Autism Spectrum Disorder (ASD) is a severe neurodevelopmental disorder characterized by impairment in social interaction, language deficits and repetitive and restricted behaviors. ASD has a very strong genetic factor. However, the underlying genetic mechanism is still elusive. In addition, genetics of ASD in East Asian populations has not been well explored and there is no reported genome wide genetic study for ASD in East Asian populations. To address these issues, in the current study, a two-stage genome wide association study (GWAS) and genome wide copy number variation (CNV) study were conducted in East Asian populations.

In the discovery stage of GWAS, 166 ASD families (500 subjects) and 425 healthy controls were genotyped with the Affymetrix Genome-Wide Human SNP array 6.0 chip (Affymetrix, Santa Clara, CA, USA). A novel method which combines the P value of family based association study (FBAT) and case-control was used in prioritizing SNPs for replication study. A total of 86 SNPs were selected for replication and genotyped in independent 205 Japanese ASD trios and 410 Taiwanese ASD trios. A number of nominal associations were observed within or near genes including GPC6, NTN4, JARID2, AIM2, ROR2, CNTN4. No single SNP was found to be associated with ASD at a genome-wide significance level (P=5×10-8) in both FBAT and case-control analysis. Furthermore no SNP was replicated either in the Japanese or the Taiwanese trios. The results indicated that ASD has a very complicated genetic architecture and the current sample size might not be sufficient to detect common variants with modest effect.

In the genome wide CNV study, after stringent quality control, a total of 158 cases, 306 parents and 349 controls from the Japanese population were used for CNV calling by three algorithms including PennCNV, QuantiSNP and Birdsuite. Ten putative rare de novo CNVs were identified and were validated by qPCR assay. Eight CNVs out of 10 from 8 ASD probands were confirmed. In which, a 5Mb de novo 15q11.2 duplication was found. The frequency of this CNV is 1/158 in the Japanese population, which is similar to the frequency reported in Caucasians (1%). Two novel genes were indentified in the de novo loci and have critical functions in central nervous system. These two genes may serve as novel ASD candidate genes for further functional study. In addition, 8 Inherited CNVs (5 duplication and 3 deletions) were found overlap with known ASD loci or ASD candidate genes. Of these CNVs, 16p11.2, NLGN4X and MCPH1 are already well-established in association with ASD. 22q11.21 and 17p12 have been reported in a variety of neuropsychiatric disorders including ASD. PRKCB1 and SLC25A12 were both ASD candidate genes but no CNV have been previously reported affecting these two genes. The current CNV study confirmed a number of known ASD loci and also identified several novel de novo loci which may contain ASD candidate genes.

In conclusion, GWAS and genome-wide CNV analyses were done to explore the genetic architecture of ASD in East Asian populations. The results suggest that common variations may have a more subtle role in the pathogenesis of ASD than expected. Also the lack of association is likely to have been caused by insufficient statistical power of the small sample size. Future GWAS with large sample size were recommended to confirm the findings of the current study. By CNV analysis, a number of rare inherited or de novo CNVs have been identified in ASD patients. The results confirmed the significant contribution of CNV in the etiology of ASD and also identified novel loci which worth further study. These results provide a comprehensive genetic landscape of ASD in East Asian populations and also shed new light into the pathogenesis of ASD.

To better understand the underlying genetic architecture of autism spectrum disorder (ASD), especially in East Asian populations, a two-stage genome wide association study (GWAS) and genome wide copy number variation (CNV) study were performed using 2780 individuals. The following results were obtained.

(1) In the discovery stage of GWAS, after stringent sample quality control procedures, a total of 156 cases and 427 controls were used for case-control analysis and 156 cases and 295 parents were available for family based association test (FBAT) analysis. After SNP quality control, 588,073 SNPs remained for further analysis. The estimated inflation factor (λ) was 1.0069 and 1.0186 for FBAT and for case-control respectively, which indicated negligible population stratification. A number of nominal associations were observed within or near genes including GPC6, NTN4, JARID2, AIM2, ROR2, CNTN4. Of these genes, JARID2 and CNTN4 are known ASD candidate genes. The other genes might be novel ASD candidate genes. No single SNP was found to be associated with ASD at a genome-wide significance level (P=5×10-8) in both FBAT and case-control analysis.

(2) After the integration of case-control and FBAT, a combined P value was re-assigned to each SNP. Based on the ranking of the combined P value and other criteria, 86 SNPs were chosen for replication in the second stage in independent Japanese ASD trios (n=205) and Taiwanese ASD trios (n=410). No SNP was replicated either in the Japanese or the Taiwanese trio set. The results indicated that ASD has a very complicated genetic architecture. The lack of association was also likely to be caused by insufficient sample size and limited probe density of the genotyping platform.

(3) In the CNV analysis, three algorithms including PennCNV, QuantiSNP and Birdsuite were utilized for CNV calling to increase detection accuracy and to reduce false positive calls. A total of 27,720 CNVs were detected by all three algorithms in all subjects including ASD family samples and controls. Subsequently in a total of 5,653 CNVs identified in probands, 311 putative CNVs were suggested as de novo by the trio calling algorithm in PennCNV software (5.5%). Ten CNVs were not present in the Database of Genomic Variants (DGV hg18 ver.10) and were regarded as rare de novo CNVs. All 10 de novo CNVs were validated by SYBR Green-based quantitative PCR assay. Eight CNVs out of 10 from 8 ASD probands were confirmed (80%). Besides the well-known 15q11.2-13.1 duplication, other 7 regions were identified and experimentally confirmed, providing novel clues to the pathogenesis of ASD. Two novel genes Gene A and Gene B were indentified in the de novo loci and serve as novel ASD candidate genes for further functional study.

(4) Eight inherited CNVs (5 duplication and 3 deletions) in ASD probands were found to overlap with known ASD loci or ASD candidate genes. The maternal transmissions were observed three times more frequently than paternal transmissions (6:2), which suggests that females might be protected from such genetic risks, in a way unknown yet. Of these CNVs, 16p11.2, NLGN4X and MCPH1 have already been well-established to be associated with ASD. 22q11.21 and 17p12 have been reported in a variety of neuropsychiatric disorders including ASD. PRKCB1 and SLC25A12 were both ASD candidate genes but no CNVs affecting these two genes have previously been reported. Of all the above CNVs, NLGN4X is of particular interest. The deletion was transmitted from unaffected mother to the male proband, which is in an X-linked recessive inheritance mode. To confirm this deletion, PCR was carried out and NLGN4X deletion was confirmed in the proband, which indicated that a female could be a risk carrier of ASD. Collectively, 19 out of 158 patients were found to carry a potential ASD susceptible CNV (12%). The results confirmed the significant contribution of CNV in the etiology of ASD.

In conclusion, comprehensive analysis was conducted in this study to understand the genetic underpinning for ASD. The study provided a detailed genetic landscape of ASD in East Asian populations for the first time. The results from GWAS in this study, combined with previous GWAS findings, indicate that common variants captured by the current genotyping platform may play a more subtle role in the etiology of ASD than expected. In contrast, CNVs were found with a more critical role in the pathogenesis of ASD. The data generated in this study provide a solid basis for future study and contributes important new data to the ASD scientific research community. Considering the importance and novelty of the work, it is worthy of the award of Doctor of Philosophy to the candidate.