Genetic evidence of gender difference in autism spectrum disorder supports the female-protective effect

Abstract

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder with a male-to-female prevalence of 4:1. However, the genetic mechanisms underlying this gender difference remain unclear. Mutation burden analysis, a TADA model, and co-expression and functional network analyses were performed on de novo mutations (DNMs) and corresponding candidate genes. We found that the prevalence of putative functional DNMs (loss-of-function and predicted deleterious missense mutations) in females was significantly higher than that in males, suggesting that a higher genetic load was required in females to reach the threshold for a diagnosis. We then prioritized 174 candidate genes, including 60 shared genes, 91 male-specific genes, and 23 female-specific genes. All of the three subclasses of candidate genes were significantly more frequently co-expressed in female brains than male brains, suggesting that compensation effects of the deficiency of ASD candidate genes may be more likely in females. Nevertheless, the three subclasses of candidate genes were co-expressed with each other, suggesting a convergent functional network of male and female-specific genes. Our analysis of different aspects of genetic components provides suggestive evidence supporting the female-protective effect in ASD. Moreover, further study is needed to integrate neuronal and hormonal data to elucidate the underlying gender difference in ASD.

Conflict of interest statement

The authors declare that they have no conflict of interest.

Figures

Fig. 1. Mutation load of functional classes of DNMs in the coding region.

a Mutation load per person in ASD versus control group. b Mutation load per person in male ASD subjects versus male controls. c Mutation load per person in female ASD subjects versus female controls. d Mutation load per person in male ASD subjects versus female ASD subjects. Mutation types are displayed by class. p-values were calculated by Fisher’s exact test. The “p.adjust” function in R was employed to calculate the corrected p-values for multiple comparisons, *adjusted p < 0.05, **adjusted p < 0.01, ***adjusted p < 0.001, N.S. not significant. The error bars represent 95% confidence intervals for the mean rates.

Fig. 2. Association between putative functional DNMs and clinical phenotypes in the SSC data set.

a The distribution of verbal IQ by gender. b Putative functional DNMs per person by gender in ASD patients with IQ < 80 and IQ ≥ 80. c The distribution of restricted repetitive score by gender. d Putative functional DNMs per person by gender in ASD patients with restricted repetitive score < 4 and restricted repetitive score ≥ 4. e The distribution of diagnostic month by gender. f Putative functional DNMs per person by gender in ASD patients with diagnostic month < 120 and diagnostic month ≥ 120. The p-values of comparison of distributions by gender were calculated by t test. The p-values of comparison of mutation burden by gender were calculated by Poisson test. *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant.

Fig. 3. Co-expression in the three subclasses of candidate genes across gender.

Comparisons of the number of co-expressed genes in the three subclasses of candidates genes across gender from (a) the FC subregion, including DFC (dorsolateral prefrontal cortex), MFC (anterior cingulate cortex), OFC (orbital frontal cortex), and VFC (ventrolateral prefrontal cortex); b the SC subregion, including AMY (amygdaloid complex), HIP (hippocampus), MD (mediodorsal nucleus of thalamus), and STR (striatum); c the SM subregion, including A1C (primary auditory cortex), M1C (primary motor cortex), S1C (primary somatosensory cortex), and V1C (primary visual cortex); d the TP subregion, including IPC (posteroinferior parietal cortex), ITC (inferolateral temporal cortex), and STC (posterior superior temporal cortex). p-values were calculated by the pairwise Wilcoxon test. *p < 0.05, **p< 0.01, ***p < 0.001, N.S. not significant.

Fig. 4. Functional network in sex-specific genes.

Based on co-expression data from BrainSpan, 103 candidate genes formed a large interconnected functional network, mainly involving the following major functional blocks: cell–cell communication, chromosome organization, nervous system development, regulation of cellular process, and regulation of developmental process, distinguished by different fill colors of the nodes. The sex-specific genes are marked by different border colors of nodes. Different line types between nodes represent the interactions existing in female-brain samples or male-brain samples or in both female and male-brain samples. The top right image shows the distribution of co-expressed genes (genes with |R| > 0.80) among sex-specific genes in female and male-brain samples. p-values were calculated by the pairwise Wilcoxon test.