Viral Load Affects the Immune Response to HBV in Mice With Humanized Immune System and Liver

Background & Aims Hepatitis B virus (HBV) infects hepatocytes, but the mechanisms of the immune response against the virus and how it affects disease progression are unclear. Methods We performed studies with BALB/c Rag2–/–Il2rg–/–SirpaNODAlb-uPAtg/tg mice, stably engrafted with human hepatocytes (HUHEP) with or without a human immune system (HIS). HUHEP and HIS-HUHEP mice were given an intraperitoneal injection of HBV. Mononuclear cells were isolated from spleen and liver for analysis by flow cytometry. Liver was analyzed by immunohistochemistry and mRNA levels were measured by quantitative reverse transcription polymerase chain reaction (PCR). Plasma levels of HBV DNA were quantified by PCR reaction, and antigen-specific antibodies were detected by immunocytochemistry of HBV-transfected BHK-21 cells. Results Following HBV infection, a complete viral life cycle, with production of HBV DNA, hepatitis B e (HBe), core (HBc) and surface (HBs) antigens, and covalently closed circular DNA, was observed in HUHEP and HIS-HUHEP mice. HBV replicated unrestricted in HUHEP mice resulting in high viral titers without pathologic effects. In contrast, HBV-infected HIS-HUHEP mice developed chronic hepatitis with 10-fold lower titers and antigen-specific IgGs, (anti-HBs, anti-HBc), consistent with partial immune control. HBV-infected HIS-HUHEP livers contained infiltrating Kupffer cells, mature activated natural killer cells (CD69+), and PD-1+ effector memory T cells (CD45RO+). Reducing the viral inoculum resulted in more efficient immune control. Plasma from HBV-infected HIS-HUHEP mice had increased levels of inflammatory and immune-suppressive cytokines (C-X-C motif chemokine ligand 10 and interleukin 10), which correlated with populations of intrahepatic CD4+ T cells (CD45RO+PD-1+). Mice with high levels of viremia had HBV-infected liver progenitor cells. Giving the mice the nucleoside analogue entecavir reduced viral loads and decreased liver inflammation. Conclusion In HIS-HUHEP mice, HBV infection completes a full life cycle and recapitulates some of the immunopathology observed in patients with chronic infection. Inoculation with different viral loads led to different immune responses and levels of virus control. We found HBV to infect liver progenitor cells, which could be involved in hepatocellular carcinogenesis. This is an important new system to study anti-HBV immune responses and screen for combination therapies against hepatotropic viruses.

See Covering the Cover synopsis on page 1459.
BACKGROUND & AIMS: Hepatitis B virus (HBV) infects hepatocytes, but the mechanisms of the immune response against the virus and how it affects disease progression are unclear. METHODS: We performed studies with BALB/c Rag2 -/-Il2rg -/-Sirpa NOD Alb-uPA tg/tg mice, stably engrafted with human hepatocytes (HUHEP) with or without a human immune system (HIS). HUHEP and HIS-HUHEP mice were given an intraperitoneal injection of HBV. Mononuclear cells were isolated from spleen and liver for analysis by flow cytometry. Liver was analyzed by immunohistochemistry and mRNA levels were measured by quantitative reverse transcription polymerase chain reaction (PCR). Plasma levels of HBV DNA were quantified by PCR reaction, and antigenspecific antibodies were detected by immunocytochemistry of HBV-transfected BHK-21 cells. RESULTS: Following HBV infection, a complete viral life cycle, with production of HBV DNA, hepatitis B e (HBe), core (HBc) and surface (HBs) antigens, and covalently closed circular DNA, was observed in HUHEP and HIS-HUHEP mice. HBV replicated unrestricted in HUHEP mice resulting in high viral titers without pathologic effects. In contrast, HBV-infected HIS-HUHEP mice developed chronic hepatitis with 10-fold lower titers and antigenspecific IgGs, (anti-HBs, anti-HBc), consistent with partial immune control. HBV-infected HIS-HUHEP livers contained infiltrating Kupffer cells, mature activated natural killer cells (CD69þ), and PD-1þ effector memory T cells (CD45ROþ).
Reducing the viral inoculum resulted in more efficient immune control. Plasma from HBV-infected HIS-HUHEP mice had increased levels of inflammatory and immunesuppressive cytokines (C-X-C motif chemokine ligand 10 and interleukin 10), which correlated with populations of intrahepatic CD4þ T cells (CD45ROþPD-1þ). Mice with high levels of viremia had HBV-infected liver progenitor cells. Giving the mice the nucleoside analogue entecavir reduced viral loads and decreased liver inflammation. CONCLUSION: In HIS-HUHEP mice, HBV infection completes a full life cycle and recapitulates some of the immunopathology observed in patients with chronic infection. Inoculation with different viral loads led to different immune responses and levels of virus control. We found HBV to infect liver progenitor cells, which could be involved in hepatocellular carcinogenesis. This is an important new system to study anti-HBV immune responses and screen for combination therapies against hepatotropic viruses. W ith over 350 million people infected by hepatitis B virus (HBV) and despite an efficient prophylactic vaccine, HBV prevalence is on the rise, constituting a major global health burden. The sequelae of HBV infection include acute and chronic hepatitis, which may lead to the development of fibrosis, cirrhosis, and hepatocellular carcinomas (HCC). The clinical course depends in part on the age of the host: most neonatally acquired infections develop into chronicity, whereas in adults they are mostly self-resolving. 1 Current treatment strategies lead to viral suppression, but rarely establish a "functional cure" with HBsAg (hepatitis B surface antigen) loss. Furthermore, these life-long therapies do not eradicate the virus, nor do they efficiently train the immune response to control viremia, since viral rebound is frequent following discontinuation of treatment. 2,3 HBV is not directly cytopathic for hepatocytes, disease pathophysiology is conditioned by the ensuing anti-viral immune response. 4 The breadth and scope of this response is essential for viral eradication; therefore, understanding the mechanisms that lead to viral clearance or persistence represents an essential goal.
Viral clearance is the result of a coordinated immune response initially mediated by Kupffer, natural killer (NK), and antigen-presenting cells, leading to robust and polyclonal CD4 þ and CD8 þ T-cell responses and the development of neutralizing anti-HBs antibodies that ensures protective immunity to functionally cured patients, despite the maintenance of covalently closed circular DNA (cccDNA). 4,5 In contrast, chronic viremia is associated with increased immunosuppressive cytokines (TGF-beta, IL-10) and regulatory T cells, combined with functional exhaustion of virusspecific T cells expressing programmed death-1 (PD-1), cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) and tumor necrosis factor (TNF)-related apoptosis-inducing ligand death receptor (TRAIL-R2), targeting them for NK cell-mediated deletion. 4,[6][7][8] In chronic HBV (CHB) patients, an increased risk of progression to cirrhosis and HCC has been correlated with the viral load; however, the mechanisms driving these virally induced pathologies are unclear. [9][10][11] Deciphering the cross-talk between the immune system and HBV-infected hepatocytes has been hampered by the viruses' restricted tropism to humans and chimpanzees. Although HBV transgenic or transduced mice, and HBVinfected human liver chimeric mice, are valuable for exploring the viral life cycle and testing antiviral compounds, they have limited immune responses and cannot assay immunotherapeutic approaches because significant differences between mice and humans predominate. 12,13 To establish an immunocompetent animal model for hepatotropic infections, several mouse models harboring both a humanized immune system and human hepatocytes have been recently described. [14][15][16][17][18][19] Although HBV infection of these models resulted in immune responses, low levels of liver chimerism limited viral output and the analysis of chronic pathophysiologic responses. 15,19 Previously, we established a dually humanized mouse model in BALB/c Rag2 -/-Il2rg -/-Sirpa NOD Alb-uPA tg/tg recipients, stably engrafted with a humanized immune system and human hepatocytes (HIS-HUHEP mice). 18 In this study, we demonstrate that HIS-HUHEP mice are susceptible to HBV, developing a full viral life cycle (HBeAgþ, HBsAgþ, cccDNAþ) similar to CHB patients. By varying the viral inoculum in HIS-HUHEP mice, we found that distinct immune responses were obtained that led to different levels of viral control. Subsequent biomarker analysis revealed a cluster of effectors that may be useful to further dissect the complexity of HBV immune control. Nucleos(t)ide analog (NUC) treatment of chronically infected mice efficiently reduced viral loads, resulting in the resolution of liver inflammation.

Materials and Methods
Generation of Humanized Mice, HBV Infection, and NUC Treatment HUHEP and HIS-HUHEP mice were established in BALB/c Rag2 -/-Il2rg -/-Sirpa NOD uPA tg/tg male and female mice as previously described. 18 Mice with >100 mg/mL human Albumin (hAlbumin) and >10% hCD45þ cells in peripheral blood mononuclear cells (PBMC), including human T and B cells, were HBV infected intraperitoneally at 15 (±3) weeks with either 1x10 7 or 1x10 9 HBV genome equivalents (GE) purified from concentrated supernatants of the HepG2.2.15 stably producing cell line (HBV genotype D subtype ayw). 20,21 NUC-treated mice had been inoculated with 10 7 HBV GE for 14 ±1 weeks prior to receiving Entecavir at 0.3 mg/kg/day either injected intraperitoneally or dispensed in the drinking water (Baraclude; Bristol-Myers Squibb, Princeton, NJ). Animals were housed in isolators under pathogen-free conditions with humane care. Experiments were approved by an institutional ethical committee at the Institut Pasteur (Paris, France) and validated by the French Ministry of Education and Research (MENESR #02162.02).

Enzyme-Linked Immunosorbent Assay (ELISA) Analysis
Species-specific enzyme-linked immunosorbent assays (ELISA) of plasma samples for human albumin, human IgM, and human IgG were performed as previously described. 18 M65 ELISA was performed according to manufacturer's specifications (M65 EpiDeath ELISA; Peviva, Nacka, Sweden) on plasma samples pre-and post-HBV infection, or the equivalent time points for controls. M65 levels were normalized to hAlbumin levels for each time point. Human cytokines and chemokines were quantified in mouse plasma with the Human Cytokine Magnetic 25-plex Panel (catalog # LHC0009M; Life Technologies, Carlsbad, CA) according to manufacturer's specifications and analyzed with MAGPIX (Luminex, Austin, TX).

BACKGROUND AND CONTEXT
Although HBV infects hepatocytes, evasion and modulation of the immune response is key to the subsequent physiopathology and the mechanisms of disease progression are unclear.

NEW FINDINGS
Chronic HBV immunopathology is modeled in a dually humanized mouse developing a complete viral life cycle and immune responses proportionally to the dose of viral inoculum.

LIMITATIONS
HBV-infected mice do not develop fibrosis.
IMPACT Liver-specific immune responses are investigated during chronic HBV and innovative therapeutic approaches can be tested in this small animal model. dye (eBioscience). Intracellular labeling was performed as previously described 18 (see Supplementary Materials and Methods). Acquisitions were performed using BD Fortessa and LSRII flow cytometers (Becton Dickinson, Franklin Lakes, NJ). Analysis was performed with FlowJo Version 8.8 (TreeStar, Ashland, OR) and Prism 6 (GraphPad, La Jolla, CA).

Immunofluorescence Analysis
Liver cryostat sections were prepared as previously described 22

Antigen-Specific Antibody Responses
Clinical assays for HBsAb and HBcAb have detection thresholds well above the IgG antibody titers developed in HIS-HUHEP mice. HIS-HUHEP mouse plasma was assayed for antigen-specific antibodies by immunocytochemistry on transfected BHK-21 cells producing either HBs or HBc protein, or b-galactosidase as a control. 24 BHK-21 cells were incubated with humanized mouse sera previously normalized to 80 mg/ mL total human IgG followed by 2-fold serial dilutions (details provided in Supplementary Materials and Methods).

Gene Expression Analysis by Quantitative
Reverse-Transcription Polymerase Chain Reaction cDNA was prepared from 200 ng of total liver RNA with 2 steps using Superscript III RT (Invitrogen) for reverse transcription followed by TaqMan PreAmp Master Mix (Applied Biosystems) for target amplification with all TaqMan gene expression assays pooled at 0.2x. Subsequently, specific gene expression was performed using BioMark 48.48 Dynamic Arrays (Fluidigm, San Francisco, CA) following manufacturer's protocols. For each sample, 3 independent experiments were analyzed with 2 housekeeping genes (HPRT and GAPDH). Human specificity of each gene expression assay was verified on C57BL/6 mouse liver samples.

Statistical Analysis
Data sets were tested with 2-tailed unpaired Student t tests or Mann Whitney U tests, correlations were analyzed with Pearson's c 2 test using Prism version 6 (GraphPad Software, San Diego, CA). Significant P values are shown as: *P < .05, **P < .005, and ***P <.0005.

Control of Viremia in Immunocompetent Humanized Mice Chronically Infected With HBV
We analyzed the outcome of HBV infection after 'high dose' inoculation (10 9 HBV DNA copies) in HUHEP and HIS-HUHEP mice. Both models could be productively infected by HBV over the 4-5 month time course of the experiment ( Figure 1, individual animals are shown in Supplementary  Figure 1). While high levels of viral replication were evident in both models, viral progression was markedly different in the presence of human immune cells. Interestingly, viremia increased unchecked in HUHEP mice, whereas it stagnated in HIS-HUHEP mice, resulting in 10-fold lower viral titers in HIS-HUHEP compared with HUHEP mice (respectively, up to z10 8 vs 10 9 HBV DNA copies/mL); Figure 1A and 1B). We determined whether viral loads correlated with hAlbumin levels in HUHEP vs HIS-HUHEP models. In HUHEP mice, this correlation was significant (r 2 ¼0.51), whereas in immunocompetent humanized HIS-HUHEP mice there was no correlation (r 2 ¼0.07) ( Figure 1C and 1D). Moreover, virus progression (ratio HBV DNA/hAlbumin) increased proportionally over time in the HUHEP mice (r 2 ¼0.72), but not in the HIS-HUHEP model (r 2 ¼0.03) ( Figure 1E). The difference between HUHEP and HIS-HUHEP models was highly statistically significant (P < .0001) indicating that HBV infection was controlled in HIS-HUHEP mice, presumably through anti-HBV human immune responses. The cross-talk between infected hepatocytes and immune effectors may differ according to the viral load; chimpanzees inoculated with low or high doses of HBV had distinct intrahepatic immune responses, resulting in either a delayed or rapid viral clearance. 25 We therefore analyzed HIS-HUHEP mice infected with a lower dose (10 7 HBV DNA copies) to assess the impact of HBV load on viral progression. Following infection with the lower inoculum, viremia was globally reduced (maximum z10 7 HBV DNA copies/mL), more efficiently controlled, and 1 HIS-HUHEP mouse cleared the infection ( Supplementary Figures 2 and 3).
We characterized the viral life cycle in these HBV-infected humanized mouse models. Both HBeAg and HBsAg were within clinical ranges and correlated to HBV DNA viral titers ( Figure 1F and 1G). cccDNA levels were higher in 'high dose' compared with 'low dose' inoculated mice (respectively, mean 1.9 vs 0.06 copies/cell; Figure 1H and 1I). However, the replicative activity (intrahepatic total HBV DNA/ cccDNA) was similar for both inocula ( Figure 1J) and comparable to non-treated CHB patients. 26 In our previous work, we found that human hepatocyte graft function, measured as plasma hAlbumin, generally increased over time in non-infected HUHEP and HIS-HUHEP mice. 18 Interestingly, hAlbumin also increased in HBVinfected HUHEP mice, but this failed to occur in HBV-infected HIS-HUHEP mice ( Figure 1, Supplementary Figures 1, 2, and 3). In uPA-based models, active tissue remodeling of the host, independently of the grafted cells, precludes use of ALT measurements (alanine aminotransferase) due to its lack of species specificity. 27 Hepatotoxicity results in soluble and fragmented Cytokeratin-18 measurable in the plasma with a human-specific antibody M65 by ELISA. 28 In non-infected HIS-HUHEP mice, M65 was stable over time, yet levels increased following HBV infection, indicating that human hepatocyte cytolysis was occurring ( Figure 1K). No significant differences were observed in control or HBV-infected HUHEP mice (Supplementary Figure 2).

Cellular Immune Responses in HIS-HUHEP Mice With Chronic HBV-Infection
To assess human immune responses to HBV infection in situ, we immunostained liver sections of infected and control HIS-HUHEP mice. Most human hepatocytes were infected (hAlb þ HBcAg þ ) and strong inflammation (hCD45) composed of T cells (hCD3) and Kupffer cells (hCD68) was observed in the parenchyma of 'high dose' infected animals ( Figure 2). Inflammatory foci of hCD45 þ cells were grouped adjacent to HBcAg þ hepatocytes. Specifically, CD3 þ T cells were concentrated within clusters of infected hepatocytes, whereas CD68 þ Kupffer cells were present throughout  the liver parenchyma and intertwined between infected hepatocytes ( Figure 2). In 'low dose' infected mice, fewer hepatocytes were HBcAg þ , and a diffuse recruitment of CD3 þ T cells and CD68 þ Kupffer cells to the liver was apparent (Supplementary Figure 4). Histologically, liver fibrosis was not observed and a-smooth muscle actin staining was normal (data not shown). Human CD45 þ cellularity was significantly increased in the liver of HIS-HUHEP mice following HBV infection ( Figure 3A). HBVinfected mice demonstrated a striking increase in the numbers of hepatic NK cells (hCD45 þ CD3 -NKp46 þ ) (for low and high dose, respectively, in the liver, 12x and 16x; in the spleen, 3x and 7x) ( Figure 3B). More NK cells were activated (CD69 þ ) and showed a mature phenotype (CD56 þ CD16 þ ) in HBV-infected mice compared with controls ( Figure 3C and 3D). Liver-derived NK cells from high-dose HBV-infected mice produced significantly more IFN-g and TNF-a than from control mice after stimulation with IL-12, IL-15, and IL-18, suggesting heightened responsiveness subsequent to viral stimulation in vivo ( Figure 3E). A similar trend was observed in the spleen, although the differences were not significant.
Persistence of viral antigens during chronic hepatitis induces dysfunctional hyporesponsive T cells with increased expression of inhibitory molecules PD-1 and CTLA-4. 8,29,30 We found that HBV-infected HIS-HUHEP mice had increased numbers of CD3 þ T cells in the liver (low dose, 7x; high dose, 6x) and spleen compared with controls ( Figure 4A). Following infection, CD4 þ T H cells were expanded in both the liver (5x for both low and high doses) and spleen. Furthermore, CD8 þ T C cell numbers increased massively in the liver (low dose, 9x; high dose, 6x), which is consistent with their role as key effectors in HBV clearance ( Figure 4B). Strikingly, in the liver of high-dose inoculated mice, the majority of T cells shifted from a naïve (CD45RA þ ) to a memory phenotype (CD45RO þ ) ( Figure 4C), resulting in the generation of an increased pool of effector memory cells (CD8 þ CD45RO þ HLA-DR þ ) (Supplementary Figure 5). Liverderived memory CD4 þ and CD8 þ cells in high-dose inoculated mice showed an exhausted phenotype with increased expression of PD-1 ( Figure 4D). Interestingly, the frequency of PD-1 þ memory CD4 þ CD45RO þ T cells correlated to viral loads in the liver of low-dose inoculated mice ( Figure 4E). We next analyzed PD-1 ligand (PD-L1) expression in the liver of HIS-HUHEP mice. In non-infected mice, PD-L1 was expressed by few human hepatocytes (hAlbþ) as well as non-parenchymal cells (hAlb-) (Supplementary Figure 6). During chronic HBV infection, PD-L1 þ cells were far more abundant and distributed among immune cells as well as infected and non-infected hepatocytes (HBcAg þ or HBcAg -) (Supplementary Figure 6). These results suggest that in the context of chronic HBV infection, enhanced expression of PD-L1 by hepatocytes and non-parenchymal cells may engage PD-1 on T cells, thereby contributing to their exhaustion.

Antigen-Specific Antibody Responses in HBV-Infected HIS-HUHEP Mice
Absolute numbers of splenic B cells (CD19 þ CD20 þ ) were unchanged after 'low dose' or 'high dose' HBV infection (data not shown). Total hIgM levels were similar for control and 'high dose' infected mice, whereas in the 'low dose' inoculated mice antibody titers increased at 7-9 week post-infection (wpi) ( Figure 5A). Class-switched hIgG antibodies increased significantly in both the 'low dose' and 'high dose' infected animals compared with controls, with an accelerated maturation in the 'low dose' inoculated mice ( Figure 5A). We assessed anti-HBs (HBsAb) and anti-HBc (HBcAb) specific IgG antibody responses using a previously validated, highly sensitive immunohistochemistry assay. 24 Both HBsAb and HBcAb responses were detected in the plasma in a majority of HBV-infected mice (low dose, 75%, n¼4; high dose, 55%, n¼9) ( Figure 5B and 5C). HBV-specific antibody responses developed in a slightly higher fraction in 'low dose' vs 'high dose' infected mice. In line with previous reports, 31 polyreactive IgGs could be detected in a few control HIS mice (16%, n¼19). These results indicate that HIS-HUHEP mice generate HBV-specific IgG antibodies in response to infection.

Human Biomarker Disease Profiles for Chronic HBV in Humanized Mice
Although HBV is considered a "stealth" virus, modest elevations in cytokine and chemokine levels are observed in HBV-infected patients during the acute phase or hepatic flares. [32][33][34][35] To identify serum biomarkers in HBV-infected mice, plasma was screened by multi-analyte profiling. In non-infected HUHEP mice, very low levels of IL-8, CCL4, IFN-a, CXCL9, and CXCL10 could be detected that did not significantly increase following HBV infection (Supplementary Figure 7). Similarly, in non-infected HIS-HUHEP mice, IL-1b, IL-12, IL-1Ra, IL-2R, CCL5, and IFN-g, could be detected but did not significantly change postinfection. In contrast, HBV-infected HIS-HUHEP mice had significantly increased levels of inflammatory mediators IL-6, IL-8, CCL2, CCL3, CCL4, TNF-a, as well as IFN-a and the interferon stimulated genes CXCL9 and CXCL10 ( Figure 6A). Furthermore, the immunosuppressive cytokine IL-10 was also upregulated ( Figure 6A). Biomarker profiles from low-dose inoculated mice had a moderate inflammatory signature, with significant increases observed only in IL-6, IL-8, and CCL2. Interestingly, plasma levels of CXCL10, IL-10, and to a lesser extent IFN-a, correlated with the expression levels of PD-1 on memory CD4 þ T cells ( Figure 6B).
We further analyzed gene expression profiles in the liver. Expressions of pro-inflammatory molecules CCL3, CCL4, IL-8, IL-18, CXCL9, CXCL12, and of acute phase response genes IL-6, C-reactive protein, and serum amyloid A2, were upregulated in HBV-infected HIS-HUHEP mice ( Figure 6C and data not shown). Although a weak IFN-a response was detected in the plasma of low-dose infected HIS-HUHEP mice, a clear increase in interferon-stimulated genes ISG56, and RIG-I was observed in the livers ( Figure 6C). Previous studies showed that NK and T cells eliminate HBV-infected hepatocytes via Fas (CD95) and TRAIL-mediated apoptosis. 32,36 In the liver of HIS-HUHEP infected mice, we also observed increased expression of Fas/CD95, TRAIL, and TRAIL-R ( Figure 6C). HBV antigens act as immunosuppressors by inhibiting the activity of innate sensors Toll-like receptor (TLR) 7 and 9 that recognize viral nucleic acids. 37,38 Similarly, TLR7 and TLR9 expression were downregulated in HIS-HUHEP-infected livers ( Figure 6C). Moreover, immunosuppressive mediators (IL-1Ra and PD-L1) were also overexpressed in HBV-infected mouse liver ( Figure 6C). In HUHEP livers following HBV infection, IL-18 and CXCL-12 were not detected and C-reactive protein, SAA-2, ISG56, RIG-I, FAS, TRAIL, TRAIL-R, TLR7, TLR9, IL-1Ra, and PD-L1 were not significantly modulated (data not shown).

HBV-Infected Liver Progenitor Cells
Inflammatory cytokines (IL-6, TNF-a, IFN-g) produced by macrophages or activated T and NK cells have been shown to induce proliferation of liver-progenitor cells  (LPC) in response to liver damage. 22,39,40 The more aggressive HCC subtypes express LPC markers (EpCAM, CK7, CK19, CD133) 41,42 and increased incidences of HCC have been correlated to high viral loads in CHB patients. 9 EpCAM þ Alb þ and CK7 þ Alb þ LPCs were observed in noninfected HIS-HUHEP mice, as expected in the context of liver regeneration because of continuous expression of the uPA transgene (Supplementary Figure 8A). 43 Interestingly, in HBV-infected mice, some LPC were HBc þ and the more dedifferentiated progenitor-like CK7 þ Albumin -HBc þ cells were only observed in the high-dose infected mice (Supplementary Figure 8B and 8C). These HBc þ LPCs could originate from infected hepatocytes that had dedifferentiated, or may have been infected after their dedifferentiation to LPCs. Some LPCs localized near CD3 þ T cell clusters, suggesting inflammation may impact on their development (Supplementary Figure 8D). 44 Testing Anti-Viral Therapies in HBV-Infected

HIS-HUHEP Mice
To determine whether suppression of viremia modifies intrahepatic immunophenotypes, we treated chronically HBV-infected HIS-HUHEP mice with the nucleoside analogue entecavir (ETV). ETV efficiently reduced viral loads (> 3 logs) in HBV-infected mice; viremia was undetectable in 2 of 4 treated mice ( Figure 7A and Supplementary Figure 9). ETV-treated mice showed decreased liver inflammation with reduced monocytes and NK cells, including proinflammatory NK cells (CD56 þ CD16 þ ), at levels comparable with noninfected mice (Figure 7B, C, and D). CD4 þ and CD8 þ T cellularity diminished and returned to a more naïve phenotype (CD45RA þ ), while the frequency of PD-1 þ memory CD4 þ CD45RO þ cells tended to decrease ( Figure 7E and Supplementary Figure 9). Among immunoregulatory mechanisms induced by HBV infection, T REG accumulation in the liver of CHB patients dampens the antiviral immune response. 8 Interestingly, whereas HBV-infected HIS-HUHEP mice had more T REG CD4 þ CD25 þ Foxp3 þ cells in the liver than controls, they were reduced after ETV treatment, and T REG cellularity correlated with viral loads ( Figure 7F). Thus, ETV treatment resulted in decreased viral loads and restored naïve immune phenotypes in the liver.

Discussion
HIS-HUHEP mice constitute a robust small animal model to investigate bi-directional interactions between the immune system and the liver during hepatotropic infections by HBV. The strength of this model relies on the combination of stable hepatic chimerism that supports high levels of HBV replication, coupled with multilineage development of myeloid and lymphoid cell subsets recruited to the liver following infection. Although HIS and HUHEP grafts are not HLA matched, non-infected HIS-HUHEP mice showed no signs of graft vs graft responses. No inflammation or hepatotoxicity was observed, most likely because of the timing of grafts and the tolerant liver environment. 18 Nevertheless, HBV-infected HIS-HUHEP mice developed chronic hepatitis that persisted for several months with viral titers, HBe and HBs antigenemia, cccDNA, and anti-viral immune reactions that were similar to those observed in CHB patients. Inoculation with lower or higher doses in HIS-HUHEP mice resulted in viremia similar to patient samples categorized as moderate vs elevated. 45 Taken together, HIS-HUHEP mice represent a versatile system in which the molecular and cellular mediators of human antiviral immune responses can be studied. In this report, we investigated how viral loads affected the quality of the anti-HBV immune response and the outcome of viral infections.
The cellular distribution and phenotypes of immune mediators in the liver are distinct from those found in the periphery, thus analysis at the site of infection is key to deciphering anti-viral immunity. 4 In HBV-infected HIS-HUHEP mice, Kupffer cells swarmed the liver, while NKand T-cell recruitment increased proportionally to the dose of viral inoculum. Chemokine attraction has been suggested to play a role in this process via CXCL9 and CXCL10/CXCR3 interactions during hepatic flares in patients. 35 We found that hepatic MIP-1a, MIP-1b, CXCL9, and CXCL10 levels were increased in chronically infected mice, which could participate in continuous monocyte, NKcell, and T-cell recruitment to the liver. Intrahepatic NK cells were activated (CD56 þ CD16 þ CD69 þ ) and functionally producing IFN-g and TNF-a ex vivo following re-stimulation. The role of NK cells in mediating viral clearance vs persistence is still unresolved: they are essential in killing virally infected cells, yet they can eliminate HBV-specific T cells via TRAIL-mediated pathways, thereby contributing to pathogen maintenance. 7,46,47 Liver-resident CXCR6þ NK cells have elevated TRAIL expression that may play a role in modulating T-celldependent HBV immunity. 48 Because NK-and T-cell recruitment were positively correlated with viral infection = Figure 6. Biomarker analysis in HBV-infected HIS-HUHEP mice. (A) Plasma from HIS-HUHEP control (n¼11) and HBV-infected mice (inoculum 10e7, n¼5; or 10e9, n¼10) at endpoint were analyzed using human cytokine multiplex assay. No crossreactivity with mouse cytokines was detected. Plasma levels of IL-1b, IL-1Ra, IL-2, IL-2R, IL-4, IL-12, IL-13, IL-15, IFN-g, GM-CSF, and RANTES were similar between control and HBV-infected mice; while IL-5, IL-17, and Eotaxin were not detected (data not shown). (B) Correlation of PD-1 þ memory CD4 þ CD45RO þ T cells with either IFN-g, or IP-10/CXCL10, or IL-10 plasma cytokines quantified in (A). Each dot represents a mouse: grey or black inoculated with, respectively, HBV 10e7 or HBV 10e9. (C) RT-qPCR analysis of liver samples from HIS-HUHEP control (n¼14) and HBV-infected mice (inoculum 10e7, n¼5; or 10e9, n¼9). Fold changes in gene expression of HBV-infected compared with control mice are shown. Data was normalized to the internal control human GAPDH (hGAPDH) to account for differences in humanization levels on triplicate samples. Dotted line indicates fold change of 1. Histograms show the mean and SEM. Data from 14-20 wpi. Statistical significance: Mann Whitney U tests. in HIS-HUHEP mice, it seems unlikely that NK cells were eliminating hepatic HBV-reactive T cells in this context. Still, HIS-HUHEP mice should offer a useful tool to further study intrahepatic NK-cell responses during different phases of HBV infection where NK cells may modulate anti-viral responses.
A broad and robust T-cell response is essential for viral clearance, yet in the context of CHB patients, excess stimulation by viral antigens induces hypo-responsive PD-1 þ T cells. 8,11,30 We found that PD-1 þ T cells were similarly enriched in the livers of HBV-infected HIS-HUHEP mice. Infected hepatocytes and antigen-presenting cells constituted a source of PD-L1 expression that may create an immunosuppressive environment and promote PD-1 mediated T-cell exhaustion. Interestingly, in low-dose inoculated and ETV-treated mice, the frequency of PD-1 þ memory CD4 þ T cells correlated with viral loads. This result suggests that low viral loads can be effectively controlled via anti-viral T cells up to a threshold level. Once this viral load is exceeded, T-cell responses may become ineffective, in part through PD-1 induction. Accordingly, in high-dose infected HIS-HUHEP mice, excessive viral production may overwhelm T-cell responses with uniform conversion to PD-1 þ phenotype. While PD-1 blockade can improve anti-HBV T-cell responses in vitro, 11,30 this does not occur in all patients. Other co-inhibitory receptors, including CTLA-4, Tim-3, LAG-3, and 2B4, may have non-redundant roles in regulating HBV-specific T-cell responses. 29 CD3 þ T cells were specifically recruited to HBV-infected human hepatocyte foci (but not uninfected human hepatocytes), suggesting that human T cells are activated in a virusspecific fashion. However, because of the HLA-mismatched setting of these HIS-HUHEP mice (and eventually PD-1dependent exhaustion), we were unable to detect HLAspecific T-cell responses (data not shown), which could have added to the usefulness of the model. Future studies in HLA transgenic and haplotype-matched immune system/hepatocyte HIS-HUHEP mice may help to address this issue. Still, antigens from virally infected hepatocytes may be cross presented by dendritic cells via MHC class I to CD8 þ T cells, and in chronic HBV, CD14 þ monocyte-derived dendritic cells carry HBV antigens to induce CD8 þ T-cell responses. 49 These mechanisms may allow HBV proteins to gain access to the MHC class I pathway to generate virus-specific T-cell responses in our model.
The balance between immunosuppressive and stimulatory signals regulates immune cell recruitment and activation in the liver of CHB patients and contributes to the pathophysiology during hepatic flares. This interplay was mirrored in HBV-infected HIS-HUHEP mice that had elevated pro-inflammatory mediators along with immunoregulatory cytokines in the plasma and T REG recruitment to the liver. Plasmatic IL-10, CXCL10, and IFN-a levels correlated with the frequency of intrahepatic PD-1 þ memory CD4 þ T cells, which also correlated to HBV viremia, suggesting that the viral antigen load is able to fine tune the immune response via multiple pathways. These secreted biomarkers could facilitate the non-invasive evaluation of intrahepatic immune responses. Antiviral treatments with NUC analogs control viral replication yet do not deplete cccDNA pools and rarely lead to functional cures. However, long-term treatments lead to improved clinical outcomes with reversion of hepatitis, fibrosis, and cirrhosis in CHB patients. HBVspecific polyfunctional T-cell responses and quiescent NK-cell phenotypes are partially restored in peripheral blood mononuclear cells from treated patients, yet the mechanisms regulating viral-host responses following NUC treatments are unclear. 47,50 ETV treatment of HBV-infected HIS-HUHEP mice reduced viral loads and restored naïve immune profiles, with diminished liver infiltration of monocytes, inflammatory NK cells, CD4 þ and CD8 þ T and T REG cells. These results suggest that HBV viral loads are sensitively detected by the immune system and demonstrate the proof-of-concept utility of HIS-HUHEP mice for evaluating therapeutic strategies.
Because of the lack of robust prognostic biomarkers of HCC development and the invasiveness of repeated biopsy sampling in human liver, the evolution of HBV-related pathology to HCC development is still obscure. HCC tumor samples from CHB patients express elevated levels of LPC gene transcripts (EpCAM, CK19, AFP) and these LPCs are tumorigenic in immunodeficient NOD/SCID mice, suggesting that they may constitute tumor-initiating cells. 42,51 Hepatic ectopic lymphoid structures have been associated to LPCs with poor HCC prognosis. 44 Following 4 months of high viremia and chronic inflammation in HIS-HUHEP livers, HBV-infected LPCs were observed in inflammatory sites, although HCC did not develop in our model. The human timeframe for HCC development (several decades) would be difficult to match in a mouse model. Nevertheless, chronically infected HIS-HUHEP mice may help to define tumorigenic pathways that are relevant to HCC initiation and progression.
The lack of effective therapeutic strategies to eliminate HBV, or induce a sustained anti-viral immune response off treatment, remains a major health problem. Adults co-infected with HIV and HBV have a higher incidence of liver-related mortality, although the underlying mechanisms remain obscure. Our HIS-HUHEP model could offer a means to understand disease pathophysiology in this context. Taken together, humanized immune system and liver mouse models provide a potential platform to test innovative therapeutic anti-HBV strategies that combine direct-acting antivirals with immunomodulatory drugs to assess and eventually harness the full potential of the human innate and adaptive immune response to cure HBV infections.

Supplementary Material
Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/ j.gastro.2017.08.034.