An improved droplet digital PCR assay reveals the stability of intrahepatic hepatitis B virus cccDNA.

The persistence of covalently closed circular DNA (cccDNA) poses a major obstacle to curing chronic hepatitis B (CHB). Here, we used droplet digital PCR (ddPCR) for cccDNA quantitation. ddPCR measured a less than two-fold difference in the intrahepatic cccDNA content more accurately than conventional real-time PCR (qPCR), (R 2 =0.9416 and R 2 =0.8963, respectively) and had also higher sensitivity and specicity than qPCR. The results of ddPCR correlated closely with serum HB core-related antigen and HB surface antigen (HBsAg) in 24 HBV-infected human-liver-chimeric mice (PXB-mice). We demonstrated that the total cccDNA content did not change during liver repopulation, although the cccDNA content per hepatocyte was reduced in PXB-mice after entecavir treatment. In the 6 patients with HBV-related hepatocellular carcinoma, ddPCR detected cccDNA in both tumor and non-tumor tissues. In 13 HBeAg-negative CHB patients with pegylated interferon alpha-2a, cccDNA contents from paired biopsies were more signicantly reduced in virological response (VR) than in non-VR at week 48 (p=0.0051). Interestingly, cccDNA levels were the lowest in VR with HBsAg clearance but remained detectable after the treatment. Collectively, ddPCR revealed that cccDNA content is stable during hepatocyte proliferation and persists at quantiable levels, even after serum HBsAg clearance. Linear HBV DNA fragments (nt 1372-2187) containing cccDNA-specic forward and reverse primer regions digested by NcoI and EcoRI from HBV genotype C2/Ce DNA genome (AB246345) were analyzed with and without PSAD treatment and conducted qPCR and ddPCR assays. When the linear HBV DNA in all dilution series (ranging 10 7 -10 5 copies/assay) without PSAD treatment was analyzed by the cccDNA specic primer-probe set, both ddPCR and qPCR was able to accurately measure total DNA (cccDNA and HBV DNA containing rcDNA and ssDNA) (R2=0.998 and R2=0.998, respectively). In contrast, when the serially diluted samples were measured by the cccDNA specic primer-probe set with PSAD, ddPCR was able to accurately measure cccDNA at a concentration of approximately 1/45 of total DNA in all dilution series (R2=1), but qPCR was measure cccDNA at a concentration of approximately 1/32 of total DNA in all dilution series (R2=1). These results suggest that the ne sensitivity of cccDNA-specic ddPCR, excluding linear HBV-DNA containing the amplicon and integrated cccDNA to 1/45 after PSAD treatment. The limits of quantitative value (LOQ) and detection value (LOD) for HBV-cccDNA-specic ddPCR were determined using serial dilutions of total DNA extracted from the livers of HBV-infected and non-infected chimeric mice (Fig. 2E). The LOD and LOQ in mouse liver were 1 copy/ ddPCR assay and 20 copies/ ddPCR assay, respectively, and those were 5.9 copies/qPCR assay and 60.2 copies/qPCR assay, respectively. Collectively, these results suggest that both ddPCR and qPCR detect intrahepatic cccDNA with high sensitivity, but ddPCR is more accurate than qPCR in measuring less than two-fold differences.


Introduction
Approximately 257 million people are estimated to be chronically infected with hepatitis B virus (HBV) worldwide and 887,000 deaths every year are attributed to sequelae of HBV infection [1]. Although nucleos(t)ide analogs (NAs) inhibiting viral DNA synthesis are approved antiviral therapies for chronic HBV (CHB) infection [2,3], they cannot eliminate the virus due to the persistence of covalently closed circular DNA (cccDNA) in the hepatocytes.
Currently, intrahepatic cccDNA is detected by quantitative real-time PCR (qPCR) or Southern blotting (SB). However, the amount of cccDNA in human hepatocytes is rather low, limiting the use of SB [4,5]. Because cccDNA is very stable, its reduction may be slow. Therefore, a more precise measurement is desirable to evaluate the impact of new anti-HBV drugs on cccDNA. Droplet digital PCR (ddPCR) provides an accurate copy number of a single gene sequence [4,5]. Several studies have been systematically compared the quantitative accuracies of ddPCR and qPCR in detecting intrahepatic cccDNA, but the kinetic analysis of cccDNA using ddPCR and its clinical application is limited. Meanwhile, clinical studies revealed the potential of serum hepatitis B surface antigen (HBsAg) and hepatitis B core-related antigen (HBcrAg) levels as surrogate biomarkers for intrahepatic cccDNA in CHB patients, but little information is available regarding the extent to which HBsAg marker re ects the cccDNA content in the liver in the absence of antibody response to HBsAg under immune suppression, as well as the possibility of HBsAg synthesized from integrated HBV DNA needs to be considered [6][7][8][9]. Moreover, the impact of cell division on intrahepatic cccDNA content remains contentious, partly due to the lack of a highly quantitative methodology.
Here, we developed a new assay to quantitate cccDNA content using ddPCR. This ddPCR assay could measure cccDNA more accurately and with greater speci city and sensitivity than qPCR. The amount of Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js cccDNA was stable during hepatocyte proliferation and that the amount of cccDNA were shown to be reduced upon PEG-IFN treatment in vivo.

Materials And Methods
Ethics statement.
This study was performed in accordance with the relevant national guidelines and regulations.
Chimeric mice with human hepatocytes.
Severe combined immunode ciency mice transgenic for the urokinase-type plasminogen activator gene (uPA wild/+ /SCID+/+ mice), with their livers replaced by human hepatocytes, (human-liver-chimeric mice) were purchased from PhoenixBio Co., Ltd. (Hiroshima, Japan). These mice were infected with sera obtained from human hepatocyte chimeric mice previously infected with HBV subgenotype C2/Ce, as described in a previous report [10]. 24 HBV-infected chimeric mice were under various conditions at the time of autopsy, and seven mice showed low HBV infection e cacy due to ETV treatment during the acute phase of infection (Table S1). The in-life phase of this study was outsourced to PHOENIXBIO Co., Ltd. The protocol of this study was reviewed by Ethics Committee of PHOENIXBIO, approved on June 15, 2017(Resolution No.: 1910 prior to the start of the in-life phase on June 16, 2017 and carried out in compliance with the ARRIVE guidelines. More detailed procedures are given in the Supplementary Information.

Patients.
A total of 6 HCC patients and 13 hepatitis B e antigen (HBeAg)-negative patients with CHB were enrolled in this study. Liver tumor tissues and the corresponding non-tumor tissues were collected from ve patients with HBV infection (3 HBsAg-positive and 2 HBsAg-negative/anti-HBc-positive) and one patient without HBV infection (HBsAg-negative and anti-HBc-negative) at Tokushima University Hospital and Nagoya City University Hospital in Japan. Thirteen HBeAg-negative patients who received 48 weeks of PEG-IFN-alfa2a (180 µg/week) monotherapy at the King Chulalongkorn Memorial Hospital, Bangkok, Thailand, were enrolled into this study, and paired liver tissues were collected before and after PEG-IFN treatment. Virological response (VR) was de ned as an HBV DNA level below 2,000 IU/mL at 48 weeks after treatment. Thirteen patients with HBeAg-negative CHB who received 48 weeks of PEG-IFN monotherapy were categorized into 3 groups: non-VR (n=6), VR without HBsAg clearance (n=4) and VR with HBsAg clearance (n=3). At baseline, there was no signi cant difference in the levels of HBV DNA, HBsAg, or HBcrAg in the serum, or the levels of total cccDNA and cccDNA per cell in the liver, among the 3 groups (p=0.26473, p=0.64468, p=0.69439, p=0.30080, and p=0.5565, respectively) ( Serum HBV DNA titers were measured in chimeric mice by real-time PCR, as described previously [11,12]. Hepatitis B surface antigen (HBsAg) and hepatitis B core-related antigen (HBcrAg) were measured by chemiluminescent enzyme immunoassays using commercial kits (FUJIREBIO Inc., Tokyo, Japan), as described previously [10,13]. The detection limits of the HBsAg and HBcrAg assays are 0.005 IU/ml and 3 logU/ml, respectively. The HBV DNA titers were measured in patients using the Abbott Real Time HBV assay or the TaqMan polymerase chain reaction assay (COBAS TaqMan, ROCHE MOLECULAR SYSTEMS [lower detection limit: 20 IU/ml]) or Abbott RealTime HBV assay (lower detection limit: 10 IU/ml). More detailed procedures are given in the Supplementary Information.
Isolation of genomic DNA from the livers of chimeric mice.
Genomic DNA was isolated from the livers of chimeric mice using the phenol/chloroform method, as described previously [14]. Protein-free DNA extraction from the mouse livers was carried out using a modi ed Hirt extraction procedure [15]. The genomic DNA and protein-free DNA were digested with DNase-free RNase and restriction enzyme (Hind III) before analysis. More detailed procedures are given in the Supplementary Information.
Quantitation of HBV cccDNA and relaxed circular (rc) DNA by ddPCR.
Total liver DNA treated with plasmid-safe ATP-dependent DNase (PSAD) (EPICENTRE, Madison, Wisconsin, USA) was used as template inputs for either qPCR or ddPCR ampli cation. The cccDNA speci c primersprobe set for cccDNA ampli cation designed as shown in Fig. 1A, B was used for both ddPCR and qPCR assays. After generating reaction droplets, intrahepatic cccDNA was ampli ed using a C1000 touch Thermal Cycler (BIO-RAD LABORATORIES, Hercules, California, USA). In some cases, intrahepatic cccDNA values were normalized by the cell number measured by the hRPP30 copy number variation assay (BIO-RAD LABORATORIES, Pleasanton, California, USA) [16]. Of note, hRPP30 levels were separately determined using DNA that was not treated with PSAD, which digests genomic DNA. Since the amount of DNA genome diluted to various concentrations in the range of 125-1,000 ng/assay in a vial accurately measured the number of cccDNA (R 2 value greater than 0.99), up to 1,000 ng of cccDNA per vial was measured in HCC case. More  Genomic DNA and protein-free DNA was isolated from the livers of chimeric mice infected with HBV. The DNAs were digested with the restriction enzyme, Hind III, that does not cleave HBV, and analyzed by Southern blotting. The signals were analyzed using an ImageQuant LAS 4000mini (GE HEALTHCARE, United Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Kingdom). To con rm the nature of cccDNA, the samples were subsequently digested with Xho I to linearize the DNA. Agarose gel electrophoresis and Southern blotting were performed as described previously [17]. Statistical Analysis.
The coe cient of determination (R 2 ) was assessed by linear regression analysis. The correlation between two continuous variables was analyzed using Spearman's rank correlation coe cient. The Mann-Whitney Utest was used to examine statistical difference between the VR and non-VR groups. The limits of quantitative value (LOQ) and detection value (LOD) for HBV-cccDNA-speci c ddPCR were determined using serial dilutions of total DNA extracted from non-infected and HBV-infected livers (chimeric mice or human) using classical method. The LOD was de ned as the lowest concentration at which 50% of hit rate of cccDNA detection by probit analysis. The LOQ was de ned as the coe cient of variation (CV) lower than 0.2 at the levels of cccDNA/assay using regression analysis. The LOD and LOQ in the mouse liver were 1 copies/assay and 20 copies/assay, respectively. The LOD and LOQ in the human liver were 4 copies/assay and 15 copies/assay, respectively. The Statistical analyses were performed using BellCurve for Excel (version 3.21) (SOCIAL SURVEY RESEARCH INFORMATION Co., Ltd., Tokyo, Japan).

Results
Optimization and validation of ddPCR for the detection of HBV cccDNA.
In the HBV genome, the incomplete plus strand has a variable 3'-end but a de ned 5'-end at the position of 1592 near DR2, while the complete minus strand has de ned 5'-and 3'-ends with a terminal redundancy of 9 bases. There is a gap at the position of 1826 near DR1 ( Fig. 1A and B). To discriminate between the relaxed circular DNA (rcDNA) and the cccDNA of the HBV genome, we designed the primer to bind the negative strand on one side of the gap, and the probe to bind the same strand on the other side. Therefore, the forward and reverse primers, 5'-ACGGGGCGCACCTCTCTTTACGCGG-3' [nt: 1519-1543] and 5'-CAAGGCACAGCTTGGAGGCTTGAAC-3' [nt: 1862-1886], were designed against the minus and plus strand gaps of rcDNA and selected to speci cally amplify DNA fragments from replication intermediate cccDNA but not from viral genome DNA. A TaqMan probe, which binds to DNA when PCR is performed, is composed of FAM and MGB quencher (Fig. 1B). To establish an assay to detect HBV cccDNA using ddPCR technology, we used total DNA extracted from the livers of human-liver-chimeric mice that had been infected with HBV ( Fig.   1). Digestion of DNA by a restriction enzyme is believed to improve template accessibility [18]. To test the effect of restriction enzyme digestion on the e cacy of cccDNA detection, temperature gradient ddPCR was performed in the presence of Hae III with fteen restriction sites in the HBV genome outside the cccDNA ampli cation region on the sample DNA treated with PSAD. Xho I, which does not recognize the cccDNA ampli cation region but cuts at one site in the HBV genome, was used for comparison. As shown in Supplementary Fig. S1A and B, Hae III digestion provided better separation of positive and negative signals, and the results were less affected by the temperature as the signal intensity of positive droplet populations started to decrease at the annealing temperature below 58.5℃ for Xho 1 and 55.3℃ for Hae III. Based on these results, we decided to use Hae III digestion and set the annealing temperature at 61.2℃. Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js ddPCR accurately measures less than two-fold differences in cccDNA content.
To determine the performance of the ddPCR method for HBV cccDNA quanti cation, the plasmid containing the HBV genotype C2/Ce DNA genome (AB246345) was serially diluted from 10 5 to 10 1 copies/assay and determined by ddPCR. As a results, the cccDNA-speci c ddPCR showed an excellent linear correlation between expected and observed HBV cccDNA copies/assay (R 2 =0.9997).
To compare the sensitivity and accuracy of cccDNA detection between ddPCR and qPCR, we serially diluted total DNA extracted from the livers of HBV-infected chimeric mice from 20 to 0.156 ng/assay and conducted ddPCR and qPCR. As shown in Fig. 2A, the results of both ddPCR and qPCR displayed high goodness of t (R 2 =0.9935 and R 2 =0.9685, respectively). In addition, the results of ddPCR and qPCR exhibited a near-perfect correlation (R 2 =0.9958, p=2.12x10 -70 ) (Fig. 2B). These results suggest that both ddPCR and qPCR are highly sensitive and su ciently accurate to detect 2-fold differences of the intrahepatic cccDNA. In contrast, when the samples were diluted to various concentrations, ranging between 5 and 8 ng/assay, ddPCR but not qPCR was able to accurately measure the differences (R 2 =0.9416 and R 2 =0.8963, respectively) (Fig. 2C). In addition, ddPCR was found to more accurately detect less than 2-fold difference than qPCR in other assay ranges (R 2 = 0.9980 and R 2 =0.9972, respectively at higher assay range;12-20 ng/assay and R 2 =0.9974 and R 2 =0.9946, respectively at lower assay range; 4-6 ng/assay).
Additionally, the sensitivity of cccDNA detection was compared between ddPCR and qPCR. Linear HBV DNA fragments (nt 1372-2187) containing cccDNA-speci c forward and reverse primer regions digested by NcoI and EcoRI from HBV genotype C2/Ce DNA genome (AB246345) were analyzed with and without PSAD treatment and conducted qPCR and ddPCR assays. When the linear HBV DNA in all dilution series (ranging 10 7 -10 5 copies/assay) without PSAD treatment was analyzed by the cccDNA speci c primer-probe set, both ddPCR and qPCR was able to accurately measure total DNA (cccDNA and HBV DNA containing rcDNA and ssDNA) (R2=0.998 and R2=0.998, respectively). In contrast, when the serially diluted samples were measured by the cccDNA speci c primer-probe set with PSAD, ddPCR was able to accurately measure cccDNA at a concentration of approximately 1/45 of total DNA in all dilution series (R2=1), but qPCR was measure cccDNA at a concentration of approximately 1/32 of total DNA in all dilution series (R2=1). These results suggest that the ne sensitivity of cccDNA-speci c ddPCR, excluding linear HBV-DNA containing the amplicon and integrated cccDNA to 1/45 after PSAD treatment. The limits of quantitative value (LOQ) and detection value (LOD) for HBV-cccDNA-speci c ddPCR were determined using serial dilutions of total DNA extracted from the livers of HBV-infected and non-infected chimeric mice (Fig. 2E). The LOD and LOQ in mouse liver were 1 copy/ ddPCR assay and 20 copies/ ddPCR assay, respectively, and those were 5.9 copies/qPCR assay and 60.2 copies/qPCR assay, respectively. Collectively, these results suggest that both ddPCR and qPCR detect intrahepatic cccDNA with high sensitivity, but ddPCR is more accurate than qPCR in measuring less than two-fold differences.
Speci c detection of cccDNA by ddPCR.
The speci city of cccDNA detection was compared between ddPCR and qPCR. Serially diluted total DNA extracted from the sera of HBV-infected chimeric mice, from 0.32 to 0.04 µl/assay, was analyzed by ddPCR and qPCR. Several studies reported that total DNA in the sera and the culture medium could contain a large amount of HBV DNA and a small amount of cccDNA [19,20,21], indicating that the cccDNA-speci c ddPCR of the serum should generate a very low signal. As references, the serum HBV DNA levels of the HBV-infected chimeric mice were measured using a primer-probe set that could detect both cccDNA and rcDNA (herein referred to "set 1") [12]. As shown in Fig. 2D, the results of both ddPCR and qPCR using the set 1 primerprobe set showed high concordance (R 2 =0.9994 and R 2 =0.9906, respectively). When these serum samples were analyzed by the cccDNA speci c primer-probe set without PSAD treatment, both ddPCR and qPCR was able to accurately measure cccDNA at a concentration of approximately more than 1/600 of total DNA in all dilution series, ranging between 0.32 and 0.04 µl/assay (R 2 =0.9022 and R 2 =0.9601, respectively).
Moreover, when the serial diluted serum samples were measured by the cccDNA speci c primer-probe set with PSAD treatment, ddPCR was able to accurately measure cccDNA at a concentration of approximately 1/1000 of total DNA in all dilution series, ranging between 0.32 and 0.04 µl/assay (R 2 =0.9915), but qPCR was unstable at low cccDNA concentrations (R 2 =0.9259, Fig. 2D). Because expressing uPA in the liver of the uPA/SCID mice with human hepatocytes induces a su ciently damaged liver (Tateno et al., PLoS One, 2015), we suggest that the cccDNA in hepatocytes could ow out to the blood and the cccDNA could be detectable in sera of the chimeric mice. Unfortunately, although we performed Southern blotting (SB) experiments, the mice serum volumes are too small to detect any signals after PSAD treatment; we were unable to provide data whether the detected signal was cccDNA or rcDNA in false-positive by ddPCR. Therefore, we tried to determine the cccDNA in sera from 7 inactive HBV carries with normal liver function using the HBV cccDNA-speci c ddPCR with PSAD treatment. The assay was conducted in triplicate. The cccDNA signal was undetectable in all samples with HBV-DNA levels up-to 5.1 log IU/mL, suggesting that the speci city of the ddPCR assay for cccDNA was 100% (0/21), and the assay could detect cccDNA in mice serum. Finally, we compared ddPCR and SB analysis using DNA isolated from two chimeric mice in the chronic phase of HBV infection with or without ETV treatment ( Supplementary Fig. S2). As shown in the Supplementary Information, these values were closely correlated with the cccDNA content detected by the SB assay.
Collectively, these results suggest that ddPCR could detect intrahepatic cccDNA with higher speci city and sensitivity than qPCR, depending on the cccDNA-speci c ddPCR, especially after PSAD treatment.
The intrahepatic HBV cccDNA level correlates strongly with serum HBV parameters.
To determine the extent to which the intrahepatic cccDNA level is re ected by the HBV markers in the serum, the cccDNA contents were measured in 24 HBV-infected chimeric mice during 2 conditions of infection; spreading phase/partially infected mice (n=8, low HBV group) and stable phase/fully infected mice (n=16, high HBV group), and then correlated with the serum HBsAg and HBcrAg levels in each condition. As shown in Fig. 3A and B, the intrahepatic cccDNA levels were more signi cantly correlated with both the serum HBsAg level (R 2 =0.5639, P=3.5442x10 -5 ) and the HBcrAg level (R 2 =0.6703, P=2.0217x10 -4 ) in HBV high group Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js compared with the low HBV group (R 2 =0.4091, p=0.00745 and R 2 =0.7093, p=0.01912, respectively), due to the small number of mice in Low HBV levels group. These results suggest that both HBcrAg and HBsAg are excellent surrogate serum markers of intrahepatic cccDNA.
Hepatocyte proliferation has little impact on the total amount of cccDNA in the liver.
Having established a highly quantitative cccDNA-speci c ddPCR assay, we examined the impact of hepatocyte proliferation on the stability of cccDNA. All six chimeric mice undergoing hepatocyte repopulation were inoculated with HBV (1x10 6 copies per mouse) 4 weeks after transplantation, at which time the reconstitution rate was less than 30%, based on serum human albumin (hALB) levels (Fig. 4A) [22]. As shown in table S1, the 6 mice derived from this experiment were indicated with an asterisk (week4#1: * , week4#2: † , week10 ETV #1: ‡ , week10 ETV #2: § , week10 control #1: || , week10 control #2: ¶ , respectively.). Serum HBV DNA levels were measured in all mice 4 weeks later, ranging between 1.7 to 6.8 x10 5 copies/ml (Fig. 4B). Two chimeric mice were sacri ced on week 4 and two of the remaining four were administered ETV orally at 0.2 mg/kg/day for 4 weeks. The other two mice served as controls. The HBV DNA levels had already decreased after one week of ETV treatment and continued to decrease during the treatment, reaching a nadir on week 10 (1.6 log 10 copies reduction) (Fig. 4B). In the ETV-treated chimeric mice, serum HBsAg and HBcrAg levels peaked at week 6 and week 5, respectively, and then decreased slightly ( Fig. 4C and   D). These markers continued to increase in the control group during the same period (Fig. 4B, C and D).
Serum human albumin levels increased in both groups (Fig. 4E), re ecting liver repopulation, but the body weights did not change (Fig. 4F). As shown in Fig. 4G and H, HBV rcDNA, ssDNA and cccDNA were readily detectable in the control mice at week 10, but not in the ETV treated mice, indicating the low sensitivity of SB.
We then analyzed the levels of intrahepatic cccDNA at weeks 4 and 10 after HBV infection (i.e. immediately before the ETV treatment (week 4) and two weeks after the termination of ETV treatment (week 10)) ( Fig. 4I). Before ETV treatment, 96.0-142.0 copies of cccDNA were detected in 1,000 ng of total liver DNA. Interestingly, almost the same amounts of cccDNA were detected in the ETV-treated mice at week 10 after HBV infection, consistent with the serum HBcrAg and HBsAg levels. In contrast, cccDNA levels in the control mice increased markedly to 9,800 and 19,800 copies/1,000 ng), a greater than 100-fold increase in 6 weeks. To determine the amount of cccDNA/hepatocyte, intrahepatic hRPP30 was quanti ed at the corresponding time points. As shown in Fig. 4J, the hRPP30 level increased by approximately 2.6 and 2.3-fold in the ETV-treated and control mice, in 6 weeks. The results are consistent with the increase of serum hALB levels (Fig. 4E) and suggest the ongoing human hepatocyte repopulation of the chimeric mouse livers. Consequently, the cccDNA level per hepatocyte decreased from 0.016 and 0.024 copies/cell before ETV treatment to 0.011-0.012 copies/cell after ETV treatment (Fig. 4K), while it increased to 1.361 and 1.886 copies/cell in the control mice.
Collectively, these results suggest that the total amount of cccDNA in the liver is not affected by hepatocyte repopulation, although the cccDNA level/hepatocyte appears to be reduced after ETV treatment. Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Intrahepatic HBV cccDNA in patients with HBV-associated hepatocellular carcinoma.
To apply this ddPCR-based method to clinical samples, we measured the amount of intrahepatic cccDNA and RPP30 in tumor and non-tumor tissues from ve HBV-related HCC patients using the ddPCR assay. As shown in Table 1, varying levels of intrahepatic cccDNA were detected in three HBsAg-positive patients among the ve HBV-related HCC patients. In two HBsAg-positive, HBcrAg-positive patients (patients 3 and 4), the cccDNA levels were 0.00933-0.05081 copies/cell in non-tumor tissues, and 0.93442-1.17975 copies/cell in tumor tissues. In the HBsAg-positive, HBcrAg-negative patient (patient 2), cccDNA was detected weakly at 0.00013 copies/cell in tumor tissue, while it was undetectable in non-tumor tissue. In two HBV-related HCC patients whose HBV infections had resolved, cccDNA levels were below the detection limit. Intrahepatic RPP30 in all 6 HCC patients, including the negative control, were 36,100-109,500 and 51,650-104,000 copies/1000 ng in non-tumor and tumor tissues, respectively. These data showed that ddPCR could detect cccDNA in tumor and non-tumor tissues.
The impact of PEG-IFN treatment of patients with HBeAg-negative CHB on intrahepatic cccDNA levels.
To investigate the effect of PEG-IFN on intrahepatic cccDNA, we determined the levels of intrahepatic cccDNA in patients with HBeAg-negative CHB at baseline and 48 weeks after PEG-IFN treatment (Fig. 5,   Supplementary Fig. S3 and Table S2). Thirteen patients with HBeAg-negative CHB who received 48 weeks of PEG-IFN monotherapy were categorized into 3 groups: non-VR, VR without HBsAg clearance and VR with HBsAg clearance. At baseline, there was no signi cant difference in the levels of HBV DNA, HBsAg, and HBcrAg in the serum, or the levels of total cccDNA and cccDNA per cell in the liver, among the three groups (Supplementary Table S2). As shown in Fig. S3 and Table S2, serum HBV DNA levels in the non-VR group either did not change or decreased temporarily at the end of PEG-IFN treatment, only to rebound thereafter. In the VR without or with HBsAg clearance groups, HBV DNA levels decreased continuously after PEG-IFN treatment ( Supplementary Fig. S3). Serum HBsAg levels in the non-VR and VR without HBsAg clearance groups were 2.8±1.1 (1.32-3.26) and 1.1±0.9 (0.40-2.38) log IU/mL, respectively, at week 48, whereas HBsAg levels were undetectable at weeks 12 and 24 in the VR with HBsAg clearance group ( Supplementary Fig. S3).
Serum HBcrAg levels in the non-VR group could be quantitated (4.3±1.4 (2.8-6.6) Log U/mL), while those in the VR without and with HBsAg clearance groups were mostly under the quantitation limit at week 48 ( Supplementary Fig. S3). Interestingly, the levels of intrahepatic cccDNA and cccDNA per cell were signi cantly lower in the VR groups than the non-VR group at week 48 (p=0.0160 and p=0.0432, respectively), associated with a signi cant reduction of cccDNA per cell (p=0.0051) (all data shown in Fig. 5A-F and Supplementary Table S2). These signi cant changes in cccDNA levels in the VR groups were associated with a higher peak of ALT than the non-VR group (p=0.0165) (Fig. 5A-G). In particular, in the VR with HBsAg clearance group, intrahepatic cccDNA levels at week 48 after PEG-IFN treatment were lower than those in the VR without HBsAg clearance group (p=0.036), in association with the higher ALT peak (p=0.060) (Fig. S4A-B), although HBcrAg levels in most VR patients were undetectable, due to the limited sensitivity. Remarkably, cccDNA levels could be readily quantitated even in patients (11 and 12) with HBsAg clearance after PEG-IFN treatment ( Fig. 5C and

Discussion
In this study, we developed a ddPCR assay for the quantitation of HBV cccDNA, with more sensitivity, speci city and accuracy than qPCR ( Fig. 1 and 2). The speci city of ddPCR for cccDNA was validated by the SB assay results (Fig. 3). The intrahepatic cccDNA was closely correlated with both HBsAg and HBcrAg in the sera of HBV-infected chimeric mice (Fig. 3). Notably, the total amount of cccDNA in the repopulating liver did not change in the absence of active HBV DNA replication during ETV treatment, although the cccDNA content/hepatocyte was reduced by hepatocyte proliferation (Fig. 4). To our knowledge, this is the rst study indicating the stability of cccDNA in the proliferating liver, and this novel nding could be obtained because of the high accuracy of ddPCR. The ddPCR was also able to quantitate cccDNA in CHB patients with HCC and those with HBsAg clearance after PEG-IFN treatment (Table 1). Importantly, ddPCR readily quantitated the intrahepatic cccDNA content, even after HBsAg clearance (Fig. 5).
It is generally believed that the amount of cccDNA should be reduced to a very low level to cure HBV infection. Thus, a reliable assay to measure cccDNA is required to evaluate the therapeutic e cacy of new anti-HBV treatments. While qPCR has a theoretical limit of quanti cation between two samples is 2-fold (corresponding to a single PCR cycle threshold), ddPCR counts the total number of individual target molecules in digital form. Therefore, ddPCR is a new technique with improved accuracy over qPCR [23].
Furthermore, ddPCR was reported to be less variable between experiments than qPCR, allowing long-term monitoring [24]. Several recent studies have applied the ddPCR technology to measure HBV cccDNA levels in cultured cells and the livers of CHB patients, and demonstrated the higher sensitivity of ddPCR than qPCR [25,26,27].
As explained above, several studies both ddPCR and qPCR technologies have already utilized to measure cccDNA using cccDNA-speci c primers and probes. However, we believe this is the rst study that applied the ddPCR technology to test the impact of hepatocyte proliferation on the stability of cccDNA. We found that the cccDNA content in the liver is extremely stable during 2-3 times of cell division, and this discovery was possible because of the ability of ddPCR to detect differences less than two-fold. Previous study has already used the ddPCR method to study the effect of nucleoside analogue on intrahepatic cccDNA levels but did not address the issue of IFN-treated patients [25]. We subsequently showed that our ddPCR can be applied to human samples including HCC and IFN-treated patients.
Although we could not distinguish whether the signals detected in serum of the chimeric mice were false positive rcDNA or cccDNA by SB, the cccDNA signal was undetectable in sera from all 7 inactive HBV carriers with HBV-DNA levels up-to 5.1 log IU/mL, suggesting that the speci city of the ddPCR assay for cccDNA was 100%, and the assay could detect cccDNA in mice serum. In the current study, the novel ddPCR assay enabled more speci c detection of small amounts of cccDNA, even in the presence of large amounts of HBV DNA, indicating more accuracy, speci city and sensitivity than qPCR (Fig. 2C, D and E). Combined with the previous nding that ddPCR is more reproducible than qPCR [24], ddPCR might be more appropriate for measuring cccDNA levels than qPCR in clinical settings, because most anti-HBV drugs do not directly target cccDNA and, therefore, require long-term treatment to achieve a meaningful cccDNA reduction.
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js The high accuracy, speci city and sensitivity of ddPCR allowed us to determine the impact of hepatocyte proliferation on cccDNA. Surprisingly, the total amount of cccDNA in the liver showed no reduction, even though the number of hepatocytes doubled in the absence of active HBV replication (Fig. 4I, week 4 vs. week 10 with ETV). Because of the small number of mice involved, however, these results are not statistically signi cant. Nevertheless, the data suggest that cccDNA is extremely stable during hepatocyte repopulation.
Our ndings differ from those recently reported by Allweiss et al.; total cccDNA levels in the liver appeared to be reduced by the proliferation of hepatocytes [28]. This discrepancy presumably re ects different experimental conditions. In their study, HBV-infected hepatocytes were transplanted to uPA-SCID mice, and the baseline cccDNA levels were measured almost immediately after the transplantation of HBV-infected hepatocytes under conditions leading 1,000 times hepatocyte division. In contrast, as HBV could e ciently infect with the chimeric mice with a replacement rate of more than 30% of human hepatocytes in our PXBmice (personal communication), we inoculated HBV at 4 weeks after liver transplantation and conducted the NA administration test with reference to human ALB value that predicts the replacement rate of human hepatocytes. Hence, our experiments would have only 2-3 times hepatocyte divisions, the stability of cccDNA during multiple hepatocyte divisions needs to be investigated in the future. This notion is in line with the observations in NA treated woodchucks chronically infected with woodchuck hepatitis virus [29] and growing ducklings infected with duck HBV [30]. These observations support the hypothesis that some fraction of cccDNA was distributed to the daughter cells of infected hepatocytes (Supplementary Fig. S6A). Alternatively, it is possible that the liver reconstitution after HBV infection mainly re ected the proliferation of uninfected hepatocytes, while the majority of HBV-infected hepatocytes did not proliferate during the same period of time ( Supplementary Fig. S6B). Because of the small number of mice involved in this study, we did not conclude the signi cance of cccDNA maintenance in cell division, and further studies are required to distinguish these alternatives.
Secreted HBcrAg is considered a good surrogate marker for intrahepatic cccDNA. Also, middle and large HBsAg might re ect the activity of cccDNA, although HBsAg is partially driven from integrated HBV DNA [31]. Patients with low serum HBsAg and HBcrAg levels have been reported to have a low risk of relapse after cessation of NAs therapy [8,32,33,34,35]. Recent studies indicated that serum HBcrAg was a better surrogate marker of intrahepatic cccDNA than serum HBsAg [36,37,38]. Quantitation of cccDNA by ddPCR recapitulated those previous ndings, as serum HBcrAg, but not HBsAg, was also signi cantly correlated with intrahepatic cccDNA in HBeAg-negative CHB patients (p=0.00056, Supplementary Fig. S5). In human samples, HBcrAg was more closely correlated with cccDNA than HBsAg. Besides, a good positive correlation between serum HBsAg and cccDNA was obtained in chimeric mice with immunode ciency (Fig. 3). This discrepancy presumably re ects the presence of anti-HBs and/or the presence of HBsAg-producing viral integrations in human samples. As expected, the more profound reduction of cccDNA in the VR than non-VR groups was associated with higher peak ALT activities after PEG-IFN treatment (Fig. 5), indicating a role of cell death in elimination of cccDNA. Interestingly, Lucifora et al. suggested that APOBEC3 family cytidine deaminase A3A mediates the IFN-a induced cccDNA degradation [39].
In conclusion, a sensitive, speci c, highly quantitative ddPCR was developed for the measurement of cccDNA in the liver. This new method revealed that intrahepatic cccDNA is stable during hepatocyte proliferation and Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js persists at quanti able levels, even after serum HBsAg clearance. The use of ddPCR will provide valuable information on the development of new drugs against CHB infection.   were measured using a primer-probe set that detects both cccDNA and rcDNA. Serum cccDNA levels were determined using the cccDNA-speci c primers and probe set by ddPCR or real-time PCR, with PSAD treatment. (E) Intrahepatic HBV DNA and cccDNA levels in the liver of chimeric mouse infected with HBV were measured using various amounts of DNA ranging from 20 to 0.02 ng/assay by ddPCR or real-time PCR. Each assay was conducted in triplicate. Target concentrations were repeatedly and independently tested three times.

Figure 3
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Correlation between intrahepatic cccDNA and serum HBV markers. Twenty-four HBV-infected chimeric mice under various conditions were divided into the HBV infection as during spreading phase of partially infection (n=8; #1-2, #13-16, #19, #20) and the stable phase of fully infection (n=16; #3-12, #17-18, #22-24). The cccDNA contents were correlated with (A) serum HBsAg and (B) serum HBcrAg in each group. Characteristics of each mouse were described in Table S1.   Whitney U-test was used to examine statistical difference between the two groups. The mean values of the two groups were depicted by red dots, and were 217.7 (30.7-674.0) in the non-VR group and 39.8 (0. 4-256.8) in the VR group. (E) Each dot represents the average of duplicate tests for the copy number of cccDNA per hepatocyte of each patients. The Mann-Whitney U-test was used to examine the statistical difference between the two groups. The mean values of the two groups were depicted by red dots, and were 0.088 (0.013-0.293) in the non-VR and 0.0206 (0.0002-0.0659) in the VR groups. (F) The reduction of cccDNA/cell (%) from baseline to the end of treatment was calculated in each patient, and the statistical difference was determined by the Mann-Whitney U-test. Red dots depict the mean values in the two group. (G) The peak levels of ALT during PEG-IFN treatment were 88.5 (58-111) in the non-VR and 222.0 (93-500) in the VR group.
Comparison between two groups was performed by the Mann-Whitney U test.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.