Allelic contribution of Nrxn1α to autism-relevant behavioral phenotypes in mice

Copy number variations (CNVs) in the Neurexin 1 (NRXN1) gene, which encodes a presynaptic protein involved in neurotransmitter release, are some of the most frequently observed single-gene variants associated with autism spectrum disorder (ASD). To address the functional contribution of NRXN1 CNVs to behavioral phenotypes relevant to ASD, we carried out systematic behavioral phenotyping of an allelic series of Nrxn1 mouse models: one carrying promoter and exon 1 deletion abolishing Nrxn1α transcription, one carrying exon 9 deletion disrupting Nrxn1α protein translation, and one carrying an intronic deletion with no observable effect on Nrxn1α expression. We found that homozygous loss of Nrxn1α resulted in enhanced aggression in males, reduced affiliative social behaviors in females, and significantly altered circadian activities in both sexes. Heterozygous or homozygous loss of Nrxn1α affected the preference for social novelty in male mice, and notably, enhanced repetitive motor skills and motor coordination in both sexes. In contrast, mice bearing an intronic deletion of Nrxn1 did not display alterations in any of the behaviors assessed. These findings demonstrate the importance of Nrxn1α gene dosage in regulating social, circadian, and motor functions, and the variables of sex and genomic positioning of CNVs in the expression of autism-related phenotypes. Importantly, mice with heterozygous loss of Nrxn1, as found in numerous autistic individuals, show an elevated propensity to manifest autism-related phenotypes, supporting the use of models with this genomic architecture to study ASD etiology and assess additional genetic variants associated with autism.


Introduction
Neurexins are presynaptic adhesion molecules that play an essential role in trans-synaptic signaling, cell adhesion, and neurotransmitter release [1][2][3]. There are three evolutionarily conserved Neurexin genes, NRXN1, NRXN2, and NRXN3, that are expressed in the mammalian brain, each of which is transcribed from two distinct promoters that generate a longer alpha isoform (α) and a shorter beta (β) isoform. These isoforms share identical intracellular domains, but different extracellular structures [2]. The NRXN1 locus contains an additional third promoter internal to the α and β isoforms that gives rise to a gamma isoform (Neur-exin1γ), composed of the same intracellular domains as the other two isoforms, but only having the membrane-proximal sequences of the extracellular domain [4]. Although all three neurexin genes have been associated with psychiatric disorders, NRXN1 variants account for the majority of disease burden, and NRXN1 variants appear to be the most frequently observed single-gene variants associated with autism spectrum disorder (ASD) [5,6].
NRXN1 is one of the longest human genes, with 24 exons spanning over 1 Mb on chromosome 2. This lengthy locus is particularly susceptible to non-recurrent copy number variations (CNVs) [2]. Varying in size and location across the NRXN1 locus, these non-recurrent CNVs are associated with a variety of psychiatric and neurodevelopmental phenotypes, such as developmental delay, intellectual disability, schizophrenia, and autism [2]. Notably, most of the reported CNVs are heterozygous, with only 11 individuals carrying documented homozygous deletions thus far, all of which show significant development delay and are identified as having Pitt-Hopkins syndrome [2,7]. Additionally, exonic deletions, particularly in the 5' end of NRXN1, appear to be more pathogenic than intronic deletions [6]. This supports the idea that disruption of the NRXN1 alpha isoform (NRXN1α) is more frequently associated with clinical phenotypes than disruptions of other isoforms, as the promoters of the NRXN1β and NRXN1γ isoforms are located downstream in intron 17 and intron 23 of the alpha isoform, respectively [2,4]. Nevertheless, deletions in the 3' end of the NRXN1 gene, including those occurring within the NRXN1β isoform, have also been associated with clinical manifestations [6,8]. Together, these findings highlight the genetic complexity at the NRXN1 locus and support the need to systematically assess the neurobiological effect of CNVs in NRXN1 that vary in location across the genomic locus-5' versus 3' or exonic versus intronic-in eliciting behavioral phenotypes.
The first Nrxn1 mouse model, in which the promoter and exon 1 is deleted and thus abolishing Nrxn1α transcription, was developed two decades ago [9]. Several behavioral phenotyping studies reported that mice carrying Nrxn1α promoter/exon 1 homozygous deletion (often referred to as knockout, KO, and named as ΔExon1/ΔExon1 hereafter) exhibit multiple behavioral phenotypes relevant to autism, including alterations in social affiliative behaviors, aggression, locomotion, self-grooming, motor learning, and value-based decision making [10][11][12][13]. Although mice with Nrxn1α promoter/exon 1 heterozygous deletion (ΔExon1/+) were included in several studies, the phenotypes reported in those studies were inconsistent with each other and sex as a variable was often not considered [11,12,[14][15][16]. Moreover, the effects of genomic position of a CNV deletion across the NRXN1 locus on isoform expression and subsequent autism-associated behavioral phenotypes remain to be assessed.
Given that sleep and circadian disturbances are some of the most common comorbidities in ASD, occurring in 50-80% of autistic children [17][18][19][20], the assessment of circadian abnormalities in autism models is of paramount importance. Common sleep problems in ASD include delayed sleep onset, shorter sleep duration, multiple and prolonged night awakenings, and daytime tiredness [20][21][22]. A recent meta-analysis of 49 peer-reviewed articles and 51 independent samples found that sleep problems are most strongly associated with externalizing symptoms, internalizing symptoms, and executive functioning in autistic individuals [23]. Thus, identifying a translation model that captures the sleep/circadian phenotypes similar to those observed in ASD may elucidate the biological underpinnings of these phenotypes. However, this clinically relevant phenotype has not been assessed in Nrxn1α mouse models.
To dissect the genetic complexity of autism at the NRXN1 locus, we have collected and generated an allelic series of mouse models carrying distinct CNVs across the Nrxn1 locus, including ΔExon1, that has been studied previously but not systematically for ΔExon1 heterozygotes, and two new mouse models: the first bearing a deletion of Nrxn1α exon 9 that disrupts Neurexin 1α protein translation (named as ΔExon9 hereafter), and the second carrying an ASD proband-associated~20kb deletion in intron 17 of Nrxn1α, upstream of the Nrxn1β promoter (named as ΔIntron17 hereafter). Given that heterozygous CNVs in NRXN1 are the most observed genetic architecture associated with ASD in humans, we performed systematic phenotyping of mice carrying heterozygous ΔExon1, ΔExon9 or ΔIntron17, in comparison to their respective wild-type (WT) littermate controls. We have also included homozygous deletions of exon 9 or intron 17 in the same cohort of study as these two mouse models are newly generated. To address sex as a variable and examine behavioral phenotypes relevant to clinical symptoms of ASD, both male and female mouse models were subjected to the same set of behavioral tests, including elevated zero maze to assess anxiety-related behaviors, open field test to measure locomotor activity, three chamber approach for social choice and social memory, rotarod for motor coordination and motor skill learning, running wheels for circadian activity, and resident intruder test for aggressive types of behaviors. We find that heterozygous loss of Nrxn1α, as shown in ΔExon1/+ and ΔExon9/+ mice, impairs social memory in male mice and enhances repetitive motor coordination in both sexes. Homozygous loss of Nrxn1α, as shown in ΔExon9/ΔExon9 mice, impairs social memory, alters circadian activities, and enhances motor coordination in both sexes, but selectively elicits enhanced aggression in males and reduced affiliative social interactions in females. In contrast, locomotor function and anxiety-related behaviors were not affected in ΔExon1 and ΔExon9 mouse models, regardless of sex or heterozygous or homozygous deletions. Mice with ΔIntron17 do not show any behavioral abnormalities in these tests. Our findings demonstrate a vital role for Neurexin1α in regulating circadian activities, for the first time, and underscore the significance of Nrxn1α gene dosage, genomic positioning of CNV deletions, and sex in eliciting ASD-relevant behavioral phenotypes.

Nrxn1 allelic series in mice-expression of Nrxn1 isoforms
In this study, we employed three mouse models with distinct CNVs across the Nrxn1 locus to assess the functional significance of genomic positioning of those CNVs. The first was a newly acquired mouse model from the MRC Mary Leon Center carrying the deletion of exon 9 in Nrxn1α (Nrxn1 ΔExon9/+ or Nrxn1 ΔExon9/ΔExon9 ; subsequently referred to as ΔExon9/+ or ΔExon9/ΔExon9), an exon that is conserved across humans and mice [24]. Deletion of exon 9 introduces a premature stop codon in exon 10, thus disrupting Neurexin1α protein translation at its N-terminus. To measure the extent to which Nrxn1 mRNAs, particularly the three different isoforms, are affected by exon 9 deletion, we carried out RT-qPCR to measure the relative Nrxn1α isoform-specific mRNA levels in cortical tissues of ΔExon9 mice using primers located in different conserved exons ( Fig 1A). Specifically, we used primer pair E9α-F and E10α-R to detect exon 9 deletion, finding that this PCR product was reduced by roughly 50% in heterozygous mice (ΔExon9/+) and by nearly 100% in homozygous mice (ΔExon9/ΔExon9), regardless of sex (Fig 1B), confirming the successful deletion of exon 9 in this ΔExon9 model.
Notably, RT-PCR with primers located in exon 14 and exon 15 (primer pair E14α-F and E15α-R; Fig 1A), two conserved exons across humans and mice [24], shows that relative Nrxn1α mRNA levels are significantly decreased to about 50% of WT levels in ΔExon9/+ mice and nearly eliminated in ΔExon9/ΔExon9 mice, in both males and females ( Fig 1C). In contrast, the relative mRNA expression for both Nrxn1β and Nrxn1γ isoforms are not affected regardless of heterozygous or homozygous deletion of exon 9 in both sexes (Fig 1D and 1E). These results suggest that the premature stop codon in exon 10 triggers nonsense-mediated mRNA decay (NMD), a known surveillance mechanism that targets mRNAs containing a premature stop codon for rapid degradation [25,26]. Thus, NMD minimizes potential production of truncated protein products of Neurexin 1α. Unfortunately, the lack of specific antibodies against Neurexin 1α limited our ability to examine Neurexin 1α protein levels in this study. Given the premature stop codon/NMD triggered by deletion of exon 9 and the resulting elimination of Nrxn1α mRNAs (Fig 1C), Neurexin1α protein translation is likely abolished and mice with ΔExon9 are therefore considered as a legitimate loss-of-function model to study the function of Nrxn1α.
To determine whether a CNV in an intron near the 3' end of the Nrxn1 gene elicits behavioral phenotypes, we also generated a mouse model carrying a~20 kb deletion at intron 17 of Nrxn1α (Nrxn1 ΔIntron17/+ or Nrxn1 ΔIntron17/ΔIntron17 ; subsequently referred to as ΔIntron17/+ or ΔIntron17/ΔIntron17) that is similar to a CNV identified in an individual on the autism spectrum (in the Autism Spectrum Program of Excellence (ASPE) cohort [27]). Upon examining the expression levels of all three different Nrxn1 isoforms in this mouse model, we found that both Nrxn1α and Nrxn1γ transcript levels were unaffected in these mice (Fig 1F and 1H). However, Nrxn1β levels were slightly reduced in homozygous male mice (ΔIntron17/ ΔIntron17), compared to WT mice ( Fig 1G). This reduction may be partially related to the fact that the 20kb intronic deletion is approximately 80kb upstream from the promoter of Nrxn1β, which is located within intron 17 of Nrxn1α [9,28,29]. However, the slight reduction of Nrxn1β isoform expression is not observed in female mice carrying the same intronic homozygous or heterozygous deletions ( Fig 1G). Nevertheless, this mouse model allows us to probe Three different promoters driving Nrxn1 isoform-specific transcription are indicated with differently colored arrows. The Nrxn1β specific exon is labeled with a brown vertical line and the Nrxn1γ specific exon is labeled with a green vertical line. Translation start site, ATG, for each isoform-specific protein translation, was labeled under each isoform-specific exon. The positions of primers used for quantitative RT-PCR are indicated below enlarged exons (black bars) with red arrows. The genomic positions for promoter/exon 1 deletion (ΔExon1), exon 9 deletion (ΔExon9), and a~20kb deletion in intron 17 (ΔIntron17) are marked with blue bars on top of the Nrxn1 gene. B-E. Quantitative measurement of Nrxn1 isoform specific mRNAs in the ΔExon9 mouse line (top row for males and bottom row for females). (B) Nrxn1α mRNAs containing the exon 9 sequence is decreased to about 50% in ΔExon9/+ mice and 100% in ΔExon9/ΔExon9 mice compared to +/+ WT controls (Exon9 labelled as ΔEx9). (C) Nrxn1α mRNAs are reduced to about 50% in ΔExon9/ + and nearly 100% in ΔExon9/ΔExon9 mice compared to +/+ WT controls. Nrxn1β (D) and Nrxn1γ (E) mRNAs are not altered in mice carrying ΔExon9/+ or ΔExon9/ΔExon9 in comparison to +/+ (WT). For panels of ΔExon9 mice, behavioral phenotypes associated with Nrxn1 distal intronic deletion found from an ASD proband.
Previously, multiple studies have examined the behavioral phenotypes in mice carrying homozygous Nrxn1α promoter/exon 1 deletions [9][10][11][12][13][14][15][16]. However, heterozygous ΔExon1 mice have not been systemically studied in both males and females. Given the large number of CNVs occurring near the 5'-end of NRXN1 locus in humans with autism and that nearly all of those CNVs are heterozygous, we have included this mouse model, particularly the heterozygotes (ΔExon1/+), in our allelic series as well.

Normal locomotion and anxiety-like behaviors in Nrxn1 mutant mice
To systematically characterize the contribution of gene dosage and genomic position of CNV deletions in NRXN1 to autism-relevant behavioral phenotypes in both males and females, we performed a battery of behavioral tests (S1 Fig) in adult mice (� 3 months of age) from these three mouse models with WT, heterozygous, and homozygous deletions of specific parts of Nrxn1. Given that anxiety is a comorbid condition in a subset of individuals on the autism spectrum, we explored whether the Nrxn1 mutant mice display alterations in anxiety-related behaviors in the elevated zero maze and spontaneous activity in the open field test. The elevated zero maze is sensitive to anxiolytic treatments, which increase time spent in the open areas of the maze. In the open field, the natural proclivity of a mouse to explore a novel environment is assessed, providing data on ambulation/locomotor function and thigmotaxis. Thigmotaxis refers to the tendency to move along the periphery of the arena and increased thigmotaxis is considered an anxiety-related phenotype. We found that all three mouse models, ΔExon1, ΔExon9, and ΔIntron17, exhibited similar anxiety-like behaviors as WT littermates on the elevated zero maze and open field tests, regardless of sex or genotype (S2 and S3 Figs). Mice carrying heterozygous ΔExon1 show slightly increased locomotor activity in male carriers (S2G Fig), but not in females (S2K Fig), and no locomotion difference was found in mice carrying ΔExon9 or ΔIntron17 (S2 and S3 Figs). This is in stark contrast to other mouse models of neurodevelopmental disorders, such as those for Rett syndrome that show robust reductions in locomotor activity and those for CDKL5 deficiency disorder that exhibit consistent hyperactivity across similar behavioral tasks [30][31][32][33]. Together, the absence of significant alterations in locomotor or anxiety-related behaviors in these Nrxn1 mouse models provides a behavior control and allows us to appropriately interpret other motor-related behavioral phenotypes.

Impairments of social function in Nrxn1α mutant mice
Since persistent deficits in social interactions are key diagnostic criteria for ASD, we next examined social behavior in Nrxn1 mutant mice using the three-chamber social approach and the resident-intruder test. There are two phases of the social approach test in which behaviors are assessed: (1) a social choice phase that is indicative of social interaction in which the animal explores cylinders containing either a novel, inanimate object or novel conspecific mouse, and male +/+, n = 9; male ΔExon9/+, n = 10; male ΔExon9/ΔExon9, n = 9; female +/+, n = 11; female ΔExon9/+, n = 14; female ΔExon9/ΔExon9, n = 11. One-way ANOVA test with Dunnett's multiple comparison test was used to analyze the results. F-H. Quantitative measurement of Nrxn1 isoform specific mRNAs in the ΔIntron17 mouse line (top row for males and bottom row for females). Nrxn1α (F), Nrxn1β (G) and Nrxn1γ (H) mRNAs are not altered in mice carrying ΔIntron17/+ or ΔIntron17/ΔIntron17, in comparison to +/+ (WT) (Intron17 labelled as In17). For panels of ΔIntron17 mice, n = 10 in each genotype group. One-way ANOVA test with Dunnett's multiple comparison test was used to analyze the results. For all panels in Fig 1,  (2) a social novelty phase in which the animal has the choice to explore cylinders containing either a familiar, conspecific mouse or a novel, conspecific mouse [34,35]. None of the mouse models exhibited impairments in the social choice phase of the three-chamber social approach paradigm, showing comparable time spent sniffing the social cylinder across different genotypes, regardless of Nrxn1α expression status, and indicating normal olfactory function in discriminating a social cue as well (Figs 2A, 2C and S4; S1-S3 Tables). On the other hand, mice with reduced Nrxn1α expression, such as those carrying heterozygous or homozygous ΔExon9 or heterozygous ΔExon1, displayed a reduction in time spent exploring the novel mouse during the social novelty phase, particularly in male carriers, indicative of an impairment in preference for social novelty or social memory [34,35] (Fig 2B and 2D; S1 and S2 Tables). In contrast, male and female mice carrying intronic deletions do not show any deficits in these three chamber social tests (S4 Fig; S3 Table).
The resident-intruder assay is a test in which a sex-matched young intruder mouse is introduced into a singly housed mouse's home cage to assess their interactive behaviors, including aggressive interactions, such as barbering, and passive interactions, such as chasing, following, and anogenital sniffing. Neither ΔExon9/+, ΔExon1/+, nor ΔIntron17/+, ΔIntron17/ΔIntron17 mice displayed alterations in aggressive behaviors and passive interactions in the residentintruder test (Figs 3 and S5; S2 and S3 Tables). However, mice with ΔExon9/ΔExon9 exhibited sex-specific alterations with males demonstrating increased aggressive behaviors and females ΔExon9/+ and ΔExon9/ΔExon9 male mice spend significantly less time than +/+ males sniffing the novel mouse (left), and no difference was found in females (right) in the social novelty phase. Male +/+, n = 14; male ΔExon9/+, n = 18; male ΔExon9/ΔExon9, n = 9; female +/+, n = 18; female ΔExon9/+, n = 17; female ΔExon9/ΔExon9, n = 11. Data were analyzed using one-way ANOVA test with Tukey's multiple comparison test. C. No genotype difference was observed in time spent sniffing social cylinder among ΔExon1 male (left) and female (right) mice in the social choice phase. D. ΔExon1/+ male mice spend significantly less time than +/+ males sniffing the novel mouse (left), and no difference was found in females (right) in the social novelty phase. Male +/+, n = 18; male ΔExon1/+, n = 12; female +/+, n = 14; female ΔExon1/+, n = 13. Data were analyzed using the Mann-Whitney U test. For all panels in Fig  demonstrating reduced passive social interactions (Fig 3A-3C). Furthermore, in the ΔExon9 mouse strain there was a significant effect of genotype on time the resident mouse spent with the intruder mouse (S1 Table). Together, these findings suggest that Nrxn1α dosage plays a critical role in altering particular social behaviors. For example, decreased preference for social novelty is observed in mice with reduced Nrxn1α expression, including heterozygous deletion of exon 1 or exon 9, as well as homozygous deletion of exon 9 (ΔExon1/+; ΔExon9/+, ΔExon9/ ΔExon9). However, the manifestation of aggressive behaviors requires a homozygous loss of Nrxn1α. These observations support that a dose-dependent reduction in Nrxn1α expression leads to multiple impairments in social function that are relevant to autism.

A reduction in Nrxn1α expression alters circadian locomotor and bout activity
Sleep disturbances are one of the most common comorbidities in autism and estimated to occur in up to 80% of autistic children [17][18][19][20]. Reported sleep disturbances in autism include perturbed sleep-onset, impaired sleep maintenance, and/or alterations in waking time [17][18][19][20]. Furthermore, disturbances in circadian rhythms have recently been reported in a mouse model of neurodevelopmental disorders induced via prenatal maternal immune activation [36]. As this clinically relevant phenotype has not been previously studied in Nrxn1 deficient animal models, we sought to systemically characterize endogenous circadian activity cycles and assessed multiple parameters, including both parametric and nonparametric traits of circadian activity using running wheel cages in which mice are individually housed and wheel revolutions are continually collected over 4 weeks of time (S1 Fig). In the first two weeks, mice were exposed daily to 12 hours of light and 12 hours of dark (L/D phase), which mimics the naturally occurring light-dark signal from the environment. In the last two weeks, mice were exposed to constant darkness (D/D phase), which enables the assessment of circadian behavior (or the free-running rhythm) without the influence of light entrainment (Figs 4A and 4B; S1, S9A and S9B).
During the L/D phase, we found that only mice with homozygous loss of Nrxn1α demonstrated significantly decreased running wheel activity, with ΔExon9/ΔExon9 mice exhibiting a reduction specifically during the dark period, which is the active period of nocturnal mice (Figs 4Cs, 4D and S9C). Notably, reduced overall activity was also observed in the most active 10 hours of the day (so called M10 [37][38][39]), but not the least active 5 hours of the day (so called L5 [37][38][39] Assessment of the endogenous free-running circadian period reveals that reduced Nrxn1α gene expression lengthens the endogenous period in a sex-dependent manner, as alterations were observed in male ΔExon9/ΔExon9 mice and female ΔExon1/+ mice (Figs 4E and 4F and S9D; endogenous periods are reported in figure legends; S1-S3 Tables) Relatedly, examination of the phase shift of activity onsets between the L/D and D/D periods led to the observation that lower Nrxn1α gene expression is associated with a reduction in phase shift (a phase delay), meaning the onset of activity across the D/D phase is significantly shifted later in time compared to the onset of activity across the L/D phase; a sex-specific effect was observed in phase shift as well (Figs 4G and 4H, S9E; S1-S3 Tables). We next measured intradaily variability, a non-parametric measurement of rest-activity rhythm fragmentation that assesses the extent and frequency of transitions between rest and activity within a 24-hour period. High intradaily variability is indicative of an increase in activity episodes during typically restful periods like sleep, and/or an increase in rest episodes during typically active periods [37][38][39]. This type of high intradaily variability has been reported in age-related conditions, such as Alzheimer's disease [37,40,41], and in individuals associated with reduced social interaction, poor cognitive and motor performance [42,43]. We found that only mice with homozygous loss of Nrxn1α (ΔExon9/ΔExon9) exhibited increased intradaily variability during the L/D phase and/or D/D phase, particularly in females (Figs 4I and 4J; S6C, S6F; S9F and S9G; S1-S3 Tables). We also examined interdaily stability, a non-parametric measurement of how similar rest-activity patterns are from one day to the next, of rest-activity synchronization to the 24-hour light-dark cycle. This synchronization occurs through external inputs-somatosensory, social effects, and physical activity-to the main circadian oscillator, the suprachiasmatic nucleus [37][38][39]. A reduction in interdaily stability has been observed in aging, cognitive disorders, and dementia [40,44,45]. We observed that female mice with homozygous loss of Nrxn1α exhibited reduced interdaily stability, in contrast to heterozygous and WT mice (Figs 4K and 4L; S9H; interdaily stability can only be measured in the L/D phase; S1-S3 Tables).
Given the observed increase in intradaily variability, we next examined bouts of activity (Figs 5A and 5B; S10A and S10B) and found that both male and female mice with homozygous loss of Nrxn1α (ΔExon9/ΔExon9) showed an increase in the number of daily bouts, but bouts were shorter in duration with fewer revolutions per bout, particularly in the L/D phase ( Fig  5C, 5E and 5G; S1-S3 Tables). Similar alterations in bouts of activity were also found in the D/    Tables). Notably, heterozygous ΔExon1 male mice displayed an increase in the number of daily bouts with less activity per bout in only the D/D phase (Figs 5D, 5F and 5H; S8D-S8F). In contrast, mice with ΔIntron17 did not show any circadian phenotypes (S9 and S10 Figs). Taken together, we found that mice with complete loss Nrxn1α expression display robust alterations in circadian activity, a clinical comorbidity that is commonly observed in autism, as shown in multiple measurements. Mice with partial loss of Nrxn1α expression in Exon9/+ and Exon1/+ show statistically significant changes in a few measurements depending on sex. These findings underscore key roles for Nrxn1α gene dosage and sex in the manifestation and extent of circadian activity alterations.

Reduced Nrxn1α expression enhances performance on the rotarod
Given that restricted, repetitive patterns of behaviors are key diagnostic criterion in autism, we assessed the extent to which these allelic series of Nrxn1 mutant mice display alterations on the accelerated rotarod test, a motor coordination and motor skill learning task that serves as a proxy for acquired repetitive behaviors [46]. Previous studies have demonstrated that homozygous deletion of the promoter and exon 1 of Nrxn1α (ΔExon1/ΔExon1) display enhanced performance on the rotarod task [10]. Expanding upon this previous finding, we found that heterozygous ΔExon1 mice (ΔExon1/+) also exhibited enhanced rotarod performance: compared to control mice, ΔExon1/+ mice demonstrated similar performance during the initial trials (trials 1-3), but improved performance on subsequent trials in which acceleration stays the same (trials 4-6), implicating enhanced motor skill learning in these mice (Fig 6C). Notably, with increased acceleration on the high speed rotarod following two days on the rotarod with regular speed, ΔExon1/+ mice show persistently decreased latency to fall across multiple trails over two days, suggesting improved motor coordination in these heterozygous ΔExon1 mice (Fig 6D). Similarly, mice with either heterozygous or homozygous deletion of exon 9 of Nrxn1α (ΔExon9/+ and ΔExon9/ΔExon9) also exhibit enhanced motor skill learning and motor coordination (Fig 6A and 6B). In contrast, mice with either heterozygous or homozygous deletion in intron 17 of Nrxn1α do not display alterations in rotarod test (S11 Fig). Taken together, our findings suggest that the formation of repetitive motor patterns are sensitive to Nrxn1α gene dosage-with approximately 50% reduction of Nrxn1α via heterozygous deletion of either exon 9 or promoter/exon 1 leading to this phenotype, regardless of testing in males or females.

Discussion
Non-recurrent CNV deletions in NRXN1 are among the most common genetic variants in ASD, underscoring their clinical relevance. However, the extent to which specific variants, differing in their positions across the NRXN1 locus, contribute to behavioral traits associated with autism remains unknown. This study was designed to investigate the effect of three CNV deletions that vary in location across the NRXN1 locus on behavior phenotypes relevant to autism in both male and female mouse models. We found that CNV deletions affecting Nrxn1α expression, such as heterozygous ΔExon1, heterozygous or homozygous ΔExon9, but not ΔIntron17, were associated with autism-relevant behavioral phenotypes. This could potentially be related to the fact that beta isoforms are not conserved across vertebrates, in contrast to alpha isoforms which exhibit identical intronic architecture in mice and humans [2,47]. These findings align with current genetic studies reporting the majority of CNV deletions in NRXN1 identified in autism clinical cases occurs near the 5' end of the NRXN1, thus likely affecting expression of the NRXN1α isoform [2,4,6,8]. Consistently, we found that deletion of exon 9, an exon conserved in human and mice, results in a premature stop codon and triggers NMD, thus leading to disruption of Nrxn1α expression as well.
Given that in humans, homozygous deletions are rare and NRXN1 heterozygous deletions are associated with ASD clinical phenotypes, behavioral phenotypes observed in heterozygous mice are particularly relevant to translational studies. In this direction, we identified two reproducible behavioral phenotypes in heterozygous mice across two independent mouse ��� p<0.001. (H) If groups were normally distributed (D'Agostino & Pearson test) and didn't have significantly different variance (F test), the unpaired T test was used. If samples were not equally distributed or had different variances, Welch's T test was used. The number of animals used for each group in this test is the same as those in Fig 2. Data are represented as mean ± SEM in all graphs.
https://doi.org/10.1371/journal.pgen.1010659.g005 models, ΔExon1/+ and ΔExon9/+. In both heterozygous mice, Nrxn1α mRNAs were reduced to about 50%. One behavioral alteration found in this study was an impairment in social interactions, with males in both mouse models exhibiting impairments in social novelty seeking or social memory, resolving an inconsistent phenotype from previous studies [12]. The other behavioral alteration observed in this study was enhanced motor coordination, a proxy for acquired repetitive behaviors [46]. We found that this was the most robust phenotype, consistent across two mouse models and present in both males and females, particularly by the high speed rotarod test. Notably, enhanced performance on the high speed rotarod was previously reported in a cohort of mixed male and female mice on mixed Sv129/C57BL/6 genetic background with homozygous ΔExon1. This phenotype is consistently observed in our ΔExon1 and ΔExon9 mice on C57BL/6 background in both sexes. Importantly, these two behaviorsaltered social novelty preference and repetitive motor behaviors-are relevant to two core autism diagnostic domains-(1) impairment in social interaction and (2) restricted, repetitive patterns of behaviors. Thus, mice with heterozygous ΔExon1 or ΔExon9 can be valuable models for translational studies of core symptoms in autism.
Our study also documented sex-dependent behavioral traits associated with Nrxn1α deficiency in known sexually dimorphic behaviors, as homozygous ΔExon9 males exhibited enhanced aggression but females displayed reduced affiliative social behaviors. This is in line with sex-dependent behavioral responses that have been reported previously in the Nrxn1α ΔExon1 deletion models [11,12,[14][15][16]. Growing evidence suggests ASD may be under-identified and under-diagnosed in females [48], with three to four times more males being diagnosed with autism than females [49]. Our findings show that several autism-relevant behaviors differ between males and females. This agrees with prior work that has reported appreciable sex-specific clinical behaviors [50], such as the male bias towards enhanced restricted, repetitive behavior on the autism spectrum [51][52][53][54][55]. The molecular and cellular mechanisms underlying sex-specific differences in autism-relevant behaviors remains unknown, though several models have been proposed to date [55][56][57][58][59][60]. One working model relevant to our findings comes from a study by Werling and colleagues in which sexually dimorphic gene expression was assessed in neurotypical and autistic human neocortical tissues throughout development. It was found that although ASD risk genes did not display sex-differential expression, genes with sexually dimorphic expression patterns display enhanced dimorphic expression in autistic human neocortical tissue [56], suggesting that sexually dimorphic processes, circuits, and behaviors might be susceptible to autism risk variants. Our study emphasizes the importance of focusing on sex-specific behavioral phenotypes when conducting biomedical and translational studies of Nrxn1 deficiency, as well as when carrying out clinical diagnosis and measurement of treatment outcomes.
Notably, the two behavioral phenotypes observed in mice with heterozygous loss of Nrxn1α were more pronounced in mice carrying homozygous loss of Nrxn1α, suggesting a Nrxn1α gene dosage-dependent effect on these behaviors. A few other autism-related behaviors, however, were only observed with homozygous loss of Nrxn1α, such as increased aggression in males, reduced affiliative social behaviors in females, and altered circadian locomotor and bout activities in both sexes. These findings raise a high likelihood that additional genetic variants in the same or different pathways as NRXN1 may be contributing to human phenotypes. This is particularly evident in the case of an ASD individual carrying a~20kb deletion in the intron 17 of NRXN1, as mice bearing heterozygous or homozygous deletion of similar genomic region have yet to exhibit any behavioral abnormality. Though this could be related to the genomic architecture and fundamental differences between human and mice, our findings support an oligogenic model where multiple gene variants, heterozygous in nature, work together and enhance susceptibility to autism. Given that Neurexin 1α is a presynaptic adhesion molecule that plays a critical role in synapse formation, maintenance, and plasticity, variants in NRXN1 could lead to alterations in synaptic pathways, and the presence of additional genetic variants within individuals carrying NRXN1 CNVs could amplify synaptic aberration, resulting in significant perturbations of neurotransmitter release, Ca2+ signaling, glutamate receptor composition, CASK signaling, or excitatory synaptic strength, as impairments in these pathways have been reported in human induced pluripotent stem cell (iPSC) models of NRXN1 CNV, mouse models of Nrxn1 [61][62][63][64][65][66][67], and other CNVs associated with neuropsychiatric disorders such as 1q21.1 [68]. At present, two functional groups of genes, encoding synaptic proteins or epigenetic factors, are genetically linked to autism. One interpretation we previously reported was that synaptic genes, such as NRXN1, harbor broad enhancer-like chromatin domains (BELD) for the facilitation of high level and persistent transcription over time, thus mutations in chromatin genes likely disrupt the BELD feature and indirectly impair synaptic gene function [69]. Through two or more genetic hits, though each affects a single allele encoding synaptic function or chromatin regulation, the combination of both results in significant impairment at the level of synapses, and ultimately elicits autism [69].
In summary, through behavioral phenotypic analysis of an allelic series of Nrxn1α mouse models, we show that heterozygous loss of Nrxn1α elicits autism-relevant behavioral phenotypes in two core autism diagnostic domains. Various sleep traits were also found to be altered by a reduction in Nrxn1α expression, consistent with sleep disturbance comorbidity in autism. These findings support the face validity and the translational utility of Nrxn1α haploinsufficiency mouse models, which is important given that heterozygous NRXN1 variants are the most frequently observed single-gene variants associated with autism. Future work is needed to determine the molecular pathways by which Nrxn1α interacts with other gene variants to shape autism-relevant behaviors, the developmental trajectory of those autism-relevant behaviors, and the reversibility of autism-related behavioral phenotypes, at least in animal models. Such future studies may reveal new avenues for mechanism-based therapeutic development to improve the overall quality of life for autistic individuals.

Ethics statement
Animal procedures were conducted in accordance with the ethical guidelines of the National Institute of Health (NIH) and approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. This study did not include research with human subjects.

Animals
Mice were group housed in cages of 2 to 5 on a 12-hour light/12-hour dark cycle with food and water provided ad libitum. The mice used in this study for behavioral testing were between 3-5 months of age, including both males and females.
Mice with Nrxn1α promoter and exon 1 deletion (ΔExon1) were described previously [9] and have been maintained in C57BL/6J background. Mice with Nrxn1α exon 9 deletion (ΔExon9) were generated by crossing an exon 9 floxed allele of Nrxn1α (Nrxn1 tm1a(KOMP)Wtsi from MRC Mary Lyon Center, Harwell, UK) [13] with mice carrying UBC-CreERT2 [70]. An unexpected leaky activity of Cre in male gametes [33] carrying both floxed exon 9 of Nrxn1α and UBC-Cre-ERT2 leads to a germline loss of exon 9 (ΔExon9). The deletion of exon 9 was confirmed by PCR analysis using primers flanking the deleted region and within the exon 9 sequence. To study a CNV identified in an individual on the autism spectrum [27], mouse homologue of the~20 kb deleted region at intron 17 of Nrxn1α was identified and deleted using the CRISPR/Cas9-medicated genomic editing approach. Two sgRNAs (5' AATATGTGGGCAAGCTGGGT TGG 3' and 5' GAAATGGTACCTTTGATCTA AGG 3') flanking the deletion region in intron 17 of Nrxn1α were injected together with Cas9 protein into 1-cell zygote of C57BL/6J/SJL genetic background. The target deletion was confirmed by PCR and sequencing analyses using primers flanking the deleted region and deletion carriers (ΔIntron17) were back crossed to C57BL/6J for 5 more generations to collect littermates for behavioral phenotyping.
To generate experimental animals used in this study, heterozygous males were bred with heterozygous females to generate mice with homozygous (ΔExon9/ΔExon9; ΔIntron17/ ΔIntron17) or heterozygous (ΔExon9/+; ΔIntron17/+) deletions, as well as WT littermates (+/+). We noted that mice carrying homozygous deletion of exon 9 (ΔExon9/ΔExon9) were significantly underrepresented with WT:Het:Homo ratio as 48:90:29, in contrast to the expected ratio of 42:83:42, indicating sub-viability in mice carrying a complete loss of Nrxn1α. To generate mice carrying ΔExon1, heterozygous carriers of ΔExon1 were bred with WT mice to collect heterozygotes and WT for experiments described in this study.

Behavioral assessment
All animal behavioral testing was performed blinded to genotype. Mice were habituated to the testing room for at least 30 minutes before testing. Testing was performed in the same order for each animal over approximately 6 weeks (S1 Fig), beginning with the elevated zero maze assay, open-field test, social choice and preference for social novelty tests, high speed rotarod assay, circadian wheels assay, and resident-intruder test. Blood and brain tissues were collected afterwards for verification of genotypes.

Elevated zero maze
The elevated zero maze was conducted as previously described [33,71,72]. It consists of a circular shaped, partially walled maze platform. Two opposite quadrants of the maze are enclosed whereas the other two are open. Mice were placed in one of the closed quadrants and their movement traced over the course of 5 min. The movement of the animals were recorded by a high-definition digital camera for offline analysis with video tracking software (ANY-maze, Stoelting Co.). Analysis included the quantification of percent of time spent in open arms and the number of entries. An entry was defined as 90% of the mouse body within a quadrant of the maze.

Open-field test
Locomotor activity was measured similarly as previously described [33,71,72] via an openfield test where mice were individually placed into, and allowed to explore, a simple novel arena for a total of 15 min. Horizonal activity, vertical activity, and center activity was collected via infrared beam breaks. The 15 min trial was binned into 1 min epochs to assess habituation of activity as the animal became familiar to the arena.

Social choice and preference for social novelty
The social choice test was carried out in a three-chambered apparatus, as previously described [33,71,72], that consisted of a center chamber and two end chambers. Before the start of the test and in a counter-balanced sequence, one end chamber was designated the social chamber, into which a stimulus mouse would be introduced, and the other end chamber was designed the nonsocial chamber. Two identical, clear Plexiglas cylinders with multiple holes to allow for air exchange were placed in each end chamber. In the habituation phase of the test (Phase 1), the experimental mouse freely explores the arena with empty cue cylinders in place for 10 min. In the social choice phase of the test (Phase 2), an age-matched stimulus mouse (M1) (adult, gonadectomized A/J mice) was placed in the cylinder in the social chamber while an inanimate object was simultaneously placed into the other cylinder in the nonsocial chamber. The social novelty phase immediately followed. In the social novelty phase (phase 3), the object used in phase 2 was replaced by a novel mouse (M2). The experimental mouse was tracked during a 10 min trial as it explores M2 and the familiarized mouse (M1) used in the choice phase. Image analysis software (ANY-maze) was used to determine the time of cue exploration and visits to each cue (snout of experimental mouse within 1 cm of cue cylinder) in all phases. The data was verified with manual video review and scoring.

High speed rotarod
Mice were placed on an accelerating rotarod apparatus for 12 trials (three trials a day over four consecutive days) with at least 20 min of rest between the trials. On the first 2 consecutive days, mice were trained on the rod as it accelerated from 4-40 rpm, then on the next 2 days the rate was increased to 8-80 rpm [46]. Each trial lasted for a maximum of 5 min. Latency to fail in adjusting cadence (either falling from or grasping onto the rod for a full rotation) is recorded for each trial.

Circadian wheels assay
Mice are single housed with free access to running wheels as previously described [36,73]. After habituation to the new housing conditions, wheel revolutions are collected to obtain a measure of voluntary wheel running in a 12 hr/12 hr light/dark cycle. After 14-16 days, mice are maintained in constant darkness to determine their endogenous period and phase shift, in addition to wheel running activity. Wheel revolutions and non-parametric analysis was obtained and analyzed with ClockLab software (Actimetrics, Willmette, IL) [36,74].

Resident-intruder test
Similarly as previously described [75], mice were single housed for at least ten days to establish territoriality in their cages. Young intruder mice of the same sex, 6-8 weeks of age, were introduced into the experimental mouse cage followed by video recording for 10 min. Specific interactions by the resident experimental mouse are counted and timed in a 10 min trial. The specific interactions were classified as 1) Chasing/following with anogenital sniffing (cumulative time) and 2) Latency to the first aggressive behavior (grooming/barbering intruder mouse) and total duration of aggression.

Statistics
For behavioral assays, we chose similar sample sizes for all behavioral experiments. The number of mice used in each group was predetermined based on our previous studies [33,71,72]. All data sets were analyzed using the Shapiro-Wilk test for normality. For 2-sample comparisons, data sets with normal distributions were analyzed for significance using the unpaired t test, whereas data sets with non-normal distributions were analyzed using the Mann-Whitney U test. For multiple-comparisons, ordinary one-way ANOVA was performed with Tukey's multiple-comparison tests. For high speed rotarod test, two-way repeated measures ANOVA was conducted for the appropriate data sets with Dunnett's multiple-comparison tests. The quantitative RT-PCR results were analyzed using one-way ANOVA. All tests were 2-tailed. ΔExon9/+ and ΔExon9/ΔExon9 mice show normal locomotor activity in the open field test (X, Y-axis beam breaks, %Center beam breaks, and X,Y-axis beam breaks over time) in both males (E-F) and females (I-J). ΔExon1/+mice male mice exhibit slightly increased overall X,Yaxis beam breaks compared to +/+ (G), but no significant differences in %Center beam breaks or beam breaks over time (H). Female ΔExon1/+ mice show normal locomotor activity compared to +/+ (X,Y-axis beam breaks, %Center beam breaks, and X,Y-axis beam breaks over time; K-L). One-way ANOVA test with Tukey's multiple comparison test was used to analyze the ΔExon9 data and Mann-Whitney U test was used to analyze the ΔExon1 data, �� p<0.01. The number of animals used in these tests is the same as those in (A-D). Data are represented as mean ± SEM in all graphs.