Disentangling the genetic causes of autism

BackgroundBy suspecting that the scarce disease association results obtained by several genome-wide association studies, even in highly heritable diseases, could be due to wrong assumptions inferred from reference genomes but not true in affected individuals, we re-analyzed data from the Autism Genome Project consortium starting from the genotype calling phase. MethodsWe identified patterns of raw genotype intensities associated to the disease, mainly lack of hybridization. In light of it, we re-designed the polygenic models of individual risk and were able to explain most of the variance - AUC above 0.9 - in two independent datasets of whole exome sequencing (WES) provided by the Autism Sequencing and Human Autism Genetics consortia, by measuring differences in missing rates and Mendelian inconsistencies. FindingsAnalytical results when using a missingness test, pointed out to several highly significant positions, most of them having the autistic individuals as the group with the unknown variants. Results from Mendelian tests under our polygenic models, showed high accuracy and robustness as well, revealing a surprising pattern of higher rates of Mendelian inconsistencies -when considering rare variants- in unaffected offspring than in affected ones, even with affected and unaffected children belonging to the same family, and confirm autism as highly heritable. Missingness tests performed on copy number variation (cnv)s in sex chromosomes within array-based datasets, revealed that, in males, a missing cnv only in one gene (IL3RA) within pseudoautosomal region (PAR) 1 was able to explain the whole variance (AUC = 1) while in females it was also necessary but not enough another missing cnv, in gene IL9R, within PAR 2. Another cnv deletion, within a gene in chromosome Y very close to IL3RA, PRKY, is also highly associated to autism, and explains the large rates of aneuploidies in sex chromosomes of autistic individuals. Expression heatmaps performed with genes within autosomes found associated to autism, reveal cerebral cortex and testis as the regions most targeted by these genes. We also analyzed sequenced genotypes of Asperger and PDD-NOS individuals and found out that they are closer to individuals with typical development (AUC around 0.7) than to strict autistic ones (AUC above 0.9). InterpretationsAlthough the origin of autism may be associated to deletions of a cnv at IL3R in males (common macrocephaly in autism may be explained by this), the strong linkage disequilibrium with PRKY, the close interaction between autosomes and sex chromosomes through evolution and large epistatic effects, result in several associated genes in autosomes strongly expressed in cerebral cortex and testis. In light of these results, autism may be confirmed as a spectrum of traits with thousand genome-wide genetic variants but with the origin in the deletion of a cnv in IL3RA in chromosome Y (PAR1), which may cause macrocephaly. Very often it is accompanied by the deletion of cnvs in PRKY, because of its strong linkage disequilibrium with IL3RA, which may cause, at least long-range, aneuplodies. The disease is transmitted to females because of PAR1 recombination but it also needs other genes in sex chromosomes (IL3RA together with IL9R in PAR2 are not enough) for the disease onset in females. Missingness tests showing almost no genetic differences between parents and affected offspring in autosomes (AUC around 0.5) seem to confirm that autism is a complex disease that requires not only the presence of the appropriate genetic variants, thousand of them perhaps, but environmental factors for the disease onset.