Increased c-Fos expression in lateral habenula during social transmission of negative valence in prairie voles

Abstract

Background

Social learning is the process of acquiring social skills, new information, or associating negative or positive valence to a context through the observation of others and through direct social interaction with others. Neurodevelopmental disorders such as autism spectrum disorder show deficits in social salience and reciprocal affective responses. Social learning implicates the basolateral amygdala (BLA), anterior cingulate cortex (ACC), and anterior insula (AI). The lateral habenula (LHb), a brain area renowned for its role in negative reinforcement learning, has not been yet extensively studied in the domain of social learning.

Methods

We developed a fear conditioning by proxy paradigm called ‘social transmission of negative valence’ (STNV) and tested prairie voles on the task. Observers experienced negative social conditioning through a proxy cage mate that served as the demonstrator during retrieval of a cued fear memory. Observers went through a social memory recall session 24 h after observation. We measured observers’ freezing time, self-grooming, rearing, and ultrasonic vocalizations emitted as signs of distress. We also quantified immediate early gene translation as a proxy for neural activity using c-Fos immunochemistry 80 min after observing demonstrators going through memory recall.

Results

Socially-conditioned observers that were exposed to the fear-conditioned demonstrators displayed increased freezing time, self-grooming, and rearing during social recall sessions compared to control observers. They also displayed higher ultrasonic vocalization frequency compared to controls. Socially-conditioned observers showed increased c-Fos expression in the LHb, BLA, ACC and AI compared to controls.

Limitations

The c-Fos findings are correlational and additional experiments involving chemo- or optogenetic inhibition or excitation of LHb neurons in observers are necessary for causality confirmation.

Conclusions

We found that the LHb is co-activated with other key areas during social learning in prairie voles. These findings extend the traditional view of the LHb as an area involved in negative reinforcement learning and position it as a critical area for social affect. This offers a fresh perspective on the neural mechanisms of social affect and opens a new line of inquiry into brain dysfunction of social salience in neurodevelopmental disorders.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13229-026-00712-5.

Background

Social learning is a key element of our daily life and its dysfunction in neurodevelopmental disorders, such as autism spectrum disorder (ASD), can have a significant impact on the capacity to interact with others and to navigate social challenges. Although prevalence of ASD has increased substantially in the past decade with recent estimates from the Centers for Disease Control and Prevention of 1 in 31 children in the United States having autism [1], research on treatment has been hindered, in part, by a lack of knowledge in the brain and molecular underpinnings of the complexity and heterogeneity of social dysfunctions [2, 3]. In addition, there are relatively few paradigms that measure social learning via an interaction between partners in translational animal models.

Prior research has used social learning paradigms that hold significant translational relevance for ASD, and other disorders characterized by empathy impairments, mainly supporting the concept of social transmission of aversive stimuli via observational learning. Observational fear learning (OFL) or social fear learning (SFL) are well-established, across-species, paradigms that seek to measure transmission of stress between demonstrators (receiving aversive stimuli such as a shock, paired with conditioned stimuli such as a tone) and observers watching this classical fear conditioning. During OFL paradigms in rodents, social stress transmission is depicted by increased freezing behavior in observers witnessing demonstrators receiving aversive stimuli [4–10] and increased freezing behavior in a recall session 24 h after transmission of stress [7]. OFL studies regarding social transmission of pain and analgesia in rodents [11, 12] have also resulted in similar distressed phenotypes, such as increased freezing levels, displayed by both the affected demonstrator and observer. Rearing, self-grooming, and ultrasonic vocalizations (USVs) are classic behavioral markers of acute stress, anxiety, and conflict in rodents but are underexplored in OFL experiments [13–17].

The social transmission of fear by proxy (STF) paradigm in rats [10, 18–22] also measures social learning but through indirect exposure by the transmission of subtle social signals, where observers watch their partner freeze to the tone during memory recall instead of observing them during shock reception and fear acquisition. During a socially acquired threat recall session the following day, observer rats displayed increased freezing to the conditioned stimulus without the presence of demonstrators, as a sign of social transmission of the acquired threat.

In terms of neural correlates implicated in observational social learning or social transmission of threat or pain, the anterior cingulate cortex (ACC), anterior insula (AI), and basolateral amygdala (BLA) are among the main areas of interest that showed increased activation and connectivity in response to the observation of stressed conspecifics in rodents [23–44]. Pharmacological manipulation and inactivation of the ACC showed its essential role in the acquisition phase of OFL, whereas the BLA is involved in acquisition and recall [7, 45, 46].

We propose that these areas, the ACC, AI, and BLA, are part of a social affect salience (SAS) network, and are activated in response to the pain or stress of others in humans [47, 48]. These areas are involved in detecting salient social events and in generating and regulating emotional responses and associated values to these events. They are also part of the salience network at resting-state functional connectivity [11, 49–55].

Another area of particular interest that has significant connections with both the AI and ACC, is the lateral habenula (LHb) that plays an important role in processing negative motivational information [56, 57]. The LHb has been extensively studied in the context of anti-reward processing and negative reward prediction [58]. Studies in macaques [59] and rodents [60] have identified activation of the habenula in the absence of a reward, or presence of a punishment, confirming its role in learning and motivation [61–65]. Additional involvement has been suggested in the pathophysiology of conditions such as major depressive disorder [62], substance use disorder, schizophrenia, and bipolar disorder due to dysregulated reward circuitry [66].

Despite the known role of the habenula in learning and emotional processing and its connection to the SAS network, there is less investigation of its implication in social learning. We hypothesize that the LHb will be activated during social learning in negative valence contexts in rodent species along with the other known SAS brain areas, specifically during the observation of distress or negative valence in others. Understanding these pathways and behavioral phenotypes associated with subtle social learning would advance our understanding of the neural basis of social deficits in animal models of ASD.

Inspired by the fear by proxy paradigm, here we present an adapted version termed ‘social transmission of negative valence’ (STNV), in line with the nomenclature of the National Institute of Mental Health research domain criteria [67], which aims to assess STNV in observers through witnessing the conditioned partner going through stress during a negative valence memory recall session. We also assessed social memory in observers 24 hours after observation. We decided to test STNV in prairie voles (Microtus ochrogaster), a highly social species, due to their display of highly complex, species-specific social behaviors which include pair bond formation/maintenance, bi-parental care, social loss, and consolation [68–74]. Prairie voles are distinctive from other rodents as they are known for their strong pair bond formation, cooperative parenting, and aggression toward strangers, making them unique rodent models for studying social behavior, empathy, and consolation [68–74].

We predicted that prairie vole observers would show increased freezing, self-grooming, rearing, and modified USV frequencies during negative valence social recall as a demonstration of social transmission of negative valence. We also conducted a control experiment by reuniting demonstrators and observers in their home cage between day 1 and day 2 to examine if observers would still show an increase in freezing during social memory recall.

Notably, socially-conditioned observers exhibited increased freezing, self-grooming, rearing and higher USVs as well as higher c-Fos expression in the LHb, ACC, BLA, and AI. Observers did not exhibit an increase in freezing during the control condition in which observers and demonstrators were reunited, implying that social buffering might influence social learning. Our study demonstrates that prairie voles as highly translational animal models for neurodevelopmental disorders such as ASD and suggests that the lateral habenula is involved in social affect salience processing.

Methods

Animals

Prairie voles (Microtus ochrogaster) used in this experiment were sexually naïve wild-type males and females, nine weeks of age. Voles were raised in a breeding colony at the University of Toledo, weaned 21 days after birth, and socially housed in same-sex duos or trios on a 12:12 h cycle, alternating light:dark. Water and food were provided during their course of life. All breeder animals and experimental subjects were less than five generations from the wild.

All the housing, breeding, handling, and experimental procedures were approved by the University of Toledo Institutional Animal Care and Use Committee (IACUC) and were conducted in compliance with the Animal Welfare Act and the National Research Council’s Guide for the Care and Use of Laboratory Animals.

Social experiments used same-sex adult pairs, all housed together since weaning (9 weeks old). In trio cages, the third animal was removed from the enclosure at least one week before testing, allowing the remaining couple to habituate to the housing change. Adult subjects tested as solo subjects for validation studies were individually housed since weaning. An equal number of male and female subjects were used for all experiments.

Experimental procedure: social transmission of negative valence

Social transmission of negative valence (STNV) is a behavioral assay that aims to measure the transmission of threat and anxiety in “observers” from the observation of a stressed conspecific, or “demonstrator” (Fig. 1). We developed this adapted version of a previously validated social transmission of threat paradigm used in rats and the experiment was conducted on a 2-days period [18]. We accounted for consoling behavior in prairie voles and for social buffering and therefore separated siblings and single-housed them between day 1 and day 2 for both groups.

Fig. 1.

Social transmission of negative valence design. Experimental design of the task and controls

The two-day STNV paradigm assesses social learning in prairie voles by measuring anxiety phenotypes such as freezing levels, grooming, rearing, and USVs in observers.

Day 1, session 1: classic threat conditioning

Pair housed siblings with individual ear tags were brought into the testing room at least 30 min before testing, and the home cage was placed in a sound-attenuating isolation booth (ROOM, Brooklyn, NY). One animal was randomly designated as the observer while the other was designated as the demonstrator for the remainder of testing. The demonstrator was transferred to the shock-capable side of a two-sided modular threat conditioning cage in a sound-attenuating box (Coulbourn, Harvard Apparatus) located within a sound-attenuating booth (ROOM), separated by a transparent barrier from the shock-incapable side. Demonstrators in the experimental group were first exposed to 300 s of habituation with no tones and no shocks, followed by 15 consecutive tones (30 s, 6 KHz, 80–84 dB) each preceding a mild foot shock (1 s, 1 mA) with an inter-tone interval of 120 s. Control demonstrators underwent the same protocol with no shocks delivered. For context, we used a clear barrier, cinnamon scent, bar grid floor, and house lights (Context A).

The primary outcome measure was freezing during the tones, with freezing during habituation as a secondary measure. We recorded and analyzed freezing time using the automated program FreezeFrame (Harvard Apparatus, Holliston, MA). Freezing was counted when no animal movement was detected above threshold for more than 1 s. To eliminate demonstrators not responding normally to test conditions, a planned outlier test was performed on average freezing across 15 tones. Demonstrators with an average freezing greater than two standard deviations away from the mean were eliminated from subsequent analyses along with their matched observers.

Day 1, session 2: social transmission of negative valence

The demonstrator was kept in the experimental cage following threat conditioning. The observer was brought into the same cage on the shock-incapable side of the transparent barrier to exclude physical contact and to allow sensory observations. The observer and demonstrator were then exposed to a 300 s habituation period followed by five consecutive tones (30 s, 6 KHz, 80–84 dB) with an inter-tone interval of 120 s, for a total of 930 s. Freezing time was recorded and analyzed as before for the observer and demonstrator, with freezing during the tones as the primary outcome measure.

Following the STNV session, observers were returned to their home cage, and demonstrators were isolated overnight in a separate cage (Experiment 1, Primary Experiment, Fig. 2) or both demonstrators and observers were returned to their home cage (Experiment 2, Consoling Control, Fig. 3). We separated the observer and demonstrator to prevent social buffering resulting from consoling behavior in the home cage [75], which we hypothesized would reduce stress in subsequent recall sessions and attenuate social memory.

Fig. 2.

Social transmission of negative valence. a-b: Threat conditioning in demonstrators. (a) Demonstrators exposed to tone-shock pairings on Day 1, session 1, developed a conditioned freezing response over the course of 15 tones, as compared to no-shock controls. (b) Demonstrators exposed to tone-shock pairings on Day 1, session 2 demonstrated a conditioned freezing response to tones while across a transparent barrier from an observer. Error bars show SEM. c-d: Social threat conditioning in observers and emotional contagion. (c) Observers across a transparent barrier from fear-conditioned demonstrators on Day 1, session 2 did not acquire a within-session freezing response to tones. (d, e) On Day 1, session 2 on Day 2, observers’ freezing correlated significantly with demonstrators’ freezing for the experimental condition but not in the control condition. (f, g) observers of fear-conditioned demonstrators showed a conditioned freezing response to the tone. Error bars show SEM. e- f: Negatively valenced behaviors in observers. During the threat recall session on Day 2, observers of fear-conditioned demonstrators showed an increase in negatively valenced behaviors, including (h) self-grooming and (i) rearing. Error bars show SEM

Fig. 3.

Characteristics of ultrasonic vocalizations in STNV and CTC. a-c: Social transmission of negative valence in observers (primary experiment): during the threat recall session on Day 2, observers of fear-conditioned demonstrators showed varying USV characteristics, including (a) higher principal frequency, (b) increased average call length, (c) decreased average power. Error bars show SEM; *p < 0.05. d-f: Classical threat conditioning in prairie voles (validation experiment): during the validation experiment, conditioned prairie voles exhibited (d) higher principal frequency but did not display change in (e) average call length or (f) average power. Error bars show SEM; *p < 0.05

Day 2: social threat memory

Twenty-four hours after Day 1 testing, the observer was re-introduced to the “safe” side of the experimental cage with a metal barrier separating the cage in half. Context B consisted of a metal barrier, peppermint scent, and infrared lights. The observer was exposed to a 300 s habituation period followed by five consecutive tones (30 s, 6 KHz, 80–84 dB) with an inter-tone interval of 120 s. Freezing was recorded and analyzed as before, with freezing during the tones as a primary outcome measure.

The session was video recorded and self-grooming duration and rearing frequency were quantified as primary outcome measures of negative valence. These behaviors were manually rated by a blinded experimenter using a behavioral coding system (The Observer XT, Noldus, Wageningen, the Netherlands). Self-grooming bouts were counted if the animal groomed itself for more than 3 s. Rearing consisted of an animal standing on its rear limbs, extending its forelimbs.

Finally, ultrasonic vocalizations (USVs) were recorded during this session in a randomly selected subset of observers using an Ultravox microphone (Noldus, Wageningen, the Netherlands). The microphone was placed on the top of the experimental chamber. The microphone’s gain was reduced to 54% of its maximum capacity. Vocalizations were analyzed using the DeepSqueak software using a rat matrix with frequency cut-offs between 20 and 120 Hz and a score threshold of 0.5. Based on a validation study (below), the primary outcome measure was principal frequency (Hz).

USV classical threat conditioning (CTC) validation study

A validation experiment was conducted to better characterize vocalizations during stress and fear learning in prairie voles [76–78]. Singly housed voles (N = 10 experimental, N = 6 controls) underwent classical threat conditioning exactly as described above for Day 1 but without the second STNV session. Subjects were tested for threat memory as described for Day 2. Vocalizations were recorded during the threat memory test as described above.

Study design, sample size, and eliminations

This study includes data from two experiments on social transmission of negative valence. The experimental unit for social experiments was the pair, which consisted of two co-housed animals, a “demonstrator” and an “observer.” In Experiment 1, observers and demonstrators were housed separately following testing on Day 1; in Experiment 2, observers and demonstrators were housed together following testing on Day 1. Experiment 1 included two groups: pairs with a shocked demonstrator (N = 32) and no-shock controls (N = 30). Following the first session of Day 1, one pair from each group was eliminated as outliers based on the demonstrator’s response to threat conditioning, leaving N = 31 pairs in the shock group and N = 29 control pairs for subsequent sessions. Experiment 2 also included pairs with a shock demonstrator (N = 22) and no shock controls (N = 21). One pair was eliminated as an outlier following the first session of Day 1, and an additional two pairs were eliminated following a technical error during Session 2 that prevented the collection of data, leaving N = 20 pairs in the shock group and N = 20 control pairs for subsequent sessions.

In the USV validation study, N = 10 single prairie voles experienced tone-shock pairings along with N = 6 no-shock controls. The following day, USVs were recorded from all subjects during a recall test. Subsequently, as part of Experiment 2, N = 9 observers per group were selected at random to have USVs recorded during the recall test on Day 2.

Blinding

Fully automated scoring was used for measures of freezing behavior and USVs. For manual scoring of self-grooming and rearing, video files were renamed by research staff not involved in testing, and behaviors were manually rated by a blinded experimenter using pre-defined criteria.

Immunohistochemistry

Aiming to explore the brain correlates of social learning, we quantified immediate early gene translation as a proxy for neural activity using c-Fos immunochemistry in a new group of voles. Exactly 80 min following the end of the first day of STNV, observers (control demonstrator N = 10; shocked demonstrator N = 10) were euthanized by isoflurane overdose and immediately perfused with 30 mL of sterile phosphate-buffered saline (PBS), followed by 20 mL of 1% paraformaldehyde (Sigma-Aldrich, St. Louis, MO) in PBS then 50 mL of 4% paraformaldehyde with 0.125% glutaraldehyde in PBS. Perfusions were done using a homemade perfusion pump set to a constant rate of 6 mL/min. Brains were dissected and submerged in a 4% paraformaldehyde solution in PBS for 24 hours and then transferred to a 30% sucrose solution in PBS for cryoprotection. Following the cryoprotection, perfused brains were cut into 40 µm sections using a cryostat (Leica CM3050 S, Deer Park, IL) and stored in PBS at 4 °C until stained.

Immunochemistry:

Brain sections were washed 5 x 5 min in Tris-buffered saline (TBS), incubated for 10 min in 1% sodium borohydride (NaBH4), and washed 3 x 5 min in TBS. Slices were incubated in a 0.2% Triton X-100 solution (Sigma-Aldrich) in TBS containing 10% bovine serum albumin (BSA; Sigma-Aldrich) for 1 h at room temperature. Sections were then incubated overnight in a TBS solution containing 0.05% Triton X, 2% BSA, and a primary rabbit polyclonal anti-Fos antibody (Cell Signaling, Cat #: 2250, 1:1000 dilution) while being rocked at 4°C. The next day, sections were washed 3 x 5 min in 0.05% TBS. Then they were incubated in TBS, 0.05% Triton X, and 2% BSA containing an Alexa Flour 488 goat anti-rabbit secondary antibody (Sigma-Aldrich) for 2 h at room temperature, protected from the light. Sections were washed 5 x 5 min in 1X TBS before being mounted and cover slipped using VectaShield media with DAPI (4′,6-diamidino-2-phenylindole; Sigma-Aldrich).

Imaging and counting:

Expression of c-Fos was quantified using immunofluorescence per nuclear area, to represent the Fos immunostaining across the entire brain region rather than a representative 20x image. This method takes advantage of the linear relationship between DAPI intensity and cell number [79-]. Sections were imaged at 4x magnification (Cytation5, Agilent, Santa Clara, CA) in the Integrated Core Facilities at the University of Toledo. Images were then analyzed for fluorescence intensity bilaterally at five target areas in the brain (anterior insula (AI), anterior cingulate cortex (ACC), basolateral amygdala (BLA), medial habenula (MHb), and lateral habenula (LHb) using the MCID software (Microcomputing Imaging Device, GE Healthcare Life Sciences, Marlborough, MA). Brain regions were identified and manually outlined bilaterally by matching the DAPI image to the mouse brain atlas. The optical density (OD) of fluorescence within the outlined area was then quantified for both the fluorescent antibody (Channel 1) and DAPI (Channel 2). We then calculated the ratio of these ODs for each outlined area, representing c-Fos immunofluorescence per nuclear area. From 1-5 measurements of OD ratio per brain region and side were collected from serial sections; these measurements were analyzed for coefficient of variation, and measurements outside of 30% coefficient of variation were removed. The average OD ratio measurements across sections was used as the primary outcome measure of interest. We added an example of 20x magnification images showing the merged DAPI staining and GFP (c-Fos) immunofluorescence from ACC, anterior insular cortex, amygdala, and habenula (Fig. S1, Supplementary file 1). 20x images were not used for quantification.

Statistics and reproducibility

All experimental data were parametrically distributed and had homogeneous variance, and parametric statistics were used in all analyses. For all ANOVAs, a sex factor was initially included; since no significant main effects or interaction effects of sex were found, the sex factor was removed from the final analyses. For both observers and demonstrators, threat acquisition was tested using a repeated measures ANOVA with post-hoc t-tests, with time as a within-subject factor and group as a between-subject factor (demonstrators, day 1 session 1, 2 × 15 ANOVA with 15 tones; observers, day 1 session 2, 2 × 5 ANOVA with 5 tones). Threat recall was assessed in demonstrators and observers using t-tests on average freezing across 5 tones (demonstrators, day 1 session 2; observers, day 2). We used Pearson correlations to examine the correlation in freezing time between observers and demonstrators in each group separately. We used t-tests to compare self-grooming duration and rearing frequency between the stress and control groups. The ultrasonic vocalization studies also used t-tests to primary frequencies between stress and control groups. For immunohistochemistry, we used separate two-factor ANOVAs (factors of group (between) and side (within) with post-hoc t-tests to compare c-Fos immunofluorescence per nuclear area in the left and right ACC, AI, BLA, MHb, and LHb.

We overlayed individual data points on bar graphs of the main results in a supplementary figure for more details about sample distribution (Fig. S2, Supplementary file 1).

Results

Social transmission of negative valence experiment

Classical fear learning in demonstrators

In the first session of STNV, demonstrators underwent classic fear conditioning, following a habituation period, in which they received 15 consecutive presentations of a tone paired with mild foot shocks (Fig. 1). Control demonstrators underwent an identical procedure without the presence of shocks. Demonstrators in the shock group (N = 32) and the no-shock control group (N = 30) did not differ in baseline freezing during habituation (t-test, t (60)=0.251, d = 0.06, p = 0.80) signaling that both groups had similar stress levels. Demonstrators in the shock group successfully acquired a within-session conditioned response over the course of fifteen tones relative to no-shock controls, as evidenced by increased total freezing during fifteen tones (ANOVA, main effect of group, F(1,60) = 42.8, η2 = 0.42, p < 0.001) and a difference in the 15-tone freezing response curve (ANOVA, group-time interaction, F(14,47) = 4.4, η2 = 0.57, p < 0.001, Fig. 2a), findings that corroborate the growing literature evidence of classical fear learning in rodent species (Fig. 2a). Additional information can be found in Supplementary file 2.

Social threat acquisition in observers

During the second session of Day 1, observers were placed across a transparent barrier (permitting sight and sound) from demonstrators during a habituation period followed by 5 presentations of the conditioned tone with no shocks. Observers did not differ in their baseline freezing during habituation in both groups (t-test, t (58)=0.074, d = 0.02, p = 0.94), suggesting similar baseline levels of stress between the two groups. Demonstrators in the shock group responded to the conditioned tone with elevated freezing relative to controls (t-test, t(58) = 2.7, d = 0.70, p = 0.009; Fig. 2b). During the threat recall session and while demonstrators in the shock group responded to the tone, observers did not show elevated freezing responses in comparison to observers of no-shock control demonstrators (t-test, t (58)=1.2, d = 0.31, p = 0.23; Fig. 2c). Nonetheless, we observed a correlation between freezing levels in observers and demonstrators in the experimental condition (Fig. 2d), but not in the control condition (Fig. 2e). Freezing in observers and demonstrators in the shock group went from uncorrelated during tone 1 (Pearson’s r = −0.054, p = 0.77) to correlated during tone 5 (Pearson’s r = 0.45, p = 0.011) and these two correlations were significantly different (Fisher’s transformation, p = 0.04). Additional information can be found in Supplementary file 2. This change in behavior in observers suggests a degree of emotional contagion during the social threat acquisition session.

Social transmission of negative valence (STNV) in observers

Observers were tested 24 hours after witnessing the shocked or unshocked demonstrators to measure their responses to the conditioned tone, including freezing, self-grooming, and rearing behaviors (Fig. 1). Observers in the shock group responded to the conditioned tone with elevated freezing compared to controls (t-test, t (58)=2.0, d = 0.52, p = 0.049; Fig. 2f,g), demonstrating the recall of a socially transmitted threat response. Observers in the shock group also showed a significant increase in time spent self-grooming (t-test, t = 2.2, d = 0.56, p = 0.035; Fig. 2h) and rearing bouts (t-test, t = 2.4, d = 0.63, p = 0.019; Fig. 2i) compared to observers in the control group, suggesting an increase in negatively valenced behaviors [61–64]. Additional information can be found in Supplementary file 2.

Ultrasonic vocalization (USVs) in observers as another indicator of STNV

USVs were measured as an additional marker of stress response in observers. We recorded USVs in a subset of observers during social threat recall on Day 2. Observers previously paired with demonstrators in the shock group produced USVs at a higher principal frequency than those paired with naïve demonstrators (one-tailed t-test, t (14)=2.0, d = 0.99, p = 0.035; Fig. 3a). This increase in principal frequency may reflect stress in observers from the shock group, consistent with changes in fear-conditioned demonstrators. Exploratory analysis of other call features also revealed shortening of average call length (t (14)=2.8, p = 0.015) and a decrease in average power (t(14)=2.9, p = 0.012) in observers paired with conditioned demonstrators (Fig. 3b,c), differences that were not observed when recording from directly fear conditioned voles. Additional information can be found in Supplementary file 3.

This may suggest that observers in the shock group produce different types of calls that are selective to socially transmitted negative valence. Further replication is needed to characterize USVs frequencies and calls during social transmission of negative valence.

Given that this task was not performed previously in prairie voles and that stress-responses USVs are not well established in these species, we conducted a control experiment during which we measured USVs during classical fear learning in a new subset of prairie voles to examine the type of calls and frequencies prairie voles exhibit during general stress and to examine whether it corroborates with what was found during STNV. In this control experiment, we exposed prairie voles to tone-shock pairs (or tone-only controls) and, 24 hours later, recorded USVs produced during a fear memory recall session. Prairie voles conditioned to the tone-shock produced USVs at a higher principal frequency (t-test, t (14)=2.6, d = 1.3, p = 0.022; Fig. 3d), suggesting that higher call frequencies are associated with negative valence and may represent generalized stressed calls. No other call features were reported as statistically significant during this CTC experiment (Fig. 3e,f). Additional information can be found in Supplementary file 4.

Control experiment for STNV

Impact of consoling behavior on social transmission of negative valence

In a separate consoling control (CC) experiment, sibling voles went through the same STNV paradigm as described above, except that they were both returned to the home cage following the 2 sessions on Day 1. We predicted that this housing condition would allow for social buffering behaviors in the home cage, including consoling behavior [75] and subsequently reducing the stress in demonstrators and the social transmission of negative valence in observers. As before, demonstrators in the shock group (N = 22) did not differ from controls (N = 21) in baseline freezing during habituation (t-test, t (41)=1.15, d = 0.35, p = 0.26), they successfully acquired a within-session conditioned freezing response (ANOVA, main effect of group, F(1,41) = 17.9, η2 = 0.30, p < 0.001; group-time interaction, F(14,28) = 3.1, η2 = 0.61, p = 0.006; Fig. 4a;), and they demonstrated the conditioned freezing response during session 2 (t-test, t (38)=3.4, d = 1.1, p = 0.002; Fig. 4b). As predicted, observers did not show any differences in within-session freezing response (ANOVA, main effect of group, F (1,38)=0.93, η2 = 0.024, p = 0.34; group-time interaction, F(4,35) = 0.79, η2 = 0.0.047, p = 0.79; Fig. 4c) or freezing to the tone on Day 2 ((t-test, t (38)=0.45, d = 0.14, p = 0.66; Fig. 4d), suggesting that reunion with the partner did enhance social buffering and reduced subsequent social learning. Supplementary file 2 contains the raw data of this experiment.

Fig. 4.

Consoling control for STNV. a-d: Social buffering and social threat conditioning. (a) As before, demonstrators exposed to tone-shock pairings (A) developed a conditioned freezing response and (b) demonstrated that conditioned freezing response while across a transparent barrier from an observer. (c) The observer did not acquire a within-session freezing response. (d) On Day 2, following unhindered overnight cohabitation between demonstrators and observers, observers of fear-conditioned demonstrators did not show a conditioned freezing response to the tone. Error bars show SEM; **p < 0.01

Neural correlates of STNV

We examined variations in c-Fos expression between observers of control (N = 10) and shocked (N = 10) demonstrators following STNV on Day 1, focusing on five key areas: the anterior cingulate cortex (ACC), anterior insula (AI), basolateral amygdala (BLA), and the medial and lateral habenula (MHb and LHb) (Fig. 5).

Fig. 5.

Immunohistochemistry. Representative images from one observer from each group showing DAPI staining and GFP (c-Fos) immunofluorescence in (a) anterior cingulate cortex, (b) anterior insular cortex, (c) amygdala, and (d) medial (left circle) and lateral (right circle) habenula. Dashed circles represent the quantified area for each region for both blue and green channels. Solid lines are 250 µm scale bars. (e) Relative differences in brain region-specific activity in observers of shocked demonstrators as compared to control observers. Bars show the quantitative differences in region-specific cellular activity, as measured by c-Fos immunofluorescence per nuclear area. *p < 0.05, **p < 0.01

Observers in the shock group showed an increase in c-Fos immunofluorescence in the ACC (ANOVA, main effect of group, F (1,18)=0.049, η2 = 0.20, p = 0.049), AI (ANOVA, main effect of group, F(1,17) = 7.3, η2 = 0.30, p = 0.015), BLA (ANOVA, main effect of group, F(1,16) = 9.1, η2 = 0.36, p = 0.008), and LHb (ANOVA, main effect of group, F(1,17) = 5.5, η2 = 0.24, p = 0.031), but not the MHb (ANOVA, main effect of group, F(1,17) = 0.006, p = 0.94) (Fig. 5), corroborating the hypothesis that LHb can be part of the SAS network.

Exploratory analysis on laterality in each region found that only the AI showed significant lateralization (ANOVA, group × side interaction, F(1,17) = 5.27, η2 = 0.24, p = 0.035), with the right AI activated more than the left in observers in the shock group (t-test, left vs. right AI in shock group, p = 0.026) but not different than the left in observers in the control group (p > 0.05). The right AI was more activated in observers in the shock group than in control observers (t-test, control vs. shock in right AI, p = 0.0061). The left AI was not different between the two groups (p > 0.05). Additional information can be found in Supplementary file 5.

Discussion

Here, we used a highly social animal model, prairie voles, to investigate the behavioral and neuronal correlates of social learning, a phenomenon that is deficient in autism spectrum disorder and other neurodevelopmental disorders [2, 3]. We first found that prairie voles can acquire threat and associate a neutral stimulus to aversive experience based on the observation of conditioned demonstrators undergoing distress (exhibiting freezing behavior) during a threat recall session. During the observation of stressed demonstrators, even though observers did not show an overall average increase in the level of freezing compared to control observers, their freezing level correlated significantly and positively with demonstrators’ freezing behavior in the experimental (shock) condition, and not in the control condition. This signifies that a synchronization occurred between the two and that there is a transmission of social information and that freezing in demonstrators is modulating in a timely and precise matter the behavior of observers.

During memory recall, prairie vole observers also showed an increase in classic negatively valenced responses, including increased rearing, increased self-grooming, and altered USVs. These results showed that both the threat value, through direct experiences, and the negative valence, through indirect experiences, of a cue can be socially transmitted through the observation of subtle social cues from a prairie vole demonstrator.

Self-grooming has been linked to the hypothalamus-amygdala axis where studies have found a correlation between increased activation of the amygdala and increased self-grooming in rodents exhibiting anxious responses [81–84]. Rearing has also been shown to be stimulus sensitive and increases in response to stress [13]. Therefore, the significant increases in rearing and self-grooming behaviors by the observers paired with conditioned demonstrators seen in STNV are indicative of threat acquisition and transmission of negative valence from their partners. It could also reflect anxiety or anticipation or conflict [81–84].

In addition to behavioral markers, social transmission of negative valence was also demonstrated by differential USVs in observers paired with demonstrators under distress in comparison to observers paired with non-stressed demonstrators. During social threat recall on day 2, socially conditioned observers exhibited higher frequency USVs on average compared to controls. This higher USV frequency was also exhibited in a separate group of prairie voles (in a separate experiment) that underwent themselves, as demonstrators, a classical fear learning task, during fear recall session in comparison to controls, signaling that this elevated frequency is an indicator of elevated stress levels. The characteristics of vocalizations serve as a vital communication tool between members of species and can vary widely when compared to those of other species. Although USVs in prairie voles have been understudied in OFL experiments, changes in specific features of their USVs have previously been documented as indicators of distress. For example, a positive correlation between corticosterone levels and number of calls has been reported in prairie vole pups in response to social isolation [85]. Additionally, pup vocalizations have been shown to be uniquely potentiated during reunions with mothers, but not fathers, in pair-bonded, bi-parental rearing conditions [86], supporting the notion that vole vocalizations vary in both quality and quantity in response to varying environments. Further promoting the association between elevated corticosterone and increased fear, one study demonstrated that sibling-bonded voles exhibited increased heart rates and circulating corticosterone when observing a sibling subjected to a physical stressor [87]. Although USVs were not measured, these findings support the ability of voles to vicariously experience biologically measurable stress and strengthens conclusions of studies where heart rate or corticosterone levels were correlated with varying USV characteristics. Notably, a higher frequency and a lower vocal prosody were associated with increased heart rates [88], highlighting that USVs reflect the automimic state.

In addition to higher principal frequency of USVs, socially-conditioned observers exhibited shorter USV call lengths and reduced average power as compared to controls. These differences in length calls and power were not observed in prairie voles undergoing classical fear learning and foot shocks. It is possible that these indicators are specific to distress in social context and maybe selective indicators of stress for others instead of stress for self. Further replications are necessary to conclude if these exploratory findings on shortness in call lengths and a reduction in call power are generalized markers of distress in response to the stress of others in prairie voles and in other species with highly translational values.

To corroborate further that these changes in behavioral and ultrasonic vocalizations are specific markers of negative valence triggered by the distress of others and not by random factors, we conducted an additional control experiment in which we manipulated the distress levels in demonstrators by adding a social buffering condition. We showed that by reuniting observers and their distressed partners in their home cage after classical fear learning and threat recall on day 1, socially-conditioned observers did not show increased freezing behavior in response to the conditioned context on day 2 in comparison to controls. Reunion, which is well documented in the literature as being associated with consoling behavior [75], is likely to have acted as social buffering and therefore reduced the levels of distress in the partner, which would have led to a more neutral response during social threat recall on the second day. Lack of direct consoling data (due to logistic difficulties) is one limitation of the study and future studies using STNV in voles can provide further validation to the role of social buffering in altering social transmission of negative valence.

At the neuronal level, we conducted c-Fos immunofluorescence after observation of the stressed partner to assess the brain activity of key areas of interest as stated in the hypothesis, including LHb, BLA, AI and ACC, as part of the SAS network. These brain circuits are interconnected and involved in emotional processing, decision-making, and reinforcement learning especially in relation to negative valence [89].

In line with the hypothesis, we found elevated neuronal activity in these areas in socially-conditioned observers in comparison to control observers, indicating that the LHb can play an important role in social learning.

Most literature focuses on the habenula in non-social reinforcement learning and negative valence [58, 90–92], where the lateral habenula, specifically, has shown to be the central hub for aversive and impulsive action integration [59, 93]. The LHb neurons receive afferent input from the limbic system and basal ganglia, helping to modulate motivation and emotional information [94]. It then exerts inhibitory control over midbrain dopaminergic (ventral tegmental area, substantia nigra pars compacta) and serotonergic (raphe nuclei) centers through activation of gamma-aminobutyric acid (GABA) neurons. Due to this, LHb stimulation is typically associated with behavioral responses to aversive stimuli and suppression of reward-related activity [58, 93, 95–99]. This leads to the performance of avoidance responses such as social withdrawal and reduced motivation where in fact overactivation has been correlated with depressive-like phenotypes: anhedonia, behavioral despair, and heightened anxiety [90, 91]. On the contrary, suppression of the LHb increases dopamine turnover, leading to rewarding phenotypes, facilitating approach behaviors, reduced anxiety, and diminished avoidance responses [93, 98].

Disruptions within the LHb circuitry are noted to contribute to the dysregulation of underlying impairments in learning, decision-making, and affective processing; all of which are critical for socially based behaviors [95, 97, 100]. Although a more detailed exploration of this connectivity is necessary for application across diverse psychiatric disorders, a study comparing magnetic resonance imaging in humans found that across all ages included, the habenula was larger in those diagnosed with ASD when compared to controls [101]. Other literature shows experimental models in adolescent rats that have shown the exclusion from social play induces anxiety-like phenotypes and increases c-Fos expression within the LHb, implicating its role in emotional consequences of early social stress [102]. A more recent study on social fear conditioning in mice found that neurons projecting from the LHb to the medial prefrontal cortex were highly activated during social fear, and that inhibiting this LHb–medial prefrontal cortex pathway significantly reduced fear responses [103]. Another interesting study showed that serotonin release was elevated in the LHb during emotional contagion in mice and serotonin supports vicarious emotions leading to resilience [104]. These recent studies corroborate our findings with regards to the role of the LHb in social learning.

Additionally, the higher activation of the ACC and AI in observers of stressed voles is in accordance with previous literature involving these areas of social learning and emotion contagion. Shank3 mutant mice, a model of autism, display structural and functional impairments in the ACC, associated with deficits in social interaction [105, 106]. Studies in humans, macaques, and rodents show that the ACC is essential for the social success of an individual in a group environment [107–110]. They have also shown that the ACC, AI, and BLA are jointly activated in empathy-related or emotion-contagion studies [48, 111–113], reinforcing the idea of complementary motor and sensory system feedback [114–116] as well as connections through the SAS network. These findings contribute to our understanding of stress-induced c-Fos activation patterns in specific brain regions, shedding light on neural responses to social stressors and implications in those diagnosed with autism.

Limitations

We lack explicit consoling data in the control experiment to show that social buffering (duration of consoling behavior or allogrooming) is the cause of lack of freezing during social threat recall in the behavioral control experiment. Future experiments should include strangers to study the behavioral and physiological responses in observers to the stress of unfamiliar others. Another limitation of the experiments in this manuscript is that the findings are correlational with c-Fos and not causal. We need further experiments involving optogenetics or other approaches to inhibit or excite LHb neurons in observers to examine if this brain area is essential for social learning. Despite the translational aspect of this work and its relevance to neurodevelopmental disorders, replicating these experiments in genetic animal models relevant to autism such as Shank3 knockout mice would strengthen the pathophysiological relevance of the lateral habenula dysfunction or deficits in other areas of the SAS network to neurodevelopmental disorders.

Conclusions

Our findings suggest that prairie voles are capable of socially transmitting information and can acquire new information through observation of subtle social cues. We discovered that the lateral habenula area is involved in social transmission of negative valence. Further molecular and pharmacological manipulations confirming the essential role of this area in the SAS network is crucial. In humans, investigating the role of the lateral habenula in social learning and in neurodevelopmental disorders such as autism can be also critical as it can shed light to new targeted brain circuitry or areas implicated in social affective and social salience processes.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ACC

Anterior cingulate cortex

AI

Anterior insula

ASD

Autism spectrum disorder

BLA

Basolateral amygdala

LHb

Lateral habenula

MHb

Medial habenula

SAS network

Social affect salience network

STNV

Social transmission of negative valence

USVs

Ultrasonic vocalizations

Author contributions

D.G-P. conceptualized the study, conducted the experiments and data curation, performed data analysis and co-wrote the manuscript. J.H-K. conducted the visualization, organization, and formatting of data and results, co-wrote the manuscript, and contributed to the interpretation of findings. S.P. and B.A.H. conducted the experiments, data curation, and data analysis. J.P.B. conducted/supervised experiments, conducted analysis, and co-wrote the manuscript. E.A. conceptualized the study design, supervised the study, conducted analysis, performed visualization, and co-wrote the final version of the manuscript.

Funding

This work was supported by ProMedica Health System Foundation (207007).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval

All the housing, breeding, handling, and experimental procedures were approved by the University of Toledo Institutional Animal Care and Use Committee (IACUC) and were conducted in compliance with the Animal Welfare Act and the National Research Council’s Guide for the Care and Use of Laboratory Animals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.