Evidence of an Infectious Virus Reservoir in Suppressed Chronic Hepatitis B Patients

Current approved therapies for Hepatitis B Virus (HBV) effectively control viral replication but are rarely curative. Patients on suppressive therapy still harbor cccDNA and cessation of therapy leads to rapid viral rebound 1-3. Even after long-term therapy of nucleos(t)ide analog (NA) treatment, the majority of chronic hepatitis B (CHB) patients have detectable but not quantiable HBV DNA 4. While these data suggest viral suppression is not absolute with current NA therapies, it has not been determined whether the detectable HBV DNA represents infectious virus. Here, we show that serum from patients on suppressive therapy with viral loads below the limit of quantitation still contain infectious circulating virus. We used a chimeric mouse model with a humanized liver to demonstrate HBV-infected human serum can establish infection similar to that in humans. We determined the limit of detection of the model using diluted patient serum, then measured the infectivity of serum from patients suppressed on therapy. Our results support the hypothesis that residual virus can serve as a reservoir to maintain a chronic infection in patients on therapy. These results highlight the urgent need for orthogonal therapies to eliminate low levels of circulating infectious virus which may minimize ongoing chromosomal damage to the liver and potentially lead to a cure. levels of serum HBsAg and HBV RNA, and low-to-detectable levels of intrahepatic cccDNA, suggesting that intrahepatic cccDNA is still transcriptionally active. The low-level viremia in NA-treated patients may serve as a viral reservoir leading to the de novo formation of cccDNA and maintenance of chronic HBV infection in the liver. The continued low-level viral replication would explain the viral rebound that is observed upon discontinuation of NA therapy. Moreover, the residual viremia may also explain the continued risk, although low, for the development of long-term HBV-related complications, such as HCC, in virally suppressed patients. Studies have demonstrated that patients on NA therapy that achieve functional cure, dened as HBsAg loss, have further reduction in HCC risk compared to those that remain virally suppressed 20,21 . However, HBsAg loss is uncommon in NA-treated patients 22,23 . These results highlight the need for orthogonal approaches (i.e., capsid modulators, RNAi, or immune modulation) in addition to NA therapy to achieve HBV functional cure.


Introduction
Hepatitis B is a global epidemic and a major cause of liver disease, including cirrhosis and liver cancer. The WHO estimates approximately 257 million people suffer from CHB worldwide and viral hepatitis accounted for 1.34 million deaths in 2015 (66% due to HBV) 5 . Widespread vaccination programs have been implemented to control the global prevalence rate of HBV. However, the vaccine does not currently prevent all new infections and is not helpful to patients that already have chronic infection 6,7 . Current direct-acting antiviral treatments for CHB target the HBV polymerase. With time, these drugs lead to a suppression of HBV replication in most patients 8 . Although viral suppression can be achieved in the majority of patients on long-term NA therapy, treatment rarely results in functional cure (e.g HBsAg seroconversion) and patients continue to be at-risk for hepatocellular carcinoma (HCC) 9-12 . We previously reported that after 240 weeks of TDF or TDF/FTC treatment the majority of CHB patients had detectable (i.e., PCR positive) but not quanti able (DNQ) HBV DNA (LLOQ ≤29 IU/mL) 13 . Moreover, HIV-HBV co-infected patients on long-term TDF therapy still have very low but detectable levels of intrahepatic cccDNA 14 . Taken together, these data suggest that viral suppression is not absolute with the current NA therapies, whether DNQ serum HBV DNA contains infectious virus is unknown. To determine if serum from patients with detectable but not quanti able HBV DNA contains infectious virus capable of infecting new hepatocytes, we utilized the uPA/SCID chimeric mouse model. Using this mouse model, we assessed the infectivity of serum from patients prior to and after long-term NA therapy.

Results And Discussion
To determine if human serum containing HBV was capable of infection in a mouse model of hepatitis B virus infection, patient isolates from Gilead study GS-US-203-0101 were evaluated in a uPA/SCID mouse model of HBV. SCID mice harboring the urokinase plasminogen activator gene and lacking an intact mouse liver are xenotransplanted with human hepatocytes (Figure 1). Mice with reconstituted human livers can be infected with HBV and although the mice lack a functional complete immune system, the disease course is similar to CHB in humans 15,16 . In this model, serum HBV DNA, HBsAg and HBeAg levels reach levels similar to those seen in CHB patients. Ultimately, the mice are unable to resolve the infection without intervention 17 . To determine if sera from patients could infect uPA/SCID mice, we identi ed a set of patients who had been infected with HBV genotype B and C and had high viral loads at baseline (mean viral load of 1.5 x 10 9 copies/mL) ( Figure 2, Supplementary Table 1). Baseline sera from each patient were inoculated by tail vein injection into uPA/SCID mice (n=2 mice/patient serum sample) and monitored for the presence of infection. Mouse sera were analyzed biweekly for HBV DNA by quantitative PCR and HBsAg/HBeAg by ELISA for 112 days. Maximal viremia and antigen production were achieved at between 40 -60 days post-infection (average 5.5 x 10 8 copies/mL HBV DNA, 4.5 x 10 4 IU/mL HBsAg, 3.3 x 10 3 cut-off index HBeAg) and continued to increase until mice were sacri ced at day 112 post-infection ( Figure 3). Therefore, baseline patient sera with high viral titers were able to establish infections in uPA/SCID mice.
To determine the lower level of sensitivity of the uPA/SCID mouse model, one patient from each genotype (B or C) was selected for further analysis. Baseline patient sera were serially diluted such that individual mice received 100 uL of diluted sera containing 100 (n=4), 10 (n=7), 5 (n=3) or 1 (n=3) genome equivalents (GE). Mice were followed for signs of infection for 112 days by monitoring mouse sera biweekly for HBV DNA by quantitative PCR and HBsAg/HBeAg by ELISA. Mice became viremic when inoculated with human sera containing 100 (4/4, 100%), 10 (6/7, 85%), and 5 (2/3, 67%) GE ( Figure 4A -C). Mice inoculated with sera containing 1 GE failed to become viremic after 112 days of observation (0/3, 0%) for either genotype tested. Mice inoculated with more virus became viremic faster, with the median time to viremia was 64.5, 94.5, and 96 days for 100 GE, 10 GE, and 5 GE, respectively. As no mice became viremic after inoculated with 1 GE, we can deduce that > 1 and < 5 GE is su cient for establishing infection in the uPA/SCID mouse model.
Chronic NA therapy is rarely curative as it is hypothesized viral suppression is not complete. Despite reaching clinical suppression (DNQ = <29 IU/mL), viral nucleic acids are still detectable in patient serum. Thus, it is likely there are still low levels of virus being produced, maintaining a pool of infected hepatocytes, even on NA therapy. To con rm if patients suppressed on NA therapy still have low levels of circulating virus, we infected mice with patient serum with viral DNA below the limit of quantitation (Table 2). Patients selected for this analysis were infected with either genotype B or C isolates that had been treated with TDF alone or TDF/FTC for 208 weeks ( Figure 2, Supplementary Table 1). All patients were positive for HBeAg and HBsAg at week 208. The nine patients all achieved viral suppression (≤29 IU/mL) at a median of 72 weeks and were classi ed as having DNQ levels of HBV DNA. Mice (n=14) were inoculated with serum from week 48 and week 96 time points on suppression from 7 different patients. Mouse sera were analyzed biweekly for HBV DNA by quantitative PCR and HBsAg/HBeAg by ELISA. While select mice inoculated with sera from patients near the LLOQ became viremic, none of the mice inoculated with sera below the LLOQ became viremic after 168 days ( Figure 6).
Given that each 100 uL inoculation of sera from a patient whose DNA level is below LLOQ (29 IU/mL or 165 copies/mL) may contain less than or equal to 16.5 copies and that every inoculation is subject to the jackpot effect, a separate group of uPA/SCID mice were tested with serial inoculations. Mice were inoculated ve times over a two-week period with sera from patients taken 12-32 weeks on NA treatment with HBV DNA ≤29 IU/mL (Supplementary Table 1). In contrast to the mice receiving single inoculations, mice that received multiple inoculations of sera from patients on NA therapy with DNQ HBV DNA resulted in viremia for 9/31 (29%) samples at a median of 66 days post-infection ( Figure 7A -C).
NAs have been shown to be safe and effective in improving in ammation and brosis, preventing liver failure, reducing the rate of HCC, and improving overall quality of life for patients with HBV. Even after years of NA treatment, the majority of CHB patients still have detectable but not quanti able HBV DNA and cessation of NA therapy often leads to rapid viral rebound, suggesting the existence of a viral reservoir 4 . Here, we show that sera from patients on long-term NA therapy are capable of infecting naïve hepatocytes in the uPA/SCID mouse model. At the time of this study, only tissue-culture derived laboratory strains or serum from chimpanzees had been used to infect uPA/SCID mice [17][18][19] . Serum from multiple patients and various timepoints below the LLOQ was able to give rise to viremia in mice. In some cases, serum from patients suppressed on therapy for as long 48 weeks (median 34 weeks, range 24 -48 weeks) was able to cause viremia in mice, indicating that those patients on therapy still contained infectious virus. This is the rst demonstration that NA therapy does not fully suppress viral replication and the residual viral DNA likely represents infectious virus. Consistent with this data, patients who are virally suppressed on NA therapy maintain high-to-detectable levels of serum HBsAg and HBV RNA, and low-to-detectable levels of intrahepatic cccDNA, suggesting that intrahepatic cccDNA is still transcriptionally active. The low-level viremia in NA-treated patients may serve as a viral reservoir leading to the de novo formation of cccDNA and maintenance of chronic HBV infection in the liver. The continued low-level viral replication would explain the viral rebound that is observed upon discontinuation of NA therapy. Moreover, the residual viremia may also explain the continued risk, although low, for the development of long-term HBV-related complications, such as HCC, in virally suppressed patients. Studies have demonstrated that patients on NA therapy that achieve functional cure, de ned as HBsAg loss, have further reduction in HCC risk compared to those that remain virally suppressed 20,21 . However, HBsAg loss is uncommon in NA-treated patients 22,23 . These results highlight the need for orthogonal approaches (i.e., capsid modulators, RNAi, or immune modulation) in addition to NA therapy to achieve HBV functional cure. Buti, M. et al. Long-term clinical outcomes in cirrhotic chronic hepatitis B patients treated with tenofovir disoproxil fumarate for up to 5 years. Hepatology international 9, 243-250, doi:10.1007/s12072-015-9614-4 (2015 Treatment TDF  FTC-TDF  TDF  FTC-TDF  TDF  TDF  FTC-TDF  FTC-TDF  FTC-TDF   Week 4  104000  269000  727000  573000  145000  408000  13800000  617000  185000   Week 8  12600  15400  279000  80100  45300  120000  2690000  169000  47700   Week 16  2440  447  55800  6750  8950  29700  570000  16300  8120 Week 24

Patient Serum Selection
Nine patients who were infected with HBV genotype B and C viruses were selected from a phase 3 clinical study (GS-US-203-0101) that evaluated TDF alone or in combination with FTC for the treatment of CHB 1-4 . None of the nine patients achieved HBsAg seroconversion. Samples with HBV DNA levels below 169 copies/mL (determined by the COBAS TaqMan assay) were de ned as HBV DNA target detected (TD). Of the nine patients selected, only one patient did not become suppressed on treatment. All patient samples were sent to PhoenixBio by Covance biological repository. Unless otherwise speci ed, inocula were thawed once, aliquoted once to limit freeze thaw cycles, and maintained at -80°C. Patient serum was kept frozen until the day of the study and thawed at 37°C. All mice received intravenous injections (IV) of inocula via the tail vein using disposable 1.0 mL syringes with permanently attached needles. For dilution studies, patient serum was diluted in serum of cDNA-uPA wild/+ /SCID mice that had not been transplanted with human hepatocytes to achieve the desired target concentration. The desired target concentration was validated by quantitative real-time PCR with at least 5 technical replicates.

Study oversight
All patients provided informed consent. The study was approved by the Institutional Review Board at both participating sites and was conducted in compliance with the Declaration of Helsinki, Good Clinical Practice guidelines, and local regulatory requirements.
The study was designed and conducted by the sponsor in collaboration with the principal investigators. The sponsor collected the data and monitored the study conduct. The investigators, participating institutions, and sponsor agreed to maintain con dentiality of the data. All authors had access to the study data and reviewed and approved the nal manuscript.
Frozen human hepatocytes (donor BD195, Corning Incorporated, Tewksbury, MA, USA) were thawed and transplanted into 2-to 4- week-old uPA/SCID mice by splenic injection. Mice were selected for studies if they contained hepatocytes with an estimated replacement index greater than 70% based on the blood concentration of human albumin (>8.5mg/mL) at one week prior to study onset. General health observations including weight were monitored weekly. Within the different groups, serum HBV DNA sampling frequency ranged between 30 minutes and 9 days with sampling in the different groups beginning as soon as 1 minute postinfection and continuing through 63 days after inoculation. All animal protocols described in this study were performed in accord with the Guide for the Care and Use of Laboratory Animals and approved by the Animal Welfare Committee of Phoenix Bio Co., Ltd.
All mice were housed individually and maintained in accordance with the Animal Ethics Committee of PhoenixBio (resolution #1856, 1916, 2104, 1943, 2102, and 2103). In total, all studies described herein included 94 uPA/SCID mice with humanized livers divided into 4 experimental groups.
Blood was collected from all animals under iso urane (ISOFLURANE Inhalation Solution [P zer], Mylan, Osaka, Japan) anesthesia via the retro-orbital plexus/sinus using Intramedic™ Polyethylene Tubing (Becton Dickinson and Compound, NJ, USA) at each time point. A maximum of 40 µL was obtained at each timepoint. For blood collection at terminal sacri ce, a minimum of 400 µL was collected by cardiac puncture and exsanguination. Of the blood collected, 2 uL was reserved for human albumin measurement.
Remaining blood samples were left at room temperature for at least 5 minutes to coagulate and then centrifuged at 13200 x g at 4°C for 3 minutes to obtain serum.

Quanti cation of Human Serum Albumin
Blood obtained was diluted in saline and the clinical chemistry analyzer (BioMajesty™ Series JCA-BM6050, JEOL Ltd., Tokyo, Japan) was used to measure the albumin concentration using latex agglutination immunonephelometry test (LX Reagent "Eiken" Alb II; Eiken Chemical Co., Ltd., Tokyo, Japan). Levels of human serum albumin were measured as soon as 1 minute after inoculation and then at weekly intervals.
Extraction and Quanti cation of Serum HBV DNA DNA was extracted from 10 µL (5 uL x 2) of serum by the Smitest Ex-R&D Nucleic Acid Extraction Kit (Medical & Biological Laboratories Co, Ltd, Nagoya, Japan) and was dissolved in 20 µL of nuclease-free water (Life Technologies Japan Ltd., Tokyo, Japan). HBV DNA copy numbers were determined by quantitative real-time PCR (qRT-PCR) as reported previously. HBV DNA extracted from the serum of an HBV-infected uPA/SCID mouse was used as the DNA standard. The copy numbers in the standard were determined in reference to a synthetic HBV DNA template. The range of standards used was 4 x 10 4 to 2 x 10 9 copies/mL.
The lower limit of quanti cation is 4 x 10 4 copies/mL. The lower limit of detection is 8 x 10 2 copies/mL. Real-time quantitative PCR was performed using the TaqMan Fast Advanced Master Mix (applied Biosystems, Thermo Fisher Scienti c Inc.) and ABI Prism 7500 sequence detector system (Applied Biosystems) using a forward probe (HBV target location 166-186: 5' -