Article Text


Original article
Study of hepatitis E virus infection of genotype 1 and 3 in mice with humanised liver
  1. Ibrahim M Sayed1,2,
  2. Lieven Verhoye1,
  3. Laurence Cocquerel3,
  4. Florence Abravanel4,5,6,
  5. Lander Foquet1,
  6. Claire Montpellier3,
  7. Yannick Debing7,
  8. Ali Farhoudi1,
  9. Czeslaw Wychowski3,
  10. Jean Dubuisson3,
  11. Geert Leroux-Roels1,
  12. Johan Neyts7,
  13. Jacques Izopet4,5,6,
  14. Thomas Michiels8,
  15. Philip Meuleman1
  1. 1Center for Vaccinology, Ghent University, Ghent, Belgium
  2. 2Microbiology and Immunology Department, Faculty of Medicine, Assuit University, Assuit, Egypt
  3. 3U1019—UMR 8204—CIIL-Centre d'Infection et d'Immunité de Lille, Univ. Lille, CNRS, Inserm, CHU Lille, Institut Pasteur de Lille, Lille, France
  4. 4INSERM U1043, IFR-BMT, CHU Purpan, Toulouse, France
  5. 5Université Paul-Sabatier, Toulouse, France
  6. 6Laboratory of Virology, CHU Purpan, Toulouse, France
  7. 7Rega Institute, KUL, Leuven, Belgium
  8. 8de Duve Institute, Université Catholique de Louvain, Brussels, Belgium
  1. Correspondence to Professor Dr Philip Meuleman, Center for Vaccinology, Ghent University, University Hospital Ghent, MRB2, De Pintelaan 185 B-9000 Gent, Belgium; philip.meuleman{at}


Objective The hepatitis E virus (HEV) is responsible for approximately 20 million infections per year worldwide. Although most infected people can spontaneously clear an HEV infection, immune-compromised individuals may evolve towards chronicity. Chronic HEV infection can be cured using ribavirin, but viral isolates with low ribavirin sensitivity have recently been identified. Although some HEV isolates can be cultured in vitro, in vivo studies are essentially limited to primates and pigs. Since the use of these animals is hampered by financial, practical and/or ethical concerns, we evaluated if human liver chimeric mice could serve as an alternative.

Design Humanised mice were inoculated with different HEV-containing preparations.

Results Chronic HEV infection was observed after intrasplenic injection of cell culture-derived HEV, a filtered chimpanzee stool suspension and a patient-derived stool suspension. The viral load was significantly higher in the stool compared with the plasma. Overall, the viral titre in genotype 3-infected mice was lower than that in genotype 1-infected mice. Analysis of liver tissue of infected mice showed the presence of viral RNA and protein, and alterations in host gene expression. Intrasplenic injection of HEV-positive patient plasma and oral inoculation of filtered stool suspensions did not result in robust infection. Finally, we validated our model for the evaluation of novel antiviral compounds against HEV using ribavirin.

Conclusions Human liver chimeric mice can be infected with HEV of different genotypes. This small animal model will be a valuable tool for the in vivo study of HEV infection and the evaluation of novel antiviral molecules.


Statistics from

Significance of this study

What is already known on this subject?

  • Hepatitis E virus (HEV) infection is an emerging problem in industrialised countries.

  • Chronic HEV infection is reported among immunocompromised patients, mainly solid organ transplant patients, which then become at risk of developing severe liver disease and cirrhosis.

  • Ribavirin is the first choice of treatment for chronic HEV infection, but recently viral isolates were identified that harbour a G1634R mutation in their polymerase. This mutation results in more efficient replication of the virus, thereby reducing ribavirin therapy efficacy.

  • The study of HEV is hampered by the absence of robust cell culture systems and small animal models.

What are the new findings?

  • Immune-deficient human liver chimeric mice can be infected with HEV of different genotypes and different sources.

  • The infection evolves towards chronicity and influences the host's gene expression profile. Many interferon-induced genes become activated.

  • The human liver chimeric mouse is validated for the identification of novel antiviral strategies.

How might it impact on clinical practice in the foreseeable future?

  • This report highlights the use of mice with humanised liver as a new small animal model for the study of HEV infection.

  • This model will allow the identification of novel antiviral targets and the preclinical in vivo evaluation of novel antiviral compounds, as alternative treatment against ribavirin-resistant HEV isolates.


The hepatitis E virus (HEV) is a small icosahedral, enterically transmitted virus that belongs to the Hepeviridae family.1 It has a positive-sense single-stranded (ss)RNA genome of about 7.2 kb, which comprises three open reading frames (ORF1, ORF2 and ORF3). ORF1, located at 5′ region of the viral genome, encodes a polyprotein with domains endowed with methyl transferase, putative cysteine protease, helicase and RNA-dependent RNA polymerase activities.1 ,2 ORF2 is located mainly in the 3′ region of the genome and encodes the viral capsid. ORF3, which overlaps with both ORF1 and ORF2, encodes a cytoskeleton-associated phosphoprotein that is involved in virion morphogenesis and egress of virus from cells.

Viruses that can infect humans are classified into genotype (gt) 1–4 of the 7 genotypes of the Orthohepevirus A genus. Every year, approximately 20 million people become infected with HEV worldwide. Most infections occur in low-income and middle-income countries and are mainly caused by gt1 viruses. Overall mortality rates range from 0.5% to 4%, but increased mortality is observed in young infants under 2 years of age and peak in pregnant women who are in their third trimester (10–20%).1

HEV seems to be emerging in the Western world, especially HEV of gt3.3 While it is considered that most gt3 infections in industrialised countries are caused by the consumption of contaminated produce, transmission through blood transfusion has recently been described.1 ,3 ,4 Most infected individuals will spontaneously clear the infection with no or only mild symptoms. Only immune-compromised people (organ transplant recipients, patients with HIV, patients with leukaemia) may evolve towards chronicity and are at risk of developing severe liver disease and cirrhosis. Chronic HEV infection can be cured using ribavirin (RBV) but recently, viral isolates with lower RBV sensitivity have been identified,5 necessitating the discovery of novel antiviral molecules. The study of HEV replication has been hampered by the lack of an efficient cell culture system. Recent reports show that certain HEV isolates, mainly of gt3 and gt4, were adapted to grow in vitro in cell lines such as HepG2/C3A, PLC/PRF/5 and A549.6 ,7 However, gt1 viruses that account for most HEV infection worldwide only poorly replicate in cell culture.8

Macaques and chimpanzees are the only animals susceptible to infections with all HEV genotypes,9 ,10 while pigs can be used for the propagation of HEV gt3 and gt4 only.11 ,12 A small animal model for the study of HEV is still lacking. Here, we report on the evaluation and characterisation of HEV infection in human liver chimeric mice. These mice are generated by transplanting primary human hepatocytes into immune-deficient urokinase-type plasminogen activator-severe combined immunodeficiency (uPA-SCID) mice that suffer from a transgene-induced liver disease. Several weeks after transplantation, the major part of the mouse liver is repopulated by functional human hepatocytes, which renders these chimeric mice susceptible to human-specific hepatotropic pathogens such as hepatitis B virus, hepatitis C virus and Plasmodium falciparum.13–16 We show here that these humanised mice are also susceptible to HEV infection and that they can be used for the preclinical in vivo evaluation of novel antiviral compounds. In addition, we have observed that HEV infection directly influences the host's transcriptome.

Material and methods

Production, infection and treatment of human liver chimeric mice

Approximately 106 primary human hepatocytes (donor HH223, Corning, The Netherlands) were transplanted into homozygous uPA+/+-SCID mice as described previously.13 ,17 ,18 Human albumin concentration, determined by ELISA (Bethyl Laboratories, Montgomery, Texas, USA), was used as marker to assess liver chimerism.

Humanised mice were inoculated with different HEV preparations, mainly of gt3 (obtained from an infected patient and cell culture-derived HEV (HEVcc)) and gt1 (strain Sar-55; stool suspension obtained from an infected chimpanzee10). One particular group of mice was inoculated with sodium deoxycholate and trypsin-treated cell culture supernatant (see online supplement). All other inoculations were performed with non-treated virus preparations. Stool and plasma samples were collected on a weekly basis. Detection and quantification of viral RNA in mouse plasma and 10% (w/v) mouse stool was done using RT-qPCR (see below). Alanine aminotransferase (ALT), aspartate aminotransferase (AST) and bilirubin levels were quantified using the cobas 8000 system (Roche Diagnostics). Non-transplanted mice inoculated with these preparations served as negative control. Passage experiments were performed by inoculating naïve humanised mice with EDTA plasma or a 10% (w/v) 0.22 µm filtered stool suspension obtained from a Sar-55-infected humanised mouse. Inoculations were performed via oral and intrasplenic routes.

RBV was dissolved in phosphate-buffered saline solution and, using 20G plastic feeding tubes (Instech Laboratories, The Netherlands), administered orally at a single daily dose of 25, 50 or 100 mg/kg. Control mice received vehicle only.

Quantification of HEV RNA

HEV RNA levels were quantified via RT-qPCR using primers (5′-GGTGGTTTCTGGGGTGAC-3′ (F) and 5′-AGGGGTTGGTTGGATGAA-3′ (R)) and a probe (5′-FAM-TGATTCTCAGCCCTTCGC-TAMRA-3′) that target a conserved 70 bp region in the ORF2/3 overlap.19 A standard curve was generated and calibrated against the WHO 1st international standard (Paul-Ehrlich Institute, Germany).20 HEV RNA was extracted from plasma and faeces using the NucliSENS easyMAG system (Biomérieux, France) and a one-step real-time RT-qPCR was performed on the LightCycler480 (Roche, Belgium) using the Roche LightCycler Multiplex RNA virus master mix. The limit of quantification (LOQ) of our assay on undiluted samples was 40 IU/mL and equals the 100% limit of detection (LOD).

A detailed description of the production of HEVcc and the comprehensive analysis of infected humanised livers and plasma samples can be found in the online supplement.


Infection of humanised mice with HEVcc

HepG2/C3A cultures were transfected with in vitro transcribed capped RNA of the gt3 HEV isolate Kernow C1-P6. Cell culture supernatant was harvested and administered to human liver chimeric mice. Initial experiments involving oral (n=2) and intrasplenic (n=6) administration of up to 1.1×106 IU HEVcc did not result in a detectable infection.

Next, we injected supernatant HEVcc preparations that underwent sodium deoxycholic acid and trypsin treatment, thereby mimicking the process viral particles undergo after shedding into the bile duct and passage through the duodenum. Although intrasplenic injection of 3.3×105 IU/mouse (n=2) did not result in infection, one out of three mice inoculated with 3.2×106 IU/mouse showed signs of active infection (figure 1A). HEV RNA was only detectable in 10% stool suspensions starting from week 3 until week 9 postinoculation, after which the viral titre decreased. This decrease coincided with a decline in human albumin level, while ALT, AST and bilirubin levels were not affected, indicating spontaneous failure of the human liver graft rather than viral-induced hepatotoxicity.

Figure 1

Infection of humanised mice with cell culture-derived virus. Human liver chimeric mice were inoculated intrasplenically with (A) supernatant treated with sodium deoxycholic acid and trypsin or (B) lysate of hepatitis E virus (HEV)-infected cell cultures (genotype 3, strain Kernow C1-P6). Viral RNA was quantified at different time points in 10% (w/v) faecal samples (black) and plasma (red). Solid lines represent data obtained from humanised mice, while dotted lines show data obtained from inoculated non-humanised control mice. The different symbols used indicate data obtained from different animals and allow the comparison of corresponding faecal and plasma data of a single animal. LOQ, limit of quantification.

HEV RNA could also be detected in the stool of two out of three humanised mice injected in the spleen with lysate from an infected cell culture (1.7×106 IU/mouse). Virus was detected from week 1 onward and titres steadily increased reaching levels up to 6.2×104 IU/mL (figure 1B). In one animal, we observed a decrease in viral load after 12 weeks of infection, which again corresponded with a decrease in human albumin. ALT, AST and bilirubin levels did not significantly change in both infected mice. Sequence analysis of the virus present in the stool of these two mice (collected 12 weeks after infection) showed complete conservation of the 58 amino acid-long human S17 ribosomal sequence that is inserted in the ORF1 region of this isolate (data not shown).

HEV RNA could not be detected in the plasma of any of the infected humanised mice, nor in the plasma or in the faeces of non-humanised control mice inoculated with the same viral preparations.

Infection of humanised mice with chimpanzee-derived HEV

A 10% (w/v) stool suspension obtained from a chimpanzee infected with HEV strain Sar-55 was used as inoculum for gt1 infection studies. Two out of two humanised mice that were intrasplenically inoculated with 3×106 IU/mouse became robustly infected. HEV RNA was detected both in faeces and plasma starting from week 1 postinoculation and reached levels exceeding 107 and 106 IU/mL, respectively (figure 2A). Viral titres in stool suspensions were in general about 10-fold to 100-fold higher than in plasma. HEV RNA could not be detected in stool and plasma of non-transplanted mice inoculated with the same preparation.

Figure 2

Infection of humanised mice with human-derived or chimpanzee-derived hepatitis E virus (HEV) preparation. Human liver chimeric mice were inoculated intrasplenically with a stool suspension of an infected chimpanzee containing 3×106 IU (A) or 3×105 IU (B) HEV RNA (genotype 1, strain Sar-55). An additional group of mice was inoculated via the same route with a faecal suspension of a patient infected with HEV of gt3f (C). Viral RNA was quantified at different time points in 10% (w/v) faecal samples (black) and plasma (red). Solid lines represent data obtained from humanised mice, while dotted lines relate to data obtained from inoculated non-humanised control mice. The different symbols used indicate data obtained from different animals and allow the comparison of corresponding faecal and plasma data of a single animal. LOQ, limit of quantification.

A 10-fold lower dose of chimpanzee-derived Sar-55 (3×105 IU/mouse) also resulted in infection of all inoculated humanised mice (n=5) (figure 2B). HEV RNA could be detected in the stool of all mice from week 1 postinoculation and viral content sharply rose over time, indicating the robustness of the infection. HEV RNA could initially only be detected in plasma of three mice from week 2 onwards, but became detectable in the remaining two mice at weeks 5 and 6, respectively (figure 2B). Overall, there was a considerable delay in the kinetics of the viral infection compared with the mice that received a 10-fold higher inoculum dose.

Human albumin, ALT, AST and bilirubin levels were monitored at regular time points throughout the study and were not affected by the viral infection (data not shown).

Infection of humanised mice with patient stool and plasma

To investigate whether humanised mice could also be infected with patient-derived HEV of gt3, we inoculated two chimeric mice with a human stool suspension containing 3.95×105 IU/mouse. Both mice became infected as evidenced by the presence and increase of viral RNA in faecal suspensions. While the viral load in stool ultimately reached levels at around 105 IU/mL, HEV RNA could only be detected in the plasma of one of the two mice and at a later time point (figure 2C). Human albumin, ALT, AST and bilirubin levels were monitored at regular time points throughout the study and appeared not to be affected by the viral infection (data not shown). Non-humanised mice injected with the patient stool suspension and a single highly humanised mouse inoculated with plasma obtained from the same patient (dose 6.5×104 IU/mouse) never showed any signs of viral infection.

Passage experiments in humanised mice

To investigate the infectious nature of the viral particles identified in the faeces of Sar-55-infected humanised mice, a 10% filtered stool suspension was prepared and subsequently administered to naïve humanised mice. All mice (n=3) inoculated with 2.28×105 IU Sar-55 became infected (figure 3A). HEV RNA could be detected in stool and plasma of all mice from the first week postinoculation onwards and the viral titre increased over time indicating replication of the virus in vivo. After 4–5 weeks of infection, stool samples contained up to 1.2×107 IU/mL of HEV RNA, whereas plasma levels ranged from 6.9×103 to 1.2×106 IU/mL. One week after inoculation with a mouse Sar-55 faecal suspension, a minute amount of HEV RNA, close to the LOD of the assay, could be detected in the faeces of one non-humanised control mouse. Since no viral RNA could be detected in all subsequent samples, we assume this was either residual input RNA and/or contamination due to cohousing of this mouse with the three other humanised mice that already exhibited a high level of infection at that time point (figure 3A).

Figure 3

Hepatitis E virus (HEV) Sar-55 serial passage and titration experiments in humanised mice. A 10% stool suspension was prepared from a Sar-55-infected humanised mouse and used to inoculate naïve humanised mice. Each mouse was injected intrasplenically with a suspension containing 2.28×105 (A), 2.5×104 (B) or 1.2×103 (C) IU HEV RNA. Viral RNA was quantified at different time points in faecal samples (10% w/v suspensions, black symbols) and plasma (red). Solid lines represent data obtained from humanised mice, while dotted lines relate to data obtained from inoculated non-humanised control mice. The different symbols used indicate data obtained from different animals and allow the comparison of corresponding faecal and plasma data of a single animal. LOQ, limit of quantification.

A humanised mouse was also intrasplenically inoculated with plasma obtained from an HEV-infected mouse. One week after inoculation (105 IU HEV RNA), the faeces scored HEV RNA negative (<2×102 IU/mL), but at weeks 2 and 3, a specific signal corresponding to respectively 3.98×103 and 4.6×103 IU/mL was obtained. This mouse died a few days later, precluding further follow-up.

To determine the minimal infectious dose of the Sar-55 preparation in humanised mice, additional infection experiments were performed. Two mice were challenged intrasplenically with mouse stool containing 2.5×104 IU/mouse and readily became infected. HEV RNA could be detected in the stool of both mice from week 1 postinfection onwards, while viremia was delayed until week 3 postinfection. The stool viral titre of this group was lower than the titre observed in the three mice receiving 2.28×105 IU HEV, probably due to lower inoculum dose (figure 3B).

A further reduction of the inoculum dose to 1.2×103 IU/mouse resulted in infection of only one out of two mice. In the infected mouse, HEV RNA could be detected in stool but the titre was low (up to 104 IU/mL) (figure 3C). Viremia could only be detected 5 weeks after inoculation. The second mouse died 3 weeks after inoculation, but no HEV RNA could be detected in the stool during this period (figure 3C).

Finally, three humanised mice were inoculated orally with 2.28×105 IU/mouse, a dose that was previously shown to be 100% infectious via intrasplenic route. In none of these mice, HEV RNA became detectable during the 8-week observation period. Interestingly, intrasplenic inoculation of a lower dose (2.5×104 IU/mouse) into one of the unsuccessfully orally challenged mice resulted into infection, indicating that intrasplenic inoculation is much more efficient than oral gavage in establishing HEV infection in humanised mice.

In order to identify potential adaptation of the virus during passage to humanised mice, several faecal samples were selected for sequencing of the ORF2 and ORF3 region of the viral genome. These included samples from the first passage (3 samples collected at week 4 (#2) and week 10 (#1) after inoculation), second passage (inoculum used: passage 1, week 10; samples analysed: week 5 and week 9 after infection) and third passage (inoculum used: passage 2, week 9; sample analysed 5 weeks after infection). Population sequencing showed that all the viruses present in these samples had a completely identical sequence as the one present in the original chimpanzee inoculum.

Detection of viral RNA in the liver of infected humanised mice

Total RNA extracts were prepared from the liver of mice infected with either HEV of gt1 strain Sar-55 (n=2), patient-derived HEV of gt3f (n=1) or HEV p6 produced in cell culture (n=1). The mean HEV RNA load in a 30 mg fragment was 1.31×107 IU, 1.03×107 IU, 6.88×105 IU and 1.47×104 IU, respectively. These levels correlate with the viral titre in stool and plasma of these infected mice and indicate a higher level of viral replication of gt1 virus compared with gt3 viruses.

Using strand-specific PCR, we were able to unambiguously show the presence of replicative-intermediate negative-strand HEV RNA in the livers of infected humanised mice, but not in corresponding faecal samples (figure 4). Semiquantitative analysis indicated that the (+)/(−) strand RNA ratio was at least 1000 to 1 (data not shown).

Figure 4

Detection of (+) and (−) sense hepatitis E virus (HEV) RNA in the liver of infected humanised mice. Plus strand-specific (lanes 2–10) and minus strand-specific (lanes 11–19). RT-PCR was performed on liver extracts obtained from mice infected with HEV Sar-55 (lanes 2 and 11), patient gt3f (lanes 3 and 12) or gt3 HEVcc (lanes 4 and 13). Liver extracts from non-inoculated humanised mice (lanes 5 and 14), a stool sample from a gt1-infected mouse (lanes 6 and 15) and a gt3f-infected patient (lanes 7 and 16), extracts from gt3-infected HepG2 cells (lanes 8 and 17), in vitro transcribed (−) sense RNA (lanes 9 and 18) and non-template control (lanes 10 and 19) were included as positive and negative controls. Minus strand HEV RNA could specifically be detected in liver extracts of infected mice and infected cell lysate, while it was absent in corresponding faecal samples. Lane 1 shows the pattern of a 100 bp DNA ladder.

Detection of viral protein in the liver of infected humanised mice

Paraffin-embedded liver tissue sections from HEV-infected humanised mice, non-infected mice and mice infected with other hepatotropic pathogens (HBV, HCV and P. falciparum) were stained with a polyclonal antibody against HEV-ORF3. A clear and strong signal was detected specifically in the bile canalicular network formed by the human hepatocytes inside infected mouse livers (figure 5). Areas occupied by mouse hepatocytes stained completely negative, as did livers from non-infected mice and mice infected with other pathogens. The intensity of ORF3 staining in the liver correlated with the viral titre determined at time of sacrifice. The presence of ORF3 in infected livers was confirmed using western blot analysis (figure 6). Interestingly, the molecular weight of ORF3 protein present in mouse liver was slightly higher compared with that present in the lysate of an infected cell culture.

Figure 5

Visualisation of ORF3 protein in hepatitis E virus (HEV)-infected humanised mouse livers using immunofluorescence. Within the liver of a chimeric mouse, areas repopulated by human hepatocytes can be easily discerned from residual mouse parenchyma by the larger size and paler colour of the former (A). Liver sections prepared from a non-infected humanised mouse (B), an HCV-infected mouse (D–F) and an HEV Sar-55-infected mouse (C, G–K) were stained with a polyclonal antibody against HEV ORF3 protein (red fluorescence) and an antibody that specifically binds human CEACAM-1, a marker for bile canaliculi (green fluorescence). Panels (D) and (G) only show the CEACAM-1 staining, while panels (E) and (H) only show ORF3 staining. An overlay of CEACAM-1 and ORF3 staining is shown in panels (F) and (I). ORF3 could be specifically observed in areas repopulated with human hepatocytes, while mouse areas stained negative. The staining pattern indicates the presence of HEV ORF3 in the bile canalicular network formed by human hepatocytes. Panels (J) and (K) represent higher magnifications of HEV-infected areas stained with anti-ORF3 (red), anti-CEACAM-1 (green) and Hoechst (blue).

Figure 6

Visualisation of ORF3 protein in hepatitis E virus (HEV)-infected humanised mouse livers using western blot analysis. Protein extracts from the liver of two Sar-55-infected mice (lane 1 and 2) and one uninfected control mouse (lane 3) were separated on a 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane. ORF3 protein was visualised by overnight staining with a polyclonal antibody. A protein extract from an HEV-infected cell culture was used as a positive control (lane 4).

Density gradient analysis of HEV particles

Given the differences in infectivity between faeces-derived and plasma-derived virus and the different cell culture-derived preparations, we analysed the buoyant density of the viral particles using an iodixanol gradient. While most of the viral particles present in the mouse plasma had a buoyant density of 1.11 g/mL, the density of the majority of faecal-derived particles was about 1.16 g/mL (figure 7A). Our buoyant density gradient analysis also showed that HEVcc isolated from culture supernatant had a similar density profile as plasma-derived virus, while HEVcc from cell lysate resembled more the virus present in mouse stool (figure 7B). HEVcc that was treated with sodium deoxycholate/trypsin seemed to contain two populations, one with a low and one with a high buoyant density (figure 7B).

Figure 7

Density gradient analysis of different hepatitis E virus (HEV) preparations. (A) Viral particles present in the plasma (black line) and faeces (red line) of infected mice were separated in 12 equal fractions using iodixanol gradient centrifugation. The density as well as the HEV RNA content was determined in each fraction. (B) Three different preparations of cell culture produced virus were analysed: supernatant (black), treated supernatant (green) and cell lysate (red). After iodixanol gradient centrifugation, each preparation was separated into 12 fractions and both the density and HEV RNA content was determined. Since the amount of virus that was layered on the gradient differed from sample to sample, panel (A) and (B) show the relative HEV RNA content of each fraction normalised to the level of the fraction containing the highest viral load. HEVcc, culture-derived HEV.

HEV influences host genome expression

To evaluate whether HEV infection has an influence on the gene expression of the infected hepatocyte, the gene expression profile of liver tissue of an uninfected mouse was analysed and compared with that of two animals that were infected for 4 and 9 weeks with gt1 Sar-55. This analysis was restricted to genes involved in the immune response and only human-specific primer/probe sets were used that did not cross-react with mouse genes. At week 9 of infection, expression of CXCL9 and CXCL10 was increased 24.3-fold and 8.7-fold, respectively. CXCL9 and CXCL10 are two chemokines involved in leucocyte stimulation and trafficking, and adhesion molecule expression (table 1). Also a twofold to threefold increase of HLA-A, HLA-B, HLA-F and HLA-J was observed after 9 weeks of infections. In addition, several other interferon-induced genes were upregulated, such as IFI27, IFI44L, IFIT1, IFI2, IFIT3, ISG20, OAS2, OASL, RSAD2, TAP1 and TRIM22 (table 1). Overall, a more pronounced effect was observed at week 9 of infection compared with week 4. Numerous genes involved in the innate immune response were not detectably affected (<twofold) by the infection. These include USP18, TLR3, ISG15, STAT1, STAT2, OAS1, IRAK1, MX1, MAVS, etc (table 1).

Table 1

Host gene expression analysis of HEV-infected livers

RBV treatment of HEV-infected mice

Finally, we wanted to validate the human liver chimeric mouse model for the evaluation of novel antiviral compounds. Since RBV is currently the first-choice therapy for chronic HEV infection, we used this molecule for antiviral therapy of infected chimeric mice.

First, a dose-escalation experiment was conducted in two infected chimeric mice. One mouse was used as control and received vehicle only while the second animal was given a 1-week oral RBV treatment of 25 mg/kg once a day. As shown in figure 8A, a slight decrease in stool viral titre was observed (−0.51×log10). An additional 2-week treatment at 50 mg/kg had no major effect. A further dose increase to 100 mg/kg resulted in a sharp 1.60×log10 drop in viral titre within 6 days. However, since this dose was not well tolerated due to anaemia, we decided to treat a larger mouse cohort with a daily RBV dose of 50 mg/kg.

Figure 8

Ribavirin (RBV) treatment of hepatitis E virus (HEV)-infected mice. (A) A pilot study was performed to determine the optimal RBV dose. Repetitive dose-escalation indicated that the highest response was observed at 100 mg/kg daily. Because this dose was poorly tolerated, 50 mg/kg was used in the subsequent studies. (B) Five mice were treated with RBV (red lines) and viral RNA was quantified at different time points in the faeces (solid line) and blood (dotted line). Two infected mice received vehicle only and served as negative controls (black lines). Upon RBV treatment, a considerable drop in viral load was observed, both in the plasma and faeces. In one animal (indicated with a rhombus), the drop in faecal viral titre was much less pronounced, although there was a 3×log10-fold drop in viremia. An increase in RBV dose to 100 mg/kg (blue line) resulted in a 0.72×log10-fold drop in faecal HEV titre within 5 days. Two mice were followed-up after therapy cessation and a rebound in viral titre was observed. (C) Overall, a 14-day RBV treatment at 50 mg/kg resulted in an average drop in faecal viral titre of 1.64×log10 IU/mL (n=3).

A total of five HEV-infected humanised mice received a 2-week RBV therapy (figure 8B). One mouse died after 7 days of treatment, whereas a second mouse died after 12 days, most likely as a consequence of RBV-induced anaemia. In the latter case, a 0.66×log10 reduction in faecal viral titre compared with baseline was observed. While the faecal viral titre in two control mice remained stable (average change of −0.13 and +0.21×log10 IU/mL after 7 and 14 days, respectively), a more than 10-fold decrease (n=4) was observed after 10 days in treated mice (figure 8C). An average decrease of −1.79×log10 IU/mL (n=3) was observed at the end of treatment (EOT). In one animal, no change in faecal viral titre was observed between days 10 and 14. However, an increase of RBV dose to 100 mg/kg resulted in an additional −0.72×log10 IU/mL within 5 days (figure 8B). RBV therapy did not have a negative impact on human albumin levels, indicating absence of RBV-induced hepatotoxicity (data not shown).

The decrease in faecal viral content coincided with the decline in viremia (figure 8B). The two animals that died before EOT experienced a 0.78 to 2.47×log10 drop in viremia, reaching levels below the LOQ. Of the three remaining mice, two reached a plasma viral load at or below LOQ at EOT (−1.29 and −3.04×log10 IU/mL), while the drop in the third animal was less pronounced (−0.57×log10). A less pronounced ORF3 staining was also observed in the liver of a RBV-treated mouse (data not shown). Follow-up was performed on two mice that reached EOT. After cessation of therapy, the faecal viral titre increased 1.35 and 2.19×log10-fold, respectively, while viremia became detectable but unquantifiable in one mouse and increased at least 1.68×log10-fold in the other mouse.


The lack of a small animal model and efficient cell culture systems has hampered the study of HEV replication and pathogenesis. Since HEV appears to emerge in industrialised countries and viral isolates with decreased sensitivity to RBV become prominent, these tools will be increasingly needed for the identification and preclinical evaluation of novel antiviral compounds.

Here, we show that human liver chimeric mice are suitable for the study of HEV replication, virus–host interactions and the evaluation of antiviral therapies. To evaluate the susceptibility of humanised mice to HEV infection, viral preparations of different origins have been examined. The initial infection studies were performed with cell culture-derived virus of gt3, strain Kernow C1 P6. Only 3 out of 15 challenged mice became infected and viral titres were low in stool specimens and negative in plasma samples. The low viral replication of this particular strain may be explained by its adaptation to grow in cell culture and/or to a recombination event resulting in an insertion of 174 ribonucleotides (58 amino acids) of the human ribosomal protein S17 gene into the viral genome.6 Interestingly, infections were only successful if the mice were challenged with viral particles that were pretreated with trypsin and sodium deoxycholic acid (1/5); or that originated from cell lysates (2/3). This is most likely related to the differences in particle composition between intracellular and secreted virus. It has been reported that secreted viral particles are associated with lipids, while intracellular particles and particles treated with trypsin/sodium deoxycholic acid, which mimics the transformation secreted particles undergo during passage through the bile duct and intestine, are not associated with lipids.7 ,21–24 Our density gradient analyses of the different viral inocula confirm this hypothesis and highlight that differences in particle origin and composition may have a profound impact on viral infectivity. Further studies are needed to ascertain in detail the interplay between particle composition and viral infectivity.

Infections with patient (gt3f) and chimpanzee (gt1, Sar-55) stool suspensions were more successful. All intrasplenic inoculations lead to active infection of the humanised mice, evidenced by the presence and increase over time of HEV RNA in the stool; and of viremia in those mice with highest faecal titre. The viral content was always much higher in mouse stool compared with the plasma, like it has been observed in chimpanzees, cynomolgus monkeys, pigs and humans.25–28 HEV RNA could also be detected much earlier in the stool than in plasma, which is in line with what was observed in experimentally infected chimpanzees.25 The infection was further corroborated by the presence of both positive-strand and negative-strand viral RNA and viral protein in the infected mouse liver and also by successful serial passage of the infection from an infected humanised mouse to a naïve one. Western blot analysis indicated that the molecular weight of ORF3 protein present in mouse liver was slightly higher compared with that present in lysate of an infected cell culture. To investigate whether this was instigated by adaptation (mutation) of the virus during passage in humanised mice, we sequenced the complete ORF3 gene of the viral genome. However, the ORF3 region showed complete conservation (data not shown), thereby excluding genome mutation as the cause of this different migration pattern. Alternatively, the difference may be explained by differences in post-translational modifications, but further study is needed to corroborate this.

Comparison of peak viral titres between gt1 and gt3 infections indicated much higher replication of the former. This is in agreement with previous publications and clinical data indicating that gt1 viruses are more virulent than gt3 viruses.10 ,29 ,30 Therefore, the low replication rate in humanised mice observed with cell culture-derived gt3 virus may be related to cell culture adaptation, and may (in part) also be attributed to certain intrinsic characteristics of gt3 viruses.

Our attempt to infect humanised mice using HEV-positive patient plasma was unsuccessful. This may be related to the physical composition of the particles and/or to the presence of HEV-specific antibodies (10.8 U/mL), as determined by the Wantai HEV IgG-ELISA.31 Using a slightly higher dose of HEV-positive mouse-derived plasma, which is devoid of antibodies, an HEV infection could be established in a naive mouse, but the magnitude and kinetics of the infection were much inferior to what was observed in mice that were inoculated with a similar or lower dose of the same viral strain derived from faeces. This result confirms the risk of HEV transmission by transfusion of HEV-positive blood products, especially during the early phase of infection where the blood donor has not yet developed anti-HEV antibodies or when they are still low,4 but indicates that the infectivity of viral particles in plasma appears to be much lower than that of faecal-derived particles. It has been reported that faecal viral particles have a higher buoyant density (1.27–1.28 g/mL) than HEV particles present in plasma (1.15–1.16 g/mL),23 a phenomenon that is most likely due to the association of lipids to the latter. This was also corroborated by our density gradient analysis of infected mouse plasma and faeces. The viral particles present in faeces clearly had a higher buoyant density than those present in the mouse plasma. Similar to the different virus preparations originating from cell culture, high-density particles seem to be highly infectious, while low-density particles appear to be less infectious.

We also attempted to infect humanised mice via the oral route, the presumed natural route of HEV transmission. However, while a single oral dose of 2.28×105 IU/mouse was not successful, a 10-fold lower intrasplenic dose was sufficient to establish an infection in the same mouse. This indicates that the failure of oral transmission was not related to the level of mouse liver chimerism. Similar findings were reported by Tsarev et al, who showed that oral inoculation of cynomolgus monkeys required at least a 10 000-fold higher dose than intravenous inoculation to establish an HEV infection.32 Likewise, pigs could be experimentally infected with a single intravenous injection of as little as 102 genome equivalents, while multiple high-titered oral inoculations, containing more than 106 genomes, established infection in only one third of the animals.33 ,34 These data clearly show that, although it is considered to be the natural way of infection, transmission via the oral route is not efficient. On the other hand, we cannot rule out that oral transmission requires certain essential (human) receptors that are lacking or are different in the mouse intestine, which is not humanised in this chimeric animal model.

We also evaluated the direct influence of HEV infection on gene expression in the infected hepatocytes. For this purpose, these humanised mice are a unique tool because they lack a functional adaptive immune system, which otherwise also would influence the host's gene expression profile. Our data show that a large number of ISGs are directly upregulated upon infection. These include chemokines, molecules involved in antigen presentation and several proteins known to inhibit viral translation and replication. All the genes, except for TAP1 and OASL, which were upregulated in our mouse model were also upregulated in chimpanzees that were infected with the same viral isolate.35 ISG15, STAT1 and Mx1 were upregulated in chimpanzees, but are here reported as unaffected, mainly because their fold change was below 2 (1.90, 1.62 and 1.88, respectively). IFI6 was highly upregulated in chimpanzees (up to 10-fold), but remained completely unaffected in mice. This gene plays a critical role in the regulation of apoptosis, but the cause of this discrepancy remains unclear. One must also keep in mind that the chimpanzee data do not only reflect gene expression changes inside the hepatocyte that are directly triggered by the viral infection, since it also encompasses the influence of and changes in activated cells of the adaptive immune system. In addition, the changes may also be the consequence of differences between human and chimpanzee hepatocytes. Nevertheless, despite this robust activation of the innate immune response, the viral infection is not spontaneously cleared. Further study is needed to evaluate whether HEV can specifically interfere with the action of certain ISGs.

While the absence of an adaptive immune response allows us to study the changes in host gene expression that are directly provoked by the viral infection, it evidently prevents the study of cell-mediated immune responses. Novel animal models possessing both a human liver and human immune system are currently in development, but still suffer from a lack of or suboptimal cross-talk between the liver and immune compartment.15 Nevertheless, the activity of (neutralising) antibodies and specific T cell clones can be studied in this model using passive immunisation and adoptive transfer experiments, respectively.

The lack of adaptive immune system in our model probably explains why we observed a chronic-type infection with continuous increase to a high viral titre, both for gt1 and gt3 infections, rather than an acute self-resolving course with clear peaks during the early phase of the infection. This indicates that the adaptive immune system must be the major contributor to control and clearance of an acute HEV infection in humans. Likewise, the lack of a direct hepatotoxic effect of the virus in our model indicates that the liver disease observed during HEV infection is mainly caused by destruction of liver tissue by the activated T lymphocytes.36

Recently, Geng et al reported that urine from HEV-infected individuals and monkeys contained viral protein and RNA.37 The ratio of viral capsid protein to RNA was much higher in urine than in plasma and faeces, indicating the presence of especially free antigen and/or empty viral particles. Nevertheless, transmission experiments in cynomolgus monkeys showed that HEV RNA-positive urine samples did contain infectious particles. However, it remains unclear whether viral replication in kidney cells is a prerequisite and whether infiltration of inflammatory immune cells in the kidney is a contributing factor. Our mouse model could contribute to the study of the relationship between HEV infection and renal disease. Similar studies could be performed to investigate (1) the presence of HEV RNA in urine of humanised mice, (2) the characteristics of (sub)viral particles in urine, (3) the necessity of viral replication and protein expression in kidney cells and (4) the potential contribution of the adaptive immune system (which are absent in the humanised mouse model).

Finally, we aimed to validate the chimeric mouse model for the evaluation of novel antiviral strategies. RBV is currently the first-choice antiviral treatment in chronically infected patients with HEV.38 A therapy of at least 3 months is usually sufficient to eliminate the virus.39 ,40 We were able to show a dose-dependent reduction of viral load upon RBV treatment. A 2-week RBV therapy resulted in an average 1.79×log10-fold reduction in faecal viral load, which coincided with an even more pronounced reduction in viremia. During therapy, human albumin levels remained relatively constant, indicating that the drop in viral load was specifically caused by the antiviral activity of RBV and was not the result of unspecific hepatotoxicity. This is also supported by the viral rebound that was observed upon cessation of RBV therapy. A few mice died during RBV treatment, which was not surprising since RBV is well known to induce haemolytic anaemia, especially at high doses as used here in our study (50–100 mg/kg). At the end of therapy, the plasma viral titre was below the LOD, while HEV RNA still could be detected in mouse stool. These results are similar to clinical data. Ambrosioni et al reported that during prolonged RBV therapy of a leukaemia patient with chronic HEV, viral RNA became undetectable in the plasma after 6 months, while viral particles were continuously shed in the faeces for an additional 6 months.41 In another study, 3 months of RBV therapy of 24 organ transplant patients chronically infected with HEV led to 100% clearance of the virus from plasma, while HEV RNA could still be detected in stool in 25% of the patients.40 Abravanel et al reported that 9 out of 24 patients showed a rebound after cessation of a 3-month RBV therapy.40 At the end of the treatment, HEV RNA was still detectable in the stool of six out of these nine patients, while it was undetectable in the plasma.

We here report the establishment and characterisation of HEV infection in mice with humanised liver. This model is well-suited for the in vivo study of the fundamental aspects of the HEV life cycle and the interactions between the virus and its host. Furthermore, it can be used for the preclinical in vivo evaluation of novel antiviral therapies.


The authors are grateful to Dr Suzanne U Emerson (Molecular Hepatitis Section, Laboratory of Infectious Diseases (LID), National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH)) and Dr Robert H Purcell (Hepatitis Viruses Section, LID, NIAID, NIH) for providing essential reagents; Muriel Minet for expert technical assistance in immunostainings and Professor Dr Véronique Stove for the quantification of liver enzymes and bilirubin levels.


View Abstract


  • Twitter Follow Lander Foquet at @lfoquet

  • Contributors IMS, LV, LC, LF, CM, YD, AF, CW, TM and PM designed the experiments. IMS, LC, FA, LF, CM, AF, CW, JD, GL-R, JN, JI, TM and PM analysed data. IMS, LV, FA, LF, CM and YD performed experiments. IMS and PM wrote the manuscript. PM conceived and supervised the study.

  • Funding This study was funded by the Ghent University (Concerted Action Grant 01G01712 and IRO project MODEL-HEPE), the Lille 2 University (IRO project MODEL-HEPE), The Research Foundation—Flanders (FWO-Vlaanderen; project G0D2715N), the Agency for Innovation by Science and Technology (IWT SBO project HLIM-3D), the Belgian Science Policy Office (BELSPO; IUAP P7/47-HEPRO-2 and P7/45-BELVIR), the ‘Agence Nationale de Recherches sur le Sida et les hépatites virales’ (ANRS) and the European Union (FP7, HepaMab), IMS is a recipient of a PhD fellowship provided by the Egyptian Government. LF was supported by a PhD fellowship of the Agency for IWT. YD received a PhD fellowship from the FWO.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.