MAVS-dependent host species range and pathogenicity of human hepatitis A virus

See allHide authors and affiliations

Science  30 Sep 2016:
Vol. 353, Issue 6307, pp. 1541-1545
DOI: 10.1126/science.aaf8325


Hepatotropic viruses are important causes of human disease, but the intrahepatic immune response to hepatitis viruses is poorly understood because of a lack of tractable small- animal models. We describe a murine model of hepatitis A virus (HAV) infection that recapitulates critical features of type A hepatitis in humans. We demonstrate that the capacity of HAV to evade MAVS-mediated type I interferon responses defines its host species range. HAV-induced liver injury was associated with interferon-independent intrinsic hepatocellular apoptosis and hepatic inflammation that unexpectedly resulted from MAVS and IRF3/7 signaling. This murine model thus reveals a previously undefined link between innate immune responses to virus infection and acute liver injury, providing a new paradigm for viral pathogenesis in the liver.

Although viral hepatitis is an important cause of human morbidity worldwide, there are no small- animal models that accurately recapitulate liver disease caused by any of the five responsible viruses (1, 2). Previous studies have relied heavily on nonhuman primates, especially chimpanzees (3, 4), to investigate pathogenesis and immune responses to hepatitis viruses. This has handicapped efforts to understand host responses within the unique immunologic environment of the liver (5, 6). Recent NIH policies effectively eliminate the use of chimpanzees in such studies (7), intensifying the need for alternative models. Here, we report a murine model that recapitulates many features of human infection with hepatitis A virus (HAV), a hepatotropic picornavirus (genus Hepatovirus) that circulates in blood as quasi-enveloped, membrane-cloaked virions and is shed in feces as naked, nonenveloped particles (8).

Like hepatitis B (HBV) and hepatitis C (HCV) viruses, the host range of HAV is considered restricted to humans and nonhuman primates (2, 9). However, successful adaptation to growth in murine and guinea pig cells suggests a broader host range (10, 11). Closely related viruses have also been discovered recently in bats, rodents, shrews, and hedgehogs, with phylogenetic evidence suggesting past shifts among host species (12). HAV replication is strongly suppressed by type I interferon (IFN) (13), but HAV—like HCV—blunts IFN responses in human cells by expressing proteinases that degrade MAVS and TRIF, adaptor molecules involved in the induction of IFN (13, 14). As a result, infected chimpanzees demonstrate limited type I IFN responses (4). Because the sequences targeted in human MAVS and TRIF are not conserved in small mammals (fig. S1A), the inability of HAV to infect these species could stem from a failure to disrupt IFN responses.

To test this hypothesis, we intravenously inoculated Ifnar1−/−Ifngr1−/− double knockout (DKO) mice, which lack receptors for both type I and type II IFN, with wild-type human HAV (15). These mice proved highly permissive for infection, developing multiple features of acute hepatitis A in humans (4, 16): fecal HAV shedding, low-grade viremia, and elevated serum alanine aminotransferase (ALT) activity (Fig. 1A). Multifocal inflammatory cell infiltrates, often surrounding necrotic or apoptotic hepatocytes, were present in liver at 37 to 41 days post-inoculation (dpi) together with HAV RNA (Fig. 1B and fig. S2A). Fecal shedding of infectious virus was confirmed by three subsequent passages in DKO mice, each leading to intrahepatic HAV RNA, fecal virus shedding, and elevated ALT (Fig. 1C and fig. S2B). Antibodies to HAV were detectable at 28 dpi (fig. S3A). A fifth serial passage used fourth-passage liver extract as inoculum. Unlike nonenveloped virions present in feces (density ~1.23 g/cm3) (8), ~65% of liver-derived virus was membrane-associated (~1.11 g/cm3) (fig. S4). This inoculum rapidly induced ALT elevation with impressive fecal shedding and intrahepatic HAV RNA abundance (Fig. 1, C and D).

Fig. 1 HAV infection in DKO (Ifnar1−/−Ifngr1−/−), Ifnar1−/−, and Ifngr1−/− mice.

(A) Representative course of infection in a DKO mouse inoculated intravenously with 107 genome equivalents (GE) of human HAV (chimpanzee fecal extract). (B) Hematoxylin and eosin–stained liver from representative infected (top) and control (bottom) DKO mice at 41 dpi showing inflammatory infiltrates; scale bar, 50 μm. Inset shows apoptotic hepatocytes; scale bar, 12.5 μm. (C) Summary of serial passage of HAV in DKO mice showing intrahepatic HAV RNA (left) and fecal HAV RNA (right), source and magnitude of HAV inocula (ptF, chimpanzee feces; mF, DKO mouse feces; mL, DKO mouse liver), and day of harvest. Data are means ± SEM or range; n = 2 or 3 animals (shown by open circles in each bar). *P < 0.05, ***P < 0.001 for p1 versus p5 [one-way analysis of variance (ANOVA)]. (D) Fecal HAV RNA (top) and serum ALT (bottom) in DKO, Ifnar1−/−, Ifngr1−/−, and wild-type (WT) BL6 mice challenged with fourth-passage DKO liver extract (2.6 × 108 GE). Data are means ± SEM; n = 4. *P < 0.05, ***P < 0.001 for Ifnar1−/− versus DKO (ANOVA). (E) Intrahepatic HAV RNA in WT, DKO, Ifnar1−/−, and Ifngr1−/− mice at 127 dpi (mean ± range, n = 2). (F) Intrahepatic HAV RNA in Ifnar1−/− mice infected with fourth-passage liver extract. Symbols represent individual mice. Dotted horizontal lines in panels indicate level of detection (RNA) or upper limit of normal (ALT).

Like DKO mice, Ifnar1−/− animals shed virus and developed ALT elevation when challenged with liver-derived virus, whereas type II IFN receptor Ifngr1−/− knockouts and wild-type mice showed no evidence of infection (Fig. 1, D and E). The rapid induction of disease in this experiment, compared with slower onset in early DKO passages (Fig. 1A), resulted from a higher inoculum titer rather than viral adaptation to mice. Only a single nonsynonymous nucleotide substitution occurred in the viral sequence over four mouse passages (table S1). Infection persisted in Ifnar1−/− and DKO mice for more than 3 months (Fig. 1D). Declining serum ALT and fecal virus shedding over this period of time suggested slow immune control in both types of mice, but histopathologic lesions persisted throughout (fig. S2C). As in chimpanzees (4), HAV RNA copy numbers remained high in liver after fecal virus shedding had terminated (Fig. 1, E and F).

These data suggest that the capacity of HAV to evade type I IFN responses defines its host range. However, DKO mice were resistant to challenge with either fecal or liver-derived virus administered by oral gavage, possibly reflecting a greater role for type III IFN in the gut (17, 18) or absence of an essential receptor. Rag1−/− and NSG mice lacking adaptive immunity were resistant to intravenous virus challenge (Fig. 2A and fig. S5A), further highlighting the importance of innate immunity in control of HAV. Because HAV-encoded proteinases disrupt IFN responses by degrading human MAVS and TRIF (13, 14), we challenged Mavs−/− and Trif−/− mice to ascertain whether signaling through these adaptor molecules restricts replication. Mavs−/− mice were highly permissive for HAV, shedding 10 times as much virus as DKO or Ifnar1−/− mice (Fig. 2A), whereas Trif−/− mice were nonpermissive (Fig. 2A and fig. S5, B and C). Thus, MAVS-mediated type I IFN responses block HAV replication in wild-type mice. Consistent with this finding, HAV 3ABC, a proteinase that degrades MAVS in human cells (13), did not cleave murine MAVS (fig. S1B).

Fig. 2 HAV infection in Ifnar1−/− and Mavs−/− mice.

(A) Fecal HAV RNA on day 7 and day 14 after intravenous challenge of different genetically deficient mice. Data are means ± SEM; n = 3 to 5. (B) Viral RNA in livers of Ifnar1−/− and Mavs−/− mice at 15 and 63 dpi. Data are means ± SEM; n = 2 to 5 as shown. *P < 0.05, ***P < 0.001 (two-sided t test). (C) Serum ALT at 7 and 14 dpi in genetically deficient mice, with expanded low ALT range at right. Data are means ± SEM; n = 5. **P < 0.01, ***P < 0.001 for combined day 7 and day 15 data (two-sided Mann-Whitney test); ns, not significant. (D) Immunohistochemical staining of cleaved caspase 3 in liver from representative (top) Ifnar1−/− versus (bottom) Mavs−/− mice at 15 dpi; scale bar, 100 μm. Inset shows an apoptotic hepatocyte; scale bar, 12.5 μm. (E) Tissue distribution of HAV RNA in infected Ifnar1−/− and Mavs−/− mice. Data are means ± SEM; n = 3 or 4. ***P < 0.001 (multiple t test with false discovery rate of 1%). (F) Fecal virus shedding in infected Ifnar1−/− or Mavs−/− mice over 56 days of infection. Data are means ± SEM; n = 3 to 5.

Intrahepatic HAV RNA copy numbers were higher in Mavs−/− mice than in Ifnar1−/− mice by a factor of 10 (Fig. 2B), with the majority of Mavs−/− hepatocytes containing HAV RNA (fig. S5, D and E, and table S2). Nonetheless, Mavs−/− mice developed neither ALT elevation (Fig. 2C) nor hepatic inflammation (Fig. 2D). Immunohistochemical staining for activated caspase 3 (Fig. 2D) and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assays (fig. S6A) revealed numerous apoptotic hepatocytes in infected Ifnar1−/− liver, but none in Mavs−/− tissue. Apoptotic cells in Ifnar1−/− and DKO mice were surrounded by inflammatory infiltrates in proximity to cells containing HAV RNA (Fig. 2D and fig. S6B). The activities of caspases 8 and 9 (and of caspase 3) were slightly increased in infected DKO and Ifnar1−/− liver (fig. S6C), but cleaved caspase was not detected in immunoblots; only ~1% of hepatocytes were apoptotic (fig. S6D). These data show that apoptosis and inflammation result from a MAVS-dependent but IFN-independent mechanism. MAVS-mediated apoptosis has been recognized previously, but its role in vivo is uncertain (19, 20).

Virus was largely restricted to the liver in Mavs−/− mice: HAV genomes were less abundant in spleen by a factor of 400 and were less abundant in lung by three orders of magnitude (Fig. 2E). Viral RNA was more abundant in spleens of Ifnar1−/− mice, possibly reflecting sequestration of virus released from damaged hepatocytes. Little virus was present in ileum or colon of either knockout, indicating that fecal shedding originates in the liver, as in primates (21, 22). Thus, HAV is highly hepatotropic in mice. Viral shedding persisted unabated for 56 days in Mavs−/− mice with only minimal ALT increases (Fig. 2F and fig. S6E). Rare, isolated apoptotic hepatocytes were observed in only two of five mice at 63 dpi. The appearance of antibodies to HAV was delayed in Mavs−/− mice (fig. S3B), but virus-neutralizing activities were comparable to Ifnar1−/− mice at 63 dpi.

Irf3−/− and Irf7−/− mice lack transcription factors downstream of MAVS that drive type I IFN expression (23). These mice supported only limited HAV replication, whereas Irf3−/−Irf7−/− double knockouts shed virus and accumulated intrahepatic HAV RNA levels equivalent to those of Mavs−/− or Ifnar1−/− mice (Fig. 2A, fig. S5A, and table S2). This is consistent with redundant roles for IRF3 and IRF7 in the control of flaviviruses (24, 25). Serum ALT elevations were minimal in infected Irf3−/−, Irf7−/−, and Irf3−/−Irf7−/− mice (Fig. 2C and fig. S7A). Irf7−/− and Irf3−/−Irf7−/− livers contained rare apoptotic hepatocytes (fig. S7B), possibly reflecting activation of IRF5 (25). However, a general lack of pathology in infected Irf3−/−Irf7−/− animals mirrored the absence of disease in Mavs−/− mice.

Hepatocyte apoptosis is known to drive inflammation within the liver (26). The livers of Ifnar1−/− mice at 7 dpi showed a factor of 55 increase in F4/80+CD11b+ macrophages and a factor of 13 increase in NK1.1+ natural killer (NK) cells, whereas CD4+ and CD8+ T cells were increased by factors of only 3 to 5 (Fig. 3A). Increases in CD3+CD4CD8 and γ/δ T cells did not achieve statistical significance. Immunohistochemistry confirmed a mixed cellular infiltrate (Fig. 3B). Luminex assays showed increased hepatic CCL3 (MIP-1α), CCL5 (RANTES), and CXCL10 (IP-10) protein, but not IFN-γ, tumor necrosis factor–α (TNF-α), or interleukin (IL)–1β, IL-2, or IL-6 (Fig. 3C). Similarly, serum IFN-β was markedly elevated (>10 ng/ml) in infected DKO and Ifnar1−/− mice (Fig. 3D), but enzyme-linked immunosorbent assays (ELISAs) for IFN-γ, TNF-α, IL-1β, and IL-6 were negative. Nonetheless, reverse transcription polymerase chain reaction demonstrated HAV-induced intrahepatic transcripts for multiple cytokines and chemokines in DKO and Ifnar1−/− mice, but not in Mavs−/− mice (Fig. 3E and fig. S8A). CCL2 (MCP-1) and CCL5 mRNA responses were maximal at 7 dpi, whereas CCL3, IFN-γ, and TNF-α mRNAs peaked at 15 dpi (Fig. 3E) despite the absence of detectable protein in serum or liver. Diminishing chemokine and cytokine responses at 28 dpi (Fig. 3E) correlated temporally with a decline in fecal virus shedding by two orders of magnitude. NLRP3 inflammasome-related transcripts were not increased (fig. S8B).

Fig. 3 Cellular and cytokine response to HAV infection in DKO and Ifnar1−/− mice.

(A) Estimated intrahepatic leukocyte numbers in naïve and infected Ifnar1−/− mice at 7 dpi. Data are means ± SD; n = 5 (mean ALT = 372 IU/liter). **P < 0.01, ***P < 0.001 (two-way ANOVA with Tukey multiple comparison test). (B) Dual immunohistochemical staining of infected Ifnar1−/− liver for CD4 (magenta) and CD8 (brown) showing a mixed cellular infiltrate at 14 dpi. Scale bar, 10 μm. (C) Relative increase in liver cytokine levels in HAV-infected DKO mice (Luminex assay) with ALT >200 IU/liter. Data are means ± range; n = 2. (D) Serum IFN-β measured by ELISA at 7 dpi. Data are means ± SD; n = 4. (E) Relative increase in intrahepatic cytokine and chemokine mRNA abundance in Ifnar1−/− mice. Data are means ± SEM; n = 4 or 5. (F and G) Immunoblots of phospho-Ser396 and total IRF3 (F) and ISG15 (G) in livers from HAV-infected and naïve DKO mice; β-actin was included as a loading control. (H) Intrahepatic transcripts of IRF3-regulated ISGs, ISG15, IFIT1 (ISG56), CCL5 (RANTES), and ISG20 (not directly regulated by IRF3) in HAV-infected DKO (n = 4) and Mavs−/− (n = 3) mice and naïve animals at 18 to 28 dpi. *P < 0.05 (t test).

IFN-β transcription is coordinately regulated by IRF3/7 and nuclear factor (NF)–κB (27). Phosphorylation of IRF3 confirmed IRF3 activation in infected Ifnar1−/− mice (Fig. 3F), and IFN-stimulated genes (ISGs) such as ISG15, IFIT1, and CXCL10 that are directly regulated by IRF3 (28) were induced (Fig. 3, C, G, and H). IRF3 similarly regulates CCL5 transcription (29), explaining prominent and early CCL5 expression by HAV-infected hepatocytes in Ifnar1−/− but not Mavs−/− mice (Fig. 3, C and E, and fig. S8C). The phospho-p65 component of NF-κB was not measurably increased (fig. S8D).

Several possible mechanisms could account for apoptosis induced through a MAVS-IRF3/7 pathway (fig. S8E). First, CCL5 expression could recruit cytotoxic lymphocytes to the liver, resulting in death receptor–mediated apoptosis (30, 31). However, depletion of CD4+ or CD8+ T cells had no impact on acute (7 dpi) disease in Ifnar1−/− mice (fig. S9). Moreover, virus-specific T cell responses were minimal in Ifnar1−/− and Mavs−/− mice (fig. S10). Depletion of NK1.1+ NK cells similarly failed to reduce liver injury (fig. S11). This argues against primary death receptor–mediated apoptosis. Clodronate depletion of macrophages prior to infection also had no effect on viral replication or inflammation (fig. S12).

Alternatively, apoptosis could be induced by ISGs that are directly regulated by IRF3 (28). The functions of these proteins are only partly understood, but IFIT2 (an ISG that is transcriptionally regulated by IRF3) is known to trigger mitochondrial apoptosis in human cells (28, 32). IRF3 similarly regulates PMAIP1, a proapoptotic BH3-only protein (33). Both Ifit2 and Pmaip1 transcripts were induced early, and this induction was more extensive in Ifnar1−/− mice than in Mavs−/− mice (fig. S13). IRF3 can also induce apoptosis through a transcription-independent mechanism involving a direct interaction with mitochondrial Bax (34). However, this would not explain the rare apoptotic hepatocytes observed in Irf3-deficient mice (fig. S7B).

Although many details remain to be resolved, our data show that Ifnar1−/− mice provide a useful model that recapitulates many aspects of type A hepatitis in humans. Despite heroic efforts, such a model has proved elusive for HBV or HCV infection (2). Our results suggest that HAV host species range is dictated largely by its capacity to evade MAVS-mediated type I IFN responses, and reveal an unexpected role for MAVS signaling in virus-mediated liver injury. Such signaling leads to IRF3/7-dependent, but IFN-α/β– and IFN-γ–independent, hepatocellular apoptosis with a secondary inflammatory response (fig. S8E). This may explain why HAV and HCV have evolved independently to target MAVS for degradation. Disrupting innate immune signaling upstream of IRF3/7 not only limits IFN-mediated antiviral responses, but also restricts inflammation within the liver, delays antiviral antibody responses, and slows viral clearance (Fig. 2, D and E, and figs. S3B and S6E). IRF3, activated through STING as a result of endoplasmic reticulum stress, has been implicated recently in acute ethanol-induced hepatitis (35), suggesting a common final pathway for toxin- and virus-induced liver injury. Our findings establish the critical importance of innate immune responses in control of viral infection in the liver, and provide a paradigm for HAV pathogenesis that is likely relevant to other hepatotropic human viruses.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Tables S1 and S2

References (3652)

References and Notes

  1. See supplementary materials on Science Online.
Acknowledgments: We thank D. Yamane, K. McKnight, T. Benzine, L. Wai, D. Hilliard, and M. Chua for helpful discussions and technical assistance; R. Lanford (Texas Biomedical Research Institute) for chimpanzee-passaged HAV; and C. Walker (Research Institute of Nationwide Children’s Hospital) for the generous gift of HAV peptides. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. DNA sequences are deposited in GenBank with accession numbers KX343014, KX343015, KX343016, KX343017, and KX343018. Irf3−/−, Irf7−/−, and Irf3−/−Irf7−/− mice were provided under a materials transfer agreement from Tokyo University; Mavs−/− mice were provided under a materials transfer agreement from the University of Washington. Supported by NIH grants R01-AI103083 and U19-AI109965 (S.M.L.), NIH grants R01-AI074862, R21-AI117575, and R56-AI110682 (J.K.W.), and NCI Center Core Support Grant P30-CA016086 to the Lineberger Comprehensive Cancer Center.

Stay Connected to Science

Navigate This Article