Neurobehavioral Traits in Family Members Inform GENE Discovery in ASD

Background:  Autism spectrum disorder (ASD) is highly prevalent and has a complex genetic architecture. The ASD phenotype is multi-dimensional and variable.Changes in diagnostic criteria have contributed to an increased phenotypic heterogeneity. We aim to identify a narrower ASD phenotype based on core ASD features. Furthermore, it has been demonstrated that neuropsychiatric (NPD) and neurodevelopmental (NDD) disorders are part of a connected molecular system. There may be up to 4000 genes contributing to their etiology. Individuals with ASD harbor several different risk alleles and symptoms overlap across NPD and NDD.They are not single biological identities, rather a spectrum of conditions.

Objectives: We hypothetize that inherited molecular complexity featuring the presence of several NPD and NDD within the same family, together with sub clinical ASD features among family members, concentrates autistic liability across generations with common genetic drivers.

Methods:  We developed a Quantitative Autism Score (QAS) using items from the ADI-R which consistently discriminate ASD from non-ASD, occur early in development and remain stable throughout changing diagnostic criteria. We then divided our sample (520 ASD individuals) in 2 groups, according to their family history of NPD, NDD and sub clinical autistic features in first degree relatives. Our first group consisted of 185 individuals with ASD from families with a high burden of NPD, NDD and sub clinical autistic features in first degree relatives. Our comparison group consisted of 335 individuals with ASD from families with a very low or no burden of NPD, NDD or subclinical autistic features in first degree relatives. The outcome measure was the QAS score to try to capture a more homogeneous phenotypic manifestation. We conducted a SKAT-O gene-based test on WES data available; we analyzed the 2 groups separately.

Results: in the first group analyzed, LMAN1L, GREB1L and EIF4A2 were significant (p-value=9.7E-06; p-value=6.4E-05 and p-value=6.8E-05 respectively).LMAN1L was also the most significant gene when only predicted damaging variants were analyzed (P-value=2.4E-05). LMAN1L exerts its function in glycoprotein transportation and organelle targeting, an important mechanism that if defective, increases unfolded proteins within cells. EIF42A was reported to be overexpressed in fronto-temporal dementia and down-regulated in schizophrenia. In the comparison group PMFBP1, MIS18A and RPL14 were significant (p-value=3.2E-06; p-value=7.2E-05 and p-value=9.9E-05 respectively). When only predicted damaging variants were analyzed, SLC35F1 was the most significant gene (p-value=5.7E-05). PMFBP1 is involved in the general organization of the cellular cytoskeleton and localized in the 16q22,1-q22.3 CNV-enriched region in ASD cases. RPL14 is reported to interact with other genes in mice models of ASD.

Conclusions: our two groups consisted of individuals with a very homogenous phenotypic manifestation of ASD as measured by our newly developed QAS. They differed in their autistic liability derived from family history. Two different sets of genes were found significant, pointing towards the presence of different mechanisms/different genes leading to ASD. Refining the ASD phenotype using the QAS and considering family liability allowed for identification of several potential risk genes. This novel approach is the first step towards dissecting a polygenic and multi-dimensional condition to clarify its underlying biology.