No Evidence of Excess of De Novo Mutations in Autistic Children from Multiplex Families

Background:  The presence of an excess of de novo single nucleotide variants (SNVs), insertion/deletion (INDEL) and copy number variants (CNVs) has been widely reported in the analysis of exome sequencing in families with a single affected child (simplex families). The overall picture remains unclear as to whether there is a general excess of de novo variants or whether the increase in frequency is particularly marked in gene-disrupting mutations. In addition, it is not known if this excessive mutation rate is also observed in families with 2 or more affected children (multiplex families).

Objectives:  We aimed to investigate whether affected probands and affected siblings in multiplex families showed an excess of de novo mutations compared to the overall mutation rate in unaffected children. 

Methods:  We performed whole exome sequencing in 124 individuals from 26 families with 2 or more affected children (range 2-3) from the Autism Genetic Resource Exchange (AGRE). Poorly performing variant calls were identified and removed in Golden Helix SVS. Variants which caused Mendelian errors in SNV and INDEL variants were identified using PLINK. Error rates for the offspring in each pedigree were calculated in R and compared both within and across families using a T test.

Results:  Two families were removed from the analysis due to incomplete sequencing at the time of analysis or sample identity mismatches leading to incomplete family data. There were 49 affected and 22 unaffected offspring. After quality control measures to remove poor quality calls in both the SNVs and INDELs, 2.2% of all variants were inconsistent in at least one independent pedigree. Of the inconsistent variants, 15.1% of SNVs and 22.7% of INDELs generated Mendelian errors in more than one pedigree. These variants were removed as probable sequencing artifacts before analysis. In the remaining data, there were 440902 SNVs and 62262 INDELs. For the SNVs, the mean Mendelian error rate in affected offspring was 5.9x10^-5 (range 9.1x10^-6 to 4.2x10^-4) and the mean rate for unaffected offspring was 4.1x10^-5 (range 1.1x10^-5 to 1.7x10^-4). In the INDELs, the mean Mendelian error rate for affected offspring was 1.5x10^-4 (range 4.8x10^-5 to 4x10^-4) and for unaffected offspring the mean rate was 1.1x10^-4 (range 0 to 2.4x10^-4). Affected individuals did not show markedly higher rates of Mendelian inconsistency than unaffected individuals in SNVs (p=0.2) or INDELs (p=0.06).  

Conclusions:  Although de novo mutations have been reported as important in families with a single affected child, these families with multiple affected children do not seem to show this same pattern of excessive de novo variants in affected compared to unaffected individuals. In addition, some of the observed inconsistencies were seen in multiple independent pedigrees, suggesting that these are sequencing artifacts rather than genuine de novo mutations. Even in the possible de novo variants observed in only a single pedigree, it is likely that some of these will also be artifact. Work is ongoing to examine whether the pattern of de novo variants in cases is clustered more in gene disrupting mutations compared to unaffected individuals in these families.