Enrichment of fetal and maternal long cell‐free DNA fragments from maternal plasma following DNA repair

Abstract Objective Cell‐free DNA (cfDNA) fragments in maternal plasma contain DNA damage and may negatively impact the sensitivity of noninvasive prenatal testing (NIPT). However, some of these DNA damages are potentially reparable. We aimed to recover these damaged cfDNA molecules using PreCR DNA repair mix. Methods cfDNA was extracted from 20 maternal plasma samples and was repaired and sequenced by the Illumina platform. Size profiles and fetal DNA fraction changes of repaired samples were characterized. Targeted sequencing of chromosome Y sequences was used to enrich fetal cfDNA molecules following repair. Single‐molecule real‐time (SMRT) sequencing platform was employed to characterize long (>250 bp) cfDNA molecules. NIPT of five trisomy 21 samples was performed. Results Size profiles of repaired libraries were altered, with significantly increased long (>250 bp) cfDNA molecules. Single nucleotide polymorphism (SNP)‐based analyses showed that both fetal‐ and maternal‐derived cfDNA molecules were enriched by the repair. Fetal DNA fractions in maternal plasma showed a small but consistent (4.8%) increase, which were contributed by a higher increment of long fetal cfDNA molecules. z‐score values were improved in NIPT of all trisomy 21 samples. Conclusion Plasma DNA repair recovers and enriches long cfDNA molecules of both fetal and maternal origins in maternal plasma.

DNA fragmentation during karyorrhexis, followed by the formation and production of apoptotic bodies. [21][22][23] Commonly used sequencing library preparation methods employ the repair of DNA ends only, leading to the loss of damaged cfDNA fragments such as those with single-strand nicks. 20,24,25 Therefore, the detection of these damaged cfDNA molecules by the current paired-end (PE) massively parallel sequencing (MPS) platforms is challenging.
To salvage the damaged DNA molecules, the use of PreCR repair mix, a commercial DNA repair kit, has been shown to be effective in recovering damaged DNA in forensics, 26,27 archeology, 28 and molecular diagnostics. 29 We hypothesized that the PreCR repair mix could also be used to repair cfDNA in maternal plasma. Repaired maternal plasma DNA with higher DNA integrity might enhance the success for downstream analysis. Due to the fact that fetal-derived cfDNA molecules are shorter and more fragmented than the maternal ones, 2,30 we proposed that fetal cfDNA molecules possess more DNA damages. By repairing these damaged fetal cfDNA molecules, the fetal DNA fraction may be enriched.
In this study, we applied PreCR repair treatment on cfDNA from maternal plasma of first and third trimester pregnancies. We studied and compared the size profiles and fetal DNA fractions between repaired cfDNA samples and their sham controls. We also evaluated the impact of PreCR repair treatment on the performance of NIPT on samples collected from trisomy 21 pregnancies.  (Table S1). Plasma was isolated from 20 mL of EDTA-anticoagulated maternal peripheral blood as previously described. 2 For third trimester cases, 1 cm 3

| Targeted capture enrichment and MPS
Targeted capture of sequencing libraries was performed as described. 32 The capture probes (Roche Nimblegen) predominantly targeted sequences on Chr6 and ChrY and were designed for a previous study on noninvasive prenatal assessment of congenital adrenal hyperplasia. 32 Briefly, capture probes were designed to target a size of approximately 6-Mb region, with each approximately 3 Mb flanking CYP21A2 gene located on 6p21.3. For ChrY, a total size of approximately 1 Mb targeting ChrY unique regions was designed for detecting fetal signal in the maternal plasma. All repeat regions were excluded from the probe designing process. An average sequencing depth of 152 times per base (ranging from 82× to 210×) was obtained.
Target capture enrichment efficiency was validated by quantitative real-time PCR. All libraries were sequenced with a PE format of 75 bp × 2 on a NextSeq 500 System (Illumina). Adaptor sequences and low-quality bases (ie, quality score less than five) were removed, and sequencing reads were aligned to the nonrepeat-masked human reference genome (hg19) using the Short Oligonucleotide Alignment What is already known about this topic?
• Most of the cell-free DNA (cfDNA) fragments in maternal plasma have sizes less than 200 bp, with fetal molecules being shorter than maternal ones.
• Occasional no call for noninvasive prenatal testing (NIPT) can be caused by insufficient fetal DNA fraction.
What does this study add?
• Repair of cfDNA by PreCR repair mix can recover a subset of long (>250 bp) cfDNA molecules.
• Both fetal and maternal long cfDNA are enriched by PreCR repair treatment.
• Mild but consistent increments in fetal DNA fractions after PreCR repair, which are contributed by higher enrichment of long fetal cfDNA molecules.
• PreCR repair treatment improves NIPT of trisomy 21 by elevating z scores resulting in better discrimination of aneuploid from euploid samples. Program 2 (SOAP2). 33 Up to two nucleotide mismatches but not indels were allowed for each member of the PE reads.  Single-nucleotide polymorphisms (SNPs) were called by the Birdseed v2 algorithm with a confidence score cutoff 34 of 0.15. The genotypes of the CVS and placentas were compared with those of the mothers to identify the fetal-specific and maternal-specific SNP alleles. An SNP was considered as fetal-specific if it was homozygous in the mother and heterozygous in the fetus and the reverse for maternal-specific SNPs. The fetal DNA fractions were deduced as described. 2 Briefly, fetal DNA fraction ( F ) was deduced by the allelic ratio between a fetal specific SNP allele (p) and a common SNP allele (q) shared by the mother and the fetus using the following formula 5 : In the size-fractionated fetal fraction analysis, fetal DNA fragments were divided into 10-bp bins as previously described, 31 with modified size range of analysis. For detection of trisomy 21, overrepresentation of Chr21 of each sample was quantified by the z score using the following formula: where GRchr21 is the genomic representation (GR) of chromosome 21.  fetal-or maternal-specific SNP alleles were separated and considered as fetal-or maternal-specific cfDNA molecules, respectively. 31 We observed a similar trend in repaired cfDNA molecules from both fetal and maternal origins (Figures 2A,B  These results suggested that there was a relative higher enrichment  3.2 | Fetal DNA fraction characteristics of sham-or PreCR-repaired cfDNA libraries from first and third trimester maternal plasma samples

| Statistical analysis
As the fetal DNA fraction is an important parameter for successful NIPT, an increase in this metric might potentially benefit the sensitivity of NIPT, especially in cases with insufficient fetal contribution. 35 Therefore, we characterized and compared the overall fetal DNA fractions of maternal plasma from first and third trimester samples after sham and PreCR repair treatments. We observed a consistent 4.4% increase of fetal DNA fraction (12.5%-13.1%) for first trimester samples and 5% increase of fetal DNA fraction (28.5%-30.0%) for third trimester samples after the PreCR repair treatment compared with their sham controls ( Figure 3A). Statistical analysis showed that such increments were significant (both P = 5.12E-03; Wilcoxon signed rank test). We speculated that the source of such mild fetal DNA fraction increment was contributed from the newly repaired fetal long cfDNA molecules. To prove this, we applied size-fractionated fetal fraction analysis on both sham-and PreCR-repaired subjects. Fetal DNA fragments were divided into 10-bp bins, ranging from 0 to 600 bp. Fetal DNA fractions within each bin were then deduced separately. 31 We observed that such fetal DNA fraction increment was mainly contributed by the increase of fetal cfDNA molecules with the size of around 300 bp ( Figures 3B and S3). Taken together, these results suggested a higher enrichment of long fetal cfDNA molecules over their maternal counterparts in maternal plasma.
As fetal cfDNA molecules are the minority species in the maternal plasma, most of the data obtained from the total maternal plasma cfDNA sequencing are dominated by maternally derived cfDNA molecules. To obtain increased sequencing depth of fetal cfDNA molecules specifically, we used a target capture approach with probes hybridizing to sequences on Chr6 and ChrY. 32  3.3 | Enrichment of cfDNA molecules longer than 250 bp from maternal plasma after PreCR repair treatment As the major size differences between sham and repair cfDNA resided on the cfDNA molecules longer than 250 bp, detection of such molecules by conventional next-generation sequencing (NGS) platforms, such as the Illumina platform, might not be optimal due to their short read lengths. A third-generation sequencing platform from Pacific Biosciences, which employed the SMRT technology, can achieve a sequencing read lengths exceeding 20 kb. 36 We hypothesized that using such a technology, we might improve our ability to observe the cfDNA molecules longer than 250 bp in the maternal plasma.

| Improved NIPT performance of trisomy 21 after PreCR repair treatment
As the repaired cfDNA sample has more intact long cfDNA molecules that are analyzable from the fetus, we aimed to demonstrate its potential application for prenatal diagnosis. We recruited five pregnancies with trisomy 21 fetuses and compared the chromosomal representation of Chr21 with the 10 first trimester euploid cases, which would be the most clinically relevant samples for NIPT. We observed that after PreCR repair treatment, Chr21 z scores for the trisomy 21 cases showed an average increase in 17.3% (ranging from 14.4% to 19.8%) and exhibited better separation from the z scores of the euploid samples ( Figure 6). These results suggested that PreCR repair treatment may improve NIPT performance of trisomy 21 pregnancies.  The PreCR repair mix we used in this study has been shown to repair a variety of DNA damages, including DNA nicks, gaps, modifications of DNA molecules such as oxidation, and deamination, loss of DNA bases, and pyrimidine-dimer formation. 26,27 In forensics, it was reported that the PreCR repair mix could restore short tandem repeat profiles from UV-damaged DNA samples 26 and artificially degraded DNA, mimicking the exposure of native DNA to oxidizing agents, hydrolytic conditions, and ionizing radiation. 27 In archeology, it was used to recover heavily damaged ancient DNA samples. 28 A similar repair procedure was applied in molecular diagnostics using formalinfixed paraffin-embedded samples. 29  to 3.6%) after PreCR repair treatment ( Figure 1A,B). However, this enrichment became marginal (2.5%, from 6.0% to 6.2%) in the target capture setting ( Figure 4B). We believe that this was due to capture efficiency differences between different samples.
The ability of the PreCR repair mix to restore damaged DNA opens up many potential applications. In prenatal diagnosis, our data demonstrated that PreCR repair treatment may contribute to improved NIPT performance for trisomy 21 screening ( Figure 6). z scores of repaired trisomy 21 samples were better separated from those of the euploid samples ( Figure 6). This is especially important for a case where the Chr21 z score is borderline and close to the euploid-aneuploid cutoff without repair treatment. Future investigation is needed to investigate the performance of this approach in larger scale studies and if such improvement can be extended to the detection of other chromosomal aneuploidies. Beyond enhancing the statistical confidence of discrimination of euploidy and aneuploidy, repair of cfDNA may improve the call rate of NIPT. In particular, scenarios with nonreportable NIPT results caused by insufficient fetal DNA fraction [42][43][44][45][46] or poor cfDNA quality might be improved by DNA repair. Examples of such scenarios might include NIPT in very early pregnancies and samples obtained from pregnant women with high body mass indices. 12,[47][48][49] Furthermore, the ability of cfDNA repair to reveal more long analyzable cfDNA molecules in plasma might open up the possibility of using NIPT for long genomic targets, eg, sequences involved in triplet repeat disorders such as the fragile X syndrome (FXS). In FXS, the length of the CGG tandem repeats in patients with fully mutated FMR1 alleles 50,51 is longer than 600 bp.
In conclusion, this study has revealed a preferential recovery of long cfDNA molecules in maternal plasma after PreCR repair treatment. Small but consistent increment of overall fetal DNA fraction is contributed by higher fetal-derived cfDNA molecules enrichment.
Practically, this PreCR repair treatment is a convenient single-step procedure with low unit cost (~$10 per reaction). We hope that the data presented here might catalyze further research to translate these observations into clinical enhancements in NIPT.

ACKNOWLEDGMENTS
We thank Dr Ji Lu, Mr Wenlei Peng, and Ms Carol Szeto for their technical assistance in tissue genotyping and data analysis.