Sexual dimorphism of AMBRA1-related autistic features in human and mouse

Ambra1 is linked to autophagy and neurodevelopment. Heterozygous Ambra1 deficiency induces autism-like behavior in a sexually dimorphic manner. Extraordinarily, autistic features are seen in female mice only, combined with stronger Ambra1 protein reduction in brain compared to males. However, significance of AMBRA1 for autistic phenotypes in humans and, apart from behavior, for other autism-typical features, namely early brain enlargement or increased seizure propensity, has remained unexplored. Here we show in two independent human samples that a single normal AMBRA1 genotype, the intronic SNP rs3802890-AA, is associated with autistic features in women, who also display lower AMBRA1 mRNA expression in peripheral blood mononuclear cells relative to female GG carriers. Located within a non-coding RNA, likely relevant for mRNA and protein interaction, rs3802890 (A versus G allele) may affect its stability through modification of folding, as predicted by in silico analysis. Searching for further autism-relevant characteristics in Ambra1+/− mice, we observe reduced interest of female but not male mutants regarding pheromone signals of the respective other gender in the social intellicage set-up. Moreover, altered pentylentetrazol-induced seizure propensity, an in vivo readout of neuronal excitation–inhibition dysbalance, becomes obvious exclusively in female mutants. Magnetic resonance imaging reveals mild prepubertal brain enlargement in both genders, uncoupling enhanced brain dimensions from the primarily female expression of all other autistic phenotypes investigated here. These data support a role of AMBRA1/Ambra1 partial loss-of-function genotypes for female autistic traits. Moreover, they suggest Ambra1 heterozygous mice as a novel multifaceted and construct-valid genetic mouse model for female autism.


INTRODUCTION
Autism spectrum disorders (ASD) are extremely heterogeneous neurodevelopmental conditions, affecting~1% of the population. Typical, shared symptoms range from social communication and interaction deficits, including decreased attraction by and compromised reading of social signals, restricted interests, repetitive behaviors or pronounced routines, to reduced cognitive flexibility. 1-4 Early brain enlargement [5][6][7][8] and predisposition to epileptic seizures 9,10 are among the reported non-behavioral characteristics found in this disorder category. Causes likely converge at the synapse, as indicated by mutations of synaptic genes or by mutations causing quantitative alterations in synaptic protein expression, half-life or degradation, and are reflected by a virtually autism-pathognomonic neuronal excitation-inhibition dysbalance. 4,[11][12][13][14][15] Neuroligin-4 mutations, for instance, belong to the most common causes of monogenetic heritable autism. 16 Construct-valid and face-valid mouse models of autism, building on monogenetic grounds, have helped in approaching the underlying common biology. 17,18 The estimated heritability of autism approximates 90%. We note, however, that monogenetic forms including copy number variations altogether account for o 20%, leaving~80% of cases unexplained, which also enter the final common pathway of disease expression. [1][2][3] Importantly, normal genetic variants, mainly single nucleotide polymorphisms (SNPs), likely contribute to the manifestation of autistic phenotypes. This is indicated by the results of genome-wide association (GWAS) and respective polygenic-risk studies on autism, 2,19,20 but even more so by phenotype-based genetic association studies (PGAS), reporting an accumulation of 'unfortunate' normal variants, so-called proautistic genotypes, to be associated with increasing severity of autistic traits. [21][22][23] In fact, phenotypical continua of autistic features from health to disease suggest underlying mechanisms of quantitative rather than qualitative nature. 21,24 Genetic modifiers, like protective genes and environmental (co-)factors, mainly those acting during uterine and early postnatal development may also modulate autism severity. 25,26 ASD has an overall gender distribution of~4:1 males/ females. 1,2,27 Remarkably, little research has focused on the reasons for this disparity. Better understanding of this difference, however, could lead to major advancements in the prevention or treatment of ASD in both genders. 28 We previously reported that an Ambra1 (activating molecule in Beclin1-regulated autophagy) partial loss-of-function genotype is associated with the autism-like behavior in female mice. The restriction to the female gender of autism generation by a defined genetic trait has thus far remained unique. 29 Ambra1 is a positive regulator of a principal player in autophagosome formation, Beclin1. Importantly, autophagy has already been linked to autism in recent work. 30,31 Moreover, Ambra1 is involved in other developmentally relevant processes in the nervous system and in neuronal function. 32,33 Although homozygosity of the Ambra1 null mutation causes embryonic lethality, heterozygous mice with reduced Ambra1 expression appear completely normal at first view. 33 Only upon comprehensive behavioral characterization, a striking autism-like phenotype of Ambra1 +/ − females emerges. This trait is quantifiable by the autism severity composite score, which even allows a behaviorbased genotype predictability of 490%. 29,34 As first mechanistic hint explaining the prominent gender difference, stronger reduction of Ambra1 protein in the cortex of Ambra1 +/ − females was found. 29 Until now, no association of AMBRA1 genotypes with autistic features has been described in humans, therefore still questioning the construct-validity of our mouse model. 29 However, a recent GWAS on schizophrenia identified a genetic risk variation on chromosome 11 (11p11.2) in a region containing AMBRA1. 35 Schizophrenia and ASD show considerable syndromic overlap, including deficits in social cognition and communication, 24,36 and at least a subgroup of schizophrenia is also regarded as a disorder of the synapse. 37 The present study has therefore been designed to explore whether any autism-relevant phenotype association with normal AMBRA1 genotypes would emerge in humans, thereby supporting the construct-validity of our Ambra1 +/ − mouse model. In addition, we aimed at defining potential further characteristics of ASD in Ambra1 +/ − mice, namely (1) decreased interest in social odors, as highly relevant social signals in mice, (2) increased epileptic predisposition as in vivo readout of neuronal excitation-inhibition dysbalance and (3) early brain enlargement, as recognized in human autism. Indeed, we show here that also in humans, an AMBRA1 genotype, the intronic SNP rs3802890-AA, located in a long non-coding (lnc) RNA, is linked to autistic features and characterized by partial loss-of-function in females. Moreover, we demonstrate in Ambra1 +/ − mice prepubertal brain enlargement. Only in female mutants, we see loss of interest in sex pheromones and altered seizure propensity.

MATERIALS AND METHODS
In all experiments, the experimenters were unaware of genotypes ('fully blinded').

Human studies
Discovery: schizophrenia subjects and healthy controls (GRAS). The Göttingen Research Association for Schizophrenia (GRAS) data collection consists of 41200 deep-phenotyped patients, diagnosed with schizophrenia or schizoaffective disorder (DSMIV-TR 38 ), recruited across Germany since 2005. 39,40 Diagnosis is based on a comprehensive examination, lasting for at least 4 h (the examination often took much longer, with breaks in between, dependent on the patient's condition). It is guided by the GRAS manual, which contains standardized interviews, psychopathology and neuropsychology testing. Moreover, careful study of all the medical discharge letters and charts of every single individual aids in assessing longitudinal aspects of the diagnosis as well. 39,40  Phenotyping. For quantification of autistic phenotypes, we used the Positive and Negative Syndrome Scale (PANSS) 42 -based autism severity score (PAUSS) 24 with slight modifications (Figure 1a), available for 1067 patients. For the replication sample, the Instrumental Support Index (ISI) was taken as proxy, indicating quality of instrumental and emotional support, 43 expected to be low in autistic individuals. 23 It was crossvalidated with PAUSS in GRAS subjects, with social support operationalized as self-reported number of individuals a person can rely on in case of emergency. 23 For both measures of social support (intercorrelation 0.77), higher score values (z-transformed; range: 1.5-9) represent higher social support, that is, lower autistic features.
Based hereon, only rs3802890-A/G remained for analysis.
In silico analyses. Genome sequences were established according to latest available releases (human-genome vs32-2015; mouse-genome 2016). Iterative sensitive sequence comparisons were conducted 47 and evaluated including detailed genome and transcriptome mapping. Expression of the rs3802890-containing RNA region was derived from latest largest collection of ESTs available at NIH. 48 LncRNA matches were also established according to latest human lncRNA release at NCBI. RNA folding used RNAfold. 49 For demonstrating rs3802890-A/G differences, thermodynamic ensemble structures drawing encoded base-pair probabilities were used. Protein binding regions were calculated using RNAanalyzer 50 and CatRapid, 51 coding potential was calculated using Genscan. 52 AMBRA1 and EST TCAAP2E6309 mRNA expression. Peripheral blood mononuclear cells (PBMC) were isolated from morning blood, obtained via phlebotomy into CPDA-vials (Citrate-Phosphate-Dextrose-Adenine, Sarstedt, Germany), applying standard Ficoll-Paque-Plus isolation procedure (GE-Healthcare, Munich, Germany). Total RNA extraction was done using miRNeasy Mini-kit (Qiagen, Hilden, Germany). For reverse transcription, 1 μg of cDNA was applied using a mixture of oligo(dT)/hexamers, dNTPs, DTT and 200U SuperscriptIII (Life Technologies, Darmstadt, Sexual dimorphism of AMBRA1-associated autistic traits M Mitjans et al Germany). AMBRA1 RNA expression was measured using quantitative realtime PCR. The cDNA was diluted 1:25 in 10 μl reaction-mix, containing 5 μl of SYBR-green (Life Technologies) and 1pmol/primer: Technical triplicates were run on LightCycler480 (Roche-Diagnostics, Mannheim, Germany). Relative AMBRA1 expression was calculated using the threshold-cycle method (LightCycler480 Software1.5.0SP3-Roche) and normalization to GAPDH. EST TCAAP2E6309 RNA expression was measured using traditional PCR. Extracted RNA, cDNA synthesized with oligo-dT primers with/without hexamers, or genomic DNA were used as template with the following primers: EST TCAAP2E6309-Fw: 5′-GGCAGAGCAGAATGGATAGACA-3′ EST TCAAP2E6309-Rv: 5′-AACGCCTGTTATCTGGGATCA-3′ Mouse studies Investigations were carried out in agreement with guidelines for welfare of experimental animals, issued by the Federal Government of Germany and Max Planck Society, approved by local animal care and use committee (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, Oldenburg, Germany).
Mouse line and housing. Ambra1 mutation in mice is caused by a truncated, non-functional Ambra1 protein via insertion of a gene-trapping vector into the murine Ambra1 gene. 33 Ambra1 wild-type (WT, Ambra1 +/+ ) and heterozygous (Ambra1 +/ − ) littermates of both genders with 499% C57BL/6 N background were used (male Ambra1 +/ − × female WT-C57BL/ 6N). Genotyping was performed as described. 29   C3H mice of opposite gender and another box with only fresh bedding. WT mice typically prefer used bedding containing pheromones.
Magnetic resonance imaging for morphometry. Mice were anesthetized (5% isoflurane), intubated and kept at 1.75% isoflurane/5% oxygen by active ventilation with constant respiratory frequency (85 breaths per minute; Animal-Respirator-Advanced, TSE-Systems). Magnetic resonance imaging (MRI) was performed at 7 and 9.4T (Bruker Biospin MRI, Ettlingen, Germany). Radiofrequency excitation and signal reception were accomplished with use of a birdcage resonator (inner diameter, 72 mm) and a four-channel phased-array surface-coil, respectively. T2-weighted MRI data were acquired with three-dimensional fast spin-echo MRI sequence (repetition-time TR = 3.5 s, effective echo-time TE eff = 55 ms, 12 differently phase-encoded echoes, 56 min measuring time) at isotropic spatial resolution of 100 μm. From these datasets, polygonal surface models of selected brain structures were generated by importing DICOM images into AMIRA (Visage-Imaging, Berlin, Germany). Structures of interest (whole brain, hippocampus, cerebellum, olfactory bulb, ventricles) were manually and semi-automatically labeled with segmentation editor on three-dimensional label fields (80 horizontal, 192 coronal, 144 sagittal slices).

RESULTS
A normal AMBRA1 genotype, rs3802890-AA, is associated with autistic traits predominantly in female schizophrenic individuals Only one directly genotyped and-according to our PGAS SOP 23 -suitable SNP, AMBRA1-rs3802890-A/G, was available in our array. Thus, rs3802890 is not associated with the schizophrenia diagnosis. Next, PGAS was performed with rs3802890-AA/-GG and PAUSS 24 as quantitative measure of autistic traits (Figure 1a). In previous work, we have demonstrated that autistic features cross diagnostic borders and can be quantified not only in ASD, but also in schizophrenia and other diseases as well as in healthy individuals. 21,23,24 For quantification, we developed the PAUSS, a dimensional instrument based on PANSS, 42 capturing the continuous nature of autistic behaviors. 24 PGAS revealed an association: AA carriers display higher PAUSS scores than GG subjects (P = 0.039). Interestingly, when separating genders, the PAUSS association remains significant only for females and, likewise, most PAUSS sub-items show this trend in females but not males (Figures 1b and c).
A role of AMBRA1-rs3802890-AA for female autistic features is confirmed in a general population sample Even though this highly targeted approach to an association of only one available AMBRA1 SNP with autistic traits in schizophrenic individuals was already encouraging, we aimed at replication of this finding in a general population sample. For this, a social support score, derived from ISI, 43 was used as proxy phenotype, expected to be low in individuals with autistic features. 23 Crossvalidation of social support (operationalized as the self-reported number of individuals a person can rely on in case of emergency) with PAUSS in the discovery sample (GRAS) yielded a high negative Spearman rank correlation (Figure 1d), underlining the relevance of this proxy for autistic features. Again, the social support score disclosed a genotype effect (rs3802890-AA/-GG) in both discovery and replication sample, more pronounced in females (Figures 1e and f). Thus, in two independent human cohorts, a single normal variant in the AMBRA1 gene, rs3802890, is associated with autism-related behaviors predominantly in females.
Consequence of AMBRA1-rs3802890-A versus G on mRNA expression in human PBMC and in silico prediction of potential underlying mechanisms In some subjects, PBMCs were available for AMBRA1 mRNA analysis. Although female GG (N = 14) versus male GG (N = 33) carriers had higher expression levels (0.0059 versus 0.0045 AMBRA1/GAPDH; P = 0.05), AA carriers of both genders (males N = 35; females N = 33) did not differ in their level (males 0.0049; females 0.0046; P = 0.62), which was comparable to that of male GG carriers. Comparing both genders, AMBRA1 mRNA expression in PBMC of AA (risk SNP) relative to GG carriers is reduced in females but not males (P = 0.017; Figure 1g), possibly indicating partial loss-of-function of AMBRA1 in AA females. This relative reduction found in PBMC of women resembles the situation in cortex of female mice: normal WT females have higher Ambra1 expression than WT males, whereas Ambra1 +/ − females show stronger relative Ambra1 reduction compared to Ambra1 +/ − males. 29 A detailed map of human AMBRA1 gene is explained in Figure 2a. Exploring the location of the intronic rs3802890, a similarity to expressed sequence tags (EST) from NCBI database arises (Figure 2a). The predicted RNA folding of the transcribed EST TCAAP2E6309, covering the SNP region, is remarkably influenced by the presence in rs3802890 of G versus A (Figure 2b). As we find EST TCAAP2E6309 RNA expressed in PBMC and other human tissues (Figure 2c), relevance of this lncRNA for AMBRA1 mRNA or protein levels may be assumed.
Together, these data suggest that AMBRA1 likely shapes autistic behavior also in humans in a sexually dimorphic way. This acrossspecies unique female autism generation by a defined genetic trait that appears to cause partial loss-of-function encouraged us to continue searching for further autism-specific readouts in our Ambra1 +/ − mouse model.
Pheromone-based social preference is reduced in Ambra1 +/ − females Sex pheromones have an important role in social behavior throughout the animal kingdom. [59][60][61][62] Mice typically favor a social context that contains pheromones of the opposite gender. In autistic phenotypes, social interest, approach and communication as well as understanding of social signals are compromised. [1][2][3][4]62 We therefore designed a novel intellicage set-up to test pheromone preference as potential autism-relevant readout in Sexual dimorphism of AMBRA1-associated autistic traits M Mitjans et al Ambra1 +/ − mice. Upon free choice between a social box containing used bedding from mice of the opposite gender and another social box with fresh bedding only, WT male and female and Ambra1 +/ − male mice behave as expected, namely choose to stay longer in the respective 'pheromone box'. In contrast, Ambra1 +/ − females fail to show this preference (Figure 3).
MRI analysis reveals brain enlargement in Ambra1 +/ − mice Brain enlargement has been described both in children and adults with ASD. [5][6][7][8] Recently, brain volume overgrowth in children was linked to the emergence and severity of autistic social deficits. 7 We measured by high-resolution MRI (T2-weighted) brain dimensions in Ambra1 +/ − versus WT mice. Whole brain and hippocampus were enlarged in male and female mutants at postnatal day 23 (around puberty). Cerebellum was increased in female Ambra1 +/ − mice only (Figure 4). Sizes of olfactory bulb and ventricles in Ambra1 +/ − mice were similar to WT. Repeated examination of females at age 13 months revealed persistence of the increased brain dimensions (Figure 5a). We note that Ambra1 +/ − mice are the first autism model showing autism-typical brain enlargement. Regarding this particular readout, genders were comparable, uncoupling in this model autism-like behavior (only females) from brain dimensions.
Female Ambra1 +/ − mice show altered seizure propensity Another frequently observed trait, connected with autistic behaviors, not only in syndromic forms of autism, is epileptic seizures. 9,10 Most likely, seizure predisposition reflects the autismpathognomonic neuronal excitation-inhibition dysbalance. 4,[11][12][13][14][15] In our Ambra1 +/ − model, prepubertal female mutants displayed reduced response to PTZ, namely longer latency to the first wholebody seizure and decreased seizure score compared to female WT (Figure 5b). This early resistance turned into the opposite response at older age: Number of tonic-clonic seizures and duration of whole-body seizures were enhanced in 3-months (data not shown) and 13-months-old Ambra1 +/ − females versus WT, also resulting in reduced survival (Figures 5c and d). Male mutants did not differ from WT at any time point investigated. Together, these data support Ambra1 +/ − mice as multidimensional model of human autism.

DISCUSSION
We previously reported in female Ambra1 +/ − mice a discrete behavioral trait, reminiscent of human ASD. 29 In the present study, we extend this finding, showing for we believe the first time that AMBRA1 may-in likewise sexually dimorphic manner-be relevant also in humans for the expression of a female autistic a b c   phenotype. The female preponderance, unique thus far in autism genetics, may even help illuminating some general molecular underpinnings of gender susceptibility to brain disease. Remarkably, in a highly targeted association approach, using the only AMBRA1 SNP available for analysis, rs3802890-A/G, we find in two independent populations, the GRAS sample of schizophrenic individuals and the SHIP sample of general population subjects, relevance for this marker regarding autistic traits in women. Partial loss-of-function, reflected by a relative decrease in AMBRA1 mRNA levels in PBMC of female risk genotype (AA) carriers, may suggest an underlying autism-causing mechanism similar to that in heterozygous mice where Ambra1 reduction was stronger in female than male mutant cortex. 29 The question of how AMBRA1/Ambra1 reduction may influence synaptic function, thereby causing the autism-pathognomonic neuronal excitation-inhibition dysbalance, 4,11-15 still remains unanswered. We may, however, speculate that reduced autophagy at synaptic terminals, 30,31,63 likely more pronounced in Ambra1 +/ − females, 29 influences synaptic protein turnover and function in a gender-specific manner. In fact, females may be particularly sensitive to reduced autophagy as suggested also by a recent paper reporting higher basal autophagy activity in the brain of neonatal female as compared to male rats. 64 In this sense, AMBRA1/Ambra1 adds to the number of proteins shown to underlie sexually dimorphic effects on the brain. 65,66 In a first in silico search for mechanisms, we saw that the lncRNA, covering the SNP region, shows highly diverse folding upon presence of G versus A allele. This pronounced structural effect may influence AMBRA1 mRNA and/or protein stability and will be subject for further investigation.
Returning to the Ambra1 +/ − mouse model, we extend our earlier findings 29 to crucial, additional autistic features, so far not systematically addressed in genetic models of autism, namely early brain enlargement, altered propensity towards epileptic seizures and reduced pheromone preference.
Brain enlargement is a consistently reported feature in human autism, both in adults and children. [5][6][7][8] Already upon first description of autism, increased size of the head was observed in affected children. 67 The substrate underlying the enlarged brain has remained obscure, and the Ambra1 +/ − mouse model may now help to approach this question. Interestingly, we found Ambra1 +/ − -associated brain enlargement in both genders, thus uncoupled from the predominantly female behavioral phenotype. This finding may be important in connection with recent suggestions, based on genetically undefined autistic children, where brain volume overgrowth was linked to the emergence and severity of autistic social deficits, predicting a later autism diagnosis based on MRI deep-learning algorithms. 7 This obvious discrepancy should stimulate further investigations considering gender, genetics and biological ASD subgroups.
Olfactory deficits in human autism have been reported, even though the literature is scarce, heterogeneous and inconsistent, likely reflecting subject selection and assessment biases or other methodological limitations, including statistical power issues. 68 We note, however, that compared to humans, pheromone preference in mice represents a more prominent component of their social behavioral repertoire, thus more vulnerable to be disturbed in autism-like phenotypes of this species.
To conclude, our data suggest a fascinating sexual dimorphism regarding the role of the autosomal AMBRA1/Ambra1 gene for autistic traits across species. In humans, it will for instance be interesting to systematically screen ASD patients for AMBRA1 mutations, particularly female autists. Ambra1 +/ − mice may serve as a novel multilayered construct-valid genetic model of human autistic phenotypes.

CONFLICT OF INTEREST
The authors declare no conflict of interest.