Induction of reversible bidirectional social approach bias by olfactory conditioning in male mice

Abstract

Social avoidance is a common component of neuropsychiatric disorders that confers substantial functional impairment. An unbiased approach to identify brain regions and neuronal circuits that regulate social avoidance might enable development of novel therapeutics. However, most paradigms that alter social avoidance are irreversible and accompanied by multiple behavioral confounds. Here we report a straightforward behavioral paradigm in male mice enabling the reversible induction of social avoidance or approach with temporal control. C57BL/6J mice repeatedly participated in both negative and positive social experiences. Negative social experience was induced by brief social defeat by an aggressive male CD-1 mouse, while positive social experience was induced by exposure to a female mouse, each conducted daily for five days. Each social experience valence was conducted in a specific odorant context (i.e. negative experience in odorant A, positive experience in odorant B). Odorants were equally preferred pre-conditioning. However, after conditioning, mice sniffed positive experience-paired odorants more than negative experience-paired odorants. Furthermore, positive- or negative-conditioned odorant contexts increased or decreased, respectively, the approach behavior of conditioned mice toward conspecifics. Because individual mice undergo both positive and negative conditioning, this paradigm may be useful to examine neural representations of social approach or avoidance within the same subject.

INTRODUCTION

Abnormal social approach and avoidance are hallmarks of social anxiety and autism spectrum disorders (ASD), and have been identified across most neuropsychiatric disorders (Hoertnagl and Hofer 2014). Social deficits result in significant impairment in both personal and vocational function (Couture et al. 2006), and while behavioral interventions can be effective for some individuals (Taylor et al. 2015), pharmacotherapies for social deficits in clinical studies have had mixed results (Guastella and Hickie 2016; Berry-Kravis et al. 2017), with no currently approved agents. Challenges in therapeutic development for social deficits include the difficulty of modeling a complex human behavior in animals, the effects of developmental processes, and an incomplete understanding of neuronal circuitry and brain regions governing social processes. Novel approaches are necessary to overcome these barriers.

Most studies assessing a social behavior phenotype involve manipulations of a specific gene, neuronal subpopulation, or circuit, and then evaluation of the effects on performance in social assays compared to controls. Studies of ASD risk genes (Won et al. 2012) and oxytocin (Dolen et al. 2013) exemplify this ‘reverse’ approach. Less common is the forward approach, in which a model organism with a specific alteration in social behavior is subject to analysis to identify altered genes, neuronal activity, or circuit connectivity. While the forward approach was once common in genetic and biochemical studies, the explosion of techniques to measure and manipulate neuronal activity and connectivity involved in specific behaviors (Gunaydin et al. 2014; Yang W and Yuste 2017) makes this approach particularly important for studies of social behavior.

Here we sought to develop a straightforward method by which approach behavior of wildtype male mice could be reversibly increased or decreased within-subject in a temporally-controlled manner. The advantage of such an approach would be to enable within-subject comparisons of neural representations influencing social behaviors. Many social and non-social behavioral paradigms can reduce social approach (Toth and Neumann 2013), for example, chronic mild stress (Kompagne et al. 2008; Venzala et al. 2013) and social defeat (Kudryavtseva et al. 1991; Krishnan et al. 2007; Henriques-Alves and Queiroz 2015). However, social deficits resulting from these models develop during the training paradigm and are persistent, thereby eliminating the possibility of within-subject comparisons, and between-subject comparisons between trained mice and controls suffer from numerous confounds such as altered locomotor behavior, anxiety, and exploratory drive (Allsop et al. 2014).We took advantage of olfactory associative learning, whereby rodents efficiently learn to associate a distinct odor with an unconditioned stimulus, that has been previously described primarily in rodent fear conditioning (Otto et al. 1997; Paschall and Davis 2002; Jones et al. 2005), and in reproductive behavior primarily in rats (Kippin et al. 2001; Coria-Avila et al. 2005; Ismail et al. 2008). We hypothesized that pairing a distinct odorant context with a social interaction of a positive or negative valence might result in enhanced or reduced social approach behavior, respectively, in the conditioned contexts. Using repeated interaction with a female mouse as the positive experience and repeated brief social defeat as the negative experience, we demonstrate differential influences on social behaviors only in the conditioned contexts, without development of important behavioral confounds.

MATERIALS AND METHODS

Animals

Male and female C57BL/6J mice (age 10–12 weeks upon arrival) were purchased from Jackson Laboratory. All mice were group housed, 4–5 mice per cage. CD-1 retired breeders were purchased from Charles River and housed singly. All mice were housed under standard conditions (temperature: 21 ± 2 °C, 12-hour light-dark cycle, food and water available ad libitum). All procedures were approved by the Yale University Institutional Animal Care and Use Committee.

Odorants

Odorants used were imitation almond extract (Badia, Doral, FL), and banana extract (McCormick, Hunt Valley, MD). All odorants were diluted with distilled water at the ratio noted in the individual behavioral tests below.

Behavioral training and testing

General

Prior to all behavioral training or testing, mice were habituated in the behavior room for at least 30 min. All experiments were conducted during the light cycle. All training sessions to develop odorant associations were conducted in separate rooms between odorants to eliminate cross-contamination.

Olfactory habituation/dishabituation test

Habituation to the novel odors was performed essentially as described (Yang M and Crawley 2009). After acclimatizing to the testing room, male C57BL/6J mice were placed individually in clean standard mouse cages, and were exposed for at least 30 min to one clean cotton swab suspended into the cage through the top food rack. This procedure reduces the novelty of the cotton swab and the cage. After habituation, the swab was replaced with a new swab dipped in distilled water and mice were exposed to a freshly dipped swab 3 times for 2 min each time. A new swab was then dipped in almond extract (diluted 1:100 in distilled water) and presented to the mouse 3 times for 2 min each time. Finally, the procedure was repeated with swabs dipped in banana extract (diluted 1:100 in distilled water), again 3 times for 2 min each presentation. A fresh swab was used for every presentation. The time between each swab presentation was 1 min. The time spent sniffing the swab was quantified as the time the mouse spent with its nose within the immediate proximity of the swab, meaning one head length or less away from the odor with nose directed toward the swab. Time was recorded using a standard stopwatch.

Olfactory preference test

Male C57BL/6J mice were acclimatized to the testing room, then placed individually in clean standard mouse cages in which two cotton swabs were suspended from the empty food rack and tested for baseline odor preference, essentially as described (Zou et al. 2015). Cotton swabs were situated approximately 7 cm apart from each other and oriented symmetrically about the long axis of the mouse cage. After habituation to the new cage and swabs for at least 30 min, both swabs were removed and replaced with 2 new swabs, one dipped in almond extract and the other dipped in banana extract, both diluted 1:100 in distilled water. The odorants were presented for 3 min, and the time spent smelling each odorant was quantified with a stopwatch. The side of the cage in which each odorant was presented was counterbalanced across mice.

Pairing of positive and negative social experience with olfactory contexts (Figure 2)

Figure 2.

Conditioning paradigm and testing schematic. Prior to conditioning, mice are tested in the 3-chamber maze to assess pre-conditioning odorant sniffing preference. The olfactory conditioning paradigm is then conducted daily for 5 days. C57BL/6J mice experience a negative valence interaction (interaction with an aggressive CD-1 male mouse) and a positive valence interaction (interaction with a C57BL/6J female mouse) for 10 mins each, in the presence of either almond or banana scent. Post-tests are conducted beginning in the afternoon of day 5 following the final conditioning paradigm, and were continued up until day 9. In addition to counterbalancing the odorant-interaction pairings and the order in which interactions occurred each day (depicted in the schematic), we also alternated the order of interactions experienced by any individual mouse (example for an individual mouse: day 1: female interaction followed by CD-1 interaction, day 2: CD-1 interaction followed by female, and so forth).

Positive social experience: interaction with female mouse

A gauze pad saturated in almond or banana extract diluted 1:20 in distilled water was placed under the bedding of a clean standard mouse cage for 1 hr to permeate the cage with odorant. A C57BL/6J male (test mouse) and a similar aged female C57BL/6J mouse were then placed in the odor-saturated cage, and the mice were permitted to interact for 10 min. Following odor exposure and interaction, the mice were returned to their respective home cages and returned to the vivarium. Estrus phase was not determined for interacting females.

Negative social experience: interaction with aggressive male mouse

Retired CD-1 male breeders were chronically single housed and allowed to habituate to the testing room. A gauze pad saturated in almond or banana odorant extract diluted 1:20 in distilled water was placed into the home cage of the CD-1 mouse for ~1 hr. A C57BL/6J mouse was then placed in the odor-containing home cage of the CD-1 male mouse and allowed to interact and experience multiple bouts of offensive aggression for 5 min in a modified social defeat paradigm (Mineur et al. 2013; Mineur et al. 2016). These interactions were monitored in real-time to ensure that no wounding occurred. After 5 min of direct interaction, a wire mesh barrier was introduced between the mice, with no further contact allowed, but continued experience of the stressful environment and odorant for an additional 5 min. During these encounters, the intruder (C57BL/6J test mouse) demonstrated stereotyped defeat postures, including standing on hind legs with paws held in front of the body curled downward following an attack. With subsequent time within an interaction session and across sessions, the intruder would display this defeat posture simply upon CD-1 approach without even requiring attack (anticipatory defeat posture). Following the interactions, mice were then returned to their home cage and place back in the vivarium.

Conditioning protocol

To develop an association between an odorant and a specific valence social interaction, mice were repeatedly exposed to positive and negative valence interactions in distinct odorant contexts. Each C57BL/6J test mouse experienced both the positive and negative social interaction experience daily for 5 days. For an individual test mouse, the odorant-interaction pairing was always the same (for example, interaction with the female mouse always occurred in almond context and interaction with aggressive CD-1 always occurred in the banana context). Female and CD-1 mice were each allocated to a specific odorant, meaning each interaction in which they participated was always in almond context or always in banana context. The specific CD-1 or female mouse used in each encounter with the test mouse was randomly chosen. The first interaction of the day was conducted in the early morning, with the second interaction of the day in the late morning or early afternoon. Post-conditioning tests (three-chamber test, dyadic interactions, and tail suspension test, all described below) were conducted beginning in the afternoon following the final conditioning interaction and continued for up to five days following the final day of conditioning.

Counterbalancing

To reduce the likelihood of order effects, we counterbalanced the odorant-interaction pairings (for example, the number of test mice experiencing the female interaction in an almond context was similar to the number of test mice experiencing the female interaction in a banana context, and the number of test mice experiencing the CD-1 interaction in an almond context was similar to the number of test mice experiencing the CD-1 interaction in a banana context). We also counterbalanced the order in which the interactions occurred each day (for example, the number of test mice each day experiencing the female interaction first was similar to the number of test mice experiencing the CD-1 interaction first). Finally, the order in which each individual mouse was exposed to the female mouse and CD-1 was alternated daily (for example, on day 1 of the protocol, a given test mouse experienced the female interaction first, followed by the CD-1 interaction second. Then on day 2 of the protocol, the given test mouse would experience the CD-1 interaction first and the female interaction second).

Three-chamber odorant preference test

The three-chamber odorant preference test was conducted before and after the olfactory conditioning paradigm. After habituation to the testing room, each C57BL/6J mouse was allowed to explore a standard three-chamber arena (Noldus, Leesburg, VA), in which inverted cups with several holes were placed in the two lateral chambers adjacent to an empty middle chamber. After the 5-min exploration period, the cups were lifted and a gauze pad soaked in almond or banana extract diluted 1:20 with distilled water was placed under each cup, such that one side of the arena contained almond odorant and the other contained banana odorant. The mouse was allowed to explore freely for an additional 5 min. Behavior was video recorded. The side of the cage in which the odorants were placed was counterbalanced across mice. For both the habituation and test periods, the time spent in each maze compartment, the time spent sniffing each cup (quantified as the time spent in which the mouse nose was oriented toward the odor source and less than one head length away), and the distance traveled by the mouse were quantified using Ethovision XT (Noldus). Mice displaying ≥ 80% preference for a single side prior to the odor association pairing were excluded from analysis (n = 4). The amount of time sniffing the positively- or negatively-paired odorant (or the odorant that would become positively- or negatively-paired, in the case of the test performed prior to beginning the conditioning paradigm) was divided by the total time spent sniffing either odorant.

Dyadic interactions

A gauze pad soaked in 1:20 dilution of almond or banana extract was placed under the bedding in a clean standard mouse cage for at least 30 min to permit cage permeation. Interaction pairs, consisting of two conditioned C57BL/6J test mice or one conditioned C57BL/6J test mouse and an unconditioned C57BL/6J male mouse of similar age were then gently placed in the cage and allowed to interact for 10 min. For each mouse, the number of distinct approaches made toward the conspecific were recorded. An approach was defined as a specific episode of behavior where the subject mouse was moving its entire body, headfirst, toward its interaction partner that culminated in sniffing or direct contact with the interaction partner.

Tail suspension test

Mice were gently suspended by the tail over a clean basin containing gauze soaked in 1:20 dilution of either almond or banana extract for 6 min, behavior videotaped, and the time spent immobile was quantified.

Data analysis

All data are represented as mean ± standard error of the mean (s.e.m.) unless otherwise stated. Comparisons between two groups were performed using t tests (paired or unpaired as appropriate), while one-way analysis of variance (ANOVA) and tests for linear trend were used to investigate comparisons between 3 or more groups. Repeated-measures ANOVA, with ‘odorant’ and ‘presentation order’ as within-subject factors was used to evaluate the habituation-dishabituation test. Two-way mixed ANOVA, with ‘conditioning’ as a between-subject factor and ‘context’ as a within-subject factor, was used to evaluate the conditioning paradigm during conspecific interactions. Geisser-Greenhouse correction was used to correct for violation of sphericity. In the event of a significant overall ANOVA (i.e. F statistic achieving p < 0.05), Sidak’s post-tests were performed. Data were organized using Microsoft Excel version 16, and statistical analyses and plots performed using GraphPad Prism version 8.

RESULTS

We first identified odorants that could be easily discriminated, but equally preferred, so they could be associated with positive or negative social experiences without inducing a preference on their own. Male C57BL/6J mice were able to discriminate between water, almond extract, and banana extract in an odor habituation-dishabituation test, as shown by a main effect of odorant presentation order (two-way repeated-measures ANOVA: F(2, 28) = 73.2, p < 0.0001) with post-hoc analyses demonstrating significant differences between the initial odorant presentation and two subsequent presentations (habituation), and significant differences between the final odorant presentation and the initial presentation of the new odorant (dishabituation; Figure 1A). There was no main effect of odorant (F(2, 28) = 0.05, p = 0.95) and no presentation order x odorant interaction (F(4, 56) = 1.41, p = 0.24). In the odorant preference test, mice demonstrated no preference for the almond or banana odorants (paired t test: t(14) = 0.18, p = 0.86; Figure 1B). These data support the use of almond and banana extract for subsequent conditioning experiments.

Figure 1.

C57BL/6J mice discriminate almond from banana odor but do not demonstrate an odor preference. A. Mice (n = 15) were repeatedly and sequentially presented with cotton swabs dipped in water, almond extract, or banana extract (1:100 dilutions), and the amount of time spent sniffing was quantified. Mice habituate to repeated presentations of the same odor, and dishabituate when presented with a novel odorant, demonstrating intact discrimination. B. Mice (n = 15) were given a choice between swabs dipped in almond and banana extracts (1:100 dilutions) and time spent sniffing each swab was quantified. No significant differences were observed in time spent sniffing either odorant. Data are expressed as mean ± s.e.m. ****p < 0.0001.

We next developed a novel conditioning paradigm to pair a specific odorant with a positive or negative social experience so we could influence social approach behavior in a temporally-defined and reversible manner. Chronic social defeat paradigms can reduce social approach in the defeated mouse, but also induce additional stress-related behaviors that can persist long after the initial defeat paradigm (Hollis and Kabbaj 2014). In contrast, subthreshold social defeat paradigms do not result in persistent behavioral alterations in wild type mice following training (Krishnan et al. 2007). We hypothesized that pairing brief defeat with an odor might provide a method to reactivate the neural representation of the defeat experience, and might therefore be sufficient to induce social avoidance in the context of the defeat-paired odor, but not once the odor was withdrawn. Conversely, exposure to a female conspecific is a positive social experience (Choe et al. 2015), and can reverse stress-induced behaviors (Ramirez et al. 2015). We therefore hypothesized that reactivation of this experience with a paired odor might enhance social approach behavior.

To test this hypothesis, we developed a conditioning paradigm during which male C57BL/6J mice repeatedly interacted with an aggressive CD-1 male mouse in a limited social defeat protocol, and repeatedly interacted with a female C57BL/6J. Each social interaction occurred in either an almond or banana odorant context. Both positive valence (female interaction) and negative valence (CD-1 social defeat) interactions were performed daily for five days, followed by behavioral post-tests to quantify the efficacy of the conditioning paradigm (Figure 2). To reduce the influence of interactive or order effects, we counterbalanced the odorant-interaction valence pairing as well as the order in which the mice experienced the interactions each day (detailed in methods).

To determine whether this conditioning paradigm resulted in differential valuation of an odorant over baseline levels, we quantified the time spent sniffing the odorant paired with the positive female interaction experience versus the time spent sniffing the odorant paired with the negative social defeat experience using a three-chamber maze (Figure 3) (Beny and Kimchi 2016). We compared performance in this task in subjects pre- and post-conditioning. The conditioning paradigm resulted in a significant increase in the time spent sniffing the positively-conditioned odorant (or, equivalently, a significant reduction in the amount of time spent sniffing the negatively-conditioned odorant; paired t test: t(10) = 2.29, p = 0.045; Figure 3A). Prior to conditioning, time spent sniffing each odorant was no different than chance (one-sample t test vs. 50%: t(10) = 1.00, p = 0.34), while after conditioning, mice developed a preference for sniffing the positively-conditioned odorant and an aversion to sniffing the negatively-conditioned odorant that were significantly different from chance (one-sample t test vs. 50%: t(10) = 2.27, p = 0.046). In contrast to the time spent sniffing, mice did not demonstrate a significant preference for the chamber in which the conditioned odorant was placed (paired t test: t(10) = 0.74, p = 0.47; Figure 3B), nor did conditioning result in a chamber preference significantly different than chance (one-sample t test vs. 50%: t(10) = 1.4, p = 0.19), suggesting that direct contact with the odorant is necessary for forming the neural representation of valence, and also suggesting that the odorants did not diffuse to the other chambers of the apparatus. Finally, locomotor behavior during the three-chamber task was not significantly different before or after conditioning, importantly demonstrating that the conditioning paradigm did not induce changes in general locomotion or overall propensity to explore (paired t test: t(10) = 1.72, p = 0.12; Figure 3C).

Figure 3.

Associative learning of odorant-social interaction pairings of positive or negative valence induced differential valuation of conditioned odors, without general locomotor or exploratory changes. A. Subject mice were tested in a three-chamber apparatus. The lateral chambers contained a porous cup with gauze soaked in diluted almond or banana extract, while the middle chamber was empty. Mice were tested prior to the conditioning paradigm (“pre-conditioning”), and then following conditioning (“post-conditioning”). The percent of time spent sniffing the odorant that was positively-conditioned was increased following the conditioning paradigm as compared to before the conditioning paradigm. Furthermore, mice sniffed the positively-conditioned odorant at a percentage greater than chance only after conditioning. B. Conditioning did not induce a significant preference in time spent in the positively-conditioned odorant chamber, nor did it increase the time spent in the chamber to greater than chance. C. Locomotion, quantified by distance traveled during three-chamber maze exploration, did not change significantly after conditioning. Data are expressed as mean ± s.e.m. n = 11 mice for all panels. #p < 0.05 one-sample t test versus 50%.

We next tested whether social interactions were altered as a result of the conditioning paradigm. As an initial assessment, we paired 2 conditioned subject mice in a clean, neutral mouse cage in the presence of a specific odorant context. One of the mice in the interaction was positively-conditioned to the odorant, while the other mouse in the interaction was negatively-conditioned to the odorant, thus maximizing the power to observe a differential effect in the interaction (Figure 4A). Interactions between differentially-conditioned mice were repeated such that each mouse experienced an interaction with a novel conspecific in the presence of its positively-conditioned odorant, negatively-conditioned odorant, and in the absence of an odorant. Novel pairs of conditioned mice were used each time to limit social habituation. Two-way mixed ANOVA of number of approaches demonstrated a significant conditioning x context interaction (F(2, 24) = 33.3, p < 0.0001; Figure 4B). Post-test analysis demonstrated significant differences in number of approaches in the context of almond odorant (p = 0.0062), and banana odorant (p = 0.0094), but not in interactions conducted in the absence of odorant (p = 0.97).

Figure 4.

Olfactory conditioning differentially modifies social approach behavior. A. Schematic of post-conditioning social interaction test. Pairs of mice with opposite conditioning experiences interacted in a clean, neutral mouse cage in the presence of almond odorant, banana odorant, or no odorant. B. The number of approaches made by each mouse during the interactions depicted in (A) were quantified during a 10-min period (n = 7 mice per each conditioning group). C. The number of approaches made by a conditioned subject mouse toward an unconditioned male conspecific was quantified during a 10-min period (n = 5 for banana-positive/almond-negative and n = 10 for almond-positive/banana-negative). D. The number of approaches shown in (C) was collapsed across odorants into interaction contexts. “Positive context” denotes that the interaction occurs in the presence of the positive experience-paired odor for the subject mouse. “Neutral context” denotes interaction in the absence of an odorant. “Negative context” denotes that the interaction occurs in the presence of the negative-experience-paired odor. There was a significant difference between number of approaches in the presence of the positive- and negative-interaction-paired odors, as well as a significant linear test for trend (p = 0.0023) across the three ordered contexts (n = 15 per context). E. The number of approaches made by the unconditioned male conspecific mice toward conditioned mice do not differ across the three odorant contexts. Data are expressed as mean ± s.e.m. **p < 0.01.

These results suggest an effect of conditioning on motivation for social approach. However, because the interactions were between differentially-conditioned pairs, the finding may be explained by mice in the presence of their positively-conditioned odorant approaching more, mice in the presence of their negatively-conditioned odorant approaching less, or both. To determine whether both positive- and negative-conditioning paradigms were successful within-subject, we repeated this dyadic paradigm, however now testing a conditioned mouse interacting with a novel, unconditioned male conspecific. Number of approaches were quantified in the context of the positively-conditioned odorant, negatively-conditioned odorant, and in the absence of an odorant. In this paradigm, two-way mixed ANOVA again revealed a significant conditioning x context interaction (F(2, 26) = 4.66, p = 0.019; Figure 4C). We quantified the results of these interactions by examining the number of approaches made by conditioned mice in “positive” interaction contexts (exposure to the positive experience-paired odor), “neutral” interaction contexts (exposure to no odorant), or “negative” interaction contexts (exposure to the negative experience-paired odor). One-way repeated-measures ANOVA demonstrated a significant main effect of interaction context (F(1.90, 26.57) = 5.61, p = 0.010; Figure 4D). Post hoc-tests revealed a significant difference between positively-conditioned and negatively-conditioned contexts (p = 0.0061), as well as a significant linear test for trend across the three ordered contexts (F(1, 28) = 11.21, p = 0.0023). One-way ANOVA demonstrated no difference in the number of approaches made by the unconditioned mice toward the conditioned mice across the three odorant contexts (F(2, 42) = 0.22, p = 0.81; Figure 4E). These data suggest that reactivation of the olfactory association with both the positive social interaction and negative social interaction has independent effects on social approach.

Finally, we tested whether the odorant conditioning paradigm influenced performance in the tail suspension test (TST), a test of behavioral despair with relevance for depression-like behavior. We found no effect when comparing the time spent immobile in the presence of the positively-conditioned odorant versus the negatively-conditioned odorant in the TST (unpaired t test: t(13) = 0.48, p = 0.64; Figure 5). These results suggest the influence of odorant conditioning on social approach are unlikely to be explained as a component of a larger depression-like phenotype induced by social defeat.

Figure 5.

Olfactory conditioning does not result in changes in immobility in a measure sensitive to chronic stress. Mice underwent olfactory conditioning and were then tested in the tail suspension test in the presence of the odor paired with positive social experience (n = 7) or paired with negative social experience (n = 8). The time spent immobile during the 6-min test was quantified, and there was no significant difference between the two groups. Data are expressed as mean ± s.e.m.

DISCUSSION

In this study we designed and tested a novel behavioral paradigm to study social approach and avoidance. We used olfactory conditioning to pair an inherently positive or negative social experience with a distinct odorant. Our 5-day conditioning paradigm resulted in differential valuation of two odorants that were previously equally preferred. Social approach was significantly increased only when trained mice interacted in the context of an odorant previously paired with a positive social experience. In contrast, exposure to the negatively-conditioned odor decreased social interactions. Importantly, exposure to the odorants that resulted in changes in social behavior did not influence locomotor or exploratory behavior, or behavior in a test sensitive to antidepressant treatment. Altered social behavior was temporally bounded by odorant exposure, and could be interleaved to alter behavior bidirectionally, as each animal received both positive and negative conditioning and reacted appropriately to each stimulus. We therefore propose this paradigm as a straightforward method in which to influence social approach or avoidance within-subject with temporal control, and potentially as a tool to study neural representations of social motivation

In rodents, direct connections between olfactory sensory areas and brain regions critical for both memory and emotional processing makes olfactory conditioning particularly powerful for this form of emotional recall. Specifically, direct connections exist between the olfactory bulb and the piriform cortex (Haberly and Price 1977), amygdala (Price 1973), and hippocampus via the entorhinal cortex (Haberly and Price 1977; Wilson and Steward 1978; Mouly et al. 1998; Biella and de Curtis 2000). These regions are densely interconnected as well (Brothers and Finch 1985; McDonald and Mascagni 1997). Several studies demonstrate the importance of many of these regions in valence encoding (Han et al. 2009; Redondo et al. 2014; Gore et al. 2015; Tye 2018). Consistent with these anatomical and functional connections, removal of the olfactory bulbs, bulbectomy, has long been shown to have critical effects on social, aggressive, and affective behaviors (Neckers et al. 1975; Song and Leonard 2005; Jimenez-Sanchez et al. 2016). Direct social defeat (Matsuda et al. 1996) or exposure to olfactory cues from the bedding of the defeat aggressor mouse (Bourne et al. 2013) increase neuronal activation in many of these regions. It is possible that exposure to the conditioned odorants in our paradigm activates many of these regions important for memory and emotional processing, with downstream connection to other brain regions known to be important in social behaviors.

Our work using bidirectional olfactory conditioning builds upon a substantial literature of unidirectional olfactory conditioning. Several studies in rats have demonstrated the utility of olfactory-conditioned partner preference to influence mating behavior in male (Kippin et al. 2001; Ismail et al. 2008) and female rats (Coria-Avila et al. 2005). Intriguingly, odorant conditioning has also been used to induce both an interactive and sexual preference for same-sex rats when olfactory conditioning was performed in conjunction with administration of the dopamine agonist quinpirole, in male (Triana-Del Rio et al. 2011; Cibrian-Llanderal et al. 2012; Coria-Avila et al. 2018) and female (Tecamachaltzi-Silvaran et al. 2017) rats. In male mice, social avoidance behavior toward females could be induced by pairing female pheromones with highly aversive lithium chloride (LiCl) for three consecutive days (Beny and Kimchi 2016). Conversely, pairing an unconditioned odor with a female mouse resulted in significantly greater approach to the positively conditioned odor, and this effect required oxytocin signaling in the piriform cortex, while non-social learning did not require oxytocin (Choe et al. 2015). In the Choe et al. study, odorants paired with estrus/proestrus females were more rewarding than odorants paired with diestrus females, however, other studies have shown that interaction with a female mouse confers positive valence without determination of estrus phase or confirmation of mating (Redondo et al. 2014; Ramirez et al. 2015). It is notable that most investigation regarding social behavior and olfactory conditioning have studied rats and focused primarily on reproductive behaviors. Our paradigm may be of particular utility to the study of social processes because of our use of social interaction as the unconditioned negative stimulus, as opposed to an aversive pharmacological treatment such as LiCl. However, it should be noted that the use of sexual and aggressive interactions as unconditioned stimuli followed by group housing may influence dominance hierarchies in grouped mice and shift subsequent behaviors (Annas et al. 2013; van den Berg et al. 2015). Olfactory cues have also been successfully used as conditioned stimuli in several other behaviors relevant to neuropsychiatric disorders in rats and mice. Examples of conditioned responses include freezing (Cousens and Otto 1998; Jones et al. 2005; Pavesi et al. 2012; Carew et al. 2018; Ross and Fletcher 2018), fear-potentiated startle (Paschall and Davis 2002; Jones et al. 2005), and shock avoidance and appetitive behaviors (Choi et al. 2011; Meissner-Bernard et al. 2019).

Complex social behavior results from integration of multimodal sensory information, motivational states, and motor programs (Anderson 2016). Identification of neural circuit activity specifically governing social approach and avoidance can be confounded by circuits necessary for social approach but with little specificity, such as those supporting locomotion. Commonly, brain states are inferred from observed behavior, and lead to comparison in circuit or neuronal activity while the mouse is engaged in one behavior (for example, “exploration”), versus another (for example, “freezing”). One drawback to this approach is the difficulty in disentangling the neural signature underlying the motivation of the behavior from the signature of the behavior itself. The current protocol may facilitate onset of neural activity relevant to social approach or avoidance in response to exposure to a conditioned odorant, and therefore serve as an alternative to measurement of neural activity during social behavior itself. This implementation might enhance the specificity of future studies geared toward identifying the neural circuitry of social behaviors across different brain regions.

In conclusion, we have developed a novel olfactory conditioning paradigm by pairing odorant contexts with social interactions of opposing valence. This novel paradigm is unique in that individual mice learn associations of both positive and negative valence, and reactivation of these neural representations results in alterations in social approach within each animal. This work advances a rich literature of olfactory conditioning that was previously focused on unidirectional pairing. The ability to quantify differential activation of specific neurons, brain regions, or circuits within-subject mediating social approach and avoidance may be an important model for the identification of therapeutic targets to improve social deficits.