR primary cultured neurons infected with AAV-Cre and AAV-sgTph2 or AAV-sgNTC, t₄ = 5.027, P = 0.0073, n = 3, biological replicates. m, Left, duration that subjects spent in chamber with a novel object (no) or nm and d scores (t₁₆ = 0.6332, P = 0.5355, Tph2: n = 10; NTC: n = 8). Right, Duration that subjects spent in chamber with a nm or familiar mouse (fm) and d scores (t₁₆ = 1.171, P = 0.2587, Tph2: n = 10; NTC: n = 8). Statistical tests, duration (c, d, f, g, i, m): two-tailed paired Student’s t-test. (c, f, g) d scores and e, h: two-way ANOVA with Šidák’s post-hoc test. (d) d scores: Kruskal-Wallis test with Dunn’s post-hoc test. j, one-way ANOVA with Tukey’s post-hoc test. k, Mann-Whitney test. l, (m) d score: two-tailed unpaired Student’s t-test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m. Scale bars, 200 μm. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Fig. 3. Opposing social valence regulation by DR 5-HT neurons and PVT NT neurons converging in the vCA1.

a,d,j, Schematic and representative image of virus injection in the DR (a,j) and PVT (d). b,e, Left, duration in chamber with ntrl partner or chamber with neg valenced partner (Sert-Cre (b); NT-Cre (e)). Right, d scores (b, F_3,36 = 0.0446, P = 0.9873, NpHR3.0 n = 12; eYFP off n = 16, on n = 15; e, F_3,35 = 6.717, P = 0.0011, NpHR3.0 n = 14; eYFP: n = 12). c,f, Left, duration in chamber with ntrl partner or chamber with positive valenced partner. Right, d scores (c, F_3,39 = 7.304, P = 0.0005, NpHR3.0 n = 13; eYFP P = 0.553, off n = 17, on n = 16; f, P = 0.512, NpHR3.0 n = 14. eYFP off n = 11, on n = 12). g, Schematic of experiment. h,i, Left, time courses of average GRAB sensor transient z scores event-locked to direct interactions. Right, quantification of peak z score during interaction (h, F_2,23 = 9.181, P = 0.0012, ntrl, n = 8; neg, pos, n = 9; i, F_2,27 = 13.87, P = 0.000072, n = 10). k, Left, duration in chamber with neutral partner or chamber with negative valenced partner and d scores (t₁₆ = 1.1, P = 0.2874, Tph2: n = 10; NTC: n = 8). Right, duration in chamber with neutral partner or chamber with positive valenced partner and d scores (t₁₆ = 3.335, P = 0.0042, Tph2 n = 10; NTC n = 8). l, Representative images of virus expression in the vCA1 with TPH2 immunohistochemistry shown in magenta. Quantification of the percentage of the TPH2 area (t₁₆ = 3.752, P = 0.0017, Tph2: n = 10; NTC: n = 8). Statistical tests: duration (b,c,e,f,k); two-tailed paired Student’s t-test, Wilcoxin rank test was only used for YFP off duration (b,f); d score (b,c,e), two-way ANOVA with Šidák’s post hoc test; d score (f), Kruskal–Wallis test with Dunn’s post hoc test; one-way ANOVA with Tukey’s post hoc test (h,i); d score (k,l), unpaired Student’s t-test. Data are mean ± s.e.m. Scale bars, 200 μm (a,d,j), 100 μm (l). Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

To address whether PVT NT neurons exert opposing effects, we injected AAVs expressing Cre-dependent NpHR3.0 in the aPVT and implanted optic fibre cannulas above the vCA1 to silence PVT NT inputs in the vCA1 (Fig. 3d). Optogenetic inhibition during initial interaction and three-chamber testing specifically impaired negative social valence association (Fig. 3e) without affecting social memory (Extended Data Fig. 5f) and positive social valence (Fig. 3f). Again, light-off conditions in the NpHR3.0 cohort or light on and off conditions in the eYFP cohort did not alter social memory and valence (Fig. 3e,f). Baseline sociability and RTPP were not significantly affected by those manipulations (Extended Data Fig. 5g,h). To evaluate NpHR3.0-mediated silencing, we expressed Cre-dependent NpHR3.0eYFP and RCaMP2, a calcium indicator, in the DR or PVT of the respective Cre-driver lines and implanted optogenetics fibre cannulas above the injection sites as well as fibre photometry cannulas above the vCA1. Light stimulation of NpHR3.0 strongly diminished terminal activity during social interactions in both cohorts (Extended Data Fig. 5i).

Our results predict enhanced 5-HT release into the vCA1 during an appetitive social interaction and increased NT release during an aversive one. Therefore, we expressed GRAB_5HT3.5 or GRAB_NT1.0 (ref. ⁹) in the ventral hippocampus and placed fibre implants adjacent to the vCA1 to record 5-HT or NT activity while the subject was exposed to sequential interactions with a same-sex age-matched conspecific, a potential mate and an aggressor (Fig. 3g). Robustly elevated 5-HT activity was detected only during the appetitive interaction, whereas interactions with the neutral and aversive partners produced smaller increases in 5-HT activity (Fig. 3h). By contrast, NT signals showed a marked increase only during an aversive interaction, with neutral and rewarding social encounters only generating moderate NT activity (Fig. 3i). When we expressed a membrane-bound enhanced green fluorescent protein (EGFP) as a control, we did not detect any significant changes in fluorescence between different social encounters (Extended Data Fig. 5j).

Because DR 5-HT neurons also corelease glutamate into the vCA1 (ref. ²³), we used CRISPR interference (CRISPRi) combined with a viral intersectional strategy to specifically knockdown the 5-HT producing enzyme tryptophan hydroxylase 2 (encoded by Tph2) in vCA1-projecting DR neurons. Initially, we validated knockdown efficiency by expressing AAVs carrying three single-guide RNAs (sgRNAs) targeting Tph2 (AAV-sgTph2) or non-targeting control sgRNAs (AAV-sgNTC) along with AAV-Cre in the DR of dCas9-KRAB mice, where dCas9 expression can be turned on by Cre recombination. Immunostaining for TPH2 in the DR revealed a roughly 75% reduction in fluorescence intensity in sgTph2-expressing mice compared to non-targeting controls (Extended Data Fig. 5k). To further quantitatively assess knockdown efficiency, we performed quantitative PCR (qPCR) on DR primary cultured neurons infected with AAV-Cre and AAV-sgTph2 or AAV-sgNTC, which revealed a 98% reduction in Tph2 messenger RNA (mRNA) level after knockdown (Extended Data Fig. 5l). We then injected AAV-sgTph2 into the DR and CAV2-Cre into the vCA1 of dCas9-KRAB mice (Fig. 3j). As a control, we expressed sgNTC in the DR and CAV2-Cre in the vCA1 in a separate cohort of mice. Knockdown of Tph2 in vCA1-projecting DR neurons only ablated positive social valence but not negative social valence, social memory and sociability (Fig. 3k and Extended Data Fig. 5m). No changes in those behaviours were detected in control mice expressing non-targeting control sgRNA (Fig. 3k and Extended Data Fig. 5m). To further confirm knockdown, we assayed TPH2 expression in DR terminals in the vCA1 of the behavioural cohort and found a roughly 80% reduction in TPH2-positive terminals in sgTph2 mice compared to non-targeting controls (Fig. 3l).

5-HT1_BR and NT1R potentiate vCA1 cells

Our data indicate that 5-HT and NT promote opposing social valence. Since DR 5-HT and PVT NT neurons can corelease other neuromodulators^9,23, we assessed whether perturbing 5-HT and NT action is sufficient to block social valence evaluation by directly inhibiting 5-HT and NT receptors in the vCA1. To uncover receptor subtypes involved, we locally infused receptor (R) antagonists (A) into the vCA1 20 min before social memory tests (Fig. 4a). NTR1-A (SR48692) but not NTR2-A (NTRC 824) micro-infusion blocked negative social valence (Fig. 4b). Although the NTR1-A (SR48692) did not influence positive social valence, a promiscuous 5-HTR-A (methiothepin) impaired it (Fig. 4c). Next, we targeted the 5-HT1_BR, as it is known to be involved in sociability and social memory when released into the nucleus accumbens³¹ and medial septum⁷, respectively. Indeed, local infusion of the 5-HT1_BR-A (NAS-181) blocked positive (Fig. 4c) but not negative social valence (Fig. 4b). To examine the specificity of this receptor, we also infused an antagonist against 5-HT1_AR (NAD299). The results show a P value of 0.07, suggesting that 5-HT1_AR is most probably not involved in positive social valence regulation (Fig. 4c). Neither NTR1-A nor 5-HT1_BR-A infused into the vCA1 had detectable effects on social memory (Fig. 4d). When comparing discrimination scores, NTR1 antagonist micro-infusion into the vCA1 seem to specifically impair negative social valence, whereas 5-HT1_BR antagonist infusion specifically blocked positive social valence (Extended Data Fig. 6a).

Fig. 4. Regulation of opposing social valence by vCA1 5-HT1_BRs and NTR1s.

a, Schematic and representative image. b–d, Left, duration in chamber with ntrl or neg valenced partner (b, saline t₁₁ = 2.98, P = 0.0125; NTR1-A t₁₁ = 0.1196, P = 0.9070; NTR2-A t₁₁ = 2.853, P = 0.0157; 5-HT1_BR-A t₁₁ = 3.035, P = 0.0113, n = 12), ntrl or pos valenced partner (c, saline t₁₂ = 2.447, P = 0.0308; NTR1-A t₁₂ = 2.295, P = 0.0405; 5-HTR-A t₁₂ = 0.4592, P = 0.6543; 5-HT1_BR-A t₁₂ = 0.8693, P = 0.4017; 5-HT1_AR-A P = 0.071, n = 13), or fm or nm (d, saline P = 0.0002, n = 14; NTR1-A t₁₁ = 2.918, P = 0.014, n = 12; 5-HT1_BR-A t₁₂ = 2.307, P = 0.03967, n = 13). e–g, Time courses of GCaMP6f transient z scores event-locked to direct interactions. Right, peak z score quantification (e, F_3,32 = 6.938, P = 0.001, n = 9; f, F_3,32 = 17.1, P = 8.32 × 10⁻⁷, n = 9; g, F_3,20 = 14.47, P = 3.94 × 10⁻⁶, n = 9). e, Left, schematic and representative image. h,i, Schematic of experiment and pie charts of cell responses (h, n = 10, i, n = 11) with representative traces. j, Left, sample traces of vCA1 pyramidal neuron spiking in response to drug. Right, quantification of spiking (P = 3 × 10⁻²¹, predrug n = 14, PD n = 11, CP n = 8). CP versus baseline results have an extra * at 60, 100–140 pA. k,l, Left, summary time course of EPSCs and IPSCs. Right, average PSC (percentage of control) with sample traces at time points 1 and 2 (k, t₁₁ = 2.92, P = 0.0139, EPSC n = 6, IPSC n = 7; l, t₁₄ = 3.076, P = 0.0082, EPSC, IPSC n = 8). Scale bars: 200 μm (a,e); 20 ms (x axis), 25 pA (y axis) (k,l); 100 ms (x axis), 10 mV (y axis) (h,i). Statistical tests: two-tailed paired Student’s t-test (b–d); one-way ANOVA with Tukey’s post hoc test (e–g); two-way ANOVA mixed-effects with Tukey’s post hoc test (j); two-tailed unpaired Student’s t-test (k,l); Wilcoxon rank test was only used for NAD299 (b) and saline (c). Data are mean ± s.e.m. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Extended Data Fig. 6. NTR1 agonist increases sEPSC amplitude, while 5-HT1_BR agonist enhances sIPSC frequency.

a, Discrimination scores between different drug infusions into the vCA1 from Fig. 4b,c where negative social valence (sv) and positive sv were tested (negative sv: F_2,26 = 3.353, P = 0.0436, n = 12; negative sv: F_2,32 = 1.567, P = 0.2196, n = 13). b, c, Percentage of cells from Fig. 4k,l which showed a change or no change in their evoked EPSC/IPSC amplitudes after bath-application of PD149163 (PD) or CP93129 (CP). d, Left, recordings of spontaneous EPSCs (sEPSCs) before and after bath-application of PD and cumulative probability plot of sEPSC amplitudes with representative traces (1 of 10 cells) above (t₉ = 2.628, P = 0.0274, n = 10). Scale bars, x-axis = 50 ms, y-axis = 10 pA. Right, cumulative probability plot of sEPSC inter-event intervals with representative traces (1 of 10 cells) above (t₉ = 1.333, P = 0.2154, n = 10). Scale bars, x-axis = 0.50 s, y-axis=10 pA. e, Left, recordings of sIPSCs before and after bath-application of PD and cumulative probability plot of sIPSC amplitudes with representative traces (1 of 8 cells) above (t₇ = 0.6492, P = 0.5369, n = 8). Right, cumulative probability plot of sIPSC inter-event intervals with representative traces (1 of 8 cells) above (t₇ = 1.823, P = 0.111, n = 8). f, Left, recordings of spontaneous IPSCs (sIPSCs) before and after bath-application of CP and cumulative probability plot of sIPSC amplitudes with representative traces (1 of 9 cells) above (t₈ = 0.1498, P = 0.8846, n = 9). Scale bars, x-axis = 50 ms, y-axis = 10 pA. Right, Cumulative probability plot of sIPSC inter-event intervals with representative traces (1 of 9 cells) above (t₈ = 2.94, P = 0.0187, n = 9). Scale bars, x-axis = 0.50 s, y-axis = 10 pA. Scale bars, x-axis = 50 ms, y-axis = 10 pA. g, Left, recordings of sEPSCs before and after bath-application of CP and cumulative probability plot of sEPSC amplitudes with representative traces (1 of 10 cells) above (t₉ = 0.4582, P = 0.6577, n = 10). Right, Cumulative probability plot of sEPSC inter-event intervals with representative traces (1 of 10 cells) above (t₉ = 0.2687, P = 0.7942, n = 10). Scale bars, x-axis = 50 ms, y-axis = 10 pA. h, Left, Schematic and representative images (1 of 4 mice) of in-situ hybridization in the vCA1. Right, pie chart of cells positive and negative for Gad2 and Htr1b along with a smaller pie chart and bar graph showing the proportion of Gad2+ cells positive and negative for Htr1b. n = 4. Scale bar, 50 μm. Statistical tests: a, one-way ANOVA with Dunnett’s post-hoc test, d–g, two-tailed paired Student’s t-test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m.

Source data

Our data indicate that NTR1-A and 5-HT1_BR-A can regulate vCA1 neuronal activity during differently valenced social interactions. To evaluate this hypothesis, we expressed a calcium indicator GCaMP6f in the vCA1 and recorded vCA1 neuronal activity during sequential novel object, neutral, negatively and positively valenced social interactions. Like previous studies^3,34, we observed a robust increase of vCA1 activity, during all types of social interaction compared to object interaction (Fig. 4e). Administration of NTR1-A specifically decreased the activity during an aversive social interaction (Fig. 4f), whereas 5-HT1_BR-A administration only attenuated the activity during a rewarding one (Fig. 4g). To assess the effect of NTR1-A on PVT inputs, we expressed channelrhodopsin2 (ChR2) in the aPVT of C57BL/6J mice and performed ex vivo optogenetic recordings in the vCA1. Light stimulation evoked action potentials in half of vCA1 cells and bath application of the NTR1-A (SR48692) significantly reduced responses (Fig. 4h). To evaluate DR inputs, we used the same approach and expressed ChR2 in the DR. Light stimulation induced responses in more than 90% of vCA1 cells and 5-HT1_BR-A (NAS-181) application strongly diminished responses (Fig. 4i).

To interrogate how 5-HT and NT release may cause an increase in vCA1 activity during different social encounters, we performed ex vivo whole-cell recordings to assess the excitability of vCA1 cells in response to NTR1 (0.5 µM PD149163) and 5-HT1_BR (5 µM CP93129) agonists. Bath application of both NTR1 (PD) and 5-HT1_BR (CP) agonists increased vCA1 neuronal spiking in response to depolarizing current pulses (Fig. 4j). To address if both receptors work through the same or different synaptic mechanisms, we further assessed evoked and spontaneous postsynaptic currents. Because 5-HT1_BRs are commonly presynaptic and inhibit neurotransmitter release³⁵ but NTR1s have been reported to be postsynaptic and excitatory^36,37, we predicted differential synaptic mechanisms. Consistent with our hypothesis, bath application of the specific NTR1 agonist, PD149163, increased evoked excitatory postsynaptic currents (EPSCs) and had minimal impact on evoked inhibitory postsynaptic currents (IPSCs) (Fig. 4k). Further, a higher percentage of cells showed a change in evoked EPSC compared to IPSC amplitude in response to PD149163 (Extended Data Fig. 6b). Likewise, bath application of the 5-HT1_BR agonist, CP93129, decreased evoked IPSCs but had no significant effects on evoked EPSCs (Fig. 4l), which suggests that the increase in excitability is due to depression of inhibitory inputs onto the vCA1 pyramidal neurons. Consistently, a greater proportion of cells showed a change in IPSC compared to EPSC amplitude in response to CP93129 (Extended Data Fig. 6c). To assess pre- versus postsynaptic mechanisms, we measured spontaneous postsynaptic currents (sPSCs). Bath application of PD149163 increased sEPSC amplitude, but not frequency (Extended Data Fig. 6d) and had minimal effects on sIPSC amplitude and inter-event interval (Extended Data Fig. 6e). By contrast, CP93129 increased sIPSC frequency, but not amplitude (Extended Data Fig. 6f), and did not influence sEPSC amplitude and frequency (Extended Data Fig. 6g). The results show that NT increases excitability by directly exciting vCA1 pyramidal neurons, whereas 5-HT acts through disinhibition by decreasing the inhibitory input onto them. To assess whether inhibitory neurons in the vCA1 express 5-HT1_BRs, we performed in situ hybridization with probes targeting the γ-aminobutyric acid-expressing (GABAergic) neuron marker, Gad2 and 5-HT1_BR (Htr1b). Quantification of Gad2⁺ cells revealed that 44% express Htr1b (Extended Data Fig. 6h).

Next, we tested whether positive and negative social valence flexibly mapped onto the same, overlapping or distinct populations of cells. Therefore, we tagged cells activated by a negative or positive social experience with tdTomato and used cFos immunostaining to label re-activated cells in the same or different social context (Extended Data Fig. 7a). Quantification of tdTomato-positive cells stained with cFos antibody revealed a 64–77% overlap when animals were exposed to the same social context, whereas the overlap decreased by roughly 40% when the second context had the opposite social valence (Extended Data Fig. 7b). We reasoned that these distinct cell populations encoding opposing social valence could be due to regionally separated innervation of differential neuromodulator inputs and/or heterogeneity in receptor expression. Hence, we examined DR and PVT innervation of the vCA1 by expressing GFP in the aPVT and mCh in the DR of the same brains and found overlapping but not identical areas of innervation (Extended Data Fig. 7c). Then, we evaluated receptor composition of cells encoding positive or negative social valence by using in situ hybridization with probes targeting NTR1 (Ntsr1) and 5-HT1_BR (Htr1b) in TRAP2;Ai14 vCA1 cells that were labelled by tdTomato after a rewarding or aversive social experience (Extended Data Fig. 7d). Quantification of tdTomato cells positive for Ntsr1, revealed that roughly 80% of cells participating in negative social valence express Ntsr1, whereas only roughly 45% of cells participating in positive social valence showed expression (Extended Data Fig. 7e). There is also a trend towards a decrease, from roughly 40 to 30%, in the proportion Htr1b-expressing cells active during an aversive compared to an appetitive social interaction (Extended Data Fig. 7e). The data suggest that separate populations of vCA1 pyramidal neurons may be preferentially potentiated for distinct valence associations.

Extended Data Fig. 7. DR and PVT axon innervation as well as 5-HT1_BR and NT1R expression in the vCA1 are not homogenous.

a, Schematic of one experimental condition and representative images (1 of 2 for neg-neg; 1 of 3 for neg, neg-pos, pos) displayed as merge with DAPI, tdTomato and cFos (Alexa 488). neg: t₃ = 11.84, P = 0.0013, neg-neg: n = 2, neg-pos: n = 3, pos: t₄ = 4.546, P = 0.0104, n = 3. Scale bars, 100 μm. b, percentage of tdTomato (tdT) and cFos double positive cells. F_3,7 = 26.29, P = 0.0003, neg-neg: n = 2, neg-pos, pos-pos, pos-neg: n = 3. c, Left, schematic of experiment and representative images (1 of 2 mice) of dual injection sites in the DR and aPVT. Right, representative images of axon innervation in the vCA1. n = 2. Scale bars, 200 μm. d, Schematic of experiment. e, Left, representative images of in situ hybridization in the vCA1 (1 of 3 mice). Right, quantification of triple (above) and double (below) positive cells (F_3,8 = 34.723, P = 0.00006, n = 3). Scale bars, 100 μm. Statistical test, one-way ANOVA with Tukey’s post-hoc test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Social valence inception by 5-HT and NT

Having established that social valence is encoded in cells that are potentiated by 5-HT and NT, we proposed that endogenous release of those neuromodulators can attribute valence to the memory of a conspecific. To determine whether we could invert the polarity of social valence, we expressed Cre-dependent ChR2-mCh in DR 5-HT neurons of Sert-Cre mice. Stimulation of DR 5-HT inputs in the vCA1 during the aggressor interaction disrupted the preference for the non-aggressor in the subsequent valence test, which was normal without light stimulation (Fig. 5a, left). Because NT is still released during the aversive social interaction, it is possible that release of both 5-HT and NT signals contradictory valence and causes the ‘impairment’ phenotype. Blocking NT signalling might shift the balance in favour of positive social valence. Consistent with this prediction, simultaneous NTR1-A (SR48692) administration and 5-HT release during an aversive social interaction completely switched social valence, whereas administration of NTR1-A without light stimulation inhibited positive social valence (Fig. 5a, right). 5-HT release during the entire duration of the sociability and social memory tests caused no behavioural effects (Extended Data Fig. 8a,b). Likewise, this manipulation did not significantly influence RTPP (Extended Data Fig. 8c), nor male and female sexual behaviours (Extended Data Fig. 8d,e).

Fig. 5. DR 5-HT and PVT NT released into the vCA1 induce opposing social valence.

a, Top, schematic of experiment. Bottom left, schematic and representative image of injection site. Bottom right, duration in chamber with ntrl or neg valenced partner and discrimination scores (F_2,24 = 3.253, P = 0.0483, n = 11). b, Top, schematic of experiment. Bottom left, schematic and representative image of injection site. Bottom right, duration in chamber with ntrl or pos valenced partner and discrimination scores (F_2,21 = 11.11, P = 0.0004, n = 10). c, Schematic of experiment. d–g, Left, duration in chamber containing mouse (m) 1 or object (o) 1 without previous light pairing, or in chamber containing m2 or o2, with previous light stimulation during the 10-min social interaction phase. Right, discrimination scores (d,f) mouse (d, t₁₁ = 3.733, P = 0.0039, off n = 12, on n = 11; f, t₉ = 2.483, P = 0.0348, n = 10), (e,g) object (e, t₁₁ = 1.432, P = 0.1801, n = 12; g, t₉ = 0.9567, P = 0.3637, n = 10). m1/o1 and m2/o2 are the same mice and objects in the same experiment, but not between experiments. h, Schematic of experiment. i, Left, duration in chamber with ntrl or pos valenced partner. Right, discrimination scores (F_2,25 = 5.504, P = 0.0105, saline n = 10, CP, WT, n = 9). j, Left, sample traces of vCA1 pyramidal neuron spiking. Right, quantification of spiking (P = 2 × 10⁻¹⁵), saline, CP n = 15, WT n = 14). Asterisks are depicted for CP versus pre-CP. Wild-type (WT) versus pre-CP **40 pA, ***80–160 pA. Scale bars: 200 μm (a,b,h); 100 ms (x axis), 20 mV (y axis) (j). Statistical tests: duration (a,b,d,f,g,i), d scores; two-tailed paired Student’s t-test (d–g); one-way ANOVA with Tukey’s post hoc test (a,b,i); two-way ANOVA mixed-effects with Tukey’s post hoc test (j); Wilcoxin rank test was used for NTR1-A off duration (a) and duration (e). Data are mean ± s.e.m. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

Extended Data Fig. 8. Excitation of DR 5-HT and PVT NT inputs in the vCA1 does not alter social memory nor innate valence.

a, Duration that subjects spent in chamber with a novel object (no) or novel mouse (nm) and discrimination (d) scores (t₁₁ = 0.1971, P = 0.8474, n = 12). b, Duration that subjects spent in chamber with a nm or familiar mouse (fm) and d scores (t₁₁ = 0.7047, P = 0.4956, n = 12). c, Average time spent in light paired or unpaired chambers in initial and reversal trials (t₁₁ = 0.7037, P = 0.4963, n = 12). d, Number of times male subjects attempted to mount females in 5 min (t₅ = 0.3492, P = 0.7412, n = 6). e, Female subjects interacted with male partners in their home cage. d score of duration female subjects spent in the compartment with male partners or in the empty compartment (t₅ = 0.2499, P = 0.8126, n = 6). f, g, Duration that subjects spent in chamber with (f) no or nm (t₉ = 0.5448, P = 0.5991, n = 10) and (g) fm or nm (t₉ = 0.1835, P = 0.8584, n = 10) and d scores. h, Average time spent in light paired or unpaired chambers in initial and reversal trials (t₉ = 0.7234, P = 0.4878, n = 10). i, Time subjects spent in open and closed arms of the elevated plus maze (F_1.3,12.1 = 159.5, P = 8 × 10⁻⁹, n = 10). j, Schematic of experiment. Scale bar, 200 μm. k, l, Duration in chamber with (k) no or nm (F_2,24 = 0.607, P = 0.5531, n = 9) and (l) with nm or fm (F_2,24 = 2.207, P = 0.1319, n = 9) and d scores. Statistical tests: a – h, two-tailed paired Student’s t-test. i, (k, l) d scores, one-way ANOVA with Tukey’s post-hoc test. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Mean values are depicted. Error bars denote s.e.m. Brain image: Jonas Töle/Wikimedia (CC0 1.0).

Source data

To test bi-directionality, we expressed Cre-dependent ChR2-mCh in the aPVT of NT-Cre mice and placed optical fibres adjacent to the vCA1. When the light was off, saline administered mice showed normal positive social valence, whereas activation of NT inputs in the vCA1 during the appetitive social interaction impaired it (Fig. 5b, left). Further, 5-HT1_BR-A administration with simultaneous NT release during the rewarding social interaction inversed social valence, whereas 5-HT1_BR-A administration alone in the light-off condition merely disrupted positive social valence (Fig. 5b, right). Sociability, social memory, RTPP nor anxiety were influenced by NT release into the vCA1 (Extended Data Fig. 8f–i). By manipulating 5-HT and NT release in combination with receptor antagonists, we were able to invert the polarity of social valence with the same social partner during repeated exposure. Moreover, these results also indicate that 5-HT and NT induce opposite valence signals that are integrated for social valence calculation as presence of both ablated positive and negative social valence assignment.

Next, we asked whether valence can be conferred to a neutral social stimulus by manipulating 5-HT and NT signalling. To test this, we allowed subjects to interact with two novel mice under a cup in separate chambers, where light stimulation was paired with one chamber only and memory was tested without stimulation after a 5-min interval (Fig. 5c, left). To assess the valence associated to the conspecific and not the chamber, we swapped the location of the light-paired mouse in half of the trials. When the light was off in session 1, Sert-Cre mice did not show a preference for either of the familiar mice under a cup in session 2 (Fig. 5d, left). However, activation of 5-HT inputs in the vCA1 during the initial interaction generated preference for the previously light-paired mouse in the following memory test (Fig. 5d, right). To address whether novelty, regardless of social interaction, drives the behaviour, we used the same experimental set-up replacing novel mice with novel objects (Fig. 5c, right). Light pairing with an object did not evoke preference in the subsequent memory test (Fig. 5e). To examine whether NT release can also convey negative social valence to an otherwise more neutral stimulus, we manipulated NT release in the same experimental set-up (Fig. 5c). Activation of NT inputs in the vCA1 caused avoidance of the previously light-paired mouse (Fig. 5f), but not object in the second session (Fig. 5g).

Given social cognitive capabilities are commonly impaired in autism spectrum disorder (ASD)³⁸, we asked whether our findings can be leveraged to specifically rescue positive social valence in heterozygous Shank3^Δ4-22+/− mice. Because sociability is not impaired in Shank3^{Δ4-22−/−} mice³⁹, but Shank3 mutations disrupt social memory^40,41, we surmised that Shank3^Δ4-22+/− mice would show a similar deficit. Consistent with previous reports³⁹, Shank3^Δ4-22+/− and wild-type littermates showed normal sociability after saline micro-infusion into the vCA1, which was not altered by infusion of 5-HT1_BR agonist, CP93129 (Extended Data Fig. 8j,k). By contrast, social memory was impaired in Shank3^Δ4-22+/− mice compared to wild-type littermates and CP93129 infusion did not rescue the deficits (Extended Data Fig. 8l). To activate 5-HT1_BRs during an appetitive but not neutral social interaction, we exposed the mice to the neutral social interaction immediately after CP93129 infusion, before the drug had time to diffuse. After a 15-min incubation, we exposed the subjects to an appetitive social interaction and then tested for social valence (Fig. 5h). Notably, 5-HT1_BR agonist infusion rescued positive social valence in Shank3^Δ4-22+/− mice to a level comparable to wild-type mice (Fig. 5i). To assess whether the 5-HT1_BR agonist acts through increasing excitability of vCA1 cells, we performed current-clamp recordings in the vCA1 of Shank3^Δ4-22+/− and wild-type littermates. Indeed, spike numbers were lower in Shank3^Δ4-22+/− mice compared to wild-type littermates, but bath application of the 5-HT1_BR agonist rescued the decrease in excitability in the mutant mice (Fig. 5j).

Discussion

Our results demonstrate that two neuromodulatory inputs, DR 5-HT and PVT NT, converge in a single brain region, the vCA1, to balance social valence by means of distinct mechanisms. 5-HT signals in the vCA1 were specifically elevated during a rewarding social interaction, whereas NT activity increased only during an aversive one. Our findings suggest that whereas 5-HT potentiates vCA1 pyramidal neurons by reducing the inhibitory inputs through presynaptic 5-HT1_BRs, NT increases excitability by directedly activating NTR1 on the pyramidal neurons. Moreover, our data indicate that 5-HT and NT affect distinct or possibly overlapping populations of vCA1 neurons, as cells activated by rewarding and aversive social experiences differ in their 5-HT1_BR and NTR1 composition in that they can be preferentially potentiated by 5-HT or NT (Extended Data Fig. 9). These results not only extend previous findings suggesting distinct vCA1 cell populations are activated by appetitive and aversive stimuli²⁴, but also reveal that neuromodulator receptor expression can be a strong determinant in valence information routing. In addition, the physiological and behavioural effects of 5-HT and NT contrast non-social valence assignment in the amygdala, where a single neuromodulator gates directionality independent of receptor expression⁹. Our experiments show that 5-HT conveys positive, and NT negative valence association with a conspecific, which can be flexibly switched within minutes. Thus, activation of 5-HT1_BRs specifically restored positive social valence assignment as well as vCA1 excitability in an ASD mouse model haplo-sufficient for Shank3 Exon 4-22. This suggests that enhanced 5-HT1_BR activity can compensate for the reduction in synaptic plasticity caused by mutation of Shank3, which might be a common mechanism applicable to many ASD-risk gene mutations affecting synaptic transmission.

Extended Data Fig. 9. Model illustrating physiological changes in the vCA1 in response to NT and 5-HT.

DR 5-HT and PVT NT neurons innervate overlapping but not identical areas within the vCA1. During an aversive social interaction NT is released from PVT neurons into the vCA1 and binds to NTR1 receptors on vCA1 pyramidal neurons, which increases excitability. During an appetitive social interaction 5-HT is released from DR neurons into the vCA1 and binds onto 5-HT1_BRs on GABAergic interneurons and inhibits their activity. By releasing the local inhibition onto vCA1 pyramidal neurons, the net excitatory drive (likely from the dCA2⁴) onto the pyramidal neurons is increased. As there is heterogeneity in the vCA1 cell population, a subset of cells can be preferentially potentiated by NT due to higher expression of NTR1s or preferentially potentiated by 5-HT as they receive inputs from interneurons with higher expression of 5-HT1_BRs. Overall, the increase in excitability in vCA1 pyramidal neurons in response to either NT or 5-HT balances social valence. The right insert highlights the converging circuitries from the PVT and DR to the vCA1, which mediates the plasticity. Credit: SciStories.

Given that information of conspecific identity is routed from the dCA2 to the vCA1 (refs. ^4,42,43), we postulate that 5-HT and NT increase excitability of a subset of vCA1 neurons so that incoming identity information is preferentially mapped onto the potentiated population and thereby overlayed with valence. Since manipulation of 5-HT and NT release into the vCA1 lead to distinct behavioural outcomes, it is possible that information stored in the 5-HT and NT potentiated vCA1 neurons is differentially routed to promote appropriate action-selection. This represents a basic mechanism defining how the brain processes and updates many characteristics of a given entity. Further studies investigating the downstream target regions receiving inputs from those cells will thus be informative.

None of our projection- and neuromodulator-specific manipulations affected sexual behaviours, anxiety, baseline sociability and RTPP, indicating that innate valence is not affected. As manipulations of 5-HT and NT in other brain regions have been found to influence one or many of those behaviours^9,31,44,45, our findings extend growing evidence that behavioural effects of the neuromodulators are highly region specific. Assuming the high regional heterogeneity of vCA1 neurons along with expression of various neuromodulator receptors^46,47, it is possible that further neuromodulatory signals such as oxytocin, norepinephrine or dopamine^48,49, are multiplexed to promote more specific types of social representation. This converging neuromodulatory motif has the potential to integrate complex information into a behaviourally relevant outcome and may represent a shared principle used by behaviours including social dominance, cooperation and foraging. In sum, we have identified physiological correlates of a neural competition mechanism for flexible valence association, which is capable of integrating social history. Our findings indicate that further simultaneous exploration of several neuromodulators in converging circuit motifs will not only provide critical insights into complex social behaviours but also reveal potential therapeutic targets such as 5-HT1_BR and NTR1, which can be leveraged to improve social cognitive deficits in neuropsychiatric disorders including ASD and schizophrenia.

Methods

Mice

Female and male C57BL/6J (Jackson Laboratory, 664) and heterozygous transgenic TRAP2;Ai14 (ref. ⁵⁰), Amigo2-Cre² (B6.Cg-Tg(Amigo2-cre)1Sieg/J; Jackson Laboratory, 30215), Sert-Cre (Mouse Mutant Resource and Research Centers, stock number 017260-UCD, strain code: Tg(Slc6a4-cre)ET33Gsat/Mmucd), NT-Cre⁴⁴ (a gift from S. Russo) and, Shank3Δ4-22 (ref. ³⁹) (B6.Cg-Shank3^tm2.1Bux/J; Jackson Laboratory, 32169), dCas9-KRAB (Rosa26-LSL-dCas9-KRAB (Jackson Laboratory, 033066) mice were used as experimental subjects. TRAP2;Ai14 mice were generated by crossing TRAP2 (STOCK Fos^{tm2.1(icre/ERT2)Luo/J}; Jackson Laboratory, 30323) and Ai14 (B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze/J}; Jackson Laboratory, 7914) mice. Male retired breeder CD-1 (Charles River, ICR), female Swiss-Webster (Charles River, CFW) as well as male and female C57BL/6J mice were used as partner mice.

Except for singly housed aggressors, mice were housed with 2–5 mice per cage and weaned at 21 days old. All behavioural experiments were conducted with 7–14-week-old mice. Animals were maintained on a 12-h light–dark cycle at roughly 21 °C with 50% humidity and food and water ad libitum. Behavioural experiments were performed during the same circadian period (07:00–19:00). Experiments were conducted in accordance with the National Institutes of Health Guide for Care and approved by the Use of Laboratory Animals and the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee. Sample sizes were not predetermined by statistical analysis but based on previous experience with the variance of the assays. The sequence of control and manipulated conditions were shuffled.

Stereotactic injections and cannula implantation

Mice (4–8 weeks old) were anaesthetized with intramuscular injections of a drug mixture containing ketamine (100 mg kg⁻¹) and xylazine (5 mg kg⁻¹) and heads were then positioned on a stereotaxic apparatus (David Kopf Instruments) for virus injections (0.2–0.5 μl) at a flow rate of 0.1–0.25 μl min⁻¹ with a microinjection pump (Harvard Apparatus). The following bregma coordinates from dura were used to target the virus solution filled glass cannula: dCA2 (bilateral, anteroposterior −1.6 mm; mediolateral ±1.6 mm; dorsoventral 1.7 mm), vCA1 (bilateral, anteroposterior −3.16; mediolateral ±3.1; dorsoventral −4.55), DR (unilateral, anteroposterior −4.36; mediolateral 0; dorsoventral −3.1), aPVT (unilateral, anteroposterior −0.4; mediolateral 0; dorsoventral −3.8), median raphe (unilateral, anteroposterior −4.4; mediolateral 0; dorsoventral −4) and reuniens (bilateral, anteroposterior −0.2; mediolateral ±0.85; dorsoventral −4.3).

Adeno-associated viruses (AAVs) used for stereotaxic injections were purchased from Addgene, WZ Biosciences and the Stanford Neuroscience Gene Vector and Virus Core and included: AAV_DJ-hSyn-EGFP, AAV_DJ-hSyn-hM4Di-mCh, AAV_DJ-hSyn-DIO-hM4D(Gi)-mCh, AAV_DJ-hSyn-mCh, AAV_DJ-EF1α-DIO-eYFP, AAV_DJ-hSyn-ChR2-eYFP, AAV_DJ-hSyn-ChR2-mCh, AAV_DJ-hSyn-GCaMP6f, AAV_DJ-EF1α-DIO-RCaMP2, AAV_DJ-EF1α-DIO-NpHR3.0eYFP, AAV_DJ-EF1α-DIO-ChR2-eYFP, AAV_DJ-EF1α-DIO-ChR2-mCh, AAV_DJ-CMV-DIO-EGFP, AAV_DJ-EF1α-DIO-mCh, AAV_DJ-EF1α mCh-IRES-Cre-WPRE, AAV₉-hSyn-EGFP-CAAX, AAV₉-hSyn-GRAB_5-HT3.5 and AAV₉-hSyn-GRAB_NT1.0. AAV titres ranged from 1 × 10¹² to 2 × 10¹³ gc ml⁻¹. CAV2-Cre (ref. ⁵¹) was purchased from Plateforme de Vectorologie de Montpellier. Synapsin-driven lentiviral NLS-GFP-Cre, AAV-CAG-FLEx-TC, AAV-CAG-FLEx-G and rabies virus (Janelia Research Campus) were used for monosynaptic tracing. Behavioural experiments with cell body manipulations were performed 3–4 weeks after virus injections, whereas the viral particles were allowed to incubate 4–8 weeks for manipulation of axon terminals.

Optic fibres (400-µm core, numerical aperture (NA) 0.5; RWD Life Science) for optogenetic light delivery were implanted in the vCA1 bilaterally at anteroposterior −3.16; mediolateral ±3.1; dorsoventral −4.4. A 26-gauge guide cannula (Plastics One) for drug micro-infusion was implanted at anteroposterior −3.16; mediolateral ±3.1, so that the 33-gauge infuser insert reached a depth of dorsoventral −4.5. Miniature screws (thread size 00–90 × 1/16; Antrin Miniature Specialties and McMaster-Carr), light-cured dental adhesive cement (C&B Metabond, Parkell) and resin (Geristore A&B paste, DenMat or Ortho-Jet powder&liquid, Lang Dental) were applied to secure implants to the skull. A small percentage (roughly 5%) of mice were excluded from behavioural analysis based on either off-target transgene expression or inaccurate implant placement.

Intraperitoneal injection and cannula infusion of drugs

For this, 4-hydroxytamoxifen (4-OHT, Sigma H6278-50MG) was administered by means of intraperitoneal (i.p.) injection at 50 mg kg⁻¹ for 15 min after pairing with partner mice. CNO (Tocris Biosciences 4936) was also delivered through i.p. injections at 10 mg kg⁻¹ for 30 min before behavioural experiments. Drugs used for cannula microinjection experiments were infused in a total volume of 200–400 nl at a speed of 200 nl min⁻¹ through an injector cannula connected to a micro-infusion pump (World Precision Instruments). The following drugs and concentrations were used: methiothepin mesylate salt (Millipore Sigma M149, 0.2 μg), NAS-181 (Tocris Biosciences 1413, 2 μg), NAD-229 (Tocris Biosciences 3282, 3.5 μg), SR48692 (Tocris Biosciences 3721, 5 pg), NTRC 824 (Tocris Biosciences 5438, 22 pg) and CP93129 dihydrochloride (Tocris Biosciences 1032, 0.5 μg). After completion of the infusion, injection infusers stayed for 2 min until removal and sociability and social memory assays were carried out 20 min later. NAS-181 (10 mg kg⁻¹) and SR48692 (3 mg kg⁻¹) were delivered through i.p. injections 30 min before fibre photometry experiments.

TRAP2;Ai14 behavioural experiments

Three groups of TRAP2;Ai14 mice (6–8 weeks old) were habituated to i.p. saline injections and empty cages for two consecutive days. On test day, group 1 subjects were introduced to an age-matched same-sex conspecific in a new cage, group 2 to a sex-matched aggressive male CD-1 or female CFW mouse in the aggressor’s home cage and group 3 to a potential mate (opposite sex) in either the female partner’s home cage or the female subject’s home cage. After 15 min of interaction, subjects were administered 4-OHT (50 mg kg⁻¹) and interaction was continued for 1 h before subjects returned to their home cages. Ten days later, mice were either perfused with 10% neutral buffered formalin (NBF) for histological analysis or brains were removed to be fresh frozen for in situ hybridization experiments.

Three-chamber behavioural tests

Depending on the experiment, test mice were habituated for 2 days to i.p. injections, infuser inserts or the fibre optic patch cord as well as the three-chamber apparatus containing two empty inverted metal grid pencil cups (10 cm diameter) in the outer two chamber for 5 min. The three-chamber apparatus (60 cm long by 23 cm wide by 26 cm high) was constructed of 0.3-cm-thick white opaque acrylic sheets, with two outer chambers (23 cm long by 23 cm wide) and a centre chamber (15 cm long by 23 cm wide) divided by 15-cm-long barriers extending from opposite ends of the walls. White opaque acrylic dividers were used to block entrance to the two outer chambers from the centre chamber. Partner mice were also habituated under the cup for 5 min for 2 days. On test day, subjects were placed into the centre chamber for 2 min before the barriers were lifted, whereafter the subjects were able to freely explore all three chambers. Four variations of this assay were performed to test social memory. The location of partner mice and objects in either chamber was counterbalanced between trials in all tests.

In the neutral social memory test, the first 10-min session is also a sociability test, in which the time spent in the chamber with a same-sex age-matched novel mouse or an object is measured similarly to previously described methods^3,5,7. After a 10-min interval, in which subjects were separated from the contents of the two outer chambers, the barriers were lifted and the subjects were allowed to freely explore the three chambers for 5 min in a second session. This time, the one of the outer two chambers contained a novel mouse and the other one a familiar mouse (same mouse from previous session) under a cup.

In the negative social memory test, subjects interacted directly or 5 min with either a same-sex aggressor in the aggressor’s home cage (similar to a resident–intruder test^17,18) or a same-sex non-aggressor in a new cage. After a 5-min interval, those subjects that interacted with an aggressor were exposed to a non-aggressor for 5 min and vice versa. The sequence was counterbalanced between trials. After another 5-min interval, the subject was placed into the centre chamber for 2 min, whereafter the barriers were lifted and the subject was allowed to freely explore the three chambers containing the aggressor and the non-aggressor under a cup for 5 min. Owing to the sharp drop in the discrimination scores after a short separation time^3,7, a 5-min interval was chosen to shorten the separation time after the first interaction. Singly housed (at least 10 days) male retired breeder CD-1s were used as male aggressors and pair-housed female CFWs were used as female aggressors¹⁸. CD-1s and CFWs were screened for their level of aggression as previously described¹⁷ and only mice that attacked within 10 s and more than three times within 5 min qualified as aggressors. Retired breeder CD-1s and female CFW group-housed with same-sex conspecifics were used as non-aggressors. Only mice that did not attack in a 5-min interaction qualified as non-aggressors.

In the positive social memory test, subjects interacted directly with either an opposite-sex conspecific potential mate in the potential mate’s home cage or an opposite-sex conspecific neutral partner in a new cage for 5 min. After a 5-min interval, the subject that interacted with one type of partner was exposed to other for 5 min, whereby the sequence was counterbalanced between trials. After another 5-min interval, social memory was tested in a three-chamber apparatus for 5 min with one outer chamber containing the potential mate and the other containing the neutral partner under the cup. In the case of male subjects, the experimenter held the potential mate female partner to simulate the mating posture¹⁹ with its rear close to the male’s snout three times for 3 s. In the case of female subjects, a barrier was placed into their home cage so the timing and pacing of the interaction with the potential mate male partner could be controlled. The cardboard barrier had an opening at the bottom, small enough for the female, but not the male, to easily escape to the other side of the cage. The interactions with the neutral partner were without lordosis posture and barrier in a new cage. It is important to note that the term neutral is used to describe a more neutral or less positive stimulus compared to the potential mate but not an absolute neutral stimulus.

We used two variations of the RTPP test. The basic RTPP was performed as previously described³¹, in which light stimulation was paired with one of the two outer chambers while the subject was allowed to freely explore the three chambers for 15 min. Next, light stimulation was paired with the opposite outer chamber while the subject freely explored all three chambers again for 15 min. In variation 1, the two outer chambers each contained a new mouse and light stimulation was paired with one of the two chambers, while the subjects freely explored all three chambers for 10 min. After a 10-min interval, the barriers were lifted and the subjects were allowed to move freely between the three chambers, where the outer chambers contained the two previously acquainted mice (now familiar mice) under a cup. During this 5-min period, no light stimulation was applied. In variation 2 of this assay, two new objects instead of mice were used.

Recorded videos in the three-chamber apparatus was analysed using a video tracking system (BIOBSERVE, v.3.01), which automatically tracked the location of the subject mouse. The time spent in either of the two outer chambers were scored for the duration of the tests. The following were used to calculate the discrimination scores: sociability test ((time in novel mouse chamber − time in object chamber)/(time in novel mouse chamber + time in object chamber)); neutral social memory test ((time in novel mouse chamber − time in familiar mouse chamber)/(time in novel mouse chamber + time in familiar mouse chamber)); negative social memory test ((time in non-aggressor chamber − time in aggressor chamber)/(time in non-aggressor chamber + time in aggressor chamber)); positive social memory ((time in potential mate chamber − time in neutral mouse chamber)/(time in potential mate chamber + time in neutral mouse chamber)); RTPP mouse/object ((time in previously light-paired mouse/object chamber − time in no-light mouse/object chamber)/(time in light-paired mouse/object chamber + time in no-light mouse/object chamber)). For all experiments, male and female subject mice were used in roughly equal numbers. Test mice were excluded from the analysis (less than 2% of the total), if they were not attacked by the aggressor, if the non-aggressor attacked more than three times in 5 min, or if they spent the entirety of the assay in one chamber only.

Optogenetic stimulation

Mice were habituated for 2 days to behavioural chambers with optic fibres connected to light-emitting diodes (LEDs) in off states. A 545-nm Dual-LED (Prizmatix) was connected to the optical implants via a rotary joint and a fibre optic patch cord (Prizmatix) to photostimulate NpHR3.0. A 450-nm Dual-LED (Prizmatix) was used to photostimulate ChR2. The LEDs were adjusted to around 15 mW for axon terminal stimulation using a digital power meter console (ThorLabs) and the frequency was controlled by Pulser Plus (Prizmatix). A cycle of 8 s on and 2 s off was used to stimulate NpHR3.0 to avoid tissue overheating. ChR2 was stimulated at 20 Hz with a 5-ms pulse duration.

CPP

Similar to previously described⁵², mice were tested for baseline preference for two differently textured clear acrylic floors by spending 15 min in a white opaque polycarbonate chamber (23 cm long by 23 cm wide by 26 cm high, 0.3 cm thick) on both floors (23 cm long by 13 cm wide by 2.5 cm high, 0.3 cm thick). The ‘grid’ floor was constructed of seven evenly spaced apart rails, which were 2.5 cm in width. The hole floor contained 4 × 9 round holes 8 mm in diameter, evenly spaced apart. On day 1, mice were either administered saline or CNO (10 mg kg⁻¹) through i.p. injections 30 min before conditioning to one of the floors for 15 min. On the following day, mice that received saline on the previous day, received CNO and vice versa. 30 min later, animals were conditioned to the alternate flooring for 15 min. On day 3, test mice were placed into the chamber containing both types of flooring and time spent on either type was automatically scored with video tracking software (BIOBSERVE).

Elevated-plus maze

Mice were placed into the centre of the elevated-plus maze (74 cm long by 74 cm wide by 94 cm high; Med Associates Inc.) with white floors (each arm 35 cm long by 6 cm wide with 19 cm high walls on closed arms) for 10 min and location of the mice were scored with video tracking software (BIOBSERVE).

Male and female sexual behavioural tests

Male subjects were placed for 5 min into a female conspecific’s home cage with one female present. The female was held by the experimenter for 3 s in a mating posture, so that the male could sniff the female’s rear. Male mounting attempts during the 5 min of free interaction were manually quantified. As females do not show mounting behaviour, a different test was used to assess female receptiveness to males. Male partners were placed into the female subject’s home cage, which was divided into two compartments by a barrier, small enough for the female to easily escape. Escape times and duration of the female in each compartment were manually scored.

Fibre photometry

AAV_DJ-CaMKII-GCaMP6f, AAV₉-hSyn-EGFP-CAAX, AAV₉-hSyn-GRAB_5-HT3.5 (ref. ⁵³) or AAV₉-hSyn-GRAB_NT1.0 (ref. ⁹) was injected into the vCA1 (unilateral, anteroposterior −3.16; mediolateral ±3.1; dorsoventral −4.55), and fibre optic implants were secured above the injection site (unilateral, anteroposterior −3.16; mediolateral ±3.1; dorsoventral −4.4) in stereotaxic surgeries. Two to three weeks later, mice were habituated to the behavioural set-up and tested on the following day with simultaneous video and fibre photometry acquisition. Test mice interacted with a new object, a novel same-sex age-matched conspecific, a sex-matched aggressor or a potential mate for 5 min with an interval of 5 min, whereby the sequence of the interaction was counterbalanced between subjects. The interaction with the aggressor and female potential mate was in the partner’s home cage, whereas the interaction with the object and same-sex conspecific was in a new cage. Female subjects interacted with the potential mate in their own home cage. Synapse software, which controlled an RZ10X LUX-I/O processor (Tucker-Davis Technologies), was used for data acquisition. To stimulate Ca²⁺-dependent and isosbestic emission, GCaMP6f was excited by frequency-modulated built-in 465- and 405-nm LEDs (RZ10X LUX-I/O processor), respectively. All optical signals were band-pass-filtered with a fluorescence mini cube (Doric), emission was measured with built-in photosensors in the RZ10X LUX-I/O processor and the signal was digitized at 6 kHz. A previously described custom MATLAB_R2022b (MathWorks) code⁷ was used for signal processing. To debleach, fibre photometry raw signals were fitted with a mono- or bi-exponential decay function, and the resulting fluorescence trace was z scored. Corresponding videos were manually analysed by frame in MATLAB to identify the time of physical contact between test mouse and partner mouse and object. Peristimulus time histograms were constructed by averaging 7 s of non-overlapping epochs from the z scored trace, where a time of 0 represents the time of contact. The maximal z score between 0 and 4 s was used as peak z scored fluorescence.

Molecular cloning and AAV production

We generated a multi-guide CRISPRi AAV vector starting with pX552 (Addgene, 60958) as the backbone, which we then digested with NotI-HF (NEB). Between the NotI sites, we inserted two gene fragments to generate the AAV backbone pAAV mU6-sgRNA-CR1 EF1α-EGFP-W3-SV40. The first fragment comprised a mouse U6 (mU6) promoter, multiple cloning site for protospacers, constant region (CR1) and an EF1α promoter derived from pU6-sgRNA EF1α-puro-T2A-BFP (Addgene, 60955) but engineered to include BbsI sites flanking the mU6-sgRNA-CR1 cassette. The second fragment contained EGFP, W3 terminator and SV40 poly(A) signal. Both were assembled into the NotI-digested pX552 backbone using a three-part Gibson assembly (HiFi NEB). Subsequently, sgRNA arrays comprising a mU6 promoter, sgRNA protospacer sequence of interest and CR1 cassette, synthesized as gene fragments, were inserted into the backbone with a Golden gate approach using the engineered BbsI (NEB) sites in accordance with the manufacturer’s instructions. To ensure maximum knockdown efficiency, the vectors consist of three tandem sgRNA cassettes containing the protospacers predicted to have highest activity according to the Weissman laboratory’s V2 CRISPR sgRNA algorithm⁵⁴.

AAV particles were produced using a previously described protocol with minor modifications⁵⁵. One T175 flask of HEK293T cells (American Type Culture Collection) per construct was transfected with 87.6 µg PEI (1 µg ml⁻¹; 3:1 PEI: DNA; PolySciences 24765) along with 15.4 µg of pAdΔF6, 7.3 µg of pAAV2/9n (pAdΔF6 and pAAV2/9n were gifts from J. M. Wilson; Addgene, catalogue nos. 112867 and 112865, respectively) and 6.5 µg of pAAV mU6-sgRNA-CR EF1α-EGFP-W3-SV40 plasmids. Viral supernatant was harvested 72 h posttransfection, along with cells and mixed thoroughly with 0.1 volumes of chloroform. NaCl was added to a final concentration of 1 M, mixed thoroughly, and the sample centrifuged for 5 mins at 3,000g, 4 °C. The aqueous phase was retained, mixed with 9.4 ml of 50% PEG-8000, incubated on ice for 1 h and centrifuged for 30 min at 3,000g, 4 °C. Pellets we resuspended in dPBS, mixed with 5 µl benzonase (EMD Millipore, 71205) and 1 mM MgCl₂, and incubated at 37 °C for 30 min. Postbenzonase treatment, samples were mixed thoroughly with an equal volume of chloroform and centrifuged for 5 min at 3,000g, 4 °C. Samples were buffer exchanged into dPBS and concentrated to roughly 50 µl using 100-kDa cut-off cellulose centrifugal filters (Millipore Sigma, UFC8100).

Immunohistochemistry

For the cFos immunohistochemistry experiment, mice were subjected to an interaction with an aggressor (CD-1/CFW) or a potential mate for 1.5 h and immediately perfused with chilled 10% NBF. After overnight incubation in 10% NBF, 40-μm slices were prepared on a Leica VT 1000S vibratome and collected in 24-well plates filled with phosphate buffered saline (PBS). After a 1.5-h blocking in 5% normal goat serum, slices were incubated with 1:1,000 diluted primary antibody and 5% normal goat serum at 4 °C for 12 h. After three 5-min washes with PBS, slices were incubated for 1.5 h in secondary antibody, subsequently washed three times in PBS for 15 min and Fluoromount-G mounting medium was used for 4,6-diamidino-2-phenylindole (DAPI) staining. The following primary antibodies were used: 1:1,000 diluted cFos antibody (Synaptic Systems, 226 008) and 1:500 diluted TPH2 antibody (Abcam, ab184505). The following secondary antibodies were used: Alexa 488 (1:500 dilution), Alexa 568 (1:1,000 dilution) and Alexa 633 (1:1,000 dilution).

Validation of knockdown efficiency

AAV_DJ-EF1α mCh-IRES-Cre-WPRE and AAV₉-EF1α-sgTph2-EGFP or AAV₉-EF1α-sgNTC-EGFP were injected into the DR (unilateral, anteroposterior −4.36; mediolateral 0; dorsoventral −3.1) of dCas9-KRAB mice. Three weeks later, immunohistochemistry was performed as described above. Imaging was conducted on a Zeiss LSM 780 confocal (×20 objective; NA 0.8; diode laser 405 nm, argon laser 458/488/514 nm, diode-pumped solid-state laser 561 nm and HeNe laser 594/633 nm; detectors 32 channel GaAsP, T-PMT (transmitted light), 2-PMT and Airyscan detector; bit depth of images was 8 bit). TPH2 channels were background subtracted, and the same cut-off threshold was applied for all replicates. Polygons were manually drawn around the DR area and particles within the area of interest were analysed. All infected DR cells received sgTph2 or non-targeting guides. To evaluate cell body TPH2 expression in the DR, we analysed fluorescence intensity. An average percentage fluorescence intensity was calculated from the sgNTC replicates and the remaining ratio was used for calculation of the percentage TPH2 fluorescence intensity and knockdown efficiency.

Behavioural cohorts were injected with AAV₉-EF1α-sgTph2-EGFP or AAV₉-EF1α-sgNTC-EGFP into the DR and CAV2-Cre into the vCA1. Six weeks later, behavioural experiments were performed followed by perfusion and immunohistochemistry. Imaging and background subtraction was performed as described above. Polygons were manually drawn around the vCA1 and particles in within the area were analysed. In the vCA1 only the DR terminals received sgTph2 or non-targeting sgRNA. Therefore, we analysed the area of the TPH2+ terminals in the vCA1. An average percentage of TPH2+ area within the vCA1 area was calculated from the sgNTC replicates and the remaining ratio was used for calculation of percentage TPH2+ area.

Primary neuronal cultures

Primary neuronal culture was performed as previously described⁵⁶. In brief, DR nuclei were dissected from P0 pups in ice-cold Hank’s buffered saline solution (HBSS) buffer (Thermo Fisher Scientific, 88284) supplemented with 2 mM Ca²⁺ and 0.5 mM EGTA under a dissection hood and incubated in Neuronal Isolation Enzyme (Thermo Fisher Scientific, 88285) and HBSS for 20 min, washed twice with HBSS and dissociated in prewarmed Neurobasal-A Medium (Thermo Fisher Scientific, 10888022) with 2% FBS (Avantor, 97068-085) by gentle trituration. Cell yield and viability were determined by a hemocytometer and trypan blue (Thermo Fisher Scientific, 15250061) staining, respectively. Neurons were plated on a poly-d-lysine (Thermo Fisher Scientific, A3890401) precoated 24-well plate. After 24 h, half of the serum medium was replaced with an equivalent volume of Serum-free Neurobasal-A Medium supplemented with 0.5 mM l-glutamine (Thermo Fisher Scientific, A2916801) and 0.5% B-27 (Thermo Fisher Scientific, 17504044). On day 3, half of the old medium was replaced with fresh medium supplemented with 4 mM Ara-C (Sigma, C6645), and neurons were infected with AAV constructs expressing control sgRNAs or Tph2 sgRNAs. After infections, the primary neuronal cultures were maintained for 1 week with medium changes every 3 days.

qPCR with reverse transcription

RNA extraction was performed using a Quick-RNA MicroPrep Kit (Zymo Research, R1050) according to the manufacturer’s protocol. Total RNA was reverse transcribed using the High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific, 4387406). The resulting complementary DNA (cDNA) was then used for qPCR using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, A25742) with gene-specific primers. Data were recorded using a QuantStudio 7 Flex Real-Time PCR system (Thermo Fisher Scientific). Tph2 expression was normalized to Actb. The relative changes in expression were calculated using the 2^-ΔΔCt method. Primers used in the studies include: Tph2 (mouse): forward, 5′-GCAAGACAGCGGTAGTGTTCT-3′; reverse, 5′-CAGTCCACGAAGATTTCGACTT-3′; Actb (mouse): forward, 5′-GGCTGTATTCCCCTCCATCG-3′; reverse, 5′-CCAGTTGGTAACAATGCCATGT-3′.

Ex vivo electrophysiology

Here, 4–8-week-old C57BL/6J, Shank3^Δ4-22+/− and wild-type littermates were used for whole-cell recordings. Mice were euthanized and brains were sliced in a sucrose cutting solution containing (in mM): 228 sucrose, 26 NaHCO₃, 11 glucose, 2.5 KCl, 1.2 NaH₂PO₄, 7 MgCl₂ and 0.5 CaCl₂ on a vibratome (Leica VT1200 S). Slicing and recording solutions were continuously equilibrated with 95% O₂ and 5% CO₂. Coronal vCA1 slices (300 μm) were transferred to a slice holding chamber with artificial cerebrospinal fluid containing (in mM): 119 NaCl, 26 NaHCO₃, 11 glucose, 2.5 KCl, 1.2 NaH₂PO₄, 1.3 MgCl₂ and 2.5 CaCl₂ (osmolarity 289–295) for 30 min at 32 °C and then further equilibrated for 30 min at room temperature. Next, brain slices were placed in a recording chamber perfused with 28–30 °C artificial cerebrospinal fluid and visualized with a ×40 water-immersion objective on an upright fluorescent microscope (BX51WI; Olympus) equipped with infrared-differential interference contrast video microscopy and epifluorescence (CoolLED). Whole-cell current-clamp recordings (pipette opening 4–6 MΩ) were performed with pipettes filled with (in mM): 130 C₆H₁₁KO₇, 5 KCl, 10 HEPES, 0.6 EGTA, 2.5 MgCl₂, 4 Mg₂ATP, 0.4 Na₃GTP, 10 phosphocreatine (pH 7.25; osmolarity 290). Whole-cell voltage-clamp recordings (pipette opening 3–4 MΩ) of IPSCs were performed with pipettes filled with (in mM): 80 CsCl, 65 CsMeSO₄, 8 NaCl, 10 HEPES, 0.25 EGTA, 2 Mg₂ATP, 0.3 Na₃GTP, 0.1 spermine, 7 phosphocreatine (pH 7.34; osmolarity 300). No CsCl but 140 CsMeSO₄ was used for whole-cell voltage-clamp recordings of EPSCs. Series and input resistance were monitored with −4 mV, 70-ms pulse delivered through the recording pipette and experiments were excluded from the analysis if series resistance varied by more than 15%.

Current-clamp recordings were carried out to assess intrinsic cell excitability with a series of incremental rectangular depolarizing current pulses (20 pA, 500 ms) injected into vCA1 pyramidal cells. Next, the same procedure was performed in the same cell after either CP93129 dihydrochloride (Tocris Biosciences 1032, 5 µM) or PD149163 (Sigma, 0.5 µM) was bath applied for 5 min. Following 10 min of wash-out, recordings were carried out in the same cell in presence of the second drug, whereby the sequence of drug application was counterbalanced between cells. Only one cell was recorded per slice. Current-clamp recordings in the PVT was carried out with 20-pA depolarizing current pulses and in the DR with 25-pA depolarizing current pulses. Pipettes were filled with in (mM): 130 C6H11KO7, 10 KCl, 10 HEPES, 0.2 EGTA, 4 Mg2ATP, 0.5 Na3GTP, 10 phosphocreatine. CNO (Tocris Biosciences 4936, 5 µM) was bath applied for 5 min. Evoked EPSCs and IPSCs were recorded from vCA1 cells with a bipolar stimulating electrode (fabricated from platinum and iridium wire) placed near the recording pipette. Baseline evoked and spontaneous EPSCs were recorded in voltage-clamp settings at −70 mV the presence of picrotoxin (50 μM). Baseline evoked and spontaneous IPSCs were recorded at −70 mV in the presence of NBQX (10 μM) and D-AP5 (50 μM). CP93129 dihydrochloride (5 µM) or PD149163 (0.5 µM) were bath applied after an initial baseline recording. Summary graphs of the effects of CP93129 and PD149163 over time in evoked recordings were generated by averaging 1-min bins as a percentage of the averaged 4-min baseline. Cumulative probability graphs were generated from all spontaneous events before and after drug application. Synaptic responses were recorded in the vCA1 8 weeks after stereotaxic injections of AAV_DJ-hSyn-ChR2-eYFP into the DR or PVT. To photostimulate ChR2 in vCA1 axons and/or terminals, 470-nm 0.1–5.0-ms light pulses from the pE-300^ultra (CoolLED) were delivered by means of a ×40 water-immersion objective to the whole slice. NAS-181 (20 µM) or SR48692 (0.5 µM) were bath applied following an initial baseline recording. Recordings were made using a MultiClamp 700B amplifier (Molecular Devices), digitized at 10 kHz with the Digidata 1320A or 1440A data acquisition system (Molecular Devices) and analysed with Clampfit v.10.7 software (Molecular Devices).

In situ hybridization and quantification

The experiment was performed according to the manufacturer’s manual⁵⁷ using the RNAscope Multiplex Fluorescent v.2 kit (Advanced Cell Diagnostics). Briefly, brains were fresh frozen on dry ice and stored at −80 °C until 15-μm slices were prepared and collected directly on Superfrost plus microscopy slides using a cryostat. Following fixation in prechilled 4% paraformaldehyde, slices were washed in PBS and dehydrated gradually in 50%, 70% and two times 100% ethanol (5 min each). After a barrier was created with a hydrophobic pen, slides were incubated with roughly five drops of RNAscope hydrogen peroxide for 10 min, subsequently washed with distilled water and incubated with Protease IV for 30 min, then washed twice with PBS. The probes mCh-O3 (513201), Mm-Slc17a6-C2 (319171-C2), Mm-Gad2-C2 (439371-C2) and Mm-Nts-C3 (420441-C3) were warmed up to 40 °C in a water bath, and 1 volume of C2 and C3 were diluted in 50 volumes of C1 to create the probe mixture. In the hybridization step, slices were incubated with the probe mixture for 2 h at 40 °C in a HyEZ Oven and rinsed twice at room temperature in Wash buffer (50× RNAscope Wash buffer diluted in distilled water). Next, slides were incubated with roughly six drops of RNAscope Multiplex FL v2 Amp1 for 30 min at 40 °C in the HyEZ Oven and washed twice with wash buffer at room temperature. This procedure was repeated with RNAscope Multiplex FL v2 Amp2 (30 min) and RNAscope Multiplex FL v2 Amp3 (15 min). Slides were then incubated with RNAscope Multiplex FL v2 HRP-C1 in the HyEZ Oven at 40 °C for 15 min, rinsed twice in wash buffer at room temperature and incubated in a 1:1,000 diluted Opal 520 dye for 30 min at 40 °C and subsequently rinsed twice in wash buffer. This step was repeated with HRP-C2, HRP-C3 and matching Opal dyes 570 (1:3,000 dilution), 650 (1:1,000 dilution). Finally, Fluoromount-G mounting medium was used for DAPI staining.

Imaging was performed on a Zeiss LSM 780 confocal (×20 objective; NA 0.8; diode laser 405 nm, argon laser 458/488/514 nm, diode-pumped solid-state laser 561 nm and HeNe laser 594/633 nm; detectors 32 channel GaAsP, T-PMT (transmitted light), 2-PMT and Airyscan detector; bit depth of images, 8 bits). A Fiji (ImageJ v.1.52p) macro was used for automated image analysis, which was performed blinded. Based on pixel-intensity threshold, DAPI channels were background subtracted, binarized and converted to masks. Polygons were manually drawn around the vCA1 area and particles in the remaining channels were automatically analysed within the masked areas inside the polygon. The cut-off threshold was visually determined by comparing ascending intensity values with marked location in the images to eliminate background particles.

Statistical methods and reproducibility

For behavioural experiments and analysis, the experimenters were blinded to the virus injection the animals received. Prism10 (GraphPad) was used for statistical analysis. Individual data points are identified by sex in the source data. Parametric statistical tests were only chosen for normally distributed samples. Otherwise, non-parametric statistical tests were performed. One-way analysis of variance (ANOVA) with Tukey’s or Dunnett’s multiple comparison post hoc test was used to determine significance for several treatment comparisons and two-way ANOVA with Tukey’s or Sidak’s multiple comparison post hoc test for several group comparisons and across many time points. P values found in figure legends refer to ANOVA results of the whole group, whereas the asterisks between individual samples represent results of the multiple comparison post hoc tests. A two-tailed paired Student’s t-test was applied for within-group comparison of two treatments and unpaired test for comparison between two groups. When normality was not assumed, the Wilcoxon signed rank test was used for within-group comparison of two treatments, Mann–Whitney test for between group comparison and Kruskal–Wallis with post hoc Dunn’s test for multiple comparisons. All tests are two-sided. NS, not significant. *P < 0.05, **P < 0.01, ***P < 0.001. In all figures, data are shown as mean ± s.e.m.

Representative data are one of: Fig. 1c, 17 mice; Fig. 1d, 12 mice; Fig. 2a, 13 mice; Fig. 2e, 11 mice; Fig. 3a, 12 mice; Fig. 3d, 14 mice; Fig. 3j, 10 mice; Fig. 3l, 10 mice (sgTph2) and 8 mice (sgNTC); Fig. 4a, 12 mice; Fig. 4e, 9 mice; Fig. 4h, 10 cells; Fig. 4i, 11 cells; Fig. 4j, 14 cells (predrug), 11 cells (NTR1 (PD)), 8 cells (5-HT1_BR (CP)); Fig. 4k, 6 cells (EPSC), 7 cells (IPSC); Fig. 4l, 8 cells; Fig. 5a, 11 mice; Fig. 5b, 10 mice; Fig. 5h, 10 mice and Fig. 5j, 15 cells (Shank3) and 14 cells (wild-type).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-025-08809-2.

Author contributions

X.W. designed the experiments, interpreted the results and wrote the paper, which was edited by all authors. X.W., J.M.Z., H.L. and A.S. performed surgeries and behavioural experiments. J.M.Z. and A.S. performed fibre photometry recordings, histology and micro-infusion. T.O. designed the CRISPRi knockdown strategy. H.L. and T.O. cloned the sgRNA vectors and made AAVs. H.L. performed CRISPRi validation and associated histology. H.L. and R.A. performed RNAscope experiments. R.A. and Y.C. performed ex vivo electrophysiology recordings. R.A. performed cFos immunohistochemistry. X.W., J.M.Z., H.L., R.A., A.S. and Y.C. analysed data. X.W. provided all supervision.

Peer review

Peer review information

Nature thanks the anonymous reviewers for their contribution to the peer review of this work.

Data availability

Sequences of the custom vector and sgRNAs used for CRISPRi knockdown experiments, detailed statistical analysis between sexes and raw histology images are available at Zenodo (10.5281/zenodo.14814653)⁵⁸. All other raw and/or processed data generated during this study are available from the corresponding author upon reasonable request. Source data are provided with this paper.

Code availability

MATLAB codes used in this study are available at Zenodo (10.5281/zenodo.14814674)⁵⁹.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41586-025-08809-2.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-025-08809-2.