UNRAVELING THE ROLE OF DE NOVO STRUCTURAL VARIANTS IN SCHIZOPHRENIA THROUGH COMPREHENSIVE WHOLE GENOME SEQUENCING WITH LONG-READ AND SHORT-READ TECHNOLOGIES

Backgrounds

Genetic liability to schizophrenia involves various types of mutations from across the allele frequency spectrum and distributed across the genome. Findings from studies focusing on different types of mutations in schizophrenia converge partially on the same biological processes, while also providing complementary insights. This underscores the importance of studying the full spectrum of mutation. However, currently widely used genotyping technologies, microarrays and short read sequencing (SRS), have limited ability in detecting medium-sized structural variations (SVs) (e.g. 50-2000bp) compared to long-read sequencing (LRS), which is relatively new and has been rarely applied to genetic studies of schizophrenia so far, suggesting an opportunity to leverage this more comprehensive approach to uncover additional sources of genetic variation that may contribute to the disorder.

Methods

Utilizing both 20X LRS and 30X SRS, we performed comprehensive whole-genome analysis on 40 Han Chinese parent-offspring trios. We called single nucleotide variants (SNVs), insertions and deletions (indels), and SVs utilizing multiple algorithms. Our primary focus was on the detection and validation of de novo mutations (DNMs). Comparative analysis between LRS and SRS was conducted to assess their respective abilities in detecting SVs and de novo SVs. Subsequently, we annotated the de novo mutations and delved into their potential mechanisms in schizophrenia through mining public databases and conducting functional experiments. Finally, we compared the diagnostic yield of our approach to previous studies employing whole exome sequencing or whole genome sequencing using SRS.

Results

Our analysis identified an average of 71.55 DNMs per proband, including 12 de novo SVs. Notably, four of these de novo SVs were detected by more than three out of four algorithms employed for LRS, whereas none were detected by any of the four algorithms utilized for SRS. In addition, our analysis revealed a 2.8Mb region exclusively accessible via by LRS and not SRS. LRS demonstrated exceptional performance in phasing, while the call sets derived from both LRS and SRS exhibited comparable levels of Mendelian consistency. Of particular interest in our study is a de novo 11kb deletion encompassing the last intron, last exon, and 3’ UTR of PPP3CA. Through experimental investigations, we discovered a significant reduction in PPP3CA protein levels in blood cells from the schizophrenia patient harboring this DNM. Similar reductions in PPP3CA protein levels were also observed in HEK293T cell lines carrying a comparable mutation, indicating that the down regulation of PPP3CA results from the identified de novo SV. Subsequently, in mice model with targeted knockdown of PPP3CA in excitatory neurons within the hippocampus, we observed alterations indicative of schizophrenia-like behavior and impaired cognitive function. Furthermore, our study revealed a slight enhancement in diagnostic yield when employing identical diagnostic criteria, as demonstrated in two comparative analyses.

Conclusion

Our findings underscore the superior performance of LRS over SRS in identifying risk mutations associated with schizophrenia. Moreover, our study implicates PPP3CA in the pathogenesis of schizophrenia, demonstrating reduced expression in excitatory neurons within the hippocampus, which correlates with schizophrenia-like behavior and impaired cognitive function.