Multiplex gene editing suppresses random integration of hepatitis B virus DNA in chronically infected liver

Abstract

Gene editing technologies have opened the possibility of directly targeting viral DNA for therapeutic applications. In chronically infected hepatocytes with hepatitis B virus (HBV), covalently closed circular DNA (cccDNA) serves as the master template for viral transcripts and gene products. In the present study, we evaluated anti-HBV multiplex gene editing with the CRISPR-Cas9 endonuclease from Staphylococcus aureus (SaCas9) using primary human hepatocytes (PHHs) and HBV mouse models. Nonviral delivery of SaCas9-encoding mRNA and a pair of HBV-targeting guide RNAs (gRNAs) substantially reduced viral biomarkers and intrahepatic HBV DNA copies in vitro and in vivo. Hybridization capture sequencing analyses showed that small insertions and deletions (indels) and structural variants including excisions and inversions of the viral sequences were introduced in the residual HBV DNA. These assays also demonstrated that transient expression of the HBV-targeting SaCas9 significantly suppressed random integration of HBV DNA into the host genome, while leading to no detectable increase in chromosomal translocations involving viral copies. Lastly, our gene editing approach blocked viral rebound after stopping treatment with a nucleos(t)ide analogue (NA), entecavir. Our results suggest that anti-HBV multiplex gene editing removes viral DNA from chronically infected hepatocytes, potentially reducing the risk of hepatocarcinogenesis associated with HBV DNA integration.

Graphical abstract

The CRISPR-Cas9 multiplex gene editing system designed to cut two conserved sites of hepatitis B virus (HBV) DNA was evaluated in vitro and in vivo. This approach reduced secreted viral biomarkers, intracellular HBV DNA and its chromosomal integration without detectable effects on chromosomal translocations involving HBV sequences.

Introduction

HBV is a hepatotropic DNA virus that carries a 3.2-kb partially double-stranded, relaxed circular DNA (rcDNA) genome in the capsid proteins surrounded by a lipid envelope layer embedded with three isoforms of the HBV surface antigen (HBsAg).¹ HBV enters liver hepatocytes by binding to the sodium taurocholate cotransporting polypeptide (NTCP) via the unique, extended N-terminal peptide carried by the largest HBsAg isoform. The rcDNA, enclosed within capsid proteins, is transported into the nucleus via the cytoplasmic trafficking machinery, where host DNA replication and repair proteins convert it into cccDNA. This form of HBV DNA persists in the nucleus and serves as a template for 5 types of mRNA encoding a total of 7 viral gene products. Among these viral transcripts, pregenomic RNA (pgRNA) is used as the template of reverse transcription by the viral polymerase during encapsidation.

HBV infection can lead to chronic hepatitis B (CHB), which affects an estimated 250 million people worldwide.^2,3 Its persistence significantly increases the risk of developing serious liver diseases, including cirrhosis, liver failure, and hepatocellular carcinoma (HCC). HBV infection is the leading cause of HCC, accounting for more than 50% of cases globally. HBV DNA integration is one of the mechanisms that potentially contribute to hepatocarcinogenesis. Integrated HBV DNA (intDNA) copies have been found near HCC-associated genes, including TERT and CCNE1, which encode telomerase reverse transcriptase and cyclin E1, respectively.^4,5 Viral DNA can be inserted into the host genome early during infection, with the frequency of this event estimated to range from 0.01% to 0.1% under previously tested experimental conditions.⁶ The number of integrated HBV DNA copies has also been shown to increase over the course of chronic infection.⁷ Double-stranded linear DNA (dslDNA), a by-product of viral replication, serves as the preferred substrate for integration within the nucleus.^8,9 NAs, the current standard of care for CHB, inhibit viral reverse transcription and reduce viral load by several orders of magnitude in patients with CHB.^10,11 While this oral medication is well tolerated and significantly slows the progression of liver disease, it achieves a functional cure in only a small proportion of patients. A functional cure is defined as sustained loss of HBsAg and HBV DNA levels below the lower limit of quantitation (LLOQ) for 24 weeks after treatment cessation,^12,13 and it has been established as the preferred primary endpoint in clinical trials evaluating new therapies for CHB.

We previously aimed to eliminate human immunodeficiency virus (HIV) and herpes simplex virus type-1 (HSV-1) using multiplex gene editing with CRISPR-Cas9, demonstrating that systemic delivery of AAV vectors carrying the SaCas9 gene and a pair of gRNAs effectively blocked viral rebound and reactivation in preclinical models.^14,15 We further examined the HIV-targeting EBT-101 therapy in the clinic and confirmed its safety and tolerability (NCT05144386). These results suggest that directly removing viral DNA with CRISPR-Cas9 is a feasible strategy for treating chronic infections. A similar therapeutic approach potentially offers a curative option for other infectious diseases, including CHB. Systemic delivery of lipid nanoparticles (LNPs) encapsulating gene-editing payloads has been shown to achieve targeted gene disruption in liver hepatocytes.^16,17 Earlier studies have demonstrated that CRISPR-Cas9 can access intracellular HBV DNA to suppress viral transcription and protein expression,^{18,19,20,21,22,23} and two gene editing-based therapies for CHB have recently advanced to clinical trials (NCT06680232; NCT06671093). While anti-HBV gene-editing therapy holds promise for improving the lives of patients with CHB, potential safety concerns include an increased risk of HBV integration following the linearization of episomal HBV DNA, as well as unintended structural variants such as chromosomal translocations resulting from cleavage of intDNA. To investigate to what extent nuclease-mediated HBV DNA editing alters chromosomal DNA, we first selected a pair of gRNAs designed to target well-conserved sites of the viral genome and examined the anti-viral activity of SaCas9 complexed with each of the paired gRNAs in HBV-infected cells and in two CHB mouse models. We analyzed DNA edits introduced by our therapeutic method in PHHs and mouse tissue samples by deep sequencing. In addition, we investigated the efficacy of our gene-editing therapy in combination with an NA. These results provided valuable insights into not only the molecular mechanisms of anti-HBV multiplex gene editing but also its impact on chromosomal integrity.

Results

Identifying a pair of guide RNAs that direct CRISPR-Cas9 to conserved sites of the HBV genome

HBV genomic sequences exhibit high diversity due to the error-prone nature of reverse transcription, as the viral polymerase lacks proof-reading activity. Viral sequences that differ by more than 7.5% are classified as distinct genotypes, which display geographically varied distributions.²⁴ Through in silico analysis of viral sequences curated from HBVdb,²⁵ we identified a pair of sites suitable for targeting with SaCas9 (see “materials and methods” for details). One target site (SaCas9 gRNA 1 target site) is located within the region encoding the viral polymerase and major surface antigen, while the other (SaCas9 gRNA 2 target site) lies within the core antigen gene. Both sites were highly conserved, with no nucleotide substitutions or gaps in approximately 90% or more of viral sequences assigned to each of the eight genotypes (A–H) in the database (Figure S1A). We next searched for homologous sequences in the human reference genome and found only a few sites that differed from these viral target sequences at 3 or fewer nucleotide positions within the 21-nt protospacer (Table S1).

To assess the on-target editing activity of SaCas9 with the selected pair of gRNAs, we generated an HBV reporter cell line with a chromosomally integrated DNA sequence containing the two HBV target sites, spaced approximately 400 base pairs apart (Figure S1B). Transfection with SaCas9-encoding mRNA and the paired gRNAs resulted in excision of the intervening region between the two target sites in approximately 60% of the chromosomally integrated reporter constructs in the cell line (Figure S1C). We also sequenced the unexcised target sites and found that indels were present in 45% and 30% of the PCR-amplified regions targeted by SaCas9 gRNA 1 and SaCas9 gRNA 2, respectively (Figure S1D). Since these target sites are located in the open reading frames of viral genes, both excisions and indels are predicted to disrupt viral protein expression.

To investigate the specificity of gene editing with the selected gRNAs, we first nominated potential off-target sites in the human genome using GUIDE-seq.^26,27 The HBV reporter cell line was transfected by electroporation with SaCas9-encoding mRNA, either of SaCas9 gRNA 1 or SaCas9 gRNA 2, and a double-stranded oligonucleotide (dsODN). We then used next-generation sequencing to identify dsODN insertion sites across the human genome. The majority of sequence reads containing the dsODN were mapped to the HBV on-target sites within the chromosomally integrated construct, while only a small number were detected in human genomic sites with relatively high sequence similarity to the intended target sites (Figures S2A and S2B). A striking difference between read counts assigned to the on-target sites for the two gRNAs was attributed to varied efficiencies of dsODN insertion (data not shown), as reported previously.²⁶ To assess whether SaCas9 complexed with each of the paired gRNAs edited the sites nominated by GUIDE-seq, we quantified indels at 3–4 nominated sites using targeted amplicon sequencing in the same reporter cell line transfected with SaCas9-encoding mRNA and one of the gRNAs. These sites were selected based on read counts from the GUIDE-seq assays and the number and position of base pair mismatches and gaps, which have been shown to influence SaCas9 endonuclease activity.²⁸ Indels were detected at most of the integrated on-target sites of the reporter construct in the edited cells, compared to less than 0.1% in the control samples, confirming robust editing at the intended target sites by our anti-HBV gene-editing nuclease under the conditions tested (Figures S2C and S2D). In contrast, very little change in the frequency of indels was observed at any of the nominated sites between the treated and control samples. These results suggest that the selected gRNAs specifically direct SaCas9 to the viral target sequences.

Multiplex gene editing-mediated suppression of HBV replication in HBV-infected primary human hepatocytes

To examine whether our anti-HBV editing approach inhibited viral replication in primary human hepatocytes (PHHs), we transfected HBV-infected PHHs with SaCas9-encoding mRNA and the pair of HBV-targeting gRNAs at 4 days post-infection. Viral antigens secreted into the culture medium and intrahepatic HBV DNA were then analyzed at 8 and 11 days post-infection (day 9 and day 12) as shown in Figure 1A. Chemiluminescent immunoassays showed that concentrations of HBsAg and HBeAg in the culture media of the edited samples harvested on day 9 were reduced by 79% and 75%, respectively, compared to those transfected with a control gRNA (Figures 1B and 1C). We observed further reduction of HBsAg and HBeAg from the baseline in edited samples on day 12. In addition, intrahepatic HBV DNA copies were reduced by 58% and 71% in cells transfected with the HBV-targeting gRNAs on day 9 and day 12, respectively (Figure 1D). To assess the cell viability, we quantified extracellular human albumin present in the harvested media and verified that its secretion was minimally affected by expression of the anti-HBV SaCas9 (Figure 1E). These results suggest that our gene editing treatment promotes degradation of viral DNA and suppression of viral gene expression in hepatocytes.

Figure 1.

Anti-HBV multiple gene editing in HBV-infected PHHs

(A) A schematic of the experimental procedures. Primary hepatocytes were infected with HBV one day after cell seeding (day 1), and transfection with SaCas9-encoding mRNA and paired gRNAs was performed on day 5. The culture media were exchanged on days 6 and 9 (B, C) concentrations of HBsAg (B) and HBeAg (C) were quantified by chemiluminescent immunoassays. The control cultures were transfected with the non-targeting (NT) gRNA instead of the pair of HBV-targeting gRNAs (gRNA pair). (D) Intrahepatic copies of total HBV DNA were quantified by digital PCR with DNA samples extracted from cells harvested on day 9 or day 12. (E) Human albumin levels were measured by ELISA using the same set of culture media for the HBsAg and HBeAg assays. Data shown are mean values for three biological replicates and error bars show standard errors of the means.

To characterize sequence alterations in the residual HBV DNA recovered from the edited cells, we performed hybridization capture sequencing using HBV-specific capture probes and processed the raw sequencing data through our customized bioinformatic pipeline (see “materials and methods” for detail). Sequencing depth was relatively uniform across the HBV genome, and its coverage relative to the internal control, human GAPDH, was reduced in the edited samples (Figure S3). This is in line with reduced viral copies by anti-HBV gene editing shown in Figure 1D, suggesting that the HBV sequences were recovered at a similar efficiency in all the sequencing libraries. Among the viral DNA that escaped clearance by degradation, we detected indels at 1.8% and 1.1% of the HBV on-target sites for SaCas9 gRNA 1 and SaCas9 gRNA 2, respectively (Figure 2A). Additionally, we identified structural variants consistent with excision and inversion of the viral DNA between the two target sites (Figures 2B and 2C). These results indicate that our HBV-targeting endonuclease can generate double-strand breaks at the intended HBV DNA target sites that persist in human hepatocytes. Cleaved DNA products may be primarily degraded before they can be repaired by endogenous DNA end-joining mechanisms.

Figure 2.

Sequence analysis of HBV DNA copies recovered from HBV-infected PHHs by short-read hybridization capture sequencing

(A) Frequency of indels were calculated by dividing sequence reads containing indels at each on-target site by those carrying its flanking regions. (B, C) Sequence reads representing excisions (B) or inversions (C) of sequences between the two intended target sites were identified, and the percentage of such editing events were calculated using the number of total reads containing sequences adjacent to either of the intended cut sites. (D) Junctions between HBV DNA and human chromosomes were mapped on the HBV reference genome (GenBank ID: U95551). The red arrows indicate the on-target sites. The blue arrows show one of the two direct repeat, DR1 (the nucleotide positions 1826–1836), which is proximal to the termini of dslDNA. (E) The number of chimeric sequence reads containing HBV- and human chromosome-derived sequences was normalized with the average sequence depth of the human GAPDH that served as an internal control. Data shown are mean values for three biological replicates and error bars show standard errors of the means.

HBV is an oncogenic virus that potentially induces chromosomal abnormalities through random DNA integration.²⁹ A small number of sequence reads derived from these genetic events were found by hybridization capture sequencing in both control and edited samples collected 11 days post-HBV infection. As described previously,^5,8,30 chromosomal insertion site breakpoints were distributed throughout the HBV genome, with increased read counts around direct repeat 1 (DR1), located upstream of the viral core antigen gene (Figure 2D). This region is proximal to the predicted termini of dslDNA,³¹ which has been thought to serve as the primary source of randomly integrated HBV DNA. The relative number of HBV-chromosome junctions, normalized to the average sequence coverage of the internal control gene (human GAPDH), was lowered by treatment with the pair of HBV-targeting gRNAs by 60% (Figure 2E). This suggests that fragmentation of the HBV genome by our gene editing treatment does not lead to an overall increase in HBV DNA integration into the host genome. Targeted cleavage of HBV DNA might facilitate degradation of episomal viral DNAs, which in turn decreased the frequency of random HBV DNA integration. In the edited samples, we observed modest spikes of the HBV-chromosome junctions near the two intended target sites (see “discussion” for details). HBV insertion sites were found across all human chromosomes and distributed similarly in both the control and treated samples (Figure S4). Many of these sites were identified in single consensus sequence reads, suggesting that they originate from integrated copies carried by single cells. This may reflect the limited expansion of PHHs harboring integrated HBV DNA under the experimental conditions, resulting in a heterogeneous pattern of integration.

While we did not detect off-target editing at several nominated sites in our reporter cell line (Figure S2), chromatin states might be different in other cell types and alter the gene editing profile. If the HBV-targeting SaCas9 generates double-strand breaks at unintended target sites in human chromosomes, HBV DNA fragments may be captured by these DNA ends. We searched homologous sites to the HBV on-target sites in the vicinity of HBV-chromosome junctions detected by the hybridization capture sequencing. We identified three sites that were also nominated by GUIDE-seq (Table S2). Each site was found only in a single consensus read in one of the biological triplicates from the edited samples. We next performed targeted amplicon sequencing to further investigate editing of these sites. A slight increase in the frequency of indels was observed at the two sites but was not statistically significant over the background (Figure S5). It should be noted that liposomes were used to deliver the SaCas9-encoding mRNA and a pair of the gRNAs into HBV-infected PHHs, while LNP-based delivery is the gold standard for in vivo gene editing. This discrepancy may slightly influence the pharmacokinetics and pharmacodynamics of our anti-HBV gene editing reagent.

In vivo editing of HBV DNA in the AAV-HBV mouse model

We next studied anti-HBV activity of our editing approach in AAV-HBV mice, which were generated by transduction with a recombinant AAV8 vector carrying a 1.3-fold excess length of the genotype-D HBV genome.^32,33 This synthetic viral sequence supports transcription of viral mRNA and expression of viral proteins in mouse hepatocytes, allowing viral replication. Clinical biomarkers such as HBV DNA, HBsAg, and HBeAg can be detected in blood at stable levels, while re-entry of secreted virions into mouse hepatocytes is limited. To transiently introduce anti-HBV editing in mouse hepatocytes, we intravenously administered LNPs encapsulating SaCas9-encoding mRNA and the pair of HBV-targeting guide RNAs 5 weeks post-AAV-HBV administration. We observed an LNP dose-dependent reduction in serum HBsAg, HBeAg, and HBV DNA (Figures 3A–3C). To verify the molecular mechanisms underlying the viral suppression, we quantified intrahepatic HBV DNA. The assays showed that total HBV DNA significantly decreased in the livers of mice receiving the LNPs (Figure 3D), suggesting that LNP-mediated delivery of our HBV-targeting SaCas9 inhibited HBV replication by removal of viral DNA from hepatocytes.

Figure 3.

LNP-mediated delivery of the anti-HBV gene-editing therapy in the AAV-HBV mouse model

(A-C) Serum levels of HBsAg (A), HBeAg (B), and HBV DNA (C) were assayed weekly following intravenous administration of PBS (control) or LNPs encapsulating SaCas9-encoding mRNA and the pair of the HBV-targeting gRNAs at two dose levels (low dose, 1.4 mg/kg; high dose, 2.9 mg/kg). (D) Total HBV DNA copies were quantified by digital PCR in liver samples harvested from each mouse 5 weeks after LNP administration. Data shown are mean values for six biological replicates and error bars show standard errors of the means.

To analyze sequence alterations of viral DNA derived from the synthetic AAV construct and its chromosomal insertion sites in vivo, we performed hybridization capture sequencing on viral DNA extracted from three representative liver samples from each of the PBS control and high-dose groups, selected based on intrahepatic HBV DNA copy numbers (Figure S6A). The whole HBV genome was covered at 310,000 × and 210,000 × on average in the control and edited samples, respectively, and the sequence depth was greater in the redundant regions on the AAV-HBV vector (nucleotide positions 1070–1990) (Figure S6B). The read coverage patterns were slightly different between the viral DNA collected from the HBV-infected PHHs and AAV-HBV mice, suggesting that artificial forms of viral DNA derived from the recombinant AAV vector may be present in this mouse model. Consistent with Figure S6A, relative sequence coverage of HBV DNA to the internal control, mouse Gapdh, was reduced in livers treated with the LNPs (Figure S6C). Indels were detected in about 20% of each intended target site (Figure 4A). Excisions and inversions of sequences between the two intended target sites were also observed in 20% and 10% of total sequence reads spanning at least one of the two intended cut sites (Figures 4B and 4C).

Figure 4.

Analysis of sequence alterations in HBV DNA in the AAV-HBV mouse model by short-read hybridization capture sequencing

(A) Frequency of indels at the intended target sites on HBV DNA extracted from the control and edited liver samples of the AAV-HBV. (B, C) Frequency of excisions (B) and inversions (C) of the intervening HBV sequence between the two target sites for SaCas9 gRNA 1 and SaCas9 gRNA 2. The percentages of indels and structural variants were calculated as described in the legend of Figure 2D. Breakpoints of HBV and mouse chromosomal sequences were mapped on the HBV reference genome (GenBank ID: U95551). The red arrows indicate the on-target sites. (E) The number of chimeric sequence reads containing HBV-chromosome junctions was normalized with the average sequence coverage of the mouse Gapdh locus to estimate relative frequency of chromosomal HBV DNA insertion. Data shown are mean values for three biological replicates and error bars show standard errors of the means.

HBV-chromosome junctions of chimeric sequence reads were mapped throughout the viral genome (Figure 4D). Consistent with Figure 2D, two peaks were observed in proximity to the termini of dslDNA in the untreated samples.^6,8 Two additional spikes were found in the vicinity of the start and end positions of the 1.3 × oversized HBV genome (the positions 1070 and 1990) inserted between the AAV inverted terminal repeats (ITRs). These insertion events may be facilitated by the molecular machinery of recombinant AAV vector random integration.³⁴ As detected in PHHs edited with the HBV-targeting SaCas9, chimeric sequence reads containing regions flanking the HBV DNA on-target sites were slightly increased compared to surrounding areas of the HBV genome in samples treated with the LNPs (Figure 4D). In contrast, the relative number of total HBV insertion sites was reduced in edited samples when compared to the controls (Figure 4E). Viral DNA integration sites were detected across all mouse chromosomes (Figure S7), with most events supported by only a single consensus read, suggesting that most hepatocytes harboring chromosomally integrated HBV DNA did not undergo clonal expansion during the course of the experiment.

Nuclease-mediated editing of the chromosomally inserted DNA in the transgenic HBV mouse model

To assess DNA repair outcomes following double-strand breaks generated at HBV target sites inserted in chromosomal DNA in vivo, we intravenously administered our anti-HBV gene editing payload encapsulated in LNPs to a transgenic (Tg-HBV) mouse model. This model was generated by microinjecting a synthetic DNA sequence containing a 1.3-fold overlength of the genotype-A HBV genome. Viral transcripts from this transgenic construct are predicted to be the primary source of viral protein expression and replication. HBV DNA and HBsAg were secreted into the bloodstream at high levels in the hemizygous mice, similar to those observed in a previously described transgenic animal model.^35,36 A dose-dependent reduction in serum HBsAg and HBV DNA levels, as well as intrahepatic HBV DNA copies, was observed in mice treated with LNPs (Figures 5A–5C). These results suggest that LNP-mediated delivery of the HBV-targeting SaCas9 efficiently targeted both the chromosomally integrated HBV transgene construct and episomal viral DNAs (transported from capsids to the nuclei) in the Tg-HBV mice.

Figure 5.

In vivo editing by systemic LNP delivery in the Tg-HBV mouse model

(A, B) Dose-dependent reduction in serum HBsAg (A) and HBV DNA (B) was observed in the transgenic mice treated with two different quantities of LNPs encapsulating the SaCas9-encoding mRNA and the pair of the HBV-targeting gRNAs (low dose, 1.6 mg/kg; high dose, 3.4 mg/kg). (C) Total HBV DNA copies were quantified by digital PCR in liver samples harvested from each mouse 5 weeks after LNP administration. Data shown are mean values for six biological replicates and error bars show standard errors of the means.

Given that all the HBV promoters were inactive in the stomach of the Tg-HBV mouse model, digital PCR should only amplify DNA from the integrated Tg-HBV construct and not from replicating HBV. Digital PCR was performed using custom-designed primers and a probe targeting the non-redundant region of the 1.3 × HBV genomic construct. Analysis of DNA extracted from the stomach tissue of a PBS-treated (control) mouse suggested that each cell harbored approximately two integrated copies of the full-length viral genome (Figure S8A). To determine the configuration of the Tg-HBV construct and its chromosomal insertion site, we performed long-read hybridization capture sequencing with DNA extracted from stomach tissues of three individual mice. The hybridization capture sequencing indicated that the transgenic construct was inserted in the reverse orientation of the mouse chromosome 3 (Figure S8B). It was composed of 2.3 × direct repeats of the genotype-A HBV genome containing a total of five target sites for the paired HBV-targeting gRNAs in addition to the partial plasmid vector backbone used for DNA cloning and 240 base pairs of HBV-derived sequence, which was in the reverse orientation to the other 2.3 × HBV genomic sequence (HBV inverted repeat [HBV-IR]). Neither of the target sites for the paired gRNAs were present in HBV-IR.

To examine structural variants introduced by the HBV-targeting SaCas9, we analyzed HBV DNA extracted from liver samples of five mice per group by long-read hybridization capture sequencing again. Excisions and inversions of the intervening sequences between the 5 on-target sites were detected in the treated mice (Figure 6A). HBV pgRNA transcribed from the 2.3 × overlength HBV DNA is predicted to produce rcDNA and dslDNA through reverse transcription within the viral capsid. These viral DNA molecules can be shuttled into the nuclei via intracellular recycling and randomly integrated into chromosomal DNA in the liver of this mouse model. As expected, HBV-chromosome junctions were found at distinct chromosomal locations from that of the Tg-HBV construct. Chimeric reads representing de novo HBV DNA insertion sites were reduced by up to 80% in the LNP-treated groups, as previously observed in the HBV-infected PHHs and AAV-HBV mouse model (Figure 6B). This reduction may be primarily attributed to suppression of viral replication, leading to lower levels of episomal viral DNA and consequently fewer random HBV DNA integration events. HBV-associated chromosomal translocations have been observed in liver biopsies from CHB patients.^7,37 After excluding chromosome 3 junctions associated with the Tg-HBV construct, we identified an extremely small number of sequence reads containing two distinct chromosomal regions interrupted by HBV DNA in both the control and treated samples. The relative frequency of chromosomal translocations involving such randomly integrated HBV DNA was not substantially altered by anti-HBV multiplex gene editing (Figure 6C). While the Tg-HBV mouse model was specifically chosen to examine editing of intDNA, we ultimately decided to exclude sequence reads containing the original Tg-HBV construct from this analysis. The structure of the Tg-HBV construct was found to be substantially different from that of integrated HBV DNA copies isolated from HBV-infected hepatocytes,^7,37 and it is therefore difficult to conclude primary factors that contributed to chromosomal rearrangements with this synthetic DNA sequence. Taken together, these results suggest that transiently induced double-strand breaks on intrahepatic HBV DNA do not increase the risk of chromosomal aberrations caused by translocations in a relatively short period.

Figure 6.

Analysis of structural variants, random HBV DNA insertions and chromosomal translocations by long-read hybridization capture sequencing

(A) The Tg-HBV mice were found to carry the 2.3 × tandem repeats of the genotype-A HBV genome in its chromosome 3. Excisions and inversions of the chromosomally integrated HBV genomic sequences between the 5 on-target sites were quantified in liver samples from mice of the control and LNP-treated groups. (B,C) Chimeric sequence reads containing de novo HBV DNA insertion sites that were distinct from the Tg-HBV construct (B) and chromosomal translocations associated with these randomly integrated HBV sequences (C) were counted and normalized with the average sequence coverage of the internal control gene locus, mouse Gapdh. Data shown are mean values for five biological replicates and error bars show standard errors of the means.

Anti-HBV gene editing in a combination with NA therapy

NAs have been well tolerated and have served as the standard of care for CHB for decades.^10,11 To evaluate the efficacy of our anti-HBV gene-editing product in combination with an NA, AAV-HBV mice were treated with entecavir (ETV) via drinking water for 2 weeks (days 0–14) and were dosed with LNPs encapsulating SaCas9-encoding mRNA and the pair of the HBV-targeting gRNAs on days 0 and 14. No clinical signs of distress, pain, and overall health were observed following LNP administration. As expected, ETV treatment produced little change in HBsAg levels, whereas HBsAg reduction increased in a dose-dependent manner with LNP administration (Figures 7A and 7B). Up to 3 weeks after the start of the treatment (day 21), serum HBV DNA levels continued to decrease in all ETV-treated groups, relative to those receiving LNPs alone (Figures 7C and 7D). The intracellular concentration of ETV seemed to be high enough to block reverse transcription for up to 7 days after 2 weeks of ETV treatment. HBV DNA levels then rebounded in the ETV-only group but remained suppressed in mice receiving both ETV and LNPs. These results suggest that the combination therapy may control viral load more effectively than either monotherapy.

Figure 7.

Antiviral effects of HBV-targeting gene-editing in combination with entecavir in the AAV-HBV mouse model

LNPs encapsulating SaCas9-encoding mRNA and the pair of HBV-targeting gRNAs were administered twice over a 2-week interval (low dose, 1.0 mg/kg; high dose, 2.0 mg/kg), as indicated in the graphs. Half of the mice received entecavir in drinking water for 2 weeks. (A, B) Serum HBsAg levels were measured in mice treated with LNPs alone (A) and in those receiving both LNPs and entecavir (B). (C, D) Serum HBV DNA levels were monitored in mice treated with LNPs alone (C) and in those receiving the combination therapy (D). Data shown are mean values for six biological replicates and error bars show standard errors of the means.

Discussion

NAs inhibit reverse transcription during encapsidation, significantly reducing viral load. However, low-level viral replication can still be detected in patients undergoing long-term NA therapy, leading to the persistence of cccDNA and a slow decline of serum HBsAg levels.³⁸ Gene editing is considered a promising approach to achieving functional cure of CHB, since it potentially targets all forms of HBV DNA including cccDNA. While cccDNA is assembled into a minichromosome with histone and non-histone proteins,^39,40,41 previous studies have shown that this nucleosome-like structure is unlikely to block access of CRISPR-Cas9 proteins to cccDNA and other forms of HBV DNA.^{19,20,21,22,23} In this study, we demonstrated that co-delivery of SaCas9-encoding mRNA and a selected pair of HBV-targeting gRNAs reduced serum levels of HBsAg and HBV DNA, as well as intrahepatic HBV DNA copies, both in PHHs and in HBV mouse models. Viral suppression was evaluated using identical SaCas9-encoding mRNA and a pair of HBV-targeting gRNAs encapsulated by slightly different LNP formulations, which may result in slightly different pharmacokinetics, pharmacodynamics, and distribution of the anti-HBV gene editing payload in vivo. This discrepancy does not seem to significantly affect evaluation of the anti-viral efficacy in hepatocytes, as we observed a similar degree of reduction in serum biomarkers. No cccDNA or only very low levels have been detected in the AAV-HBV or Tg-HBV mouse models,^32,33,35 and viral transcripts and proteins are primarily derived from the episomal AAV construct or the chromosomally integrated HBV transgene, respectively. Recombinant AAV vectors have been thought to exist primarily as episomal forms of monomeric and concatemeric circular DNAs in a chromatin-like structure.⁴² The AAV-HBV vector that is stably maintained in mouse hepatocytes may be chromatinized in a similar fashion to HBV cccDNA, although a slightly different set of host proteins may be involved in its epigenetic regulation. Our anti-HBV therapeutic approach, which transiently cleaves the viral target sequences, resulted in stable suppression of serum HBV biomarkers under the conditions tested. While direct targeting of cccDNA is critical to achieve long-term viral control without indefinite treatment, its demonstration is not straightforward as no standard assay protocol is available.⁴³ Additionally, other nuclear-localized HBV DNA species can serve as the source of cccDNA, and removal of them may also lead to a reduction of cccDNA within the period that allows for maintenance of primary hepatocytes in a culture. A liver-humanized, immunocompromised mouse serves as an excellent model for testing treatments for CHB, as it recapitulates the full course of HBV infection for extended time compared to PHHs. We previously attempted to deliver mRNA to human hepatocytes of a chimeric mouse liver using conventional LNPs, which adsorb serum apolipoprotein E to bind the low-density lipoprotein receptor and enter cells via endocytosis. However, we found that LNP uptake by human hepatocytes was inefficient in these chimeric liver animal models.⁴⁴ As a result, we opted to use the two alternative mouse models to evaluate our anti-HBV multiplex gene editing approach in vivo, despite their inability to fully recapitulate iterative HBV infection. Liver-targeting LNPs have been known to deliver a gene editing payload into a few other organs such as spleen and adrenal glands with greatly reduced efficiency.¹⁷ Since LNP distribution potentially affects both anti-HBV efficacy in the primary target tissue and safety risk in off-target cell types, selection of LNP formulations with enhanced targeting efficiency and specificity would further improve the risk-benefit balance of our anti-HBV treatment.

To gain molecular insights into the outcomes of our HBV-targeting multiplex gene editing approach, we performed target-enriched sequencing of viral and host genomic DNA from HBV-infected PHHs and two HBV mouse models using hybridization capture. Detection of indels and structural variants (excisions and inversions) at the HBV on-target sites indicated that SaCas9 generated double-strand breaks on both episomal and integrated forms of the viral DNA. These DNA edits were detected at a higher frequency in the AAV-HBV mouse model compared to HBV-infected PHHs. This may be attributed, in part, to differences in the abundance and composition of intrahepatic HBV DNA species between the two models, which could influence overall DNA editing efficiency. As the majority of intrahepatic HBV DNAs were episomal, excision, and inversion events detected in our short-read hybridization capture sequencing assays were likely to be the evidence of recircularization of fragmented HBV DNAs by the host DNA repair machinery. An earlier study has demonstrated that Cas9-induced strand breaks at two HBV DNA sites can give rise to compact cccDNA-like molecules by rejoining free DNA ends at its two on-target sites.²¹ Such DNA copies have been shown to serve as a template for viral transcription, while pathological significance of viral variant proteins expressed from the shuffled DNA sequences remains to be determined. A dose-dependent increase in frequency of intended structural variants was observed in the Tg-HBV mouse model, suggesting that the HBV-targeting SaCas9 could access the chromosomally integrated HBV sequence in vivo. However, we found that this transgenic construct differed significantly from typical intDNA in both size and configuration, and the effects of nuclease-induced strand breaks on the integrated Tg-HBV sequence may need to be interpreted cautiously.

A previous study has shown that linearization of HBV DNA by ARCUS nucleases (which is an engineered variant of the homing endonuclease I-CreI) promotes its insertion in chromosomal sites, including potential off-target sites in HBV-infected PHHs.⁴⁵ The consequences of cutting HBV DNA seem to highly depend on the characteristics of the editing nucleases. The same research group has recently shown that the version of HBV-targeting ARCUS nucleases evaluated in preclinical development of PBGENE-HBV neither introduces off-target editing nor promotes HBV DNA integration significantly over the background.⁴⁶ Our analysis, based on a series of hybridization capture sequencing experiments, indicated that our anti-HBV treatment suppressed random HBV DNA integration in HBV-infected PHHs and CHB mouse models. We also found that the number of break points representing dslDNA insertions, which were mapped near DR1, was notably reduced. In contrast, we observed a slight increase in integration events forming HBV-chromosome junctions near the two intended target sites compared to adjacent regions of the HBV genome. The corresponding sequence reads can be derived either from integration of episomal HBV DNA following cleavage or from complete deletion of the HBV DNA sequences extending outward from an intended target site within pre-existing intDNA following cleavage. Since our sequencing analyses only indicated the consequences of DNA repair that occurred in the edited samples, it is difficult to determine the origins of HBV-chromosome junctions around the HBV on-target sites. Chromosomal insertion of viral DNA fragments generated by a gene-editing nuclease may lead to expression of truncated HBsAg variants, which have been suggested to promote hepatocarcinogenesis.^47,48,49 One of our intended target sites is situated within the first transmembrane helix (TH1) of the major (small) HbsAg.^50,51 While truncations in this region have not been associated with an increased risk of liver disease progression, we compared the relative frequencies of chimeric sequence reads containing breakpoints within the TH1 domain between control and edited samples and confirmed that our gene editing approach did not enhance integration of such partial HBsAg genes (Figure S9A). The second target site is located within the precore region. Although the consequences of N-terminal truncation in this region are not well characterized, our anti-HBV gene-editing approach appeared to reduce such genetic events (Figure S9B). A reduced number of HBV-chromosome junctions were found in the short-read sequencing data from PHHs compared to those from the AAV-HBV mouse model, likely due to a shorter period of viral replication.

Chromosomal translocations mediated by intDNA represent another safety concern. HBV-associated chromosomal translocations have been detected in approximately one-third of CHB patients, even in non-HCC liver biopsies, according to published studies.³⁷ By long-read hybridization capture sequencing, we did not detected a clear difference in frequency of chromosomal translocations involving randomly integrated HBV DNA between control and treated liver samples from the Tg-HBV mouse model. This suggests that the transient increase in strand breaks induced by our therapeutic approach is unlikely to disrupt host DNA repair pathways, which efficiently join DNA ends to prevent chromosomal abnormalities. However, it is important to note that only a few, if any, consensus sequence reads were assigned to this type of translocation in each replicate of sequencing libraries. This indicates that the assay may require further optimization to reliably detect such extremely rare genetic events.

Our study demonstrated that multiplex gene editing against intrahepatic HBV DNA greatly suppressed viral transcription, protein expression, and replication from baseline levels. To evaluate potential safety concerns associated with the cleavage of both episomal and integrated HBV DNA copies, we conducted hybridization capture sequencing analyses. These assays indicated that our anti-HBV SaCas9 did not increase the risk of off-target effects in preclinical settings. Gene editing is currently the only approach to directly remove viral DNA that is present in patients with CHB. In addition to two classes of FDA-approved treatments, NAs and interferon α, a number of new drugs that target different aspects of the viral life cycle and host immune responses are in development.⁵² Our gene editing treatment appeared not only to reduce serum HBsAg levels but also to prolong NA-mediated viral load suppression in the AAV-HBV mouse model. This suggested that reduction in replicative intermediates localized in the nuclei by NA treatment might increase targeting efficiency of our HBV-targeting SaCas9 to AAV-derived, cccDNA-mimic molecules which supported viral replication. While the ideal goal of our approach is to achieve a functional cure for CHB through a one-time monotherapy, repeated administration or combination with other therapeutic strategies is also feasible and may significantly enhance clinical outcomes.

Materials and Methods

Cell culture

The cell line 293FT was purchased from Thermo Fisher Scientific (Waltham, MA, USA), and grown in DMEM supplemented with GlutaMAX, sodium pyruvate and 10% fetal bovine serum. Primary human hepatocytes (PHHs) were purchased from BioIVT (Westbury, NY) and were derived from a male human liver (Cat #: M00995-P, Lot #: WWL). PHHs were thawed using cryopreserved hepatocyte recovery medium and plated in William’s E medium with primary hepatocyte thawing and plating supplements. PHHs were cultured in William’s E medium (Gibco) with primary hepatocyte maintenance supplements (Gibco) and 10% FBS (Gibco).

Preparation of genomic DNA

Genomic DNA was extracted from PHH cells using a Maxwell RSC Blood DNA Kit (Promega) with the following protocol: PHH cells were harvested, washed with PBS, then incubated at room temperature for 10 min in 300 μL PBS with 10 μL RNase A. Then, 300 μL lysis buffer with 30 μL proteinase K was added, and samples were incubated at room temperature for 10 min. Samples were then incubated at 80°C for 30 min prior to loading into the Maxwell RSC blood DNA cartridge and ran on the Maxwell RSC instrument. Genomic DNA was extracted from 293FT cells using a Maxwell RSC Cultured Cells DNA Kit (Promega) according to the manufacturer’s protocol.

Genomic DNA was extracted from mouse tissues using a MasterPure complete DNA and RNA Purification Kit (Biosearch Technologies) with the following protocol: ∼5 mg tissue samples were homogenized with a pestle in 300 μL of TE (10 mM Tris-HCl pH 8.0, 10 mM EDTA) in 1.5 mL eppendorf tubes, then 300 μL 2× tissue and cell lysis solution and 2 μL of proteinase K was added and samples were incubated at 56°C for 1 h. Then, 2 μL RNaseA was added, and samples were incubated at 37°C for 30 min. Subsequent steps were performed according to the manufacturers’ protocol.

Production of the SaCas9-encoding mRNA and HBV-targeting gRNAs

The open reading frame of the SaCas9-encoding sequence optimized for human codon usage with a reduced number of thymidine bases was synthesized and inserted between the human β-globin 5′ UTR and 3′ UTR connected to a polyadenine tail flanked by BspQI target sites of a standard cloning vector. In vitro transcription by the T7 RNA polymerase was performed using the BspQI-linearized plasmid and N1-methylpseudouridine-5′-triphosphate instead of uridine-5′-triphosphate, and the cap-1 structure was added with the vaccinia capping enzyme and 2′-O-methyltransferase.

All gRNAs used in this study were generated by chemical synthesis. Each of the HBV-targeting spacers was 21 nucleotides in length and connected to the following backbone sequence: GUUUUAGUACUCUGGAAACAGAAUCUACUAAAACAAGGCAAAAUGCCGUGUUUAUCUCGUCAACUUGUUGGCGAGAUUUU. Methyl groups were added to 2′ hydroxyl groups of three ribose moieties at each end, and phosphorothioate modification was introduced at the three terminal positions. The entire nucleotide sequences of these gRNAs are shown in a PCT patent application WO2025235552A1. The non-targeting (NT) gRNA carried the following 18-nt spacer upstream of the backbone sequence: GGAGACCCGAGGGUCUCU. The protospacer for this control gRNA and the protospacer adjacent motif (PAM) were not aligned well with any sites in the HBV and human genomes.

Generation of the HBV reporter cell line

The HBV reporter cell line, which was derived from 293FT (Thermo Fisher Scientific), was generated as described previously.¹⁵ Briefly, the synthetic sequence illustrated in Figure S1B was synthesized and cloned in the pUC57-Kan vector at GenScript. The enhanced green fluorescent protein (EGFP) gene fused to the puromycin-resistant gene via the self-cleaving peptide P2A-coding sequence was under the control of the murine stem cell virus promoter with a reduced number of CpG sites,⁵³ and transcription of the downstream sequence including the synthetic exons and the blue fluorescent protein (BFP) and blasticidin-S deaminase (BSD) genes was activated by the human EF1α core promoter. A synthetic exon containing multiple stop codons (exon 2) was inserted between the two HBV on-target sites and caused premature translation termination. Nuclease-induced double-strand breaks at the two intended target sites placed in intronic regions facilitated excision of exon 2, leading to expression of the downstream gene products, BFP and BSD. While excision of the intervening sequence between the two HBV-on-target sites could be examined using BFP expression as a readout, frequency of excision was estimated by multiplex digital PCR in this study. The entire sequence was flanked by the two AAVS1 CRISPR-Cas9 target to be linearized for targeted knock-in via non-homologous end joining with Streptococcus pyogenes Cas9 ribonucleoprotein complex (SpCas9 RNP). Five hundred thousand 293FT cells were transfected with 0.5 μg of the reporter construct and 20 pmol of SpCas9 RNP by electroporation using 4D-Nucleofector System (Lonza) in 20 μL of the SF Cell Line Nucleofector Solution under the CM-130 program. Cells with the chromosomally integrated reporter construct were selected in the presence of 2 μg/mL puromycin, and single clones were isolated by limiting dilution. Biallelic knock-in was confirmed by multiplex digital PCR.

Evaluation of intended editing in the HBV reporter cell line

To quantify excision and indels, 1.2 × 10⁵ HBV reporter cells were seeded per well in a 24-well plate and transfected with 0.5 μg each of SaCas9-encoding mRNA and the pair of HBV-targeting gRNAs using Lipofectamine MessengerMAX. The cells were harvested 5 days post-transfection and genomic DNA was extracted. To calculate frequency of excision, the intervening sequence between the two on-target sites and the downstream sequence encoding BFP were quantified by multiplex digital PCR. Indels at the target sites that were not removed by excision were assayed by Sanger sequencing of PCR-amplified fragments.

GUIDE-seq

GUIDE-seq was performed and analyzed as previously described, with minor modifications.^15,26 Briefly, the HBV reporter cell line (200,000 cells) was electroporated with 1 μg each of SaCas9-encoding mRNA and one of the two gRNAs, along with 15 pmol of double stranded oligonucleotide (dsODN), using the 4D-Nucleofector System (Lonza) and the SF Cell Line Nucleofector Kit in a 20 μL reaction volume under the CM-130 program. The cells were harvested 4 days post-transfection for DNA extraction followed by library preparation.

Targeted amplicon sequencing

To quantify indels at the nominated sites identified by GUIDE-seq, the HBV reporter cell line (200,000 cells) was transfected with 1 μg each of the SaCas9-encoding mRNA and either of the two gRNAs by electroporation using 4D-Nucleofector System (Lonza) in 20 μL of the SF Cell Line Nucleofector Solution under the CM-130 program. Sequences spanning the HBV on-target sites and nominated sites selected by GUIDE-seq were PCR-amplified (see Table S3 for primer sequences) using Q5 High-Fidelity 2× Master Mix (New England Biolabs) and sequenced using a 2 × 150-bp paired-end run on the MiSeq system (Illumina). PCR amplification of the three nominated sites and sequencing were performed in the same procedures to analyze unintended editing in the DNA samples extracted from PHHs 11 days post-HBV infection. Targeted amplicon sequencing analysis was performed using CRISPResso2⁵⁴ as described previously.¹⁵

Hybridization capture sequencing

Library preparation and target capture for HBV sequences were performed using SureSelect XT HS2 reagents (Agilent Technologies) according to the manufacturer’s protocol. Capture probes were designed to target 2 HBV strains (GenBank: U95551 and AY128092) and two regions of human GAPDH (GenBank: NC_000012.12, 6537589–6538000 and 6534291–6534579) that share high sequence identity with mouse Gapdh. The probe sequences are shown in Table S4. The captured libraries were pooled and sequenced using a 2 × 150-bp paired-end run on the NovaSeq X Plus platform (Illumina). Long-read hybridization capture sequencing was performed according to Agilent’s Target Capture Long-Read Sequencing Using the Agilent SureSelect XT HS2 Target Enrichment System application note with the following modifications. After captured-library amplification, 49 μL of each library was used as input for steps 4 through 7 of PacBio’s preparing whole genome libraries using the HiFi prep kit 96 protocol. The nuclease-treated libraries were then cleaned up an additional time using the SMRTbell cleanup beads and eluted in 26 μL. The purified libraries were then annealed with sequencing primers, bound with sequencing polymerase, then cleaned up using the SMRTbell cleanup beads and eluted in 50 μL. The libraries were then pooled and sequenced on the PacBio Revio system. The capture probes used were the same as described above.

HBV infection and transfection of PHHs

Stock vials of HBV genotype D, produced using the HepAD38 cell line derived from HepG2 cells, were purchased from ImQuest BioSciences. Primary human hepatocytes (PHHs) obtained from BioIVT were seeded at 3.5 × 10⁵ live cells per well in a 24-well plate. After 24 h, HBV was added at 500 genome equivalents per cell in the presence of 4% polyethylene glycol 8000 (PEG 8000) (Sigma-Aldrich) and 2% DMSO (Corning). After incubation at 37°C for 16 h, the cells were thoroughly washed to remove the residual virions and maintained in the presence of 2% DMSO. Four days after infection, cells were transfected with 0.5 μg of SaCas9-encoding mRNA and 0.25 μg each of the pair of HBV-targeting gRNAs using Lipofectamine MessengerMAX and washed on the following day. Four days and 7 days after transfection, supernatants were collected, and cells were harvested for gDNA extraction.

DNA quantification by digital PCR

DNA isolated from cells and mouse tissues was quantified using the QIAcuity Probe PCR Kit (Qiagen) on a QIAcuity digital PCR instrument (Qiagen), following the manufacturer’s instructions. Digital PCR reactions (40 μL) were performed using 1, 5, or 25 ng of isolated DNA as a template. For DNA isolated from 293FT cells, primer-probe sets specific for the synthetic Exon 2 region (forward primer: ATTCCGGATAAATCCATCTACT, probe: 5′6-FAM-TGCCTTTCT/ZEN/CTCCACAGGTGTCCA-3IABkFQ, reverse primer: GGTCTTACTGACATCCACTT) and BFP region (forward primer: CGTGCTGTTCCACGTATG, probe: 5′HEX-TACCGGCTG/ZEN/GAACGCATCAAGGAA-3′IABkFQ, reverse primer: CGCCAAGAACCTGAAGATG) of the chromosomally integrated reporter construct were used to quantify frequency of excision. For DNA isolated from PHHs, primer-probe sets specific for HBV DNA (previously described)⁵⁵ and RPPH1 (forward primer: GGAGAGTAGTCTGAATTGGGTTATG, probe: 5′Cy5-ACCTCACCT/TAO/CAGCCATTGAACTCAC-3′IAbRQSp, reverse primer: GGAGCTTGGAACAGACTCAC, reference NC_000014.9) were used to quantify total HBV DNA per cell. For DNA isolated from mouse tissues, primer-probe sets specific for HBV DNA (same as above) and TERT (forward primer: TCCATACATGCACAGGGTGT, probe: 5′Cy5-CCCTTTCAG/TAO/CCAACCCTCCGGT-3′IAbRQSp, reverse primer: TTGAGGTTGGCCCTCTGTAG, reference NC_000079.7) were used to quantify total HBV DNA per cell.

HBsAg and HBeAg analysis

HBsAg and HBeAg levels in cell culture medium were measured with HBeAg- and HBsAg-specific chemiluminescence assays (CLIA) (Ig Biotechnology). Fifty microliters (50 μL) of culture supernatant and standard curve calibrators were incubated with 50 μL enzyme conjugate (HRP labeled anti-HBs or anti-HBe monoclonal antibody in PBS containing casein, BSA, and ProlClin 300) in a 96-well test plate, shaken on a plate shaker for 60 s and then incubated for 60 min at 37°C. Following incubation, the contents of the plate were removed, and the plate was washed six times with 350 μL of wash solution (PBS-Tween, 1×). Substrate solution (hydrogen peroxide in buffer solution) was prepared and 50 μL was added to each well. The plate was incubated for 10 min in the dark, then read on a FlexStation 3 plate reader (Molecular Devices) measuring the Relative Light Units (RLUs) of each well. From the RLUs, the amount of HBeAg (PEIU/mL) or HBsAg (IU/mL) in each sample was determined.

LNP formulation

SaCas9-encoding mRNA and the pair of HBV-targeting gRNAs were mixed at 2:1:1 (in weight) and formulated in four lipid components using a microfluidic instrument at the N/P ratio of 6:1. For the AAV-HBV mouse model, SM-102, DSPC, cholesterol, and DMG-PEG2000 dissolved in ethanol at 50:10:38.5:1.5 (in molar) were used. The aqueous phase was substituted with 1 × PBS using a centrifugal filter. After concentration, the LNPs were sterilized through a 0.2-μm filter. For the Tg-HBV mouse model, the lipid solution containing LP-01, DSPC, cholesterol, and DMG-PEG2000 at 50:9:38.5:3 (in molar) was used to prepare LNPs. After dialysis in 50 mM Tris and 45 mM NaCl, pH 7.5 overnight, the LNPs were concentrated using a centrifugal filter and sterilized through a 0.2-μm filter. Sucrose was added to 5% (v/v) before storage at −80°C. LNPs used to investigate anti-HBV gene-editing therapy in combination with treatment with entecavir in the AAV-HBV mouse model (Figure 7) were formulated using FL-0207 as described in a PCT patent application WO2022230964A1. The resulting LNPs were suspended in 20 mM Tris buffer (pH 7.4) with 8% sucrose and stored at −80°C until use. The characteristics of the LNPs tested in this study are summarized in Table S5.

Ethic statement and animal care

All mice used in the AAV-HBV and Tg-HBV studies, which received a single LNP dose, were housed in individually ventilated cages at WuXi AppTec’s animal facility in Shanghai, China. Animal care and euthanasia were conducted following standard operating procedures approved by the WuXi Institutional Animal Care and Use Committee (IACUC; protocol numbers ID01-013-2021v1.0 and ID01-SH002-2023v1.2). The animal facility and IACUC were fully accredited by Association of Assessment and Accreditation of Laboratory Animal Care International (AAALAC).

Mice used to compare the efficacy of our gene-editing approach with or without ETV treatment were housed in a Yecuris Corporation IACUC accredited facility. General procedures for animal care and housing followed the Guide for the Care and Use of Laboratory Animals (National Research Council), Yecuris IACUC Policy, and Yecuris General Mouse Handling Care and Euthanasia. All animal handling was performed by trained technicians.

In vivo editing in the AAV-HBV mouse model

Each of the 5–6 weeks old C57BL/6 J male mice were transduced with 1 × 10¹⁰ vg of the AAV-HBV vector (derived from the genotype-D ayw) by tail vein injection. Serum samples were collected by submandibular bleeding to assay serum level of HBV DNA, HBsAg, and HBeAg weekly from 3 weeks post the AAV administration until termination of the in-life stage. HBV DNA was isolated with QIAamp 96 Blood DNA Kit (Qiagen) following the manufacturer’s instructions and its copy number was determined by quantitative PCR using TaqMan Universal PCR Master Mix (Thermo Fisher Scientific). HBsAg and HBeAg were assayed using HBsAg and HBeAg ELISA kits (Autobio). Mice were divided into 3 groups to minimize differences in body weights and serum HBV DNA, HBsAg and HBeAg among the groups. PBS or 35 μg or 70 μg of the anti-HBV gene editing payload encapsulated by the LNPs were administered into mice of each group 5 weeks post AAV administration. All the mice were euthanized 5 weeks after LNP administration, and the partial left liver lobe from each mouse was harvested, flash-frozen in liquid nitrogen, and stored at −80°C until further use.

In vivo editing in the Tg-HBV mouse model

Tg-HBV female mice were purchased from Beijing Vitalstar Biotechnology Co. Ltd. This mouse model was generated through pronuclear microinjection of the linearized DNA fragment carrying the 1.3-fold oversized HBV genome (GenBank: AF305422.1, subtype adw2, genotype A) into the fertilized egg of a C57BL/6 mouse. Serum samples were collected by submandibular bleeding to assay serum level of HBV DNA and HBsAg one week prior to LNP administration. HBV DNA and HBsAg were quantified as described above. The assay results were used to divide the mice into three groups for equalizing body weights, serum HBV DNA, and HBsAg among the groups. PBS or 30 μg or 60 μg of the anti-HBV gene editing payload encapsulated by the LNPs were administered into mice of each group by tail vein injection at 8–9 weeks old. Each mouse was weekly bled to assay serum HBV DNA, HBsAg, and HBeAg. All the mice were euthanized 5 weeks after LNP administration, and tissue samples including the partial left lobe of liver and part of stomach from each mouse were harvested, flash-frozen in liquid nitrogen, and stored at −80°C until further use.

Anti-HBV gene editing in the AAV-HBV mice treated with ETV

Approximately 7-week old C57BL/6 J male mice were transduced with 5 × 10¹⁰ vg of the AAV-HBV vector (derived from the genotype-D ayw) by retro-orbital sinus injection. Serum samples were collected via the retro-orbital route to assay serum HBV DNA, HBsAg, and HBeAg levels 25 days after administration of the AAV-HBV vector. The assay results were used to divide 36 of these mice into six groups to equalize serum HBsAg and HBV DNA levels across them. PBS or the LNP-encapsulated anti-HBV gene editing payload was administered at 1 or 2 mg/kg into mice of each group by retro-orbital sinus injection at 5 and 7 weeks after the AAV-HBV delivery. Mice of three groups received entecavir at 5 μg/mL in drink water during the 2-week interval of the LNP dosing. Blood was collected weekly to quantify serum HBsAg and HBV DNA levels.

In silico HBV viral target site search

HBV viral sequences were downloaded from HBVdb²⁵ for genotypes A to H. The total number of sequences included were: 970 genotype A, 1,916 genotype B, 2,523 genotype C, 1,174 genotype D, 356 genotype E, 289 genotype F, 48 genotype G, and 28 genotype H. Customized python script was used to search for target sites with a 21-nt protospacer followed by the PAM sequence of NNGRRT. Target sites from all HBV sequences were compiled, and the occurrence of each target site was computed within each genotype (Figure S1A). The occurrence of a target site represents the percentage of genomes within a certain genotype containing that target site. Those with over 80% occurrences across all genotypes were selected for downstream analysis.

Homologous site search in the human reference genome

Human genomic sites that shared high sequence identity with the selected gRNA on-target sequences were identified using a Nextflow pipeline as described previously.¹⁵ Briefly, CasOffinder⁵⁶ was used to search for homologous sites in human genome (hg38_analysis_set) with 21-nt protospacers and PAM of NNNNNN, allowing differences at up to total six positions including up to two gaps. Duplicated sites were identified by sharing the same predicted PAM locations and removed from the result. For each unique predicted PAM location, the homologous site in human genome with minimal number of mismatches and gaps was kept. Mismatches within PAM region against NNGRRT were calculated and added to the total number of mismatches. The sites that differ at up to total six positions in the alignment with the corresponding HBV on-target site are shown in Table S1.

Data analysis for short-read hybridization capture sequencing with DNA samples from HBV-infected PHHs and AAV-HBV mice

Sequence reads obtained from short-read hybridization capture analysis were analyzed using a customized Nextflow pipeline developed in house. Demultiplexed sequencing reads were first preprocessed and trimmed for sequencing adapters using FASTP.⁵⁷ Unique molecular identifier (UMI) sequences were extracted, and reads were collapsed into UMI consensus sequences using FGBIO recommended pipeline.⁵⁸ Editing frequencies around gRNA target sites were analyzed and quantified using CRISPResso2.⁵⁴ Sequence reads containing at least 19 nucleotides mapped on two distinct regions of the reference sequences were identified as structural variants using FGSV,⁵⁹ and those with guanine stretches longer than nine nucleotides were excluded from the subsequent analyses, reads containing ITR sequences were excluded. Reads with excision or inversion were identified if DNA break point was within 25 nt of both gRNA cut sites. Sequencing depth at regions of interest was computed using Mosdepth.⁶⁰ Host genomic sites with potential HBV integration were searched for homology against each gRNA sequence using Calitas⁶¹ with ±60 bp window. Reference sequences used during this analysis include human genome (hg38 analysis set), mouse genome (mm39), HBV reference AYW (GenBank: U95551.1), and AAV2 ITR sequence.

Data analysis for long-read hybridization capture sequencing with DNA samples from Tg-HBV mice

Sequence reads obtained by long-read hybridization capture sequencing were analyzed using a customized Nextflow pipeline developed in house. Briefly, demultiplexed PacBio sequencing reads were quality checked using NanoPlot⁶² and pre-processed to remove Illumina sequencing adapters using LIMA.⁶³ Duplicated sequencing reads were identified and removed using Pbmarkdup.⁶⁴ Three-nucleotide sequences from each end of sequencing reads were extracted as UMI sequences. Reads were aligned to reference genomes using bwa⁶⁵ and ngm_lr.⁶⁶ Reference genomes include mouse genome (mm39) and HBV reference AYW (GenBank: U95551.1), and pK18msr (GenBank: AB694752.1). Sequencing depth of region of interested was computed using bwa mapped reads with Mosdepth.⁶⁰ Reads aligned to HBV reference genome were extracted and further characterized for read composition. HBV containing reads were blasted against the same reference genome set using command line blast+.⁶⁷ Customized python script was used to remove duplicated reads sharing the same UMI sequences. Reads with potential integration or translocation were identified by having sequencing from host genome wrapping around HBV sequences, gaps between host sequence and HBV sequence less than 50 nt, and blast identity greater than or equal to 95%. The host genomic sites with potential HBV integration or translocation were searched for sequence homology against gRNAs using Calitas⁶¹ with ±60 bp window. Customized python scripts were used to quantify reads with excision or inversion with ±20 nt window near gRNA target sites.

Analysis of results from GUIDE-seq

GUIDE-seq analysis was performed using in house pipeline as described previously.¹⁵ Briefly, sequencing reads were demultiplexed using deML.⁶⁸ Identification of dsODN integration sites was conducted with the following modifications from the previous publication²⁶: Genomic sites that are homologous to an intended target site were searched using Calitas⁶¹ within 60 base pairs from dsODN integration sites. Mosdepth tool⁶⁰ was used to calculate base coverage at the identified sites. The sites with total number of differences less than or equal to 7 in the alignment with the HBV on-target sites were selected for downstream analysis.

Data and code availability

All data shown in the main text and the supplemental information are available from the corresponding authors upon request.

Acknowledgments

We thank FUJIFILM Toyama Chemical Co., Ltd. for technical support in LNP formulation. The studies described in this manuscript was conducted with only an internal fund of Excision BioTherapeutics.

Author contributions

S.S., W.H., J.G., and R.T. designed the experiments; S.S., A.A., and R.T. performed the experiments; W.H. analyzed the NGS raw data. All the authors wrote and reviewed the manuscript.

Declaration of interests

All the authors were employees of Excision BioTherapeutics Inc. and received salaries and stock options.