Sex-Dependent Social and Repetitive Behavior and Neurochemical Profile in Mouse Model of Autism Spectrum Disorder

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by deficits in social interaction, impaired communication, and repetitive behaviors. ASD presents a 3:1 ratio of diagnosed boys and girls, raising the question regarding sexual dimorphic mechanisms underlying ASD symptoms, and their molecular basis. Here, we performed in vivo proton magnetic resonance spectroscopy in juvenile male and female Tsc2+/− mice (an established genetic animal model of ASD). Moreover, behavior and ultrasonic vocalizations during social and repetitive tasks were analyzed. We found significant sexual dimorphisms in the levels of metabolites in the hippocampus and prefrontal cortex. Further, we observed that female mutant animals had a differential social behavior and presented an increase in repetitive behavior. Importantly, while mutant females displayed a more simplified communication during social tasks, mutant males exhibited a similar less complex vocal repertoire but during repetitive tasks. These results hint toward sex-dependent alterations in molecular and metabolic pathways, which can lead to the sexual dimorphic behaviors and communication observed in social and repetitive environments.


Introduction
Autism spectrum disorder (ASD) is a neurodevelopmental condition affecting 1 in 59 children [1]. Its diagnosis requires the individual to present deficiencies in social communication and interaction, and restricted/repetitive behaviors (RRB) [2]. Interestingly, ASD presents a male bias, with a male to female ratio of 3:1 [3]. However, the underlying mechanisms for sex bias in ASD remain largely unknown. Therefore, it is important to explore the effect of biological sex on ASD-related pathways and consequent behavioral phenotypes.
Further, the abnormalities of speech produced by ASD individuals [30] reveal impaired communication as a hallmark of the disease. Rodents are able to communicate through ultrasonic vocalizations (USVs), which are produced by juvenile and adult animals in social contexts [15]. Importantly, USVs have also been observed in non-social situations, such as during the exploration of a novel environment or as a consequence of stress [31,32]. Several works have revealed abnormal communication in animal models of ASD (e.g., [11,[33][34][35][36][37][38]), even though the molecular mechanisms underlying this manifestation also remain unknown.
Alterations in prefrontal cortex (PFC) circuits are reported to cause social deficiencies [39][40][41], while deficient hippocampal structure and function are related to abnormal cognitive mechanisms [42]. Indeed, an aluminum-induced model of cholinergic deficiency presented altered acetylcholine levels in the PFC and hippocampus, along with reduced social skills and cognitive function [43]. Further, reduced levels of myo-inositol in the PFC were reported in rats exhibiting high-impulsivity behavior [44], as well as increased lactate in socially isolated rats [45]. These results underscore the role of specific brain structures and metabolites in the development of ASD manifestations. However, scarce studies address both the behavior and metabolic hallmarks of ASD-namely, altered social and stereotyped behaviors and specific molecular and metabolic signatures in ASD individuals, and only a few dispersed connections have been described, especially regarding biological sex.
Here, we performed in vivo proton magnetic resonance spectroscopy to analyze the levels of several metabolites in the PFC and hippocampus, in juvenile male and female mice, using a genetic model of ASD, tuberous sclerosis complex 2 (Tsc2 +/− ) mice. Additionally, we carried social and RRB-eliciting tasks. We found altered cortical gamma-aminobutyric acid (GABA)/glutamate (Glu) ratio in mutant females. Further, it was observed sexual dimorphisms in the levels of total choline (tCho), N-acetyl-aspartate + N-acetyl-aspartylglutamate (NAA + NAAG), taurine (Tau), alanine (Ala), and inositol (Ins) in the hippocampus and prefrontal cortex (PFC). These neurometabolic alterations may be somehow associated with the sex-dependent behaviors and communication found in both social and RRB-eliciting tasks, thus deserving further research.

Behavioral Testing
To test whether the observed metabolic profile was accompanied by core ASD manifestations in Tsc2 +/− mice, we performed behavioral tests, namely, social behavioral test and RRB-eliciting test.

Social Behavioral Test
Overall, these results indicate a differential social interaction behavior in female Tsc2 +/− , who also present vocalization deficiencies during such behavior. It is also interesting to note the sexual dimorphism observed in vocalization complexity, where no differences were found between males.

RRB Behavioral Test
RRBs are among the core symptoms of ASD and are widely observed in ASD animal models. Here, we observed an increased number of total buried marbles by female Tsc2 +/− (female WT = 4.500 ± 0.5955 marbles buried; female Tsc2 +/− = 6.207 ± 0.4043 marbles buried; p = 0.0244; Figure 5a). These results indicate an increased repetitive, stereotyped behavior displayed by transgenic females. Regarding USV production, there were no significant differences regarding USV length, principal frequency, or frequency amplitude (Supplementary Figure S1d-f). However, the analysis on the complexity of the USVs produced during this task revealed that differences in the vocal repertoire were found between WT and transgenic males but not females. There was a significant interaction between call category and genotype (F (6, 57) = 3666; p = 0.0038) and an effect of call category (F (2, 57) = 46,13; p < 0.0001). Indeed, Tsc2 +/− males emitted significantly more single USVs than their WT littermates (male WT = 41.19 ± 14.65%; male Tsc2 +/− = 89.52 ± 6.801%; p = 0.0063), thus displaying reduced vocal complexity (Figure 5b). This seems to indicate that, contrary to social tasks, RRB-eliciting tasks promote alterations in vocal complexity in Tsc2 +/− males.

Discussion
Studies on region-specific metabolic profiles in ASD poorly address the behavioral phenotypes of the condition. Importantly, most studies carried out thus far give little-tono relevance to biological sex. Here, we investigated in detail the metabolic signatures of the hippocampus and PFC of male and female Tsc2 +/− mice, an established genetic model of ASD. We found region-specific sexual dimorphisms in several metabolites. Indeed, an increase in Ala, Tau, and Ins levels in the PFC and a decrease in NAA+NAAG and Tau

Discussion
Studies on region-specific metabolic profiles in ASD poorly address the behavioral phenotypes of the condition. Importantly, most studies carried out thus far give little-to-no relevance to biological sex. Here, we investigated in detail the metabolic signatures of the hippocampus and PFC of male and female Tsc2 +/− mice, an established genetic model of ASD. We found region-specific sexual dimorphisms in several metabolites. Indeed, an increase in Ala, Tau, and Ins levels in the PFC and a decrease in NAA + NAAG and Tau levels in the hippocampus of mutant females were observed. Additionally, while transgenic males differed in cortical Ala and GSH and hippocampal tCho, Ins, Lac, and Tau, transgenic females showed important differences in cortical GABA/Glu ratio relative to WT.
Given these results, we performed behavioral and communication analyses in another set of juvenile mice to investigate whether the alterations observed in the brain translated into abnormal phenotypes. In this sense, we carried out social and repetitive tasks, to investigate into core features of ASD. Moreover, we recorded the ultrasonic vocalizations produced by the animals during the above-mentioned tasks to assess communication ability. The classification algorithm used [46] allows the determination of vocal complexity: Single USVs are the least intricate vocalizations, not having sudden frequency changes, such as multi-syllabic USVs, or simultaneous independent sounds at different frequencies such as stacked USVs. Our data indicate that transgenic females showed altered sociability, as well as decreased vocal complexity, hinting that altered social behavior may promote or be linked to vocal alterations as well.
In this work, the phenotypic disturbances displayed by transgenic females were also present in RRB-eliciting tasks. RRBs are associated with anxiety [47][48][49], as sameness of behavior may represent a strategy to better control the individual's surroundings and avoid uncertainty, and thus temporarily reduce anxious feelings [50]. Interestingly, the marbleburying test, such as the one performed in this work, has shown controversy, as some reports have described reduced and not increased burying activity [24,[51][52][53]. However, this behavior may relate to the avoidance of novel objects [54]. Therefore, we posit that the altered sociability of transgenic females may also represent a stress-coping strategy. In fact, increased anxiety paired with reduced hippocampal NAA + NAAG in Tsc2 +/− females is in line with previous findings, in which NAA levels diminished in the hippocampus following fear conditioning [55]. A reduction in the levels of this metabolite in the hippocampus and cortex after an acute cold swim stress test has also been described [56].
Importantly, even though mutant females also displayed increased repetitive, stereotyped, anxious behavior, this was not accompanied by alterations in communication in RRB-eliciting tasks. Indeed, while social tasks elicited communication differences between females, RRB-eliciting tasks promoted communication discrepancies between males. It has been described that the USV call rate of male adult mice is influenced by social isolation and is dependent on behavioral states [57]. Interestingly, a study using only female CBA/CaJ mice showed that social isolation alters USV perception, also indicating that vocal communication may be dependent on experience and/or emotional state [58]. However, currently, a gap exists regarding USV production during RRB-eliciting tasks, as most studies focus on neonatal age or social contexts. Our data give important insights into the factors that may be governing communication and suggest that they could be sex dependent as well as environment specific.
Interestingly, our results demonstrate an overall greater severity in ASD-like phenotype in mutant females, rather than males. This seems to contradict the notion that males are more affected by the disorder, both by prevalence and by liability, as there seem to exist sex-differential factors that may potentiate risk in males and protect females [59]. However, this does not appear to be the case when ASD in TSC individuals is specifically considered [60]. Indeed, the syndrome was observed to be more pervasive in females, both in humans [61] and in animal studies [62]. This is in accordance with our results and further validates our animal model for the research of ASD in TSC individuals.
Region-specific expression patterns of metabolic proteins have already been described in the PFC and hippocampus of male mice susceptible and not susceptible to social stress. This indicates that specific cortical and hippocampal metabolic signatures may be underlying sociability and anxiety [63]. Indeed, region-specific mechanisms of excitation/inhibition imbalance have been previously described [64]. This is in line with our results, which extend this matter by considering biological sex. Indeed, we found genotype-dependent differential GABA/Glu ratio in the PFC only in females, who also display differences in repetitive behavior and in social behavior and communication. This suggests that a sex-dependent and increased cortical excitatory tone may be driving the observed core features of ASD and their corresponding sexual dimorphisms in this particular period of neurodevelopment.
Region-specific alterations seem to be associated with distinct behavioral outcomes. While hippocampal studies demonstrate that metabolic alterations or deficits result in depressive symptoms and learning deficiencies, studies on cortical metabolic profiles show that altered metabolite changes result in altered social, impulsive, or anxious behavior. In the hippocampus, chronic restraint stress has been shown to alter cholinergic signaling [65], and conversely, choline supplementation was beneficial for the reversal of nicotine-induced learning deficits [66]. Further, treatments with creatine and taurine yielded antidepressant effects, with increased hippocampal catecholamine and serotonin levels [67]. This is in agreement with our data on taurine levels, where transgenic females displayed a reduction in the hippocampus and decreased vocal complexity, which could be linked to cognitive disabilities. Interestingly, transgenic females showed increased taurine levels in the prefrontal cortex, which indicates that behavioral phenotypes are dependent on the metabolic profiles of a specific brain region. Additionally, a study found that caffeine plus taurine supplementation resulted in greater activity [68]; in this sense, the increased hippocampal taurine in transgenic males we observed could be translated into their higher number of social interactions. On the other hand, this augmented activity could be an expression of increased hyperactivity caused by decreased inositol in the prefrontal cortex of transgenic males, as previously shown [44]. Moreover, a study on mice with impaired glutathione synthesis found an association between decreased myo-inositol concentration and social isolation [69]. This is in accordance with our findings, as transgenic females showed increased inositol levels and sociability.
Overall, this work sheds light on the sexually dimorphic, region-specific metabolic profiles that may be mediating sex-dependent behaviors and communication displayed in different environments. More studies focusing on the role of biological sex are necessary in order to further understand how male and female metabolic signatures differ, and how that translates into abnormal ASD-like behavior.

Animals
Tsc2 +/− mice were generated and backcrossed to a C57BL/6N. Mice for experiments were generated by crossing Tsc2 +/− animals with mice with C57BL/6J background. All pups were housed together with the dam until postnatal day 21 (P21) and segregated by sex from P21 onwards. All animals were maintained in a housing room with a 12 h light-12 h dark cycle, at 21 ± 2 • C. Identification of pups was performed at P4 with permanent tattoos on the toes. Tail tips were collected at P4 for posterior genotyping. All experiments were carried out in accordance with the European Union Council Directive (2010/63/EU) and the National Regulations, approved by the Internal Review Board of ICNAS and conducted under the authority of the Project License (1/2017).

Magnetic Resonance Spectroscopy
For localized 1 H-magnetic resonance spectroscopy ( 1 H-MRS), data were collected in a volume of interest placed on the hippocampus and pre-frontal cortex (PFC). B0 map was acquired before spectroscopy, and shims were optimized through a MAPSHIM voxel. Spectra were acquired using a point-resolved spectroscopy (PRESS) sequence with outer volume suppression (OVS) and VAPOR water suppression [70,71]. The following parameters were used: TR = 2500 ms, TE = 16.225 ms, number of averages = 720, 3 flip angles = 90 • , 142 • , 142 • , bandwidth = 5000 Hz, number of acquired points = 2048 yielding a spectral resolution of 1.22 Hz/pt. The total acquisition time was 30 min. Before each spectrum, we acquired an unsuppressed water spectrum at the same voxel location (TE = 16.225 ms, TR = 2500 ms, 16 averages, scanning time = 40 s) (Supplementary Figure S2). Data analysis of 1 H-MRS spectra was performed using linear combination modeling LCModel (Stephen Provencher Inc., Toronto, ON, Canada) [72]. Metabolite quantification was performed applying the internal water reference method. Concentrations in millimole units were calculated for metabolites, and results are presented in arbitrary units (a.u.). Only metabolites with Cramér-Rao bounds < 20% were considered for statistical analysis. A timeline of spectroscopic experiments is represented in Figure 6.
the National Regulations, approved by the Internal Review Board of ICNAS and conducted under the authority of the Project License (1/2017).

Magnetic Resonance Spectroscopy
For localized 1 H-magnetic resonance spectroscopy ( 1 H-MRS), data were collected in a volume of interest placed on the hippocampus and pre-frontal cortex (PFC). B0 map was acquired before spectroscopy, and shims were optimized through a MAPSHIM voxel. Spectra were acquired using a point-resolved spectroscopy (PRESS) sequence with outer volume suppression (OVS) and VAPOR water suppression [70,71]. The following parameters were used: TR = 2500 ms, TE = 16.225 ms, number of averages = 720, 3 flip angles = 90°, 142°, 142°, bandwidth = 5000 Hz, number of acquired points = 2048 yielding a spectral resolution of 1.22 Hz/pt. The total acquisition time was 30 min. Before each spectrum, we acquired an unsuppressed water spectrum at the same voxel location (TE = 16.225 ms, TR = 2500 ms, 16 averages, scanning time = 40 s) (Supplementary Figure S2). Data analysis of 1 H-MRS spectra was performed using linear combination modeling LCModel (Stephen Provencher Inc., Toronto, Canada) [72]. Metabolite quantification was performed applying the internal water reference method. Concentrations in millimole units were calculated for metabolites, and results are presented in arbitrary units (a.u.). Only metabolites with Cramér-Rao bounds < 20% were considered for statistical analysis. A timeline of spectroscopic experiments is represented in Figure 6.

Ultrasonic Vocalization Recordings and Analysis
A 55 cm × 50 cm × 70 cm (H × D × W) anechoic chamber was assembled with 1.5 cm thick acrylic sheets and fully covered with absorbing foam on the inside, to block external sound. All USVs were acquired using an ultrasound recording system with Avisoft CM16/CMPA condenser microphone placed 10 cm above container wall top, Ultra-soundGate 416H amplifier, and Avisoft Recorder software (Avisoft Bioacoustics, Glienicke/Nordbahn, Germany). Sonograms were generated, with 512 point FFT, 16-bit format,

Ultrasonic Vocalization Recordings and Analysis
A 55 cm × 50 cm × 70 cm (H × D × W) anechoic chamber was assembled with 1.5 cm thick acrylic sheets and fully covered with absorbing foam on the inside, to block external sound. All USVs were acquired using an ultrasound recording system with Avisoft CM16/CMPA condenser microphone placed 10 cm above container wall top, UltrasoundGate 416H amplifier, and Avisoft Recorder software (Avisoft Bioacoustics, Glienicke/Nordbahn, Germany). Sonograms were generated, with 512 point FFT, 16-bit format, a sampling frequency of 250 kHz, a time resolution of 1 ms, a frequency resolution of 488 Hz, and an overlap of 50%. USVs were further analyzed using MATLAB toolbox DeepSqueak version 2.6.1., which allowed the extraction of individual mouse USV calls by applying the Mouse Call_Network_V2 neural network, with a chunk length analysis of 6 s, an overlap of 0.1 s, a high-frequency cut-off of 125 kHz, and no score threshold. The principal frequency, frequency range, and duration of each call were extracted, and call rate (number of USVs produced over time) was calculated. Each USV was manually classified into three classes. The characteristics of each USV class are as follows: single: one waveform only present in the sonogram; multi-syllabic: two or more waveforms, with no interval in time between them and no temporal overlap; stacked: two or more waveforms with temporal overlap.

Video Recordings
The juvenile play test and marble-burying test were video recorded with a Logitech C170 video camera (Logitech Europe S.A., Lausanne, Switzerland) on top of the chamber under dim red light (5 lux). All video recordings were manually analyzed by the operator.

Juvenile Social Play Test
On P24, animals were isolated for 24 h prior to the social task and housed in a new clean mouse cage with fresh bedding, to potentiate their social behavior. On P25, two same-sex, same-genotype animals were placed on a clean, standard cage in the anechoic chamber [73]. The vertical walls of the cage were further extended with acrylic sheets to avoid the animal exiting the apparatus. The animals were allowed to interact for 30 min. At the end of the experiment, both animals were returned to their home cage. Video recordings were analyzed by an operator blind to sex and genotype, to identify social behavior. Actions considered as social behaviors included investigative interactions (sniffing, following, mutual circle), affiliative interactions (group sitting, allogrooming, push under), play solicitation (crawl, push past, approach), and agonistic interactions (threat, aggression, defense, flight, submission) [74,75]. Total and average social time were manually tracked with a stopwatch, and the number of social interactions was registered. A timeline of behavioral experiments is represented in Figure 7.
On P24, animals were isolated for 24 h prior to the social task and housed in a new clean mouse cage with fresh bedding, to potentiate their social behavior. On P25, two same-sex, same-genotype animals were placed on a clean, standard cage in the anechoic chamber [73]. The vertical walls of the cage were further extended with acrylic sheets to avoid the animal exiting the apparatus. The animals were allowed to interact for 30 min. At the end of the experiment, both animals were returned to their home cage. Video recordings were analyzed by an operator blind to sex and genotype, to identify social behavior. Actions considered as social behaviors included investigative interactions (sniffing, following, mutual circle), affiliative interactions (group sitting, allogrooming, push under), play solicitation (crawl, push past, approach), and agonistic interactions (threat, aggression, defense, flight, submission) [74,75]. Total and average social time were manually tracked with a stopwatch, and the number of social interactions was registered. A timeline of behavioral experiments is represented in Figure 7. Figure 7. Timeline of behavioral experiments in this study. All animals were identified on postnatal day (P)4 with permanent tattoos on toes, and tails were collected for genotyping. On P24, animals were socially isolated in individual clean cages and juvenile social play test was carried out on P25. On P40, marble-burying test was performed.

Marble-Burying Test
On P40, each animal was placed on a clean, standard mouse cage filled with a 5 cm layer of sawdust and with 12 marbles equidistantly placed on top of the sawdust, on a 3 × 4 arrange. The vertical walls of the cage were further extended with the application of acrylic sheets, to avoid the animal exiting the apparatus. The cage was inside the anechoic chamber. The animal was allowed to interact with marbles for 30 min. At the end of the experiment, the animal was returned to its home cage. Video recordings were analyzed by an operator blind to sex and genotype, to register the number of marbles buried at 5, Figure 7. Timeline of behavioral experiments in this study. All animals were identified on postnatal day (P)4 with permanent tattoos on toes, and tails were collected for genotyping. On P24, animals were socially isolated in individual clean cages and juvenile social play test was carried out on P25. On P40, marble-burying test was performed.

Marble-Burying Test
On P40, each animal was placed on a clean, standard mouse cage filled with a 5 cm layer of sawdust and with 12 marbles equidistantly placed on top of the sawdust, on a 3 × 4 arrange. The vertical walls of the cage were further extended with the application of acrylic sheets, to avoid the animal exiting the apparatus. The cage was inside the anechoic chamber. The animal was allowed to interact with marbles for 30 min. At the end of the experiment, the animal was returned to its home cage. Video recordings were analyzed by an operator blind to sex and genotype, to register the number of marbles buried at 5, 10, 15, 20, 25, and 30 min of the experiment [76]. A timeline of behavioral experiments is represented in Figure 7.

Statistical Analyses
All data were analyzed using GraphPad Prism version 8.0.1 (GraphPad Software, San Diego, CA, USA). Outliers were identified as values outside of mean ± 3SD interval and excluded from statistical tests. Two-way ANOVAs followed by Sidak's multiple comparisons tests (USV analyses) or Kruskal-Wallis, followed by Dunn's multiple comparison tests (MRS and behavioral analyses), were conducted to compare experimental groups. All effects are reported as significant at p < 0.05. Error bars are given as SEM.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/metabo12010071/s1, Figure S1: Localized 1 H NMR voxel in the prefrontal cortex and hippocampus, Figure S2: Length, principal frequency and amplitude-frequency of USVs produced during social behavior task and RRB-eliciting task, Table S1: 1 H-magnetic resonance spectroscopy (MRS) analysis of cortical neurochemical profile,