NEW YORK (GenomeWeb News) – A University of Washington-led team reported online today that it has successfully sequenced the genome of an 18.5-week-old fetus using non-invasive methods.
"Although technical and analytical challenges remain, we anticipate that non-invasive analysis of inherited variation and de novo mutations in fetal genomes will facilitate prenatal diagnoses of both recessive and dominant Mendelian disorders," University of Washington genome sciences researcher Jay Shendure, the study's senior author, and colleagues wrote in Science Translational Medicine.
"Given our interest in applying genomics to Mendelian disorders — and the fact that, collectively, the thousands of Mendelian disorders for which we understand the molecular basis add up to quite a bit more than Down syndrome, in terms of their incidence — it seemed like a logical thing to think about," Shendure told GenomeWeb Daily News.
Shendure and his colleagues sequenced and genotyped the male fetus by doing deep sequencing on DNA circulating from maternal blood plasma, which contains fetal DNA. The analysis relied on additional information provided by genome sequencing on both parents, combined with the creation of a genome-wide haplotype map for the mother.
At almost three million sites in the genome where one or both of the parents were heterozygous, researchers correctly predicted the fetus' inherited alleles just over 98 percent of the time. Although specificity was more limited for predicting new mutations not present in either parent, the fetal genome also helped to identify the bulk of the non-inherited mutations that were found in the child's genome after birth.
Through shallower genome sequencing studies on a second family, the team showed that it could also tease apart many inheritance patterns in the genome of an 8.2-week-old fetus, though they could not discern de novo mutations.
In both families, the availability of maternal haplotype information proved critical to analyses of variant inheritance. And, researchers noted, the accuracy and level of information available on the fetal genome is expected to improve further when haplotype information is available for both parents.
"If you had to have haplotype information for one of the two parents, you would want it to be the mom because … the plasma that we sequenced from her is a mixture of her own DNA, her own sequences, and a small minority from the fetus," the study's first author Jacob Kitzman, a graduate student in Shendure's University of Washingon lab, told GWDN.
"But just as it helped the [inheritance] predictions for her, it would help the predictions for the father," he added.
The presence of cell-free fetal and maternal DNA in each pregnant woman's circulating blood plasma has raised the possibility of replacing some pre-natal tests that rely on invasive and potentially risky procedures such as amniocentesis or chorionic villus sampling with non-invasive, blood-based testing.
"Obviously we're not the first to be thinking about non-invasive fetal diagnostics," Shendure said. "There are several companies looking at trisomies this way now and taking advantage of this cell-free DNA, a fraction of which comes from the fetus."
For the most part, though, he and his co-authors explained, most of the applications of this cell-free fetal DNA to date have focused on developing non-invasive ways to detect extra chromosome copies, such as trisomy 21, or to test for conditions caused by single-gene mutations.
A team of researchers from China and the US that included Sequenom Chief Scientific Officer Charles Cantor and Dennis Lo, a member of that company's prenatal clinical advisory board, did demonstrate in Science Translational Medicine that it was possible to predict fetal genotypes by bringing together sequence data from DNA in maternal plasma with parental haplotype information.
Because that work relied, in part, on information obtained from chorionic villus sampling and focused on a set of common SNPs present on commercial arrays, though, those involved in the new study decided to take a slightly different tack.
Their approach involved harnessing a haplotype-resolved genome sequencing method that Shendure, Kitzman, and colleagues described previously in Nature Biotechnology to help come up with a fetal genome sequence that's complete enough to offer information on both inherited variations and de novo mutations.
In the first family assessed, researchers used the Illumina HiSeq 2000 to sequence fetal and maternal DNA in the mother's plasma sample to an average depth of 78 times using less than five nanograms of the plasma-derived DNA.
The identification and interpretation of fetal sequences in that plasma sample was informed by SNP patterns in the father's genome — which was shotgun sequenced to an average depth of 39 times using genomic DNA from a spit sample — and using haplotype-resolved genome sequence data on the mother.
Her whole blood sample was sequenced to a depth of 32 times, on average, and the team sorted out her haplotype patterns by using information at almost two million heterozygous SNPs.
To look at the accuracy of their fetal genome sequence, which represented a male fetus at 18.5 weeks gestation, researchers also did 40X genome sequencing on genomic DNA from cord blood after the child was born.
The team was able to accurately infer fetal genotype at millions of SNPs, looking at plasma DNA features such as allele frequencies in relation to maternal haplotype block patterns, the presence or absence of alleles at paternal-only alleles, and so on. The method was especially good at identifying inherited fetal variants at sites in the genome where the father was homozygous, the mother was heterozygous, and researchers had access to complete maternal haplotype information.
Sites where they ran into trouble trying to infer fetal genotype included a subset of heterozygous sites not phased in the maternal genome. There, the team did not try to distinguish inherited variants in the fetus, since accuracy at such sites was expected to be low.
The fetal sequence also helped in identifying 39 of the 44 de novo mutations uncovered in the infant's genome after birth, though researchers noted that the specificity is still quite low when analyzing these non-inherited alterations.
Through their analysis of second family, combined with simulations using data from the first, the researchers looked at how the accuracy of their fetal genome inferences varied depending on factors such as sequence depth, the fraction of fetal DNA in maternal plasma, and maternal haplotype block length.
The second family assessed in the study was sequenced at lower depth, involved a fetus at an earlier stage of gestation, and relied on maternal plasma that contained a lower fraction of fetal DNA. There, researchers could still discern patterns for many of the inherited variants, but were less equipped to predict de novo mutations or to untangle fetal genotype at the same level of resolution.
Meanwhile, simulation studies on the first family indicated that the ability to detect inheritance could be improved by doing deeper sequencing and/or having access to the longer paternal haplotype blocks.
"We would have done a lot better if we had blood from Dad," Shendure explained. "We only had saliva, so we had to rely on a site-by-site method rather than a haplotype-based method."
Indeed, researchers say haplotype-resolved sequencing on both parents — combined with even deeper sequencing — will be needed for fetal genome sequencing and interpretation down the road, particularly if and when the approach does move from the research stage to the clinic.
"In the long run, you want really good haplotype-resolved genomes for both parents and you want ultra-deep coverage of plasma," Shendure said.
Going forward, increased sequencing depth and more extensive haplotyping on both parents are also expected to help in untangling other types of genetic variation in the fetal genome, including CNVs, insertions and deletions, structural rearrangements, and more.