ABSTRACT
Background & Aims Integrated hepatitis B virus (HBV) DNA, found in >85% of HBV-associated hepatocellular carcinomas (HBV-HCC), can play a significant role in HBV-related liver disease progression. HBV-host junction sequences (HBV-JS’s), created through integration events, have been used to determine HBV-HCC clonality. Here, we investigate the feasibility of analyzing HBV integration in a noninvasive urine liquid biopsy.
Approach & Results Utilizing an HBV-targeted NGS assay, we first identified HBV-JS’s in 8 HBV-HCC tissues and designed short-amplicon junction-specific PCR assays to detect HBV-JSs in matched urine. We detected and validated tissue-derived junctions in 5 of 8 matched urine samples. Next, we screened 32 urine samples collected from 25 HBV-infected patients (5 with hepatitis, 10 with cirrhosis, 4 with HCC, and 6 post-HCC). Encouragingly, all 32 urine samples contained HBV-JS’s detectable by HBV-targeted NGS. Of the 712 total HBV-JS’s detected in urine, 351 were in gene-coding regions, 11 of which, including TERT, had previously been reported as recurrent integration sites in HCC tissue and were found in urine of cirrhosis or HCC patients only. The integration breakpoints of HBV DNA detected in urine were found predominantly (∼70%) at a previously identified integration hotspot, HBV DR1-2.
Conclusions HBV viral-host junction DNA can be detected in urine of HBV-infected patients. This study is the first study to demonstrate the potential for a noninvasive urine liquid biopsy of integrated HBV DNA to monitor HBV-infected patients for HBV-associated liver diseases and the efficacy of antiviral therapy.
INTRODUCTION
Hepatitis B virus (HBV) integration can occur early in the course of infection (1-3). Integration events are known to cause insertional mutagenesis, chromosomal instability, activation of oncogenes, and inflammation, and thus can play a critical role in liver disease progression and the development of hepatocellular carcinoma (HCC) (4). Integrated viral DNA has been detected in more than 85% of HBV-associated HCC (HBV-HCC) cases (5-8). Although HBV integration sites have been observed throughout the human genome, integration into known HCC driver genes (such as TERT, MLL4 and CCNE1) has been reported repeatedly in patients with HCC (9).
One outcome of HBV DNA integration is the creation of a unique HBV-host junction sequence (HBV-JS) representing a specific molecular signature of the infected cell. HBV DNA integrations have been used to determine clonality of recurrent HCC tumors in tissues (10, 11). Recent studies have shown that integrated HBV DNA can be detected in the circulation (12-15). We have previously demonstrated that urine of HCC patients contains cell-free DNA (cfDNA) with HCC-associated modifications (16-19), whereas urine of HBV-infected patients contains fragmented (but not intact) HBV DNA (20). We hypothesize that urine can serve as a noninvasive liquid biopsy for the investigation of HBV integration in HBV-infected patients.
In this proof-of-concept study, we assessed the feasibility of detecting HBV-JS’s in urine from HBV-infected patients. We first identified HBV-JS’s in HCC tissue followed by detection in matched urine by sensitive junction-specific PCR assays. Next, we detected HBV-JS’s in urine from HBV-infected patients with hepatitis, cirrhosis, and HCC, including those undergoing recurrence monitoring (post-HCC), using an HBV-targeted NGS assay. These findings support the potential of urine as a noninvasive liquid biopsy for integrated HBV DNA, which can be used to monitor the dynamics of HBV integration in the infected liver and assess HBV treatment efficacy and disease progression.
PATIENTS AND METHODS
Study subjects and specimens
Two studies, A and B, were performed. Study A specimens included archived DNA isolated from eight pairs of matched HBV-HCC formalin-fixed paraffin-embedded tumor tissue and urine samples. The specimens were obtained from patients in stages I-II undergoing surgery (Table 1) at the National Cheng-Kung University Medical Center, Taiwan, as described previously (17, 19). Study B was performed on urine samples collected from 25 HBV-infected patients, including 5 with hepatitis, 10 with cirrhosis, 4 with HCC, and 6 post-HCC, at the Thomas Jefferson University Hospital (Table 2). Five patients (B7, B15, B17, B19, and B20) provided more than one urine collection over the course of the study, resulting in a total of 32 urine samples. All patient samples were collected with written informed consent and in accordance with the guidelines of the Institutional Review Board.
Additionally, eight HBV-negative normal urine donors ranging from 23 to 60 years of age from four male and four females served as a normal urine donor cohort.
Urine collection and DNA isolation
Urine samples were collected and total urine DNA isolated as previously described (21). For study A, cfDNA was obtained by isolating the <1kb fraction from total urine DNA using carboxylated magnetic beads, a method previously developed by our laboratory (22). For study B, the JBS urine cfDNA isolation kit (JBS Science Inc, PA) was used according to the manufacturer’s specification.
HBV-JS detection by HBV-targeted NGS assay and validation by PCR-Sanger sequencing
Tissue DNA was fragmented by sonication and subjected to Illumina NGS library preparation as previously described (23). Ten cycles of library amplification with Herculase II Fusion polymerase were performed (Agilent Technologies, Santa Clara, CA). Urine cfDNA samples underwent NGS library preparation using the ThruPLEX Tag-Seq kit (Cat# R400585, Takara Bio, Mountain View, CA), which is designed for adaptor ligation to double-stranded DNA and contains unique molecular tags (UMTs), according to manufacturer specifications. Library DNA was subjected to an HBV-targeted NGS assay following the manufacturer’s instructions (JBS Science Inc, PA). The captured library DNA was pooled for NGS analysis on the Illumina MiSeq platform (Penn State Hershey Genomics Sciences Facility at Penn State College of Medicine, Hershey, PA) or on the NovaSeq platform (Psomagen Inc, Rockville, MD).
The ChimericSeq analysis software we developed previously (24) was used to detect HBV-JS’s in NGS datasets. For NGS libraries containing UMTs, UMT family consolidation and consensus read generation were performed using the software package Connor (25). Briefly, UMT-containing reads were aligned to human and HBV reference genomes, and paired reads were grouped into families based on the mapping coordinates and combined UMT sequence. For each UMT family, a consensus read pair was generated. The consensus reads were then processed by ChimericSeq for HBV-JS identification.
To validate junctions identified by NGS, junction-specific primers were designed for the host and viral sequences for tissue and urine (Supplemental Table 1). Tissue primers generating amplicons larger than 100 bp for Sanger sequencing, while urine primers were designed to target the same junction sequence generating amplicons of 65 bp or less. The junction-specific PCR was performed with Hotstart Plus Taq Polymerase (Qiagen, Valencia, CA). PCR products were visualized by agarose gel electrophoresis on a 2.2% FlashGel DNA Cassette (Lonza Group, Basel, Switzerland) and subsequently subjected to Sanger sequencing, a nested PCR reaction, or restriction endonuclease (RE) digestion using enzymes obtained from New England Biolabs (Ipswich, MA), according to the manufacturer’s specifications.
Data availability
The sequence data for the subjects studied in this work who consented to data archiving have been deposited in the European Genome–Phenome Archive (EGA), www.ebi.ac.uk/ega, which is hosted by the European Bioinformatics Institute (EBI) www.ebi.ac.uk (accession no. pending).
Statistical analysis
The Chi-square test was performed for categorical variables. For all statistical tests, the level of significance was defined as p < 0.05 to determine whether differences were significant.
RESULTS
Detection of HBV integration in urine DNA by junction-specific PCR assays
To investigate if HBV-JS’s can be detected in urine, we identified eight pairs of archived DNA samples from matched tissue and urine from HBV-HCC patients (study A, Table 1), with each urine DNA isolate extracted from at least 1 mL of urine. We first detected HBV-JS’s in tissue DNA using an HBV-targeted NGS assay as described in Material and Methods. HBV integrations were detected in all eight HCC tumor DNA samples (Table 3). The integrated regions included the TERT gene, the most frequently reported integrated host gene (patient A5) (9). The junctions were validated by Sanger sequencing (Supplemental Fig. 1) using primers listed in Supplemental Table 1.
To detect HCC tissue-derived HBV junction DNA in urine, we designed a junction-specific PCR assay for each junction (Supplemental Table 1). The amplicons were designed to be short since urine cfDNA is mostly fragmented to less than 250 bp (21). Matched tissue and urine DNA samples were subjected to junction-specific PCR amplification. PCR products of expected size were generated from five of the eight matched urine samples tested. To determine if the PCR products generated from the urine and tissue in each of the five tissue-urine pairs were identical, we used either two-step nested PCR followed by RE digestion of PCR products (patients A5, A6, and A8; Fig. 1A) or Sanger sequencing when no suitable RE site was available in PCR products (patients A1 and A2; Fig. 1B-D). Four (patients A2, A5, A6, and A8) of the five PCR product pairs contained sequences indistinguishable by the above methods, indicating that the HCC-tissue-derived HBV-JS was present in the matched urine.
Interestingly, the sizes of the junction PCR products from patient A1 were different between tissue and urine (Fig. 1C). Sanger sequencing analysis identified a 23-nucleotide (nt) insert indicating a re-arrangement between chromosomes 5 and 10 in the urine DNA sample. To determine if the urine-identified integration site could be found in tumor tissue DNA, we designed a primer across the chimeric chromosome 5/10 sequence (Fig. 1D) to specifically amplify the 23-nt-insert-containing HBV-JS species. Remarkably, the tripartite HBV-JS was successfully amplified and validated in the matched tissue DNA sample (Fig. 1D). In total, we were able to detect and verify five of the eight HCC-tissue-derived HBV-JS’s in matched urine samples.
Detection and characterization of HBV-JS’s in urine of HBV-infected patients by an HBV-targeted NGS assay
After detecting liver-derived HBV-JS’s in urine of HBV-HCC patients by junction-specific PCR assays, we next assessed if it was feasible to analyze HBV integrations in urine without prior knowledge of junction sequences from tumor tissue and before disease progression to HCC. A total of 32 urine specimens were collected from 25 HBV-infected individuals, including 5 patients with hepatitis, 10 with cirrhosis, 4 with HCC, and 6 post-HCC (study B; summarized in Table 2). Five patients provided more than one urine specimen in the course of the study. Urine DNA was isolated from HBV-infected patients (Table 2) and normal donors, enriched for HBV sequences, and subjected to NGS followed by HBV-JS identification as described in Materials and Methods. Encouragingly, all 32 HBV-infected urine samples contained HBV-JS chimeric reads after UMT consolidation with at least 2 supporting reads (SR), a threshold applied for stringent HBV-JS identification and consistent with other reports (26-29).. Additionally, there were no detectable HBV-JS chimeric reads of ≥2 SR in the normal urine donor cohort.
Of all integration junctions identified in HBV-infected patient urine, a total of 712 HBV-JS’s from 31 urine samples These included 84, 366, 41, and 221 junctions detected in 4 hepatitis, 12 cirrhosis, 6 HCC, and 9 post-HCC urine samples, respectively. Table 4 lists the major HBV-JS’s of each individual urine sample, defined as the two most abundant HBV-JS’s supported by at least 3% of total junction reads. As expected, we detected integrations in both gene-coding and non-coding regions in all disease categories. The latter included repeat regions, such as LINEs, SINEs, and simple repeats (30), that have been previously reported as HBV integration sites. Among the targets of the 351 unique integrations in gene-coding regions detected in urine, 11 genes had been previously reported in liver tissue from patients with severe liver diseases, such as cirrhosis and HCC. These genes are AC007880.1, ADAM12, ARSA, ATF6, CLEC2L, FOCAD, PPP2R2C, TNFRSF9, TXNDC16, VEGFC, and TERT. Most interestingly, all the genes, except for AC007880.1 and ARSA, have been associated with the development of HCC (ADAM12 (31), ATF6 (32), CLEC2L (33) TERT, VEGFC (34)) or other cancers (FOCAD (35), PPP2R2C (36), TNFRSF9 (37), TXNDC16 (38)).
Distribution of integrated HBV breakpoints detected in urine
We next examined the HBV genome breakpoints of the 712 junction sequences detected in urine. The HBV sequences of these junctions were plotted based on their locations in the HBV genome (Fig. 2A) and summarized in Fig. 2B. Approximately 70% of HBV DNA breakpoints detected in urine were clustered in the HBV DR1-2 region, a known integration hotspot (8) (Fig. 2B). The second most frequent of the previously reported HBV breakpoint regions, PreS (8), was also identified in urine, with higher frequencies found in hepatitis (7.1%), cirrhosis (10.5%), and post-HCC (7.3%) as compared with HCC (4.9%). However, the association of integrated HBV breakpoint location (DR1-2, PreS1/2, or other) with disease type was not statistically significant (χ2(6, 712)=6.3; p=0.39) (Fig. 2B).
Profiling of detectable HBV integrations sites in serially collected urine samples
Urine cfDNA containing HBS-JS’s may reflect the HBV integration profile of the infected liver. For a preliminary assessment of the possibility of using HBS-JS’s for disease monitoring, we collected serial urine samples from five patients (B7, B15, B17, B19, and B20) and compared the HBV-JS’s detected in different collections (Table 4). Patients B15 and B19, both with cirrhosis, had two collections one week apart and showed consistent integration in SINE and simple repeat regions, respectively, although they do not occur at the same integration breakpoint. Similarly, patient B20 (post-HCC) had four collections at 3-month intervals, all showing integration in non-coding regions (e.g., simple repeats), although the specific sites varied. Patient B17 (HCC) had two urine collections 2 months apart, with different junction sequences detected. Patient B7, with cirrhosis, had two urine samples collected 7 months apart, with integrations initially detected in a non-coding region followed by integrations in gene-coding regions. Interestingly, patient B7 developed HCC 4 years after the second collection. This initial comparison suggests that different HBV integrations can be captured in urine through serial monitoring. Collectively, our data demonstrate the feasibility of analyzing HBV integration using urine liquid biopsy.
DISCUSSION
In this proof-of-concept study, we provide evidence for noninvasive detection of integrated HBV DNA in urine of HBV-infected patients with liver disease ranging from hepatitis to HCC. The novelty of this study is threefold. First, it is the first study to report detection of integrated HBV DNA in urine. Second, it includes the first comparative analysis of HBV integration in hepatitis, cirrhosis, HCC, and post-HCC using solely ectopic samples. Third, this study provides the first unambiguous evidence that urine contains liver-derived DNA, by taking advantage of the liver tropism of HBV and unique HBV-JS’s as molecular signatures of infected hepatocytes. Taken together, our results demonstrate the potential of urine liquid biopsy for integrated-HBV burden and HBV-related liver disease monitoring.
Our previous study demonstrated that only fragmented HBV DNA was detected in urine of patients with HBV infection even when the blood viral load was as high as 108 copies/mL (20). Urine cfDNA is mostly mono- and dinucleosomal in size and largely derived from apoptotic cells (21). Therefore, HBV-sequence-containing urine cfDNA likely originates from degradation of host-integrated HBV genomes protected by nucleosomes in apoptotic HBV-infected hepatocytes. Hepatocytes from the HBV-infected liver have a higher turnover rate than hepatocytes from uninfected normal liver (39). Thus, urine provides a unique biopsy source to noninvasively detect integrated HBV DNA and monitor infected-cell burden, neither of which is achievable with current clinical HBV tests.
As expected, HBV genome integration breakpoints detected in urine were found to be mostly (>70%) in the HBV DR1-2 region, consistent with previous HBV-HCC tissue studies (8). Additionally, 11 HBV-JS’s identified in urine and reported previously in HCC tissue were all found in patients with cirrhosis or HCC. Urine DNA is likely to reflect the heterogeneity of HBV-JS’s in the liver more accurately compared with a single tumor biopsy or resection. HBV-JS’s have been detected in plasma of patients with HCC in recent studies (12-15). However, because urine cfDNA analysis is not confounded by the possible presence of replicating HBV or virions (20), urine is likely to provide a more sensitive liquid biopsy than blood for assessing integrated HBV DNA sequences in infected patients. Although we did not directly compare matched urine and plasma for HBV-JS detection due to plasma unavailability, the large number of HBV-JS’s we detected (712 in 32 urine samples) supports the notion that urine may be a more sensitive liquid biopsy for integrated HBV DNA.
Characterization of integrated HBV DNA and its urinary levels may provide insight into clonal expansion and turnover of infected hepatocytes, HCC risk stratification, monitoring recurrence post-treatment, and treatment efficacy. Frequent monitoring via urine liquid biopsy would be particularly advantageous in assessment of HBV-JS dynamics. The detection of HBV-host sequences and a chromosome rearrangement in HCC urine echoes our recent progress in developing a sensitive, noninvasive, urine-based HCC screening test (16, 19, 40). Further study with a much larger sample size is needed to determine the detection sensitivity and urine quantity needed for robust, reliable HBV-JS profiling and its clinical applications. In the present proof-of-concept study, we demonstrated that detection of HBV-JS’s in urine is feasible and holds potential for application in assessing HBV-related liver disease in infected patients.
Data Availability
The sequence data for the subjects studied in this work who consented to data archiving have been deposited in the European Genome Phenome Archive (EGA), www.ebi.ac.uk/ega, which is hosted by the European Bioinformatics Institute (EBI) www.ebi.ac.uk (accession no. pending).
Footnotes
Financial Support: This work was supported by the National Institute of Health grants R43 CA165312 (WS), R44 CA165312 (SL and WS), and R43 CA192507 (WS).
slin{at}jbs-science.com, etrauger{at}gmail.com; bensong721{at}gmail.com; mhoffman1996{at}gmail.com; fsong{at}jbs-science.com, robinsu328{at}gmail.com; egcthom{at}gmail.com, yu-lan.kao{at}bblumberg.org, yixiao.cui{at}bblumberg.org, ying-hsiu.su{at}bblumberg.org, ttchang{at}mail.ncku.edu.tw, lyj007{at}mail.ncku.edu.tw, grace.park{at}jefferson.edu, hie-won.hann{at}jefferson.edu.
List of Abbreviations
- HBV
- hepatitis B virus
- HBV-HCC
- HBV-infected hepatocellular carcinomas
- HBV-JS’s
- HBV-host junction sequences
- HCC
- hepatocellular carcinoma
- cfDNA
- cell-free DNA
- UMT
- unique molecular tag
- RE
- restriction endonuclease
- nt
- nucleotide