Generation of genotypes and phenotypes from sequencing data

Abstract

DNA sequencing of hundreds of thousands of human samples has become a component of research biobanks over the last several years. Unlike previous efforts using genotyping arrays and whole exome sequencing, whole genome sequencing imposes no prior expectation on captured genetic information. Whole genome sequencing thus allows for the characterization of novel human polymorphisms such as large structural variation and for the identification of incidentally captured DNA sequences that are non-human in origin. When performed at scale, human structural variation can be associated with phenotypes and the presence of non-human sequences can be considered phenotypes to then be associated with human genotypes. These concepts have been employed in separate lines of work spanning retrotransposons, human DNA viruses, and the oral microbiome. Retrotransposons comprise about 45% of the human genome, but their contributions to human trait variation and evolution are only beginning to be explored. Here, we find that a sequence of SVA retrotransposon insertions in an early intron of the ASIP (agouti signaling protein) gene has probably shaped human pigmentation several times. In the UK Biobank (n = 169,641), a recent 3.3-kb SVA insertion polymorphism associated strongly with lighter skin pigmentation (0.22 [0.21–0.23] s.d.; P = 2.8 × 10−351) and increased skin cancer risk (odds ratio = 1.23 [1.18–1.27]; P = 1.3 × 10−28), appearing to underlie one of the strongest common genetic influences on these phenotypes within European populations. ASIP expression in skin displayed the same association pattern, with the SVA insertion allele exhibiting 2.2-fold (1.9–2.6) increased expression. This effect had an unusual apparent mechanism: an earlier, nonpolymorphic, human-specific SVA retrotransposon 3.9 kb upstream appeared to have caused ASIP hypofunction by nonproductive splicing, which the new (polymorphic) SVA insertion largely eliminated. Extended haplotype homozygosity indicated that the insertion allele has risen to allele frequencies up to 11% in European populations over the past several thousand years. These results indicate that a sequence of retrotransposon insertions contributed to a species-wide increase, then a local decrease, of human pigmentation. It is largely unknown which human genetic variants shape a person’s oral microbiome and potentially promote its dysbiosis. We characterized the oral microbiomes of 12,519 people by analyzing whole-genome sequencing reads from previously sequenced saliva-derived DNA. Human genetic variation at 11 loci (10 novel) associated with differences in oral microbiome composition. Nearly all of these associations implicated candidate genes with readily interpretable functions, several related to carbohydrate availability. The strongest association (p=3.0x10-188) involved the common FUT2 W154X loss-of-function variant, which associated with the abundances of 32 bacterial species. Human host genetics also appeared to powerfully shape within-species genetic variation in oral bacteria. Variation at the 11 human loci associated with variation in gene dosages in 68 regions of bacterial genomes. Several such associations implicated interactions of bacterial proteins with histo-blood group antigens presented on host mucosal cell surfaces and salivary proteins. Common, multi-allelic copy-number variation of AMY1, which encodes salivary amylase, associated with oral microbiome composition (p=1.5x10-53) and with dentures use in UK Biobank (p=5.9x10-35, n=418k), suggesting that amylase abundance impacts oral health by influencing the oral microbiome. Two other microbiome composition-associated loci, FUT2 and PITX1, also significantly associated with dentures risk, collectively nominating numerous microbial taxa that might contribute to tooth decay. Many viruses have adapted to persist in infected humans for life. Variable host control of their abundance (or load) can lead to clearance or disease. Here, we analyzed the viral load of 31 DNA viruses in human blood and saliva using whole-genome sequencing data from UK Biobank (n=490,401), SPARK (n=12,519), and All of Us (n=414,817). Viral load varied markedly with age, time of day, and season, and was higher in men than women for most viruses. Human genetic variation at dozens of genomic loci associated with load of seven viruses: Epstein-Barr virus (EBV, 45 loci), human herpesvirus 7 (HHV-7, 24 loci), HHV-6B, Merkel cell polyomavirus, and three anelloviruses. Variation at the human leukocyte antigen (HLA) complex generated the strongest associations (p = 1.1×10-8 to 9.1×10-851). Effects of HLA alleles exhibited specificity to different viruses and varied by age and body site, and HLA-B*08:01 also exhibited a host-virus genetic interaction with EBV subtype (p = 2.3×10-44). Other human genetic effects involved genes encoding proteins that process peptides for antigen presentation, such as ERAP1 (HHV-7, p = 2.7×10-78) and ERAP2 (EBV, p = 4.6×10-111). Mendelian randomization analyses indicated that although EBV infection strongly increases risk of multiple sclerosis (MS), EBV DNA load is unlikely to further modulate risk of MS (p = 0.52). In contrast, EBV viral load exhibited a strong causal effect on increased risk of Hodgkin lymphoma (OR = 19.81 [3.04–129] per s.d. increase in EBV load, p = 1.8×10-3). This suggests that higher chronic EBV viral load increases lymphoma risk, whereas associations of EBV infection with autoimmune conditions reflect host immune responses to particular viral epitopes.