Sex-Specific Role for Dopamine Receptor D2 in Dorsal Raphe Serotonergic Neuron Modulation of Defensive Acoustic Startle and Dominance Behavior

Visual Abstract

While Drd2 is expressed in many cell types throughout the midbrain and basal forebrain, expression in serotonergic neurons is restricted to a small subset of cells resident in the dorsal raphe (DR) nucleus. In these serotonergic neurons, Drd2 expression initiates around adolescence and continues through adulthood, at which point, Drd2 transcripts are the major DA receptor mRNA detected (Niederkofler et al., 2016). Thus, Drd2-Pet1 neurons come under DRD2 and presumably DA regulation during the developmental transition to sexual maturity. Drd2-Pet1 neurons project to brain regions involved in sensory processing, defensive, and mating behaviors including auditory brainstem regions and the sexually dimorphic medial preoptic area (mPOA; Niederkofler et al., 2016). These findings led us to hypothesize that DRD2 signaling in Drd2-Pet1 neurons contributes to social and sensory alertness and defensive behavior in a sex-specific manner.
Indeed, serotonergic and dopaminergic perturbations affect social and defensive behaviors differently in male versus female rodents. Decreases in serotonergic tone associate with increased levels of aggression in males (Brown et al., 1982;Hendricks et al., 2003;Yu et al., 2014;Niederkofler et al., 2016). By contrast, lesions of the serotonergic DR in female rats decreased maternal aggression (Holschbach et al., 2018), while DR serotonergic neuron activity in female, but not male, hamsters associates with social dominance (Terranova et al., 2016). The acoustic startle reflex (ASR), an evolutionarily-conserved, defensive reflex to loud, potentially threatening stimuli (Davis et al., 1982), also shows sex-specific differences within the context of altered 5-HT levels. Reduction in 5-HT levels enhanced ASR in female but not male rats (Pettersson et al., 2016). With respect to DA, deletion of the DA re-uptake transporter gene (Dat) altered ASR only in male mice (Ralph et al., 2001). Genetic removal of the soluble form of catechol-O-methyltransferase (COMT), important for degradation of DA, enhanced the ASR and dominance behaviors in both sexes, but ASR especially in males (Tammimäki et al., 2010). Thus, serotonergic and dopaminergic neuronal systems influence social behaviors and sensory processing in sex-specific ways.
Here, we queried whether Drd2 conditional deletion in serotonergic neurons would alter aggression and social dominance behavior in males. Further, we sought to examine the role of Drd2 expression in serotonergic neurons in females with the hypothesis that other sensory or defensive behaviors would be affected, given typical lack of aggression in female mice (Lonstein and Gammie, 2002). We undertook a phenotypic analysis of mice in which we engineered Drd2 gene deletion selectively in ePet1-creexpressing serotonergic neurons (Drd2 Pet1-CKO mice). Here, we report that Drd2 Pet1-CKO males exhibited increased social dominance whereas females displayed a robust decrease in ASR. We also investigated sex differences in Drd2-Pet1 neurons at the molecular, cellular, and circuit levels, identifying differences in candidate mRNA levels, electrophysiological properties, and relative distribution densities of axonal collaterals.

Ethical approval
All experimental protocols were approved by Harvard University Institutional Animal Care and Use Committees (IACUC) and were in accordance with the animal care guidelines of the National Institutes of Health.

Image collection
All images were acquired on a Nikon Ti inverted spinning disk confocal microscope with 488-, 561-, 647-nm laser lines and Andor Zyla 4.2 Plus sCMOS monochrome camera. Images were acquired with Nikon Elements Acquisition software AR 5.02. For RNA quantification and Drd2 Pet1-CKO validation experiments, four images were taken of brain slices containing the DR: the first directly ventral to the aqueduct then one field of view below and to the left and right to capture each lateral wing.

FISH quantification
Quantification was conducted blind to sex and genotype. For Drd2 Pet1-CKO validation, all Pet11 (serotonergic) neurons within each image were identified, then the viewer outlined the DAPI-stained nuclei of each Pet11 neuron and scored the presence of Drd2 puncta as "positive" (having puncta) or "negative" (no puncta).The total number of Drd21 Pet11 neurons was then divided by the total number of Pet11 neurons to yield the "% Drd21Pet11 neurons." For quantification of Dmd, Drd2, Gad2, and Serpini1 manual counting of each mRNA punctum per cell was conducted by a trained viewer. All cells counted fit the criteria of GFP1 with a DAPI1 nucleus. The viewer outlined the GFP1 cell body in FIJI (https://Fiji.sc/; Schindelin et al., 2012) while only viewing that channel and then counted the number of distinct RNA puncta within that cell outline. Brain sections sampled were from five males and five female animals.
For quantification of Drd2-Exon7/8 and Gad2 puncta in Drd2 Pet1-CKO tissue, DR sections corresponded to interaural À0.80 to 1.04 mm and bregma À4.60 to -4.84 mm based on DAPI staining and anatomic landmarks (Franklin and Paxinos, 2008), where Drd2-Pet1 neurons are most enriched. A series of custom FIJI scripts and a CellProfiler (McQuin et al., 2018) pipeline were used to process and analyze confocal images of RNAscope FISH signal in a semi-automatic manner. Analysis was performed in 2D on maximum intensity projections of 6-mm-thick z-stacks. First, a (step 1) preprocessing FIJI script separated channels and preprocessed them for (step 2) CellProfiler to use as input to segment nuclei. The DAPI-stained channel was preprocessed by a Gaussian blur with a diameter of 18 before segmenting with the IdentifyPrimaryObjects module with a diameter range 30-100 pixels using a minimum cross entropy global thresholding strategy. Objects outside of the diameter range or those on the edges were excluded. A threshold smoothing scale of 1.3488 was used and the image was automatically declumped based on intensity values. Finally, holes were filled in the resulting label map image, which was exported for use in FIJI (step 3). In FIJI, the user manually excluded misidentified objects or added additional nuclei that were missed by the automatic detection pipeline. A highly similar script was recently published (Okaty et al., 2020), though this current script performs additional difference of Gaussian (Marr and Hildreth, 1980) based filtering for each FISH channel. For each FISH probe, after background subtraction with a rolling ball radius of 50 pixels, the image was duplicated and a Gaussian blur was performed at two different s levels, one which obscured small background pixels but preserved mRNA puncta, and a more extreme blur that only retained larger diffuse background puncta. The difference of these two images was then calculated and puncta localized using the Find Maxima function. To find appropriate settings for each FISH channel, we compared the performance of several sets of parameters to automatically detect puncta versus a hand count of puncta. We were able to achieve excellent concordance between the hand count and automatic puncta detection. Table 1 summarizes our settings and performance in a linear regression against the hand count for each FISH probe (statistics calculated in GraphPad Prism v8.4.3 and Microsoft Excel v2002).

Behavioral assays
All assays, except the resident-intruder assay, were conducted in an initial cohort of 15 control (eight males, seven females) and 11 Drd2 Pet1-CKO (six males, five females) mice. All behavioral assays were conducted at postnatal day (P)90 or later. The run order for the initial cohort was open field, elevated plus maze, tail suspension test, forced swim test, social interaction, acoustic startle response, prepulse inhibition of acoustic startle, water T-maze, contextual fear conditioning, tube test of social dominance and rotarod. An additional cohort of 16 controls (seven males, nine females) Drd2 Pet1-CKO (six males, 10 females) was run for acoustic startle response. Resident-intruder assay of aggression was conducted in three separate cohorts of mice totaling 24 control and 26 Drd2 Pet1-CKO males. The tube test of social dominance was run in the initial cohort and in the second (eight control and Drd2 Pet1-CKO males) and third (11 control and Drd2 Pet1-CKO males) aggression cohorts for a total of 24 control and Drd2 Pet1-CKO males and a separate cohort of 16 control and 18 Drd2 Pet1-CKO females. The rotarod assay was also repeated in a separate cohort of males (seven controls, six Drd2 Pet1-CKO ). Experiments were conducted between zeitgeber time (ZT)6 and ZT10, with interspersion of control and experimental animals, and assays were run and analyzed by a trained experimenter blinded to genotype. The open field test, elevated plus maze, tail suspension test, forced swim test, social interaction, prepulse inhibition of acoustic startle, water T-maze, and contextual fear conditioning were performed as previously described (Niederkofler et al., 2016). All other behavioral assays are described in detail below.

Rotarod
The rotarod apparatus (Stoelting; Ugo Basile Apparatus) contains a rotating rod set to an accelerating speed. Mice are placed onto the rod and rotation of the rod begins. When a mouse loses its balance and falls, the apparatus automatically stops and measures the latency and rotating speed at which the mouse fell. Training consisted of exposing the mice to the apparatus for 5 min at a constant speed of 4 rpm. Mice that fall during the training session are placed back on the apparatus until the training session time has elapsed. An hour following the training session, mice are placed back on the rod for a 2-min session in which speed increases steadily over 2 min from 5 to 40 rpm. If a mouse does not fall during the 2 min, the trial ends at 2 min. Each animal was tested over 3 d and the latency to fall was averaged for each mouse. This assay was conducted in 21 control mice (14 males, seven females) and 18 Drd2 Pet1-CKO (13 males, five females).

Acoustic startle response
Mice were placed in a perforated holder (acrylic cylinder with 3.2-cm internal diameter) that allowed movement to be monitored. Animal holders were placed on top of a transducer platform, measuring the active response to both weak and startle stimuli, adjacent to a speaker, within an individual acoustic chamber (Med Associates). Each session consisted of a 5 min acclimation period followed by 10 blocks of 11 trials each with white noise acoustic stimuli (20-120 dB). Each startle stimulus (20-120 dB, in 10-dB increments) was played once per block, in a quasi-random order with a variable intertrial interval of 10-20 s (average of 15 s). The duration of the stimulus was 40 ms. Responses were recorded for 150 ms from startle onset and are sampled every ms. Mice were placed back into the home cage immediately after testing. Males and female were run on different days. This assay was conducted in 30 control mice (14 males, 16 females) and 28 Drd2 Pet1-CKO (13 males, 15 females), as two separate cohorts per sex.

Tube test of social dominance
Two age-matched (;P90), weight-matched mice of the same sex are introduced into opposite ends of a clear PVC tube (30.5 cm in length with an internal diameter of 2.5 cm) allowing them to interact in the middle but not pass each other within the tube. The subordinate mouse will back out allowing the dominant mouse to pass through (Lindzey et al., 1961). For each pair, five consecutive trials were run with a maximum time of 2 min per trial. Trials ended when one mouse backed out of the tube such that all four limbs are outside of the tube which was then recorded as a "backout" for that mouse. Matches lasting .2 min were excluded from analysis and scored as a draw. Side of introduction to the tube were alternated between trials and the tube was cleaned with ethanol between each trial. Opponents were from different litters and had never been housed together. This assay was conducted in 24 Drd2 Pet1-CKO males versus 24 control males and 23 Drd2 Pet1-CKO females versus 23 control females, conducted across three cohorts of animals.

Resident-intruder assay
Drd2 Pet1-CKO or control mice were group-housed with male siblings until adulthood (P90) when they were singlehoused for one night in the test cage to establish territorial residency. On day 1, a five-week-old Swiss Webster (Charles River) male, the "intruder," was introduced to the cage divided with a clear perforated divider for 5 min. After 5 min, the perforated divider was removed, and the mice could interact for 5 min, in which the encounter was video recorded. Number of attack bites were counted by a trained, blinded viewer. The intruder mouse was introduced for 3 d to obtain an average number of attack bites per day. The intruder mouse had a lower body weight than the resident male. This assay was conducted only in males, as female laboratory-reared mice do not display territorial aggression (Palanza, 2001;Lonstein and Gammie, 2002) This assay was conducted in 26 Drd2 Pet1-CKO and 24 controls.

Auditory brainstem response (ABR)
ABRs were recorded in a separate cohort of adult mice (males: 10 control and seven Drd2 Pet1-CKO ; females: eight control and seven Drd2 Pet1-CKO ) aged P71-P102 to correspond to the age of animals in other assays. ABRs were conducted similar to (Maison et al., 2013). Mice were anesthetized by intraperitoneal injection of ketamine (100 mg/kg) and xylazine (7.5 mg/kg) and placed in a soundproof chamber on a heating pad. Acoustic stimuli were delivered using EPL Cochlear Function Test Suite (CFTS) software and analyzed using ABR peak analysis software [1.1.1.9, Massachusetts Eye and Ear (MEE)]. All ABR thresholds, amplitudes, and latencies were read by an investigator blind to mouse genotype.

Electrophysiology
Slice preparation and whole-cell patch-clamp recordings were conducted as previously described (Rood et al., 2014;Niederkofler et al., 2016). Briefly, to assess membrane and action potential (AP) characteristics a protocol of repeated sweeps of 500-ms current injections stepping in 20-pA steps from À80 to 180 pA was administered to cells in current clamp. Data were analyzed using Clampfit (Molecular Devices). Some cells included in cell property analyses were also used to generate data on the function of DRD2 receptors in the DR (Niederkofler et al., 2016). However, the intrinsic cell properties data we present in this article have not been previously published and include cells not part of the Niederkofler et al. (2016) dataset.

Projection mapping
Brain tissue from six females and five males from different litters, but with a female and male from the same litter where possible, were collected at P90 and processed as previously described (Niederkofler et al., 2016). Target region identification was based on anatomic landmarks identified by DAPI staining, anti-choline acetyltransferase (goat polyclonal anti-ChAT,1:500, AB144P; EMD234 Millipore) staining, and/or anti-tyrosine hydroxylase (rabbit anti-TH, 1:1000, AB152, Millipore) staining. Staining and imaging protocols were identical among the eleven samples analyzed.

Quantification of target innervation
Target innervation was quantified in a similar manner to (Niederkofler et al., 2016). Briefly, image stacks were acquired bilaterally per brain region analyzed for each animal using a Nikon Ti inverted spinning disk microscope with a Plan Fluor 40Â/1.3 Oil DIC H/N2 objective, 488-, 561-, and 647-nm laser lines, and Andor Zyla 4.2 Plus sCMOS monochrome camera. Images were acquired with Nikon Elements Acquisition software AR 5.02. Image stacks (.nd2 files) were imported to FIJI for analysis of axon projection area. Each stack contained 21 optical slices of 0.3 mm. Innervation density was quantified by a FIJI macro, such that all images, were treated identically, including background subtraction, thresholding and particle counting as described in (Niederkofler et al., 2016). We then divided the total area occupied by the projection signal by the total area of the 21 optical slices to obtain the percent area occupied by projection signal. This was then averaged within images of the same brain region across male or female samples. Brain regions analyzed were either those previously described to be innervated by Drd2-Pet1 neurons in males only (Niederkofler et al., 2016) or those involved in auditory processing and ASR.

Statistical analyses
Data are presented as mean 6 SEM. Statistical analyses were conducted in GraphPad Prism version 8.1. Statistical significance was determined by unpaired t test between control versus Drd2 Pet1-CKO groups or male versus female groups except where noted: open field, forced swim test, acoustic startle response, and ABR statistical significance was determined using two-way ANOVA. For the resident-intruder assay, the tube test of social dominance, and rotarod, statistical significance was determined using the non-parametric Mann-Whitney U test. A result was considered significant if p , 0.05. Detailed statistical results are reported in Table 2.
To query the behavioral requirement for Drd2 gene expression in Drd2-Pet1 neurons, we deployed the ePet-cre driver (Scott et al., 2005) to delete floxed Drd2 gene sequences (Bello et al., 2011), creating a functional null Drd2 allele selectively in Pet1 neurons (Fig. 1D), and then subjected these Drd2 Pet1-CKO mice to behavioral phenotyping. Cre-negative, Drd2 flox/flox littermates served as controls. To confirm loss of Drd2 gene expression in Pet1 neurons, we analyzed Drd2 Pet1-CKO and control Drd2 flox/flox brain tissue sections by mRNA in situ hybridization using a probe designed to detect exon 2-containing Drd2 mRNA, as exon 2 was the floxed gene portion to be excised by Cre recombination; concomitant identification of serotonergic neurons was by detection of Pet1 transcripts (Fig. 1F,G). Robust loss of Drd2 expression was observed in serotonergic neurons in both male and female mice [15.23 6 2.41% of Pet11 neurons in the DR express Drd2 transcripts in controls (n = 6), consistent with prior findings, compared with 3.87 6 0.73% in Drd2 Pet1-CKO s (n = 6), p = 0.0011, unpaired t test; Fig. 1E]. The few residual Pet11 cells harboring Drd2 transcripts likely reflects a limitation in cell capture by the ePet-cre driver. Reliable immunodetection to confirm the expected parallel loss of DRD2 protein in PET1 cells remains unavailable.
Behavioral assessments in Drd2 Pet1-CKO mice and the detection of sex-specific sensory, defensive, and social behaviors Having validated effective loss of Drd2 expression specific to Pet1 neurons in Drd2 Pet1-CKO mice, next, we screened these mice for behavioral alterations in comparison to sibling control Drd2 flox/flox (Cre-negative) mice. Locomotor behaviors were explored first because they are known to be influenced by serotonergic and dopaminergic manipulations (Baik et al., 1995;Gainetdinov et al., 1999;Holmes et al., 2003;Seo et al., 2019), and because motor alterations can affect performance in and interpretation of subsequent behavioral assays. Notably, we found no differences between Drd2 Pet1-CKO versus control mice (males or females) in the locomotor behaviors reflected in the open field and rotarod tests, such as distance traversed ( Fig. 2A) and location within the field (Fig.  2B), vertical rearing, length of time on the rotating rod (Fig. 2C), which reflects balance, coordination, physical conditioning, and motor-planning. Next, we explored measures of depression-like and anxiety-like behaviors, as they are altered in various 5-HT-pathway or DA-pathway mouse models and pharmacological manipulation of these neurotransmitter systems show positive clinical effect. (Lucki, 1998;Hendricks et al., 2003; Holmes et al.,  2003; Grace, 2016). We observed no differences in performance in the elevated plus maze (Fig. 2D), tail suspension test (Fig. 2E), or forced swim test (Fig. 2F) in Drd2 Pet1-CKO males and females compared with littermate controls. Additionally, contextual fear conditioning (Fig. 2G) and water T-maze acquisition and reversal (Fig. 2H) were not affected, suggesting no impairment of memory and learning in Drd2 Pet1-CKO mice.
Because the serotonergic and dopaminergic systems are implicated in modulating the ASR (Davis and Aghajanian, 1976;Davis et al., 1980;Davis, 1999, 2000a,b), we explored that next. The ASR is an evolutionarily conserved reflex involving rapid contraction of facial and skeletal muscles into a protective posture in response to a loud, threatening stimulus. We hypothesized that Drd2-Pet1 neurons modulate this response, given their dense projections to auditory brain regions (Niederkofler et al., 2016) and the observation that following acoustic startle, the activity of certain serotonergic neurons increases in the lateral wings of the DR (Spannuth et al., 2011), a location in which we find Drd2-Pet1 neurons. We measured startle responses to weak and startling stimuli ranging from 20 to 120 dB presented in a randomized order (Fig. 3A,B). Female Drd2 Pet1-CKO mice showed a significant decrease in ASR magnitude in response to startle stimuli (n = 15 Drd2 Pet1-CKO , n = 16 control littermates, p = 0.0011, two-way ANOVA; Fig. 3F). By contrast, the male Drd2 Pet1-CKO cohort was indistinguishable from their male littermate controls (n = 13 Drd2 Pet1-CKO , n = 14 control littermates, p = 0.7745, twoway ANOVA; Fig. 3C). To prevent habituation to the startle stimuli, the different stimulus intensities were presented in a quasi-random order with varied intertrial intervals (see Materials and Methods), and indeed, startle responses in late as compared with early trials were indistinguishable (shown at 110 dB, trial number is not significantly correlated with startle magnitude, males: controls, r = À0.1950 and Drd2 Pet1-CKO , r = 0.1360; females: controls, r = 0.1171 , except for C where, n = 21 control mice (14 males, 7 females) and 18 Drd2 Pet1-CKO (13 males, 5 females). Each symbol represents one animal, error bars represent SEM. No significant differences (p . 0.05) between Drd2 Pet1-CKO and controls were observed. No sex-specific (male vs female) phenotypes observed. For assay details, see Materials and Methods; for statistical details, see Table 2. . Drd2 Pet1-CKO females, but not males, display attenuated acoustic startle responses (ASR). A, Schematic of ASR experimental design. After an initial 5-min acclimation, mice are exposed to 10 blocks of 11 trials of auditory stimuli ranging from 20 to 120 dB in quasi-randomized order with a 10-to 20-s intertrial interval (ITI). B, Schematic of ASR measurement apparatus, mouse is placed in a perforated holding chamber atop transducer platform adjacent to speaker (for detailed description, see Materials and Methods). C, F, Averaged ASR magnitudes (mean 6 SEM) across increasing stimulus intensities in (C) male Drd2 Pet1-CKO (blue, n = 13) and controls (black, n = 14), no significant difference, p = 0.7745, two-way ANOVA and (F) female Drd2 Pet1-CKO (blue, n = 15) and controls (black, n = 16), Drd2 Pet1-CKO females display significantly attenuated ASR, p = 0.0011, two-way ANOVA. D, G, Group averaged ASR for 10 trials at 110-dB stimulus in (D) males and (G) females, demonstrates no habituation to the startle stimulus; xaxis numbers refers to trial number out of 110 total trials. E, H, No significant differences in latency to startle are observed in (E) males, p = 0.1319, two-way ANOVA and (H) females, p = 0.5452, two-way ANOVA. I, J, No significant differences in prepulse inhibition of acoustic startle are observed in (I) males (n = 8 control, 6 Drd2 Pet1-CKO ), p = 0.4325, two-way ANOVA or (J) females (n = 7 control, 5 Drd2 Pet1-CKO , p = 0.4380, two-way ANOVA).
and Drd2 Pet1-CKO , r = 0.0517, Pearson correlation; Fig.  3D,G). Further, we observed no differences in latency to startle in either females or males (Fig. 3E,H). Females were of similar mass (controls: 32.117 6 3.15 g vs Drd2 Pet1-CKO : 37.2 6 2.427 g, unpaired t test, p = 0.2031) regardless of genotype, thus differences in weight and its relative impact on transduction of the startle response via the piezoelectric platform were not a confound. While Drd2 Pet1-CKO females showed diminished response magnitudes to startling acoustic stimuli, they nevertheless expressed normal acoustic prepulse inhibition (PPI) whereby even the diminished response to startling acoustic stimuli (e.g., 120-dB stimuli) was further blunted proportionately when immediately preceded by a weak, non-startling stimulus (e.g., 65-, 75-, or 85-dB stimuli; Fig. 3I,J). Thus, sensorimotor gating, as measured by acoustic PPI, appeared relatively intact; the acoustic dysfunction instead centered on the ASR itself.
ABRs were conducted in adult mice (ages P71-P102) to align with the age at which the other behavioral assays were performed. However at such ages, C57BL/6 mice, the strain background here, exhibit some age-related hearing loss at higher frequencies (Kane et al., 2012), which we saw here at 32 kHz with two control and three Drd2 Pet1-CKO females and five control and three Drd2 Pet1-CKO males. At all other tested frequencies, the ABRs were effectively normal for both genotypes, with one exception being a Drd2 Pet1-CKO female that exhibited undetectable ABRs at 5.6 kHz, but otherwise normal responses at all other frequencies tested including 32 kHz. These findings at 32 and 5.6 kHz are likely independent of the ASR phenotype observed in females because all animals had normal hearing at 8 and 16 kHz, frequencies included in the white noise startle stimulus of the ASR test.
Next, we examined social behavior in Drd2 Pet1-CKO mice using the three-chambered test of sociability (Moy et al., 2004) that measures preference to investigate a social stimulus (a novel "stranger" mouse inside a holder) as compared with an object (an empty holder). Drd2 Pet1-CKO mice showed no alterations in sociability compared with controls and both control and Drd2 Pet1-CKO spent significantly more time investigating the stranger than the object (Fig. 5A,B). Females of both genotypes displayed preference toward the social stimuli only for the first 5 min of the assay (Fig. 5B, white bars), while males displayed this preference throughout the 10-min assay. Similar sex differences in sustained preference for the social stimulus have been described in C57BL/6J mice (Netser et al., 2017).
We assayed intermale, territorial aggression in a separate cohort of mice using a resident-intruder assay. Females were not tested, as they have been shown to display low or no aggression in most forms of this assay (Palanza, 2001;Lonstein and Gammie, 2002). We observed no statistically significant difference in number of attack bites delivered to the intruder mouse by Drd2 Pet1-CKO males (n = 26) compared with number of attack bites delivered to the intruder by controls (n = 24; Drd2 Pet1-CKO : 4.076 1.50 bites, controls: 1.77 6 0.39 bites, p = 0.6649, Mann-Whitney test; Fig. 5C) noting, however, that four Drd2 Pet1-CKO males displayed high levels of aggression.
To assay social dominance, we performed the tube test, which has relevance in females as well as males (Lindzey et al., 1961;van den Berg et al., 2015;Zhou et al., 2017). Two mice are simultaneously released into opposite ends of a clear tube of sufficiently narrow diameter that prevents mice from passing by each other and instead requires that one back out for the other, more dominant "winning" mouse, to move forward (Fig. 5D). Drd2 Pet1-CKO males won a higher percentage of trials against non-sibling, weight-matched, and genetic background-matched opponent males (shown as percent of trials won, Drd2 Pet1-CKO : 65.83 6 9%, n = 24; controls: 34.17 6 9%, n = 24; p = 0.0065, Mann-Whitney test; Fig. 5E). By contrast, we observed no difference in percent of trials won by female Drd2 Pet1-CKO mice as compared with female sibling controls (Drd2 Pet1-CKO : 48.7 6 8%, n = 23; controls 51.3 6 8%, n = 23; p = 0.8123, Mann-Whitney test; Fig.  5E).

Drd2-Pet1 neurons in males versus females exhibit differences in candidate molecular and biophysical properties but not in cell number
Given these sex-specific differences in behaviors observed in Drd2 Pet1-CKO mice, next we looked for sex-specific differences in Drd2-Pet1 cellular properties, beginning with cell number. Analyzing triple transgenic Drd2-Cre; Pet1-Flpe; RC-FrePe males versus females, we found no difference in number of GFP 1 Drd2-Pet1 neurons per brain (males: 410.40 6 55.30 cells/brain, females: 313 6 87.52 cells/brain, p = 0.4304, unpaired t test; Fig. 6A). Further, in both males and females, Drd2-Pet1 neurons distributed as expected across the rostral-caudal and medial-lateral axis of the DR.
To understand whether gene expression might differ between male and female Drd2-Pet1 neurons, we examined single-cell RNA sequencing data previously analyzed for expression of serotonergic pathway genes as validation that Drd2-Pet1 cells were indeed serotonergic (Niederkofler et al., 2016). Comparison across sex, albeit lacking statistical significance given the small sample size, highlighted four genes for further evaluation, Drd2, Dmd (encoding Dystrophin, a component of protein scaffolds in the CNS; Perronnet and Vaillend, 2010), Gad2 (encoding glutamate decarboxylase 2 involved in catalyzing the production of the neurotransmitter GABA), and Serpini1 (encoding the serine protease Neuroserpin, important for synapse formation and plasticity; Galliciotti and Sonderegger, 2006). Quantitative in situ mRNA detection using dual FISH with immunodetection on tissue sections from Drd2-Cre;Pet1-Flpe;RC-FrePe mice revealed greater abundance of average Gad2 transcripts (puncta) per cell in males versus females [Gad2: 20.46 6 2.243 in males (n = 5) vs 12.20 6 2.427 in females (n = 5), p = 0.0370, unpaired t test; Fig. 6D]. There was no difference in the percentage of Drd2-Pet1 neurons expressing Gad2 in male versus female mice (Fig.  6E). No difference in soma size (GFP-stained cell body) was observed between males and females suggesting that transcript differences were not because of larger soma volume measured (Fig. 6C). No significant differences in mRNA abundance were observed between males and females for Dmd, Drd2, or Serpini1 (see Table 2).
As a first step toward understanding whether sex-specific gene expression differences observed in wild-type mice persist or are altered in Drd2 Pet1-CKO mice, we assessed Gad2 transcript levels in Drd2 Pet1-CKO cells. In these cells, the floxed exon 2 of Drd2 is excised by Cre recombination. Therefore, to identify mutant Drd2 mRNA and thus the mutant Drd2 Pet1-CKO cells, we used a multiprobe strategy involving one probe to intact downstream exons 7 and 8 (referred to here as Drd2-E7/8), another to exon 2 (referred to as Drd2-E2), and another to either cre or Pet1. We examined expression in the DR region most enriched with Drd2-Pet1 neurons. We found Drd2-E7/81 puncta in Pet11 cells in both controls and Drd2 Pet1-CKO mice, whereas Drd2-E21 puncta were detectable in control tissue but greatly reduced in Drd2 Pet1-CKO as expected given the efficiency of Cre-mediated gene deletion ( Fig. 7A; see Drd2-E2 quantification in Fig. 1E). Drd2-E7/8 puncta were detected in 35.97 6 2.403% of Pet11 cells in control mice (n = 6) compared with 36.53 6 3.621% of Pet11 cells in Drd2 Pet1-CKO mice (n = 6; p = 0.8998, unpaired t test; Fig. 7B). Similarly, in a separate experiment using an in situ probe to cre mRNA, 34.91 6 2.238% of cre1 cells expressed Drd2-E7/8 (n = 10 mice, one-way ANOVA compared with Pet1 control and Drd2 Pet1-CKO cell expression, p = 0.9051; Fig. 7B). Next, we analyzed Gad2 mRNA transcript levels in Drd2 Pet1-CKO cells (dual Drd2-E7/81 and cre1 cells) in the DR (Fig. 7C). In males, we observed 87.44 6 3.034% of Drd2 Pet1-CKO cells were Gad21, while this percentage was 75.76 6 0.5862% in females (p = 0.0157, unpaired t test; Fig. 7D). In these Drd2 Pet1-CKO cells, there were 14.25 6 1.325 transcripts per cell in males and 10.13 6 2.074 transcripts per cell in females (p = 0.1151, unpaired t test; Fig. 7E). Because of the tightly packed distribution of cells in the DR, puncta were measured only within cre1 DAPI-stained nuclei to ensure puncta were not assigned to more than one cell. The area of nuclei did not differ between males (114.9 6 3.030 mm 2 ) and females (110.9 6 1.768 mm 2 , p = 0.3497, unpaired t test; Fig. 7F). Thus, in Drd2 Pet1-CKO males as compared with Drd2 Pet1-CKO females, a greater percentage of the Drd2-Pet1 cells harbored Gad2 transcripts; of these Gad2-expressing cells, however, transcript levels were not significantly different between males versus female Drd2 Pet1-CKO mice.
To explore potential sex differences in electrophysiological properties characterizing Drd2-Pet1 neurons, we A B C D E Figure 5. Drd2 Pet1-CKO males, but not females, display increased social dominance. A, B, Three chambered social interaction assay. No significant difference in time spent investigating a stranger mouse or an empty holder for (A) males (n = 8 controls compared with 6 Drd2 Pet1-CKO , p = 0.541, unpaired t test) and (B) females (n = 7 controls compared with n = 5 Drd2 Pet1-CKO , p = 0.358, unpaired t test). Investigation time is binned into 5-min intervals where white bars indicate first 5 min of assay and colored bars indicate last 5 min of assay. As expected, mice of both genotypes spent significantly less time investigating the empty holder than the stranger mouse noting that females of both genotypes only did so during the first 5 min of the assay. C, Resident intruder assay of aggression. No significant difference in the average attack bites per day delivered to a Swiss Webster intruder mouse was observed between Drd2 Pet1-CKO males (n = 26, 4.07 6 1.50 bites) aggression levels were not significantly different from controls (n = 24, 1.77 6 0.39 bites; Mann-Whitney, two-tailed, U = 289.5, p = 0.6649). D, Schematic of tube test (for details of assay, see Materials and Methods). Schematic created with BioRender. E, Drd2 Pet1-CKO males (n = 24) demonstrate more dominance behavior than controls (n = 24) as they displayed increased winning in the tube test (controls: 34.17 6 9% wins, Drd2 Pet1-CKO : 65.83 6 9% wins, p = 0.0065, Mann-Whitney, two-tailed, U = 166). Female Drd2 Pet1-CKO (n = 23) showed no difference in social dominance compared with controls (n = 23; controls: 51.3 6 8%, Drd2 Pet1-CKO : 48.7 6 8% wins, p = 0.8123 Mann-Whitney, two-tailed, U = 253).
Differing covariance in axonal collateral densities from Drd2-Pet1 neurons directed to auditory targets in males versus females As a first step in exploring sex differences in Drd2-Pet1 neuron circuitry that may underlie the sex-specific behavioral phenotypes in Drd2 Pet1-CKO mice, we compared relative innervation density to brain regions involved in sensory processing and social behavior in male and female mice. Boutons from Drd2-Pet1 neurons were selectively marked with a Synaptophysin-GFP fusion protein using triple transgenic Drd2-Cre;Pet1-Flpe;RC-FPsit mice (Fig. 9A,B; Niederkofler et al., 2016). At P90, the same age at which the behavioral assays were conducted, we collected brain tissue and quantified projections to the cochlear nucleus complex (CNC), superior olivary complex (SOC), lateral lemniscus (LL), inferior colliculus (IC), caudal pontine reticular nucleus (PNC; critical for ASR; Davis et al., 1982), dorsal lateral geniculate nucleus (dLGN), mPOA, medial habenula (mHb), periaqueductal gray (PAG), and dorsal paragigantocellular nucleus (DPGi; shown as percentage of target area occupied by projections; Fig. 9C). We observed no significant sex differences in the cohort average for absolute innervation density to each of these 10 brain regions. However, because we observed considerable interanimal variability in bouton densities at targets, we next explored correlation of innervation density across brain regions  (Weissbourd et al., 2014). Using pairwise correlations between auditory brain regions (Fig. 9D), we constructed a correlation matrix that shows positively correlated regions in green and negatively correlated regions in black (Fig. 9E). This visualization reveals that most auditory brain regions are positively correlated in males (SOC and LL, Pearson's r = 0.89) with only the LL and cochlear nucleus being slightly negatively correlated (Pearson's r = -0.28). Interestingly, a greater number of innervated regions were negatively correlated in females, including the CNC with both the SOC and the IC (r = À0.68 and r = À0.75, respectively), as well as PNC and IC (r = À0.67). The innervation of the PNC and SOC was significantly negatively correlated (r = À0.85, p = 0.033, twotailed test). Further, we expanded analyses to include the dLGN, a region critical for visually-cued potentiation of the acoustic startle (Tischler and Davis, 1983), and found that in females innervation of the dLGN was not strongly correlated with innervation of auditory brain regions, while in males this dLGN innervation was highly negatively correlated with both the SOC (r = À0.91, p = 0.033, two-tailed test) and the IC (r = À0.91, p = 0.034, two-tailed test), indicating that Drd2-Pet1 neuron circuitry may be set up to modulate multisensory information differently in males compared with females.

GABA and 5-HT in Drd2-Pet1 neurons
Given detection of Gad2 mRNA in Drd2-Pet1 neurons, we probed for GABA versus 5-HT immunopositivity in cell soma versus axonal boutons in males versus females. Punctate GABA immunostaining was indeed detectable in some Drd2-Pet1 neuron soma (Fig. 10A) in both males and females. Yet, in all target brain regions examined, GABA was undetectable in the GFP-marked Drd2-Pet1 boutons. Shown are representative images from the SOC (Fig. 10B) and IC (Fig. 10C), noting a GABA-positive cell body in the IC (boxed) and GABA-positive staining in the corpus callosum serving as a positive control for GABA immunodetection (Fig. 10D). By contrast, 5-HT immunostaining in Drd2-Pet1 boutons was readily detectable (representative images from the SOC and IC; Fig. 9B).

Discussion Strategy
We hypothesized that loss of Drd2 gene expression and associated DRD2 signaling normally observed in certain DR Pet1 1 serotonergic neurons (Drd2-Pet1 neurons) could impair sensory, social, and/or defensive behaviors. We used the transgenic driver ePet-cre to delete functionally critical Drd2 gene sequences selectively in serotonergic neurons, thereby abolishing transcript and DRD2 protein function, which would normally initiate in Pet1 cells during adolescence. We validated these Drd2 Pet1-CKO mice and examined behavioral responses. Further, we explored Drd2-Pet1 neurons themselves.

Main findings
Key findings include the following.
(1) Sex-specific behavioral alterations were observed in Drd2 Pet1-CKO mice. Females showed a dramatic diminution in the protective, defensive ASR as compared with Drd2 flox/flox controls, while no differences were observed in males.
(2) Drd2 Pet1-CKO males, but not females, showed increased winning in the tube test of social dominance against sex-matched and age-matched controls.
(3) No differences were observed in ABRs, in PPI of acoustic startle, locomotion, cognition, nor various affective behaviors. (4) No sex-specific differences were found in Drd2-Pet1 neuron number, soma distribution, nor in the set of efferent targets; however, within-animal correlations between efferent densities across target brain regions suggest differences by sex, thus hinting at sex-specific structural differences in Drd2-Pet1 neuronal circuitry.
(5) Drd2-Pet1 cells in males as compared with females showed longer AP durations and higher levels of Gad2 transcripts (important for GABA synthesis); Drd2 Pet1-CKO cells did not show a sex specific difference in Gad2 transcript levels, but the percentage of Drd2-Pet1 cells that were Gad2 1 in Drd2 Pet1-CKO males was slightly higher than in Drd2 Pet1-CKO females. These findings, coupled with our prior work (Niederkofler et al., 2016) implicating Drd2-Pet1 neurons in setting levels of defensive aggressive and exploratory behaviors in male mice, suggest that Drd2-Pet1 neurons may serve as a specialized neuromodulatory interface whereby DRD2 Figure 9. Drd2-Pet1 neuron axon terminals target brain regions involved in sensory processing and defensive behavior in both male and female mice. A, Intersectional genetic strategy: expression of Drd2-Cre and Pet1-Flpe transgenes results in dual recombination of intersectional allele, RC-FPSit, to label boutons of Drd2-Pet1 neurons with Synaptophysin-GFP. B, Representative images of Drd2-Pet1 boutons in the SOC and IC. GFP1 (green, marked with arrows) boutons co-localize with 5-HT (magenta) staining. DAPIstained nuclei shown in blue. Scale bar: 25 mm. C, Quantification of the percent target area occupied by projections for all ten brain regions examined (for quantification protocol, see Materials and Methods). Target areas analyzed include brain regions involved in auditory processing and social behavior including the CNC, SOC, LL, IC, PNC, mHb, dLGN, mPOA, and PAG. The DPGi was also examined. No significant differences in projection area innervation were observed between males (n = 5) and females (n = 6). D, Example graph showing correlation between innervation density of auditory brain regions differs in males compared with females. Each dot represents one animal. Values are shown as Pearson's correlation coefficient (r), and * indicates p , 0.5 in a two-tailed test. E, Pairwise correlations shown for male and female innervation density in auditory brain regions. Heatmaps represent high correlation (green) and low correlation (black) between CNC, SOC, LL, IC, and PNC.
signaling alters serotonergic neuronal activity to shape defensive, protective, and dominance behaviors in a sex-specific manner.
Protective ASR diminished in Drd2 Pet1-CKO females Defensive posturing in millisecond response to abrupt noise, be it a predator or other potential hazard, is a crucial evolutionarily conserved protective mechanism. Loss or blunting of this reflex can result in life-threatening exposure, while excessive enhancement can drive unnecessary, debilitating responses that preclude normal functioning. Thus, "tuning" of the ASR setpoint to social and environmental circumstances is likely critical for species survival and well-being. The observed ASR attenuation in female Drd2 Pet1-CKO mice suggests that Drd2-Pet1 neurons and the regulation of their activity cell autonomously by DRD2 comprises a critical modulatory node for ASR in females. Further, this node appears separate functionally from that involved in acoustic sensorimotor gating, given that acoustic PPI appeared intact in Drd2 Pet1-CKO females, and from hearing, given that ABRs were indistinguishable from controls. Thus, DRD2 signaling in Drd2-Pet1 neurons forms a functional circuit node specialized in female mice to influence startle to acoustic stimuli.
In rats, reduction of 5-HT through synthesis inhibition increases ASR in females, but not males (Pettersson et al., 2016). Predicted reciprocally is that elevated 5-HT levels might blunt ASR in females. Relating this to our findings, it is possible that Drd2-Pet1 neurons are more excitable in the absence of DRD2-mediated inhibition, resulting in increased 5-HT release, perhaps explaining the observed ASR blunting. In wild-type mice, this would predict that under conditions of DA elevation, for example through local DR DA neuron activity associated with arousal and vigilance (Cho et al., 2017), Drd2-Pet1 neuron activity would be inhibited, reducing 5-HT release and thereby tuning a more sensitive ASR, conferring a protective advantage.
The ASR circuit follows from cochlea to CNC to PNC to spinal motoneurons (Davis et al., 1982;Koch et al., 1992), and receives inputs from auditory centers such as the SOC, IC, and SC (Lauer et al., 2017). Drd2-Pet1 neurons innervate each of these areas and the PNC, and thus may impart modulation at multiple levels.

Tube test wins increased in Drd2 Pet1-CKO males
The increased winning by Drd2 Pet1-CKO males in the tube test suggests that loss of DRD2 results in an increase in or favoring of dominance behaviors, at least under these forced, one-on-one interaction conditions. We did not observe significant differences in levels of aggressive attack behaviors by Drd2 Pet1-CKO males in a resident-intruder assay. Together, these findings suggest that in wild-type mice, DRD2 signaling in Drd2-Pet1 neurons contributes to tempering certain dominance behaviors under particular conditions. Understanding how the present results align with our prior work remains a pursuit. In earlier studies using a resident-intruder assay, we observed an increase in various aggressive behaviors in mice in which Drd2-Pet1 neurons were constitutively silenced, which suggested to us that Drd2-Pet1 neuron excitation and neurotransmitter release would normally temper such behaviors (i.e., favor nonconfrontational, even submissive behaviors). Because canonical DRD2 signaling is inhibitory and, as well, appears largely inhibitory in Drd2-Pet1 neurons in slice, we predicted that loss of DRD2 signaling would enhance Drd2-Pet1 cell excitability and neurotransmitter release probability, and thus would suppress or at least not enhance dominance behaviors. Yet Drd2 Pet1-CKO males exhibited enhanced winning in the tube test. Perhaps DRD2 signaling in Drd2-Pet1 neurons results in cellular activity changes that ultimately lead to a tempering of one-onone social dominance under some conditions, while extreme, constitutive Drd2-Pet1 neuron silencing is required to prompt the opposite, in the form of aggression escalation to an intruder. Indeed, other findings also support this notion that dominance by tube test does not necessarily correlate with aggression in a resident-intruder assay (Tammimäki et al., 2010). Differences might also be explained by whether the input conditions trigger Drd2-Pet1 neurons to release 5-HT versus GABA, should the latter prove a capability, noting that Drd2-Pet1 cells express Gad2, albeit we were unable to show GABA in Drd2-Pet1 boutons, only their soma.
Interestingly, a subset of Drd2 Pet1-CKO males (four out of 26) did display increased levels of aggressive behaviors as compared with other Drd2 Pet1-CKO mice and controls, suggesting there may be other influencing variables, yet unknown. This is plausible given that mice deficient for the long isoform of DRD2 (D 2L R) are reported to show anxiety-like and depressive-like behaviors only following a stress-exposure paradigm (Shioda et al., 2019). Moreover, these stress-induced affective phenotypes in D 2L R knock-out mice were abrogated by driving D 2L R expression in DR Pet11 serotonergic neurons (Shioda et al., 2019). Together these findings suggest that the behavioral role of Drd2 expression in Drd2-Pet1 neurons may be influenced by environmental factors.
Sex-specific differences in Drd2-Pet1 neuron properties The observed sex-specific differences in Gad2 transcript levels in Drd2-Pet1 neurons may contribute to the sex-specific behavioral alterations exhibited by Drd2 Pet1-CKO mice. Gad2 expression in Drd2-Pet1 neurons is in line with prior reports showing Gad2 expression more generally in the serotonergic DR (Nanopoulos et al., 1982;Calizo et al., 2011;Shikanai et al., 2012). It may be that Drd2-Pet1 neurons can release GABA as well as or instead of 5-HT under certain conditions or at particular targets. This capacity may differ in males versus females, given our observation that in males, Drd2-Pet1 neurons harbor higher levels of Gad2 mRNA. Interestingly, Drd2 Pet1-CKO cells did not display this sex specific difference in Gad2 transcript, suggesting that Drd2 expression, or more broadly dopaminergic signaling in Drd2-Pet1 neurons, may affect Gad2 transcript levels. One potential model to be tested is if DRD2 signaling, in turn, alters levels of Gad2 expression to allow for neuronal release of GABA in addition to or instead of serotonin when behavioral or environmental conditions necessitate. Indeed, there is precedent for the differential usage of serotonin and glutamate by raphe serotonergic neurons (Liu et al., 2014;Kapoor et al., 2016;Sengupta et al., 2017;Wang et al., 2019), although GABAergic and serotonergic co-release has not been reported. AP duration measured ex vivo was longer in male versus female Drd2-Pet1 cells; this may also confer neurotransmitter release properties that could contribute to behavioral differences. Additional studies are needed to determine how Drd2 Pet1-CKO affects Drd2-Pet1 neuron electrophysiology, gene expression, or efferent targets. Such experiments may be achieved through crossing Drd2 Pet1-CKO (ePet1-Cre;Drd2 f/f ) mice to Drd2-Flpo mice (The Jackson Laboratory strain #034419 provided by Bernardo Sabatini) along with an intersectional reporter transgene which would allow for dual Cre-mediated and Flp-mediated fluorescent labeling of mutant Drd2 Pet1-CKO cells. While complex genetics, this strategy would enable mutant cell visualization for electrophysiology, single-cell RNA sequencing, and analysis of axonal projections.
In both males and females, Drd2-Pet1 neurons densely innervate auditory brainstem regions, likely modulating auditory-related processes at one or multiple of these sites. In examining Drd2-Pet1 efferents, we observed interanimal variability in regional innervation density. We speculate this may arise from subgroups within the Drd2-Pet1 neuron population that target different downstream structures. For example, some Drd2-Pet1 neurons may project specifically to the SOC while others might project specifically to the LL. If some animals have more of one subgroup than the other, averaging absolute innervation densities for each target region across all males and females may hide meaningful circuit structure. Covariance analysis of projection targets in each animal thus might hint at which brain regions come under shared regulation by Drd2-Pet1 neurons. In males, the high correlation between auditory region efferent densities suggests shared input from the same Drd2-Pet1 neurons. In females, the CNC/SOC, CNC/IC, SOC/PNC, LL/PNC, and IC/PNC combinations were more negatively correlated, suggesting there might exist a subgroup of Drd2-Pet1 neurons that targets the PNC and a different subgroup, the SOC. We speculate that in males, Drd2-Pet1 neurons contribute to a general level of serotonergic tone across the auditory brainstem, while in females, certain Drd2-Pet1 neurons selectively target and modulate specific nuclei.
In conclusion, we found that Drd2 gene expression in a specialized subset of Pet1 serotonergic neurons is required for certain defensive, dominance, and protective behaviors, involving auditory processing in a sex-specific manner. Deficits in sensory processing such as altered acoustic startle and impaired social communication and dominance behaviors manifest in human disorders including autism spectrum disorder, schizophrenia, and posttraumatic stress disorder, often in sex-specific ways (King et al., 2013;Steel et al., 2014;Matsuo et al., 2016;Thye et al., 2018) and with sex-specific differences in therapeutic outcomes (Franconi et al., 2007). The presented findings, thus, may point to novel circuit nodes of relevance to human neuropsychiatric disease.