Noninvasive whole-genome sequencing of a human fetus

Abstract

Analysis of cell-free fetal DNA in maternal plasma holds promise for the development of noninvasive prenatal genetic diagnostics. Previous studies have been restricted to detection of fetal trisomies, to specific paternally inherited mutations, or to genotyping common polymorphisms using material obtained invasively, for example, through chorionic villus sampling. Here, we combine genome sequencing of two parents, genome-wide maternal haplotyping, and deep sequencing of maternal plasma DNA to noninvasively determine the genome sequence of a human fetus at 18.5 weeks of gestation. Inheritance was predicted at 2.8 × 10(6) parental heterozygous sites with 98.1% accuracy. Furthermore, 39 of 44 de novo point mutations in the fetal genome were detected, albeit with limited specificity. Subsampling these data and analyzing a second family trio by the same approach indicate that parental haplotype blocks of ~300 kilo-base pairs combined with shallow sequencing of maternal plasma DNA is sufficient to substantially determine the inherited complement of a fetal genome. However, ultradeep sequencing of maternal plasma DNA is necessary for the practical detection of fetal de novo mutations genome-wide. Although technical and analytical challenges remain, we anticipate that noninvasive analysis of inherited variation and de novo mutations in fetal genomes will facilitate prenatal diagnosis of both recessive and dominant Mendelian disorders.

Fig. 1

Experimental approach. (A). Schematic of sequenced individuals in a family trio. Maternal plasma sequences were ~13% fetal-derived based on read depth at chrY and alleles specific to each parent. (B). Inheritance of maternally heterozygous alleles inferred using long haplotype blocks. Among plasma sequences, maternal-specific alleles are more abundant when transmitted (expected 50% versus 43.5%), but there is substantial overlap between the distributions of allele frequencies when taking considering sites in isolation (left histogram, yellow=shared allele transmitted, green=maternal-specific allele transmitted). Taking average allele balances across haplotype blocks (right histogram) provides much greater separation, permitting more accurate inference of maternally transmitted alleles. (C). Histogram of fractional read depth among plasma data at paternal-specific heterozygous sites. In the overwhelming majority of cases when the allele specific to the father was not detected, the opposite allele had been transmitted (96.8%, n=561,552). (D). De novo missense mutation in the gene ACMSD detected in 3 of 93 maternal plasma reads and later validated by PCR and resequencing. The mutation, which is not observed in dbSNP nor among >4,000 individuals’ coding exons sequenced as part of the NHLBI Exome Sequencing Project (http://evs.gs.washington.edu), creates a leucine-to-proline substitution at a site conserved across all aligned mammalian genomes (UCSC Genome Browser) in a gene implicated in Parkinson’s disease by genome-wide association studies (24).

Fig. 2

Accuracy of fetal genotype inference from maternal plasma sequencing. Accuracy is shown for paternal-only heterozygous sites, and for phased maternal-only heterozygous sites, either using maternal phase information (black) or instead predicting inheritance on a site-by-site basis (gray).

Fig. 3

HMM-based predictions correctly predict maternally transmitted alleles across ~1 Mbp on chromosome 10, despite site-to-site variability of allelic representation among maternal plasma sequences (red).

Fig. 4

HMM-based detection of recombination events and haplotype assembly switch errors. A maternal haplotype block of 917 Kbp on chromosome 12q is shown, with red points representing the frequency of haplotype A alleles among plasma reads, and the black line indicating the posterior probability of transmission for haplotype A computed by the HMM at each site. A block-wide odds ratio test (OR) predicts transmission of the entire haplotype B, resulting in incorrect prediction at 272 of 587 sites (46.3%). The HMM predicts a switch between chromosomal coordinates 115,955,900 and 115,978,082, and predicts transmission of haplotype B alleles from the centromeric end of the block to the switch point, and haplotype A alleles thereafter, resulting in correct predictions at all 587 sites. All three overlapping informative clones support the given maternal phasing of the SNPs adjacent to the switch site, suggesting that the switch predicted by the HMM results from a maternal recombination event rather than an error of haplotype assembly.

Fig. 5

Simulation of effects of reduced coverage, haplotype length, and fetal DNA concentration on fetal genotype inference accuracy, defined as the percentage of sites at which the inherited allele was correctly identified out of all sites where prediction was attempted. Heatmaps of accuracy after in silico fragmentation of haplotype blocks and (A) shallower sequencing of maternal plasma, or (B) reduced fetal concentration among plasma sequences.