On January 29–30, SFARI convened the fourth annual meeting of the SFARI Sex Differences Collaboration (SSDC) in New York City. Begun in 2021, the SSDC consists of five working groups, each taking a unique, though not mutually exclusive, approach to understanding the male bias in diagnosis of autism spectrum disorders (ASD). Understanding the biological underpinnings of how male cases outnumber female cases (as high as four to one) could provide important clues to ASD itself.

Executive Vice President Kelsey Martin, Senior Scientific Officer Alan Packer and Scientific Officer Amy Norovich welcomed over 70 attendees in-person and a similar number online. These included researchers belonging to one of five working groups, as well as external advisors.

Aravinda Chakravarti of New York University introduced his group’s work, which hypothesizes that a damaged X chromosome contributes to the male bias in ASD. In this scenario, loss-of-function mutations on the X chromosome in females with ASD could help point the way to genes involved in the male bias: While a similarly deleterious mutation could be lethal for males because of their lack of a second X chromosome to buffer its effects, milder variants in these X-linked genes could alter gene expression enough to result in ASD for males. This idea has motivated the effort to comprehensively characterize the X chromosome in males and females.

Yang Sui of Evan Eichler’s lab at the University of Washington presented evidence for sex bias in genes with an excess of de novo disruptive mutations. Specifically, short-read sequencing of over 61,000 families with a child with a neurodevelopmental disorder (NDDs has identified five genes all mapping to the X chromosome where de novo mutations are significantly more likely to occur in females than in males.¹ In contrast, no male-biased genes are found on the X chromosome, which suggests that such mutations tend to be lethal in males. To understand the involvement of structural variants in the sex bias, long-read sequencing data from the entire genomes of females with autism were filtered with a human pangenome reference; this helped nominate pathogenic candidates that had been missed by short-read sequencing approaches for genes such as SYNGAP1 and MECP2, as well as other loci.²

Because hypomorphic X-linked genes that cause syndromic autism in females could contribute to the ASD sex bias in males, Huda Zoghbi of Baylor College of Medicine focused on regulatory regions of the X chromosome. Zoghbi and her team have been mapping transcription factor binding sites and their relationship to ASD variants with massively parallel reporter assays. For example, they have identified a variant in the promoter for MECP2, the gene underlying Rett syndrome, which decreases MECP2 levels in functional assays. Other regulatory elements are being systematically evaluated, including microRNA binding sites on the X chromosome. This has revealed 387 gene targets, 81 of which have some evidence supporting their association with autism. Mutations in these regulatory regions can have profound impacts: Blocking a mutated microRNA binding site in MECP2 in mice resulted in a doubling of MECP2 protein levels in the cortex. Zoghbi suggested that modifying gene expression through these regulatory elements could have translational benefits, even if boosting levels of a partially functioning protein.

Kelsey Hennick of Tomasz Nowakowski’s lab at the University of California, San Francisco used single nucleus multiome profiling to identify both differentially expressed genes and differentially accessible chromatin regions in the developing human brain. Although epigenomic and transcriptomic signatures were not correlated, she found that genes with female-enriched expression in developing human cortex are enriched for genes associated with ASD/NDDs.³ To find other means of gene control, she employed an in silico method to predict regulatory relationships of female-biased genes.⁴ This analysis identified a female-enriched regulatory network involving the transcription factor MEF2C at the center. Most of its targets are female-biased genes, and functional assays found that MEF2C did indeed bind to them.

Motivated by the prenatal testosterone surge experienced by males in utero as a way to explain the ASD sex bias, Hanna Berk-Rauch of New York University explored the effects of sex hormones on gene expression. Low levels of androgen receptors are expressed in the fetal cortex, and a similar pattern can be recreated in cortical organoids and neurons derived from pluripotent stem cells (iNeurons). When organoids or cultured neurons are exposed to androgen, subtle changes in gene expression were detected that influence progenitor cell maturation. Notably, genes whose expression was downregulated are involved in neurodevelopmental functions, whereas those that were upregulated are involved in metabolism. Androgen-upregulated genes were enriched in male-biased differentially expressed genes, which includes many genes associated with ASD.

Led by Stephan Sanders of the University of Oxford, this collaboration investigates ways in which the ASD sex bias results from an interaction between the neurobiology of ASD and sex differences in the brain mediated by sex hormone receptors. Emilie Wigdor of the University of Oxford presented her work on molecular profiling of single cells extracted from postmortem cortex of people with and without autism, subdivided by sex. Differentially expressed genes were found in those with ASD compared with controls, and these genes were enriched for known ASD- and NDD-associated genes. Sex differences were also observed, but these were underpowered given the small number of female samples. Measures of chromatin accessibility also highlighted differences between ASD and controls (differential chromatin accessibility), particularly in excitatory neurons. This highlighted the transcription factor RFX3, whose binding motifs dominated areas of differential chromatin accessibility; of note, the gene RFX3 is associated with ASD itself.

Vikaas Sohal of the University of California, San Francisco presented his findings on subcortical influences on social information processing in mice. Specifically, he found that neurons in the bed nucleus of the stria terminalis (BNST) that project onto cortical neurons enhance social information processing specifically in females, which could contribute to a female protective effect in ASD. These BNST neurons also had sex-specific effects; for example, BNST inhibition suppressed socially modulated cortical neurons in males, but increased the number of socially-modulated cortical neurons in females. To understand these differences and related to findings in the human brain, Sohal and his team have used single-nucleus transcriptomics to profile BNST-to-PFC projection neurons. They found different proportions of cell types between males and females, including a female-enriched population in which ASD-associated genes are highly expressed.

Jessica Tollkuhn of Cold Spring Harbor Laboratory and Donna Werling of the University of Wisconsin—Madison teamed up to present their research into an intersection between genetic and brain-based sex differences. Tollkuhn has been characterizing the molecular consequences of hormone changes that occur in early life and puberty. She presented data from mice in puberty that showed increases in chromatin accessibility in subcortical regions related to social behavior. Specifically in male BNST, medial amygdala (MeA), and the medial preoptic nucleus (mPOA), a wave of chromatin opening occurs, and the accessible regions contain both androgen- and estrogen-responsive elements. Her lab is also generating a multiomic reference dataset of MeA for mice. So far one highlight was a population of excitatory neurons in MeA that showed expression of ASD-associated genes.

To understand how ASD-related genes interact with sex differential regions in the brain, Werling examined ASD-related gene expression in the MeA in four mouse models of ASD. Using single-nucleus transcriptomics, she generated profiles of 51,000 MeA cells from juvenile mice. This identified cell subpopulations comparable to those found in adults. For the ADNP mouse model, some cell populations showed sex differential expression of genes that included sex chromosome genes, known co-factors for hormone receptors and ASD-associated genes. This implicates ASD risk genes in sex-differential processes within subcortical cell types important for social behavior.

Devanand Manoli of the University of California, San Francisco presented data from MeA in prairie voles, which are noted for a social attachment behavior between mates called ‘pair-bonding.’ Bonded animals prefer each other over others, and will drive off novel animals. Characterizing the molecular profiles of cells in MeA neurons with single-cell transcriptomics, Manoli reported gene expression changes after bonding. For example, an excitatory neuron subpopulation in MeA showed a dramatic change in the gene encoding corticotrophin-releasing hormone binding protein (CRHBP). Blocking this protein during pair-bonding resulted in disrupted partner preference among females. Similarly, in Shank3 heterozygous mutant prairie voles, pair-bonded females, but not males, show disrupted partner preference. This suggests that a sex-differential circuit mediating social behavior also intersects with ASD neurobiology in specific neural populations in brain regions implicated in the processing of sensory cues that drive social behavior and relationships.

Led by Joseph Dougherty of Washington University in St. Louis, this collaboration examines the interactions between genes and environment to find explanations for ASD’s sex bias. One goal has been to explore the differences in rare variant occurrence between males and females. To this end, Tychele Turner of Washington University in St. Louis developed a new and improved de novo variant (DNV) caller called HAT-FLEX that offers fast and high-quality DNV detection that is allele-specific, sex chromosome–aware, and works on human and nonhuman sequencing data alike. Applying this workflow to whole exome and whole genome data from less than 40,000 parent-child sequenced trios from SFARI cohorts (SPARK and the Simons Simplex Collection),⁵ she found that males and females with ASD showed enrichments of missense and loss-of-function DNVs in the SFARI Gene set compared with unaffected controls. At a gene-specific level, sex differences in DNV occurrence occurred in X chromosome genes like MECP2 and DDX3X (females only), as well as AUTS2 and SMAD6 (males only). Furthermore, the new DNV calling strategy also revealed RNU2-2, a noncoding RNA, as a risk factor for ASD.

Lauren Weiss of the University of California, San Francisco presented preliminary evidence for two genetic subtypes of ASD based on principal components analysis (PCA) of common autosomal genetic variation. Considering over 600,000 single nucleotide polymorphisms (SNPs) in over 6000 genomes from the SPARK cohort, Weiss reported that two clusters of cases defined by autosomal SNPs emerged in PCA. One cluster was associated with an earlier age of diagnosis and a higher male-to-female ratio, whereas the other was associated with later diagnosis, reduced male-female bias, and showed increased genetic risk for other mental health conditions, similar to a previous study.⁶ The clusters were found in other ancestries, were not due to ancestry stratification and were identified in other cohorts with different ascertainment.

Because sex hormones may modulate the effects of genetic variants to contribute to male bias in ASD, Dougherty presented a method that could obtain sex hormone levels from brain tissue with mass spectrometry. In samples from the developing human brain, this found higher testosterone levels in males than females, but no difference for progesterone or estrogen. Preliminary analysis of transcriptome data from the same samples showed correlations between certain gene transcription levels and testosterone levels, particularly in deep layer cortical neurons. The genes showing the most change with testosterone, however, did not so far include known ASD-associated genes. Experiments are underway to examine these questions via measures of chromatin accessibility, in other brain regions, other time points and with other hormones.

The Simon Baron-Cohen collaboration has focused on the prenatal origins for the sex bias in ASD. Motivated by the surge in testosterone in males that occurs in the second trimester, José González Martínez of Madeline Lancaster’s lab at the University of Cambridge has been studying the effects of testosterone exposure on human cortical organoids in vitro. These organoids are developed from induced pluripotent stem cells derived from humans carrying loss-of-function mutations in ASD-associated genes. While these organoids showed altered cell type proportions compared with controls, adding androgen accentuated these changes. For example, KDM5B organoids show a loss of interneurons and an increase in excitatory cell lineages; when exposed to androgen, these shifts were more pronounced. This result is consistent with the idea that excitatory-inhibitory imbalance drives ASD. While other monogenic organoid models did not show this kind of additive effect in cell type proportions, the addition of testosterone changed the rate of maturation and excitatory profile across most of the models.

The next talks focused on the influence of gene regulation on the ASD sex bias, as examined by Jonathan Mill of the University of Exeter and his graduate student Alice Franklin. Recently the team has generated a reference resource of genome-wide DNA methylation profiles within cell types from postmortem brain samples spanning the human lifespan.⁷ They found age-related cell type–specific DNA methylation dynamics in the prenatal cortex that differed from age-related changes in the postnatal period. With respect to ASD, Mill reported that 10 percent of ASD-associated genes were linked to sites undergoing methylation changes in the developing cortex. Sex differences in methylation profiles were also detected, but this depended on cell type and chromosome type. Franklin then addressed the DNA methylation patterns found in this dataset for four cell types (neurons, oligodendrocytes, microglia and astrocytes) with an epigenome-wide association study. While no methylation differences were found when comparing ASD samples with controls, differences became apparent when examining sex-by-diagnosis interactions. The differently methylated positions showing significant interactions were concentrated in microglia and often involved the X chromosome. The affected genes included some ASD-associated genes, as well as genes related to steroid hormone regulation and intellectual disability.

Alex Tsompanidis of the University of Cambridge presented findings on gene-steroid interactions and on clinical differences in the prenatal environment hormones between the sexes, which could ultimately contribute to the ASD sex bias. Examining the placenta for differences in gene expression between males and females, he found significant overlap with known X-linked ASD-associated genes, replicating previous findings. He also showed that autism polygenic scores of the fetus were associated with gene expression changes in the placenta, which, in turn, also overlapped significantly with genes showing baseline sex differences between male and female placentas. He also presented preliminary data from a prospective study of a cohort of Norwegian mothers and their children (MoBa).⁸ Tsompanidis obtained steroid hormone data from 2,000 samples of maternal plasma taken during pregnancy, and compared these to the child’s developmental outcome. Preliminary results showed slightly higher androgens in maternal plasma from mothers whose children later developed ASD relative to those who did not.

Led by David Page of the Whitehead Institute and the Massachusetts Institute of Technology, this collaboration explores contributions of the female-specific inactive X chromosome (Xi) and the male-specific Y chromosomes to the ASD sex bias. Page’s lab has previously shown that Xi modulates gene expression across the genome,⁹ which might underlie the female protective effect if it mitigates effects of deleterious variants.¹⁰ Given this possibility, the Page lab has been characterizing the Xi and its epigenomic features, including the extent of X inactivation. While genes on Xi can escape inactivation, the Page lab showed that their transcript levels were less than that observed on the active X chromosome. For example, in cell cultures in vitro, they found that genes expressed from Xi are in fact partially repressed, as measured by the degree of their histone modifications (both active and repressive marks) and transcription levels. Xi promoters were also depleted for active histone marks, and the amount of these mark relative to repressive marks correlated with transcription levels. This effect was also observed for ASD-associated genes on the Xi.

Gesmira Molla, a graduate student in the Page lab, presented her work parsing genetic and hormonal influences on sex differences in marmosets. Marmosets lend themselves to this question because of their sibling chimerism: Marmosets are usually born as twins, and can carry cells with their sibling’s genetic makeup due to fusing of placentas during prenatal development. Their microglial cells in the brain are chimeric, which means a male marmoset can carry XX microglia derived from his sister, and vice versa. This setup can help disentangle genetic and hormonal contributors to sex differences. Molla reported that measures of gene expression and chromatin accessibility across the genome followed patterns specific to the XX or XY status of an individual cell, regardless of the anatomical sex of the marmoset, including ASD-associated KDM6A. In contrast, other genes (e.g., CAJA-E or TGFB2) showed expression that reflected the anatomical sex of the animal.

Chris Glass of the University of California, San Diego focused on the function of MEF2C in microglia, which are particularly sensitive to their environment. MEF2C encodes a transcription factor, and MEF2C haploinsufficiency causes a syndromic form of ASD. MEF2C binding sites are enriched in active enhancer regions of the genome in microglia across multiple species, including humans. In human iPSC-derived microglia in vitro, cells lacking a copy of MEF2C take on a pro-inflammatory phenotype, reduce their phagocytosis and migration, and develop lysosomal dysfunction.¹¹ When microglia lacking both copies of MEF2C are engrafted into the brains of newly born mice to see how they will fare in an in vivo environment, native neurons take on ASD-related features such as altered neuronal density and improper synaptic connections in the hippocampus. Finally, MEF2C-dependent enhancers were enriched for motifs that suggest MEF2C works to drive TGF-β signaling.

Celina Nguyen, a graduate student in Nicole Coufal’s lab at the University of California, San Diego, addressed potential mechanisms behind Y chromosome–induced changes in microglia function. The Nicole Coufal lab has previously found that as Y chromosome dosage increases, microglia decrease phagocytosis and show lysosome dysfunction. To begin to understand the mechanisms behind this, Nguyen identified enhancers across the genome that were sensitive to Y chromosome dosage in microglia. For example, increasing Y content increased transcripts for an ASD gene called ADRB2, which was also associated with increased marks of chromatin accessibility and histone marks of active enhancers. The enhancers contained sequence motifs for several transcription factors, including those on autosomes (e.g., RFX) and on sex chromosomes (ZFY and ZFX). ZFX and ZFY binding sites overlapped with a majority of these Y-sensitive enhancers, and were enriched for ASD-associated genes.

Dougherty suggested the SSDC work collaboratively on a perspective article to review the current models of sex bias in ASD explored by the collaboration. He presented a potential outline for the piece, which focused on the biological mechanisms for the sex bias, including: a latent threshold model, prenatal sex steroid theory, damaged X, epistatic models and the Xi hypothesis. The review would address how each hypothesis would affect the brain, and summarize the state of research for each. In discussion, attendees noted that the review should not give the impression that the different models of ASD sex bias are mutually exclusive. For example, useful connections may be made between the damaged X and Xi studies, which together could help explain the female protective effect. Another idea was that the review might also try to reflect the magnitude of each hypothesis’ contribution, as currently understood.

In closing, Packer and Martin endorsed the idea of the perspective piece, and thanked everyone for attending and collaborating.

1. Hennick K. et al. bioRxiv (2025) Preprint
2. Sui Y. et al. Nat. Commun. 17, 1681 (2026) PubMed
3. Hennick K. et al. bioRxiv (2025) Preprint
4. Li Z. et al. Bioinform Adv. 3, vbad003 (2023) PubMed
5. Turner T.N. medRxiv (2026) Preprint
6. Zhang X. et al. Nature 646, 1146–1155 (2025) PubMed
7. Franklin A. et al. Cell Genom. 5, 101010 (2025) PubMed
8. https://www.fhi.no/en/ch/studies/moba/for-forskere-artikler/1.about-moba–for-researchers
9. San Roman A.K. et al. Cell Genom. 4, 100462 (2024) PubMed
10. Talukdar M. and Page D.C. Nat Genet. 58, 687–694 (2026) PubMed
11. Nguyen C. et al. Nat. Immunol. 26, 1989–2003 (2025) PubMed