Release of Genotype 1 Hepatitis E Virus from Cultured Hepatoma and Polarized Intestinal Cells Depends on Open Reading Frame 3 Protein and Requires an Intact PXXP Motif

Abstract

Hepatitis E virus genotype 1 strain Sar55 replicated in subcloned Caco-2 intestinal cells and Huh7 hepatoma cells that had been transfected with in vitro transcribed viral genomes, and hepatitis E virions were released into the culture medium of both cell lines. Virus egress from cells depended on open reading frame 3 (ORF3) protein, and a proline-rich sequence in ORF3 was important for egress from cultured cells and for infection of macaques. Both intracellular ORF3 protein accumulation and virus release occurred at the apical membrane of polarized Caco-2 cells. ORF3 protein and lipids were intimately associated with virus particles produced in either cell line; ORF2 epitopes were masked in these particles and could not be immunoprecipitated with anti-ORF2.

Hepatitis E virus (HEV) remains enigmatic in spite of recent advances (see references 7 and 16 for reviews). HEV is a major cause of acute hepatitis in numerous developing countries, but hepatitis E is infrequently detected in industrialized countries even though seroprevalence rates of anti-HEV as high as 20% in these countries have been reported. Although hepatitis E normally is a self-limited acute disease, recent studies have identified it as an emerging cause of chronic hepatitis in immunocompromised patients. Whereas contaminated drinking water is the source of most infections in developing countries, the sources in industrialized countries are not fully evaluated, but many, if not most, infections appear linked to eating undercooked meat, especially pork. These differences in epidemiology may reflect the fact that most infections in developing countries are caused by genotypes 1 and 2 while those in industrialized countries are mainly due to genotypes 3 and 4.

HEV was initially classified as a calicivirus, but subsequent sequence analysis suggested that it was more closely related to the enveloped rubella virus. However, although HEV may be associated with lipids under some conditions (22), HEV virions do not possess an envelope. Four genotypes of HEV that infect humans have been identified (4). Genotypes 1 and 2 infect primates exclusively, whereas genotypes 3 and 4 are zoonotic and commonly also infect swine and rarely other nonprimates. Recent identification of a strain infecting farmed rabbits in China suggests that other reservoirs may exist (32).

The capsid protein encoded by open reading frame 2 (ORF2) is able to form infectious virus particles, but these particles remain cell associated. The crystal structure of a truncated recombinant protein has been solved, but the size of the protein in mature virions is unknown (11, 15, 28, 31). The virus is not cytopathic, and it is unclear how it gets out of cells.

The 7.2-kb genome of HEV is a capped mRNA that contains three ORFs that encode proteins involved in replication (ORF1), a capsid protein (ORF2), and a small protein of only 113 to 114 amino acids (ORF3). All but the 5′ terminus of ORF3 is overlapped by ORF2, and both proteins are translated from the same bicistronic subgenomic RNA (10). When overexpressed in cell culture, ORF2 is glycosylated, and ORF3 is phosphorylated (26); this phosphorylated ORF3 protein binds to nonglycosylated ORF2 protein in cell culture, but phosphorylation is not required for infection of macaques (9). The virus has been exceedingly difficult to propagate in cell culture, but recently Okamoto and colleagues reported the successful adaptation of both a genotype 3 and a genotype 4 strain to efficient growth in cultures of PLC/PRF/5 hepatoma or A549 lung cells (23, 24).

The tiny ORF3 protein is particularly intriguing because it has a significant impact on virus propagation through mechanisms that have yet to be defined. Data from experiments performed with overexpressed ORF3 protein have suggested that, among other things, ORF3 may interact with cellular proteins, including signaling proteins containing Src homology 3 domains (14), bikunin (27), hemopexin (21), and microtubule proteins (13), and it may function to modulate the acute-phase disease response (3), protect cells from mitochondrial depolarization (18), and enhance expression of glycolytic pathway enzymes (17). Yet within transfected hepatoma cells in culture, virions of an ORF3 null mutant of genotype 1 were assembled in the absence of ORF3 protein and were infectious for naïve hepatoma cells (6) although this same ORF3 null mutant was unable to mount a detectable infection in rhesus monkeys (8). Also, swine transfected with genotype 3 mutant genomes encoding a truncated ORF3 protein did not get infected, indicating that an intact ORF3 protein is needed for infectivity in vivo (12). This lack of infectivity in vivo is possibly explained by the recent demonstration that the ORF3 protein of genotype 3 virus is important for export of virions out of cultured cells in vitro (30); however, this dependence on ORF3 for virion egress has not been confirmed in vivo or for strains of the other three genotypes.

The four major genotypes of human HEV appear to segregate naturally into two distinct groups. One group contains genotype 1 and 2 strains that lack a zoonotic component and are spread mainly via contaminated water; in contrast, the second group contains genotype 3 and 4 strains which are able to cross species boundaries and are zoonotic since humans have been infected as a result of eating undercooked meat (16, 25). The molecular basis for the two groupings is unknown, and much more extensive comparative analyses are required to determine which variables are epidemiologically relevant. Here, for lack of an efficient cell culture system for genotype 1 or 2 strains, we have utilized an infectious cDNA clone of a genotype 1 strain in order to explore the role of the ORF3 protein in this group.

MATERIALS AND METHODS

Cells.

Caco-2 (HTB37) cells and HepG2/C3A (CRL-10741) cells were purchased from the American Type Culture Collection (Manassas, VA). S10-3 cells are a sublcone of Huh-7 cells isolated in-house by limiting dilution; Huh-7 cells were originally isolated in Japan (19). Cells were grown as monolayers in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (S10-3 and HepG2/C3A) or 20% (Caco-2) fetal bovine serum (ultralow immunoglobulin G[IgG]; Invitrogen) that had been heat inactivated at 56°C for 30 min. Generally, 0.1 mg of gentamicin ml⁻¹, 100 U of penicillin ml⁻¹, and 0.1 mg of streptomycin ml⁻¹ were included. Stocks were maintained at 37°C; infected or transfected cultures were incubated at 34.5°C. The Caco-2 cell subclone C25j (see Results) was plated on flasks or slides coated with rat tail collagen type 1 (Millipore). Caco-2 cell monolayers were incubated with EGTA prior to trypsinization to aid in dispersal.

Plasmids.

Plasmids were described previously (8). pSK-E₂ (GenBank accession number AF444002) encodes the infectious HEV Sar55 strain. The ORF3 null mutant of pSK-E₂ was generated by mutating two nucleotides in the second codon in ORF2 while preserving the amino acid in ORF2 and introducing a termination codon in ORF3 (9); it encodes authentic ORF2 protein and an ORF3 peptide of 5 amino acids initiated at the third Met codon in ORF3. ORF3 point mutations altered the middle nucleotide (nucleotide position 5426 of Sar55) of a proline codon within a PXXP motif and did not change the amino acids encoded by ORF2. The entire sequence of each mutant was verified.

Microscopy.

Since ORF2 and ORF3 proteins are not produced in the absence of viral RNA replication (5), their detection by immunofluorescence microcopy serves as an indirect assay for replication. The same primary and secondary antibodies and blocking solutions were used for indirect and confocal microscopy. For indirect microscopy, cells in eight-well chamber glass slides (Lab-Tek II CC²) were fixed with acetone and washed in phosphate-buffered saline (PBS). A mixture of rabbit polyclonal antibody specific for ORF3 protein and chimpanzee polyclonal antibody specific for ORF2 protein in blocking solution (0.5% bovine serum albumin [BSA], 0.5% dry skim milk, 0.1% Triton X-100 in PBS) and samples were incubated at room temperature for 20 min. The antibodies were described previously (9). The anti-ORF2 was derived from a chimpanzee that had been infected with the Sar55 strain of HEV and boosted 3.5 years later by inoculation with the Mex 14 strain. The reciprocal anti-ORF2 enzyme-linked immunosorbent assay (ELISA) titer of 10,000 increased 5-fold after the boost. The anti-ORF3 was produced by immunizing a rabbit with a synthetic peptide comprising amino acids 91 to 123 of the human HEV strain Sar55. After a wash in PBS, slides were incubated for 20 min at room temperature with a mixture of Alexa Fluor 488-conjugated goat anti-human IgG (Molecular Probes) and Alexa Fluor 568-conjugated goat anti-rabbit, as described previously. Samples were washed in PBS, Vectashield (Vector Laboratories) was added, and the slides were viewed by indirect immunofluorescence microscopy with a fluorescein isothiocyanate (FITC) filter set for Alexa Fluor 488 (green), a rhodamine filter set for Alexa Fluor 568 (red), and the 40× objective of a Zeiss Axioskop 2 Plus fluorescence photomicroscope.

For confocal microscopy, S10-3 and Caco-2 cells were grown on transwell inserts (Corning, Lowell, MA) and triple labeled for ORF2 (green), ORF3 (red), and cell nuclei (with 4′,6′-diamidino-2-phenylindole [DAPI]; blue). The cells were washed briefly in PBS, fixed in 4% formaldehyde, and permeabilized in 0.2% Triton X-100. Cells were blocked for 30 min at room temperature in blocking solution, incubated in a mixture of primary antibodies for 45 min at room temperature, rinsed in PBS, and incubated with secondary antibodies for 45 min at room temperature. Cells were rinsed in PBS and mounted on a glass slide with Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA). Images were collected along the z axis using a Leica SP5 inverted five-channel confocal microscope (Leica Microsystems, Exton, PA), 63× oil immersion objective, and a zoom of ×4.92. Fluorochromes were excited using an argon laser at 488 nm for Alexa Fluor 488 and a helium-neon laser at 594 nm for Alexa Fluor 568. DAPI was excited using a 405-nm-diode UV laser. To minimize cross talk between channels, the three wavelengths were collected separately (Alexa Fluor 488 emission was collected between 496 and 580 nm, Alexa Fluor 568 emission was collected between 602 and 700 nm, and UV was collected between 415 and 470 nm) and later superimposed. Images were processed using Leica LASAF (version 2.2.1) software, Imaris (version 6.3.1; Bitplane AG, Zurich Switzerland), and Adobe Photoshop CS, version 8.0 (Adobe Systems).

In vitro transcription and transfection.

Plasmids were linearized at a unique BglII site located immediately downstream of the poly(A) tract of the HEV sequence. Capped transcripts were synthesized with a T7 riboprobe in vitro transcription system (Promega) in the presence of cap analog, as previously described (9). Each 50-μl reaction mixture contained 10 μl of 5× transcription buffer, 5 μl of 100 mM dithiothreitol, 2 μl of 40 U/ml RNasin, 5 μl of the nucleoside triphosphates (5 mM [each] ATP, CTP, and UTP and 0.5 mM GTP), 5 μl of 5 mM 3′-O-Me-m⁷G(5′)pppG (Ambion), and 2 μl of 20 U/μl T7 polymerase. The mixtures were incubated at 37°C for 1.5 h. The integrity and yield of transcripts were determined by electrophoresis of 2 μl of reaction mixture on a nondenaturing agarose gel. Transcription mixtures were cooled on ice and then mixed with a liposome mixture (20 μl of 1,2-dimyristyloxypropyl-3-dimethyl-hydroxy ethyl ammonium bromide and cholesterol [DMRIE-C; Invitrogen] per ml of OptiMEM [Gibco]) for transfection. A total of 23 μl of RNA mixture was diluted with 1 ml of the liposome mixture and added to a T25 flask containing washed S10-3 cells at 40 to 50% confluence or Caco-2 cells at 10 to 20% confluence. Flasks were incubated at 34.5°C for 5 h.

Medium was aspirated and replaced with growth medium. Cells were trypsinized and split into two T25 flasks and one to two wells of an eight-well chamber slide 1 to 2 days prior to immunostaining.

Density gradients.

OptiPrep (AxisShield, Norway), a 60% solution of iodixanol in water, was diluted with OptiMEM medium (Gibco) to prepare high- and low-density solutions. Preformed gradients were made with a gravity flow gradient maker containing 2.5 ml of 10% iodixanol in the rear chamber and 2.5 ml of 36% iodixanol in the forward chamber. A 100-μl virus sample was layered on top, and tubes were centrifuged at 4° for 17 h at 35,000 rpm in an SW55 Beckman rotor. Fractions (10 drops or 500 μl) were collected by aspiration from the bottom; the pellet fraction was resuspended in 500 μl of PBS. Aliquots (100 μl) were weighed to determine density. A total of 200 μl of each fraction was extracted with Trizol LS (Invitrogen) according to the manufacturer's directions, precipitated with isopropanol, and resuspended in 10.4 μl of water. Five microliters of each sample was assayed for HEV and hepatitis C virus (HCV) RNA by virus-specific reverse transcription-PCR (RT-PCR).

Preparation of cell lysates and culture medium.

Confluent monolayers of cells in a T25 flask were trypsinized and centrifuged in a 5415C Eppendorf centrifuge in a 2.0-ml Sarstadt tube for 2 min at 2,000 rpm. Supernatant was aspirated, and the cell pellet was stored at −80°C. Frozen pellets were extracted at room temperature by adding 0.9 ml of water per T25 pellet and vortexing vigorously until the pellet dispersed, and the solution became cloudy. The sample was vortexed once or twice more in the next 10 min, 0.1 ml of 10× concentrated PBS was added, and debris was removed by centrifugation at 16,000 × g for 2 min. If the lysate was treated with NP-40, Complete protease inhibitor (Roche) was dissolved in the water, and centrifugation was decreased to 200 × g for 5 min. The supernatant was removed and layered over 3.5 ml of a 13.3% solution of iodixanol made by mixing 2 ml of stock OptiPrep (AxisShield, Norway) with 7 ml of DMEM. A 300-μl pad of stock OptiPrep was layered on the bottom, and the mixture was centrifuged at 30,000 rpm for 2 h in an SW55 Beckman rotor. The bottom 500 μl, which contained semipurified virus particles, was collected. Western blot analysis demonstrated that this fraction contained mostly full-length ORF2 protein, whereas shorter forms remained in the top layer.

Medium from transfected cell cultures was filtered through a 0.45-μm-pore-size polyvinylidene difluoride (PVDF) Millex-HV filter (Millipore) and then frozen at −80°C. Samples were thawed at room temperature.

Infectivity assay.

In our hands, HepG2/C3A cells were infected much more efficiently than S10-3 or Caco-2 cells, so they were used for infectivity determinations. Confluent monolayers of HepG2/C3A cells were trypsinized and diluted 1 to 4 in growth medium, and 0.4 ml was added per well of an eight-well glass chamber slide coated with rat tail collagen. The slides were incubated at 37°C for 1 day. Duplicate 100-μl samples of inoculum were coded and used to replace the medium in wells of the chamber slide. Slides were incubated at 34.5°C in a 5% CO₂ atmosphere for 5 h; liquid was aspirated and replaced with 0.4 ml of growth medium containing antibiotics and 2% dimethyl sulfoxide (Hybri-Max; Sigma). After 5 to 6 days at 34.5°C, cells were fixed with acetone and stained for immunofluorescence microscopy, and the number of cells positive for ORF2 protein was tabulated. The code was not broken until all samples in the experiment had been scored.

Transfection of rhesus macaques.

The transfection method was as described previously (9). Briefly, capped in vitro transcripts were injected into multiple sites in the liver of a rhesus macaque by percutaneous intrahepatic injection guided by ultrasound. Each mutant was tested in two macaques, and weekly serum samples were collected. Alanine amino transferase (ALT) levels were determined at Anilytics, Gaithersburg, MD. HEV antibodies were detected with a sensitive in-house ELISA. Viral genomes were quantified by HEV-specific real-time RT-PCR from total RNA extracted from 100 μl of serum with Trizol LS (Invitrogen). Genome fragments amplified by RT-PCR from serum were purified after agarose gel electrophoresis (QiaQuick; Qiagen) and sequenced directly with an automated sequencer.

The animals were housed at Bioqual (Rockville, MD). The housing, maintenance, and care of the animals met or exceeded all requirements for primate husbandry as specified in the Guide for the Care and Use of Laboratory Animals (20).

Immunoprecipitation.

For treatment with nonionic detergent, NP-40 (Fluka) was first added to the virus preparation at a ratio of 71 μl of 25% NP-40 per ml of virus. Thirty microliters of semipurified intracellular virions, filtered cell culture medium, or 10% stool suspension was mixed with 20 μl of the same chimpanzee anti-ORF2 or rabbit anti-ORF3 used for immunofluorescence microscopy except that the chimpanzee plasma had been clotted with calcium, and the expressed liquid was used. Rabbit serum raised against a glycoprotein of hepatitis C virus was used as a negative control. Antibody-antigen mixtures in triplicate were incubated at 37°C for 1 h and then at 4°C overnight. A 200-μl slurry of washed GammaBind Plus beads (GE Healthcare, Sweden) in PBS was added, and the samples were rocked for 1 h at room temperature. Beads were washed with PBS five times and resuspended in 100 μl of PBS prior to extraction of RNA with Trizol LS (Invitrogen). Extracted RNA genomes were quantified by RT-PCR (TaqMan). A mixture of 30 μl of original sample and 70 μl of PBS was extracted to estimate the number of virions present at the start. The Student's t test (unpaired with 4 degrees of freedom) was performed to determine whether observed differences were significant; P values less than 0.05 were considered significant.

RESULTS

Sar55 replicates in Caco-2 cells.

Previously, we reported that genotype 1 recombinant Sar55 genomes, when transfected into S10-3 hepatoma cells, replicated and assembled into infectious virions, but virus egress was limited, and the majority of virus remained intracellular (6). Therefore, in a search for a more productive cell line, we transfected the Caco-2 cell line of intestinal cells with Sar55 genomes transcribed in vitro. Caco-2 cells were selected since they support the replication of hepatitis A virus (2), the other enterically transmitted hepatitis virus, and, given the fecal-oral transmission route of HEV, intestinal cells were a conceivable substrate for HEV infection in vivo. Additionally, HEV has been detected in intestinal epithelial cells of pigs experimentally infected with genotype 3 HEV (29) and in chickens infected with avian HEV (1).

S10-3 and Caco-2 cells were infected in parallel with the same batch of in vitro transcribed Sar55 genomes. An aliquot of cells was immunostained for ORF2 protein on day 6 to estimate transfection efficiency, and the rest of the cells were lysed on day 7 and tested for infectious intracellular virus by inoculation of HepG2/C3A cells. Immunostaining detected three to four times more transfected S10-3 cells (∼40%) than transfected Caco-2 cells (∼10 to 15%), but Caco-2 cell lysates contained over twice as many (8,270 versus 3,590) infectious virus particles per ml than S10-3 cells (data not shown). These results suggested that Sar55 replicated and/or assembled more efficiently in the Caco-2 cells.

In order to determine if infectious virus was released from the Caco-2 cells, they were transfected with Sar55 genomes, and one half of the culture medium was collected periodically for infectivity assays. An aliquot of cells immunostained for ORF2 on day 4 posttransfection indicated that approximately 5% of the Caco-2 cells had been transfected in this experiment. A low number of infectious extracellular viruses was detected on the first day medium was collected (day 4), and the numbers increased until day 11 and then plateaued (Fig. 1).

FIG. 1.

Infectious virus is released into the medium of cultured Caco-2 cells. One half of the medium from cells transfected with Sar55 genomes was collected on the days shown, and duplicate samples were plated on HepG2/C3A cells. Foci were detected by immunofluorescence microscopy with anti-ORF2 and manually counted. FFU, focus-forming unit.

In an attempt to increase the efficiency of virus production, Caco-2 cells were cloned by limiting dilution, and individual clones were tested for transfectability and virus release. There were large differences among the initial seven cell lines isolated, and transfectability ranged from 1% to 15% of the cells. A second limiting dilution selection resulted in a clone, C25j, that displayed up to 30% transfectability and released 8.7 × 10³ infectious viruses per ml into the medium on day 8. This cell clone was used in all the following experiments.

ORF3 protein is required for efficient release of Sar55 from Caco-2 cells.

Yamada et al. (30) recently demonstrated that ORF3 protein is critical for release of genotype 3 HEV from PLC/PRF/5 hepatoma cells and from A549 lung cells. In order to determine if ORF3 protein was required for release of genotype 1 virus also, cultures of the C25j clone of Caco-2 cells were transfected with genotype 1, wild-type Sar55 genomes or with those of a derivative ORF3 null mutant that does not synthesize ORF3 protein but does produce infectious intracellular virions (6).

At 9 days posttransfection, approximately 20% of the cells in each transfected culture stained positive for ORF2 protein, demonstrating that wild-type and mutant viruses were actively replicating in a similar number of cells (Table 1). Medium harvested on day 29 (7 days after a complete medium replacement) was assayed for infectious viruses and genome copy number. The level of extracellular infectious wild-type virus was 5.4 times that of the ORF3 null mutant, and the wild-type genome copy number was 10.5 times that of the mutant (Table 1), demonstrating that ORF3 protein facilitated release of infectious virus in this system. These data also indicated that only one in 400 to 800 viral genomes in the medium was infectious in vitro and suggested that the specific infectivity of the ORF3 null mutant was equal to, if not higher than, that of the wild-type virus.

TABLE 1.

Virus release into medium

Virus or mutant	% of cells transfecteda	Amt of infectious virusb	Genome copy no.c	Specific infectivityd
Wild type	20	3,800 (600)	3,054,840	803
ORF3 null	20	700 (110)	287,400	410

^aEstimated percentage of C25j cells stained for ORF2 protein.

^bMean number of ORF2-positive foci (standard deviation) per ml of medium plated on HepG2/C3A cells.

^cDetermined by real-time RT-PCR.

^dCalculated by dividing the genome copy number by amount of infectious virus.

ORF3 protein enhances virus release from C25j and S10-3 cells.

ORF3 protein contains a highly conserved PXXP motif that was suggested to interact with Src homology 3 domains in proteins involved in the host cell signaling pathways (14). Therefore, it was of interest to compare the phenotype of the ORF3 null mutant with that of mutants in which a full-length ORF3 protein was produced but in which this motif was altered. Mutational options were limited because of the sequence overlap with ORF2. The second proline in the PSAPPLP sequence of ORF3 was changed to Q or to L by mutating nucleotide (nt) 5426C to A or T, respectively, to yield PSA(Q/L)PLP. Neither mutation changed the overlapping ORF2 amino acid sequence. As expected, similar to the ORF3 null mutant, following transfection both point mutants replicated in S10-3 cells, and cell lysates of the transfected cells contained intracellular virions that were infectious for naïve cells (data not shown). However, in contrast to the null mutant in which no ORF3 protein was detected, ORF3 protein was easily visualized by immunofluorescence microscopy of cells transfected with either point mutant.

Cultures of C25j Caco-2 cells and S10-3 hepatoma cells were each individually transfected with viral genomes from a set consisting of wild-type Sar55, three ORF3 mutants (null, 5426A, and 5426T), and GAD (a polymerase mutant unable to synthesize viral RNA). Transcription mixtures were treated with DNase just prior to transfection in order to decrease the background for RT-PCR.

An aliquot of transfected cells from each culture was immunostained for ORF2 protein on days 6 and 21 to compare the extent of transfection. The number of ORF2-positive C25j cells was low enough to permit manual counting of foci; in contrast, so many S10-3 cells were ORF2 positive that counting was not an option, and the percentage of ORF2-positive cells was estimated visually (Table 2). Note that although the numbers of infected C25j foci or S10-3 cells decreased between days 6 and 21(due to either cell death or release from the monolayer), there were similar decreases of approximately 2-fold for both the wild type and the mutants, suggesting that the wild type was not more virulent under these conditions. Since a well of cultured S10-3 cells contained approximately 1 × 10 ⁵ cells and since transfection efficiencies were from 10 to 20%, the number of cells transfected ranged from 41- to 269-fold greater than the corresponding number of transfected C25j cells on day 6 (Table 2). Within each cell line, immunofluorescence microscopy for ORF2 protein indicated that transfection efficiencies for the wild-type and ORF3 mutants were roughly equivalent, whereas the GAD mutant was completely negative, thus confirming its inability to replicate.

TABLE 2.

Transfection efficiency in different cell lines

Virus or mutanta	No. of foci in C25j cells at:b		% S10-3 cells transfected at:b		S10-3/C25j ratioc
	Day 6	Day 21	Day 6	Day 21
GAD	0	0	0	0
Wild type	300	153	10	5	41
ORF3 null	183	92	10	5	68
ORF3 5426A	156	86	15-20	5	135
ORF3 5426T	92	51	20	10	269

^aThe set of five viral genomes was transcribed and transfected into the indicated cell line with the same reagent at the same time. GAD, polymerase mutant with active site mutated; ORF3 null, mutant unable to synthesize the ORF3 protein; 5426A and 5426T, point mutations that eliminate PXXP motif.

^bCells immunostained for ORF2 protein.

^cRatio of transfected S10-3 cells to C25j cells based on 123,750 S10-3 cells per well.

Although the transfection efficiencies for each cell line were comparable within the set of four replication-competent genomes, the quantity of viral RNA in the medium at later times was significantly greater for the wild-type than for any of the four mutants (Fig. 2). Because the GAD mutant could not synthesize viral RNA, it served to establish the baseline for residual RNA transcripts introduced during the transfection step. In both cell lines (Fig. 2), the quantity of residual GAD in vitro transcripts consistently decreased with each subsequent time point. RNA levels in the medium of cells transfected with the ORF3 null mutant or with the two ORF3 point mutants decreased to the same extent and in parallel with that of the GAD mutant, demonstrating that none of the three ORF3 mutants was able to effectively exit from the cells. In contrast, wild-type virus RNA levels in the medium remained high, suggesting that wild-type virus exited from the cells relatively efficiently. Therefore, ORF3 protein is important for egress of genotype 1 virus from both C25j cells and S10-3 cells. Note that even though there were many more transfected S10-3 cells than transfected C25j cells, the amounts of wild-type viral RNA accumulating in the culture medium were comparable for the two cell lines, thus reinforcing the conclusion that virus exited more efficiently from the C25j cells.

FIG. 2.

ORF3 protein is needed for efficient release of genotype 1 HEV from cells. Intestinal (C25j) or hepatoma (S10-3) cells were transfected with Sar55 wild-type (WT) and mutant genomes. Real-time RT-PCR was performed to quantify viral genomes in the medium; points represent the mean from two independent, parallel cultures. GAD, nonreplicating polymerase mutant; DEL3, ORF3 null mutant; 5426T and 5426A, ORF3 point mutants.

ORF3 point mutations inhibit infection in vivo.

Previously, we reported that the ORF3 null mutant was unable to infect rhesus monkeys (9). Since the two ORF3 point mutants exhibited a phenotype of inefficient release from cells that was similar to that of the ORF3 null mutant, it was of interest to determine whether either point mutant could infect rhesus monkeys.

Two rhesus monkeys were each transfected intrahepatically with in vitro transcripts of the point mutants. Neither monkey transfected with the 5426A (P/Q) mutant seroconverted to anti-HEV, indicating that neither was infected (data not shown). However, both monkeys transfected with the 5426T (P/L) mutant did get infected. Rhesus CK8B seroconverted at week 7 posttransfection and had a peak viremia of 22,160 genomes/100 μl of serum (Fig. 3). RT-PCR was performed on the serum from week 6, and the product spanning nucleotides 4300 to 6850 was directly sequenced to provide the nucleotide consensus. The mutated nucleotide at 5426T had completely reverted back to the wild-type C. A single silent mutation at nucleotide position 6571 was the only other change detected. The second rhesus, CJ7R, did not seroconvert until week 12, and the only viremia detected was approximately 11 genomes/100 μl of serum at weeks 10 and 11 (Fig. 3). An RT-PCR product amplified from week 11 serum spanned nucleotides 4000 to 6900 and was directly sequenced; the only change was that nucleotide 5426T had completely reverted back to the wild-type C in this monkey also. This selection of the revertant in both monkeys is a compelling indication that this region of ORF3 plays a critical role in infection. It is not known why the 5426A mutation did not revert also, but this difference may reflect the fact that an A back-reversion to the original C would be a purine-to-pyrimidine transversion mutation, which occurs less frequently than transition mutations such as the T-to-C, pyrimidine-to-pyrimidine mutation that did take place. Mfold plots (http://helixweb.nih.gov/nih/nih-mfold) indicated that the C residue which was mutated was not base paired and that replacement of C with either A or T did not affect the putative structure or free energy of this region. Whatever the reason, the results suggest that the two point mutations produced similar phenotypes of noninfectivity in vivo that were overcome only by reversion to the wild type.

FIG. 3.

Infection of rhesus macaques by 5426T ORF3 point mutant. Animals were transfected intrahepatically with in vitro transcripts of an HEV mutant. Serum samples were collected weekly, and virus genomes were quantified by HEV-specific real-time RT-PCR.

Confocal microscopy.

Since the ORF3 point mutants did produce ORF3 protein, it was possible to perform confocal microscopy and compare the distribution of wild-type and mutant ORF3 proteins relative to ORF2 protein. Neither wild-type nor either mutant ORF3 protein colocalized with ORF2 protein to any significant degree in either C25j or S10-3 cells; however, the distributions of ORF3 protein differed dramatically between the two cell lines.

The z sections of formaldehyde-fixed cells demonstrated that the wild-type and the two point mutants of ORF3 protein formed a large blob in the majority of S10-3 cells; in contrast, all three versions of ORF3 protein were localized into a continuous layer at the apical membrane of the C25j cells (Fig. 4). In z sections of C25j cells, the number of cells displaying an apical layer of ORF3 protein was 20/20 for wild-type, 13/15 for 5426A, and 13/14 for 5426T.

FIG. 4.

Representative confocal images taken along the z axis of transfected cells. ORF2 protein (green) and ORF3 protein (red) in a confluent monolayer of C25j cells (upper panel) and S10-3 cells (lower panel) on a transwell membrane are shown. Nuclei (blue ovals) and the transwell membrane (blue bottom layer) are stained with DAPI. The three wavelengths were collected separately and superimposed. All images were taken with a 63× oil objective. Viruses from left to right in each row are wild-type, 5426A, 5426T.

Polarity of egress.

Caco-2 cells cultured on a transmembrane can exhibit a polarized phenotype, and viruses growing in polarized cells can exit via either the apical or basolateral surface in a virus-dependent pathway (2). Transfected C25j cell cultures containing about 30% ORF2-positive cells were plated on transmembrane inserts in a 24-well culture plate and incubated until a confluent monolayer was formed. The medium was then completely replaced and harvested separately from the upper and lower chambers 5 days later. Focus assays determined that the mean number of infectious virions in triplicate samples from the lower chamber was 50 ± 30 per ml compared to 4,460 ± 400 per ml from the upper chamber. Therefore, infectious viruses were released almost exclusively from the apical side of the cell, the region where the majority of ORF3 protein was found to accumulate.

Association of lipid with HEV.

An initial comparison found that Sar55 virions in the bile of an infected cynomolgus monkey banded near the bottom of an OptiPrep gradient in fraction 2 of 20 at a density of 1.1284 g/ml, whereas Sar55 virions in rhesus monkey feces banded near the top of a parallel gradient in fraction 15 of 18 at a density of 1.0482 g/ml (data not shown). These results were surprising because they suggested that virus in the feces had a lower buoyant density than in the bile even though HEV is believed to be shed into the intestinal tract via the bile duct. Wild-type Sar55 and ORF3 null viruses released into C25j cell culture medium following transfection and wild-type Sar55 excreted in feces of a rhesus monkey were centrifuged on preformed OptiPrep density gradients to compare their buoyant densities. Hepatitis C virions released from transfected Huh 7.5 cells were included as an internal marker. In each case, the HEV banded at a density between 1.06 and 1.05 g/ml, i.e., at a lighter density than the majority of HCV (Fig. 5) and similar to that found for virus in the feces (1.0482). Since HCV is enveloped, these results suggested that, in all three cell culture samples and in the feces, HEV was associated with lipids.

FIG. 5.

Iodixanol gradients of Sar55 HEV. Wild-type (WT) or ORF3 null mutant (DEL3) virions released into the medium of cultured C25j cells or into feces of an infected rhesus macaque were layered over preformed gradients and centrifuged to determine buoyant density. Hepatitis C virions were included as an internal marker. Fractions were weighed, and density is plotted as a dotted line; viral genomes were quantified by virus-specific RT-PCR. HCV values are not shown, but the HCV peak is indicated by an arrow. The buoyant density is shown above the HEV peak.

The association of lipid with cell culture-generated virions was confirmed by immunoprecipitation experiments. Antibody specificity was demonstrated by precipitation of the wild type, but not the ORF3 null mutant, by anti-ORF3 and by precipitation of virus in the feces by anti-ORF2. However, precipitation by anti-ORF3 required the addition of nonionic detergent, whereas precipitation by anti-ORF2 did not (Fig. 6). The decrease in fecal virus precipitation by anti-ORF2 in the presence of NP-40 did not occur if the fecal components were first removed by centrifugation of the virus through dilute OptiPrep as was done for the cell lysates. In a separate experiment comparing genome precipitation by anti-ORF2, with or without NP-40, directly from the fecal suspension, as in the experiment shown in Fig. 6, or after centrifugation, addition of NP-40 decreased precipitation from the feces by 43% (P = 0.007), whereas precipitation of the semipurified virus increased by 32% in the presence of NP-40 (P = 0.057) (data not shown). Sar55 wild-type virus released into medium of either C25j or S10-3 cells displayed similar reactivities with anti-ORF3, and 24 to 29% of the viral genomes were precipitated in the presence of detergent. (Fig. 7A and B). Intracellular wild-type virus particles semipurified from a cell lysate of transfected S10-3 cells were treated or not with nonionic detergent; these particles, like those released into the medium, also contained ORF3 protein that became significantly more accessible to antibody by treatment with nonionic detergent since the amount of precipitated virions increased from 2% to 11% of total (Fig. 7C). Very surprisingly, polyclonal antibody to ORF2 protein was unable to precipitate either extracellular or intracellular virus under these conditions.

FIG. 6.

Immunoprecipitation of Sar55 in the absence (solid bars) or presence (open bars) of nonionic detergent. Equal amounts of Sar55 wild-type (WT) or ORF3 null mutant (DEL3) virions in the medium of cultured C25j cells (A and B) and virions shed into the feces of an infected macaque (C) were incubated with an irrelevant, anti-ORF2, or anti-ORF3 antibody, and immune complexes were captured on GammaBind beads. Virions bound to beads were quantified by real-time RT-PCR. WT, wild type. Values are the mean of triplicates with standard deviations shown as error bars.

FIG. 7.

Immunoprecipitation of Sar55 wild-type (WT) virions released into the medium of two different cell lines (A and B) or intracellular virions recovered by lysing cells (C). See the legend of Fig. 6 for details of immunoprecipitation.

Mutated ORF3 protein does not bind to virions.

Since the 5426A and 5426T mutants did not promote virus egress, it was of interest to determine if they behaved like the wild type or the ORF3 null mutant in regard to their association with intracellular virus particles. Medium and cells of S10-3 cultures transfected with the wild type or the three ORF3 mutants were harvested on day 8 posttransfection. Cells were lysed and treated with NP-40 prior to semipurification through iodixanol to remove soluble proteins and detergent. Because high levels of residual transfecting RNA remained in the cells, standardization of virus amounts by RT-PCR was not feasible. Indeed, RNA levels were over 10⁶ genome equivalents per sample. Therefore, the number of infectious virions in each sample was compared in a focus forming assay. Low levels of infectious virus in the medium were detected for the wild type, and an even lower level was detected for mutant 5426A; however, amounts of infectious intracellular virions were roughly comparable for all four viruses (Table 3). Therefore, aliquots of intracellular virus were subjected to immunoprecipitation. The background was high due to the transfecting RNA, but the wild type clearly differed from the other three samples (Fig. 8), with P values of 0.026, 0.028, and 0.044 for DEL3, 5426A, and 5426T, respectively. In a separate experiment, prior treatment with RNase A only partially reduced the background for anti-ORF3 but resulted in over a 3-fold difference in wild-type levels relative to mutant genomes (data not shown) instead of the 2- to 2.5-fold difference shown in Fig. 8. The two ORF3 point mutants gave patterns indistinguishable from the patterns of the null mutant (DEL3), demonstrating that only wild-type virions were precipitated with anti-ORF3. Therefore, either single point mutation at 5426 was sufficient to prevent a stable association of ORF3 protein with virions.

TABLE 3.

Infectious virus production

Virus or mutanta	Virus production in triplicate samples (no. of foci/100 μl [mean])b
	Intracellularc	Extracellulard
Wild-type	147, 143, 159 (150)	9, 6, 10 (8)
ORF3 null	70, 50, 104 (75)	0,0,0 (0)
ORF3 5426A	122, 134, 142 (133)	1, 1, 0 (1)
ORF3 5426T	66, 95, 52 (71)	0, 0, 0 (0)

^aSee Table 2 for a description of the HEVs.

^bHepG2/C3A cells were immunostained for ORF2 protein; foci in coded triplicate samples were manually counted (see Materials and Methods).

^cCells were lysed, and intracellular virions were semipurified by centrifugation and plated on HepG2/C3A cells.

^dMedium from cell cultures was plated on HepG2/C3A cells.

FIG. 8.

Immunoprecipitation of semipurified intracellular virions from detergent-treated cell lysates. S10-3 cells were transfected with genomes of wild-type (WT), an ORF3 null mutant (DEL3), or an ORF3 point mutant (5426A or 5426T). See the legend of Fig. 6 for details of immunoprecipitation.

DISCUSSION

Although the cell culture system for genotype 1 of HEV is far from robust, it does support the full replication cycle of the Sar55 strain; cell entry, genome replication, virion morphogenesis (6), and now virus egress have all been documented for this strain (Fig. 1 and 2). Therefore, a normal, albeit inefficient, progression through the virus replication cycle is assumed to initiate following transfection of viral genomes into cells, and thus processes and interactions that are detected in this system are likely to be biologically relevant. This contention is supported by the fact that many of the conclusions from this study are similar, although not totally identical, to those of Yamada et al. (30), who studied a cell culture-adapted genotype 3 strain of HEV that grows to high titers and spreads efficiently from cell to cell (23).

In the current studies, ORF3 protein of Sar55 was found to be required for efficient release of virus from cultured human cells originating from two different organs (Fig. 2). Compared to wild-type virus, the quantity of virus released into culture medium of transfected intestinal or liver cells was minimal, not only for a mutant which could not produce ORF3 protein but also for two mutants which produced full-length ORF3 protein containing a single amino acid substitution in a proposed signaling motif.

ORF3 protein was physically associated with Sar55 wild-type virions that had been released into the culture medium of either cell line but was not detected in Sar55 virions shed into the feces of an HEV-infected rhesus macaque (Fig. 6 and 7). This result was identical to that previously reported for genotype 3 virus grown in lung cells or hepatic cells and shed into human rather than macaque feces (22, 30). The association of ORF3 with virus particles appeared to occur prior to initiation of virus release since anti-ORF3 was able to capture a portion of free virus particles in a cell lysate of cells transfected with the wild type (Fig. 7). Although there was not a discernible difference in the quality of virus particles released from the two cell lines used in the current study, there was a difference in quantity: the intestinal cells released substantially more virions per transfected cell than the hepatoma cells (Table 2 and Fig. 2). Confocal microscopy demonstrated that the ORF3 protein in most of the hepatoma cells formed a large lump in the cytoplasm (Fig. 4). In contrast, the ORF3 protein in the intestinal cells accumulated in a distinct layer at the apical surface of the cells (Fig. 4), which was noteworthy because infectious viruses were released almost exclusively into the well on the apical side of transwell membranes. Therefore, it is possible that the enhanced release of virus from intestinal cells compared to hepatoma cells reflects the polarized accumulation of ORF3 protein at the apical membrane of the C25j cells. It should be noted that although the cultured hepatoma cells were not polarized, liver hepatocytes in vivo are polarized.

In this regard, it is interesting that although the two point mutants with a destroyed PXXP motif also formed a layer at the apical membrane of C25j cells (Fig. 4), virus release was not enhanced but, rather, was comparable to that for the mutant lacking ORF3 protein (Fig. 2); therefore, polarized localization of ORF3 protein is in itself not sufficient to promote virus egress. The immunoprecipitation experiments help provide an explanation for these results. Anti-ORF3 was unable to precipitate intracellular virus particles in lysates of cells infected with either point mutant although it precipitated wild-type particles under the same conditions (Fig. 8). Therefore, it appears that the PXXP motif is required for ORF3 protein binding to virus particles but not for ORF3 protein trafficking to the apical membrane. At present, it is unclear whether ORF3 protein binds directly to ORF2 or through an intermediate molecule.

Recovery of a PXXP revertant following intrahepatic transfection of macaques provides a plausible explanation for the lethal phenotype in vivo previously reported for the ORF3 null mutant. Two nucleotides were mutated to construct the ORF3 null mutant, whereas only one nucleotide was changed to produce the point mutants; therefore, reversion to the wild type would require two steps for the null mutant but only one step for the point mutant. Previous in vitro data showed that viral RNA replication and assembly of infectious virus proceeds in the absence of ORF3 protein. Therefore, null mutant or point mutant viral genomes transfected into rhesus macaque liver should initiate a subclinical infection restricted to the individual transfected cells since the ORF3 protein needed for virion egress will be absent in one case and nonfunctional in the other; however, such localized RNA replication would generate mutations, and the potential for reversion would exist and would be greater for the point mutant than for the null mutant. If a reversion occurred, as it did for the 5426T point mutant, virus exit, dissemination, and amplification would produce a normal HEV infection (Fig. 3). The fact that the viral sequence recovered from the macaques was exclusively that of the revertant confirms that the viral genome had indeed replicated and indicates that this mutant was seriously incapacitated, thus emphasizing the importance of this region in vivo as well as in vitro.

The effect of nonionic detergent on immunoreactivity of ORF3 protein was striking. Evidence that virions were associated with lipids included the demonstration of a low buoyant density of wild-type virions released into culture medium (Fig. 5) and greatly enhanced immunoprecipitation of virions by anti-ORF3 following detergent treatment (Fig. 6 and 7). These results also were in full accord with the published report that genotype 3 cell culture-produced virions were associated with lipids (30). However, we suggest that, in contrast to the studies with genotype 3 virus, lipid association extended to the low level of the ORF3 null mutant released into cell culture medium and to wild-type virus in feces since both viruses displayed a buoyant density virtually identical to that of the wild type (Fig. 5). Although the buoyant densities were similar, the association of lipid with Sar55 virions in the feces differed since ORF3 protein was not detected but ORF2 was. Finally, in our system a substantial proportion of intracellular virions appeared to be associated with lipids, based on their increased reactivity with anti-ORF3 following detergent treatment (Fig. 7). At this time, the differences between our results and those of Yamada et al. (30) concerning lipid interactions are not understood but could reflect genotype-specific differences, differences among the various cell lines (Caco-2 and S10-3 versus A549 lung and PLC/PR5/5 hepatoma), or different detergent treatments and centrifugation protocols (NP-40 versus Tween 20 and OptiPrep versus sucrose density gradients).

In the Sar55 cell culture replication system, we consistently have failed to detect significant colocalization of ORF2 and ORF3 proteins by confocal microscopy, even when both proteins were well stained and clearly evident within a cell. The present studies provide a probable explanation in that they demonstrated that cell culture-produced ORF2 protein within virus particles is masked and unable to react with antibodies to ORF2 even when the antibodies are polyclonal and presumably directed to multiple epitopes. Even though the ORF2 antibody sample and the immunoprecipitation protocol were validated by precipitation of Sar55 virions in fecal preparations, we were unable to detect significant anti-ORF2 precipitation of virus particles produced in either intestinal or hepatoma cells, whether extracellular or intracellular (Fig. 7 and 8), detergent or nondetergent treated, wild type or ORF3 null mutant. It is unlikely that ORF3 protein itself is masking the ORF2 epitopes because anti-ORF2 was unable to capture intracellular virions even when they lacked ORF3 protein; therefore, either the conformation of the ORF2 protein is different than in mature particles, or cellular components are intimately associated with the particles.

In summary, although genotype 1 and 3 HEVs have different epidemiologies and host ranges, genotype 1, like genotype 3, requires ORF3 protein involvement for cell egress and displays an unexpected association with lipids. These interactions appear to be important since they have now been observed with four quite different cell lines, and it seems reasonable to assume that genotype 2 and 4 viruses will behave similarly. The function performed by ORF3 that is critical to virus export remains undefined, but our studies with the PXXP mutants suggest that the SRC homology 3 signaling pathways needs to be explored further. Determination of the precise steps taken by the virus to exit cells should go a long way toward identifying which of the many functions attributed to ORF3 protein are actually relevant to HEV infection and disease.