Evaluation of modified Jag-seq based on synthetic oligonucleotides
We refined the Jag-seq method by removing the step of 3′->5′ exonuclease T (Exo T) treatment, thus preserving the authenticity of urinary DNA ends. The missing complements of 5′ single-stranded protruding DNA ends present in a double-stranded DNA molecule were restored using dATPs (As), dTTPs (Ts), dGTPs (Gs), and methylated dCTPs (i.e. mCs) using an end-repair process, resulting in blunt ends. Such end-repaired DNA molecules were ligated with sequencing adapters, followed by bisulfite treatment. Hence, the methylated signal at CH sites proximal to the ends of urinary DNA molecule could be used for reflecting the jaggedness of urinary DNA.
To validate this refined Jag-seq, we designed two synthetic oligonucleotides. The first synthetic oligonucleotide was a 46-bp molecule carrying a 1-nt 5′ protruding G nucleotide end, with another G nucleotide immediately preceding that jagged end (Supplementary Fig. 1a). The second synthetic oligonucleotide carried 14-nt 5′ protruding ends, with four consecutive C nucleotides at the recessed ends (Supplementary Fig. 1b). For data generated by an experimental protocol without Exo T (Fig. 2a), we observed that 99.9% of nucleotides at the first base relative to the 3′ end were determined to be Cs, whereas only 0.2% of nucleotides at the second base relative to the 3′ end were determined to be Cs. These results suggested that the 1-nt jagged end could be accurately detected according to the methylation signal introduced by the end-repair process. For the data generated by the previously published experimental protocol with Exo T4, a substantial proportion of nucleotides at the second base relative to the 3′ end was determined to be Cs (98.2%), suggesting that a proportion of 1-nt jagged ends was changed to at least 2-nt jagged ends (Fig. 2b). From the data related to the second synthetic oligonucleotide (Fig. 2d), there were 15.9% and 2.6% of nucleotides identified as Cs at the positions of 1-nt and 2-nt relative to the jagged-end starting point in data prepared by the protocol with Exo T, whereas the counterparts were nearly close to 0% (Fig. 2c). Such data suggested that Exo T would commonly introduce an extra of 1-nt jagged nucleotide in a double-stranded DNA end.
a–d Sequencing base compositions for spike-in sequence with known jagged ends. The partial sequence of the spiked-in sequences with a 1-nt and a 14-nt jagged end are indicated on the x-axis. The nucleotides denoted in the uppercase letters indicate that the sequences are in double-stranded form. The nucleotides in lowercase letters indicate that the sequences are newly filled during the end-repair process. Vertical bars with blue color and red color represent the frequencies of sequenced T and C, respectively. Sequencing results from the refined (a, c) and original version (b, d) of Jag-seq are shown.
Difference in jaggedness between plasma and urinary DNA
We applied the modified Jag-seq approach to analyze urinary DNA and plasma DNA from healthy control subjects. CH methylation levels proximal to the 3′ end of a DNA molecule (i.e. read2), which was termed Jagged index-methylated (JI-M) values, were used to reflect the jaggedness of cell-free DNA. As shown in Fig. 3a, the JI-M value of urinary DNA (median: 48.78; range: 38.08–56.47) was much higher than that in plasma DNA (median: 17.33; range: 14.22–20.62) (p value: 4×10−5, Mann–Whitney U test). The value of JI-M varied according to the sizes of molecules, showing wave-like patterns in both urinary and plasma DNA (Fig. 3b). For urinary DNA molecules, the JI-M value rapidly increased ranging from 25 to 80. The profile of JI-M over fragment sizes displayed a small peak at ~120 bp and a major peak at ~250 bp, followed by a second major peak at ~410 bp. We speculated that the cell-free DNA molecules with ~120 bp in size might not be associated with an intact nucleosome structure, whereas the cell-free DNA molecules at ~250 bp and ~410 bp might be associated with intact nucleosome structures. These two populations of cell-free DNA were previously reported to have different DNA nuclease accessibilities in plasma5. For plasma DNA, the JI-M value had a similar trend across different fragment sizes but with lower JI-M values. These results were generally consistent with the previous study1.
a, b CH methylation level in read2 (JI-M) across the different fragment sizes. c, d Average jagged end length across the different fragment sizes. The central line indicates the median value. The bottom and top of the boxes are the 25th and 75th percentiles (interquartile range). The whiskers encompass 1.5 times the interquartile range.
We could further determine the exact length of jagged ends for a subset of DNA molecules, on the basis of the CC-tag strategy published previously4. This method was to employ the methylation pattern for two consecutive cytosines in a molecule for which the first cytosine was unmethylated but the second cytosine was methylated. Such a methylation pattern enabled the determination of the starting point of a jagged end. The average jagged end length in urinary DNA (median: 27.3 nt; range: 22.5–28.6 nt) was much longer than plasma DNA (median: 16.4 nt; range: 13.2–19.0 nt) (p value: 4×10−5, Mann–Whitney U test) (Fig. 3c, d and Supplementary Fig. 2a, b).
Jagged end length periodicity of urinary cfDNA
We studied the relative frequency of jagged end length ranging from 0 to 74 nucleotides (nt). Generally, the longer jagged ends coincided with the lower abundance in urinary DNA (Fig. 4a). Interestingly, we observed that there were 10-nt periodic patterns of jagged end length of urinary cfDNA, which were much less observable in plasma DNA (Supplementary Fig. 3).
a Jagged end length distribution measured by the CC-tag strategy. b Overall jagged end length periodicity index in RCC patients and healthy controls. The central line indicates the median value. The bottom and top of the boxes are the 25th and 75th percentiles (interquartile range). The whiskers encompass 1.5 times the interquartile range. c The area under the ROC curve (AUC) of differentiating between patients with and without RCC using jagged end length periodicity index. d Jagged end length periodicity index across different fragment sizes in healthy controls and patients with RCC.
To quantitatively analyze the periodicity of jagged end length, we used the relative difference between a series of peak values and the paired trough values using the following formula:
$${mathrm{Jagged}},{mathrm{end}},{mathrm{length}},{mathrm{periodicity}},{mathrm{index}} = frac{{mathop {sum }nolimits_{i = 1}^{n = 7} frac{{2P_i – V_{il} – V_{ir}}}{{P_i}}}}{7}$$
(1)
Where the Pi is the frequency of jagged end length at a particular peak i, Vil, is the frequency of jagged end length at the left valley relative to the peak i and Vir is the frequency of jagged end length at the right valley relative to the peak i.
A higher jagged end length periodicity index indicated that there was a stronger 10-nt periodic patterns present in urinary DNA jagged ends. Such jagged end length periodicity index was shown to be higher in urinary DNA of patients with RCC (median: 1.06; range: 0.81–1.38), compared with control subjects (median: 0.72; range: 0.37–1.09) (p value: 0.01, Mann–Whitney U test) (Fig. 4b). The receiver operating characteristic curve (ROC) analysis showed that the area under the curve (AUC) was 0.86 in differentiating patients with and without RCC, based on the periodicity index of urinary DNA jagged end length (Fig. 4c). These results implied that the periodicity of urinary DNA jagged end length might be used as a biomarker for RCC detection.
Figure 4d shows that sharp oscillations of jagged end length periodicity index were observed in those urinary DNA molecules below 120 bp in both RCC patients and healthy controls, whereas such oscillations were attenuated for those molecules above 120 bp.
Effects of heparin treatment on jagged end length periodicity
The strong 10-nt periodicities present in the distribution of jagged end lengths were reminiscent of the intrinsic characteristics of helical periodicity close to 10 bp per turn on the nucleosome. We conjectured that the urinary DNA molecules would be in part associated with histone proteins. As heparin with a high affinity to all histones could disrupt the interaction between DNA and histone proteins6,7, we used heparin to treat the urinary DNA to study whether the 10-nt periodicities of jagged end lengths would be dependent on histone proteins or not.
We performed EDTA treatment and heparin treatment on urine samples for 0 h, 0.5 h, and 1 h of in vitro incubation under room temperature, followed by Jag-seq. The 10-nt periodicities were remarkably reduced for the urine samples treated with heparin even at 0 h, compared with those treated by ETDA (Fig. 5a). With the prolonged periods of heparin incubation, the periodicity index was gradually decreased (Fig. 5b). Both the JI-M value and the average jagged end length showed decreased trends with heparin treatment (Supplementary Fig. 4a, b). These data suggested that the urinary DNA would be in part associated with histone proteins.
a Jagged end length distribution in urinary cfDNA with heparin incubation treatment. The lines with red, blue, green, and purple colors represent jagged end length distribution at EDTA 0 h, heparin 0 h, heparin 0.5 h, and heparin 1 h treatment, respectively. b Jagged end length periodicity index with EDTA and heparin treatment.
Relationship between jagged ends and nucleosomal structures
We recently presented evidence that the fragmentation patterns of cfDNA were associated with nucleosomal structures in urine1,7 and plasma8,9,10. High-resolution Jag-seq could provide an opportunity to study the 3′ recessed ends and the corresponding 5′ protruding ends in the complementary strand for a cell-free DNA molecule (Fig. 6a). For a double-stranded DNA molecule with 5′ overhangs, there were two pairs of 3′ recessed ends and 5′ protruding ends, namely part A (corresponding to a lower value in the genome coordinate) and part B (corresponding to a higher value in the genome coordinate) (Fig. 6a).
a Illustration for the definition of 5′ protruding end and 3′ recessed end in part A and part B. b 5′ protruding end density and 3′ recessed end density surrounding CTCF binding sites for plasma and urinary DNA molecules in part A. c Jagged end length periodicity index in nucleosomal linker regions and nucleosomal core regions.
We sought to investigate whether the 3′ recessed end density and the 5′ protruding end density would be related to nucleosomal structures. We applied the analysis in urinary DNA molecules falling within the 1-kb upstream and downstream relative to CTCF binding sites and calculated the end density around CTCF binding sites. The nucleosome tracks around CTCF binding sites were determined by the cell-free DNA coverages (gray lines in Fig. 6b). The preferred cutting sites during cfDNA generation were indicated by the peaks of end density11.
Figure 6b shows the results of plasma DNA (upper panel) and urinary DNA (lower panel) end density derived from part A. The peaks of 5′ protruding end density and 3′ recessed end density for plasma DNA were both approximately aligned to the linker regions, indicating that generation of plasma jagged ends might be derived from the cutting of the linker regions (Fig. 6b). A small shift between the 3′ recessed and 5′ protruding end density curves was observed in plasma DNA, likely because of the presence of the jagged ends (Fig. 6b). In contrast, in urine samples, the peaks of 3′ recessed end density and 5′ protruding end density were phased in different locations. The large shift between 3′ recessed and 5′ protruding end density curves indicated that the DNA degradation from the 3′ end might be more severe in urinary DNA, even extending into the nucleosomal cores. These data highlighted the potential interaction between urinary DNA and nucleosomal cores. The patterns derived from part B were consistent with that from part A (Supplementary Fig. 5).
We also studied the jagged end length periodicity index within all the nucleosomal core regions (143 bp) and among the linker regions (20 bp), respectively. The nucleosomal cores were defined by the 143-bp regions surrounding the peak signals of nucleosome tracks present in Fig. 6c. The linker regions were defined by the 20-bp regions at each side of a nucleosomal core. The jagged end length periodicity index was found to be higher at the nucleosomal core regions (0.77) compared with the linker regions (0.68) in urine (Fig. 6c). These results further highlighted that the generation of 10-nt periodicity of jagged end length might be associated with histone proteins in urine.
