Maternal Genetic Effects On Autism Risk: Results From the Early Markers for Autism (EMA) Study

Background:  Studies of main genetic effects in autism have failed to explain the high heritability estimates for the disease suggested by classical twin studies and by correlation of broader autism phenotypes within families. Recently, researchers have begun to study the prenatal maternal environment as a risk factor in autism.

Objectives:  We hypothesize that maternal genetic effects, including an interaction between maternal and offspring genetic variation (transgenerational epistasis), may influence susceptibility to autism.  To test our hypothesis of maternal genetic effects in autism, we examined a sample of 400 mother-child pairs where the child has been diagnosed with autism and 400 matched control pairs where the child does not have autism.

Methods:  Using the Affymetrix Axiom EUR array, we genotyped our samples at approximately 600k single nucleotide polymorphisms (SNPs) and performed copy number variant (CNV) calling. We performed two novel genome-wide association studies: first, we compared the mothers who had given birth to an autistic child to the control mothers (maternal genetic main effects) at SNPs and CNVs; second, we considered the maternal-child genotype combination at each SNP marker and performed several tests to see if the relative proportion of genotype combinations differed between case and control pairs (transgenerational epistasis). Specifically, at each locus we compared case and control cohorts using three models: 1) comparing the proportion of pairs that have identical genotypes, 2) comparing the proportion where the mother possesses an allele that her child does not possess, and 3) comparing the proportion where the child possesses an allele that the mother does not possess.     

Results:  No single SNP results were genome-wide significant (P < 5 x 10^-8). For the test of maternal genetic main effects, we found variation within or nearby the following genes to be suggestive at the P < 10^-5 level: MAML2, TMEM97, and XK. For the tests of transgenerational epistasis, we found variation within or near the following genes to be suggestive at the P < 10^-5 level in at least one of our three models: C4orf37, INSC, SLC7A8, GABRB3, GOT2, PALM2, KIAA0430, SLC14A2, SEMA3C, LY86, and CDH11. In addition, we found variation in or near the following genes previously implicated in autism to be suggestive at the P < 10^-4 level in at least one of our three models: MACROD2, NRXN1, RELN, GRIN2B, NLGN1 and PCDH9. In this sample, children with autism did not show significantly more CNVs than controls, however, we found that mothers of children with autism had significantly more deletions than control mothers (P < 0.05) and that this maternal CNV effect was replicated in data from family studies of autism, where mothers showed significantly more deletions than fathers in autism families (P< 0.05) in 2 of 3 datasets.

Conclusions:  We have assessed a novel model for maternal genetic contribution to autism risk and found evidence for increased maternal CNV burden in autism families.