Macrophage Depletion Reactivates Fecal Virus Shedding following Resolution of Acute Hepatitis A in Ifnar1^–/– Mice

ABSTRACT

Although hepatitis A virus (HAV) is associated only with acute hepatitis in humans, HAV RNA persists within the liver for months following resolution of liver inflammation and cessation of fecal virus shedding in chimpanzees and murine models of hepatitis A. Here, we confirm striking differences in the kinetics of HAV RNA clearance from liver versus serum and feces in infected Ifnar1^−/− mice and investigate the nature of viral RNA persisting in the liver following normalization of serum alanine aminotransferase (ALT) levels. Fecal shedding of virus produced in hepatocytes declined >3,000-fold between its peak at day 14 and day 126, whereas intrahepatic HAV RNA declined only 32-fold by day 154. Viral RNA was identified within hepatocytes 3 to 4 months after inoculation and was associated with membranes, banding between 1.07 and 1.14 g/cm³ in isopycnic iodixanol gradients. Gradient fractions containing HAV RNA demonstrated no infectivity when inoculated into naive mice but contained neutralizing anti-HAV antibody. Depleting CD4⁺ or CD8⁺ T cells at this late point in infection had no effect on viral RNA abundance in the liver, whereas clodronate-liposome depletion of macrophages between days 110 and 120 postinoculation resulted in a striking recrudescence of fecal virus shedding and the reappearance of viral RNA in serum coupled with reductions in intra-hepatic Ifnγ, Tnfα, Ccl5, and other chemokine transcripts. Our data suggest that replication-competent HAV RNA persists for months within the liver in the presence of neutralizing antibody following resolution of acute hepatitis in Ifnar1^−/− mice and that macrophages play a key role in viral control late in infection.

IMPORTANCE HAV RNA persists in the liver of infected chimpanzees and interferon receptor-deficient Ifnar1^−/− mice for many months after neutralizing antibodies appear, virus has been cleared from the blood, and fecal virus shedding has terminated. Here, we show this viral RNA is located within hepatocytes and that the depletion of macrophages months after the resolution of hepatic inflammation restores fecal virus shedding and circulating viral RNA. Our study identifies an important role for macrophages in virus control following resolution of acute hepatitis A in Ifnar1^−/− mice and may have relevance to relapsing hepatitis A in humans.

INTRODUCTION

Hepatitis A virus (HAV) is a medically important virus that is transmitted by the fecal-oral route and causes outbreaks of acute liver disease in humans. Recent outbreaks have occurred in the United States (1), European countries (2, 3), and several Asian countries (notably, South Korea) (4, 5). Following an incubation period after ingestion of HAV, infected persons develop acute hepatitis with elevated levels of serum alanine aminotransferase (ALT). HAV infection is typically self-limiting and is cleared by the immune system in otherwise healthy individuals, resulting in lifelong immunity (6). However, monophasic relapse has been reported in as many as 3 to 20% of cases (7, 8). Using a chimpanzee model of hepatitis A, Zhou et al. suggested that relapsing hepatitis might be associated with rapid contraction of or failure to sustain HAV-specific CD4⁺ T cell responses (9). However, the mechanisms underlying recrudescent hepatitis A remain largely unknown, and the immune mechanisms by which the virus is cleared from the liver are not fully understood.

We recently developed a murine model of HAV infection that recapitulates important aspects of liver disease and viral replication kinetics seen in infected humans and chimpanzees (10, 11). Following inoculation with wild-type human HAV, type I interferon-receptor-deficient Ifnar1^−/− mice become viremic and shed virus in their feces. Infection is confined to the liver, with viral RNA present in hepatocytes, but is also found in much smaller quantities in the spleen. Appreciable levels of viral RNA are not found within intestinal tissues. Virus shed in feces is produced in infected hepatocytes and released to the gut via the biliary system (12). Infected mice develop hepatic inflammation with prominent hepatocellular apoptosis that results from interferon regulatory factor 3 (IRF3)-mediated transcriptional responses (13). Viremia and fecal shedding end by approximately 4 weeks and 16 weeks, respectively. However, intrahepatic HAV RNA remains high for up to 22 weeks, mirroring what has been observed in experimentally infected chimpanzees (14). The nature of this persistent intrahepatic RNA is unclear. One hypothesis is that the intrahepatic HAV RNA is associated with ongoing genome replication in hepatocytes in the absence of virus release, due to either a failure of assembly or viral egress from hepatocytes. An alternative hypothesis is that the RNA is associated with virions trapped in complex with neutralizing antibodies within an endosomal compartment of hepatocytes or phagocytic cells, such as Kupffer cells. These hypotheses are not mutually exclusive. The Ifnar1^−/− mouse model of HAV infection provides an opportunity to understand the immune mechanisms that control HAV infection and prevent relapse of viremia, fecal shedding of virus, and liver injury. In this study, we investigate the nature of the HAV RNA that persists within the liver after termination of viremia and fecal shedding and the immune mechanisms controlling HAV infection following resolution of acute liver injury in Ifnar1^−/− mice.

RESULTS

Persistent intrahepatic HAV RNA in Ifnar1^–/– mice.

To understand HAV clearance from the liver, Ifnar1^−/− mice were inoculated with HAV intravenously, and the kinetics of HAV RNA decline were measured in serum, feces, and tissues. The mice became viremic and developed hepatitis, as evidenced by elevated serum ALT levels, and shed virus in feces, as expected (10, 11) (Fig. 1A and B). Serum ALT levels peaked on day 7, returning to the normal range by 56 days postinfection (dpi). HAV was cleared rapidly from the blood. Serum HAV RNA fell to below detectable levels by 28 dpi, although low levels of RNA were occasionally detected out to 60 dpi. Fecal shedding of virus declined more slowly, with an estimated half-life of 4.2 days (95% confidence interval [CI], 2.3 to 6.8) in a one-phase decay model (Fig. 1A). By 112 dpi, fecal shedding had declined 300-fold from peak shedding, and by 126 dpi, it was below the limit of detection (>3,000-fold). In contrast, HAV RNA in the liver declined much more slowly, with an estimated half-life of 40 days (95% CI, 18 to 88 days). Whereas fecal viral RNA became undetectable, intrahepatic HAV RNA persisted at the last observation (154 dpi) with only a 32-fold reduction compared to the peak. These findings recapitulate earlier observations of HAV infection in chimpanzees and Ifnar1^−/− mice (10, 14). The disparity evident in the temporal decline of fecal virus shedding versus intrahepatic HAV RNA is particularly striking in view of the fact that all available evidence suggests fecally shed virus is derived from virus replicating within hepatocytes and reaches the gut via passage through the biliary system.

FIG 1.

Persistence of intrahepatic HAV RNA in Ifnar1^−/− mice. (A) Time course of infection following intravenous inoculation with 1.7 × 10⁹ genome-equivalents (GE) HAV showing HAV RNA in serum, feces, liver, and spleen. (B) Serum alanine aminotransferase (ALT) and HAV-specific antibodies measured in serum of Ifnar1^−/− mice following intravenous inoculation with HAV. (C) HAV RNA in multiple tissues of HAV-infected Ifnar1^−/− mice. HAV RNA was not detected in kidney, lung, brain, ileum, or colon at 14 and 56 days postinfection (dpi). Kidney, lung, brain, ileum, and colon were not tested at 112 and 154 dpi. Data are the mean ± standard error of the mean (SEM); n = 4 to 8 mice at each time point. Symbols represent individual mice. All results are representative of two independent experiments.

To more fully examine the tissue distribution of HAV in Ifnar1^−/− mice, we also determined the HAV RNA content of other organs in animals euthanized at 14, 56, 112, and 154 dpi (Fig. 1C). HAV RNA was detected in the spleen at all time points tested, likely reflecting sequestration of virus released from hepatocytes by phagocytic cells (10). Low levels of HAV RNA were detected in lung at 14 dpi, but these were 4 log₁₀ lower than in liver. Otherwise, HAV RNA was not detected in kidney, lung, brain, ileum, or colon. Antibodies specific for HAV were detected in serum by enzyme-linked immunosorbent assay (ELISA) at 21 dpi, coinciding with the end of viremia (Fig. 1B). Anti-HAV antibodies peaked at 42 dpi, remaining at comparable levels until 154 dpi, which was the last time point analyzed.

Cytokine and chemokine expression persists in HAV-infected Ifnar1^−/− mice after ALT normalization.

It is not clear whether the viral RNA detected in liver tissue after resolution of acute liver injury in chimpanzees and mice is associated with replicating virus or, conversely, represents residual nonreplicating RNA or virions sequestered in a cellular compartment of hepatocytes or phagocytic cells awaiting degradation. We hypothesized that actively replicating HAV RNA would generate double-stranded RNA intermediates that would activate antiviral signaling pathways. To test this hypothesis, we measured intrahepatic expression of interferon-stimulated gene (ISG) transcripts by reverse transcriptase quantitative PCR (RT-qPCR) at different time points following infection (Fig. 2A). These included transcripts encoding type I interferon (IFN; Ifnβ), multiple ISGs (Isg15, Ifit1, Ifit2), the chemokine Ccl2 (macrophage chemoattractant 1; MCP-1) and Ccl5 (CC chemokine ligand 5) (RANTES), which are regulated by IRF3 and known to be induced in Ifnar1^−/− mice infected with HAV (10, 13, 15). At 14 dpi, transcripts were elevated for each of these ISGs in most infected mice (Fig. 2A). Intrahepatic ISG expression declined over time, with substantial variation in expression levels in different animals. Some of the infected mice maintained high levels of ISG expression at later time points, while in others it returned to the normal range. A strong positive correlation existed between HAV RNA abundance and the relative mRNA expression of Ccl5 (Spearman r = 0.7075, P < 0.0001), and there were moderate positive correlations between HAV RNA and transcript levels for Ifnβ, Isg15, Ifit1, Ifit2, and Ccl2 (r = 0.4455-0.5804, P < 0.01) (Fig. 2B).

FIG 2.

Persistent activation of interferon-stimulated gene (ISG) expression in livers of HAV-infected Ifnar1^−/− mice. (A) Intrahepatic ISG expression was measured by quantitative RT-PCR and normalized to Gapdh mRNA in liver from naive or HAV-infected mice sacrificed at 14, 56, 112, or 154 dpi. Normalized ISG expression is plotted relative to mean ISG expression in naive mice for Ifnβ, Isg15, Ifit1, Ifit2 (Mcp-1), Ccl2, and Ccl5 (RANTES). At least 4 mice were analyzed at each time point. Symbols represent individual mice. The gray zone represents the range in naive mice. (B) Spearman correlation analysis of intrahepatic HAV RNA abundance and fold change in ISG expression in liver from HAV-infected Ifnar1^−/− mice at 14, 56, 112, and 154 dpi and naive Ifnar1^−/− mice. Symbols represent individual mice. Spearman r and P values are shown at the bottom of each panel. (C and D) In situ hybridization (ISH) visualization of HAV RNA (green) and Ccl5 mRNA (red) in liver tissue from naive and HAV-infected mice. Cell nuclei were counterstained with DAPI (blue). (C) Representative images of liver tissue are shown from naive and HAV-infected mice at 14, 56, 112, and 154 dpi. White arrows identify hepatocytes containing both HAV and Ccl5 RNA at 14 dpi; white arrowheads indicate hepatocytes containing HAV RNA but not Ccl5 mRNA, and the yellow arrowhead identify those containing Ccl5 mRNA but not HAV RNA. (D) Expanded images of hepatocytes in 14 dpi liver containing both detectable HAV and Ccl5 RNA (HAV/Ccl5), HAV RNA but not Ccl5 mRNA (HAV), and Ccl5 mRNA but not HAV RNA (Ccl5).

Both HAV RNA and Ccl5 mRNA were visualized by fluorescent in situ hybridization (FISH) in sections of liver tissue taken from infected animals at different times postinfection (Fig. 2C and D). HAV RNA typically appeared as punctae in infected hepatocytes, likely indicating sites of RNA replication. At 14 dpi, hepatocytes containing HAV RNA and Ccl5 mRNA were abundant and widely distributed across the liver, with both found occasionally in the same cell but more frequently in adjacent cells. Viral RNA and Ccl5 mRNA were typically detected in cells with large nuclei and extensive cytoplasm, consistent with the morphology of hepatocytes, as observed previously (10, 13). The proportion of hepatocytes containing detectable viral RNA declined over time, with reductions in the intensity of the HAV RNA punctae (Fig. 2C, compare green punctae at 14, 56, 112, and 154 dpi). By 154 dpi, only small foci of HAV RNA and Ccl5 mRNA could be identified (Fig. 2C), and these were often in close proximity to densely packed infiltrating immune cells (data not shown).

Infectivity of day 121 postinfection liver extracts containing HAV RNA.

To determine whether intrahepatic RNA is associated with infectious virus particles at late points in time postinfection, liver homogenates were subjected to isopycnic density gradient centrifugation, and fractions were analyzed by HAV-specific RT-qPCR (Fig. 3A). HAV RNA banded at two distinct densities, as previously demonstrated in mice infected with HAV, one between 1.072 and 1.117 g/cm³ (corresponding to membrane-associated or quasi-enveloped virus) and the other between 1.202 and 1.264 g/cm³ (naked or nonenveloped virions) (10, 12, 15). At the earliest time point, 14 dpi, most HAV RNA banded at a high density (fraction 18, 1.264 g/cm³), suggesting it was associated with nonenveloped virions, most likely newly replicated intracellular particles in infected hepatocytes, whereas a smaller proportion of the RNA was membrane associated and banded at a lighter density (fraction 11, 1.117 g/cm³). The density profile shifted over time, however, with the amount of membrane-free, naked virus steadily decreasing such that by 154 dpi most of the residual HAV RNA banded at a low density (fractions 8 to 9, 1.072 to 1.087 g/cm³).

FIG 3.

Buoyant density and infectivity of intrahepatic HAV RNA recovered late in infection from Ifnar1^−/− mice. (A) Buoyant density of HAV RNA following isopycnic density gradient centrifugation of liver homogenates at 14, 56, and 154 dpi. HAV RNA was detected in fractions by RT-qPCR. (B) Liver homogenate prepared from Ifnar1^−/− mice at 121 dpi, passed through a 0.45-μm filter, and subjected to isopycnic density gradient centrifugation. Fractions 11 to 14 containing HAV RNA were selected as inoculum for subsequent experiments. (C and D) Cohorts of Ifnar1^−/− mice (n = 4) were inoculated intravenously with gradient-isolated HAV RNA (9.2 × 10⁶ GE) and monitored weekly for viral RNA in (C) feces, and (D) serum ALT. (E) Neutralizing activity of serum at 121 dpi. 18f-NLuc reporter virus was preincubated with serum collected from naive or HAV-infected mice 121 dpi for 1 h before being inoculated onto Huh7.5.1 cells. Nluc activity and cell viability were measured at 48 h postinfection (hpi). Sera from four mice (no. 20 to 23) were assessed. Data are mean values from n = 3 technical replicates from a representative experiment. (F) Detection of mouse IgG by Western blotting in gradient-purified homogenates of Ifnar1^−/− mouse livers. From left to right: control IgG (purified mouse IgG); fractions 11 to 14 from panel B (~1.125 g/cm³, serving as inoculum for the experiment shown in panel C and D), fractions 16 to 18 from panel B (RNA banding at ~1.21 g/cm³, corresponding to naked virions), and fractions 11 to 14, and 16 to 18, from an uninfected Ifnar1^−/− mouse liver homogenate run in a parallel gradient with the 121 dpi sample. (G) Neutralizing activity in 121 dpi liver fractions. 18f-NLuc reporter virus expressing nanoluciferase was preincubated with fractions 11 to 14 or 16 to 18 (combined) from panel B for 1 h before being inoculated into Huh7.5.1 cells. NLuc activity and cell viability were measured at 48 hpi. Fractions 11 to 14 from the 121 dpi liver from each of four mice (no. 20 to 23) were assessed. Data are shown as the mean percent inhibition of NLuc activity ± SEM from n = 3 technical replicates from a representative experiment. Dashed horizontal lines represent 50% inhibition. (H) Protein G precipitation of immunoglobulin-coated HAV from extracts of liver tissue. After pulldown with protein G beads, RNA was extracted, and HAV RNA was quantified by real-time RT-qPCR. Cell culture-produced HAV (HM175/18f) complexed with the anticapsid mouse monoclonal antibody K24F2 was included as a positive control. Negative controls included (i) HAV (HM175/18f virus) treated with a control IgG with no specific activity for HAV and (ii) liver homogenates from naive Ifnar1^−/− mice spiked with HAV but no antibody.

To determine whether RNA banding at the lighter density in liver homogenates was associated with infectious virus, additional gradients were loaded with homogenates prepared from mice 121 dpi and passed through a 0.45-μm filter (Fig. 3B). Fractions containing HAV RNA (fractions 11 to 14, 1.105 to 1.146 g/cm³) were pooled and inoculated intravenously into naive Ifnar1^−/− mice. This inoculum contained 9.2 × 10⁶ genome equivalent (GE), which is approximately 10 to 100 times higher than the minimal infectious dose in Ifnar1^−/− mice (11). Mice were monitored weekly for 6 weeks but showed neither fecal shedding of virus (Fig. 3C) nor elevated serum ALT (Fig. 3D). No HAV RNA was detected in liver or spleen at necropsy 42 dpi, and the animals remained anti-HAV IgG negative. Thus, there was no evidence that the HAV RNA in these gradient fractions was associated with infectious virus. We considered the possibility that virus could be complexed with neutralizing antibodies in the inoculum or that neutralizing antibodies present in the gradient fractions could block the infection in naive mice. A neutralization assay based on inhibition of nanoluciferase expressed by a recombinant reporter virus, 18f-NLuc, confirmed that the infected Ifnar1^−/− mice had developed high titers of neutralizing antibodies to HAV by 121 dpi (Fig. 3E). Immunoblot analysis also indicated the presence of murine IgG in the 121-dpi gradient fractions used for the inoculum (Fig. 3F). Interestingly, IgG abundance was considerably higher in gradients loaded with liver homogenates from HAV-infected mice (121 dpi) than in those from naive mice. We confirmed that these gradient fractions contained neutralizing anti-HAV antibodies (Fig. 3G). Thus, it seems likely that the absence of infectivity was due to the presence of neutralizing antibodies in the gradient fractions.

HAV RNA was efficiently coimmunoprecipitated from day 56 liver homogenates in pulldown assays using protein G beads that bind immunoglobulin (Fig. 3H). This is consistent with the presence of anti-HAV coating naked virus particles in liver tissue and the gradient profile of the intrahepatic RNA at 56 dpi, which showed a substantial peak of RNA banding at the density of naked virus (Fig. 3A). By 121 dpi, however, there was no detectable coprecipitation of HAV RNA with IgG (Fig. 3H). At this late point in the infection, most viral RNA in the liver was associated with membranes and banded at a density similar to that of quasi-enveloped HAV (Fig. 3A and B). These membranes would be expected to block antibody binding to any capsids containing RNA (16). Nonetheless, neutralizing antibody present in the 121-dpi inoculum may have inhibited the outgrowth of any potentially infectious virus, much as low levels of antibody in immune serum globulin prevent infection in humans. Previous studies show that quasi-enveloped virus can be neutralized by antibody after undergoing clathrin-mediated endocytosis and degradation of the membrane within a late endosome/lysosome compartment (16, 17).

CD4⁺ and CD8⁺ T cells are not required for continued control of HAV following ALT normalization and cessation of fecal virus shedding.

As an alternative approach to determining whether infectious virus persists within the liver following ALT normalization and the cessation of fecal virus shedding, we immunodepleted T lymphocytes that contribute to virus control. Virus-specific CD4⁺ and CD8⁺ T cells accumulate in the liver of Ifnar1^−/− mice by 7 dpi and are essential for viral clearance (18). Immunodepleting CD4⁺ or CD8⁺ T cells at or near the peak of infection (9 dpi) results in increases in intrahepatic HAV RNA and serum ALT (18). To understand the role of T cells in continued control of HAV late in infection, after viremia and fecal shedding have ended, CD4⁺ and CD8⁺ T cells were depleted with antibodies beginning on day 119 dpi, when HAV was no longer detectable in serum or feces but remained detectable in liver. Mice received either anti-CD4, anti-CD8, or isotype control antibodies intraperitoneally every 4 to 6 days and were monitored weekly for HAV RNA in feces and serum ALT levels (Fig. 4A). Animals were euthanized after 4 weeks of antibody treatment. Analyses of T cell populations in liver and spleen confirmed effective depletion of CD4⁺ and CD8⁺ T cells (Fig. 4B and C). Despite this, there was no evidence of viral relapse, no fecal shedding, and no increases in serum ALT levels compared to isotype control antibody-treated mice (Fig. 4D and E). Serum anti-HAV levels remained constant (Fig. 4F), and the abundance of HAV RNA in liver was similar in T cell-depleted versus control animals (Fig. 4G). Thus, while CD4⁺ and CD8⁺ T cells are critical for HAV control early in infection (9 to 29 days) (18), they are not required for continued control once viremia and fecal shedding have terminated and anti-HAV is present.

FIG 4.

CD4⁺ and CD8⁺ T cell depletion in HAV-infected Ifnar1^−/− mice does not restore fecal shedding. (A) Experimental design: HAV-infected Ifnar1^−/− mice (119 dpi) were given anti-CD4, anti-CD8, or isotype control antibody intraperitoneally every 4 to 6 days. At 29 days after depletion treatment, animals were euthanized and livers and spleens were collected. (B) Sample flow cytometry data plots showing analyses of T cells in spleens and livers of mice treated with isotype control, anti-CD4, or anti-CD8 antibodies. Numbers in the quadrant represent the percentage of T cells that are CD4 or CD8 positive. (C) Summary of flow cytometry data showing the percentage of T cells positive for CD4 or CD8 in livers and spleen. (D to F) Mice were monitored weekly for (D) fecal HAV RNA, (E) serum ALT, and (F) anti-HAV IgG measured by ELISA. (G) HAV RNA in liver 29 days after depletion treatment. Data are mean ± SEM; n = 3 to 4 in each treatment group. Results are representative of two independent experiments.

Macrophage depletion restores fecal HAV shedding and circulating viral RNA months after apparent viral clearance.

HAV RNA is abundant within the liver, and to a lesser extent the spleen, throughout the course of infection in Ifnar1^−/− mice (Fig. 1A). HAV antigen, viral particles, and viral RNA have also been identified within cells resembling Kupffer cells, in tissue-resident macrophages, and in livers from infected humans and nonhuman primates (19–21). Whether virus is replicating in these phagocytic cells or simply present as a result of scavenging is uncertain. To determine the extent to which such cells contribute to the persistence of HAV RNA in the liver, we depleted phagocytic cells by treating the mice with clodronate-loaded liposomes, or as a control, phosphate-buffered saline (PBS)-containing liposomes. Phagocytosis of clodronate-liposomes, above a certain threshold, induces apoptosis and results in depletion of macrophages and other actively phagocytic cells. We carried out two, completely independent experiments, in which clodronate-liposomes were administered twice at 3- to 4-day intervals beginning 112 dpi (experiment 1) or 110 dpi (experiment 2), and mice were euthanized 8 days later (Fig. 5A). Clodronate treatment reduced the number of F4/80⁺ CD11b⁺ cells in the liver by 93%, and in the spleen by 86% (Fig. 5B and C). Additionally, mRNA transcripts from the Emr1 gene, which encodes the macrophage-specific F4/80 antigen, were decreased approximately 100-fold in liver and 17-fold in spleen (Fig. 5D). Immunohistochemical staining of macrophage marker IBA1 in liver and spleen showed abundant IBA1-expressing macrophages in the mice receiving PBS-liposomes, but not in the clodronate-treated mice (Fig. 5E). Ghost cells observed in the splenic white pulp of clodronate-treated animals were indicative of coagulative necrosis associated with macrophage depletion. Thus, clodronate treatment effectively depleted macrophages in these HAV-infected mice.

FIG 5.

Depletion of macrophages in HAV-infected Ifnar1^−/− mice transiently restores fecal virus shedding and circulating viral RNA. (A) Experimental design: clodronate or control PBS-liposomes were administered intravenously (i.v.) to Ifnar1^−/− mice beginning at 112 dpi (experiment 1) or 110 dpi (experiment 2) and every 3 to 5 days thereafter. Feces, serum, liver, and spleen were collected after 8 days of treatment. (B) Sample flow cytometry plots of cells from liver and spleen from representative HAV-infected Ifnar1^−/− mice 1 day after i.v. administration of clodronate or control PBS-liposomes in experiment 1. (C) Summary of flow cytometry data showing the percentage of F4/80⁺CD11b⁺ cells in liver and spleen. (D) Emr1 expression in the livers and spleens of the clodronate- or PBS-liposome treated mice, measured by RT-qPCR and normalized to Gapdh mRNA. (E) Immunohistochemical staining of IBA1 in macrophage-depleted Ifnar1^−/− mice. IBA1 (a macrophage marker) was labeled with peroxidase-conjugated polyclonal rabbit antibody and visualized (brown staining) with diaminobenzidine. Counterstaining was with hematoxylin. Sections shown are liver (top) and spleen (bottom) from representative HAV-infected PBS-liposome- (left) and clodronate liposome-treated (right) mice. Nuclei are blue. Arrows (inset) indicate ghost cells in the red pulp in spleen from a clodronate-liposome treated animal. (F to I) (F) Serum ALT, (G) HAV RNA in liver and spleen, (H) HAV RNA in serum, and (I) HAV RNA in feces from infected Ifnar1^−/− mice following 8 days of treatment with clodronate (clod) or control PBS liposomes late in the course of infection in two independent experiments, as shown in panel A. (J) Intrahepatic Ccl5, Cxcl9, Cxcl10, Tnfα, and Ifnγ expression in Ifnar1^−/− mice following 8 days of treatment with PBS- or clodronate-liposomes in experiment 1. Cytokine expression was measured by RT-qPCR and normalized to Gapdh mRNA. Data are the mean ± SEM; n = 4 to 6 in each treatment group. Statistical analysis was performed by a two-tailed Mann-Whitney test.

Clodronate treatment resulted in minimal increases in serum ALT levels (Fig. 5F) and no significant changes in the abundance of viral RNA in the liver and spleen (Fig. 5G). This suggests strongly that macrophages are not the major reservoir for HAV RNA persisting within the liver following ALT normalization in Ifnar1^−/− mice. Remarkably, however, clodronate treatment induced a resurgence of detectable viral RNA within the serum of most animals (Fig. 5H), as well as recrudescent, low-level fecal virus shedding (Fig. 5I). These findings were replicated in both independent experiments (Fig. 5H and I). This apparent resurgence of virus replication was associated with reductions in intrahepatic interferon-γ (Ifnγ), Tnfα, and chemokine transcript levels (Fig. 5J). In an effort to determine whether the HAV RNA reemerging in the feces was associated with infectious virus, Huh7.5.1 cells were inoculated with a fecal extract containing 2 × 10² GE HAV RNA. No HAV RNA was detected in the culture medium or cell lysates when the cultures were harvested 6 weeks later (data not shown). However, the particle/infectivity ratio of HAV is on the order of 100:1, and wild-type HAV is well known to replicate exceptionally poorly in cell culture. As an alternative approach, we prepared a filtered fecal suspension from clodronate-treated mice containing HAV at 3 × 10³ GE and used it to inoculate naive Ifnar1^−/− mice. Again, we observed no evidence of infection, including fecal shedding, serum ALT elevation, or anti-HAV IgG in serum (data not shown). Thus, although clodronate depletion of macrophages restored low-level fecal shedding of HAV, we were unable to demonstrate infectious virus in the feces. This was likely due to the low HAV RNA copy number of the inoculum, which was about 10- to 100-fold lower than the minimum infectious dose of HAV in Ifnar1^−/− mice (11). Nonetheless, the combination of circulating viral RNA in the blood and recrudescent fecal HAV shedding following clodronate treatment is highly suggestive of resurgent HAV replication and indicative of a role for macrophages in the control of HAV late in infection.

DISCUSSION

Despite a robust neutralizing anti-HAV antibody response, HAV RNA persists in the liver of Ifnar1^−/− mice for several months following the normalization of serum ALT activities and termination of fecal virus shedding (Fig. 1A and 3E) (10). This apparent persistence of virus could result in part from the lack of type I interferon signaling in these mice. Apart from its role in innate immunity, type I interferon is important for optimal B and T cell responses, and the absence of the receptor is an important caveat for any study involving this murine model of hepatitis A. However, infected Ifnar1^−/− mice generate high levels of neutralizing antibodies and an effective T cell response that has a major role in virus control (18, 22). More importantly, HAV RNA similarly persists, for 34 to 48 weeks postinfection, in chimpanzees with no defects in innate immunity (14). Similar studies have not been done in humans, but these data suggest that the prolonged presence of viral RNA within the liver following the resolution of hepatitis is a general feature of hepatitis A. In contrast, hepatitis C virus (HCV) RNA disappears rapidly from the liver of chimpanzees with acute resolving HCV infection (14). Whether the RNA that persists after resolution of hepatic inflammation is indicative of the continued presence of infectious virus is uncertain but potentially relevant to the pathogenesis of relapsing hepatitis A. Relapses occur in as many as 10 to 20% of patients with acute hepatitis A and are marked by recrudescent symptoms and signs of disease weeks after apparent resolution of the infection (7, 23). Recurrent fecal shedding of virus has been described in such patients, and viral RNA has been detected in serum (7, 8), suggesting that HAV may transiently escape from immune control.

We found that HAV RNA remains detectable within hepatocytes months after hepatitis has resolved in Ifnar1^−/− mice (Fig. 2C). Hepatocytes containing HAV RNA often expressed Ccl5, marking an active IRF3-dependent innate immune response and suggesting that the RNA may be replication competent (Fig. 2A). In isopycnic gradients, the viral RNA banded at a density of 1.07 to 1.14, which is consistent with its presence in quasi-enveloped virions, but also possibly membrane-limited replication organelles (10). Despite efficiently depleting CD11b⁺F4/80⁺ Kupffer cells from the liver (Fig. 5B to E), clodronate-liposome treatment did not significantly reduce the abundance of intrahepatic HAV RNA (Fig. 5G). This is consistent with HAV RNA being found primarily in hepatocytes (Fig. 2C) and argues against the persisting viral RNA residing within Kupffer cells. However, the most notable finding to emerge from these studies was the recrudescent fecal virus shedding and reappearance of viral RNA in serum that followed clodronate-mediated depletion of phagocytic cells (Fig. 5H and I). While the reappearance of circulating viral RNA could have resulted from clodronate-induced macrophage death with concomitant release of sequestered virus, perhaps in apoptotic bodies or even as naked RNA, this would not explain the resumption of fecal virus shedding. Appreciable quantities of virus are not found in the intestinal tissues of Ifnar1^−/− mice (Fig. 1C), or nonhuman primates, and most evidence suggests that fecal HAV shedding requires active transport of newly replicated virions across the apical membrane of hepatocytes into biliary canaliculi (12, 24). A more plausible interpretation of these data is that macrophage depletion unmasked the continued presence of replicating HAV within the liver, allowing for a resumption of replication and leading to a low but detectable level of viremia and fecal virus shedding. Nonetheless, the relatively constant level of intrahepatic HAV RNA found under these conditions suggests that most viral RNA remains dormant and unperturbed by macrophage depletion.

Cytotoxic CD8⁺ T cell responses are critical for control of hepatitis B virus (HBV) and HCV infections in the liver (25, 26). Depletion of CD8⁺ T cells in HBV- or HCV-infected chimpanzees results in prolonged or persistent infection (27, 28). CD8⁺ T cells have also been suggested to play a role in clearance of infected hepatocytes from the liver in acute hepatitis A. CD8⁺ T cells isolated from blood and liver of HAV-infected patients have been shown to lyse autologous fibroblasts infected with HAV (29, 30). However, more recent studies of cellular immune responses in chimpanzees indicate a dominant role for CD4⁺ helper cells in resolution of HAV infection. Functional CD4⁺ T cell responses more closely correlate temporally with clearance of infected hepatocytes than CD8⁺ T cell responses (9). In Ifnar1^−/− mice, the depletion of either CD4⁺ or CD8⁺ T cells at early times (9 to 29 days) postinfection enhanced the abundance of virus within the liver and promoted fecal virus shedding (18). In contrast, we found here that antibody-mediated depletion of CD4⁺ and CD8⁺ T cells at late time points (119 to 147 days), after fecal shedding had terminated, neither increased intrahepatic RNA nor restored viremia and fecal shedding (Fig. 4). These data are not inconsistent with a role for T cells in clearance of HAV-infected hepatocytes from the liver, as demonstrated by earlier studies (9, 18), but suggest that CD4⁺ and CD8⁺ T cell responses are not essential for continued virus control once the virus has been cleared from the blood and no longer shed in feces. In sharp contrast, depleting macrophages restored both fecal virus shedding and, likely, viremia (Fig. 5), suggesting that macrophages play a role in preventing virologic relapse, long after resolution of acute hepatitis. In addition to physically removing and sequestering virions by phagocytosis, macrophages may secrete antiviral cytokines that suppress replication in hepatocytes. Clodronate depletion of macrophages led to significant decreases in Ifnγ and Tnfα transcripts in liver tissue. Notably, IFN-γ has previously been shown to inhibit HAV replication in vitro (31). Overall, our data suggest a model in which replication-competent viral genomes may persist within hepatocytes long after resolution of the acute disease, held in check by macrophages signaling locally to infected hepatocytes to inhibit HAV RNA replication.

MATERIALS AND METHODS

Mice.

Mice were bred and housed at the National Institute of Infectious Diseases, Japan, in accordance with the policies and guidelines of the Institutional Animal Care and Use Committee (IACUC). Ifnar1^−/− mice were provided by S. Morikawa of the National Institute of Infectious Diseases, Japan (32). All experiments involving mice were approved by the IACUC of the National Institute of Infectious Disease, Japan.

Hepatitis A virus (HAV) infectious challenge.

Mice were infected at 6 to 10 weeks of age by intravenous inoculation of a homogenate of liver from a Mavs^−/− mouse infected with HM175 virus tenth murine passage, 1.7 × 10⁹ GE (genome equivalent) of HAV RNA at the National Institute of Infectious Diseases, Japan. To make the liver inoculum, liver was homogenized in PBS followed by centrifugation at 10,000 × g for 30 min. Supernatant fluids were stored in aliquots at −80°C. The abundance of HAV RNA (GE) in liver inoculum was quantified by real-time RT-qPCR. Infected mice were housed for collection of fecal pellets and periodic collection of serum samples. Tissues were harvested at necropsy and stored in RNAlater (Thermo Fisher Scientific, Waltham, MA) or fixed in 10% neutral phosphate-buffered formalin for 48 h and then stored in 70% ethanol until being processed for histology.

Alanine aminotransferase activity.

Serum alanine aminotransferase (ALT) activity was measured using the alanine transaminase colorimetric activity assay kit (Cayman Chemical, Ann Arbor, MI).

Quantitative RT-qPCR.

RNA was extracted from serum and fecal samples using the QIAamp viral RNA isolation kit (Qiagen, Valencia, CA). RNA was isolated from tissues using TRIzol reagent (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer’s protocol. RNA concentration was measured using a NanoDrop device (Thermo Fisher Scientific). HAV RNA abundance was quantified by real-time RT-qPCR using TaqMan fast virus one-step master mix (Thermo Fisher Scientific) with a QuantStudio 3 real-time PCR system (Thermo Fisher Scientific). HAV RNA levels were determined by reference to a standard curve generated with synthetic HAV RNA. Primers targeted sequences in the 5′ untranslated RNA segment of the genome: 5′-AGGGTAACAGCGGCGGATAT-3′ and 5′-ACAGCCCTGACARTCAATYCMCT-3′. The FAM (6-carboxyfluorescein)/TAMRA (6-carboxytetramethylrhodamine) probe was 5′-AGACAAAAACCATTCAACRCCGRAGGAC-3′. For quantitation of cellular mRNAs, DNA contamination in tissue RNA extracts was removed using the RNase-free DNase set (Qiagen). cDNA synthesis was carried out with the SuperScript III first-strand synthesis supermix for RT-qPCR kit (Thermo Fisher Scientific). Amplifications were carried out with Luna universal qPCR master mix (New England BioLabs, Ipswich, MA). The following primers were used: 5′-TGCGTTCCTGCT GTGCTTCTCCA-3′ and 5′-TTCTCCGTCATCTCCATAGGGATC-3′ (Ifnβ), 5′-AGCAATGGCCTGGGACCTAAA-3′ and 5′-AGCCGGCACACCAATCTT-3′ (Isg15), 5′- CAGAAGCACACATTGAAGAA-3′ and 5′-TGTAAGTAGCCAGAGGAAGG-3′ (Ifit1) (33), 5′- GGGAAAGCAGAGGAAATCAA-3′ and 5′-TGAAAGTTGCCATACAGAAG-3′ (Ifit2) (34), 5′- GCATCCACGTGTTGGCTCA-3′ and 5′-CTCCAGCCTACTCATTGGGATCA-3′ (Ccl2), 5′- AGATCTCTGCAGCTGCCCTCA-3′ and 5′-GGAGCACTTGCTGCTGGTGTAG-3′ (Ccl5) (35), 5′-TGAACGCTACACACTGCATCTTGG-3′ and 5′-CGACTCCTTTTCCGCTTCCTGAG-3′ (Ifnγ) (36), 5′-AAATGGGCTCCCTCTCATCAGTTC-3′ and 5′-CTGCTTGGTGGTTTGCTACGAC-3′ (Tnfα) (37), 5′-TGCACGATGCTCCTGCA-3′ and 5′-AGGTCTTTGAGGGATTTGTAGTGG-3′ (Cxcl9), 5′-GACGGTCCGCTGCAACTG-3′ and 5′-GCTTCCCTATGGCCCTCATT-3′ (Cxcl10) (38), 5′-TCCTGCTGTGTCGTGCTGTTC-3′ and 5′-GCCGTCTGGTTGTCAGTCTTGTC-3′ (Emr1) (39), and 5′- TTCACCACCATGGAGAAGGC-3′and 5′-GGCATCGACTGTGGTCATGA-3′ (Gapdh) (33).

In situ hybridization.

To detect HAV genomic RNA and mouse mRNAs in formalin-fixed paraffin-embedded (FFPE) tissue sections, in situ hybridization was performed using the ViewRNA ISH tissue assay (Thermo Fisher Scientific) according to the manufacturer’s protocol. The probe catalog numbers are VF1-1145 (HAV), VB6-14424 (Ccl5), VB6-12917 (albumin), and VB6-14424 (Emr1). Sections were counterstained with DAPI before mounting. Images were acquired using an oil immersion 40× or 63× lens objective on a Zeiss CellObserver SD spinning disk confocal microscope with a Yokogawa CSU-X1 scan head.

Anti-HAV IgG ELISA.

Serum anti-HAV IgG was quantified with a modification of the previously described HAV capsid antigen ELISA (10). The cell-culture-adapted HAV variant HM175/18f (40) produced in cell culture was inactivated by UV irradiation and captured by anti-HAV polyclonal antibody (ab68579, abcam, Cambridge, UK) coated on a 96-well polystyrene plate. Then, 3-fold serial dilutions of mouse serum samples starting at a 1:33 dilution were incubated in the wells of the plate at room temperature for 1 h, followed by incubation with horseradish peroxidase-conjugated goat-anti-mouse antibody (Thermo Fisher Scientific) at room temperature for 1 h. Following the addition of substrate (3,3′,5,5′-tetramethylbenzidine), the optical density at 450 nm (OD₄₅₀) was determined using a Multiskan FC (Thermo Fisher Scientific) microplate reader. Anti-HAV capsid monoclonal antibody K24F2 (Commonwealth Serum Laboratories, Victoria, Australia) and the sera from HAV-infected Ifnar1^−/− mice seroconverted to HAV were included in assays as positive controls. The sera from naive Ifnar1^−/− mice were used as a negative control.

Isopycnic gradient ultracentrifugation.

Frozen liver was homogenized in PBS (20% wt/vol), and homogenates were subjected to isopycnic gradient ultracentrifugation as described previously (10). Then, 20 fractions were collected from the top of the gradient. RNA was extracted from each fraction using the QIAamp viral RNA isolation kit (Qiagen), and the HAV RNA was quantified by real-time RT-qPCR. The density of gradient fractions was determined by refractometry. Fractions containing quasi-enveloped HAV at the appropriate buoyant densities (1.105 to 1.145 g/cm³; fraction 11 to 14 shown in Fig. 3A; 121 dpi) were combined and subjected to filtration using Amicon Ultracel-100 membranes (100K molecular weight limit) (Merck, Kenilworth, NJ) to remove iodixanol and to exchange the buffer to PBS (12). This yielded 100 μL of purified fraction, which was stored at −80°C until use. Similarly, fractions in which HAV RNA banded at the densities of naked HAV (1.202 to 1.264 g/cm³; fraction 16 to 18 in Fig. 3A; 121 dpi) were purified and stored at −80°C until use.

Immunoblots.

A total of 5 μL of the density gradient fractions of liver purified as described above was subjected to SDS-PAGE and immunoblotting using standard methods. Blots were blocked with ImmunoBlock (KAC, Hyogo, Japan) and probed with anti-mouse IgG labeled with horseradish peroxidase (1:2,000, Themo Fisher Scientific). The immune complexes were visualized by using an ECL Plus Western blotting detection system (Themo Fisher Scientific) with LAS-3000 (Fujifilm, Tokyo, Japan). Purified mouse IgG (BioXCell, West Lebanon, NH) served as the control for the immunoblot.

Immunoprecipitation of HAV.

A total of 50 μL of 20% liver tissue homogenate was lysed in 150 μL of tissue extraction reagent I (Thermo Fisher Scientific) supplemented with cOmplete EDTA-free protease inhibitor cocktail (Merck) at 4°C for 1.5 h. After centrifugation at 15,000 rpm for 10 min, the supernatant was collected and used for immunoprecipitation or extraction of RNA using the QiaAmp viral RNA isolation kit (Qiagen). The HAV RNA abundance was quantified by real-time RT-qPCR. Protein G beads (Fast Flow, GE Healthcare, Chicago, IL) were added and incubated at 4°C for 2 h. Beads were extensively washed with wash buffer (1× Tris-buffered saline [TBS], 1% TritonX-100, and 5 mM DTT supplemented with cOmplete protease inhibitor) and captured immunoprecipitates were subjected to RNA extraction for real-time RT-qPCR. The cell-culture-adapted HAV HM175/18f (40) harvested from supernatant of human liver-derived Huh-7.5 cells from Apath, L.L.C. (New York, NY), was used as the control virion. HM175/18f virus was incubated at 4°C for 2 h with either anti-HAV capsid mouse monoclonal antibody K24F2 (Commonwealth Serum Laboratories, Victoria, Australia), purified nonspecific mouse IgG (BioXcell), or liver homogenate from naive Ifnar1^−/− mice as positive and negative controls.

Neutralization assay using nanoluciferase reporter virus.

Stocks of the cytopathic cell-culture-adapted HAV variant HM175/18f carrying a Nanoluciferase (Nluc) reporter gene (41) were incubated with either serum or the density gradient fractions prepared from liver homogenates for 1 h at room temperature before infecting Huh-7.5.1 cells. Cell viability and Nluc activity in cell lysates were measured at 48 h postinfection. Cell viability was measured using CellTiter-Glo luminescent cell viability assay (Promega) according to the manufacturer’s protocol. For measurement of Nluc, cells were lysed in 1× passive lysis buffer (Promega, Madison, WI) for 30 min at room temperature. Cell lysates were transferred to an enzyme immunoassay (EIA)/radioimmunoassay (RIA) plate (Costar, 3693, Corning, AZ), and Nluc activity was detected using the Nano-Glo luciferase assay system (Promega) according to manufacturer’s instructions. Nluc luminescence was measured using a GloMax 96 microplate luminometer (Promega).

CD4⁺ and CD8⁺ T cell depletion.

To deplete CD4⁺ or CD8⁺ T cells, HAV-infected Ifnar1^−/− mice were injected intraperitoneally with 250 μg GK1.5 (anti-CD4) (BioXCell) or 2.43 (anti-CD8) (BioXCell) antibody at 119 days after virus inoculation and every 4 to 6 days thereafter. Control mice were given 250 μg rat IgG2b isotype control antibody LTF-2 (BioXCell) on the same schedule. To confirm cell type-specific depletion, spleen cells were isolated and stained for cell markers and monitored by flow cytometry as described previously (10).

Macrophage depletion.

To deplete macrophages in HAV-infected Ifnar1^−/− mice in the persistent phase, mice were intravenously injected with 200 μL clodronate liposomes (ClodLip BV, Amsterdam) at 112 days after virus inoculation and every 3 to 4 days thereafter. Control animals were similarly inoculated with an equal volume of PBS liposomes. To confirm macrophage depletion, cells were isolated from the liver and spleen and stained for macrophage markers F4/80 and CD11b and monitored by flow cytometry. FFPE livers and spleens were stained by immunohistochemistry for a macrophage marker anti-ionized calcium binding adaptor molecule 1 (IBA-1).

Immunohistochemistry.

A polymer-based detection system (Histofine simple stain mouse MAX PO; Nichirei Biosciences, Inc., Tokyo, Japan) was used for the immunohistochemical detection of a macrophage marker, IBA1, in FFPE as described previously (42). Tissue sections were deparaffinized and rehydrated prior to antigen retrieval in a retrieval solution (pH 6.0) (Nichirei Biosciences) at 121°C for 10 min in an autoclave. An anti-Iba1 rabbit polyclonal antibody (Wako Pure Chemical Industries, Ltd., Osaka, Japan) was used as a primary antibody. A rabbit Ig fraction (Dako, Agilent, Santa Clara, CA) was used as the negative control for immunohistochemistry. Peroxidase activity was detected with 3,30-diaminobenzidine (DAB; Sigma-Aldrich, St. Louis, MO), and the sections were counterstained with hematoxylin. Images were acquired using a BX53 microscope (Olympus, Tokyo, Japan).

Statistical analysis.

Statistical tests were carried out using Prism 6 software (GraphPad Software, La Jolla, CA). Unless otherwise noted, comparisons between groups used the nonparametric Mann-Whitney test or one-way or two-way analysis of variance (ANOVA). Details of specific statistical tests and experimental design are given in the relevant figure legends.