Accurate fetal chromosome dosage determination by shotgun sequencing of maternal plasma DNA without PCR amplification during library preparation.

Nextgen sequencing technologies have enabled noninvasive prenatal testing of fetal aneuploidies by shotgun sequencing of maternal plasma DNA during pregnancy (1). A statistically significant change (e.g., using z-score statistics) in the number of reads mapped to a chromosome of interest (such as chromosome 21) is used to identify fetal aneuploidy. The accuracy of the test is largely dependent on 2 factors: the variability (SD) of the percentage of reads mapped to the chromosome of interest among euploidy samples and the absolute difference between the sample being tested and the mean of the euploidy samples. We speculated that the PCR enrichment step in the library preparation before cluster generation and sequencing contributes to the variation in chromosome dosage determination. Additionally, with 3 ng of DNA (typically obtainable from about 2 mL of maternal plasma) at an average size of 150 bp per DNA fragment, around 20 billion DNA fragments were available for sequencing, suggesting PCR amplification may not be necessary. We thus modified the current protocol for library preparation from maternal plasma DNA by removing the PCR enrichment step. Maternal plasma samples of about 2.5–3.2 mL were obtained with informed consent. DNA preparation and sequencing library preparation were performed as previously described (2 ), up to the step for adaptor ligation and purification. The purified ligation products were then equally divided into 2 aliquots, one for the standard protocol with PCR enrichment (PCR ) and the other for the new protocol without PCR enrichment (PCR ). The libraries were quantified using Illumina’s quantitative PCR quantification method, using serial dilutions of the PhiX library (Illumina) for standard curve generation. The PCR libraries ranged from 62– 451 pmol/L [mean (SD), 194 (102) pmol/L]. Individual libraries were mixed in a 4-plex format of equal concentrations of 2 pmol/L for multiplexed, paired-end sequencing at 2 36 bp on the Illumina Genome Analyzer IIx. Preprocessing of the raw pairedend data from multiplexed sequencing was done using Illumina’s CASAVA software (version 1.7), followed by alignment using bowtie-0.12.1 (3 ), with only 1 mismatch allowed. Paired-end reads were used to assess DNA insert size. Uniquely aligned single reads (read 1) were used for all other analyses. The size distributions for the PCR and PCR libraries were highly correlated, with the PCR libraries having a bias against longer fragments. Given that cell-free fetal DNA was generally smaller in size than maternal DNA, fetal DNA concentrations may have been overestimated by previous shotgun sequencing data (4). To assess the variability in determining the chromosome dosages for both the PCR and PCR protocols, we randomly sampled 2.5, 3, 3.5, 4, 4.5, and 5 million reads for each library for the PCR and PCR protocols derived from 20 maternal plasma samples from euploidy pregnancies with gestational age at 14.9 (2.2) weeks (range, 12.1–19.4 weeks). The percentages of reads mapped to individual autosomes compared to total reads mapped to all autosomes were calculated. CVs for each autosome were then calculated. We repeated the random read selection 5 times at 2.5, 3, 3.5, 4, 4.5, and 5 million reads to further assess the variations due to read selections (error bar in Fig. 1). As shown in Fig. 1, the PCR protocol reduced the CVs for most autosomes, regardless of the total read numbers. At 5 million total reads, all autosomes except chromosomes 2, 10, and 14 showed significant differences in CV between the PCR and PCR protocols (P 0.05, 2-sided t-test). As reported previously (4 ), autosomes with the lowest and highest GC contents showed the highest CVs. The CVs for chromosomes 13 and 18 were reduced by about 28% and 18%, respectively, which may be clinically useful because it is known that the detection of trisomy 18 and trisomy 13 are less accurate with the shotgun sequencing approach (5). Although the CV for chromosome 21 remained low, there was an apparent increase of 14% for the chromosome 21 CV. The actual impact on the detection of trisomy 21, trisomy 18, and trisomy 13 by the PCR protocol needs to be further validated with a larger sample size. As a proof-of-principal experiment, we also sequenced 3 trisomy 21 and 3 trisomy 18 maternal plasma samples with the PCR and PCR protocols. The staff personnel for data analysis were blinded to the sample IDs. Sample IDs were unblinded only after all the chromosome mappings were finished. Correct calls were made for all samples with both the PCR and PCR protocols, using a z-score cutoff of 3 (2). In summary, we report here a simplified PCR protocol for library preparation from maternal plasma DNA for the purpose of noninvasive prenatal testing of chromosomal aneuploidies. Our data suggest that this simplified protocol improves overall reproduc1 Nonstandard abbreviations: PCR , standard protocol with PCR enrichment; PCR , new protocol without PCR enrichment. Clinical Chemistry 60:4 000 – 000 (2014) Letters to the Editor