Orthologous CRISPR/Cas9 systems for specific and efficient degradation of covalently closed circular DNA of hepatitis B virus

Abstract

Covalently closed circular DNA (cccDNA) of hepatitis B virus (HBV) is the major cause of viral persistence and chronic hepatitis B. CRISPR/Cas9 nucleases can specifically target HBV cccDNA for decay, but off-target effects of nucleases in the human genome limit their clinical utility. CRISPR/Cas9 systems from four different species were co-expressed in cell lines with guide RNAs targeting conserved regions of the HBV genome. CRISPR/Cas9 systems from Streptococcus pyogenes (Sp) and Streptococcus thermophilus (St) targeting conserved regions of the HBV genome blocked HBV replication and, most importantly, resulted in degradation of over 90% of HBV cccDNA by 6 days post-transfection. Degradation of HBV cccDNA was impaired by inhibition of non-homologous end-joining pathway and resulted in an erroneous repair of HBV cccDNA. HBV cccDNA methylation also affected antiviral activity of CRISPR/Cas9. Single-nucleotide HBV genetic variants did not impact anti-HBV activity of St CRISPR/Cas9, suggesting its utility in targeting many HBV variants. However, two or more mismatches impaired or blocked CRISPR/Cas9 activity, indicating that host DNA will not likely be targeted. Deep sequencing revealed that Sp CRISPR/Cas9 induced off-target mutagenesis, whereas St CRISPR/Cas9 had no effect on the host genome. St CRISPR/Cas9 system represents the safest system with high anti-HBV activity.

Electronic supplementary material

The online version of this article (10.1007/s00018-019-03021-8) contains supplementary material, which is available to authorized users.

Introduction

Infection with hepatitis B virus (HBV), family Hepadnaviridae, can lead to acute or chronic infections of the liver. Over 250 million people are chronically infected and approximately 1 million people die annually due to consequences of HBV infection [1, 2]. Continuous replication of HBV in human hepatocytes is associated with progressive DNA damage [3], ER stress [4], genome instability [5] and pro-oncogenic effects of HBV X protein [6]. Ultimately, chronic HBV infection develops to fibrosis and hepatocellular carcinoma [7].

HBV is a small virus that replicates via DNA and RNA intermediates, but the key factor of its chronic infection is covalently closed circular DNA (cccDNA), an episomal double-stranded form of the HBV genome that persists in the nucleus of infected hepatocytes and serves as a template for transcribing all viral RNAs, including pregenomic RNA (pgRNA) [8]. Although a number of antiviral therapeutics is applied in the clinic, these treatments fail to eliminate HBV cccDNA [9]. No cure is yet available for chronically infected patients [10–12]. Among the approaches that act directly on HBV cccDNA, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated Cas9 nucleases (CRISPR/Cas9) may represent the most promising option for highly specific and efficient depletion of the cccDNA reservoir.

CRISPR/Cas9 type II systems are amongst the most investigated in mammalian and human research. They represent the minimal CRISPR/Cas systems, operating on a principle whereby the nucleolytically active Cas9 protein is recruited to the double-stranded DNA (dsDNA) target by a small guide RNA (gRNA). Cas9 then cleaves a target site approximately three nucleotides downstream of the protospacer adjacent motif (PAM) to generate DNA double-strand breaks [13]. After cleavage, the double-strand breaks are repaired primarily by the error-prone non-homologous end-joining pathway and result in generation of insertions/deletions (indel) or single-nucleotide polymorphisms [14]. Cleavage of HBV cccDNA may result in in-frame or out-of-frame mutations, or degradation of cccDNA. The key factors for target site recognition are: (1) a gRNA sequence complementary to the target and (2) a PAM sequence downstream of the target site, which are required for the Cas9 protein to recognize and cleave the target DNA site [15].

Several attempts to increase SpCas9 specificity were successful [16–18], but resulted in reduced on-target activity. Orthologous CRISPR/Cas9 systems with longer and more restrictive PAMs are considered safer alternatives to SpCas9 [19–21]. In particular, Staphylococcus aureus Cas9 (SaCas9), the PAM sequence of which is 5′-NNGRRT-3′, showed highly efficient degradation and mutation of cccDNA while sparing the host genome; the mutation rate at predicted off-target loci was below 0.1% [22, 23]. Similarly, Cas9 orthologues from Neisseria meningitidis (NmCas9; PAM: 5′-NNNNGATT-3′) [19] and Streptococcus thermophilus (StCas9; PAM: 5′-NNAGAAW-3′ and 5′-NGGNG-3′ for CRISPR1 and CRISPR3 loci, respectively) [20] demonstrated high on-target and very low off-target activity in human cells. Thus far, all type II CRISPR/Cas9 systems have demonstrated only DNA-targeting activity except one, from Francisella novicida (FnCas9; PAM: 5′-NGG-3′), which was successfully employed to suppress hepatitis C virus (HCV) RNA in vitro [24] in a PAM-independent fashion.

Another challenge to targeting HBV cccDNA by CRISPR/Cas9 is the high heterogeneity of HBV genomes [25], complicating the design of gRNAs that would target multiple or all HBV genotypes (A–H). A number of attempts were undertaken using SpCas9 [26] and SaCas9 [22, 23], but these systems still cover only a portion of HBV genotypes and subgenotypes. Thus, identifying new gRNAs for specific Cas9 species targeting highly conserved regions of the HBV genome with minimal predicted off-target effects is necessary for CRISPR/Cas9 systems to move into clinical trials for HBV treatment.

To better describe properties of StCas9, we systematically analyzed its ability to tolerate mismatched nucleotides and cleave DNA at off-target regions. This mismatch tolerance is also important for targeting HBV genetic variants, which arise during chronic HBV infection and represent diverse genetic variants of the predominant HBV genotype (HBV quasispecies). First, single-nucleotide polymorphisms are the most common types of nucleotide substitutions within quasispecies. Second, mismatch tolerance indicates the ability of the Cas9 protein to unwind and cut double-stranded DNA at off-target regions in the host genome and, therefore, defines potential safety of the Cas9 protein. As previously mentioned, classic SpCas9 tolerates up to six mismatches, thus greatly expanding the array of potential off-target sites in the human genome [27, 28]. To this end, NmCas9 and StCas9 (CRISPR locus 3) were shown to be less tolerant of mismatched nucleotides. We analyzed mismatch tolerance of CRISPR1/StCas9 and demonstrated that it tolerated single-nucleotide mismatches and thus could effectively target single-nucleotide HBV genetic variants for degradation, whereas two- to three-nucleotide mismatches significantly impaired its activity towards HBV. Coupled with the most restrictive PAM requirements among all described systems, limited mismatch tolerance makes StCas9 much more specific than other orthologues. Deep sequencing by our group did not identify off-target mutagenesis of StCas9 at predicted loci, but detected off-target mutations generated by SpCas9.

In this work, we used four orthologous Cas9 systems—the classic SpCas9, FnCas9 capable of targeting RNA, NmCas9, and StCas9 (CRISPR1 locus, further referred to as StCas9 if not otherwise defined)—with the largest set of gRNAs for SpCas9 to date and an exhaustive set of gRNAs for NmCas9 and StCas9 that cover all conserved target regions in the HBV genome to develop an optimal approach for HBV targeting from the positions of efficacy and safety. In addition, we analyzed mismatch tolerance and off-target activity of previously not studied CRISPR1/StCas9 in respect of HBV.

Results

Potent anti-HBV activity of S. pyogenes and S. thermophilus CRISPR/Cas9 targeting conserved regions of the HBV genome

To identify CRISPR/Cas9 with the most potent anti-HBV activity, we used a well-established system in HepG2 cells and PCR-based analysis of CRISPR/Cas9-mediated inhibition of HBV transcription [29]. CRISPR/Cas9 can lead to cccDNA degradation and frameshift mutations, thereby reducing HBV transcription. The HBV genome has four overlapping open reading frames (S, P, X, and C) encoding viral proteins (Fig. S1). gRNAs with potent anti-HBV activity would reduce levels of both pgRNA, a major transcript of cccDNA, and S-RNA, an important template for the synthesis of the viral protein HBsAg. We co-transfected HepG2 cells with an HBV (genotype D)-expressing plasmid, a Cas9-encoding vector, and a PCR product containing the U6 promoter and encoding a specific gRNA (Fig. S1A). The PCR products encoded one of the 50 SpCas9/gRNAs (Fig. S1B), 5 FnCas9/gRNAs (Fig. S1C), 10 StCas9/gRNAs (Fig. S1D) or 6 NmCas9/gRNAs (Fig. S1E), with the highest predicted activity based on the Broad Institute Genetic Perturbation Platform and CCTop CRISPR/Cas9 target online calculator [30, 31]. These sets of gRNAs included all possible conserved targets for NmCas9 and StCas9 systems, and represented the largest set of SpCas9/gRNAs for HBV ever described. A brief summary of orthologous type II systems is given in Table S1. Levels of HBV pgRNA and S-RNA were measured to determine inhibition of HBV transcription by the various Cas9 systems (Fig. 1a–d; complete set of gRNAs is provided in Fig. S2). Overall, the most effective gRNA/SpCas9 and StCas9 systems exhibited similar antiviral efficacy, with SpCas9 suppressing HBV pgRNA levels by 85–93% and S-RNA levels by 85–92% (Fig. 1a) and StCas9 by up to 99% (Fig. 1d). NmCas9 was less effective, with the best-performing gRNA (Nm4) reducing pgRNA and S-RNA levels by 70% (Fig. 1b). FnCas9 had almost no effect on HBV transcription (Fig. 1c). Sp20 and Sp37, two of the most effective SpCas9/gRNAs, target highly conserved regions of the HBV genome (Fig. 1a, heatmap). The HBV region targeted by Sp20 is highly conserved in all HBV genotypes: 98% in genotype A, 83% in B, 90% in C, 91% in D, 86,4% in E, 96% in F, 100% in G, and 88% in H. The region targeted by Sp37 is highly conserved among the four most common HBV genotypes: 97% conserved in A, 81% in B, 91% in C, and 97% in D; furthermore, it is 75% conserved in E.

Fig. 1.

Anti-HBV activity of orthologous CRISPR/Cas9 type II systems. a–d Conservation of orthologous CRISPR/Cas9 targets across HBV genotypes A–H and anti-HBV activity of various gRNAs as determined by measuring pgRNA and S-RNA levels. pgRNA and S-RNA levels were normalized to Cas9 mRNA expression. Conservation of a specific locus was analyzed by directly mapping gRNA sequences in HBV genomes of different genotypes (A–H) (left side of the graphs). Intensities of red and green colors indicate low and high rates of conservation, correspondingly. ⁺p < 0.05, ^#p < 0.01, ^~p < 0.001, *p < 0.0001

Among the five StCas9/gRNAs targeting conserved regions of the HBV genome, St3 and St4 downregulated pgRNA S-RNA levels by over 87% and 60%, respectively, whereas St10 suppressed both pgRNA and S-RNA transcription by > 93% (Fig. 1d). The St4 target was 92–98% conserved among the four major genotypes, while the St3 and St10 targets were 69–98% conserved in HBV A–D (Fig. 1d, heatmap). St1 suppressed HBV transcription by over 90% and was the most effective gRNA for StCas9, but it targeted a fairly heterogenic genomic locus and was not further investigated. While we initially expected to identify gRNAs effective against each individual HBV genotype, the most potent gRNAs fortuitously targeted highly conserved regions in the HBV genome.

In summary, we identified gRNAs for SpCas9, StCas9, and NmCas9 systems that effectively target conserved regions of HBV genome. Combinations of St10 and St3 or St4 enable specific targeting of all HBV genotypes except for genotype F. To ensure the validity of these data, all experiments at the screening stage were reproduced at least five times.

Targeting with combinations of gRNAs did not significantly improve SpCas9 anti-HBV activity in our experiments (Fig. S3A). However, combining any StCas9/gRNAs consistently reduced not only pgRNA, but resulted in more decline in S-RNA levels (Fig. S3B). We speculate that dual gRNAs systems lead to more efficient generation of frameshift mutations and degradation of cccDNA, which should lead to better antiviral activity at longer periods of investigation.

To determine the antiviral efficacy at the protein level, HepG2 cells were co-transfected with an HBV-expressing plasmid and the CRISPR/Cas9 systems shown to suppress HBV transcription most potently (SpCas9 and StCas9). To strengthen PCR data, we selected SpCas9/gRNAs capable of efficient downregulation of S-RNA (Sp20, Sp39) and those resulting in moderate (Sp40) or no decline in S-RNA levels (Sp42). StCas9 system included gRNAs targeting conservative HBV genomic regions that suppressed viral transcription most potently (St3, St4 and St10). Expectedly, HBsAg levels correlated with S-RNA levels: Sp20 reduced HBsAg by 78%, Sp39 was less effective (72% decline), and Sp40 reduced HBsAg only by 40%, whereas HBsAg levels after transfection of Sp42 were not different from a mock-treated control (Fig. 2a). Similar to PCR data, St3 and St4 suppressed HBsAg levels by ≈ 50%; St10 transfection led to a 73% decline in HBsAg (Fig. 2b). HBcAg, a viral protein translated from pgRNA, and SpCas9 expression in transfected cells were then analyzed by immunofluorescence 5 days post-transfection (Fig. 2c–e). No signal was detected in the negative control cells (NT), while cells transfected with an HBV-expressing plasmid and SpCas9 with a non-coding (nc) gRNA were positively co-stained with anti-HBcAg and anti-SpCas9 antibodies. Samples with StCas9 and a negative control gRNA were negative for anti-SpCas9 signal and positive for HBcAg. In all groups with the selected gRNAs (Sp20, Sp39, Sp40, St3, St4, and St10), HBcAg expression was markedly lower compared to mock control. Percentage of HBcAg-positive cells ranged from 5 to 20% in mock-treated control groups (Fig. 2d, e). HBcAg-positive cells were not detected in Sp20, Sp40, St3 and St10 experimental groups or were very rare and weakly stained in Sp39 and St4 groups.

Fig. 2.

Inhibition of HBV protein production. HBsAg levels in culture medium after transfection of a SpCas9 and b StCas9 systems. c Reduction of HBcAg expression by various CRISPR/Cas9 systems 5 days post-transfection. HepG2 cells transfected with HBV-expressing plasmid and the indicated CRISPR/Cas9 system were co-stained for HBcAg (green) and SpCas9 (red) protein. Cell nuclei were labeled by Hoechst33342 dye (blue). Quantitative analysis of HBcAg-positive and SpCas9-positive cells after transfection of d SpCas9 or e StCas9 CRISPR/Cas9 systems from an experiment as described in c. NT not transfected

Degradation of recombinant HBV cccDNA by orthologous type II systems and effects of cccDNA methylation

The major drawback of HepG2 transfection experiments with an HBV-encoding plasmid is that cccDNA cannot be detected due to technical reasons [29]. To overcome this limitation and analyze the effects of CRISPR/Cas9 on HBV more precisely, we transfected HepG2 with a circular HBV DNA (rcccDNA) produced by in vitro ligation as shown before [32, 33]. Using described technique, HBV DNA may be ligated with a low efficacy; so we additionally verified acquired data with a novel rcccDNA generation system that lacks heterologous sequences and can be generated very efficiently in bacterial cells [34]. Generally, NmCas9 suppressed HBV transcription more actively than in a model with a plasmid HBV (Fig. 3a, b), but the magnitude of the antiviral effect imposed by Nm gRNAs was similar between the experiments with an HBV-expressing plasmid (Fig. 1b) and rcccDNA. This more robust suppression of rcccDNA transcription by NmCas9 could be due to lower numbers of rcccDNA templates and lower rates of HBV replication in the rcccDNA transfection model compared to transfection of a plasmid expressing HBV from a CMV promoter, leading to an overabundance of NmCas9 relative to rcccDNA targets. The most effective NmCas9/gRNA, Nm4, suppressed HBV transcription by more than 80% (Fig. 3a), and decreased HBV DNA by 63% and rcccDNA levels by 82% (Fig. 3b).

Fig. 3.

Antiviral activity of orthologous CRISPR/Cas9 and effects of rcccDNA methylation on anti-HBV activity. Anti-HBV effects of orthologous CRISPR/Cas9 (Nm, Fn, St and Sp) were analyzed by rcccDNA transfection. Antiviral activity of NmCas9 and FnCas9 in a HepG2 transfection model were analyzed by measuring pgRNA and S-RNA (a, c) normalized to NmCas9 and FnCas9 mRNA, and HBV DNA and cccDNA levels (b, d) relative to β-globin. e–l Comparison of HBV intermediates in HepG2 cells after transfection of unmethylated or methylated rcccDNA. Effects of St and Sp CRISPR/Cas9 on e, g, i, k pgRNA and S-RNA levels (relative to Cas9 mRNA levels), f, h, j, l HBV DNA and rcccDNA (relative to β-globin) in unmethylated and methylated rcccDNA. Asterisks indicate statistically significant differences. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

In HCV research, cytoplasmic FnCas9 promoted PAM-independent degradation of HCV RNA [24]. FnCas9 can also potentially target HBV RNA and cccDNA in the nucleus or HBV RNAs in the cytoplasm. To investigate the effects of FnCas9 on HBV, we created FnCas9-producing constructs with and without nuclear localization signals (NLS) and co-transfected these constructs with the most effective FnCas9/gRNA (Fn1) into HepG2 cells. FnCas9 without NLS does not enter the nucleus and can interact only with cytoplasmic HBV RNA.

Similar to NmCas9, FnCas9 acted more vigorously in rcccDNA-transfected model than in HBV-infected cells, resulting in statistically significant, but modest decrease in HBV transcription: 47% decline in pgRNA, 17% in S-RNA (Fig. 3c), 57% in HBV DNA and 41% in rcccDNA (Fig. 3d). However, this effect was abolished when the NLS was deleted and the FnCas9 protein could not enter the nucleus to target rcccDNA. Most likely, FnCas9 cleaves only rcccDNA and does not target HBV RNA, including pgRNA, or this effect is very weak and does not contribute to its anti-HBV activity. However, it should be noted that this low on-target activity may in part be explained by difficult self-assembly of a large protein in HepG2 cells [35]. Specific analysis of anti-HBV activity of FnCas9 ribonucleoprotein complexes is needed to unambiguously define its effect on HBV RNA and cccDNA.

It has been reported previously that highly condensed regions of the human genome are cleaved with consistently lower efficacy than genomic regions associated with euchromatin [20, 36–39]. Moreover, complex chromatin structure significantly inhibits DNA cleavage by CRISPR/Cas9 [40]. HBV cccDNA structurally resembles the architecture of genomic DNA and is wrapped around histone and non-histone proteins similarly to human genomic DNA [41]. Epigenetic modifications of HBV cccDNA have been assumed to limit cccDNA degradation, in particular by APOBEC deaminases [42]. Thus, we decided to specifically address the effects of HBV cccDNA methylation, a frequent event in chronic hepatitis B patients associated with a transcriptionally inactive state, on anti-HBV activity of StCas9 and SpCas9, which exhibited the highest antiviral activity in our screening. HBV cccDNA has four canonical CpG islands (I–IV) prone to methylation (Fig. S4A) [43, 44]. HBV cccDNA was shown to be unmethylated in cell culture, so we created rcccDNA and methylated it in vitro using S.sssI methyltransferase. The generated rcccDNA was heavily methylated by this process, as evidenced by restriction analysis using an HpaI methylation-sensitive restriction enzyme that generated a ladder of DNA fragments in unmethylated rcccDNA, while methylated rcccDNA was intact after the restriction reaction (Fig. S4B). Moreover, upon transfection into HepG2 cells, transcription of methylated rcccDNA was significantly limited: pgRNA and S-RNA levels were 8.5% and 26%, respectively, compared to levels transcribed from unmethylated rcccDNA (Fig. S4C). To target methylated rcccDNA, we selected the most effective SpCas9/gRNAs that target CpG islands in rcccDNA: Sp9, which cleaves rcccDNA in CpG island III, and Sp40, which cleaves in CpG island II (Fig. S4A). We also used St3, which targets CpG island I, and St10, which targets island II, as well as St4, which targets rcccDNA outside of CpG islands but was one of the most potent gRNA for StCas9 system (Fig. S4A). In rcccDNA transfection experiments, SpCas9 suppressed pgRNA and S-RNA levels by 40–80% (Fig. 3e) and rcccDNA by over 90% (Fig. 3f). Similar results were observed for methylated rcccDNA (Fig. 3i, j), where HBV transcription dropped by over 90% with any StCas9/gRNA used (Fig. 3g, h). StCas9 transfection resulted in over 80% decrease in HBV transcription (Fig. 3i), 90% decrease in rcccDNA and reduced HBV DNA by 40% (St4)–75% (St10) (Fig. 3j). Similar results were observed for suppression of transcription by StCas9 at methylated rcccDNA (Fig. 3k). Since only a portion of methylated rcccDNA is transcriptionally competent, the most reliable marker for comparing nucleolytic activity of CRISPR/Cas9 is the intracellular rcccDNA level. rcccDNA methylation had no effect on anti-HBV activity of St3, St10, Sp9, and Sp40 (Fig. 3h, l), as these gRNAs effectively degraded rcccDNA. St4, however, reduced methylated rcccDNA levels less prominently (Fig. 3j, l). On-target sequencing of rcccDNA in cells treated with St4 gRNA demonstrated that mutation rates were almost twofold lower in methylated than unmethylated rcccDNA (Figs. 4b, S5B; Table S2). Although St10 resulted in similar reduction of unmethylated and methylated rcccDNA (Fig. 3j, l), mutation rates in methylated rcccDNA were over 50% lower compared to unmethylated rcccDNA (Figs. 4c, S5C; Table S2). In contrast, anti-HBV activity of St3 was not affected by methylation neither at the level of rcccDNA (Fig. 3j, l), nor by rcccDNA mutation rates (Figs. 4a, S5A; Table S2). In fact, indels were detected with higher frequency in methylated rcccDNA after St3 transfection.

Fig. 4.

Deep sequencing of on-target sites in methylated and unmethylated cccDNA. Detection of indel mutations in unmethylated or methylated rcccDNA by St CRISPR/Cas9 a St3, b St4 and c St10, shown as number of mutations per 1000 reads. X-axis indicates the target regions in HBV genome, PAM protospacer adjacent motif

Impaired anti-HBV activity of St4 and reduced mutation rates in rcccDNA by St10 demonstrates that methylation may affect antiviral properties of CRISPR/Cas9. To the authors’ knowledge, this is the first indication that epigenetic modifications of HBV genome (e.g., DNA methylation) may impair antiviral activity of CRISPR/Cas9.

In summary: (a) both SpCas9 and StCas9 exhibited similar anti-rcccDNA activity with certain gRNAs; (b) the most effective SpCas9 and StCas9 systems resulted in over 90% reduction in rcccDNA; (c) methylation of rcccDNA inside or outside of CpG islands may impair rcccDNA cleavage and degradation by CRISPR/Cas9.

CRISPR/Cas9 systems effectively inhibit HBV replication in stable cell lines with integrated HBV genome

To mimic clinical infection more closely, we used two cell lines with HBV DNA integrated into the human genome: HepG2-1.1merHBV cells that activate HBV life cycle only upon doxycycline treatment, and HepG2-1.5merHBV that produce HBV constitutively from a wild-type promoter. Followed by highly efficient (> 85%; data not shown) nucleofection of HepG2 stable cell lines with CRISPR/Cas9, transfected cells were selected with blasticidin for 5 days (Fig. 5a, f). Reduction in HBV intermediates by all constructs was high, but more pronounced in HepG2-1.5merHBV cells. Sp20 gRNA was the most potent, inducing frequent indel mutations (Fig. S6) and reducing HBV transcription by over 85% (Fig. 5c) and cccDNA by over 95% (Fig. 5e). Similarly, St4 and St10 almost completely blocked HBV transcription: pgRNA levels dropped to 5.6% and 0.58%, while S-RNA to 4.4% and 0.25%, correspondingly (Fig. 5h). All StCas9/gRNAs significantly reduced cccDNA as well (Fig. 5i, j).

Fig. 5.

CRISPR/Cas9-mediated suppression of HBV replication in stable cell lines. Experimental design in a HepG2-1.1merHBV and f HepG2-1.5merHBV cell lines. Suppression of HBV transcription by b, c Sp and g, h St CRISPR/Cas9. Alterations in HBV DNA and cccDNA levels (relative to β-globin) post d, e SpCas9 and i, j StCas9 transfection. k Inhibition of HBV cccDNA degradation by NU7026. Cells were mock-transfected or transfected with CRISPR/Cas9 and Sp20 gRNA (Sp20) and treated by NU7026 for 3 days or DMSO as control. l Frequency of indel mutations in mock and Sp20 treated by DMSO or NU7026. Distribution of m insertions and n deletions by length in experimental groups. Numbers indicate the size intervals of indels. o The proposed model of HBV cccDNA degradation by CRISPR/Cas9 and inhibition of its degradation by small molecule NU7026. Asterisks indicate statistically significant differences: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Incidence of CRISPR/Cas9-induced mutations in our experimental setting did not correlate with overall antiviral activity or a decline in cccDNA/rcccDNA levels. This is explained by preferential degradation of cccDNA/rcccDNA followed by transfection of CRISPR/Cas9. This observation was reported before [22] and shown again in our recent study where we discovered that small molecule NU7026 precludes degradation of HBV cccDNA (unpublished results). To demonstrate that HBV cccDNA is actually degraded in our experimental setting, HepG2-1.1merHBV cell line was transfected with Sp20, and treated by a small molecule NU7026, an inhibitor of non-homologous end-joining of DNA double-stranded breaks [45]. Transfected cells were selected with blasticidin. Similar to our previous results, Sp20 induced a substantial decline in cccDNA, but this reduction was ≈ 3.86 times less prominent after treatment by a small molecule NU7026 (Fig. 5k). Deep sequencing revealed that the remaining cccDNA in cells treated by NU7026 was heavily mutated by CRISPR/Cas9: frequency of indels was ≈ 3 times higher compared to Sp20+DMSO group (Fig. 5l). Analysis of indel distribution by size demonstrated that Sp20 with NU7026 resulted in 2.4-fold increase in 1–10-bp insertions (Fig. 5m) and 3.5-fold increase in 1–10-bp deletions (Fig. 5n) compared to Sp20 with DMSO. Based on these results, we conclude that the majority of HBV cccDNA in our experimental setting is degraded (Fig. 5o). Inhibition of NHEJ by NU7026 rescues CRISPR/Cas9-mediated degradation of HBV cccDNA, resulting in a very efficient erroneous repair of double-strand breaks.

To conclude, SpCas9 and StCas9 effectively degraded the majority of cccDNA and the remaining cccDNA pool was composed of intact and mutated cccDNA. In this respect, frequency of mutations at target sites in HBV cccDNA is not a reliable marker to assess anti-HBV activity of CRISPR/Cas9.

Low mismatch tolerance of StCas9 system and potential escape of HBV quasispecies

In this study, we identified highly effective gRNAs for SpCas9 and StCas9 targeting conserved regions of the HBV genome. However, the absence of proofreading function of HBV reverse transcriptase [46, 47], high viral replication rates [46], and mutations introduced into the HBV genome by intracellular host restriction factors [48–50] contribute to the development of closely related but distinct HBV variants in the host, collectively known as quasispecies [51–53]. Imperfect matches between HBV quasispecies and gRNA might hamper effective cleavage of certain HBV variants by CRISPR/Cas9 [54]. Although SpCas9 demonstrates a high tolerance for single and multiple mismatches between gRNA and the target, some of the orthologous Cas9 proteins, including NmCas9, show higher specificity and do not recognize DNA as a target if it contains several mismatched nucleotides. The mismatch tolerance of CRISPR1/StCas9 system has never been studied before. To address these issues, we generated St4 gRNAs with single-nucleotide (M1–M20), double-nucleotide (M1M2–M19M20), and triple-nucleotide mismatches (M1M2M3–M18M19M20) (Fig. 6a). The M1 mutation is at the most distal nucleotide of target gRNA, while M20 lies in the vicinity of PAM. In addition to consecutive mutations, we generated double-nucleotide mutations corresponding to the most frequent variants of HBV within all HBV genotypes (M11M17 and M8M20) (Fig. 6a).

Fig. 6.

Analysis of StCas9 mismatch tolerance. a Schematic overview of generated St4 mutant gRNAs with nucleotide mismatches at the positions indicated on the bottom. In all cases, the nucleotide was changed to a complementary one to preserve CG composition. Effects of nucleotide mismatches on anti-HBV activity were analyzed by PCR measuring b, c HBV transcription, d HBV DNA and e cccDNA levels. Black squares indicate mutations at labeled sites. pgRNA/S-RNA levels were normalized to StCas9 mRNA, and HBV DNA/cccDNA levels were normalized to β-globin. ⁺p < 0.05, ^#p < 0.01, ^~p < 0.001, *p < 0.0001

Single-nucleotide mismatches were well tolerated by StCas9 and almost did not impair its anti-HBV activity. Inhibition of HBV transcription (Fig. 6b, c), and HBV DNA (Fig. 6d) and rcccDNA (Fig. 6e) replication by StCas9 gRNA mutants M1–M20 was similar to that of non-mutated St4. St4 gRNAs with mutations at several positions (M2, M4, M10, M14) decreased HBV DNA levels even further than non-mutated gRNAs, while M12, M16, and M19 mutations led to less efficient rcccDNA degradation by StCas9 (Fig. 6e). Deep sequencing revealed that all St4 gRNAs with single-nucleotide mismatches resulted in cuts in rcccDNA and indel formation, albeit with variable efficiency (Fig. S7; Table S3), except M20, which did not produce a typical indel distribution pattern. Still, frequency of indels was similar to St4 group (Table S3). In contrast, every gRNA with double- and triple-nucleotide mutations almost nullified anti-HBV activity of StCas9 by all parameters tested (Fig. 6). Deep sequencing of individual St4 gRNAs with two- (Fig. S8) and three (Fig. S9)-nucleotide mismatches revealed that certain mutations blocked nucleolytic activity of StCas9 (M1–M2, M3–M4, M5–M6, M6–M7, M8–M9, M18–M19, M19–M20, M1–M2–M3, M4–M5–M6, M8–M9–M10, M11–M12–M13, M14–M15–M16) (Figs. S8, S9; Table S3), whereas others still introduced indels into HBV rcccDNA without significant anti-HBV activity (M8–M20, M10–M11, M11–M17, M12–M13, M14–M15, M16–M17, M5–M6–M7, M8–M9–M10, M11–M12–M13, M14–M15–M16, M18–M19–M20) (Figs. S8, S9; Table S3). To conclude, single-nucleotide mismatches do not impact StCas9 anti-HBV activity; two- and three-nucleotide mismatches significantly impair antiviral potency of StCas9, but the protein still exerts its nucleolytic activity to generate indel mutations in cccDNA.

Safety and off-target activity of StCas9

The number of predicted off-target sites was very limited for StCas9/gRNAs and contained at least three to four mismatched nucleotides between a gRNA sequence and a potential target in the human genome (Table S4). In contrast, the number of potential off-target sites for HBV-targeting SpCas9/gRNAs ranges from dozens (for Sp4, Sp40, Sp46) to several hundreds (for Sp15, Sp20, Sp37 etc.). Moreover, potential off-target sites for the majority of SpCas9/gRNAs contained only one mismatch or less than three mismatches.

We selected ten off-target sites for StCas9 and ten off-target sites for SpCas9 (Fig. 7; Table S5), amplified them using high-fidelity Q5 polymerase, and deep-sequenced the amplicons. Description of selected off-target sites is provided in Table S5. We found no detectable off-target activity of StCas9 at any of the sites tested, and frequent off-target mutagenesis in four out of ten loci for SpCas9 (Fig. 7; Table S5). Frequency of mutations after St CRISPR/Cas9 transfection was the same as in mock-treated control. Analysis of indel mutations and deletion distribution in off-target sites did not reveal off-target activity for StCas9 (Fig. 7; Table S5). At the same time, SpCas9 induced frequent indel mutations (up to 4.76 per 1000 reads) in four out of ten potential off-target sites. The rest six of ten SpCas9 potential off-target sites were not proved to be statistically significant, but in general had much higher rates of mutations compared to StCas9 (Fig. 7; Table S5).

Fig. 7.

Deep sequencing of predicted off-target regions in the host genome for StCas9 and SpCas9. Frequency of indel mutations is indicated for a SpCas9 and b StCas9. Y-axis indicates off-target sites. Blue bars indicate mutation frequencies in mock-treated samples (Control), red bars indicate mutation frequencies post-transfection of HBV-targeting CRISPR/Cas9. ****p < 0.0001

Given that StCas9 on-target activity was significantly compromised by gRNAs containing two or three mismatched nucleotides, we conclude that four mismatches would abolish StCas9 activity. Indeed, no off-target mutations were detected in cells transfected with StCas9 constructs. Therefore, StCas9 with selected gRNAs can be considered a safe orthologue for HBV cccDNA targeting in humans.

Discussion

CRISPR/Cas9 system from S. pyogenes is one of the most extensively studied systems of site-specific nucleases. Numerous advantages of this system, including its flexibility, easy target search in a required region due to its short PAM motif, and robust on-target cleavage are compromised by high off-target activity [55, 56]. SpCas9 tolerates up to six mismatches between the gRNA and target DNA [19, 57]. Indeed, the CRISPR/Cas9 system can induce large unwanted deletions into the human genome that could promote chromosomal abnormalities and development of cancer [58]. A recent study identified numerous mutations generated by Cas9 nucleolytic activity in vivo [59]. Potential unwanted mutagenesis significantly hampers translation of experimental results with CRISPR/Cas9 into the clinic. Certainly, clinical application of CRISPR/Cas9 can be considered only if the risks of off-target cleavage are close to zero. Thus, identifying new CRISPR/Cas9 systems with much higher specificity is urgently needed. Esvelt et al. [60] first described orthogonal Cas9 proteins and their PAM motifs. Moreover, Muller et al. [20] and Lee et al. [19] demonstrated that CRISPR/Cas9 with longer and more restrictive PAM requirements represent safer alternatives to the classic SpCas9 system. To the best of our knowledge, only CRISPR/Cas9 from S. aureus with a PAM motif longer than the classic SpCas9 has been previously harnessed to degrade the HBV genome [22, 23].

To expand the toolkit of CRISPR/Cas9 targeting HBV, in our current work, we compared the anti-HBV activity of four orthologous CRISPR/Cas9 systems: SpCas9, StCas9, NmCas9, and FnCas9. FnCas9 exhibited only weak anti-HBV activity, possibly because of inefficient assembly of the huge FnCas9 protein in human cells [35]. In contrast to HCV studies [24], we observed poor inhibition of HBV transcription and ineffective interruption of the HBV lifecycle by FnCas9. Of the remaining Cas9 systems, NmCas9 cleared HBV at the slowest rate. SpCas9 and StCas9 effectively targeted HBV cccDNA for degradation and suppressed HBV replication. Notably, our study exhaustively analyzed gRNAs for NmCas9 and StCas9 systems, as all gRNAs targeting conserved regions of the HBV genome were considered.

It should be noted that when this study was underway, a number of investigators published SpCas9 gRNA sequences targeting HBV. The efficacy of the antiviral activity of these published gRNAs [29, 54, 61–64] was very consistent with our results. In particular, the most effective gRNA studied by Ramanan et al. [29], called Sp15 in our study, reduced both pgRNA and S-RNA transcription to 8% of mock control level. Among all published gRNA sequences for HBV targeting, Sp15 was the most effective. gRNAs Sp20 and Sp37 were first described in our study and demonstrated potent anti-HBV activity similar to Sp15. Another important parameter required for development of CRISPR/Cas9-based anti-HBV therapeutics is conservation of HBV target regions: Sp15 targets a fairly heterogenic HBV region; Sp4, Sp6, Sp20, Sp37 and Sp39 covered all major HBV genotypes (A–D) (Fig. 1a, heatmap) and effectively suppressed HBV transcription. Importantly, target regions for Sp6, Sp20 and Sp39 were very conservative in all HBV genotypes (A–H). Given low mismatch tolerance of SpCas9 and high risks of off-target effects, it is also necessary to take potential off-target sites into consideration. All the most promising SpCas9 gRNAs have off-target sites with at least three mismatched nucleotides (Table S4). Two gRNAs Sp4 and Sp20 demonstrated the most promising characteristics according to these three tested parameters. We also report that we could not reproduce the activity of two gRNAs used by Wang et al. [54], named Sp39 and Sp48 in our study.

A previous study [29] stated that the high divergence of HBV DNA sequences from human DNA opens up avenues for discovering HBV target regions disparate from the human genome. As outlined above, in silico analysis of SpCas9 gRNAs pointed to one to four potential off-target sites within human DNA that contained only one to two mismatches between gRNA, so it is highly likely to induce off-target mutagenesis even with the safest gRNAs (Sp4 and Sp20). Predicting off-target sites for CRISPR1/StCas9, on the other hand, revealed only one off-target region in the human genome for St3, that contained at least three mismatched nucleotides (Table S4). The rest of potential off-target sites for StCas9 system had a minimum of four mismatches. NmCas9 system had even better profile of predicted off-target sites: the number of off-targets was one for Nm5 and three for Nm6 with at least four mismatches. Unfortunately, NmCas9 did not demonstrate potent anti-HBV activity. This difference in the numbers of predicted off-target sites between the orthologous CRISPR/Cas9 systems is due to more restrictive PAM requirements for StCas9 and NmCas9 compared to SpCas9. It is obvious that longer PAM, such as that of StCas9, not only limits the number of potential off-target sites, but also significantly restricts the targeting range of orthologous CRISPR/Cas9 systems in the very short HBV genome. This is particularly important because multiplex gRNA targeting should improve HBV elimination and cccDNA degradation. Nevertheless, we successfully identified three gRNAs (St3, St4, and St10) potently targeting very conserved HBV regions, suggesting that the StCas9 system could be as effective at abolishing HBV as the SpCas9 system.

It is important to note that HBV cccDNA was preferentially degraded in our experimental setting. Previous experiments demonstrated that progressive decline in HBV intermediates is associated with accumulation of indel mutations in HBV cccDNA [23, 62]. However, experimental approaches in those studies were different from ours. In particular, Seeger et al. [62] reported generation of very frequent indel mutations when CRISPR/Cas9-transduced cells were then infected with HBV, i.e. this setting preferentially selects replication-competent mutant HBV strains. Our results demonstrated that treatment of cells with a small molecule inhibitor of non-homologous end-joining (NU7026) impaired degradation of cleaved cccDNA, leading to generation heavily mutated HBV cccDNA. We speculate that rapid degradation of HBV cccDNA in our study may result due to massive transcription of gRNAs from large amounts of encoding PCR products. Dramatically increased rates of gene editing by higher expression of gRNAs have been reported recently [65, 66]. Therefore, we assume that increased gRNA expression may result in degradation of HBV cccDNA instead of on-target mutagenesis. Another scenario is that preferential degradation of HBV cccDNA may be the result of intracellular immune response after transfection of double-stranded linear PCR products promoting decay of cleaved HBV cccDNA [67–69]. This observation requires further investigation and potentially can help to develop more efficacious approaches to eliminate HBV cccDNA from hepatocytes.

To consider the ability of the StCas9 system to target HBV quasispecies and to test its potential off-target activity, we generated numerous StCas9/gRNAs with nucleotide mismatches and tested their anti-HBV and nucleolytic activities (Fig. 6). Similar to previous reports, single-nucleotide mismatches did not significantly impair CRISPR/Cas9-mediated cleavage of a DNA target, suggesting that this system would effectively target many HBV variants. However, double- and triple-nucleotide mismatches nullified or drastically worsened the anti-HBV activity of CRISPR1/StCas9, while in some cases nucleolytic activity was still detected. Therefore, together with its more restrictive PAM, the low tolerance of S. thermophilus CRISPR1/Cas9 for mismatches compared to other systems makes it less likely to induce mutations in human DNA. At the same time, it raises the concern that some HBV quasispecies may escape nucleolytic cleavage by CRISPR/Cas9, and lead to incomplete cure and viral relapse after cessation of treatment, as evidenced for therapy with nucleot(s)ide analogs [70, 71]. Delivery of CRISPR/Cas9 systems with combinations of gRNAs targeting very conservative regions should solve this problem.

We showed by deep sequencing off-target regions and analyzing mismatch tolerance of StCas9 that two- or three-nucleotide mismatches significantly impair, and four mismatches preclude DNA cleavage. Sequencing of StCas9 off-target regions in the human genome revealed no off-target activity. In contrast, SpCas9 was shown to inflict frequent DNA damage in the host genome as evidenced by indel mutations detected at potential off-target regions. Taken together, these data suggest that StCas9 represents the safest and most effective orthologous CRISPR/Cas9 for targeting HBV to date.

Epigenetic modifications of HBV cccDNA add another level of complexity that might hamper effective CRISPR/Cas9 targeting [41, 72]. SpCas9 and StCas9 (CRISPR1 and CRISPR3 loci) demonstrate lower on-target activity in genes with restrictive chromatin architecture [20, 36, 37]. High-order chromatin architecture has been shown to render DNA less accessible to nucleases and to reduce insertion/deletion (indel) formation [38]. One of the crucial epigenetic modifications associated with epigenetic silencing is DNA methylation. HBV cccDNA methylation is frequently observed in chronically infected HBV patients [73–77], but cccDNA is not methylated in cell cultures [73]. Therefore, effects of cccDNA methylation on the anti-HBV activity of CRISPR/Cas9 in all published in vitro experiments have never been investigated. Rather, we analyzed CRISPR/Cas9 activity on highly methylated rcccDNA. We created hypermethylated rcccDNA in vitro and showed that transcription of methylated rcccDNA was several fold lower compared to unmethylated rcccDNA. Methylation of CpG worsened antiviral activity of several StCas9 gRNAs, although it had no significant impact on SpCas9 and St3 gRNA. Nevertheless, HBV rcccDNA levels were still significantly reduced in methylated and unmethylated states. Noteworthy, cleavage and degradation of rcccDNA appeared to be affected by methylation outside or inside of CpG islands (St4 and St10 gRNA). We assume that in case of St4, additional factors besides DNA methylation itself could contribute to decreased anti-HBV activity, like association of the rcccDNA with histones or a peculiar rcccDNA conformation. To conclude, potential anti-cccDNA agents must be tested with methylated rcccDNA templates to develop more efficacious therapeutics, as evidenced by the lower ability of St4 and St10 to target methylated rcccDNA. Overall, we provide the first examination of the effects of rcccDNA methylation on CRISPR/Cas9 targeting of HBV.

In conclusion, based on a comprehensive analysis of several promising CRISPR/Cas9 systems, our results establish CRISPR1 StCas9 system as an ideal candidate for development of HBV cure. This system can effectively target three highly conservative regions in the HBV genome, results in degradation of HBV cccDNA, has very limited tolerance to nucleotide mismatches while still targeting single-nucleotide HBV genetic variants, has few predicted off-target sites in the human genome, and does not exhibit off-target nucleolytic activity.

Materials and methods

Cell culture and transfection

The human HepG2 cell line was cultured in complete DMEM high-glucose medium with 10% FBS, 2 µM l-glutamine, and 1% penicillin/streptomycin. HepG2 cells were transfected using 7.5 mM polyethylenimine. Briefly, DNA mix containing 0.625 µg 1.1-merHBV-encoding plasmid (kindly provided by Dieter Glebe, University of Giessen) genotype D or 100 ng of synthesized HBV cccDNA, 0.625 µg of Cas9-encoding plasmid (SpCas9-EGFP, NmCas9, StCas9 or FnCas9), and 10 ng PCR product transcribed into a specific gRNA were gently added to 35 µL of NaCl solution. In parallel, a solution of polyethylenimine (5.7 µL/well) in NaCl (29.8 µL/well) was prepared and left to rest for 10 min. The two solutions were then gently mixed and incubated at room temperature for 10 min before being added to the cell culture medium. After 24 h, the cell culture medium was discarded, and the cells were washed twice in PBS and cultured in new complete medium for the next 72 h. Alternatively, HepG2 cells were transfected using Lipofectamine 2000 according to manufacturer’s protocol (Thermo Fisher Scientific). All results were reproduced at least five times. To prevent degradation of HBV cccDNA, HepG2-1.1merHBV cells were treated by small molecule NU7026 (7.5 μM) for 72 h. DMSO was used added to the mock control group.

Synthesis and methylation of recombinant HBV cccDNA

1.0-mer linear HBV genome was released from the HBV-encoding plasmid pCH-HBV by incubating with PsiI and PstI restriction enzymes in Y buffer (Sibenzyme) overnight. The HBV genome (3182 bp) was gel purified by electrophoresis and isolated using the Qiagen gel extraction kit. The linear HBV genome was religated using T4 DNA ligase (Thermo Fisher Scientific) overnight at ambient temperature, and total DNA was further isolated by isopropanol precipitation. Next, DNA ligation products were gel electrophoresed, and fragments of circular full-length HBV genome were excised and isolated using the Qiagen gel extraction kit. Alternatively, rcccDNA was generated using minicircle technology (System Biosciences). Isolated rcccDNA was methylated by M.SssI DNA methyltransferase (Thermo Fisher Scientific) according to manufacturer’s protocol. Briefly, 1 mg of DNA was incubated with the enzyme at 37 °C for 30 min and then purified by Qiagen PCR purification kit. Methylated and unmethylated rcccDNAs were incubated with a frequent-cutter methylation-sensitive Hpa II restriction enzyme (New England Biolabs). Cut and uncut rcccDNA were separated using gel electrophoresis and visualized by GelDoc XR+ (BioRad). All primers are listed in Table S6.

Nucleofection of CRISPR/Cas9

Nucleofection of stable HepG2-1.1mer/1.5merHBV cell lines (kindly provided by Dieter Glebe, University of Giessen) was performed according to manufacturer’s instructions. In brief, 1 million cells were harvested by trypsin/EDTA solution, centrifuged, washed in PBS, and resuspended with the DNA mix in SF Nucleofector solution and Supplement 1 containing 200 ng of PCR products, 4 µg of Cas9-encoding plasmid, and 800 ng of pMAX-GFP. Nucleofection was performed by LONZA Nucleofector with standard HepG2 cell-specific parameters. Cells were seeded onto six-well plates with complete medium without doxycycline (for HepG2-1.5merHBV) or with 100 ng/mL doxycycline (for HepG2-1.1merHBV). After 24 h, medium was discarded, cells were washed twice in PBS, and complete medium with blasticidin (15 ng/mL) was added for the following 120 h. All results were reproduced at least three times.

CRISPR/Cas9 constructs

HBV genome targets and potential off-target sites were assessed and gRNA was designed using the open-access web tool Broad Institute Genetic Perturbation Platform and CCTop CRISPR/Cas9 target online calculator. Conservation of gRNA sequences across HBV genotypes was assessed in Geneious using available full-genome sequences of HBV genotypes from GenBank. PCR products containing the U6 promoter and sgRNA specific for every type of Cas9 proteins were synthesized by two-step mutagenic PCR using Q5 polymerase and purified using Qiagen gel extraction kit. Multiplex CRISPR/Cas9 targeting was performed similarly with 10 ng of each PCR product for PEI-transfection of 12-well plates. Mutant gRNAs with mismatched nucleotides were generated accordingly with mutant primers at indicated positions. Results with mutant gRNAs were reproduced at least three times. The following Cas9-encoding plasmids were used: M-NMcas (AddGeneplasmid #48670), M-NM-sgRNA (AddGene plasmid #48673), M-ST1cas9 (AddGene plasmid #48669), and M-ST1-sgRNA (AddGene plasmid #48672) were gifts from George Church; Lenti-Cas9-2A-Blast (AddGene plasmid #73310) was a gift from Jason Moffat; pLX-sgRNA (AddGene plasmid #60662) was a gift from Eric Lander and David Sabatini; and PX408 Francisella tularensis subsp. Novicida Cas9 (AddGene plasmid #68705) was a gift from Feng Zhang. Lenti-ST1cas9-2A-Blast was generated by cloning ST1cas9 into AddGene plasmid #73310. All primers and probes are listed in Table S6.

Isolation of nucleic acids

At harvest, culture medium was discarded, and cells were washed twice with PBS and lysed in AmpliSens Riboprep lysis buffer (CRIE). Nucleic acids were isolated using the AmpliSense Riboprep kit (CRIE) according to manufacturer’s instructions. For isolation of RNAs, nucleic acids were treated with RNase-free DNase I (NEB) for 30 min at 37 °C, purified using the AmpliSense Riboprep kit, and reverse transcribed using AmpliSens Reverta-FL (CRIE). HBV cccDNA was isolated following the Hirt procedure as described by Cai et al. [78] followed by treatment with plasmid-safe ATP-dependent DNase (Epicentre) for 12 h at 37 °C and inactivation of enzyme at 72 °C for 15 min.

PCR analysis

In HepG2 transfection experiments, expression of HBV pgRNA and S-RNA was normalized to corresponding Cas9 mRNA expression levels. In nucleofection experiments, HBV pgRNA and mRNA of HBsAg (S-RNA) were measured relative to GAPDH mRNA. Total intracellular HBV DNA concentrations and cccDNA levels were normalized to genomic β-globin. All PCRs were performed with specific sets of primers and probes (Table S6). Relative expression levels were calculated by ΔΔCt method.

Immunofluorescence

Cells were seeded on glass coverslips, transfected using PEI, and fixed in 4% paraformaldehyde for 10 min before harvesting. Next, coverslips were washed three times in Tris–HCl (50 mM, pH 8.0) and incubated for 30 min with blocking buffer [0.02% Triton X-100, 10% horse serum, and 150 mM NaCl in Tris–HCl (50 mM, pH 8.0)]. The coverslips were then incubated with primary rabbit anti-HBc antibodies (ab115192) and mouse anti-Cas9 antibodies (ab191468) at room temperature for 1 h, washed three times for 5 min in washing buffer [0.02% Triton X-100 and 200 mM NaCl in Tris–HCl (50 mM, pH 8.0)], and incubated with secondary Alexa Fluor 488 goat anti-rabbit IgG antibodies and Alexa Fluor 594 anti-mouse IgG H&L antibodies with Hoechst33342 (1:10,000) at room temperature for 1 h. Coverslips were washed three times for 5 min in washing buffer and mounted with Fluoroshield reagent (ab104135). Images were captured on a Leica DMI6000 microscope at 20 × immersion objective.

Fluorescent microscopy and FACS analysis

Transfection and nucleofection efficiencies were analyzed by fluorescent imaging of EGFP-expressing cells in the FITC channel on an Olympus IX73 microscope. To determine transfection and nucleofection efficiency by FACS, cells were washed twice in PBS and harvested in trypsin/EDTA solution. Detached cells were resuspended in complete medium and centrifuged at 500×g for 5 min. Supernatant was discarded, and the cell pellet was resuspended in PBS and centrifuged again as above. The resultant cell pellet was resuspended in 300 μL of PBS and used for FACS analysis with FITC channel. FACS analysis was performed on Novocyte3000 (ACEA Biosciences, Inc). The results were analyzed in Novocyte software.

NGS sequencing

For analysis of off-target activity, experimental groups from two different wells were used for isolation of genomic DNA. Four experimental groups (St3, St4, Sp20 and Sp9) with corresponding mock-treated controls were used for PCR amplification of the top predicted off-targets. On-target and potential off-target regions were amplified with pairs of specific primers using Q5 polymerase; amplicons were gel purified and extracted using Qiagen gel extraction kit, quantified with a Qubit 2.0 Fluorometer (Life Technologies), and pooled in equimolar ratios. Adapters for Illumina sequencing were then attached. Libraries were sequenced with paired-end 250 reads using MiSeq instrument (Illumina). FASTQC software and Geneious software were used for quality assessment, reference alignment, discarding low-quality reads and nucleotides, and calculating indels. To determine if each off-target indel frequency is significant compared to mock-treated control samples, two-tailed P value was calculated using Fisher’s exact test.

Statistical analysis

Values were expressed as the mean ± standard deviation (SD) of triplicate experiments in SPSS software (SPSS 21.0.0.0). HepG2 transfection results were reproduced in at least five independent experiments. Results in stable cell lines and mismatch tolerance data were reproduced at least three times. One-way ANOVA and student’s T test with Tukey’s HSD post hoc test were used to compare variables and calculate P values to determine statistically significant differences in means.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

DK, SB, and AK conducted all experiments; DZ and SB generated recombinant cccDNA and created gRNAs; DZ analyzed off-target sites and designed specific primers; IG conducted sequencing; DK conceived the project; AK helped conceive experiments with mutant gRNAs; DK, DZ, SB, and AK processed the data; DK wrote the manuscript; VC guided the study and revised the manuscript.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no potential conflicts of interests. The authors have applied for patents concerning the use of Cas9 proteins and gRNAs for HBV therapy.

Funding

This work was supported by RSF Grant no. 16-15-10426.