Size-tagged preferred ends in maternal plasma DNA shed light on the production mechanism and show utility in noninvasive prenatal testing

Abstract

Cell-free DNA in human plasma is nonrandomly fragmented and reflects genomewide nucleosomal organization. Previous studies had demonstrated tissue-specific preferred end sites in plasma DNA of pregnant women. In this study, we performed integrative analysis of preferred end sites with the size characteristics of plasma DNA fragments. We mined the preferred end sites in short and long plasma DNA molecules separately and found that these "size-tagged" ends showed improved accuracy in fetal DNA fraction estimation and enhanced noninvasive fetal trisomy 21 testing. Further analysis revealed that the fetal and maternal preferred ends were generated from different locations within the nucleosomal structure. Hence, fetal DNA was frequently cut within the nucleosome core while maternal DNA was mostly cut within the linker region. We further demonstrated that the nucleosome accessibility in placental cells was higher than that for white blood cells, which might explain the difference in the cutting positions and the shortness of fetal DNA in maternal plasma. Interestingly, short and long size-tagged ends were also observable in the plasma of nonpregnant healthy subjects and demonstrated size differences similar to those in the pregnant samples. Because the nonpregnant samples did not contain fetal DNA, the data suggested that the interrelationship of preferred DNA ends, chromatin accessibility, and plasma DNA size profile is likely a general one, extending beyond the context of pregnancy. Plasma DNA fragment end patterns have thus shed light on production mechanisms and show utility in future developments in plasma DNA-based noninvasive molecular diagnostics.

Keywords: circulating cell-free DNA; liquid biopsy; nucleosome structure; prenatal diagnosis; size-based molecular diagnostics.

Fig. 1.

Analysis of fragment end sites for plasma DNA fragments. Set S and set L included the preferred end sites for short and long plasma DNA molecules, respectively. The overlapping set in the middle included the preferred end sites for both short and long plasma DNA molecules.

Fig. 2.

Application of the size-tagged preferred end sites in noninvasive prenatal testing. (A) Size distribution of the plasma DNA reads covering set S and set L preferred end sites in a maternal plasma sample. (B) Correlation between the relative abundance (S/L ratio) of plasma DNA molecules with size-tagged preferred end sites and fetal DNA fraction in 26 maternal plasma samples. (C) Comparison of relative abundance of chr21 reads between control cases and trisomy 21 cases. Only the reads covering the set S preferred end sites (median read number: 133,702) were considered in this analysis. (D) ROC comparison between reads covering set S preferred end sites and random reads for trisomy 21 testing.

Fig. 3.

Analysis of the size-tagged preferred end sites in healthy subjects. (A) Size distribution of the plasma DNA reads covering set S and set L preferred end sites in a healthy subject. (B) Comparison of the relative abundance of plasma DNA reads with set S versus set L preferred end sites (S/L ratio) in pregnant women and healthy subjects.

Fig. 4.

Distribution of the distance between any two closest preferred end sites in set S and set L preferred end sites.

Fig. 5.

Distribution of size-tagged preferred end sites around regions with well-positioned nucleosomes. (A) Snapshot of the plasma DNA coverage, set S, and set L preferred end sites. An illustration of the nucleosome arrays on chr12p11.1 region is shown. (B) Distribution of the preferred end sites surrounding the common open chromatin regions shared by placental tissues and T cells. An illustration of the nucleosome positions is shown. Red and blue lines represent set S and set L preferred end sites, respectively. The aligned nucleosome positions as plotted on the x axis are in relation to the center of the common open chromatin regions.

Fig. 6.

Positions of the plasma DNA end sites in relation to the nucleosome structure. (A) Distribution of the size-tagged preferred end sites in pregnant plasma DNA relative to the nucleosome structure. The red and blue scissors represent cutting events that would generate set S and set L preferred end sites, respectively. (B) In healthy nonpregnant subjects, the distribution of autosomal fragment ends for short and long DNA molecules in relation to the nucleosome structure. The red and blue scissors represent cutting events that would generate short and long fragments, respectively. The aligned nucleosome positions as plotted on the x axis are in relation to the nucleosome center (23). Red and blue arrows mark the peaks at ±73 bp and ±95 bp, respectively.

Fig. 7.

Analysis of the fetal-specific and maternal-specific plasma DNA end sites and chrY fragment end sites. An illustration of the nucleosomal structure is shown here. (A) Distribution of fetal- and maternal-specific preferred end sites in the nucleosome structure. (B) Distribution of the chrY fragment ends of pregnant cases and healthy male subjects in the nucleosome structure. (C) In pregnant cases and (D) healthy subjects, the distribution of chrY fragment ends for short and long DNA molecules in the nucleosome structure. The aligned nucleosome positions as plotted on the x axis are in relation to the nucleosome center (23).

Fig. 8.

Fragment size distribution from ATAC-seq data of (A) buffy coat samples and (B) placental tissues. In buffy coat samples, the transposase enzyme has mostly cut the nonnucleosome-bound DNA (e.g., the linker region). In contrast, the transposase enzyme was able to cut within the nucleosomes in the placental tissues, indicating that the nucleosome packaging in the placental tissues was not as tight as that in the buffy coat samples. Blue and red scissors denote possible cutting events in buffy coat samples and placental tissues, respectively. I, II, III, and IV represent groups of peaks that have been generated by the indicated cutting events.