Polygenic Transmission Disequilibrium Clarifies Common Variant Influences on Cognition in ASDs

Background:  Autism spectrum disorders (ASDs) are genetically and phenotypically heterogeneous. Recent studies have associated ASDs with classes of genome-wide polygenic risk (PRS) that have diverse influences on cognition (Bulik-Sullivan, 2015 Nature Genetics). For example, ASDs are (i) positively associated with polygenic risk for educational attainment, which itself is positively associated with cognition and (ii) positively associated with PRS for schizophrenia, which is negatively associated with cognition. While genetic correlation analyses are not expected to be transitive, this inconsistent network of associations has raised the potential of confounding and case heterogeneity.

Objectives:  The purpose of this analysis was to clarify common variant risk for ASDs, and to better understand the ASD subgroups for whom contributing polygenic risk factors are relevant. We developed a novel analytic technique called the polygenic transmission disequilibrium test (pTDT). Briefly, it is algebraically expected that, in any given set of parent-child trios, the mean of the offspring PRS distribution will equal the mean of the average parent PRS. This expectation is broken in the context of case ascertainment. For example, if a class of PRS (e.g., polygenic risk for educational attainment) is associated with ASD case status, we expect the PRS of ASD cases on average to significantly exceed average parent PRS. The pTDT is immune to confounding by ancestry or case ascertainment, and confers a substantial power advantage over traditional case-control polygenic risk scoring.

Methods:  We conducted pTDT using two ASD family cohorts. First, we used the Simons Simplex Collection (SSC), a resource of more than 2,500 families with a child diagnosed with ASD. Second, we used an independent Psychiatric Genomics Consortium ASD (PGC ASD) sample that consisted of 3,870 parent-child trios. The PGC ASD cohort described here does not include individuals from the SSC. Using a standard approach, we calculated common polygenic risk for ASDs, educational attainment (EA) and schizophrenia (SCZ) for all genotyped family members in the SSC and PGC ASD datasets.

Results:  We found that ASD cases over inherit polygenic risk for ASD, EA and SCZ (p < 1.00E-15 for each pTDT comparison), thereby unambiguously associating these classes of polygenic risk with ASD. In contrast, unaffected siblings on average inherit the expected polygenic burden for these three risk classes (p > 0.05 for each pTDT). ASD case over inheritance holds for ASD probands with and without intellectual disability (p < 0.0005 for each pTDT), as well as for probands with both intellectual disability and a de novo copy number or protein truncating variant (p < 0.002 for each pTDT). Finally, we find that EA PRS is positively associated with ASD case IQ, that SCZ PRS is negatively associated with ASD case IQ, and that ASD PRS is not associated with ASD case IQ.

Conclusions:  These analyses suggest that ASD-associated common variant risk is etiologically relevant across the IQ distribution in ASDs. Different types of polygenic risk have distinct associations with IQ, which suggests they operate through at least partially distinct biological pathways.