Article https://doi.org/10.

1038/s41467-024-52294-6

Sex differences in neural representations of

social and nonsocial reward in the medial
prefrontal cortex
Received: 23 April 2024 Jennifer Isaac1,2, Sonia Corbett Karkare 1,2, Hymavathy Balasubramanian1,2,
Nicholas Schappaugh2, Jarildy Larimar Javier 1,2, Maha Rashid1,2 &
Accepted: 29 August 2024
Malavika Murugan 1,2

Check for updates The reinforcing nature of social interactions is necessary for the maintenance
1234567890():,;
1234567890():,;

of appropriate social behavior. However, the neural substrates underlying

social reward processing and how they might differ based on the sex and
internal state of the animal remains unknown. It is also unclear whether these
neural substrates are shared with those involved in nonsocial rewarding pro-
cessing. We developed a fully automated, two choice (social-sucrose) operant
assay in which mice choose between social and nonsocial rewards to directly
compare the reward-related behaviors associated with two competing stimuli.
We performed cellular resolution calcium imaging of medial prefrontal cortex
(mPFC) neurons in male and female mice across varying states of water
restriction and social isolation. We found that mPFC neurons maintain largely
non-overlapping, flexible representations of social and nonsocial reward that
vary with internal state in a sex-dependent manner. Additionally, optogenetic
manipulation of mPFC activity during the reward period of the assay disrupted
reward-seeking behavior across male and female mice. Thus, using a two
choice operant assay, we have identified sex-dependent, non-overlapping
neural representations of social and nonsocial reward in the mPFC that vary
with internal state and that are essential for appropriate reward-seeking
behavior.

Animals, including humans, are capable of flexibly choosing between access to a conspecific8–12. Despite these commonalities, it is unclear
various rewarding stimuli such as food, water, and social interactions. if social and nonsocial reward-related information is encoded by the
Although considerable progress has been made toward under- same or distinct populations of neurons in various brain regions.
standing the neural substrates underlying nonsocial reward-related Some lines of evidence support the common currency view13,14, which
behaviors, less is known about social reward-related behaviors due to proposes that the same reward circuitry is used to process both
the inherent complexity of social interactions1. Across species, social and nonsocial rewards. For example, neurons in the primate
affiliative social interactions appear to engage the same neural cir- amygdala respond similarly to both juice and pictures of monkeys15.
cuits, including the canonical reward-related mesolimbic system, as In contrast, other evidence supports social specificity, the idea that
nonsocial rewards2–5. Additionally, social interactions, like nonsocial social stimuli are processed by distinct populations of neurons in the
rewards, can act as positive reinforcers, with rodents returning to the brain16–18. Thus, it remains unresolved whether social and nonsocial
context of a social interaction6,7 and engaging in operant tasks for reward-related behaviors are mediated by shared or separate

1
Neuroscience Graduate Program, Emory University, Atlanta, GA 30322, USA. 2Department of Biology, Emory University, Atlanta, GA 30322, USA.
e-mail: [email protected]

Nature Communications | (2024)15:8018 1

Article https://doi.org/10.1038/s41467-024-52294-6

neuronal populations, specifically in the medial prefrontal cor- sucrose reward-seeking behavior and a greater proportion of mPFC
tex (mPFC). neurons that were modulated by sucrose reward relative to control
The mPFC is a central node in an interconnected network of brain conditions. By tracking the same neurons across imaging sessions, we
regions that has been implicated in both nonsocial reward found that the newly recruited sucrose reward-responsive neurons
processing19–24 and a wide range of social behaviors25–30. While the role were largely derived from previously latent, reward-unresponsive
of the mPFC has been extensively studied in the context of nonsocial neurons. We also found that acute social isolation differentially affec-
reward-related behaviors19–24, less is known regarding its contribution ted reward-seeking behavior in male and female mice. However, it did
to social reward-seeking. In rodents, recent work has shown that mPFC not affect the proportion of mPFC neurons that were modulated by
neurons are active in the proximity of conspecifics and show increased social reward. Instead, social isolation altered the amplitude of the
synchrony when mice engage in affiliative social interactions25,29,31. neural responses to social but not sucrose reward in a sex-dependent
Furthermore, mPFC dysfunction is thought to underlie the social manner, with increased amplitude in socially-isolated female mice and
behavioral deficits associated with Autism Spectrum Disorders decreased amplitude in socially-isolated male mice relative to control
(ASD)32–34. These social behavioral deficits are often accompanied by conditions. Overall, these findings suggest that although male and
deficits in the evaluation and processing of nonsocial reward-related female mice behave similarly while flexibly choosing between
stimuli35–37. Consequently, determining whether social and nonsocial social and nonsocial rewards on a two choice operant assay, there are
rewards share neural substrates within the mPFC has important sex differences in how social and nonsocial rewards are represented
implications for improving therapeutic options for individuals and in how changing internal state affects these representations in
with ASD. the mPFC.
The internal state of an animal also plays a large role in deter-
mining its motivation to engage in certain rewarding behaviors and the Results
perceived value of the chosen reward38–40. While there is emerging Male and female mice demonstrate positive reinforcement for
literature to suggest that brainwide changes in neural dynamics result social and sucrose reward stimuli
from changes in internal state through different types of deprivation In order to directly compare nonsocial and social reward-related
such as thirst41 and social isolation42, how the internal state of an animal behavior and the neural circuits underlying these behaviors, we
modulates reward representations in the mPFC is poorly understood. developed a self-paced, automated two choice (social-sucrose) oper-
Despite evidence that sex differences may contribute to how animals ant assay in which mice were trained to associate two separate choice
engage with and process rewarding stimuli, sex is an often overlooked ports with access to either a sucrose reward or a same sex conspecific
component when determining how an animal’s internal state mod- (social reward) over 1 hour daily sessions (Fig. 1a–c).
ulates reward-related behaviors. In fact, recent studies have shown that Initially, mice on restricted water access are trained to associate
male and female mice exhibit divergent learning and decision-making nose-poking a choice port with sucrose reward delivery at a reward
strategies in reward-related behaviors22,43,44. Moreover, sex differences port located on the adjacent wall of the operant chamber through a
in mPFC reward representations were shown to underlie some of the series of increasingly difficult assays (Supplementary Fig. 1 and see
behavioral differences observed in nonsocial reward-seeking45. How- Methods). Mice showed a consistent decrease in the time from nose-
ever, it is unknown if these sex differences extend to social reward- poking the choice port to accessing the reward port (sucrose reward
related behaviors. Additionally, even though studies have shown dif- latency) with training (Supplementary Fig. 1h: one-way ANOVA with
ferential impacts of thirst and social isolation on reward-related post-hoc t tests, p = 4.01*10−8, for detailed statistics see Supplementary
behaviors in male versus female mice, it is unclear if there are sex Data 1). During the final training stage, a social component was added
differences in how reward representations are affected by varying to the assay to allow mice to choose between a social and sucrose
internal state. reward (Supplementary Fig. 1a, Supplementary Fig. 2). At the start of
To address these questions, we developed a two choice (social- each trial on the two choice operant assay, two choice ports on one
sucrose) operant assay in which mice can freely choose between social wall of the operant chamber illuminate (Fig. 1b, trial start). Nose-
and sucrose (nonsocial) rewards in order to directly compare these poking one of the illuminated choice ports gives mice access to 10 μl of
two types of reward-related behavior. We found that, under control 10% sucrose at the sucrose reward port for up to 8 s (Fig. 1b, sucrose
conditions (no social, food, or water deprivation), both male and trial, blue panel), while nose-poking the other illuminated choice port
female mice showed a comparable preference for social and sucrose opens an automated gate to give mice access to a novel, same sex
rewards on the two choice assay, with female mice preferring social conspecific (Fig. 1b, social trial, red panel). Both male and female mice
reward slightly more than sucrose reward. We then sought to deter- showed a decrease in their social reward latency across five training
mine if distinct ensembles of neurons in the mPFC encode social ver- sessions on the two choice operant assay, as mice learned to associate
sus nonsocial rewards. Using cellular resolution calcium imaging, we the social choice port with a social reward (Supplementary Fig. 2g, o:
identified largely non-overlapping ensembles of mPFC neurons that one-way ANOVA with post-hoc t tests, male: p = 8.19*10−6, female:
responded to social versus sucrose reward. While the neurons that p = 0.016). Male mice also showed a decrease in social reward fails,
responded to social reward displayed mostly excitatory responses, the defined as failure to enter the social zone when the gate is open, with
neurons modulated by sucrose reward exhibited both inhibitory and training (Supplementary Fig. 2h: one-way ANOVA with post-hoc t tests,
excitatory responses. Interestingly, the neurons that displayed an p = 0.0011). Additionally, both sexes showed a stable number of suc-
inhibitory response to sucrose reward were also excited in response to cessful social trials across five training sessions (Supplementary
social reward. These social and sucrose reward ensembles were less Fig. 2e, m: one-way ANOVA with post-hoc t tests, male: p = 0.11,
overlapping in female mice compared to male mice. After character- female: p = 0.98).
izing these distinct social and nonsocial reward representations in the After five training sessions on the two choice operant assay, mice
mPFC, we optogenetically manipulated mPFC neurons during the were removed from restricted water access and run for an additional
reward period of the two choice operant assay and found that it dis- five sessions (post-training, full water access). Post-training, we found
rupted reward-seeking behavior in both male and female mice. We that male mice in control conditions (not water, food, or socially
then examined how varying the internal state of the animal affected deprived) completed an equivalent number of successful sucrose and
reward representations in the mPFC by either water depriving or social trials (Fig. 1d: paired t test, p = 0.73), while female mice com-
socially isolating mice prior to imaging them on the two choice oper- pleted slightly more successful social than sucrose trials (Fig. 1h: paired
ant assay. Water deprived mice of both sexes showed increased t test, p = 4.91*10−4) at comparable choice latencies (Fig. 1e, i: paired

Nature Communications | (2024)15:8018 2

Article https://doi.org/10.1038/s41467-024-52294-6

t test, male: p = 0.12, female: p = 0.93). Male mice were slightly slower Although female mice completed slightly more successful social
to engage with social compared to sucrose reward (Fig. 1f, paired t test, trials and slightly fewer successful sucrose trials than male mice
p = 2.04*10−8), likely reflecting their longer exposure to sucrose reward (Supplementary Fig. 3i: two-factor ANOVA with sex (male/female) and
across all training assays. In contrast, female mice demonstrated trial type (sucrose/social) as factors, interaction: p = 8.00*10−4, sex:
equivalent reward latencies for both trial types (Fig. 1j: paired t test, p = 0.78, trial type: p = 0.0035, with post-hoc unpaired t tests com-
p = 0.093). Male and female mice made more sucrose reward fails than paring sex within trial type, sucrose: p = 0.011, social: p = 0.029), both
social reward fails (Fig. 1g, k: paired t test, male: p = 1.0*10−7, female: sexes completed significantly more social and significantly fewer
p = 4.10*10−6). sucrose trials when compared to partial water access conditions

Nature Communications | (2024)15:8018 3

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 1 | Male and female mice find both social and sucrose stimuli positively One-way ANOVA (p = 1.34*10−8) with post-hoc t tests (SG vs OG: p = 8.44*10−9; SG vs
reinforcing on a two choice (social-sucrose) operant assay. a, b Schematic of the ECG: p = 0.052; OG vs ECG: p = 0.10). There was no difference in sucrose choice
automated two choice (social-sucrose) operant assay in which a mouse can freely latency (m), sucrose reward latency (n) or number of sucrose reward fails (o)
choose (nose-poke) to obtain either a sucrose (right, top blue panel) or social (right, between groups. One-way ANOVA (m, p = 0.080; n, p = 0.19; o, p = 0.078). p Mice
bottom red panel) reward. c Timeline of an example behavioral session showing run on the two choice assay with a social target (SG) completed more social trials
social (red squares) and sucrose (blue squares) choices followed by the respective than mice run with an empty cage (ECG) or an object (OG). One-way ANOVA
reward consumption (red or blue arrowheads) or failure to enter the reward zone (p = 3.30*10−10) with post-hoc t tests (SG vs OG: p = 2.63*10−9; SG vs ECG:
(yellow circles) over an hour. Fully trained male mice (d) complete an equivalent p = 1.17*10−4; OG vs ECG: p = 0.87). Across all conditions, mice showed similar social
number of successful sucrose and social trials, while female mice (h) complete choice latency (q) and number of social reward fails (s). One-way ANOVA
more successful social than sucrose trials. Paired t test (male: number of trials: (q, p = 0.92; s, p = 0.44). r Mice run with a social target (SG) showed similar social
p = 0.73; female: p = 4.91*10−4). Male and female mice (e, i) show a similar choice reward latency to mice run with an empty cage (ECG) and decreased social reward
latency for sucrose and social choices. Paired t test (male: p = 0.12; female: p = 0.93). latency compared to mice run with an object (OG). One-way ANOVA (p = 2.00*10−4)
Male mice (f) were slightly faster to consume sucrose reward than social reward, with post-hoc t tests (SG vs OG: p = 1.04*10−4; SG vs ECG: p = 0.46; OG vs ECG:
while female mice (j) had equivalent latencies to consume social and sucrose p = 0.19). SG: n = 21 mice, ECG: n = 4 mice, OG: n = 10 mice, 3 sessions per mouse.
reward. Paired t test (male: p = 2.04*10−8; female: p = 0.093). Male (g) and female (k) Box plots: center line denotes median, box edges indicate the 25th and 75th per-
mice made fewer social reward fails than sucrose reward fails. Paired t test (male: centiles, and whiskers extend to ±2.7σ. *p < 0.05. Behavioral assay schematic (a)
p = 1.00*10−7; female: p = 4.10*10−6). N = 21 male mice, 12 female mice, 3 behavioral was created with BioRender.com released under a Creative Commons Attribution-
sessions per mouse. l Under full water access conditions, mice run on the two NonCommercial-NoDerivs 4.0 International License https://creativecommons.org/
choice operant assay with a social target (SG) completed fewer successful sucrose licenses/by-nc-nd/4.0/deed.en. Mouse schematic (b) adapted from Open Clipart by
trials when compared to mice from the object group (OG) and a similar number of lemmling.
successful sucrose trials when compared to mice from the empty cage group (ECG).

during training (Supplementary Fig. 3a, e: two-factor ANOVA with To further confirm the goal-directed nature of the social reward-
water condition (PW/FW) and trial type (sucrose/social) as factors; seeking behavior observed on the two choice operant assay, we
male — interaction: p = 2.63*10−26, water condition: p = 9.54*10−13, trial developed an additional multi-choice assay in which a third choice port
type: p = 1.22*10−27, with post-hoc unpaired t tests comparing water was introduced that was not associated with any reward (Supple-
condition within trial type, sucrose: p = 1.04*10−20, social: p = 2.32*10−5; mentary Fig. 5a, b). Mice run on this assay preferentially increased the
female — interaction: p = 4.93*10−20, water condition: p = 5.76*10−6, trial number of successful social but not null trials completed over time and
type: p = 3.02*10−9, with post-hoc unpaired t tests comparing water were significantly faster to enter the corresponding reward zone on
condition within trial type, sucrose: p = 4.49*10−13, social: p = 4.14*10−8). sucrose and social trials but not null trials (Supplementary Fig. 5c, d:
Additionally, male mice displayed shorter reward latencies to consume two-factor ANOVA with training day (day 1/day 9) and trial type
sucrose reward and made fewer sucrose reward fails compared to (sucrose/social/null) as factors, interaction: p = 0.0059, training day:
female mice (Supplementary Fig. 3k, l: two-factor ANOVA with sex p = 0.68, trial type: p = 5.02*10−13, with post-hoc unpaired t tests com-
(male/female) and trial type (sucrose/social) as factors, reward latency- paring trial type within training day, day 1 — suc v soc: p = 1.02*10−4, suc
interaction: p = 0.032, sex: p = 4.00*10−4, trial type: p = 1.64*10−7, with v null: p = 1.45*10−4, soc v null: p = 0.48, day 9 — suc v soc: p = 0.078, suc
post-hoc unpaired t tests comparing sex within trial type, sucrose: v null: p = 5.21*10−4, soc v null: p = 0.013; Supplementary Fig. 5e: two-
p = 1.29*10−5, social: p = 0.35; reward fails — interaction: p = 0.010, sex: factor ANOVA with choice type (sucrose/social/null) and reward zone
p = 0.0048, trial type: p = 2.79*10−15, with post-hoc unpaired t tests (sucrose/social) as factors, interaction: p = 2.44*10−13, choice type:
comparing sex within trial type, sucrose: p = 0.0073, social: p = 0.28), p = 0.41, reward zone: p = 0.10, with post-hoc unpaired t tests com-
male and female mice otherwise performed similarly when seeking paring reward zone within choice type, sucrose: p = 2.19*10−8, social:
social rewards on the two choice operant assay (Supplementary p = 1.82*10−5, null: p = 0.41). Finally, when mice were run on a social
Fig. 3j). Furthermore, female reward-seeking behavior on the two extinction assay for six consecutive sessions during which the social
choice operant assay was not affected by estrous cycle (Supplemen- target was removed after five consecutive sessions of full water access
tary Fig. 6a, b) or mouse strain (Supplementary Fig. 6c–f). These (FW) on the two choice operant assay (Supplementary Fig. 6g), mice
findings suggest that when male and female mice have ad libitum preferentially decreased the number of successful social trials and
access to water, they are motivated to seek both social and nonsocial increased the number of successful sucrose trials (Supplementary
rewards (Fig. 1, Supplementary Fig. 3). Fig. 6h, i: two-factor ANOVA with social condition (FW/extinction) and
Importantly, mice trained on the two choice operant assay with- trial type (sucrose/social) as factors, interaction: p = 0.0047, social
out a social target present (empty cage group, ECG) or with a novel condition: p = 0.55, trial type: p = 0.66, with post-hoc paired t tests
object instead of a social target (object group, OG) significantly comparing social condition within trial type social: p = 0.0295, sucrose:
decreased the number of successful empty cage/object trials, while p = 0.0063). These findings are among the first to demonstrate that
increasing the number of sucrose trials over five consecutive sessions social interaction can promote positive reinforcement of reward-
(Supplementary Fig. 4b, f: one-way ANOVA with post-hoc t tests, seeking behavior similarly to sucrose consumption in an operant assay.
number of sucrose trials — ECG: p = 0.01, OG: p = 1.25*10−7; number of
social trials — ECG: p = 5.51*10−6, OG: p = 3.0*10−4). When these mice had Cellular resolution imaging of mPFC neurons in male and female
ad libitum water access, they continued to complete significantly fewer mice during the two choice operant assay
social, but not sucrose, trials when there was no social target (empty To determine what information mPFC neurons encode during social
cage or object) compared to mice run on the two choice operant assay and nonsocial reward-related behavior, we performed cellular resolu-
with a social target (Fig. 1l, p: one-way ANOVA with post-hoc t tests, tion calcium imaging of mPFC neurons during the two choice operant
number of sucrose trials: p = 1.34*10−8; number of social trials: assay (Supplementary Movie 1). We injected an AAV expressing the
p = 3.30*10−10), which indicates that social reward-seeking behavior is calcium indicator GCaMP6f and implanted a gradient-index (GRIN)
positively reinforced by the presence of a social target and not solely lens into the mPFC of 9 male and 6 female mice (Fig. 2a). All lens-
by the gate opening or novelty-seeking behavior (Fig. 1l–s, Supple- implanted mice were then successfully trained on the two choice
mentary Fig. 4). operant assay. After 3 weeks of training, the lens was coupled to a

Nature Communications | (2024)15:8018 4

Article https://doi.org/10.1038/s41467-024-52294-6

head-mounted miniscope to record activity of GCaMP6f-expressing Neurons were classified as task-modulated if the maximum or
mPFC neurons while mice completed the two choice operant assay. minimum of their average activity differed from a null distribution
CNMFe was used to detect the activity of individual neurons46. Post- generated by randomly shuffling neural activity (see Methods, exam-
training, in full water access conditions, lens-implanted imaging mice ple task-modulated neurons Supplementary Fig. 7a) within a three
were run on a minimum of 3 daily sessions prior to imaging mPFC second window around the occurrence of a task event (trial start, social
neural activity in the two choice operant assay. We imaged a total of or sucrose choice, social or sucrose reward, Fig. 2f)20. We found that
459 neurons from 9 male mice and 570 neurons from 6 female mice in many mPFC neurons showed significantly modulated time-locked
control conditions (Fig. 2b, d). GRIN lens placement and viral expres- responses to specific behavioral events in the two choice operant assay
sion were confirmed on post-hoc histology (Fig. 2c, e). (Fig. 2g, h, Supplementary Fig. 7b–e). In fact, across all mPFC neurons

Nature Communications | (2024)15:8018 5

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 2 | Cellular resolution calcium imaging of mPFC neurons during the Coordinates are relative to bregma. f A schematic of the two choice operant assay
two choice operant assay. a Top row: Schematic of the viral strategy used to label events to which mPFC neuronal activity was aligned, from left to right, trial start,
mPFC neurons with GCaMP6f. Bottom row: Imaging setup of GRIN lens placement social choice, sucrose choice, social reward, and sucrose reward. Top row: Heat-
in the mPFC, including representative images showing GRIN lens placement (left), maps of average normalized fluorescence of all mPFC neurons that are significantly
GCaMP6f expression (middle, GCaMP6f in green; DAPI in blue) in the mPFC at modulated (excited or inhibited) by each task event in male (g) and female (h) mice.
(from left to right) 4x, 20x, and 60x magnification. Right panel shows nuclear Neurons are sorted by the time of maximum fluorescence across each task event.
exclusion of GCaMP6f. Scale bars: 500 µm, 250 µm and 25 µm. A pie chart showing Bottom row: Average normalized fluorescence traces of the neurons from the
the number of neurons recorded from each male (b, n = 459 neurons, 9 mice) and corresponding heatmap that are significantly modulated by each task event. Sha-
female mouse (d, n = 570 neurons, 6 mice). Reconstruction of GRIN lens placement ded error regions indicate ± SEM. Proportions of total recorded neurons that are
in the nine male (c) and six female (e) mice using WholeBrain software84. Each modulated by the various task events in male (i) and female (j) mice (male: n = 459
colored line corresponds to the same colored slice in the pie chart and shows the neurons, 9 mice; female: n = 570 neurons, 6 mice). Mouse schematic (f) adapted
position of the lens in the Allen Mouse Brain Common Coordinate Framework. from Open Clipart by lemmling.

recorded in male mice during the two choice operant assay, 3.27% responses across both sexes. These findings suggest that mPFC neu-
(n = 15/459) were significantly modulated by trial start, 8.71% (n = 40/ rons differentially represent social and sucrose choice in male versus
459) were significantly modulated by social choice, 9.37% (n = 43/459) female mice.
were significantly modulated by sucrose choice, 29.19% (n = 134/459) In fact, we observed that neural representations of social and
were significantly modulated by social reward and 19.17% (n = 88/459) sucrose choice in the mPFC are largely non-overlapping in both male
were significantly modulated by sucrose reward (Fig. 2i). Across all and female mice and significantly more non-overlapping than expec-
mPFC neurons recorded from female mice during the two choice ted by chance when compared to a shuffled distribution (Fig. 3e, f: %
operant assay, 1.05% (n = 6/570) were significantly modulated by trial non-overlapping, male: 83.82%, n = 57/68, shuffled: 66.37 ± 3.29;
start, 16.67% (n = 95/570) were significantly modulated by social female: 93.02%, n = 80/86, shuffled: 72.37 ± 2.81, ± standard deviation,
choice, 9.30% (n = 53/570) were significantly modulated by sucrose unpaired t test, male v shuffled: p = 1.56*10−74; female v shuffled:
choice, 35.44% (n = 202/570) were significantly modulated by social p = 3.51*10−88; see “Task modulation analysis” in Methods). Although
reward and 17.54% (n = 100/570) were significantly modulated by neural representations are non-overlapping in both male and female
sucrose reward (Fig. 2j). We found similar fractions of task-modulated mice (proportion z test, p = 0.07), we were able to decode subsequent
neurons when using an encoding model to temporally separate choice at an earlier time point in female mice compared to male mice
calcium activity kernels associated with various behavioral events (Fig. 3g: female: 3.5 s, male: 1.5 s prior to choice). We were also able to
(Supplementary Fig. 8; proportion z test, male — choice: p = 0.86, decode the choice made by the mouse from mPFC population activity
reward: p = 0.39; female — choice: p = 0.12, reward: p = 0.15). Previous with higher accuracy in female mice compared to male mice (Fig. 3h:
literature has shown that mPFC neurons respond more to action (i.e., average decoding accuracy, male: 56.09 ± 2.23, female: 67.03 ± 2.43;
nose-poke, reward consumption) than stimulus (i.e., trial start) unpaired t test, p = 0.0090). Additionally, we were able to decode the
events21. Consistent with these findings, we found that a greater pro- sex of the animal from mPFC population activity associated with
portion of mPFC neurons responded to choice and reward consump- choice on both trial types. (Fig. 3i: average decoding accuracy, social
tion than to trial start (Supplementary Fig. 9a: male — stimulus events: choice: 63.70 ± 1.03, sucrose choice: 61.68 ± 1.13; unpaired t test,
3.27%, 15/459, action events: 66.45%, 305/459; female — stimulus p = 0.19; Supplementary Fig. 9j: average F1 scores, social choice:
events: 1.05%, 6/570, action events: 78.95%, 450/570; proportion z test, 0.61 ± 0.014, sucrose choice: 0.61 ± 0.017; unpaired t test, p = 0.55).
male and female: p < 0.00001). Additionally, we found that more The largely non-overlapping nature of the choice responses was cor-
mPFC neurons were modulated by reward than choice (Supplemen- roborated by plotting the neural trajectories of mPFC population
tary Fig. 9b, c: male — choice: 18.08%, 83/459, reward: 48.37%, 222/459; activity on the first 3 principal components (PCs) in a 3 s window
female — choice: 25.96%, 148/570, reward: 52.98%, 302/570; propor- around choice port entry in both male and female mice (Fig. 3k, m:
tion z test, male and female: p < 0.00001). total variance explained, male: 59.2%, female: 57.6%) and calculating
pairwise Euclidean distances of population vectors within trial type
Sex differences in selectivity of social and nonsocial reward- versus between trial type (Fig. 3l, n). In both sexes, we found that
related neural representations the Euclidean distance within trial type was significantly less than
Given the well-defined role of the mPFC in decision-making47,48, we the distance between trial type across social and sucrose trials
next examined the responses of choice-modulated neurons during the (Fig. 3l, n: male — average distance within: 54.71 ± 4.76, between:
two choice assay. We found that choice-modulated neurons reliably 60.87 ± 5.13; paired t test, p = 5.76*10−4; female — average distance
showed peak fluorescence around the time of choice port entry and within: 49.89 ± 2.55, between: 71.93 ± 2.99; paired t test, p = 0.0048).
that the timing of peak fluorescence did not vary with reward latency Additionally, we used Mahalanobis distance to determine if social and
(Fig. 3a, c). In contrast, the timing of the peak fluorescence of reward- nonsocial choice responses were different at a population level (Sup-
modulated neurons varied with reward latency when aligned to choice plementary Fig. 9k, l). We found that there was a significantly larger
port entry (Fig. 3b, d). Furthermore, we found that a similar proportion population response to social choice compared to sucrose choice in
of neurons responded to social (n = 8.71%, 40/459) and sucrose male and female mice (Supplementary Fig. 9k, l: unpaired t test, male:
(n = 9.37%, 43/459) choice in male mice (Fig. 2i, proportion z test, p = 1.09*10−4; female: p = 0.0042). Thus, population activity even at a
p = 0.73), while a significantly greater proportion of neurons respon- single-trial level showed separable subspace representations of social
ded to social choice (n = 16.67%, 95/570) than sucrose choice and nonsocial trials. These findings suggest that mPFC neural activity
(n = 9.30%, 53/570) in female mice (Fig. 2j, proportion z test, associated with choice port entry is distinct for social and nonsocial
p = 2.15*10−4). The neurons that were excited by sucrose choice were trials despite the mice engaging in similar behaviors (nose-poking a
more selective for sucrose choice in female mice (Supplementary choice port). Consequently, neural representations of social and
Fig. 9h: paired t test, p = 3.61*10−6) relative to male mice (Supplemen- nonsocial reward-related behaviors are highly selective and sex-
tary Fig. 9e: paired t test, p = 0.45), while neurons that were excited by dependent even prior to reward consumption.
social choice (Supplementary Fig. 9d, g) as well as those that were Since we found that the largest proportion of mPFC neurons was
inhibited by sucrose choice (Supplementary Fig. 9f, i) showed similar modulated by reward (Supplementary Fig. 9b, c) and that trial type

Nature Communications | (2024)15:8018 6

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 3 | Sex differences in selectivity of social and nonsocial neural repre- entry. h mPFC population activity in female mice more accurately decoded the
sentations of choice. Normalized fluorescence of an example neuron significantly choice made on a particular trial compared to male mice. Unpaired t test
modulated by social choice aligned to social choice (a) compared to that of an (p = 0.0090). i mPFC population activity in a 3 s window around social and sucrose
example neuron significantly modulated by social reward aligned to social choice choice was equivalently sufficient to decode the sex of the animal and decoded with
(b). Normalized fluorescence of an example neuron significantly modulated by significantly higher accuracy than shuffled data (dashed line) on both trial types.
sucrose choice aligned to sucrose choice (c) compared to that of an example Unpaired t test (social versus sucrose: p = 0.19; social versus shuffled: p = 2.08*10−19;
neuron significantly modulated by sucrose reward aligned to sucrose choice (d). sucrose versus shuffled: p = 6.84*10−12). j Across all mice, mPFC neural representa-
Top row shows average fluorescence ± SEM. Bottom row shows a heatmap of tions resulted in greater decoding accuracy for reward compared to choice. Paired t
normalized fluorescence on each trial aligned to choice (dashed line at zero indi- test (p = 7.90*10−8). N = 9 male mice, 9 female mice. Trial-averaged population
cates choice onset) and sorted by reward latency (dark dotted line indicates reward neural activity traces of sucrose (blue) and social (red) trials in male (k, n = 459
onset). mPFC choice neurons are largely non-overlapping in their responsiveness to neurons, 9 mice) and female (m, n = 423 neurons, 5 mice) mice plotted on the first 3
social and sucrose choice in both male (e) and female (f) mice (% non-overlapping, PCs in state space. Arrowhead indicates direction of time. Filled green circle indi-
male: 83.82%, n = 57/68; female: 93.02%, n = 80/86). g mPFC population activity cates choice onset. Euclidean distance separating PC-projected population vectors
accurately decoded the subsequent choice earlier in female (green) mice relative to in a 3 s window around choice is significantly greater between social and sucrose
male (purple) mice indicated by shaded gray region. Unpaired t test (−3.0 to −2.5 s: trials than within each trial type in both male (l, n = 9 mice) and female (n, n = 6
p = 0.0027, −2.5 to −2.0 s: p = 0.0032, −2.0 to −1.5 s: p = 0.019). Choice decoding mice) mice. Paired t test (male: p = 5.76*10−4; female: p = 0.0048). All decoding was
accuracy in female and male mice becomes equivalent at 1.5 s before choice port significantly greater than shuffled data, indicated by a dashed line at 50%. Box plot:
entry and both are greater than chance decoding accuracy (colored asterisks and center line denotes median, box edges indicate the 25th and 75th percentiles and
bolded lines indicate where choice decoding is significantly greater than chance). whiskers extend to ±2.7σ. *p < 0.05.
Shaded error regions indicate ± SEM. Dashed line at zero indicates choice port

decoding was greater for reward than choice in male and female mice or inhibited by social reward (Fig. 4a, male: 3.73%, n = 5/134; female:
(Fig. 3j: average decoding accuracy, choice: 60.0 ± 2.18, reward: 12.38%, n = 25/202; proportion z test, male and female: p < 0.00001). In
86.38 ± 2.78; paired t test, p = 7.90*10−8), we further characterized the contrast, in all mice, more neurons were inhibited by sucrose reward
mPFC reward responses. In both sexes, we found that the majority of (Fig. 4b, male: 64.77%, n = 57/88; female: 60.0%, n = 60/100) than excited
neurons that were responsive to social reward were positively modulated by sucrose reward (Fig. 4b, male: 35.23%, n = 31/88; female: 40.0%,
or excited by social reward (Fig. 4a, male: 96.27%, n = 129/134; female: n = 40/100; proportion z test, male: p = 8.87*10−5; female: p = 0.0047).
87.62%, n = 177/202), compared to those that were negatively modulated Thus, it appears that mPFC neurons differ in their responses to social and

Nature Communications | (2024)15:8018 7

Article https://doi.org/10.1038/s41467-024-52294-6

nonsocial rewards, with a largely excitatory response to social reward and trial type (sucrose/social) as factors, social excite — interaction:
and a mixed inhibitory and excitatory response to sucrose reward. p = 0.067, sex: p = 0.17, trial type: p = 4.11*10−58, with post-hoc unpaired
Next, we examined the selectivity of mPFC reward responses by t tests comparing trial type within sex, male: p = 1.20*10−33, female:
determining the overlap between these populations. In both male p = 4.15*10−26; sucrose inhibit — interaction: p = 0.011, sex: p < 0.00001,
(Fig. 4c) and female (Fig. 4d) mice, largely non-overlapping popula- trial type: p = 0.057, with post-hoc unpaired t tests comparing trial type
tions of mPFC neurons responded to social and sucrose reward, sig- within sex, male: p = 5.48*10−18, female: p = 7.60*10−12). In contrast,
nificantly more non-overlapping than expected by chance when sucrose reward excite neurons showed significantly higher responses
compared to a shuffled distribution (% non-overlapping: male: 73.10%, to sucrose reward relative to social reward in female mice but not male
n = 125/171, shuffled: 61.15 ± 1.45; female: 87.75%, n = 179/204, shuffled: mice (Fig. 4m: two-factor ANOVA with sex (male/female) and trial type
72.20 ± 1.47, ± standard deviation, unpaired t test, male v shuffled: (sucrose/social) as factors, interaction: p = 0.0042, sex: p = 0.014, trial
p = 4.82*10−93; female v shuffled: p = 9.50*10−110, see “Task modulation type: p = 1.87*10−4 with post-hoc unpaired t tests comparing trial type
analysis” in Methods). The social and nonsocial reward ensembles in within sex, male: p = 0.58, female: p = 4.76*10−9). These sex differences
the mPFC were even more distinct in female mice relative to male mice in reward selectivity were further supported by evidence that reward
(Fig. 4e: proportion z test, p = 3.11*10−4). Thus, both male and female decoding accuracy is greater in female mice than in male mice (Fig. 5b:
mice showed largely non-overlapping social and nonsocial reward average decoding accuracy, male: 82.54 ± 3.71, female: 93.29 ± 1.47;
responses in the mPFC. unpaired t test, p = 0.022). Additionally, mPFC activity during the
We then evaluated how social and sucrose reward-responsive reward period could decode animal sex with greater accuracy on social
neurons were modulated by the alternative reward stimulus. We trials compared to sucrose trials (Fig. 5a: average decoding accuracy,
compared the social and sucrose reward responses of neurons that social: 79.45 ± 1.34, sucrose: 68.10 ± 1.74; unpaired t test, p = 1.70*10−6;
were excited by social reward (Fig. 4f, i, l), excited by sucrose reward Supplementary Fig. 7f: average F1 scores, social: 0.77 ± 0.020, sucrose:
(Fig. 4g, j, m) and inhibited by sucrose reward (Fig. 4h, k, n). Both social 0.63 ± 0.026; unpaired t test, p = 4.88*10−5). We used Mahalanobis
reward excite and sucrose reward inhibit neurons showed significantly distance to determine if social and nonsocial reward responses were
higher responses to social reward compared to sucrose reward in male different at a population level (Fig. 5c, d). We found that there was a
and female mice (Fig. 4l, n: two-factor ANOVA with sex (male/female) significantly larger population response to social rewards compared to

Nature Communications | (2024)15:8018 8

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 4 | mPFC neurons differ in social and nonsocial reward response selectivity female (right) mice. i Heatmaps of the average fluorescence of mPFC neurons in
in a sex-dependent manner. a In both male (left) and female (right) mice, a larger male (left, n = 129 neurons) and female (right, n = 129 neurons) mice that are sig-
proportion of mPFC neurons are positively modulated (excited) in response to nificantly excited by social reward aligned to social reward (top row) and sucrose
social reward (male: 28.11%, n = 129/459; female: 31.10%, n = 177/570) compared to reward (bottom row). Neurons are sorted by the time of peak response to social
those that are negatively modulated (inhibited) by social reward (male: 1.09%, n = 5/ reward. j Heatmaps of the average fluorescence of mPFC neurons in male (left,
459; female: 4.39%, n = 25/570). Proportion z test (male: p < 0.00001; female: n = 31 neurons) and female (right, n = 40 neurons) mice that are significantly exci-
p < 0.00001). b In contrast to social reward responses (a), mPFC neurons in male ted by sucrose reward aligned to social reward (top row) and sucrose reward
(left) and female (right) mice are more likely to be inhibited (male: 12.42%, n = 57/ (bottom row). Neurons are sorted by the time of peak response to sucrose reward.
459; female: 10.53%, n = 60/570) rather than excited (male: 6.75%, n = 31/459; k Heatmaps of the average fluorescence of mPFC neurons in male (left, n = 57
female: 7.02%, n = 40/570) in response to sucrose reward. Proportion z test (male: neurons) and female (right, n = 60 neurons) mice that are significantly inhibited by
p = 0.0036; female: p = 0.036). c mPFC reward neurons differ in the selectivity of sucrose reward aligned to social reward (top row) and sucrose reward (bottom
their reward responsiveness in male mice. Social excite neurons are largely exclu- row). Neurons are sorted by the time of minimum response to sucrose reward.
sive in their response to social reward (64.34%, n = 83/129), with a significantly A row on the top and bottom panels of each heatmap (i, j, k) corresponds to the
smaller subset of social excite neurons also responding to sucrose reward (35.66%, same neuron. l Comparison of the peak amplitude of responses of social reward
46/129). Proportion z test (p = 4.08*10−6). In contrast, most sucrose excite neurons excite neurons to social and sucrose reward in male (purple, n = 129 neurons,
also responded to social reward (70.97%, n = 22/31). Around half of sucrose inhibit 9 mice) and female (green, n = 129 neurons, 5 mice) mice shows that these neurons
neurons also responded to social reward (42.11%, n = 24/57). Proportion z test on average have a robust excitatory response to social but not sucrose reward in
(sucrose excite: p = 9.60*10−4, sucrose inhibit: p = 0.092). d mPFC reward neurons both sexes. Two-factor ANOVA with sex and trial type as factors (interaction:
are largely non-overlapping in their reward responsiveness in female mice. Social p = 0.067, sex: p = 0.17, trial type: p = 4.11*10-58) with post-hoc t tests comparing
excite neurons are largely exclusive in their response to social reward (80.62%, trial type within sex (male: p = 1.20*10−33, female: p = 4.15*10−26). m Comparison of
n = 104/129), with a significantly smaller subset of social excite neurons also the peak amplitude of responses of sucrose reward excite neurons to social
responding to sucrose reward (19.38%, n = 25/129). Proportion z test (p < 0.00001). and sucrose reward in male (purple, n = 31 neurons, 9 mice) and female (green,
The majority of sucrose excite neurons (70.0%, n = 28/40) and sucrose inhibit n = 40 neurons, 5 mice) mice shows that sucrose reward excite neurons are selec-
neurons did not respond to social reward (78.3%, n = 47/60). Proportion z test tive for sucrose reward in female but not male mice. Two-factor ANOVA with sex
(sucrose excite: p = 3.45*10−4, sucrose inhibit: p = 5.38*10−10). e Largely non- and trial type as factors (interaction: p = 0.0042, sex: p = 0.014, trial type:
overlapping populations of mPFC neurons respond to social and sucrose reward in p = 1.87*10−4) with post-hoc t tests comparing trial type within sex (male: p = 0.58,
both male (n = 73.10%, 125/171) and female (n = 87.75%, 179/204) mice. The popu- female: p = 4.79*10−9). n Comparison of the peak amplitude of responses of sucrose
lations are more distinct in female mice relative to male mice. Proportion z test reward inhibit neurons to social and sucrose reward in male (purple, n = 57 neurons,
(p = 3.11*10−4). f Average responses of social reward excite neurons to social (red) 9 mice) and female (green, n = 60 neurons, 5 mice) mice shows that these neurons
and sucrose (blue) reward in male (left) and female (right) mice. g Average on average have a higher amplitude response to social reward compared to sucrose
responses of sucrose reward excite neurons to social (red) and sucrose (blue) reward across both sexes. Two-factor ANOVA with sex and trial type as factors
reward in male (left) and female (right) mice. h Average responses of sucrose (interaction: p = 0.011, sex: p < 0.00001, trial type: p = 0.057) with post-hoc t tests
reward inhibit neurons to social (red) and sucrose (blue) reward in male (left) and comparing trial type within sex (male: p = 5.48*10−18, female: p = 7.60*10−12).

Fig. 5 | Distinct population-level reward representations in the mPFC. a mPFC neurons, 5 mice) mice plotted on the first 3 PCs in state space. Arrowhead indicates
population activity in a 3 s window around reward was able to decode the sex of the direction of time. Filled green circle indicates reward onset. PC1 weights are higher
animal with higher accuracy on social than on sucrose trials. Unpaired t test for both social and sucrose reward neurons compared to other neurons in both
(p = 1.70*10−6). N = 9 male mice, 5 female mice. b Decoders trained on female (n = 5 male (e, middle panel) and female (f, middle panel) mice. One-way ANOVA (male:
mice) mPFC neural reward responses decoded the reward type with higher accu- p = 9.22*10−17; female: p = 2.93*10−8) with post-hoc t tests (male — soc v other:
racy than decoders trained on male (n = 9 mice) reward responses. Unpaired t test p = 2.46*10−17, suc v other: p = 3.14*10−6, soc v suc: p = 0.076; female — soc v other:
(p = 0.022). Decoding accuracy was calculated for each animal from all recorded p = 6.21*10−7, suc v other: p = 2.11*10−5, soc v suc: p = 0.96). Euclidean distance
neurons, with a trial-matched number of sucrose and social trials. All decoding was separating PC-projected population vectors in a 3 s window around reward is sig-
significantly greater than shuffled data, indicated by a dashed line at 50%. Maha- nificantly greater between social and sucrose trials than within each trial type in
lanobis distance was greater for social than sucrose reward in male (c, n = 101 social both male (e, right panel) and female (f, right panel) mice. Paired t test (male:
trials, 153 sucrose trials, 9 mice) and female (d, 104 social trials, 83 sucrose trials, 5 p = 5.81*10−4; female: p = 0.0020). *p < 0.05. Shaded error regions indicate ± SEM.
mice) mice. Unpaired t test (male: p = 1.71*10−8; female: p = 2.07*10−5). Trial- Dashed line at zero indicates reward onset. Box plots: center line denotes median,
averaged population neural activity traces of sucrose (blue) and social (red) reward box edges indicate the 25th and 75th percentiles and whiskers extend to ±2.7σ.
trials in male (e, left panel, n = 459 neurons, 9 mice) and female (f, left panel, n = 423

Nature Communications | (2024)15:8018 9

Article https://doi.org/10.1038/s41467-024-52294-6

sucrose rewards in male and female mice (Fig. 5c, d: unpaired t test, p = 0.97, virus: p = 2.00*10−11, trial type: p = 9.23*10−114, with post-hoc
male: p = 1.71*10−8; female: p = 2.07*10−5). Sucrose and social reward unpaired t tests comparing virus within trial types, sucrose: p = 2.71*10−6,
trials also occupy distinct neural subspaces (Fig. 5e, f, left panel: total social: p = 2.75*10−6). This effect on distance traveled was not seen on
variance explained, male: 63.8%; female: 62.8%; Supplementary trials without laser stimulation in male mice or on sucrose trials without
Fig. 7g, h; Supplementary Movie 2) in both sexes as measured by laser stimulation in female mice (Supplementary Fig. 11d, i: two-factor
Euclidean distance between and within trial type population vectors ANOVA with virus (ChR2/GFP) and trial type (sucrose/social) as factors,
(Fig. 5e, f, right panel: male — average distance within: 45.63 ± 4.21, male — interaction: p = 0.15, virus: p = 0.037, trial type: p = 4.71*10−89;
between: 55.14 ± 5.19; paired t test, p = 5.81*10−4; female — average dis- female — interaction: p = 0.084, virus: p = 0.0052, trial type:
tance within: 47.16 ± 3.22, between: 67.17 ± 5.80; paired t test, p = 5.49*10−120, with post-hoc unpaired t tests comparing virus within
p = 0.0020). Across sexes, PC1 weights are higher for neurons classified trial types, sucrose: p = 0.86, social: p = 0.0076). Activating mPFC neu-
as either social or sucrose reward-responsive compared to those that rons outside of the reward period had no effect on the average distance
are reward unresponsive, further supporting the encoding of distinct traveled by mice in the operant arena (Supplementary Fig. 11f, k: two-
social and nonsocial reward information at the population level in the factor ANOVA with virus (ChR2/GFP) and stimulation (on/off) as factors,
mPFC (Fig. 5e, f, middle panel: one-way ANOVA with post hoc t tests, male — interaction: p = 0.55, virus: p = 0.28, stimulation: p = 0.74;
male: p = 9.22*10−17, female: p = 2.93*10−8). Non-overlapping social and female — interaction: p = 0.36, virus: p = 7.46*10−4, stimulation: p = 0.10),
nonsocial reward representations in the mPFC were also seen in mice which suggests that the effects on distance traveled observed during the
that were passively exposed to either a social target or sucrose solution reward period are not solely the result of increased motor behavior due
(Supplementary Fig. 10m-o) and when the nonsocial reward stimulus to mPFC activation. Additionally, we found that mPFC activation caused
was water instead of sucrose (Supplementary Fig. 10a-l). ChR2-expressing mice to spend less time in the social zone during sti-
These data suggest that in male mice, reward-responsive neu- mulated social trials compared to controls (Fig. 6g, k: unpaired t test,
rons fall into three broad categories: one that is exclusively excited male: p = 0.015; female: p = 2.69*10−6). This effect was also seen in
by social reward (Fig. 4f, i, l: left column), a second that is excited by female, but not male mice, in trials without laser stimulation (Supple-
both social and sucrose reward (Fig. 4g, j, m: left column) and a third mentary Fig. 11e, j: unpaired t test, male: p = 0.29; female: p = 8.02*10−4).
that is excited by social reward and inhibited by sucrose reward These findings are consistent with decreased social investigation
(Fig. 4h, k, n: left column). However, in female mice, there is observed with optogenetic activation of mPFC neurons in previous
increased selectivity amongst the categories of responses such that studies25,26.
the non-specific reward excite category seen in male mice is exclu- We next sought to determine if optogenetic inhibition of mPFC
sively excited by sucrose reward in female mice (Fig. 4g, j, m: right activity during the reward period of the two choice operant assay
column). The sex-dependent differences in social and nonsocial could alter reward-seeking behavior. We injected male and female
reward representations in the mPFC are also observed at the popu- mice with an inhibitory opsin (GtACR) or mCherry (control), implanted
lation level. ferrules bilaterally in the mPFC (Fig. 6l, m) and then trained these mice
on the two choice assay (Fig. 6a). Once mice were trained, we used blue
Optogenetic manipulation of mPFC neurons disrupts reward- light to inhibit mPFC neurons during the reward period on a random
seeking behavior 50% of trials over the course of 7 sessions (Supplementary Fig. 11a). We
Since we found that the largest proportion of mPFC neurons were found that inhibition of mPFC neurons during the reward period
reward-responsive during the two choice operant assay, we then wanted increased reward latency on both trial types when compared to con-
to determine if optogenetic activation of mPFC activity during the trols across sexes (Fig. 6n, r: two-factor ANOVA with virus (GtACR/
reward period could alter reward-seeking behavior. We injected male mCherry) and trial type (sucrose/social) as factors, male — interaction:
and female mice with channelrhodopsin (ChR2) or GFP (control), p = 0.67, virus: p = 0.0004, trial type: p < 0.00001, with post-hoc
implanted ferrules bilaterally in the mPFC (Fig. 6b, c) and then trained unpaired t tests comparing virus within trial type, sucrose: p = 0.0014,
these mice on the two choice assay (Fig. 6a). Once mice were trained, we social: p = 0.021; female — interaction: p = 0.83, virus: p = 2.46*10−8,
used blue light to activate mPFC neurons during the reward period on a trial type: p = 2.29*10−12, with post-hoc unpaired t tests comparing virus
random 50% of trials over the course of 7 sessions (Supplementary within trial type, sucrose: p = 5.46*10−10, social: p = 0.0014). Addition-
Fig. 11a). We found that activation of mPFC neurons during the reward ally, optogenetic inhibition caused an increase in the number of
period increased both reward latency and number of reward fails on sucrose reward fails in male mice and in both sucrose and social reward
both trial types when compared to controls across sexes (Fig. 6d, e, h, i: fails in female mice compared to controls (Fig. 6o, s: two-factor ANOVA
two-factor ANOVA with virus (ChR2/GFP) and trial type (sucrose/social) with virus (GtACR/mCherry) and trial type (sucrose/social) as factors,
as factors, male: reward latency — interaction: p = 0.21, virus: male — interaction: p = 0.50, virus: p = 0.0039, trial type: p = 0.37, with
p = 7.52*10−9, trial type: p = 5.62*10−13 with post-hoc unpaired t tests post-hoc unpaired t tests comparing virus within trial type, sucrose:
comparing virus within trial type, sucrose: p = 1.95*10−6, social: p = 0.13, social: p = 0.0084; female — interaction: p = 0.055, virus:
p = 5.51*10−5; reward fails — interaction: p = 6.57*10−6, virus: p = 2.38*10−10, p = 0.0002, trial type: p = 0.0001, with post-hoc unpaired t tests
trial type: p = 3.21*10−8 with post-hoc unpaired t tests comparing virus comparing virus within trial type, sucrose: p = 0.002, social: p = 0.026).
within trial type, sucrose: p = 2.83*10−8, social: p = 0.0042; female: Unlike optogenetic activation, we found that despite disrupting
reward latency — interaction: p = 0.064, virus: p = 4.09*10−13, trial type: reward-seeking behavior, optogenetic inhibition of mPFC neurons did
p = 2.05*10−19, with post-hoc unpaired t tests comparing virus within trial not affect the average distance traveled by male or female mice in the
type, sucrose: p = 7.92*10−10, social: p = 3.81*10−7; reward fails — interac- operant arena during or outside of the reward period across all trials
tion: p = 0.0026, virus: p = 1.14*10−9, trial type: p = 0.0002, with post-hoc with and without laser stimulation (Fig. 6p, t: two-factor ANOVA with
unpaired t tests comparing virus within trial type, sucrose: p = 9.98*10−7, virus (GtACR/mCherry) and trial type (sucrose/social) as factors,
social: p = 1.63*10−4). We also found that optogenetic activation of mPFC male — interaction: p = 0.18, virus: p = 0.16, trial type: p = 3.01*10−106,
neurons increased the average distance traveled by male and female female — interaction: p = 0.75, virus: p = 0.42, trial type: p = 1.97*10−89;
mice in the operant arena during the reward period (Fig. 6f, j: two-factor Supplementary Fig. 11n, s: two-factor ANOVA with virus (GtACR/
ANOVA with virus (ChR2/GFP) and trial type (sucrose/social) as factors, mCherry) and trial type (sucrose/social) as factors, male — interaction:
male — interaction: p = 0.19, virus: p = 1.41*10−18, trial type: p = 5.50*10−98, p = 0.69, virus: p = 0.57, trial type: p = 2.00*10−67, female — interaction:
with post-hoc unpaired t tests comparing virus within trial types, p = 0.67, virus: p = 0.20, trial type: p = 2.94*10−84; Supplementary
sucrose: p = 3.67*10−15, social: p = 3.64*10−8; female — interaction: Fig. 11p, u: two-factor ANOVA with virus (GtACR/mCherry) and

Nature Communications | (2024)15:8018 10

Article https://doi.org/10.1038/s41467-024-52294-6

stimulation (on/off) as factors, male — interaction: p = 0.38, virus: observed in female mice on trials with or without laser stimulation
p = 0.85, stimulation: p = 0.25; female — interaction: p = 0.22, virus: (Fig. 6u: unpaired t test, p = 0.31; Supplementary Fig. 11t: unpaired t
p = 0.45, stimulation: p = 0.22). In male mice, optogenetic inhibition of test, p = 0.12). Overall, these data support a causal role for the mPFC in
mPFC neurons during the reward period resulted in decreased time modulating social and nonsocial reward-seeking behavior and
spent in the social reward zone (Fig. 6q: unpaired t test, p = 0.013), an demonstrate that intact mPFC activity during the reward period is
effect that was also seen in trials without laser stimulation (Supple- crucial for animals to associate choice with reward during the two
mentary Fig. 11o: unpaired t test, p = 0.020). This effect was not choice operant assay.

Nature Communications | (2024)15:8018 11

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 6 | Activation and inhibition of mPFC neurons during the reward period (control). Example histology showing GtACR expression (GtACR in red, DAPI
disrupts reward-seeking behavior. a Experimental timeline showing training and labeling of cell nuclei in blue) and ferrule placement in the mPFC at 4x magnifica-
optogenetic stimulation schedule. b Schematic of the viral strategy (left panel) used tion (right panel). Scale bar: 500 µm. m Reconstruction of optic ferrule placement
to label mPFC neurons with either channelrhodopsin (ChR2) or GFP (control). in male (left, n = 13 mice) and female (right, n = 13 mice) mice using WholeBrain
Example histology showing ChR2 expression (ChR2 in green, DAPI labeling of cell software84. Each colored dot shows the position of the optic ferrule in the Allen
nuclei in blue) and ferrule placement in the mPFC at 4x magnification (right panel). Mouse Brain Common Coordinate Framework. Red dots indicate GtACR mice
Scale bar: 500 µm. c Reconstruction of optic ferrule placement in male (left, n = 14 (male: n = 7 mice; female: n = 7 mice), black dots indicate mCherry mice (male: n = 6
mice) and female (right, n = 14 mice) mice using WholeBrain software84. Each mice; female: n = 6 mice). Coronal slice is 1.945 mm anterior to bregma. Optoge-
colored dot shows the position of the optic ferrule in the Allen Mouse Brain netic inhibition of mPFC neurons during the reward period increases reward
Common Coordinate Framework. Blue dots indicate ChR2 mice (male: n = 8 mice; latency on both social and sucrose trials compared to mCherry controls in male (n)
female: n = 8 mice), black dots indicate GFP mice (male: n = 6 mice; female: n = 6 and female (r) mice. Two-factor ANOVA with virus (GtACR/mCherry) and trial type
mice). Coronal slice is 1.945 mm anterior to bregma. Optogenetic activation of (sucrose/social) as factors (n, interaction: p = 0.67, virus: p = 0.0004, trial type:
mPFC neurons during the reward period increases reward latency (d, h), reward p < 0.00001; r, interaction: p = 0.83, virus: p = 2.46*10−8, trial type: p = 2.29*10−12)
fails (e, i) and average distance traveled (f, j) on sucrose and social trials compared with post-hoc unpaired t tests comparing virus within trial type (n, sucrose:
to GFP controls in both male (d, e, f) and female (h, i, j) mice. Two-factor ANOVA p = 0.0014, social: p = 0.021; r, sucrose: p = 5.46*10−10, social: p = 0.0014). Optoge-
with virus (ChR2/GFP) and trial type (sucrose/social) as factors (d, interaction: netic inhibition also increases the number of social reward fails compared to
p = 0.21, virus: p = 7.52*10−9, trial type: p = 5.62*10−13; e, interaction: p = 6.57*10−6, mCherry controls in male mice (o) and both sucrose and social reward fails in
virus: p = 2.38*10−10, trial type: p = 3.21*10−8; f, interaction: p = 0.12, virus: female mice (s). Two-factor ANOVA with virus (GtACR/mCherry) and trial type
p = 1.41*10−18, trial type: p = 5.50*10−98; h, interaction: p = 0.064, virus: p = 4.09*10−13, (sucrose/social) as factors (o, interaction: p = 0.50, virus: p = 0.0039, trial type:
trial type: p = 2.05*10−19; i, interaction: p = 0.0026, virus: p = 1.14*10−9, trial type: p = 0.37; s, interaction: p = 0.055, virus: p = 0.0002, trial type: p = 0.0001) with post-
p = 0.0002; j, interaction: p = 0.97, virus: p = 2.00*10−11, trial type: p = 9.23*10−114) hoc unpaired t tests comparing virus within trial type (o, sucrose: p = 0.13, social:
with post-hoc unpaired t tests comparing virus within trial type (d, sucrose: p = 0.0084; s, sucrose: p = 0.002, social: p = 0.026). Optogenetic inhibition of
p = 1.95*10−6, social: p = 5.51*10−5; e, sucrose: p = 2.83*10−8, social: p = 0.0042; mPFC neurons had no effect on average distance traveled during social or sucrose
f, sucrose: p = 3.67*10−15, social: p = 3.64*10−8; h, sucrose: p = 7.92*10−10, social: trials in both sexes. Two-factor ANOVA (p, interaction: p = 0.18, virus: p = 0.16, trial
p = 3.81*10−7; i, sucrose: p = 9.98*10−7, social: p = 1.63*10−4; j, sucrose: p = 2.71*10−6, type: p = 3.01*10−106; t, interaction: p = 0.76, virus: p = 0.42, trial type: p = 1.97*10−89).
social: p = 2.75*10−6). Optogenetic activation of mPFC neurons also caused a Optogenetic inhibition of mPFC neurons caused a decrease in the time spent in the
decrease in the time spent in the social zone during the reward period when social zone during the reward period when compared to mCherry controls in male
compared to GFP controls in male (g) and female (k) mice. Unpaired t test (male: (q) but not female (u) mice. Unpaired t test (male: p = 0.013; female: p = 0.31).
p = 0.015; female: p = 2.69*10−6). l Schematic of the viral strategy (left panel) used to *p < 0.05. Box plots: center line denotes median, box edges indicate the 25th and
label mPFC neurons with either GtACR inhibitory opsin (GtACR) or mCherry 75th percentiles and whiskers extend to ± 2.7σ.

Optogenetic manipulation of mPFC neurons seemed to have an activity has persistent effects on social and nonsocial reward-seeking
effect on reward-seeking behavior outside of trials in which there was behavior during the two choice operant assay.
laser stimulation, which resulted in increased reward latency and
number of reward fails compared to controls on non-stimulated Reward-seeking behavior varies with the internal state of mice
social trials in all male mice (Supplementary Fig. 11b, c, l, m: two- We next asked if motivation to seek social versus nonsocial reward
factor ANOVA with virus (ChR2 or GtACR/GFP or mCherry) and trial could be modulated by changing the internal state of the animal. We
type (sucrose/social) as factors, ChR2: reward latency — interaction: changed the level of thirst that animals experienced by restricting their
p = 0.016, virus: p = 8.00*10−5, trial type: p = 4.62*10−16 with post-hoc access to water while also monitoring the activity of mPFC neurons. We
unpaired t tests comparing virus within trial type, sucrose: p = 0.074, first determined if thirsty mice (on restricted water access, RW) would
social: p = 5.31*10−4; reward fails — interaction: p = 0.96, virus: change their behavior on the two choice operant assay (Fig. 7a). We
p = 6.82*10−5, trial type: p = 0.0015 with post-hoc unpaired t tests found that all mice completed significantly more sucrose trials than
comparing virus within trial type, sucrose: p = 0.025, social: social trials when on restricted water access, compared to full water
p = 2.50*10−5; GtACR: reward latency — interaction: p = 0.0004, virus: access (Fig. 7b, f: two-factor ANOVA with water condition (FW/RW) and
p = 0.0001, trial type: p < 0.00001, with post-hoc unpaired t tests trial type (sucrose/social) as factors, male — interaction: p = 1.99*10−55,
comparing virus within trial type, sucrose: p = 0.60, social: water condition: p = 5.14*10−32, trial type: p = 1.22*10−56 with post-hoc
p = 2.14*10−5; reward fails — interaction: p = 0.22, virus: p < 0.00001, unpaired t tests comparing water conditions within trial types, sucrose:
trial type: p = 0.38, with post-hoc unpaired t tests comparing virus p = 1.16*10−38, social: p = 8.97*10−13; female — interaction: p = 1.81*10−23,
within trial type, sucrose: p = 0.027, social: p = 3.04*10−5). Female water condition: p = 2.16*10−11, trial type: p = 8.73*10−15, with post-hoc
mice across optogenetic manipulations showed increased reward unpaired t tests comparing water conditions within trial types, sucrose:
latency and reward fails on both sucrose and social trials without p = 6.51*10−6, social: p = 6.76*10−17). They also performed the task more
laser stimulation compared to controls. (Supplementary Fig. 11g, h, q, quickly with significantly decreased choice latency across all trials
r: two-factor ANOVA with virus (ChR2 or GtACR/GFP or mCherry) and (Fig. 7c, g: two-factor ANOVA with water condition and trial type as
trial type (sucrose/social) as factors, ChR2: reward latency — interac- factors, male — interaction: p = 0.88, water condition: p = 7.37*10−18,
tion: p = 0.11, virus: p = 2.61*10−9, trial type: p = 5.12*10−23, with trial type: p = 0.071 with post-hoc unpaired t tests comparing water
post-hoc unpaired t tests comparing virus within trial type, conditions within trial types, sucrose: p = 1.11*10−15, social: p = 6.99*10−7;
sucrose: p = 1.61*10-8, social: p = 6.04*10−5; reward fails — interaction: female — interaction: p = 0.29, water condition: p = 3.97*10−7, trial type:
p = 0.024, virus: p = 2.23*10−6, trial type: p = 0.83, with post- p = 0.18, with post-hoc unpaired t tests comparing water conditions
hoc unpaired t tests comparing virus within trial type, sucrose: within trial types, sucrose: p = 1.00*10−5, social: p = 0.0054) and
p = 0.044, social: p = 1.07*10−5; GtACR: reward latency — interaction: decreased sucrose reward latency (Fig. 7d, h: two-factor ANOVA with
p = 1.78*10−12, virus: p = 2.51*10−4, trial type: p = 2.24*10−14, with post- water condition and trial type as factors, male — interaction: p = 0.49,
hoc unpaired t tests comparing virus within trial type, sucrose: water condition: p = 3.48*10−9, trial type: p = 1.48*10−19 with post-hoc
p = 1.94*10−4, social: p = 4.25*10−9; reward fails — interaction: p = 0.74, unpaired t tests comparing water conditions within trial types, sucrose:
virus: p < 0.00001, trial type: p = 0.31, with post-hoc unpaired t tests p = 7.61*10−13, social: p = 0.004; female — interaction: p = 0.46, water
comparing virus within trial type, sucrose: p = 0.0076, social: condition: p = 3.10*10−10, trial type: p = 2.89*10−6, with post-hoc
p = 3.12*10−5). These data suggest that transiently altering mPFC unpaired t tests comparing water conditions within trial types,

Nature Communications | (2024)15:8018 12

Article https://doi.org/10.1038/s41467-024-52294-6

sucrose: p = 5.77*10−6, social: p = 0.014) when compared to full water male — interaction: p = 1.24*10−6, water condition: p = 2.47*10−7, trial
access. Male mice also showed decreased social reward latency on type: p = 5.31*10−9 with post-hoc unpaired t tests comparing water
restricted water access compared to full water access (Fig. 7d), while conditions within trial types, sucrose: p = 7.34*10−7, social: p = 0.36;
female mice showed increased social reward latency with water female — interaction: p = 3.88*10−5, water condition: p = 0.0016, trial
restriction (Fig. 7h). All mice on restricted water access made sig- type: p = 0.0014 with post-hoc unpaired t tests comparing water con-
nificantly fewer sucrose reward fails compared to mice on full water ditions within trial types, sucrose: p = 3.01*10−4, social: p = 0.025). This
access and made very few social reward fails overall (Fig. 7e, i: two- may be the result of increased arousal with thirst41,49. Taken together,
factor ANOVA with water condition and trial type as factors, these findings demonstrate that thirst significantly alters the reward-

Nature Communications | (2024)15:8018 13

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 7 | Sucrose reward-seeking behavior and mPFC neural representations of comparing water conditions within trial types (e, sucrose: p = 7.34*10−7, social:
sucrose reward-seeking change with restricted water access in male and p = 0.36; i, sucrose: p = 3.01*10−4, social: p = 0.025). FW: n = 21 male mice, 12 female
female mice. a Experimental timeline showing water restriction and imaging mice; RW: n = 19 male mice, 6 female mice, 3 sessions per mouse. Top row: Heat-
schedule. During restricted water access conditions (RW), both male (b) and female maps of average normalized fluorescence of mPFC neurons in male (j) and female
(f) mice completed significantly more successful sucrose trials and significantly (k) mice that are significantly modulated by sucrose choice (left, male: n = 99
fewer successful social trials than during full water access conditions (FW). Two- neurons; female: n = 159 neurons) and sucrose reward (right, male: n = 225 neurons;
factor ANOVA with water condition (FW/RW) and trial type (sucrose/social) as female: n = 250 neurons) during the two choice operant assay under RW conditions.
factors (b, interaction: p = 1.99*10−55, water condition: p = 5.14*10−32, trial type: Rows are sorted by maximum fluorescence in each heatmap. Bottom row: Average
p = 1.22*10−56; f, interaction: p = 1.81*10−23, water condition: p = 2.16*10−11, trial type: normalized fluorescence traces of the neurons from the corresponding heatmaps
p = 8.73*10−15) with post-hoc unpaired t tests comparing water conditions within in male (j) and female (k) mice that are significantly modulated by sucrose choice
trial types (b, sucrose: p = 1.16*10−38, social: p = 8.97*10−13; f, sucrose: p = 6.76*10−17, (left) and sucrose reward (right). During RW conditions, a greater proportion of
social: p = 6.51*10−6). Male (c) and female (g) mice on RW demonstrated significantly mPFC neurons in both male (l) and female (m) mice are significantly modulated by
decreased choice latency compared to FW. Two-factor ANOVA with water condi- sucrose choice (male: 22.30%, n = 99/444; female: 27.18%, n = 159/585) and sucrose
tion and trial type as factors (c, interaction: p = 0.88, water condition: p = 7.37*10−18, reward (male: 50.68%, n = 225/444; female: 42.74%, n = 250/585) when compared to
trial type: p = 0.071; g, interaction: p = 0.29, water condition: p = 3.97*10−7, trial type: full water access conditions. Proportion z test with correction for multiple com-
p = 0.18) with post-hoc unpaired t tests comparing water conditions within trial parisons (male — sucrose choice: p = 9.53*10−8, sucrose reward: p < 0.00001;
types (c, sucrose: p = 1.11*10−15, social: p = 6.99*10−7; g, sucrose: p = 1.00*10−5, social: female — sucrose choice: p = 4.22*10-15, sucrose reward: p < 0.00001). Trial-
p = 0.0054). Male (d) and female (h) mice on RW significantly decreased their averaged population neural activity traces of sucrose and social reward trials in
sucrose reward latency compared to FW. Male mice also decreased their social male (n, left panel) and female (o, left panel) mice during FW (sucrose trials: dark
reward latency on RW, while it increased in female mice. Two-factor ANOVA with blue, social trials: red) and RW (sucrose trials: light blue, social trials: pink) plotted
water condition and trial type as factors (d, interaction: p = 0.49, water condition: on the first 3 PCs in state space. Arrowhead indicates direction of time. Filled green
p = 3.48*10−9, trial type: p = 1.48*10−19; h, interaction: p = 0.46, water condition: circle indicates reward onset. Euclidean distance separating PC-projected popula-
p = 3.10*10−10, trial type: p = 2.89*10−6) with post-hoc unpaired t tests comparing tion vectors in a 3 s window around reward is significantly greater between water
water conditions within trial types (d, sucrose: p = 7.61*10−13, social: p = 0.004; restriction conditions for social than for sucrose trials in both male (n, right panel)
h, sucrose: p = 5.77*10−6, social: p = 0.014). Male (e) and female (i) mice on RW made and female (o, right panel) mice. Paired t test (male: p = 7.93*10−5; female:
significantly fewer sucrose reward fails compared to FW, while female mice on RW p = 0.0030). N = 8 male mice, 5 female mice. *p < 0.05 Shaded error regions indi-
also made significantly more social reward fails compared to FW. Two-factor cate ± SEM. Dashed line at zero indicates reward onset. Box plots: center line
ANOVA with water condition and trial type as factors (e, interaction: p = 1.24*10−6, denotes median, box edges indicate the 25th and 75th percentiles and whiskers
water condition: p = 2.47*10−7, trial type: p = 5.31*10−9; i, interaction: p = 3.88*10−5, extend to ±2.7σ.
water condition: p = 0.0016, trial type: p = 0.0014) with post-hoc unpaired t tests

seeking behavior of mice on the two choice operant assay with mice was significantly greater for social trials than for sucrose trials
preferentially seeking sucrose over social reward. (Fig. 7n, o, right panel: male — average distance between FW and RW
for sucrose trials: 21.03 ± 2.85, social trials: 40.26 ± 2.96, paired t test,
mPFC sucrose reward responses change with thirst through the p = 7.93*10−5; female — average distance between FW and RW for
recruitment of previously latent neurons sucrose trials: 26.33 ± 2.97, social trials: 44.06 ± 2.85, paired t test, p =
To determine if changes in behavior corresponded with changes in 0.0030). Despite the changes seen in neural responsiveness, mPFC
mPFC neural representations on the two choice operant assay, we activity during the reward period decoded trial type with comparably
performed cellular resolution calcium imaging of mPFC neurons when high accuracy across water access conditions (Supplementary
mice were on restricted water access (RW, male: n = 444 neurons, 8 Fig. 12m, q: average decoding accuracy, male — FW: 77.67 ± 4.55, RW:
mice; female: n = 585 neurons, 6 mice) and compared that to full water 69.84 ± 4.06; female — FW: 90.58 ± 1.26, RW: 90.11 ± 3.10; unpaired t
access (FW) conditions. test, male: p = 0.21; female: p = 0.91).
We compared the neural representations of operant assay These findings raise the question of how the mPFC is able to
events (Fig. 7j–m; Supplementary Fig. 12j–l,n–p), specifically those maintain flexible neural representations of social and nonsocial
related to sucrose reward-seeking behavior, across water access rewards that are dependent on the internal state of the animal. One
conditions. We found that across both sexes, a greater proportion of possibility is that individual mPFC neurons may change their identity
mPFC neurons were modulated by sucrose choice (Fig. 7l, m: to encode social or nonsocial reward information depending on the
male — RW: 22.30%, 99/444, FW: 9.37%, 43/459; female — RW: 27.18%, internal state of the animal. Alternatively, previously latent popula-
159/585, FW: 9.30%, 53/570; proportion z test, male: p = 9.53*10−8, tions of mPFC neurons may be activated depending on the internal
female: p = 4.22*10−15) and sucrose reward (Fig. 7l, m: male — RW: state of the animal. To distinguish between these possibilities, we
50.68%, 225/444, FW: 19.17%, 88/459; female — RW: 42.74%, 250/585, utilized the high spatial resolution of calcium imaging to track indivi-
FW: 17.54%, 100/570; proportion z test, male and female: dual mPFC neurons across water restriction conditions (Fig. 8b)50. We
p < 0.00001) in restricted water access conditions compared to full compared the activity of neurons across three conditions, two condi-
water access conditions. The patterns of selectivity for sucrose and tions in which mice had full water access and one condition of
social reward seen in male and female mice during full water access restricted water access (Fig. 8a). Across male mice, we tracked 178
were maintained during restricted water access (Supplementary mPFC neurons between the first full water access imaging session and
Fig. 12a–i). We also observed a decrease in the number of social the restricted water access imaging session (Fig. 8c, n = 8 mice), 158
reward-responsive neurons in male and female mice following water mPFC neurons between the first full water access imaging session and
deprivation (Supplementary Fig. 12l, p: male — RW: 4.28%, 19/444, the second full water access imaging session (Fig. 8d, n = 7 mice) and
FW: 29.19%, 134/459; female — RW: 25.60%, 150/585, FW: 35.44%, 202/ 164 neurons between the restricted water access imaging session and
570; proportion z test, male: p < 0.00001, female: p = 2.99*10−4). At a the second full water access imaging session (Fig. 8e, n = 7 mice). In
population level, sucrose and social reward trials in both FW and RW female mice, we tracked 226 mPFC neurons between the first full water
conditions continued to occupy distinct trajectories in the first 3 PCs access imaging session and the restricted water access imaging session
of state space (Fig. 7n, o, left panel: total variance explained, male: (Fig. 8f, n = 5 mice), 256 mPFC neurons between the first full water
55.9%; female: 52.7%) in both sexes. However, the Euclidean distance access imaging session and the second full water access imaging ses-
separating population vectors between water restriction conditions sion (Fig. 8g, n = 5 mice) and 253 neurons between the restricted water

Nature Communications | (2024)15:8018 14

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 8 | mPFC neural representations of sucrose reward change with water and sucrose reward inhibit (j) responses across both full water access (left panel)
restriction through the recruitment of neurons that were previously reward- and restricted water access imaging sessions (right panel). Pi charts depicting the
unresponsive. a Experimental timeline showing water restriction and imaging previous identity of sucrose reward excite (k) and sucrose reward inhibit neurons
schedule. b mPFC neurons were tracked across three different imaging sessions (l) in the restricted water access condition in male (left panel) and female (right
using CellReg, a cell registration software50. The top row depicts the CNMFe- panel) mice. Proportion of tracked reward-unresponsive neurons that convert to
generated field of view (FOV) of all identified neurons from each imaging session of sucrose reward excite (m) or sucrose reward inhibit (n) neurons in the subsequent
a representative animal across various water access conditions. The bottom row imaging session in male (left panel) and female (right panel) mice. Proportion z test
depicts individual neurons that were successfully tracked across all three imaging with correction for multiple comparisons (m, male — FWvFW2 (1) compared to
sessions from the same animal. Individual neurons are represented by different FWvRW (2): p = 0.011, FWvFW2 (1) compared to FW2vRW (3): p = 0.003, FWvRW (2)
colors. Venn diagrams showing the number of tracked neurons across various compared to FW2vRW (3): p = 0.64; female — 1 to 2: p = 0.036, 1 to 3: p = 0.27, 2 to 3:
water restriction conditions in male (c–e) and female (f–h) mice. Time locked p = 0.26; n, male — 1 to 2: p = 2.25*10−6, 1 to 3: p = 1.65*10−6, 2 to 3: p = 0.98; female — 1
responses (top row: average fluorescence trace, bottom row: heatmap showing to 2: p = 0.0054, 1 to 3: p = 6.37*10−8, 2 to 3: p = 0.009). *p < 0.05. Shaded error
individual trials) of example neurons that showed stable sucrose reward excite (i) regions indicate ± SEM.

access imaging session and the second full water access imaging ses- sucrose inhibit or reward-unresponsive. Neurons were considered
sion (Fig. 8h, n = 5 mice). stable if they maintained their response (excite or inhibit) to sucrose
We then characterized the identity of neurons that we tracked reward across imaging sessions (see example neurons Fig. 8i, j).
based on their reward responses in both imaging sessions. In parti- Excluding stable neurons, we found that a significantly greater pro-
cular, we identified neurons as either social excite, sucrose excite, portion of mPFC neurons that were sucrose reward-responsive during

Nature Communications | (2024)15:8018 15

Article https://doi.org/10.1038/s41467-024-52294-6

the restricted water access imaging session were previously unre- test, p = 7.58*10−11). This sex-dependent change in the peak amplitude
sponsive as opposed to converting from social excite neurons during of the response of social reward excite neurons following social iso-
the first full water access imaging session in both male (n = 71.21%, 47/ lation was specific to social reward neurons. We compared the peak
66 neurons, 8 mice) and female (n = 70.45%, 31/44 neurons, 5 mice) amplitude of sucrose reward excite (Supplementary Fig. 14a, c:
mice (Fig. 8k, l: proportion z test, male: p = 1.09*10−6; female: unpaired t test, male: p = 0.17; female: p = 0.66), and sucrose reward
p = 1.20*10−4). As a result, we next quantified the proportion of unre- inhibit neurons (Supplementary Fig. 14b, d: unpaired t test, male:
sponsive neurons that converted to either sucrose reward excite p = 0.077; female: p = 0.30) before and after social isolation and found
(Fig. 8m) or sucrose reward inhibit (Fig. 8n) neurons across imaging that there was no change in peak amplitude. To rule out the effects of
sessions. We found that a significantly larger proportion of unre- change with time on neural responses, we compared the proportion of
sponsive neurons converted to either sucrose reward excite (Fig. 8m: social reward neurons and the peak amplitude of their social reward
proportion z test with correction for multiple comparisons, responses across the partial water access (Supplementary Fig. 13) and
male — FWvFW2 (1) compared to FWvRW (2): p = 0.011, FWvFW2 (1) the full water access imaging sessions before social isolation. We also
compared to FW2vRW (3): p = 0.003, FWvRW (2) compared to observed no difference in the proportion of social reward excite
FW2vRW (3): p = 0.64; female — 1 to 2: p = 0.036, 1 to 3: p = 0.27, 2 to 3: neurons between partial water access and full water access pre-
p = 0.26) or sucrose reward inhibit (Fig. 8n: proportion z test with isolation sessions (Supplementary Fig. 14e, g: male — pre: 20.59%,
correction for multiple comparisons, male — 1 to 2: p = 2.25*10−6, 1 to 3: n = 84/408, PW: 21.28%, n = 90/423; female — pre: 35.44%, n = 202/570,
p = 1.65*10−6, 2 to 3: p = 0.98; female — 1 to 2: p = 0.0054, 1 to 3: PW: 32.37%, n = 180/556, proportion z test, male: p = 0.81, female:
p = 6.37*10−8, 2 to 3: p = 0.009) neurons across dissimilar water access p = 0.28). However, unlike with social isolation, there was no change in
conditions (full water access/full water access 2 and restricted water the amplitude of the responses of social reward excite neurons (Sup-
access) compared to similar water access conditions (full water access plementary Fig. 14f, h: unpaired t test, male: p = 0.74, female: p = 0.21).
and full water access 2). These findings indicate that the increased These data demonstrate that the sex-dependent changes in peak
proportion of sucrose reward-responsive neurons in the mPFC seen amplitude seen with social isolation are specific to social reward excite
with water restriction were driven largely by the recruitment of pre- neurons as a result of social isolation.
viously reward-unresponsive neurons and less by social reward neu-
rons changing identity or by turnover due to the passage of time. Discussion
Using an automated two choice operant assay, we found that largely
Neural representations of social reward change with social iso- non-overlapping populations of neurons in the mPFC respond to social
lation in a sex-dependent manner and sucrose reward and that these representations are more distinct in
The two choice operant assay allows for the direct comparison of female mice compared to male mice. We additionally showed that
social and nonsocial reward-seeking behavior, which enables us to manipulating mPFC activity through both optogenetic activation and
systematically change access to either type of reward outside of the inhibition during the reward period disrupts the ability of male and
assay and characterize how this affects motivation to seek either female mice to successfully engage with both types of reward,
reward. Since water restriction preferentially drove reward-seeking demonstrating that the mPFC plays a causal role in modulating social
behavior and mPFC neural representations toward sucrose reward, we and nonsocial reward-seeking behavior. Finally, we found that chan-
then wanted to determine if social isolation would similarly drive ging internal state through either water deprivation or social isolation
preference for social reward. Consequently, we socially isolated mice differentially modulates sucrose and social reward representations in a
for 7 days after they had been trained on the two choice operant assay sexually dimorphic manner. Together, these findings provide new
and imaged mPFC neurons before and after social isolation in male and insights into how social and nonsocial reward information is organized
female mice (Fig. 9a). We found that social isolation had opposing in the mPFC.
effects on male and female reward-seeking behavior. Specifically, we
found that social isolation caused an increase in the proportion of Advantages offered by the two choice (social-sucrose)
social trials completed by male mice relative to sucrose trials prior to operant assay
social isolation (Fig. 9b, two-factor ANOVA with isolation (pre/post) The use of operant assays in rodent models has significantly con-
and trial type (sucrose/social) as factors; interaction: p = 0.020). In tributed to our understanding of the neural circuitry underlying the
contrast, social isolation caused a relative decrease in the number of seeking and processing of nonsocial reward-related behaviors, such as
social trials completed by female mice compared to sucrose trials prior food rewards51–53. While we know that mice find social interactions
to social isolation (Fig. 9i, two-factor ANOVA with isolation (pre/post) rewarding, the complexity of social interactions has made under-
and trial type (sucrose/social) as factors; interaction: p = 0.0035). We standing how the brain represents social interactions challenging1.
did not observe effects of isolation on the other behavioral parameters Most studies determine whether a conspecific is perceived as
in both sexes (Fig. 9c–e,j–l). rewarding by comparing the time mice spend in the proximity of a
We next determined the effects of social isolation on mPFC neu- novel mouse relative to an object or an empty cage3,4. Although these
ronal activity. Across all mice, we found that the proportion of social assays are straightforward, ascertaining social reward-seeking beha-
reward neurons in the mPFC did not change with social isolation in vior from free ranging interactions is challenging. These assays also
male (pre: n = 20.59%, 84/408, post: n = 20.90%, 98/469, 4 mice) or rarely compare social reward to other strongly appetitive stimuli, such
female (pre: n = 35.44%, 202/570, post: n = 35.95%, 206/573, 6 mice) as sucrose. As a result, similarities and differences in how social and
mice (Fig. 9f, m: proportion z test, male: p = 0.91; female p = 0.86). nonsocial reward-seeking behaviors are encoded in the brain remain
However, when we compared the peak amplitude of social reward poorly understood. To overcome these challenges, we developed a
excite neuron responses, we found sex-dependent changes in peak low-cost, fully automated two choice operant assay in which mice can
amplitude following social isolation. Specifically, in male mice, the freely choose between social and sucrose rewards. This assay allows
peak amplitude of social reward excite neurons significantly decreased for unsupervised high throughput data collection, permitting us to
after social isolation (Fig. 9g, h: average peak amplitude, pre: collect hundreds of social and sucrose trials per day across several
0.27 ± 0.0094, post: 0.22 ± 0.0083; unpaired t test, p = 4.24*10−4), animals.
while in female mice the peak amplitude of social reward excite neu- Additionally, the operant nature of the assay allows us to con-
rons significantly increased after social isolation (Fig. 9n, o: average strain social behavior and monitor social reward-seeking with high
peak amplitude, pre: 0.23 ± 0.0069, post: 0.30 ± 0.0080; unpaired t temporal precision. Using this assay, we found that in control

Nature Communications | (2024)15:8018 16

Article https://doi.org/10.1038/s41467-024-52294-6

conditions (not social isolation or water deprivation), male mice valence, and reward duration independently and parametrically in
showed an equal preference for social and nonsocial rewards and freely behaving mice. Finally, although we did not observe strain dif-
female mice slightly preferred social rewards over sucrose rewards ferences in social and nonsocial reward-seeking behavior in female
(Fig. 1d, h; Supplementary Fig. 3i). Our results confirmed recent find- mice on our assay (Supplementary Fig. 6), others have recently found
ings that adult mice find social stimuli positively reinforcing and will strain differences in social reward-seeking behavior, with female CD1
work to gain access to a social stimulus in operant assays8–10,12. Building mice showing a stronger preference for the social reward compared to
off this work, we were able to directly compare the responses of mPFC C57BL/6J (used in this study)9. This raises interesting questions about
neurons to social and nonsocial rewards while mice demonstrated how neural circuit differences across strains and species might give
similar motivation to seek both rewarding stimuli. This approach can rise to varied behavioral outcomes.
be used to systematically map how social and nonsocial reward
representations are transformed across the reward circuitry, by ima- Social and nonsocial reward representations in the medial
ging or recording in regions such as the basolateral amygdala, ventral prefrontal cortex
tegmental area, and the nucleus accumbens during the two choice There is little consensus regarding how social and nonsocial reward-
operant assay. In the future, this assay can also be used to vary related behaviors are mediated within the same brain region. One
dimensions of social reward, such as proximity of the social reward, theory proposes that social and nonsocial behaviors are represented

Nature Communications | (2024)15:8018 17

Article https://doi.org/10.1038/s41467-024-52294-6

Fig. 9 | Social reward-seeking behavior and neural representations of social decrease in the number of successful social trials compared to sucrose trials
reward change in a sex-dependent manner following social isolation. completed after social isolation. Two-factor ANOVA with isolation and trial type as
a Experimental timeline showing social isolation and imaging schedule. b Male factors (interaction: p = 0.0035, isolation: p = 0.29, trial type: p = 0.30) with post-
mice showed a relative increase in the number of successful social trials compared hoc unpaired t tests comparing isolation within trial types (sucrose: p = 0.017,
to sucrose trials completed after social isolation. Two-factor ANOVA with isolation social: p = 0.10). Female mice did not show a difference in choice latency (j), reward
(pre/post) and trial type (sucrose/social) as factors (interaction: p = 0.020, isola- latency (k) and reward fails (l) of either trial type before and after social isolation.
tion: p = 0.97, trial type: p = 0.52) with post-hoc unpaired t tests comparing isola- Two-factor ANOVA with isolation and trial type as factors (j, interaction: p = 0.61,
tion within trial types (sucrose: p = 0.21, social: p = 0.0079). Male mice did not show isolation: p = 0.29, trial type: p = 0.41; k, interaction: p = 0.24, isolation: p = 0.13, trial
any significant difference in choice latency (c), reward latency (d) and reward fails type: p = 1.00*10−4; l, interaction: p = 0.36, isolation: p = 0.29, trial type: p = 0.0005)
(e) of either trial type before and after social isolation. Two-factor ANOVA with with post-hoc unpaired t tests comparing isolation within trial types (k, sucrose:
isolation (pre/post) and trial type (sucrose/social) as factors (c, interaction: p = 0.069, social: p = 0.81; l, sucrose: p = 0.42, social: p = 0.33). N = 6 female mice.
p = 0.69, isolation: p = 0.56, trial type: p = 0.40; d, interaction: p = 0.81, isolation: m Female mice did not show a difference in proportion of mPFC neurons sig-
p = 0.21, trial type: p = 1.08*10−6; e, interaction: p = 0.91, isolation: p = 0.81, trial type: nificantly modulated by social reward before and after social isolation. Proportion z
p = 0.0075) with post-hoc unpaired t tests comparing isolation within trial types test (p = 0.86). n Average normalized fluorescence traces of mPFC neurons sig-
(d, sucrose: p = 0.36, social: p = 0.34; e, sucrose: p = 0.95, social: p = 0.39). N = 4 nificantly excited by social reward before (red) and after (gray) social isolation in
male mice. f Male mice did not show a difference in the proportion of mPFC female mice time-locked to social reward (left panel). o Comparison of peak
neurons significantly modulated by social reward before social isolation. Propor- fluorescence before (n = 177 neurons) and after (n = 193 neurons) social isolation
tion z test (p = 0.91). g Average normalized fluorescence traces of mPFC neurons shows that neural responses to social reward increased with social isolation in
significantly excited by social reward before (red) and after (gray) social isolation in female mice (right panel). Unpaired t test (p = 7.58*10−11). *p < 0.05. Shaded error
male mice time-locked to social reward shown (left panel). h Comparison of peak regions indicate ± SEM. Dashed line at zero indicates reward onset. Error bars
fluorescence before (n = 76 neurons) and after (n = 89 neurons) social isolation indicate ± SEM. Box plots: center line denotes median, box edges indicate the 25th
shows that neural responses to social reward decreased with social isolation in male and 75th percentiles and whiskers extend to ±2.7σ.
mice (right panel). Unpaired t test (p = 4.24*10−4). i Female mice showed a relative

by discrete and mutually inhibitory neural populations54. Support for selective activity of social and nonsocial reward neurons shape reward-
this theory comes from a recent study that found that neurons in the seeking and decision-making behavior.
orbitofrontal cortex that are responsive to social interactions suppress
feeding behavior when activated16. Another study also found that dif- Effects of internal state and sex on neural representations
ferent populations within the medial amygdala represented distinct of reward
and antagonistic social and asocial behaviors55. Consistent with the There has been increasing evidence that male and female animals
idea that distinct and antagonistic populations of neurons mediate differ in their reward-seeking behavior, including differences in
social and nonsocial reward, we observed largely non-overlapping reward-associated learning strategies43,57 and sensitivity to reward
populations of neurons within the mPFC that encoded social and outcomes22,58. Studies in rodents have also shown that there are sex
nonsocial reward-related information (Figs. 3, 4 and 5). We also differences in motivation to seek social reward, with female rodents
observed non-overlapping social and nonsocial neural representations demonstrating more robust social reward-seeking behavior than male
on a passive exposure paradigm (Supplementary Fig. 10m–o). This rodents59,60. These findings raise the possibility that there might be
mPFC reward-related activity was sufficient to decode the trial type differences in how social and nonsocial rewards are represented in
during both the choice and reward periods (Fig. 3g, h and Fig. 5b), as male and female mice. In this study, we combined a two choice operant
well as, the sex of the animal (Fig. 3i and Fig. 5a). Additionally, we found assay with cellular resolution calcium imaging to identify the neural
evidence that these populations may regulate social and nonsocial substrates that might underlie some of these sex-dependent beha-
reward information in an antagonistic manner. In particular, using vioral differences.
peak amplitude analysis, we identified a subset of mPFC neurons that Although male and female mice demonstrated similar levels of
were inhibited in response to sucrose reward and excited in response motivation to seek social and sucrose rewards under control condi-
to social reward (Fig. 4h, k, n). In contrast, there were very few mPFC tions during the two choice operant assay, we observed sex differences
neurons that were inhibited by social reward (Fig. 4a). These differ- in how social and nonsocial rewards were represented in the mPFC.
ences could arise as a result of non-mutual inhibition between social Specifically, female mice had more distinct representations of social
and sucrose ensembles driven by local inhibitory interneurons in the and nonsocial reward in the mPFC (Fig. 4d, e). As a result, we were able
mPFC. Alternatively, they could result from differences in external to decode reward type with higher accuracy from the mPFC neural
inputs to the social and nonsocial reward-responsive neurons. Future activity of female mice relative to that of male mice (Fig. 5b). We were
cell-type specific imaging and optogenetic experiments are essential also able to decode the sex of the animal from the mPFC neural activity
to begin to distinguish between these possibilities. around both choice and reward (Fig. 3i and Fig. 5a). Interestingly, we
Furthermore, we showed that mPFC activity during the reward were able to decode subsequent choice at earlier time points in female
period plays a causal role in mediating reward-seeking behavior on the mice than in male mice (Fig. 3g). Additionally, sucrose reward-
two choice operant assay. We found that altering mPFC activity responsive neurons in female mice were significantly less responsive
through either optogenetic excitation or inhibition during the reward to social reward, whereas sucrose reward-responsive neurons in male
period increased reward latency and the number of reward fails across mice were equally responsive to social reward (Fig. 4m). These findings
trial type in male and female mice (activation: Fig. 6d, e, h, i and inhi- raise the intriguing possibility that sex differences in mPFC activity
bition: Fig. 6n, o, r, s). Although the role of the mPFC in nonsocial could drive divergent reward-seeking behaviors when animals choose
reward-seeking has been well established, these experiments are the between competing social and nonsocial rewards.
first to suggest a causal role for the mPFC in shaping social reward- We also found that social isolation differentially affected reward-
seeking behavior and not solely disrupting ongoing social seeking behavior in male and female mice. We observed a relative
interactions25,26. In the future, combining activity-dependent neuron increase in the number of social trials completed by male mice and a
tagging methods (e.g., TRAP-Cre56) with optogenetic tools could allow relative decrease in the number of social trials completed by female
for the selective activation and silencing of social or nonsocial reward mice compared to sucrose trials following acute social isolation
neurons. These experiments could help provide insight into how (Fig. 9b, i). These differences are consistent with previous findings that

Nature Communications | (2024)15:8018 18

Article https://doi.org/10.1038/s41467-024-52294-6

social isolation increases social approach behaviors in male mice while These findings suggest that distinct neuromodulatory and neuropep-
causing social withdrawal in female mice61. In addition, we observed tide systems may influence the changes seen in social and nonsocial
that there were sex-dependent changes in how mPFC neurons reward representations following water deprivation and social isola-
responded to social reward but not sucrose reward following social tion. Future studies are needed to determine how different neuro-
isolation. In male mice, we found that social reward-responsive mPFC modulators and hormones act on neurons in the mPFC to modulate
neurons showed a decrease in peak amplitude in response to social social and nonsocial reward representations with varying internal state
reward following social isolation (Fig. 9g, h). In contrast, in female in a sex-dependent manner.
mice, we observed an increase in the peak amplitude of social reward-
responsive neurons in response to social reward following social iso- Broader implications
lation (Fig. 9n, o). These changes in amplitude were socially specific, The ability to perceive social interactions as rewarding is necessary for
suggesting that the neural changes resulting from acute social isola- appropriate social behavior. As such, a prevailing theory for the social
tion (~1 week) act specifically on mPFC ensembles that are sensitive to dysfunction seen in ASD, known as the social motivation theory, sug-
social reward. These findings are consistent with a recent neuroima- gests that people with ASD find social interactions less rewarding76.
ging study in humans which found that acute social and food depri- However, it remains unresolved how social reward processing is dis-
vation had distinct effects on social versus nonsocial craving responses rupted in ASD. Human studies have shown that people with ASD have
in cortical regions62. Thus, acute social isolation may not affect non- impairments in processing both social and nonsocial rewards77,78, a
social reward representations in the mPFC. However, it remains finding that has also been corroborated in rodent models of ASD79–81. In
unclear if the changes observed in the neural representations of social particular, one study found decreased differentiation between social
reward in the mPFC causally contribute to the behavioral differences and nonsocial odor cue encoding in the mPFC of mice with a genetic
observed in male and female mice following social isolation. Further- model of ASD28. Another study showed that social deficits in the same
more, the cause of these sex-dependent changes in mPFC mouse model of ASD could be rescued by either increasing excitability
activity following social isolation remain unknown. Hormones63, of parvalbumin (PV) interneurons or decreasing excitability of pyr-
neuromodulators64 and neuropeptides65, all of which have been shown amidal neurons in the mPFC82. We also found that optogenetically
to change in a sex-dependent manner following social isolation, may manipulating mPFC activity during the reward period disrupted
contribute to the observed changes in neural representations of social reward-seeking behavior on the two choice operant assay (Fig. 6).
reward. Thus, further interrogation of the neural mechanisms underlying
Although we found that acute social isolation did not alter mPFC social and nonsocial reward-related behavior, specifically within the
responses to sucrose reward, there is substantial evidence that chronic mPFC, is imperative to further our understanding of diseases such as
social isolation affects nonsocial reward-seeking behaviors in ASD. The two choice operant assay described in this study offers a
rodents66–68 and humans69. Additionally, there is evidence that chronic means by which to study two competing reward-related behaviors and
social isolation affects mPFC activity61,70,71. Consequently, 7 days of to provide further insight into how the brain processes social and
social isolation may not be sufficient to produce the neural and nonsocial reward.
behavioral changes seen with more chronic forms of social isolation. In In summary, we developed a two choice (social-sucrose) operant
fact, the social homeostasis theory of social isolation postulates that assay that allowed us to characterize how social and nonsocial rewards
social isolation activates homeostatic mechanisms that drive prosocial are represented in the mPFC. Using this assay, we demonstrated that
or antisocial behavior depending on the duration of isolation72. Thus, social and nonsocial reward representations are non-overlapping and
further work is needed to systematically vary the duration of social differentially modulated by social isolation and thirst in a sex-
isolation to determine the precise timing of prosocial versus antisocial dependent manner. Finally, we showed that these representations
behaviors and the neural mechanisms underlying these changes. The are essential for appropriate reward-seeking behavior.
two choice operant assay, as presented in this study, is uniquely suited
to determine the impacts of social isolation on different reward-related Methods
behaviors, since it allows for the quantification and direct comparison Experimental subjects
of social and nonsocial reward-seeking behaviors following social For all experiments, C57BL/6J male and female mice aged 6–10 weeks
isolation. from Jackson laboratories were used. Mice were maintained on a
Unlike with social isolation, we did not find sex differences in reverse light dark cycle (12-h light off-12-h light on schedule). All
changes of reward-related behavior or mPFC reward representations experimental procedures were approved by the Emory Institutional
with water deprivation. Instead, all mice robustly changed their Animal Care and Use Committee.
behavior in favor of sucrose reward-seeking. The neural representa-
tions of reward reflected these changes in behavior, with a larger Viruses
proportion of mPFC neurons responding to sucrose choice and The viruses used in this study were purchased from Addgene: AAV5-
sucrose reward following water restriction (Fig. 7j–m). By long- Syn-GCaMP6f-WPRE-SV40 (100835-AAV5), AAV5-hSyn-hChR2(H134R)-
itudinally tracking neurons across imaging sessions, we found that this EYFP (26973-AAV5), AAV5-CAG-GFP (37825-AAV5), AAV1-CKIIa-
increase in the number of sucrose reward-responsive neurons was stGtACR2-FusionRed (105669-AAV1), and AAV5-CKIIa-mCherry
driven by the recruitment of previously reward-unresponsive neurons (114469-AAV5).
(Fig. 8k–n). These findings show that the reward selectivity of indivi-
dual mPFC neurons rarely changes with internal state (i.e., a “social Stereotaxic surgeries
neuron” rarely converts to a “sucrose neuron” in a thirsty animal), For imaging surgeries, mice aged ~6–8 weeks (n = 19 male mice, 6
demonstrating how thirst shapes reward-related neural activity in the female mice) were anesthetized with 1–2% isoflurane and placed in
mPFC. Similarly, thirst has been shown to cause widespread changes in stereotactic setup (Kopf). A microsyringe (Nanoject) was used to inject
brain activity41,73 and level of thirst is known to affect task performance 0.75 µL of AAV5-Syn-GCaMP6f-WPRE-SV40 (Addgene, injected titer of
on goal-directed behaviors49. A recent study also found that ventral 1.3 × 1013 parts/ml) unilaterally into the mPFC (1.6 mm anterior, 0.5 mm
tegmental area dopamine neurons track satiety and cause changes in lateral and 2.4 mm in depth). Mice were then implanted with a 0.6 mm
dopamine dynamics in distributed brain networks74. In contrast, a diameter, 7.3 mm length GRIN lens (Product ID:1050-004597, Insco-
separate dopaminergic system in the dorsal raphe nucleus has been pix) over the mPFC (1.8 mm anterior, 0.5 mm lateral and 2.1 mm in
shown to mediate behavioral changes following social isolation75. depth). A metal cap was cemented over the lens to protect it. Mice

Nature Communications | (2024)15:8018 19

Article https://doi.org/10.1038/s41467-024-52294-6

were group housed following lens implant. 3–4 weeks following viral or social reward consumption, the trial ends and after a 30 s intertrial
injection and lens implant, a baseplate (Product ID: 1050-004638, interval (ITI), a new trial begins with the illumination of both choice
Inscopix) attached to the miniature microscope (nVista, Inscopix) was ports. Each behavioral session is 1 h in duration. Our social operant
positioned ~0.45 mm above the lens such that blood vessels and neu- assay consists of several phases: training (4 stages, ~2 weeks) and post-
rons were in focus. The baseplate was then cemented in place using training (up to 3 weeks).
dental cement (Metabond) and a baseplate cover (Product ID: 1050-
004639, Inscopix) was secured over the baseplate to protect the lens. Training paradigms. Mice were co-housed and placed on a restricted
For optogenetic activation surgeries, mice aged ~6–8 weeks water access schedule prior to the start of training. At the end of each
(n = 14 male mice, 14 female mice) were anesthetized with 1–2% iso- day mice were provided 1.5 mL of water in addition to their con-
flurane and placed in stereotactic setup (Kopf). A microsyringe sumption during training.
(Nanoject) was used to inject 0.35 µL of AAV5-hSyn-hChR2(H134R)-
EYFP (experimental, Addgene, injected titer of 2.4 × 1013 parts/ml, n = 8 Training stage 1: Operant conditioning. In the first stage (operant
male mice, 8 female mice) or AAV5-CAG-GFP (control, Addgene, conditioning), mice were trained to associate poking the reward port
injected titer of 1.3×10^13 parts/ml, n = 6 male mice, 6 female mice) with a sucrose reward. Nose port entry triggered the port LED light to
bilaterally into the mPFC (1.6 mm anterior, 0.5 mm lateral and 2.4 mm turn on and resulted in the delivery of a sucrose reward of 10 µL of 10%
in depth). Mice were then implanted with custom-made optic ferrules sucrose in water. Each session lasted an hour and animals were run
consisting of a 2.5 mm diameter metal ferrule sleeve (Product ID: MM- through the assay once a day. Mice had to reach a minimum of 50
FER2003SS-3300, Precision Fiber Products) and a 0.3 mm diameter pokes in a single session to progress to the next training stage. If the
optic core (Product ID: FT300UMT, ThorLabs) bilaterally at a 10 degree animal did not engage with the reward port for more than 2 min after
angle into the mPFC above the viral injection (1.8 mm anterior, 0.7 mm the start of the trial, then the reward port delivered 30 µL liquid
lateral and 2.15 mm in depth). without engaging the LED. This automatic reward delivery also
For optogenetic inhibition surgeries, mice aged ~6–8 weeks (n = 13 occurred at the beginning of the first trial in each session to prime the
male mice, 13 female mice) were anesthetized with 1–2% isoflurane and animal to engage with the port.
placed in stereotactic setup (Kopf). A microsyringe (Nanoject) was
used to inject 0.35 µL of AAV1-CKIIa-stGtACR2-FusionRed (experi- Training stage 2: Single port one choice. In the second stage (single
mental, Addgene, injected titer of 1.3 × 1013 parts/ml, n = 7 male mice, 7 port one choice), mice were conditionally rewarded with 10 µL of
female mice) or AAV5-CKIIa-mCherry (control, Addgene, injected titer sucrose for poking the reward port within 8 s of the reward port LED
of 2.3 × 1013 parts/ml, n = 6 male mice, 6 female mice) bilaterally into turning on. Trials were separated by a 30 s ITI, and the animal had to
the mPFC (1.6 mm anterior, 0.5 mm lateral and 2.4 mm in depth). Mice refrain from entering the port for 3 s for the next trial to start. A trial
were then implanted with custom-made optic ferrules consisting of a was considered a sucrose reward fail if the mouse failed to enter the
2.5 mm diameter metal ferrule sleeve (Product ID: MM-FER2003SS- sucrose reward port after it had been illuminated for 8 s. Mice had to
3300, Precision Fiber Products) and a 0.3 mm diameter optic core reach a minimum of 40 successful trials during one session to proceed
(Product ID: FT300UMT, ThorLabs) bilaterally at a 10 degree angle into to the next training stage.
the mPFC above the viral injection (1.8 mm anterior, 0.7 mm lateral and
2.15 mm in depth). Training stage 3: Opposing port one choice. In the third stage
(opposing port one choice), the choice port LED furthest from the
Behavioral assays sucrose reward port (sucrose choice port) turned on to indicate trial
Automated operant assay. We developed an automated, closed-loop start. This assay is self-paced, so mice had an indefinite amount of time
operant assay to quantitatively assess social and nonsocial reward to nose-poke the illuminated sucrose choice port. A successful choice
behaviors. The assay consists of an acrylic chamber (12 × 12 × 12 inches) port poke resulted in the LED flashing on and off successively 3 times in
with social and nonsocial (sucrose) reward access on opposing sides 0.1 s intervals, after which the reward port LED turned on for up to 8 s.
and two nose-poke ports (choice ports) on the wall adjacent to both Mice had to nose-poke the reward port within this window to receive a
reward access sites (Fig. 1a). The two choice ports (Product ID: 1009, 10 µL sucrose reward. Failure to do so resulted in a reward fail and
Sanworks Bpod) are located 3 inches from the corners and 1 inch from initiation of a new trial. Trials were separated by a 15 s ITI. In addition,
the base of the chamber. The sucrose reward port (Product ID: 1009, the current trial ended if the animal engaged with the other inactive
Sanworks Bpod) is centered on one adjacent wall and contains a port instead of the sucrose choice port during the choice period. Mice
solenoid valve connected to a liquid reservoir containing 10% sucrose had to refrain from entering the sucrose choice port for 3 s for the next
in water. On the wall opposite from the sucrose reward port, a 2 × 2 trial to start. Mice had to perform 40 successful sucrose reward trials
inch square cutout provides access to the social target chamber. The to proceed to the final stage of training.
cutout is covered on the exterior side of the chamber by a flat sheet of
aluminum fixed to a conveyor belt with a stepper motor which acts as a Training stage 4: Two choice (social-sucrose) operant assay. In the
movable gate, controlled by an Arduino Uno. Behind this cutout, is a final stage (two choice operant), mice on the restricted water
smaller acrylic chamber (3 × 3 × 3 inch) with an exposed side blocked access schedule were given 500 µl of water prior to the start of each
by 6 aluminum rods (0.2 cm in diameter and 0.5 cm apart) that holds session (partial water access condition). As with the previous training
the target animal. Social target animals were non-cagemate, age- and stage, trial start was indicated by the choice port LED turning on.
sex-matched conspecifics. Nose pokes are detected by infrared beam However, unlike the previous training stage, a second choice port was
break sensors. A trial begins when both choice ports illuminate and is activated so that the animal could choose between social and non-
followed by nose-poking in either choice port to initiate social or social rewards. Nose-poking the illuminated choice port furthest from
nonsocial reward delivery. Nose-poking of the choice port furthest the sucrose reward port (sucrose choice port) resulted in the LED
from the sucrose reward port illuminates the reward port and makes flashing on and off successively 3 times in 0.1 s intervals, after which
10 µL of sucrose available upon entry into the reward port for up to 8 s. the sucrose reward port LED turned on for 8 s. Mice had to enter the
Alternatively, nose-poking of the choice port furthest from the social sucrose reward port within this window to receive a 10 µL sucrose
target chamber, opens the gate for 20 s, which allows the experimental reward. Failure to do so resulted in a sucrose reward fail. Nose-poking
animal to interact with the social target through the aluminum bars the illuminated choice port furthest from the social target chamber
without permitting entry into each other’s chamber. Following sucrose (social choice port) resulted in the LED flashing on and off successively

Nature Communications | (2024)15:8018 20

Article https://doi.org/10.1038/s41467-024-52294-6

3 times in 0.1 s intervals, after which an automated gate opened for Social extinction. Once fully trained, mice were removed from the
20 s to provide the mouse access to a social target (non-cagemate, sex- limited water schedule and had ad libitum access to water outside of
and age-matched conspecific) in a smaller chamber. The social target behavioral sessions. They were maintained on this full water access
chamber contains an open side covered with aluminum bars that are (FW) schedule and were run daily on the two choice operant assay for
0.5 cm apart. This allows mice to freely interact, smell and investigate 5 days. After 5 days of full water access, the social target was removed,
each other without permitting either animal to enter the other cham- and mice were run daily for 6 days without a social target (Supple-
ber. The gate remained open for 20 s before closing. Failure to inves- mentary Fig. 6g).
tigate the social target in this time resulted in a social reward fail. Trials
were separated by a 30 s ITI. Mice were randomly assigned to one of Definition of task events and parameters. Trial start—the time point
two configurations of the operant chamber in which the social and when both choice ports illuminate.
sucrose rewards were on different sides in order to account for base- Social choice—the time point when the experimental mouse
line side preferences that may arise. enters (nose-pokes) the social choice port after a trial starts.
Sucrose choice—the time point when the experimental mouse
Post-training: Two choice (social-sucrose) operant assay. Once fully enters (nose-pokes) the sucrose choice port after a trial starts.
trained, mice were removed from the partial water access schedule Social choice latency—the time between the trial start and social
and had ad libitum access to water outside of behavioral sessions. They choice port entry. Entry into port (nose-poke) is detected by an
were maintained on this full water access (FW) schedule and were run infrared beam break.
daily on the two choice operant assay for a week. For restricted water Sucrose choice latency—the time between the trial start and
access (RW) experiments, animals were water restricted (1.5 mL of sucrose choice port entry. Entry into port (nose-poke) is detected by
water/day) without access to 500 µL of water before sessions and run an infrared beam break.
daily on the two choice operant assay for a week. For social isolation Social reward zone—a 1 × 3 inch rectangular region within the
experiments, mice were fully trained and run daily for 1 week on the full operant arena adjacent to and centered around the social target access
water access schedule. Mice were then single housed for 1 week and point where the gate opens.
not run on the assay. After 1 week, mice were run again for one session Social reward consumption—the time point when the experi-
on the two choice operant assay. mental mouse first enters the social reward zone following a social
choice. Entry into the social reward zone is detected from the beha-
Empty cage/object control. After mice had been run through the first vioral video using Bonsai software83. A mouse is considered to have
three training stages, a subset of the mice were run on the final stage of “consumed” a social reward if it enters the social reward zone within
training with either an empty social target cage (empty cage group) or 20 s after social choice port entry.
a novel object (object group) (Supplementary Fig. 4). Mice were then Sucrose reward consumption—the time point when the experi-
removed from the limited water schedule, given ad libitum water mental mouse first enters the sucrose reward port following a sucrose
access, and run daily for several days on the full water access schedule choice. Entry into the sucrose reward port is detected by an infrared
without a social target and with either an empty cage or a novel object. beam break. A mouse is considered to have “consumed” a sucrose
One mouse from the empty cage group was excluded from full water reward if it enters the sucrose reward port within 8 s after sucrose
access analysis due to failure to engage with the task. choice port entry.
Social reward latency—the time between social choice port entry
Multi-choice operant assay. A third choice port (null port) was added and the first entry into the social reward zone.
to the two choice operant assay (Supplementary Fig. 5a). Nose-poking Sucrose reward latency—the time between sucrose choice port
the null port resulted in neither a social nor a sucrose reward (null entry and the first entry into the sucrose reward port.
trial). A subset of mice were run on this assay through similar training Social reward fail—a trial in which the experimental mouse fails to
stages as the two choice operant assay. Mice were then removed from enter the social reward zone within 20 s (duration for which the social
the limited water schedule, given ad libitum water access, and run daily gate is open) following social choice.
for 9 days on the full water access schedule. Sucrose reward fail—a trial in which the experimental mouse fails
to enter the sucrose reward port within 8 s (duration of sucrose reward
Estrous cycle monitoring. A subset of the female mice underwent port activation) following sucrose choice.
vaginal lavage after each behavioral session. The cytological evidence Successful trial—a trial in which the experimental mouse entered
from each lavage was used to determine the daily estrous phase of the respective reward zone following choice port entry while the
these female mice (Supplementary Fig. 6a, b). reward was available (20 s for social trials, 8 s for sucrose trials). On the
multi-choice assay (Supplementary Fig. 5a), all null trials were con-
Water reward control. After mice had been run through the first three sidered successful.
training stages, a subset of the mice were run on the final stage of
training in which the sucrose reward had been replaced with water Behavioral analysis
(Supplementary Fig. 10a–l). When the mouse nose-poked the non- All behavioral sessions were recorded at 40 Hz using Pylon software.
social reward port, 10 µL of water were dispensed instead of 10 µL of Following behavioral sessions, videos were analyzed and behavior was
sucrose solution. Mice were then removed from the limited water quantified. Several metrics were used to quantify mouse behavior for a
schedule, given ad libitum water access, and run daily for several days given behavioral session, including number of pokes per reward type,
on the full water access schedule while receiving a water reward when choice latency and reward latency. Poke number was defined as the
performing a nonsocial trial. total number of times that an animal nose-poked a choice port, inde-
pendent of reward consumption. Number of successful trials was
Passive exposure. Mice were placed in a behavioral arena and pas- determined as the total number of times that an animal nose-poked a
sively exposed to either a non-cagemate, sex- and age-matched social choice port and entered the appropriate reward zone when the reward
target through the opening of an automated gate or a 10 µL droplet of a was available for consumption. Choice latency was defined as the time
sucrose solution from a sucrose port (Supplementary Fig. 10m–o). between when the choice port LEDs turned on (trial start) and the
Passive exposure type was randomized and there were at least 30 s animal nose-poked either choice port. Following each choice port
between exposures. nose-poke, the chosen port would flash and the reward would be made

Nature Communications | (2024)15:8018 21

Article https://doi.org/10.1038/s41467-024-52294-6

available for consumption. Reward latency was defined as the time behavioral task parameters, such as reward latency, between (non-
from the start of the presentation of the reward to the consumption of stim versus stim trials) and across (GFP control versus ChR2
the reward. Sucrose reward consumption start was measured as the experimental) mice.
first entry of the mouse into the reward port after it was activated by a
sucrose choice port nose-poke. Social reward consumption start was Optogenetic inhibition experiments
measured as the first entry of the mouse into a 1 × 3 inch area (social Approximately 1 week post injection and ferrule implants, GtACR- and
zone) in front of the opened social gate following a social choice port mCherry-expressing mice were trained sequentially on each beha-
nose-poke. Social zone entries were tracked through behavioral video vioral assay. Before each behavioral session, optic ferrules on the
analysis using Bonsai software83. Failure to consume a reward within animal’s head were coupled to patch cords (Product ID: SBP, Doric)
the allotted time following reward presentation was considered a attached to a commutator (Product ID: FRJ_1 × 1_FC, Doric) using
reward fail. ceramic split mating sleeves (Product ID: ADAF, ThorLabs). The com-
To quantify the time to investigate a social target after the reward mutator was connected to the 473 nm laser (Product ID: BL473T8-
was presented (social latency), behavioral video data from each ses- 200FC, SLOC Lasers) via a patch cable (Product ID: M72L05, Thor-
sion were analyzed using Bonsai software83. We used the software to Labs). Mice were tethered during behavioral training to habituate them
track start and end of trials by detecting changes in light levels of the to the patch cables. Mice were run on the two choice operant assay for
choice port LEDs. Additionally, we tracked the body position (cen- 5 days on the partial water access schedule followed by 5 days on the
troid) of the mouse in the chamber and the mouse’s entry into either full water access schedule. Optogenetic activation experiments were
reward area (1 × 3 inches) around the social target and sucrose reward then performed over 7 consecutive behavioral sessions. During each
port. Reward latency was quantified as the time elapsed between 1 hour behavioral session, mice were stimulated for 8 s during the
turning off of the choice port LED and the animal’s first entrance into reward period on a random 50% of trials (473 nm, 8 s pulse duration,
the corresponding reward zone. This analysis was combined with 10 mW at fiber tip). Bpod Sanworks and a custom MATLAB program
existing session data from Bpod to quantify the social and nonsocial were used to control laser pulses. Behavioral analysis was performed
reward behavior of each mouse on all behavioral (training and post- using Bonsai and custom MATLAB software to compare behavioral
training) sessions. task parameters, such as reward latency, between (non-stim versus
stim trials) and across (mCherry control versus GtACR
Endoscopic calcium imaging experiments and analysis experimental) mice.
Approximately 1 week post injection, GRIN lens (0.6 mm diameter,
7.3 mm length) implanted mice were trained sequentially on each Analysis of optogenetic experiments
behavioral assay. Once the mice had been trained for ~2 weeks, they We tracked each mouse’s position from the time of choice port entry
were implanted with the baseplate and habituated with a tethered to the end of the trial (social trials: from nose-poke to 20 s after;
dummy microscope (Product ID: 1050-003762, Inscopix) during sucrose trials: from nose-poke to 8 s after) using its centroid coor-
behavioral sessions for at least 3 days prior to the actual imaging dinates obtained through Bonsai analysis. With these coordinates, we
session. On the day of imaging, mice were allowed to habituate to created a 2-D distribution (256×256 pixels) of each mouse’s
the scope for 10 min prior to the start of the session. During the exploration of the chamber. To quantify the distance traveled in
behavioral imaging session, the LED power was set to 0.7 and the centimeters during the reward period per trial (Fig. 6f, j, p, t and
analog gain on the image sensor was set to 1.6–1.8. Images were Supplementary Fig. 11d, i, n, s), we identified the number of non-zero
acquired at 20 Hz using Inscopix nVista software. Following imaging pixels from the time of choice port entry to the end of the trial and
acquisition, data were spatially downsampled by a factor of 4 and scaled the size of each pixel to its size in centimeters. To calculate the
motion corrected using Inscopix software. A CNMFe algorithm46 time spent in the social zone per social trial (Fig. 6g, k, q, u and
was then used to identify individual neurons and their fluorescence Supplementary Fig. 11e, j, o, t), we used Bonsai tracking analysis to
traces. The fluorescence traces were subsequently aligned to the identify the mouse’s entry into the social reward zone from the time
behavioral data and used for task modulation analysis. One animal of choice port entry to the end of the trial (20 s after social choice
was excluded from calcium imaging analysis due to failure to port entry). To control for locomotor changes with mPFC activation,
express GCaMP6f. Custom MATLAB software was used for all ima- mice were stimulated for up to 60 s in the operant arena outside of
ging data analysis. the reward period. The distance traveled (cm) outside of the reward
period was quantified in 5 s bins for the duration of the stimulation as
Optogenetic activation experiments described above (Supplementary Fig. 11f, k, p, u).
Approximately 1 week post injection and ferrule implants, ChR2- and
GFP-expressing mice were trained sequentially on each behavioral Histology and microscopy
assay. Before each behavioral session, optic ferrules on the animal’s After all imaging and optogenetic experiments were completed, mice
head were coupled to patch cords (Product ID: SBP, Doric) attached to were sacrificed and perfused with 0.5% PBS followed by 4% paraf-
a commutator (Product ID: FRJ_1×1_FC, Doric) using ceramic split ormaldehyde (PFA). The brains were dissected out and fixed in 4% PFA
mating sleeves (Product ID: ADAF, ThorLabs). The commutator was overnight at 4 °F. They were then transferred to 30% sucrose solution
connected to the 473 nm laser (Product ID: BL473T8-200FC, SLOC and allowed to fix for a minimum of 24 h. To visualize viral expression
Lasers) via a patch cable (Product ID: M72L05, ThorLabs). Mice were and GRIN lens/ferrule placement, brains were sliced at 50 µm slices
tethered during behavioral training to habituate them to the patch using an Epredia HM430 microtome and the slices containing the
cables. Mice were run on the two choice operant assay for 5 days on the target brain regions were mounted in DAPI solution (Product ID: 0100-
partial water access schedule followed by 5 days on the full water 20, Southern Biotech). Slices were imaged using fluorescence micro-
access schedule. Optogenetic activation experiments were then per- scopy (Keyence BZ-X800). Images were registered using WholeBrain
formed over 7 consecutive behavioral sessions. During each 1 hour software84 to determine GRIN lens and ferrule placement (RRID:
behavioral session, mice were stimulated for 8 s during the reward SCR_015245).
period on a random 50% of trials (473 nm, 20 Hz, 5 ms pulse duration,
1–3 mW at fiber tip). Bpod Sanworks and a custom MATLAB Data analysis and statistics
program were used to control laser pulses. Behavioral analysis was Task modulation analysis. We used a method described in ref. 20 to
performed using Bonsai and custom MATLAB software to compare classify neurons as significantly modulated by various task events (trial

Nature Communications | (2024)15:8018 22

Article https://doi.org/10.1038/s41467-024-52294-6

start, social/sucrose choice, social/sucrose reward)20. To do this, we where, T is the total duration of the recording, F ðt Þ is the z-scored
examined the normalized change in fluorescence in each peri- fluorescence of the neuron at time t, N events is the number of behavioral
behavioral task window. Given the temporal proximity of task events events being modeled, and N splines is the number of degrees of
(especially choice and reward delivery), we designated a 3 s window freedom used for the spline basis set (N splines = 10). The regression
around each task event and calculated the mean fluorescence within parameters βse and β0 are the coefficients for the sth spline for event e
that window. Peri-behavioral task windows were assigned as follows to and intercepts respectively. Lasso regularization parameter λ is
eliminate overlap between task event responses: trial start: 1.5 s before identified through minimizing the squared distance between the
and after; choice: 2.5 s before 0.5 s after; reward: 0.5 s before and 2.5 s fluorescence and the modeled data, using five fold cross-validation.
after. For excitatory responses, we compared the maximum value of Spse ðt Þ is the independent variable of the regression, and is calculated
the mean fluorescence to the maximum value expected by a shuffled as follows -
distribution (>99.5 percentile). For inhibitory responses, which
require a sustained suppression of fluorescence, we compared the X
Ts
mean value of fluorescence across the time window to that expected Spse ðt Þ = be ðtsÞ × spliness ðtsÞ ð2Þ
by a shuffled distribution (<0.5 percentile). The shuffled distribution ts = 1

was generated by circularly shifting (coded in MATLAB using circshift)

where, ts samples from 1 to Ts = 3 s, be ðtsÞ is the binary array for the
individual fluorescence traces by random values over 1000 iterations
event e, and spliness is the sth bspline basis set, created in MATLAB
and then averaging shuffled activity within each peri-behavioral time
using the create_bspline_basis function of the MATLAB package fdasrvf
window.
(https://github.com/jdtuck/fdasrvf_MATLAB).
To determine if the percent of non-overlapping reward-respon-
Considering the temporal proximity of the events, we chose 3 s
sive neurons observed in male and female mice (Fig. 4e) was greater
windows for the choice events (“Sucrose choice”/“Social choice”) as
than chance, we compared the proportion of non-overlapping reward
2.5 s before, and 0.5 s after the event; reward events (“Sucrose reward”/
responses found in male and female mice to a shuffled distribution.
“Social reward”) as 0.5 s before, and 2.5 s after; and “Trial start” to be
The shuffled distribution was generated using a conservative shuffling
1.5 s before and after the event.
method in which we randomly shuffled labels of social or sucrose
Finally, using the regression parameters βse and β0 , we calculated
reward-responsive neurons while maintaining a constant number of
the kernel for each behavioral event and neuron as
neurons (repeated over 100 iterations). The real data were then com-
pared to the shuffled data using an unpaired t test.
X
N splines
Kernelse ðt Þ = βse spliness ðt Þ ð3Þ
Tracking neurons across imaging sessions. We imaged mPFC neu-
s=1
rons on the two choice operant assay during three water access con-
ditions (two full water access and one restricted water access We used this encoding model to identify neurons that sig-
conditions) and tracked these neurons across imaging sessions as nificantly encode a behavioral event. To identify significant neurons,
described in ref. 50. The average time between imaging sessions in we compared the change in the goodness of fit of our encoding model
male mice was as follows: FWvRW: 8.13 ± 0.72 days, FWvFW2: (Eq. 1) to the fluorescence data calculated with all events (full model,
21.29 ± 1.13 days and FW2vRW: 12.14 ± 0.86 days. The average time with N events = N) to a reduced model where one of the events is
between imaging sessions in female mice was as follows: FWvRW: removed (N events = N-1). We used F-statistic as a measure of goodness
10.80 ± 0.92 days, FWvFW2: 4.0 ± 0.84 days and FW2vRW: of fit for the models and iteratively calculated this statistic by removing
7.0 ± 0.32 days. We then classified neurons based on their reward predictors of each event from the model. We then compared this value
responses (social reward excite, sucrose reward excite, sucrose reward with 500 instances of shuffled data, obtained by shuffling the fluor-
inhibit or reward unresponsive) and compared their reward response escence data in 1-s bins. We obtained a p-value by comparing the
profiles across similar (full water access and full water access 2) and F-statistic from the real data to the shuffled data. If the calculated
dissimilar (full water access/full water access 2 and restricted water p-value for the response of a neuron to the event is less than 0.01
access) water access conditions. (account for multiple comparisons), the neuron is considered to
encode the event.
Encoding model and task modulation analysis. To isolate the con-
tribution of each behavioral event (trial start, social choice, sucrose Neural trajectories for choice and reward. To determine the popu-
choice, social reward, sucrose reward) to the variation in the fluores- lation activity trajectories for sucrose and social trials, we considered all
cence activity, we used a linear encoding model as previously descri- recorded neurons across trials and reduced them to a three-dimensional
bed to estimate neural response kernels for every neuron21,22 neural subspace using Principal Component Analysis (PCA)85. For
(Supplementary Fig. 8). We modeled the z-scored fluorescence activity visualizing the male and female choice (Fig. 3k, m) and reward (Fig. 5e, f)
of each neuron (chosen as the dependent variable) as a consequence of trajectories, we calculated trial-averaged z-scored fluorescence chosen
a behavioral event (chosen as the independent variable), which was around choice or reward time windows respectively and concatenated
designed to be a set of binary arrays of the behavioral event time (1 across animals to form a three-dimensional array as [number of neurons
when the event occurs, 0 otherwise), convolved with an spline basis set x trial type x time window]. We fit a PCA (coded in MATLAB using pca) to
(with 10 splines) spanning 3 s. Thus, the independent variables of the the array after collapsing across the last two dimensions and projected
multiple-linear regression model capture time-delayed relationships the resulting dimensionality-reduced array independently onto social
between the behavioral event and the corresponding fluorescence. and sucrose trials (score output from pca). To quantify the difference
Mathematically, the regression model is set up as below - between the neural subspace occupied by these trials (Fig. 3l, n; 5e, f) we
repeated the procedure as above for each animal and calculated pair-
!2 wise Euclidean distance both within and between the first three PC
X
T NX
events X
N splines
min F ðt Þ  βse Spse ðt Þ  β0 projected vectors of social and sucrose trials31. Average Euclidean dis-
t =1 e=1 s=1 tances (Fig. 3l, n; 5e, f) are reported across 9 male mice and 6 female
! ð1Þ
NX
events X
N splines mice. Additionally, in Fig. 5e, f, we identified the contribution of social
 
+λ β  and sucrose reward neurons to the variability explained by PC1 as
se
e=1 s=1 compared to the rest of the mPFC population. Average PC1 coefficients

Nature Communications | (2024)15:8018 23

Article https://doi.org/10.1038/s41467-024-52294-6

(absolute values for the loading of PC1, coeff output from pca) are Sex decoding. To determine if sex identity could be decoded from
reported across all neurons identified as significantly encoding social neural activity during the choice and reward periods, we adjusted the
(134 neurons in male, 202 neurons in female), sucrose reward (88 neu- aforementioned logistic regression model as follows. For each trial
rons in male, 107 neurons in female) or neither (283 neurons in male, type (social or sucrose trial), we trained binary decoders to predict if the
298 neurons in female). In Fig. 7n, o, we determined if neural subspace neural responses corresponded to either a male or female mouse.
occupied by social and sucrose trials varied with changes in the internal Across the 8 male and 5 female mice (1 male and 1 female mouse were
state of the animal. We calculated trial-averaged z-scored fluorescence excluded due to insufficient number of trials), we performed 40 pairwise
chosen around reward from neurons tracked across both the FW and comparisons and calculated decoding accuracies across both unshuffled
RW access conditions (see section Tracking neurons across imaging and shuffled data. We then reported the average accuracy across the 40
sessions) that were concatenated, dimensionality-reduced, and pro- comparisons made during social and sucrose choice (Fig. 3i) or reward
jected independently onto social and sucrose, FW and RW trials as (Fig. 5a). As an additional measure of decoder performance, we also
described previously. We also quantified the Euclidean distances both reported the F1 scores (calculated as the ratio of precision and recall for
between and within (Fig. 7n, o) the PC projected vectors of the above each decoder) for decoding sex from social and sucrose choice (Sup-
four trial types and reported the average across 8 male mice and 5 plementary Fig. 9j) and reward (Supplementary Fig. 7f). Similar to our
female mice. previous decoder models, we matched the number of trials and neurons
in each comparison. Shuffling was performed as described in the
Decoding choice and reward using neural data. To decode the “Decoding choice and reward using neural data” section above.
identity of choice (Fig. 3h) or reward (Fig. 5b), we used logistic
regression (coded in MATLAB using fitclinear) for the neural data from Mahalanobis distance. To identify if neural responses to social and
the corresponding choice or reward time window. For each mouse, we sucrose rewards are uniquely represented at a population level, we
used a trial-matched number of sucrose and social trials to construct a calculated the Mahalanobis distance between the reward responses of
three-dimensional array of [number of neurons × number of trials × time all neurons to their baseline. In line with previous studies27,31,88, we
window of z-scored fluorescence data]. This ensured that each binary chose the Mahalanobis distance to quantify the difference introduced
decoder was trained with balanced labels accounting for the variability in the population response due to our two rewarding conditions, while
in the number of sucrose and social trials performed by each mouse. accounting for the underlying baseline variability in the neurons89. For
The three-dimensional array was then collapsed across the last two each trial type of social or sucrose, we calculated the Mahalanobis
dimensions to use as an input to the regression. This was done to distance as:
maximize the number of samples that could be used to train our  
decoder, as described in ref. 86. The input to the decoder was then D F ðt Þpopulation
trialtype
, F ðt Þpopulation
baseline
divided into train and test sets with a 70–30 train-test split and the r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  
>
train set was used to optimize the regression model hyperparameters = F ðt Þpopulation
trialtype
 F ðt Þpopulation
baseline
 Σ1 population
baseline  F ðt Þtrialtype  F ðt Þpopulation
baseline
using Lasso regularization. We evaluated the performance of the
ð4Þ
decoder, (reported as decoding accuracy) by applying the trained
model to predict labels for the test set. We repeated the above steps to where at time t, F ðt Þpopulation is the population response vector (number
trialtype
calculate decoding accuracy for 500 iterations in which we randomly population
sampled trials into the train and test set. The average decoding accu- of neurons × 1), F ðt Þbaseline is the average population response vector
racy for each animal was reported as an average of 500 iterations. As a at baseline, Σ1
baseline and is the covariance matrix calculated over all
comparison, we also calculated the decoding accuracy for a shuffled baseline timepoints. We calculated baseline responses for each mouse
dataset that was created by shuffling the z-scored fluorescence data in as the fluorescence average over the inter-trial-interval period of 30 s
non-overlapping 1-s bins to maintain the autocorrelation property of and time-matched to the duration of the reward. Mahalanobis
the signal but disrupt the activity-aligned neural responses, as descri- distances are calculated as averages over all social and sucrose trials
bed in ref. 87. Average decoding accuracy is reported across 9 male across male (Fig. 5c, Supplementary Fig. 9k: n = 101 social and
mice and 5 female mice (1 female mouse was excluded due to insuffi- 153 sucrose trials, 9 mice) and female mice (Fig. 5d, Supplementary
cient number of sucrose trials). Fig. 9l: n = 104 social and 83 sucrose, 5 female mice) from 3 s before to
5 s after social/sucrose reward and baseline. Boxplots on the right
Time-course of choice decoding. To determine the earliest time panel are average distances for a 3 s window around the reward.
point at which trial type (sucrose or social) can be successfully deco-
ded from neural activity prior to choice port entry, we modified the Statistical analyses. All statistical tests were performed using custom
aforementioned logistic regression model as follows. In particular, the MATLAB scripts. Unless otherwise specified, all statistical tests were
input to the decoder was the z-scored fluorescence with a time- two-sided and corrected for multiple comparisons. All the statistical
window of 6 s prior to and 2 s after choice port entry. For trials in which tests that were performed and their corresponding values can be
choice latency was <6 s, we time matched by linearly interpolating the found in Supplementary Data 1.
neural data from trial start to choice port entry. These data were then
binned into non-overlapping 500 ms bins, and decoding accuracy was Reporting summary
independently calculated for each time bin. This process was repeated Further information on research design is available in the Nature
for shuffled data, and significant choice decoding was identified using Portfolio Reporting Summary linked to this article.
a two-tailed t test to determine when the decoding accuracy in the time
bin of the original data was significantly different from the shuffled Data availability
data. To compare the time-course of choice decoding between The data shown in the figures are available in the Source Data file. The
male and female mice (Fig. 3g), we accounted for the differences in the raw and analyzed calcium imaging datasets collected in this study are
trial numbers and the number of animals in each group. We chose five too large to be made available online and are available upon request to
mice of each sex and matched the number of sucrose and social trials the corresponding author. The full water access imaging dataset is
both between and within the groups. Shuffling was performed as available here: https://figshare.com/s/794dd9190c2439310955. Beha-
described in the “Decoding choice and reward using neural data vioral data is available in the Source Data file. Source data are provided
section” above. with this paper.

Nature Communications | (2024)15:8018 24

Article https://doi.org/10.1038/s41467-024-52294-6

Code availability 22. Cox, J. et al. A neural substrate of sex-dependent modulation of

Custom MATLAB code is available online here: https://github.com/ motivation. Nat. Neurosci. 26, 274–284 (2023).
muruganlab/Isaac_et_al_2024_mPFC_sex_diff. 23. Otis, J. M. et al. Prefrontal cortex output circuits guide reward
seeking through divergent cue encoding. Nature 543,
References 103–107 (2017).
1. Chen, P. & Hong, W. Neural circuit mechanisms of social behavior. 24. Ferenczi, E. A. et al. Prefrontal cortical regulation of brainwide cir-
Neuron 98, 16–30 (2018). cuit dynamics and reward-related behavior. Science 351,
2. Dölen, G., Darvishzadeh, A., Huang, K. W. & Malenka, R. C. Social aac9698 (2016).
reward requires coordinated activity of nucleus accumbens oxy- 25. Murugan, M. et al. Combined social and spatial coding in a des-
tocin and serotonin. Nature 501, 179–184 (2013). cending projection from the prefrontal cortex. Cell 171,
3. Gunaydin, L. A. et al. Natural neural projection dynamics underlying 1663–1677.e16 (2017).
social behavior. Cell 157, 1535–1551 (2014). 26. Yizhar, O. et al. Neocortical excitation/inhibition balance in infor-
4. Hung, L. W. et al. Gating of social reward by oxytocin in the ventral mation processing and social dysfunction. Nature 477,
tegmental area. Science 357, 1406–1411 (2017). 171–178 (2011).
5. Rilling, J. et al. A neural basis for social cooperation. Neuron 35, 27. Kingsbury, L. et al. Cortical representations of conspecific sex
395–405 (2002). shape social behavior. Neuron 107, 941–953.e7 (2020).
6. Panksepp, J. B. & Lahvis, G. P. Social reward among juvenile mice. 28. Levy, D. R. et al. Dynamics of social representation in the mouse
Genes Brain Behav 6, 661–671 (2007). prefrontal cortex. Nat. Neurosci. 22, 2013–2022 (2019).
7. Pearson, B. L. et al. Absence of social conditioned place preference 29. Lee, E. et al. Enhanced neuronal activity in the medial prefrontal
in BTBR T+tf/J mice: relevance for social motivation testing in rodent cortex during social approach behavior. J. Neurosci. 36,
models of autism. Behav. Brain Res. 233, 99–104 (2012). 6926–6936 (2016).
8. Hu, R. K. et al. An amygdala-to-hypothalamus circuit for social 30. Amadei, E. A. et al. Dynamic corticostriatal activity biases social
reward. Nat. Neurosci. 24, 831–842 (2021). bonding in monogamous female prairie voles. Nature 546,
9. Ramsey, L. A., Holloman, F. M., Hope, B. T., Shaham, Y. & Venniro, M. 297–301 (2017).
Waving through the window: a model of volitional social interaction 31. Kingsbury, L. et al. Correlated neural activity and encoding of
in female mice. Biol. Psychiatry 91, 988–997 (2022). behavior across brains of socially interacting animals. Cell 178,
10. Solié, C., Girard, B., Righetti, B., Tapparel, M. & Bellone, C. VTA 429–446.e16 (2019).
dopamine neuron activity encodes social interaction and promotes 32. Castelli, F., Frith, C., Happé, F. & Frith, U. Autism, Asperger syn-
reinforcement learning through social prediction error. Nat. Neu- drome and brain mechanisms for the attribution of mental states to
rosci. 25, 86–97 (2022). animated shapes. Brain 125, 1839–1849 (2002).
11. Pierce, A. F., Protter, D. S. W., Chapel, G. D., Cameron, R. T. & 33. Willsey, A. J. et al. Coexpression networks implicate human midfetal
Donaldson, Z. R. Nucleus accumbens dopamine release reflects the deep cortical projection neurons in the pathogenesis of autism. Cell
selective nature of pair bonds. Curr Biol. 34, 519–530.e5 155, 997–1007 (2013).
(2024). 34. Brumback, A. C. et al. Identifying specific prefrontal neurons that
12. Ramsey, L. A., Holloman, F. M., Lee, S. S. & Venniro, M. An operant contribute to autism-associated abnormalities in physiology and
social self-administration and choice model in mice. Nat. Protoc. 18, social behavior. Mol. Psychiatry 23, 2078–2089 (2018).
1669–1686 (2023). 35. Greene, R. K., Walsh, E., Mosner, M. G. & Dichter, G. S. A potential
13. Sugrue, L. P., Corrado, G. S. & Newsome, W. T. Choosing the greater mechanistic role for neuroinflammation in reward processing
of two goods: neural currencies for valuation and decision making. impairments in autism spectrum disorder. Biol. Psychol. 142,
Nat. Rev. Neurosci. 6, 363–375 (2005). 1–12 (2019).
14. Willmore, L. et al. Overlapping representations of food and social 36. Kohls, G. et al. Reward system dysfunction in autism spectrum
stimuli in mouse VTA dopamine neurons. Neuron 111, disorders. Soc. Cogn. Affect. Neurosci. 8, 565–572 (2013).
3541–3553.e8 (2023). 37. Kohls, G., Antezana, L., Mosner, M. G., Schultz, R. T. & Yerys, B. E.
15. Munuera, J., Rigotti, M. & Salzman, C. D. Shared neural coding for Altered reward system reactivity for personalized circumscribed
social hierarchy and reward value in primate amygdala. Nat. Neu- interests in autism. Mol. Autism 9, 9 (2018).
rosci. 21, 415–423 (2018). 38. Anderson, D. J. Circuit modules linking internal states and social
16. Jennings, J. H. et al. Interacting neural ensembles in orbitofrontal behaviour in flies and mice. Nat. Rev. Neurosci. 17, 692–704
cortex for social and feeding behaviour. Nature 565, 645–649 (2016).
(2019). 39. Tye, K. M. Neural circuit motifs in valence processing. Neuron 100,
17. Ruff, C. C. & Fehr, E. The neurobiology of rewards and values 436–452 (2018).
in social decision making. Nat. Rev. Neurosci. 15, 549–562 40. Mosberger, A. C., de Clauser, L., Kasper, H. & Schwab, M. E. Moti-
(2014). vational state, reward value, and Pavlovian cues differentially affect
18. Lockwood, P. L., Apps, M. A. J. & Chang, S. W. C. Is There a ‘Social’ skilled forelimb grasping in rats. Learn. Mem. 23, 289–302 (2016).
Brain? Implementations and Algorithms. Trends Cogn. Sci. 24, 41. Allen, W. E. et al. Thirst regulates motivated behavior through
802–813 (2020). modulation of brainwide neural population dynamics. Science 364,
19. Starkweather, C. K., Babayan, B. M., Uchida, N. & Gershman, S. J. 253 (2019).
Dopamine reward prediction errors reflect hidden-state inference 42. Zelikowsky, M. et al. The neuropeptide Tac2 controls a distributed
across time. Nat. Neurosci. 20, 581–589 (2017). brain state induced by chronic social isolation stress. Cell 173,
20. Cameron, C. M., Murugan, M., Choi, J. Y., Engel, E. A. & Witten, I. B. 1265–1279.e19 (2018).
Increased cocaine motivation is associated with degraded spatial 43. Chen, C. S. et al. Divergent strategies for learning in males and
and temporal representations in IL-NAc neurons. Neuron 103, females. Curr. Biol. 31, 39–50.e4 (2021).
80–91.e7 (2019). 44. Greiner, E. M., Müller, I., Norris, M. R., Ng, K. H. & Sangha, S. Sex
21. Parker, N. F. et al. Choice-selective sequences dominate in cortical differences in fear regulation and reward-seeking behaviors in a
relative to thalamic inputs to NAc to support reinforcement learn- fear-safety-reward discrimination task. Behav. Brain Res. 368,
ing. Cell Rep. 39, 110756 (2022). 111903 (2019).

Nature Communications | (2024)15:8018 25

Article https://doi.org/10.1038/s41467-024-52294-6

45. Becker, J. B. & Chartoff, E. Sex differences in neural mechanisms 67. Mastrogiovanni, N. A., Wheeler, A. K. & Clemens, K. J. Social isola-
mediating reward and addiction. Neuropsychopharmacology 44, tion enhances cued-reinstatement of sucrose and nicotine seeking,
166–183 (2019). but this is reversed by a return to social housing. Sci. Rep. 11,
46. Zhou, P. et al. Efficient and accurate extraction of in vivo calcium 2422 (2021).
signals from microendoscopic video data. Elife 7, e28728 (2018). 68. Walker, D. M. et al. Crystallin mu in medial amygdala mediates the
47. Rigotti, M. et al. The importance of mixed selectivity in complex effect of social experience on cocaine seeking in males but not in
cognitive tasks. Nature 497, 585–590 (2013). females. Biol. Psychiatry.11, 895–906 (2022).
48. Euston, D. R., Gruber, A. J. & McNaughton, B. L. The role of medial 69. Heilig, M., Epstein, D. H., Nader, M. A. & Shaham, Y. Time to connect:
prefrontal cortex in memory and decision making. Neuron 76, bringing social context into addiction neuroscience. Nat. Rev.
1057–1070 (2012). Neurosci. 17, 592–599 (2016).
49. Matteucci, G. et al. Cortical sensory processing across motivational 70. Yamamuro, K. et al. Juvenile social isolation enhances the activity of
states during goal-directed behavior. Neuron 110, 4176–4193.e10 inhibitory neuronal circuits in the medial prefrontal cortex. Front.
(2022). Cell. Neurosci. 14, 105 (2020).
50. Sheintuch, L. et al. Tracking the same neurons across multiple days 71. Makinodan, M. et al. Effects of the mode of re-socialization after
in Ca2+ imaging data. Cell Rep. 21, 1102–1115 (2017). juvenile social isolation on medial prefrontal cortex myelination and
51. van Zessen, R., Phillips, J. L., Budygin, E. A. & Stuber, G. D. Activation function. Sci. Rep. 7, 5481 (2017).
of VTA GABA neurons disrupts reward consumption. Neuron 73, 72. Lee, C. R., Chen, A. & Tye, K. M. The neural circuitry of social
1184–1194 (2012). homeostasis: consequences of acute versus chronic social isola-
52. Adamantidis, A. R. et al. Optogenetic interrogation of dopaminergic tion. Cell 184, 2794–2795 (2021).
modulation of the multiple phases of reward-seeking behavior. J. 73. Eiselt, A.-K. et al. Hunger or thirst state uncertainty is resolved by
Neurosci. 31, 10829–10835 (2011). outcome evaluation in medial prefrontal cortex to guide decision-
53. Siemian, J. N., Arenivar, M. A., Sarsfield, S. & Aponte, Y. Hypotha- making. Nat. Neurosci. 24, 907–912 (2021).
lamic control of interoceptive hunger. Curr. Biol. 31, 74. Grove, J. C. R. et al. Dopamine subsystems that track internal states.
3797–3809.e5 (2021). Nature 608, 374–380 (2022).
54. Gangopadhyay, P., Chawla, M., Dal Monte, O. & Chang, S. W. C. 75. Matthews, G. A. et al. Dorsal raphe dopamine neurons represent the
Prefrontal-amygdala circuits in social decision-making. Nat. Neu- experience of social isolation. Cell 164, 617–631 (2016).
rosci. 24, 5–18 (2021). 76. Chevallier, C., Kohls, G., Troiani, V., Brodkin, E. S. & Schultz, R. T. The
55. Hong, W., Kim, D.-W. & Anderson, D. J. Antagonistic control of social social motivation theory of autism. Trends Cogn. Sci. 16,
versus repetitive self-grooming behaviors by separable amygdala 231–239 (2012).
neuronal subsets. Cell 158, 1348–1361 (2014). 77. Kinard, J. L. et al. Neural mechanisms of social and nonsocial reward
56. Guenthner, C. J., Miyamichi, K., Yang, H. H., Heller, H. C. & Luo, L. prediction errors in adolescents with autism spectrum disorder.
Permanent genetic access to transiently active neurons via TRAP: Autism Res 13, 715–728 (2020).
targeted recombination in active populations. Neuron 78, 78. Clements, C. C. et al. Evaluation of the social motivation hypothesis
773–784 (2013). of autism: a systematic review and meta-analysis. JAMA Psychiatry
57. Chen, C. S., Knep, E., Han, A., Ebitz, R. B. & Grissom, N. M. 75, 797–808 (2018).
Sex differences in learning from exploration. Elife 10, e69748 79. Lewis, E. M. et al. Parallel social information processing circuits are
(2021). differentially impacted in autism. Neuron 108, 659–675.e6
58. Dhingra, I. et al. Sex differences in neural responses to reward and (2020).
the influences of individual reward and punishment sensitivity. BMC 80. Bariselli, S. et al. SHANK3 controls maturation of social reward cir-
Neurosci 22, 12 (2021). cuits in the VTA. Nat. Neurosci. 19, 926–934 (2016).
59. Borland, J. M., Rilling, J. K., Frantz, K. J. & Albers, H. E. Sex- 81. Kim, J., Jung, M. W. & Lee, D. Reward learning improves social
dependent regulation of social reward by oxytocin: an inverted U signal processing in autism model mice. Cell Rep. 42, 113228
hypothesis. Neuropsychopharmacology 44, 97–110 (2019). (2023).
60. Vahaba, D. M., Halstead, E. R., Donaldson, Z. R., Ahern, T. H. & Beery, 82. Selimbeyoglu, A. et al. Modulation of prefrontal cortex excitation/
A. K. Sex differences in the reward value of familiar mates in prairie inhibition balance rescues social behavior in CNTNAP2-deficient
voles. Genes Brain Behav 21, e12790 (2022). mice. Sci. Transl. Med. 9, 7 (2017).
61. Tan, T. et al. Neural circuits and activity dynamics underlying sex- 83. Lopes, G. et al. Bonsai: an event-based framework for processing
specific effects of chronic social isolation stress. Cell Rep. 34, and controlling data streams. Front. Neuroinform. 9, 7 (2015).
108874 (2021). 84. Fürth, D. et al. An interactive framework for whole-brain maps at
62. Tomova, L. et al. Acute social isolation evokes midbrain craving cellular resolution. Nat. Neurosci. 21, 139–149 (2018).
responses similar to hunger. Nat. Neurosci. 23, 1597–1605 (2020). 85. Cunningham, J. P. & Yu, B. M. Dimensionality reduction for large-
63. Senst, L., Baimoukhametova, D., Sterley, T.-L. & Bains, J. S. Sexually scale neural recordings. Nat. Neurosci. 17, 1500–1509 (2014).
dimorphic neuronal responses to social isolation. Elife 5, 86. Lui, J. H. et al. Differential encoding in prefrontal cortex projection
e18726 (2016). neuron classes across cognitive tasks. Cell 184, 489–506.e26
64. Campi, K. L., Greenberg, G. D., Kapoor, A., Ziegler, T. E. & Trainor, B. (2021).
C. Sex differences in effects of dopamine D1 receptors on social 87. Engelhard, B. et al. Specialized coding of sensory, motor and cog-
withdrawal. Neuropharmacology 77, 208–216 (2014). nitive variables in VTA dopamine neurons. Nature 570,
65. Ross, A. P. et al. Sex-dependent effects of social isolation on the 509–513 (2019).
regulation of arginine-vasopressin (AVP) V1a, oxytocin (OT) and 88. Remedios, R. et al. Social behaviour shapes hypothalamic neural
serotonin (5HT) 1a receptor binding and aggression. Horm. Behav. ensemble representations of conspecific sex. Nature 550,
116, 104578 (2019). 388–392 (2017).
66. Wallace, D. L. et al. CREB regulation of nucleus accumbens excit- 89. Li, Y. et al. Neuronal representation of social information in the
ability mediates social isolation-induced behavioral deficits. Nat. medial amygdala of awake behaving mice. Cell 171,
Neurosci. 12, 200–209 (2009). 1176–1190.e17 (2017).

Nature Communications | (2024)15:8018 26

Article https://doi.org/10.1038/s41467-024-52294-6

Acknowledgements Peer review information Nature Communications thanks the anon-

We would like to thank Dr. Larry Young, Dr. Robert Liu, and Dr. Nancy ymous reviewer(s) for their contribution to the peer review of this work. A
Padilla-Coreano for their feedback on the manuscript. We would like to peer review file is available.
thank Jan Hawes, Dr. Julia Cox, and Dr. Yunmiao Wang for their technical
assistance with this project. This work was supported by National Insti- Reprints and permissions information is available at
tutes of Health grants R01MH130755 (M.M.) and F31MH133373 (J.I.); the http://www.nature.com/reprints
Emory Conte Center Pilot Project Grant (M.M.) and the Emory University
Research Committee Grant (M.M.). J.L.J. and M.M. were supported by Publisher’s note Springer Nature remains neutral with regard to jur-
HHMI Gilliam Fellowship GT15888. isdictional claims in published maps and institutional affiliations.
The mouse schematic used in the depiction of the behavioral assay
(Figs. 1b and 2f; Supplementary Figs. 1a, 4a, 5a, 5b, 8b, 10m and 11a) was Open Access This article is licensed under a Creative Commons
adapted from https://openclipart.org/detail/17558/simple- Attribution-NonCommercial-NoDerivatives 4.0 International License,
cartoon-mouse. which permits any non-commercial use, sharing, distribution and
reproduction in any medium or format, as long as you give appropriate
Author contributions credit to the original author(s) and the source, provide a link to the
J.I. and M.M. conceived the project and designed the experiments. J.I., Creative Commons licence, and indicate if you modified the licensed
S.C.K. and N.S. optimized the behavioral assay. J.I. and S.C.K., with material. You do not have permission under this licence to share adapted
support from J.L.J. and M.R., collected the data. J.I. and H.B. analyzed the material derived from this article or parts of it. The images or other third
data. J.I., S.C.K., H.B. and M.M. wrote the manuscript. party material in this article are included in the article’s Creative
Commons licence, unless indicated otherwise in a credit line to the
Competing interests material. If material is not included in the article’s Creative Commons
The authors declare no competing interests. licence and your intended use is not permitted by statutory regulation or
exceeds the permitted use, you will need to obtain permission directly
Additional information from the copyright holder. To view a copy of this licence, visit http://
Supplementary information The online version contains creativecommons.org/licenses/by-nc-nd/4.0/.
supplementary material available at
https://doi.org/10.1038/s41467-024-52294-6. © The Author(s) 2024

Correspondence and requests for materials should be addressed to

Malavika Murugan.

Nature Communications | (2024)15:8018 27