Toll-like receptor dual-acting agonists are potent inducers of PBMC-produced cytokines that inhibit hepatitis B virus production in primary human hepatocytes

Recombinant interferon-α (IFN-α) treatment functionally cures chronic hepatitis B virus (HBV) infection in some individuals and suppresses virus replication in hepatocytes infected in vitro. We studied the antiviral effect of conditioned media (CM) from peripheral blood mononuclear cells (PBMCs) stimulated with agonists of Toll-like receptors (TLRs) 2, 7, 8 and 9. We found that CM from PBMCs stimulated with dual-acting TLR7/8 (R848) and TLR2/7 (CL413) agonists were more potent drivers of inhibition of HBe and HBs antigen secretion from HBV-infected primary human hepatocytes (PHH) than CM from PBMCs stimulated with single-acting TLR7 (CL264) or TLR9 (CpG-B) agonists. Inhibition of HBV in PHH did not correlate with the quantity of PBMC-produced IFN-α, but it was a complex function of multiple secreted cytokines. More importantly, we found that the CM that efficiently inhibited HBV production in freshly isolated PHH via various cytokine repertoires and mechanisms did not reduce covalently closed circular (ccc)DNA levels. We confirmed our data with a cell culture model based on HepG2-NTCP cells and the plasmacytoid dendritic cell line GEN2.2. Collectively, our data show the importance of dual-acting TLR agonists inducing broad cytokine repertoires. The development of poly-specific TLR agonists provides novel opportunities towards functional HBV cure.

Coculturing with stimulated PBMCs inhibits HBV production from PHH. To test whether continuous production of cytokines from TLR2/7, TLR7, or TLR9 agonist-stimulated PBMCs inhibits the production of HBeAg from HBV-infected PHH more strongly than two-times addition of CM to infected cells, we cocultured TLR agonist-stimulated PBMCs with HBV-infected PHH in the Transwell system from 6-9 DPI (Fig. 4A). We found that inhibition of HBeAg in HBV-infected PHH by coculture with TLR2/7 (CL413), TLR7 (GS-9620[L]) or TLR9 (CpG-A) agonist-stimulated PBMCs did not significantly differ from inhibition following two-times addition of CM (Fig. 4B). In any case, inhibition did not exceed 70%, and CL413 was a more potent inducer of an antiviral response than CpG-A or GS-9620[L]. Previous study found that several TLR agonists can inhibit HBV replication both directly via TLR activation in PHH and indirectly via exposure to CM of stimulated innate immune cells 23 . Thus, we tested whether the TLR2/7 dual agonist CL413 can inhibit HBV replication without the indirect effect of PBMC-secreted cytokines (Fig. 4B). However, in the absence of PBMCs, CL413 did not show any antiviral activity, although it induced production of proinflammatory cytokines IL-6 (275 pg/ ml), TNF-α (84 pg/ml) and chemokine IL-8 (987 pg/ml) in HBV-infected PHH. As in the case of CM addition to HBV-infected cells, coculture of TLR agonist-stimulated PBMCs with HBV-infected PHH in the Transwell system did not result in degradation of cccDNA (data not shown). www.nature.com/scientificreports/

Anti-IFN-I receptor monoclonal antibody (IFNAR mAb) abrogates CpG-A-CM or GS-9620[L]-CMinduced inhibition of HBeAg production from HBV-infected PHH.
Subsequently, we investigated the mechanism of pDC-induced inhibition of HBeAg production in HBV-infected PHH. Due to the sensitivity of HBeAg production in HBV-infected PHH to IFN-α and IFN-λ, we examined the proportion of the inhibitory effect mediated by type I and III IFN receptors, IFNAR and IFNLR (Fig. 5A). To do so, we pretreated HBV-infected PHH with mAbs targeting IFNAR and IFNLR and determined the level of HBeAg produced upon exposure of HBV-infected PHH to CpG-A-CM or GS-9620[L]-CM (Fig. 5B). While IFNAR mAb completely abrogated the inhibitory effect of GS-9620-CM, it abrogated by only 40% the inhibitory effect of CpG-A-CM. Simultaneous blockade of IFNAR and IFNLR did not significantly increase abrogation of the inhibitory effect on HBeAg production. Production of HBeAg was also significantly inhibited by 1,000 IU of recombinant IFN-α2a (by 61.3%, p = 0.04) and IFN-λ3 (by 56.2%, p = 0.04).
Stimulated GEN2.2 pDCs show antiviral activity against HBV-infected HepG2-NTCP cells. We next compared the antiviral effect of CM from stimulated PBMCs on virus production in PHH ( Fig. 2) with that in a model comprising GEN2.2 pDCs and HBV-infected HepG2-NTCP hepatocytes 26,38,39 (Fig. 6A). To facilitate 4 days lasting coculture, which is still difficult to perform in rare and in vitro short living human primary pDCs, we performed our studies in human pDC line GEN2.2, which shares many features with human primary pDCs 26,39,40 . Production of HBeAg from HBV-infected HepG2-NTCP hepatocytes was significantly inhibited by exposure to CM from GEN2.2 cells stimulated with CpG-A (by 67%), CpG-B (by 59%), GS9620[L] (by 55%) and GS-9620[H] (by 53%) (Fig. 6B). No significant differences in inhibition of HBeAg production were observed when the antiviral effect of CM from stimulated GEN2.2 cells was compared with direct coculture of GEN2.2 and HepG2-NTCP cells (Fig. 6C). Despite the different repertoires and levels of cytokines induced by CpG-A

Discussion
In this study, we investigated the antiviral effect of CM from PBMCs stimulated with a set of agonists of endosome-localized TLRs. Our results demonstrate that synthetic TLR agonists capable of activating more than one TLR induce a broader proinflammatory cytokine spectrum and are more efficient drivers of HBV inhibition than single TLR-targeting agonists. The dual-acting TLR agonists R848 41 , CL413 35 and GS-9620[H] 36 were the bestscoring inducers of HBV inhibition. Statistical significance was a major issue in these experiments, which were performed in fresh PHH from 3 donors, each tested in 5 biological replicates. None of the selected TLR ligands reduced the level of cccDNA in HBV-infected cells. Previous findings revealed that R848 can be independently recognized by both human TLR7 and TLR8, although TLR8 is induced more efficiently than TLR7 at higher R848 concentrations 41 . A low concentration of GS-9620[L] has approximately 30-fold higher selectivity for activation of TLR7 over TLR8, with no detectable activity on other human TLRs 36 . However, GS-9620 elicits combined TLR7 and TLR8 stimulation at higher concentrations 36 . As TLR7 expression is largely restricted to pDCs in PBMC subsets [26][27][28][29][30]42 and TLR8 is predominantly expressed in myeloid DCs and monocytes 43,44 , we surmise that low concentrations of GS-9620[L] (50 nM) preferentially mediate secretion of IFN-α from pDCs, whereas high concentrations of GS-9620[H] (10 µM) preferentially stimulate secretion of inflammatory cytokines in classical myeloid DCs and CD14+ monocytes. Another dual-acting agonist, CL413, also induces both IFN-I and proinflammatory cytokines. We tested CL413 activity to analyze both a direct effect on TLR2 stimulation in PHH and an indirect effect on TLR7 stimulation in PBMCs. CL413 did not elicit a direct inhibitory effect on HBV replication in TLR2-expressing PHH. Therefore, its antiviral effect likely is conferred indirectly by triggering TLR2 and TLR7 in PBMCs. Expression of cytoplasmic TLR2 and endosome-localized TLR7 in CD14+ monocytes permits both signaling pathways to be triggered at the single-cell level 44 . In contrast, in pDCs in which TLR2 is not expressed, only TLR7 signaling can be activated by CL413. Recently, the dual-acting agonist Riboxxol, which triggers TLR2/3-mediated signaling via the IFN or NF-κB pathways, was shown to efficiently and directly suppress HBV replication in HBV-infected PHH 23 . In contrast to dual-acting agonists, single-acting ligands of TLR7 (CL264) or TLR9 (CpG-B) induced in PBMCs only moderate levels of proinflammatory cytokines with no detectable IFN-I, and CL264-CM and CpG-B-CM were associated with poor HBeAg inhibition. Robust production of proinflammatory cytokines induced in PBMCs by dual-acting agonists was associated with elevated phosphorylation of p65 NF-ĸB.
To induce a large and variable range of cytokines, we stimulated PBMCs with a larger spectrum of agonists than those used in previous studies 23, 24 . Our results indicate that inhibition of HBeAg and HBsAg production HBV-infected PHH were exposed at 3 DPI to 5 µg/ml of IFNAR mAb, IFNLR mAb or control isotype mAb. Recombinant IFN-α-2a or IFN-λ3 (1,000 IU/ml) was used as a control. Production of HBeAg was determined by ELISA 6 DPI. (B) Production of HBeAg was normalized to production by HBV-infected PHH treated with CM from unstimulated PBMCs determined by ELISA 6 DPI. The data are shown as mean ± SEM from three independent experiments with PHH from one donor (N = 1). *p < 0.05, Mann-Whitney-Wilcoxon pairwise test, p value adjusted by BH method. www.nature.com/scientificreports/ in HBV-infected PHH does not correlate with the quantity of PBMC-secreted IFN-α, but rather is a complex function of multiple secreted cytokines. We found that CpG-A-activated PBMCs produced more IFN-I and IFN-III than those stimulated with GS-9620[L], which has been tested extensively in previous HBV inhibitionrelated studies [20][21][22][23][24] . In agreement with the previous finding that IFN-I is the major component of PBMC CM responsible for inhibition of HBV production in PHH, CpG-A-CM inhibited HBeAg and HBsAg secretion more efficiently than GS-9620[L]. However, a higher concentration of GS-9620[H] (10 µM), which induced in PBMCs a broader spectrum of proinflammatory cytokines but a lower quantity of IFN-I than induced by GS-9620[L], was associated with significantly higher HBeAg and HBsAg inhibition. R848, which also induced a very broad spectrum of proinflammatory cytokines, had a similar antiviral effect as GS-9620[H]. Importance of the cytokine complexity in inhibition of HBV production is further highlighted by relatively low inhibitory activity (50 to 60%) of recombinant IFN-α-2a and IFN-λ3. Based on the blockade of IFNAR by mAb, Lucifora et al. 23 concluded that the antiviral effect of TLR1/2 and TLR3 activation in PHH was not due to type-I IFN and IL-6 production. Correlation analysis showed that in addition to IFN-α, the proinflammatory cytokines IL-6, TNF-α and IL-12 and the chemokines IL-8 and IFN-γ are major contributors to anti-HBV inhibitory activity. A statistical model that could decipher the specific combination of cytokines necessary to inhibit HBV production from infected hepatocytes would require additional measurements of the effect of CM or artificial permutations of recombinant cytokines. Although R848-CM, CL-413-CM, CpG-A-CM and GS-9620[H]-CM achieved two to fourfold greater inhibition of HBeAg than that observed with GS-9620[L]-CM in previous studies 23,24 , none resulted in reduction of cccDNA levels in freshly isolated PHH. This is compatible with recent findings showing that GS-9620[L]-CM strongly induces various IFN-stimulated genes and inhibits virus production in HBV-infected PHH without inducing APOBEC3A or the Smc5/6 complex-and without reducing cccDNA levels 24 . The importance of cccDNA and its degradation for HBV cure positions this molecule in the center of HBV research. A 2012 study reported that IFN-α inhibits cccDNA transcription by hypoacetylation of cccDNA-bound histones and reduces binding of the STAT1 and STAT2 transcription factors to the IFN-stimulated response element present in the www.nature.com/scientificreports/ HBV genome 8 . More recent studies have shown that cccDNA can be degraded in HBV-infected hepatocytes in a noncytopathic fashion during IFN-α treatment 9,10 . Production of HBeAg from the HBV-infected hepatoma cell line HepG2-NTCP was three-to fivefold less sensitive to IFN-α compared to production from HBV-infected PHH. In the presence of CM or recombinant IFN-I, residual production of HBeAg in HepG2-NTCP cells was not suppressed below 35%. The insignificant differences in inhibition of HBV production by exposure of HBV-infected HepG2-NTCP cells directly to activated GEN2.2 cells or to their CM indicate that soluble factors, and not cell-to-cell contact during coculture, plays a major role in the regulation of HBV production.
We also addressed whether IFN-I and IFN-III present in CpG-A-CM cooperate in HBV inhibition, which could explain the greater inhibitory effect of CpG-A-CM compared to GS-9620[L]-CM. Surprisingly, IFNLR targeting had no effect on HBeAg secretion, and only IFNAR mAb partially abrogated the inhibitory effect of CpG-A-CM. This inhibitory effect was not completely reverted by targeting both IFNAR and IFNLR, likely due to inefficient inhibition of IFNAR. Thus, we cannot conclude whether IFN-I is the main inhibitory driver in CpG-A-CM or if other cytokines contribute as well. Further study will be necessary to elucidate whether IFN-I signaling dominates over IFN-III signaling in PHH.
Specific and prolonged suppression of chronic hepatitis B in chimpanzee and woodchuck models by endosomal TLR7, 8, and 9 agonists led to an interest in discerning the mechanisms by which these TLR ligands elicit antiviral responses [20][21][22] . However, clinical studies with GS-9620, which showed the best antiviral effect in animal models, did not reveal a clinically significant decline of HBsAg at tolerable doses (two one-weekly doses 4 mg) in patients with chronic hepatitis B 45,46 . This dose corresponds to concentration of 4.6 ng/ml in plasma, while 50 ng/ml of GS-9620[L] were used in our experiments 47 . The dose of CpG ODNs commonly used in clinical trials was 1.5-15 μg/kg and the schedule of ODN administration ranged from weekly to monthly. In preclinical studies, much higher doses of CpG ODN (2.5 mg/kg) were administered daily to mice while 4 μg/ml of CpG-A or CpG-B were used in our experiments 48 . Also the dose of CL413 agonist (5 μg/ml) was within the range of the dose administered to mice in preclinical studies (3 μg/g) 49 . Further studies of the distinct antiviral potential of polyspecific TLR agonists are necessary to elucidate their functions, which could provide new opportunities for the development of novel strategies to achieve sustained viral clearance and provide a definitive cure for hepatitis B.  www.nature.com/scientificreports/ Detection of HBsAg and HBeAg secretion by ELISA. Cell-free supernatants from HBV-infected HepG2-NTCP cells or PHHs were collected and centrifuged at 300 × g for 5 min to remove cellular debris, transferred into clean tubes and stored at − 80 °C until antigen measurement. The titers of HBsAg and HBeAg were measured using a commercial ELISA kit (Bioneovan, Beijing, China) according to the manufacturer's instructions.
Total HBV DNA and cccDNA quantification. Total  . cccDNA quantification was performed as previously described 50 . Briefly, 1 µg of DNA was treated with 10 units of T5 exonuclease for 2 h. Then, DNA was purified using a DNA Clean and Concentrator Kit (Zymo Research). qPCR was performed with gb Elite PCR Master Mix (Generi Biotech) and specific cccDNA primers and probe. The level of cccDNA was normalized to mitochondrial-encoded cytochrome-c oxidase subunit II (MT-CO2) expression in samples without T5 exonuclease digestion. cccDNA-specific primers and MC-CO2-specific primers were used as previously described 24 . ddPCR was performed using a QX200 Digital PCR Generator and QX200 Droplet Reader (both Biorad) with cccDNA-specific primers and probe. T5 treatment and cccDNA primer specificity was confirmed by comparing the T5-treated and non-treated samples and by comparing cccDNA specific primers with total HBV DNA primers (non-specific cccDNA primers).

Statistical analysis.
Quantitative variables are expressed as means ± standard error of the mean (SEM).
Non-parametric tests were performed due to the nature of the data. First, Kruskal-Wallis was applied, followed by non-parametric post hoc pairwise multiple comparison Mann-Whitney-Wilcoxon test with p value adjustment by Benjamini Hochberg method (BH). In case of data depicted in Fig. 3  www.nature.com/scientificreports/