Conditioned place preferences induced by hearing song outside the breeding season relate to neural dopamine D₁ and cannabinoid CB₁ gene expression in female European starlings (Sturnus vulgaris)

Abstract

The affective state induced by sensory stimuli changes to adaptively modify behaviors that are critical for survival and reproduction. In European starlings, during the spring breeding season, male courtship song is rewarding to females, but only to those that possess resources that are necessary for reproduction (i.e., nesting sites). In fall, starling song is non-sexual and proposed to maintain flocks. This suggests that in fall it may be adaptive for females to be rewarded by fall rather than spring, courtship song. We used a conditioned place preference (CPP) test to evaluate song-induced affective state in fall condition females and quantitative real-time PCR to measure expression of genes that modulate affective state (CB₁ endocannabinoid and D₁ dopamine receptors) in brain regions that were previously implicated in song-induced reward (i.e., the medial preoptic nucleus (mPOA) and ventromedial hypothalamus (VMH)). Fall-condition females developed an aversion to a place that had been paired with playback of both male fall and courtship song, indicating that in general male song induces a negative affective state outside the breeding season. Song-induced aversion was stronger in birds conditioned towards an initial place preference. For mPOA, CB₁ receptor expression correlated positively with fall and spring song-induced CPP. D₁ receptor expression correlated negatively with fall (but not spring) song-induced CPP, and the ratio of CB₁ to D₁ receptor expression correlated positively with fall (but not spring) song-induced CPP. These correlations suggest that interactions between D₁ and CB₁ receptors in mPOA may play a role in modifying affective responses to song.

1. Introduction

An individual’s affective response to sensory stimuli changes to adaptively modify the production of behaviors critical for survival and reproductive success. For example, the reward value of food is high in hungry individuals, but food becomes neutral or even aversive in sated individuals [1], adaptively shifting investment in feeding behavior. With respect to social behavior, the reward value of offspring to female rats shifts such that pups are aversive to females prior to giving birth but become rewarding after parturition to promote maternal behaviors at an appropriate time [2]. Similarly, in seasonally breeding female songbirds (e.g., European starlings, Sturnus vulgaris) male courtship song is rewarding to females that possess a nesting territory but neutral or aversive to females that lack this critical breeding resource [3]. In this way individual differences in the affective state induced by male song ensure that only females with the resources needed to breed initiate breeding behaviors, which are energetically costly. Although critical for successful behavioral interactions, little is known about the neural mechanisms underlying flexibility in the affective state induced by social stimuli.

The goal of this study was to begin to identify mechanisms underlying adaptive changes in the reward value of social stimuli by studying affective states induced by distinct communication signals in female starlings in a non-breeding context. It has been shown that female starlings in a spring breeding condition discriminate between broadcast recordings of fall and spring male song, becoming more active during playback of spring compared to fall song [4]. Female starlings in spring condition also display preferences for long compared to short male songs (i.e., [5]). Outside the breeding season, during fall and winter months, male starlings sing high rates of non-sexually motivated song that is shorter and less stereotyped than spring courtship song [6, 7]. This type of song is proposed to promote safety by maintaining overwintering flocks [8–10]. Given the proposed function of fall song and given that breeding in fall would result in reproductive failure (i.e., chicks would hatch under harsh winter conditions), we proposed that for fall-condition females, male fall song would be rewarding; whereas, spring courtship song would be neutral or aversive. Although past studies show that female starlings in spring condition discriminate behaviorally between playbacks of fall and spring male song [4], the extent to which playbacks of spring and fall song differentially influence the affective state of fall condition females is not known.

Interactions between dopamine and endocannabinoids modulate affective state, motivation and the perception of reward valence [11–13], suggesting a potential role in modifying responses to social stimuli. Cannabinoids are synthesized “on demand” and released from postsynaptic membranes to primarily influence presynaptic neurons through retrograde signaling. In the central nervous system they commonly bind to presynaptic G-protein coupled CB₁ receptors (reviewed in [14, 15]), and it has been proposed that by rapidly gating dopamine release,CB₁ receptors may function to adjust behavior so that it is appropriate given changing environmental or social factors [16, 17].

Both direct and indirect mechanisms of dopamine – endocannabinoid interaction are proposed, and the manner in which these systems interact appears to differ depending on the brain region [18]. For example, studies in mammals show that in the ventral tegmental area (VTA), CB₁ receptors inhibit glutamate release [19] and inhibit GABA– and opioid-mediated inhibition of dopaminergic VTA neurons [20, 21]. This action is proposed to disinhibit activity in VTA dopamine neurons to facilitate production of motivated behaviors [22]. In contrast, in the medial prefrontal cortex CB₁ receptors located on dopaminergic presynaptic terminals are proposed to directly inhibit dopamine release ([19] and reviewed in [18]). D₁ dopamine receptor activation in tissue samples containing striatum and prefrontal cortex has also been found to alter cannabinoid concentrations, increasing 2-arachidonylglycerol and decreasing anandamide [23].

CB₁ receptors are also densely expressed in neurons that express D₁ receptors [24]. Although whether CB₁ and D₁ interactions occur in presynaptic axon terminals or postsynaptic targets was not determined, CB₁ and D₁ signal transduction pathways interact in the same neurons [25], and the absence of CB₁ receptors specifically in neurons containing D₁ receptors resulted in increases in social anxiety as well as indices of fear in conditional knock-out mice [26]. These findings suggest that a reduction in cannabinoid signaling in neurons expressing D₁ receptors induces a negative affective state, which may be reversed by increased CB₁ signaling.

The mesolimbic dopamine pathway has been the focus of most past research on the effects of dopamine and cannabinoids on motivated behaviors (e.g., [11, 13, 18, 27]), yet both CB₁ and D₁ receptors are found in multiple additional brain regions that are implicated in social and sexual behaviors (e.g., [16, 28–31]). This includes both the medial preoptic area (mPOA) and ventromedial hypothalamus (VMH), which are sites implicated in social and sexual reward [32, 33] and in female starlings are implicated in the regulation of the reward value of hearing male courtship song [3]. We used quantitative real-time PCR (qPCR) to measure expression of CB₁ and D₁ receptors in mPOA and VMH and explored the degree to which these measurements related to song-induced affective state in fall-condition female starlings. We additionally examined the ratio of CB₁:D₁ receptor expression related to song-induced affective state, with the prediction based on the conditional knockout mouse study described above [26] that a higher ratio would correspond to a more positive / less negative affective state.

2. Methods

2.1. Subjects

Twenty-four adult female European starlings were caught using baited fly-in traps in winter on a farm in Madison, WI. Following capture, birds were housed indoors in stainless steel cages (91 cm × 47 cm × 47 cm) with up to five females per cage. Each female received a numbered leg band and unique colored leg bands for identification. The breeding season in starlings is regulated by seasonal changes in photoperiod [34]. Females were captured in winter when day length was short and they were in a photosensitive condition under which they will respond to the increasing day lengths characteristic of spring with increases in estradiol and reproductive activities. After 6–8 weeks of exposure to long days starlings become photorefractory which terminates reproductive activity, and birds remain in this state until exposed to short day lengths that are characteristic of fall [35]). We used well-studied photoperiod manipulations to place females into this fall-typical state by housing them on a photoperiod of 18L:6D until they completed molt (for at least six weeks) and maintained them on this light cycle throughout the experiment. Starlings kept under this photoperiod are in a physiological state characterized by undetectable estradiol that is observed during the fall non-breeding season [36]. During the experiment, females were housed in groups of four in indoor aviaries (approx 3.5 m × 2.25 m × 2 m). Each aviary contained several perches. Food and water were available ad libitum. All animals were treated in accordance with the Guidelines of the National Institutes of Health for the Care and Use of Laboratory Animals under a protocol approved by the University of Wisconsin Institutional Animal Care and Use Committee.

2.2. Behavioral testing

One week after introducing the females to the aviaries, we ran a separate study that is not reported here in which the females tested here were exposed to playback of fall song, spring song and silence (each bird was exposed to each condition) to indirectly assay opioid release induced by hearing song (as in [37]). There is no statistical evidence that this experiment had any impact on the current study. One week after that study, we began song-induced conditioned place preference (CPP) testing. There were three phases of the CPP procedure, each occurring on a separate day: habituation, conditioning, and test day. The CPP apparatus consisted of a stainless steel bird cage (118 cm × 59 cm × 59 cm) divided into two visually distinct compartments. Each compartment contained different color cues (either red or blue construction paper on three walls and the ceiling) and a single perch.

The first phase consisted of a 30 min habituation period. A bird was removed from the aviary and placed singly into the CPP apparatus. The bird was allowed to freely move between the two compartments, and we measured the amount of time spent in each compartment to obtain a baseline preference. Following this habituation period, the bird was returned to the aviary.

On the following day, the conditioning phase was conducted. Each bird was removed from its aviary and placed singly into one compartment of the CPP apparatus. An opaque partition was placed into the apparatus so that the bird was restricted to one compartment for 15–min. During this period, the bird heard either 15–min of song playback or 15–min of silence. For one group of birds, song playback consisted of recordings of male courtship song (n = 12). For a second group of birds, song playback consisted of recordings of fall male song (n = 12). Four females were conditioned simultaneously in separate CPP compartments located in the same room, but females were visually isolated from one another. A separate set of song recordings from a separate set of males was played during each conditioning session, such that all 4 females in a given session heard the same recordings, but females in different sessions heard different recordings. All recordings were unfamiliar to the females and were of songs produced by males with whom the females had never interacted in the laboratory. It is possible that the females interacted with these males in the wild but unlikely given that several years separated trapping dates. No differences in conditioning were found in response to the specific recordings (p > 0.60 in all cases).

Following the 15–min conditioning period, the bird was returned to its home aviary. Approximately 2.5 hours later, the bird was again removed from its aviary and placed singly into the other compartment of the CPP apparatus and exposed to either song playback or silence (whichever was not presented earlier in the day) for 15–min. The compartment (i.e., blue or red) paired with song playback or silence was counterbalanced across birds. The order of presentation was also counterbalanced across birds (i.e., some birds heard silence first; some birds heard song playback first).

The extent to which each bird developed a conditioned preference was tested the following day during a 30 min test period. During this test phase, a single bird was removed from the aviary and placed into the CPP apparatus. The opaque partition was removed from the apparatus on test day, and similar to the habituation period, the bird had unrestricted access to both compartments. We measured how long (in seconds) the bird spent in each compartment. Following the test period, the bird was returned to its home aviary. The amount of time on test day that females spent in the formerly song-paired compartment of the apparatus minus the amount of time spent on the same compartment during habituation was used as a measure of CPP (i.e., song-induced reward). For this measure, 0 indicates that song had no valence relative to silence, positive numbers indicate that hearing song induced a reward state (or reduced a negative affective state), and negative numbers indicate that song induced an aversive state.

2.3. Tissue preparation for qPCR

On the day following CPP test day, birds were played either spring song playback (n = 9) or fall song playback (n = 9), consistent with what was played on conditioning day for 15–min, or were exposed to silence for 15–min (n = 6). The playback condition had no statistically significant impact on mRNA measures and we do not discuss this further. Immediately following this 15–min period, birds were rapidly decapitated. At this time all females had black beaks (characteristic of the non-breeding season [38]). Ovaries were examined and no developing follicles were present (also characteristic of the non-breeding season [39]). Brains were extracted, flash frozen using isopentane (M32631, Sigma Aldrich, St. Louis, MO, USA) cooled with dry ice, and kept at −80° C until further processing. Brains were sectioned coronally (200 μm) using a cryostat with a temperature set between −16° C and −14° C. Samples from the target brain regions were collected using a sample corer (Fine Science Tools Sample Corer Item No. 18035–02; Foster City, CA, USA). A 2 mm diameter punch was taken from VMH (1 medial punch from 1 section) and mPOA (1 medial punch from 3 sections) (Fig. 1). Tissue punches from each individual were stored in separate 1.7 mL cone-tipped tubes (one tube for each brain region) at −80° C until RNA extraction.

Figure 1.

Coronal sections showing approximate locations of 2 mm tissue punches in (A) mPOA and (B) VMH. Abbreviations: Co: optic chiasm; CoA: anterior commissure; GP: globus pallidus; HP: hippocampus; LS: lateral septum; N: nidopallium; Rt: nucleus rotundus; StL: lateral striatum.

Methods for qPCR have been published previously [40, 41]. Briefly, tissue was immersed in Purezol (BioRad product # 732688) homogenized with an electric Dremel tool and RNA was extracted with a Bio-Rad Aurum Total RNA Fatty and Fibrous Tissue Kit (Catalog No. 732–6830; Bio-Rad, Hercules, CA, USA) following manufacturer’s instructions. RNA concentration was measured with a NanoDrop system (Thermo Scientific, Wilmington, DE, USA). 100 ng of RNA was converted into single-stranded cDNA using the Invitrogen SuperScript III First-Strand Synthesis System (Catalog No. 18080–051; Life Technologies, Carlsbad, CA, USA) following manufacturer’s instructions. For qPCR analysis, cDNA was diluted to a 1:10 concentration using nuclease free water. cDNA from surrounding brain regions was pooled and used as standard tissue for qPCR analyses. We used qPCR to quantify relative gene expression for the D₁ dopamine receptor and the CB₁ cannabinoid receptor. Samples from two subjects were missing for mPOA (one was lost during extraction and one was damaged), resulting in n = 22 for this region.

2.4. qPCR analysis

We used NCBI Gene Database Primer-Blast to design primers using the zebra finch (Taeniopygia guttata) genome (for the D₁ primer) and chicken (Gallus gallus) genome (for the CB₁ primer). Each primer set had a single-peak melt curve indicating the amplification of a single gene. For each primer set, the qPCR reaction product was sequenced via Sanger sequencing at the University of Wisconsin Biotechnology Center and checked using NCBI BLAST to ensure the sequences matched the intended targets. Primer information for CB₁ and D₁ is published [16, 42] (for CB₁: Accession number: NM_001038652.1; Forward sequence: GGTCTTCTGTGGACTTAGGG; Reverse sequence: CTCCTCTATTCCTTTGTTGCTC; reaction product sequence: TAATTTTTCCCCCTCCTTTTCTTTCACATGCATATGAGACTAACAGCAACAAAGGAAT AGAGGAGA; for D₁: Accession number: NM_001243833.1; Forward sequence: ACGAGAGGAAAATGACCCCC; Reverse sequence: GTTGTAGCCTTGTGCCAGTT; reaction product sequence: GAGGAANACACGGGACAAAGGTCCACGGCCACGCCTGGATCATGGATGAAGGGCT GCCTTGGGGGGTCATTTTCCCTCTT).

To amplify cDNA, each sample (1:10 dilution) was mixed with nuclease-free H₂O, forward and reverse primers, and SsoFast EvaGreen Supermix (Catalog No. 172–5201; Bio-Rad, Hercules, CA, USA). Each plate of samples also included five standards (1:10 serial dilution) and a negative control consisting of nuclease-free H₂O in place of cDNA. Samples and standards were run in triplicate and read using the BioRad CFX96 Touch Real-Time PCR Detection System (Catalog No. 185–5195; Bio-Rad, Hercules, CA, USA). The qPCR protocol consisted of an initiation step at 95° C for 30 s, followed by 40 cycles of 95° C for 5 s, a 30–s annealing phase (temperature specific to the primer set), a 30–s elongation phase at 72° C, and a melt curve from 60° C to 88° C, 0.5 degrees for each 5–s step. After each elongation and melt curve step the plate was read and recorded. All runs followed the MIQE guidelines [43] and had an efficiency between 90–110%, an R² of at least 0.990, as well as a single-peak melt curve to verify primer specificity.

The relative levels of gene expression were determined based on the Pfaffl method (for a detailed description see [40, 44]. In brief, we set the amplification threshold at 200 RFU and for each sample, the average number of cycles that crossed the threshold (Ct) was transformed. We calculated the geometric mean of the Ct values for two reference genes (Glyceraldehyde 3-phosphate dehydrogenase [GAPDH] and hydroxymethylbilane synthase [HMBS]) in each brain region. Then, to obtain relative values, we raised the efficiency of the run for each gene (i.e., CB₁ and D₁ receptors) to the difference in the Ct value for each gene and the geometric mean of the reference gene Ct values.

3. Results

3.1. CPP and song playback

Overall, when an individual’s initial compartment preference was not considered, there was a tendency for females conditioned to fall song to have higher CPP scores than females that were conditioned to spring song (mean (sd) spring = −347.75 secs (492.27); fall = −195.00 secs (767.58), with a single sample t-test revealing that the CPP for spring song differed significantly from zero (t₁₁ = 2.45, p = 0.032), while the CPP for fall song did not (t₁₁ = 0.88, p = 0.398)). However, as detailed below, when initial compartment preference was considered, there were no significant differences found for the affective state (i.e., CPP) induced by fall or spring song playbacks, and 17 of the 24 females tested demonstrated a conditioned place aversion to the side of the apparatus that had been paired with song (i.e., they had negative CPP scores indicating that song induced a negative affective state; Fig. 2).

Figure 2.

Differences in the affective state induced by song in females conditioned against or towards initial compartment preferences. Mean (+ sem) time spent (in the absence of song playback) in the compartment of the CPP apparatus previously paired with spring song (open dots) or fall song (filled dots) minus the baseline preference for that compartment during habitation for birds conditioned against and towards initial compartment preferences. Positive CPP scores indicate that hearing song induced a positive affective state; negative CPP scores indicate that hearing song induced a negative affective state. Individual data points indicate values for individual females. Zero on the y-axis indicates no preference. *p < 0.05.

The pre-conditioning CPP compartment preferences for each bird were determined based upon which compartment of the CPP apparatus a bird spent more than 50% of its time during habituation (with 75% of the females found to prefer the red colored compartment). Irrespective of the compartment color preferred, our random assignment of birds to each conditioning compartment resulted in a balanced outcome such that approximately ½ of the birds in each group were conditioned towards the initially preferred compartment and ½ were conditioned against the initially preferred compartment (i.e., for the spring playback group, n=5 were conditioned against and n=7 were conditioned towards initial preferences; for the fall playback group, n=7 were conditioned against and n=5 were conditioned towards initial preferences). An ANOVA with song stimulus (i.e., fall or spring song) and conditioning bias (i.e., whether a bird was conditioned against or towards its initial preference) entered as independent variables and the CPP measurement entered as a dependent variable revealed a significant main effect for conditioning bias (F_{1, 20} = 12.56, p = 0.002), with CPP measures higher in birds that were conditioned against initial compartment preferences compared to birds conditioned towards initial preferences (Fig 2). No main effect was found for song stimulus (F_{1, 20} = 0.01 p = 0.913) and no main effect was detected for the song x conditioning bias interaction (F_{1, 20} = 0.48, p = 0.494; Fig. 2). A single sample t-test revealed that the CPP for birds conditioned against initial compartment preference did not differ significantly from zero (t₁₁ = 0.69, p = 0.505). In contrast, a single sample t-test revealed that CPP for birds conditioned towards the initial compartment preference differed significantly from zero (t₁₁ = 5.50, p = 0.0002). These CPP scores were negative, indicating that in this group song induced a negative affective state (Fig. 2).

3.2. CPP and mRNA

As indicated above, CPP scores were significantly lower for birds conditioned towards compared to against initial compartment biases; however, because the conditioning compartment bias was counterbalanced across birds in the spring and fall conditions, we included all birds in analyses of relationships between CPP and mRNA measures. In cases in which results were significant we report p values for analyses of birds divided based on initial compartment preferences to demonstrate that patterns were similar, though often shy of significance, likely due to a reduced sample size. We first used general linear separate-slopes models to determine whether the slopes for correlations between song-induced CPP and mRNA measures in mPOA and VMH differed significantly for birds conditioned to spring versus fall song. If slopes did not differ significantly we pooled birds that were conditioned to fall song and birds that were conditioned to spring song for subsequent correlation analyses, as the lack of difference in slopes indicated that the relationships between gene expression and CPP were not different between fall and spring song types, though we also report results of separate spring and fall analyses. (A multiple regression with all markers and brain regions entered as independent variables and CPP entered as a dependent variable was not appropriate for this analysis because variables correlated and points were missing from mPOA that then also would have been dropped from VMH). For D₁ in mPOA, standardized residual analyses identified a significant outlier (expression value = 2.40; 2 standard deviations above the mean). This outlier was removed from all analyses of D₁ expression.

A separate slopes analysis revealed no significant differences between CB₁ in mPOA for birds conditioned to spring song compared to birds conditioned to fall song (CB₁ in mPOA, F (2,18) = 2.47, p = 0.113). Therefore, we combined birds conditioned to spring and fall song for correlation analyses. A significant correlation was detected between song-induced CPP and mPOA CB₁ (r = 0.47, p = 0.027; Fig. 3A). When spring and fall song-induced CPP were analyzed separately correlations were not significant (spring: r = 0.33, p = 0.360; fall: r = 0.54, p = 0.070). When combined birds were divided based on whether they were conditioned towards or against initial side preferences p values were not significant (against: p = 0.165; towards: p = 0.341).

Figure 3.

Scatterplots showing relationships between song-induced CPP and mRNA expression in mPOA for A) CB₁ receptors, B) D₁ receptors, and C) ratio of CB₁:D₁ receptor expression for females that were conditioned to male fall (filled dots) and male spring song (open dots). For B and C, scatterplots are shown separately for females that heard male fall and females that heard male spring song. Regression line = p < 0.05.

A separate slopes model revealed that the slopes for regressions between D₁ in mPOA and CPP differed between birds conditioned to spring song compared to birds conditioned to fall song (F (2,17) = 3.82, p = 0.043; Fig. 3B). Because the slopes differed significantly, we ran separate correlation analyses for birds conditioned to spring song and birds conditioned to fall song to examine relationships between D₁ mRNA in mPOA and song-induced CPP. A significant negative correlation was detected between D₁ in mPOA and fall song-induced CPP (r = −0.63, p = 0.027) but not spring song-induced CPP (r = −0.23, p = 0.549; Fig. 3B). When the fall song conditioned birds were divided based on whether they were conditioned towards or against initial side preferences p values were significant or very close (against: p = 0.078; towards: p = 0.033).

A separate slopes analysis revealed no significant differences between D₁ in VMH or CB₁ in VMH and CPP for birds conditioned to spring song compared to birds conditioned to fall song (CB₁ in VMH, F (2,20) = 0.15, p = 0.861; D₁ in VMH, F (2,20) = 1.24, p = 0.310). Therefore, we combined birds conditioned to spring and fall song for correlation analyses. Correlations between CPP and D₁ in VMH or CB₁ in VMH were not significant (D₁ in VMH: r = −0.31, p = 0.138; CB₁ in VMH: r = −0.10, p = 0.651; Fig. 4A and B), nor were correlations significant when spring and fall song-induced CPP were analyzed separately (p > 0.17 in each case).

Figure 4.

Scatterplots showing relationships between song-induced CPP and mRNA expression in VMH for A) CB1 receptors, B) D₁ receptors, and C) ratio of CB₁:D₁ receptor expression for females that were conditioned to male fall (filled dots) and male spring song (open dots). Regression line = p < 0.05.

To gain insight into potential interactions between CB₁ and D₁ receptors and song-induced CPP, we examined relationships between CPP and the ratio of CB₁:D₁ receptor expression. A separate slopes analysis revealed a significant difference between slopes for the regression between CPP and CB₁:D₁ in the mPOA for birds conditioned to spring song compared to birds conditioned to fall song (F (2,17) = 5.53, p = 0.014). A significant correlation was detected between CB₁:D₁ in the mPOA and fall song-induced CPP (r = 0.72, p = 0.0079) but not spring song-induced CPP (r = 0.37, p = 0.321; Fig. 3C). When the combined birds were divided based on whether they were conditioned towards or against initial side preferences p values approached significance (against: p = 0.051; towards: p = 0.101). For VMH a separate slopes analysis revealed no significant difference between slopes for the regression between CPP and CB₁:D₁ for birds conditioned to spring song compared to birds conditioned to fall song (F (2,20) = 1.87, p = 0.180) and so, as in previous analyses, birds exposed to spring and fall song were pooled for correlation analysis. A significant correlation was detected between CB₁:D₁ and song-induced CPP (r = 0.41, p = 0.046; Fig. 4C); however, this relationship depends on one point with the highest CB₁:D₁. When considered separately correlations between spring and fall song-induced CPP were not significant (spring: r = 0.42, p = 0.179; fall: r = 0.39, p = 0.214). When combined birds were divided based on whether they were conditioned towards or against initial side preferences p values were not significant (against: p = 0.117; towards: p = 0.650).

4. Discussion

In this study, a majority of fall-condition female starlings (i.e., 17 of 24 birds) demonstrated an aversion for a place that had been paired previously with playback of either male fall or spring, courtship song. This suggests that hearing song induced a negative affective state in most females. In contrast, past research shows that female starlings in spring condition that possess nest sites develop positive CPPs in response to male courtship song [3]. When combined with the study on spring condition females, the present findings support the hypothesis that a reduction in the reward value of courtship song in fall may adaptively prevent females from breeding under conditions in which they are unlikely to succeed (i.e., at the onset of harsh environmental conditions in fall). This interpretation is also supported by prior research showing that male courtship song is neutral or aversive to females in spring that lack nesting sites [3]. Because females require a nesting site to breed, it appears that, similar to what we observe here for fall condition females, a reduction in the reward value of courtship song in spring birds may adaptively prevent breeding in birds that are unlikely to succeed.

In this study, females on average also demonstrated an aversion to a place that had been paired with fall song. This finding fails to support the hypothesis that an increase in the reward value of fall song may function to maintain overwintering flocks. Although past studies link fall song to a positive affective state, this state is tightly coupled to producing, rather than hearing song [9, 41, 45]. It may thus be that vocal-social interactions in flocks are rewarding, but hearing song alone is not. Our data strongly support this conclusion; however, a few studies suggest that other factors must be considered. For example, a female’s familiarity and past experience with males may impact the reward or incentive value of male song [46–48]. In the case of starlings specifically, past studies show that female starlings in overwintering flocks share song types with close neighbors [49], which suggests that neighbor song may be more biologically relevant than the unfamiliar song stimuli that we used in this study.

4.1. Initial stimulus preferences and CPP

Past CPP studies show that whether an animal is conditioned towards or against its initial stimulus/compartment preference (as determined during baseline testing) can influence the development of place preferences [50–53]. Consistent with this possibility, female starlings demonstrated a significant song-induced place aversion only when song was paired with the CPP compartment that was initially preferred (as measured during habituation). This finding is in agreement with a study that showed that the opioid leu-enkephalin induced a positive CPP in mice trained against initial preference but a negative CPP in mice trained towards the initial preference [51]. Additional past studies show that many drugs produce positive CPPs in rats trained against the initial compartment preference but no effects in rats trained towards the initial preference (reviewed in [50]).

A common explanation for the detection of a positive CPP in animals conditioned against the initial compartment preference is that pairing a presumably rewarding stimulus with an initially non-preferred stimulus simply provides greater opportunity for observing an effect (i.e., it avoids ceiling effects) [53]. In the case of the present study, our results suggest that for fall-condition female starlings, rather than inducing reward, hearing song from unfamiliar male starlings is aversive, thus flipping this logic. Specifically, it may be that pairing a presumably aversive stimulus with an initially preferred stimulus provides greater opportunity for observing an effect because it avoids floor effects. That is, it is possible that birds conditioned to the previously non-preferred compartment cannot show a decrease in CPP simply because conditioning cannot significantly decrease time spent in that compartment of the CPP apparatus. A biased design, in which animals are only conditioned against initial stimulus preferences, has been proposed as more sensitive for detecting reward when sample sizes are small [53], and it has been suggested that animals showing a greater initial aversion to a place are more likely to develop a CPP [54]. Our study suggests that in the case of aversive stimuli, it may be that conditioning animals towards initial stimulus preferences may be more sensitive for detecting aversion. Future studies that include additional control conditions (e.g., conditions in which birds are exposed to silence in both compartments of the apparatus) are needed to explore this possibility.

4.2. Patterns of gene expression in mPOA relate to song-induced CPP

Although on average females developed an aversion to a place that had been paired with male song, there was a high degree of individual variability. We examined relationships between CB₁ and D₁ receptor gene expression in the brain to provide insight into the possibility that constitutive individual differences in mRNA and subsequent protein synthesis for these receptors may influence the affective state induced by song. These relationships are correlational. Thus causal relationships are not known and third variables must be considered. However, given that interactions between CB₁ and D₁ receptors have been found to underlie affective state we propose that our findings may be revealing causal roles for these receptors in the affective state induced by song, but this must be tested in future studies.

CB₁ expression in mPOA correlated positively with the CPP measure of song-induced affective state, and similar positive relationships were found for both females conditioned to spring song and females conditioned to fall song (i.e., slopes did not differ significantly). This suggests that this relationship may reflect a role for CB₁ in mPOA related to responses to song that does not depend on the biological relevance of the stimulus; however, this interpretation is complicated by the finding that when spring– and fall-induced CPP were analyzed separately correlations were not significant. Although a role for CB₁ receptors specifically in mPOA in affective state has not been examined in prior studies, this result is consistent with a study showing that another cannabinoid marker (i.e., the endocannabinoid transporter FABP7) in mPOA relates positively to CPP associated with the production of fall song in male starlings [28]. The positive relationships between CB₁ mRNA in mPOA and CPP associated with both hearing and producing song suggest a potential role for endocannabinoids in mPOA in the affective state related to both sending and receiving communication signals that warrants additional study. The possible role of endocannabinoids in the mPOA in modulating affective state in additional contexts in other vertebrates also warrants investigation.

D₁ receptor expression in the mPOA in the fall-condition female brain was selectively associated with fall song-induced CPP. No significant correlation was identified in birds conditioned to spring song. This suggests that in fall-condition females, D₁ receptors in mPOA may selectively modify female responses to the sensory stimulus that is most relevant (i.e., fall song). The negative correlation between D₁ mRNA and fall-song induced CPP may indicate an inhibitory role for D₁ receptor stimulation in mPOA in female responses to fall song. Specifically, if dopamine acts at D₁ receptors to inhibit female responses to song, then it could be that lower levels of D₁ mRNA expression in mPOA results in fewer D₁ receptors at the protein level. This would then reduce tissue sensitivity to inhibitory effects of dopamine. This idea is supported by a study in breeding condition female starlings in which peripheral D₁ receptor agonist treatment inhibited, whereas antagonist treatment stimulated, female responses to nest boxes playing male starling song [55]. Other studies demonstrate that optimal levels of D₁ stimulation in mPOA are needed to facilitate behavior (e.g., male sexual behavior, including courtship song in male starlings [42, 56]), and it is possible that birds in fall condition that find song to be less aversive release higher levels of dopamine in mPOA compared to birds that find song to be more aversive, which results in receptor down-regulation to optimize dopamine activity to modulate song-induced affect. Site-directed gene or pharmacological manipulations are needed to address the causal nature of these relationships.

In conditional knockout mice, the absence of CB₁ receptors specifically in neurons containing D₁ receptors resulted in increases in social anxiety and fear (i.e., increases in negative affective states) [26]. Based on these findings, if the mPOA is a site in which CB₁ and D₁ receptors act on the same neurons to modify affective state, then we predicted that a higher CB₁ to D₁ (which may reflect more CB₁ receptors on D₁ containing neurons) would correspond to a more positive CPP score; whereas a lower ratio would correspond to a more aversive state. Consistent with this prediction, the CB₁ to D₁ ratio in mPOA correlated positively with fall song-induced CPP. This relationship was not observed for females conditioned to spring song, suggesting that interactions between D₁ and CB₁ receptors in mPOA may selectively underlie responses to the sensory stimulus (i.e., fall song) that is most relevant to fall-condition females. Future studies are now needed to determine the degree to which CB₁ and D₁ receptors co-localize to the same neurons and to test experimentally the effects on song-induced CPP of manipulating CB₁ and D₁ receptor expression.

4.3. Patterns of gene expression in VMH and song-induced CPP

Measures of CB₁ and D₁ receptor mRNA in VMH did not correlate significantly with either spring or fall song-induced CPP in fall condition females. However, the CB₁ to D₁ ratio correlated positively with song-induced CPP, with similar relationships (i.e., similar slopes) observed for females conditioned to fall and females conditioned to spring song. Although this correlation may reflect a role for CB₁ / D₁ receptor interactions in VMH in modifying song-induced affect, the effect was dependent upon a single data point. Furthermore, when females played spring and fall song were considered separately correlations were not significant. Without additional support, it is necessary to interpret this finding with caution.

4.4. Interpretational considerations

We interpret our results to suggest that high CB₁ and D₁ mRNA expression reflects high receptor protein synthesis and increased tissue sensitivity to cannabinoids and dopamine. Given that the mRNA measurements in our study relate to CPP scores that were induced by song played approximately 48 hours before neural tissue was collected, we do not suggest that our measurements reflect immediate song-induced changes in mRNA, which is supported by at least one study that demonstrated that changes in CB₁ mRNA in the fish brain that can be detected 1 and 4 hours after presentation of a stressor are no longer present 24 hours later [57]. Rather, we propose that our data reflect relatively stable individual differences in CB₁ and D₁ expression and behavior in fall condition females, which would suggest that these differences may in part explain an individual’s stable propensity to respond to song. However one important caveat is that we do not have information about translational and post-translational regulation of CB₁ or D₁ receptors. Future studies using measures and manipulations of CB₁ and D₁ receptors are now needed.

4.5. Conclusions

Past studies have examined differential neural responses to male courtship song in breeding and non-breeding condition female songbirds (i.e., females with high versus low concentrations of estradiol) (e.g., [55, 58–61]). In general, these studies indicate that auditory regions as well as brain regions involved in social motivation are more responsive to male courtship song in females with high estradiol. Our study in female starlings in a non-breeding state, which is characterized by low estradiol [62], indicates that these changes are accompanied by a reduction in the reward value of hearing male courtship song that may be mediated in part by dopamine and cannabinoid activity in the mPOA.

Highlights.

• Fall-condition female starlings developed an aversion for a place paired with male song

• Song-induced aversion was stronger in birds conditioned towards an initial preference

• D₁ receptor mRNA in mPOA correlated negatively with fall song-induced affect

• CB₁ mRNA in mPOA correlated positively with spring and fall song-induced affect

• D₁ and CB₁ receptors may adaptively modify affective responses to song