Autism’s polygenic architecture can be modeled as two modestly correlated genetic factors associated with age at diagnosis, each showing distinct developmental patterns and genetic relationships to other psychiatric traits, according to a study in Nature.
Using longitudinal Strengths and Difficulties Questionnaire (SDQ) data from three birth cohorts (MCS, LSAC-B, LSAC-K; ~402 autistic participants), researchers identified two socioemotional and behavioral trajectories via growth-mixture modeling. A fourth cohort (Growing Up in Ireland Child cohort) was analyzed using latent growth-curve models. One trajectory showed early childhood difficulties that remained stable or modestly attenuated (“early-childhood-emergent”); while the other began with fewer difficulties that increased in late childhood and adolescence (“late-childhood-emergent”). In MCS and LSAC-B, those on the early-childhood-emergent trajectory were more likely to be diagnosed in childhood (χ² tests), whereas LSAC-K showed no significant difference, likely because diagnoses were only recorded from age 11.
Genome-wide association studies (GWAS) using data from iPSYCH cohort and SPARK cohorts (~ 47,130 autistic participants) showed that common variants explained ~11% of the variance in age at diagnosis across independent cohorts, with little attenuation after controlling for parental sociodemographics, clinical features, or co-occurring developmental delays/conditions. Structural equation modeling across six minimally overlapping autism GWAS, supported a correlated two-factor model (rg ≈ 0.38; s.e. ~0.07 between factors).
The earlier-diagnosed factor (Factor 1) had low but significant genetic correlations with educational attainment, cognitive aptitude, ADHD, and several mental-health/related traits. The later-diagnosed factor (Factor 2) showed a similar correlation with educational attainment but higher correlations with ADHD, depression, PTSD, childhood maltreatment, and self-harm.
The iPSYCH and SPARK age-at-diagnosis GWAS were moderately correlated (rg ≈ 0.51), consistent with differences in recruitment and age structures. Across 13 autism GWAS, genetic correlations with age at diagnosis increased as the median diagnosis age increased. In sex-stratified analyses, both factors correlated more strongly with male autism GWAS than with female, especially for the earlier-diagnosed factor.
Regression analyses across birth cohorts showed SDQ latent trajectories explained ~12% (LSAC-B) to ~30% (MCS) of the variance in age at diagnosis, compared with ~5 to ~6% explained by sociodemographic variables. Analyses of rare variation in 6,206 SPARK trios found no association between age at diagnosis and rare de novo variants or inherited protein-truncating/missense variants in constrained genes, contrasting with the significant associations observed for common variants.
Limitations included reliance on the SDQ, which is not autism-specific, the focus on individuals of European ancestry, and the modest proportion of variance explained by common single nucleotide polymorphisms. The authors also noted unmeasured factors and cross-dataset variation that may influence diagnostic timing.
One author reports a speakers’ fee from the Lundbeck Foundation; others report no competing interests (see paper for details).
Source: Nature
