Protein composition of the hepatitis A virus quasi-envelope

Significance

The nonlytic cellular egress of picornaviruses in extracellular vesicles is likely to be important in disease pathogenesis, but the mechanism(s) underlying this process and the origins of the membranes surrounding virions exiting the cell are poorly understood. We describe a quantitative proteomics analysis of quasi-enveloped hepatitis A virus (eHAV) virions that shows capsids are selected as cargo for vesicular export via a highly specific process, and that infectious eHAV virions possess a host protein complement similar to that of exosomes with CD9 and DPP4 displayed on their surface. eHAV-associated proteins are highly enriched for endolysosomal components and lack markers of autophagy, suggesting an exosome-like mechanism of endosomal sorting complex required for transport-mediated eHAV biogenesis involving endosomal budding that is distinct from the autophagosome-mediated release proposed previously for enteroviruses.

Abstract

The Picornaviridae are a diverse family of RNA viruses including many pathogens of medical and veterinary importance. Classically considered “nonenveloped,” recent studies show that some picornaviruses, notably hepatitis A virus (HAV; genus Hepatovirus) and some members of the Enterovirus genus, are released from cells nonlytically in membranous vesicles. To better understand the biogenesis of quasi-enveloped HAV (eHAV) virions, we conducted a quantitative proteomics analysis of eHAV purified from cell-culture supernatant fluids by isopycnic ultracentrifugation. Amino acid-coded mass tagging (AACT) with stable isotopes followed by tandem mass spectrometry sequencing and AACT quantitation of peptides provided unambiguous identification of proteins associated with eHAV versus unrelated extracellular vesicles with similar buoyant density. Multiple peptides were identified from HAV capsid proteins (53.7% coverage), but none from nonstructural proteins, indicating capsids are packaged as cargo into eHAV vesicles via a highly specific sorting process. Other eHAV-associated proteins (n = 105) were significantly enriched for components of the endolysosomal system (>60%, P < 0.001) and included many common exosome-associated proteins such as the tetraspanin CD9 and dipeptidyl peptidase 4 (DPP4) along with multiple endosomal sorting complex required for transport III (ESCRT-III)-associated proteins. Immunoprecipitation confirmed that DPP4 is displayed on the surface of eHAV produced in cell culture or present in sera from humans with acute hepatitis A. No LC3-related peptides were identified by mass spectrometry. RNAi depletion studies confirmed that ESCRT-III proteins, particularly CHMP2A, function in eHAV biogenesis. In addition to identifying surface markers of eHAV vesicles, the results support an exosome-like mechanism of eHAV egress involving endosomal budding of HAV capsids into multivesicular bodies.

The Picornaviridae are a large and diverse family of positive-strand RNA viruses that include numerous pathogens (1). Classically considered “nonenveloped,” these viruses package their single-stranded genomes within stable icosahedral protein capsids that are released from cells following cell lysis. However, recent work has revealed that several picornaviruses gain egress from cells in a nonlytic fashion within the lumen of extracellular vesicles shed from the cell. Most notably, hepatitis A virus (HAV; genus Hepatovirus), an important cause of enterically transmitted hepatitis in humans, replicates without cytopathic effect and is released from cells as membrane-wrapped, “quasi-enveloped” (eHAV) virions similar in size and density to exosomes (2). eHAV virions have been identified in sera collected from persons with acute hepatitis A, as well as in supernatant fluids of infected cell cultures. These virions are completely cloaked in host-derived membranes that protect the capsid from neutralizing antibody until the quasi-envelope is degraded within an endosomal compartment (2, 3). Nevertheless, they have specific infectivity similar to naked, nonenveloped HAV particles in cell culture. Coxsackievirus B, poliovirus, and many other members of the genus Enterovirus typically cause lytic infections in cultured cells. However, viruses in this genus are also released within extracellular vesicles before cell lysis, often with many capsids packaged within a single vesicle (4–6). The cellular egress of these classically nonenveloped viruses in membrane-bound vesicles has blurred the distinction between enveloped and nonenveloped viruses and has important implications for pathogenesis (2).

Several lines of evidence indicate that the biogenesis of quasi-enveloped eHAV virions is dependent upon components of the endosomal sorting complex required for transport (ESCRT). For example, the ESCRT-associated protein ALIX (PDCD6IP) interacts with assembled HAV capsids, and RNAi-mediated knockdown of ALIX expression inhibits the release of quasi-enveloped virus from infected cells (2). Although the size and density of eHAV vesicles are consistent with an exosome-like origin in multivesicular bodies (MVBs) (2, 3), eHAV biogenesis remains incompletely defined and the protein composition of the membranes surrounding the quasi-enveloped capsid is unknown. In contrast, the extracellular vesicles involved in egress of enteroviruses have been suggested to be derived from autophagosomes, as they contain lipidated LC3-II, a marker of autophagy, and disrupting autophagy blocks their release (4, 6). However, the protein composition of these extracellular vesicles also has yet to be comprehensively investigated. Thus, for both hepatoviruses and enteroviruses, the cellular mechanisms underlying the biogenesis of membrane-wrapped extracellular virions remain largely obscure and the protein composition of their membranes unknown.

Here we describe the results of a quantitative proteomics study that defines the protein composition of the quasi-enveloped eHAV virion and points to an endosomal, MVB-like origin for these unusual infectious particles.

Results

We adopted a quantitative proteomics approach incorporating metabolic amino acid-coded mass tagging with stable isotopes (AACT; also known as SILAC) (7, 8) to unambiguously distinguish eHAV-associated proteins from proteins present in exosomes similar in size and density to eHAV virions. Supernatant fluids from infected Huh-7.5 human hepatoma cell cultures maintained in [¹³C₆]Lys/[¹³C₆]Arg (“heavy”; H) or [¹²C₆]Lys/[¹²C₆]Arg (“light”; L) media were subjected to low-speed centrifugation at 10,000 × g to remove large extracellular vesicles (9). eHAV virions remaining in the medium were concentrated by ultracentrifugation at 100,000 × g and then purified by isopycnic equilibrium ultracentrifugation in preformed iodixanol gradients (SI Materials and Methods for details). Fractions containing virus were identified by HAV-specific RT-PCR (Fig. 1A) and mixed with fractions of similar density from a gradient loaded in parallel with concentrated supernatant fluids from noninfected cells maintained in the opposing medium. The samples were then subjected to proteolytic digestion, and peptide fragments were identified by liquid chromatography-tandem mass spectrometry (LC-MS/MS) and quantified by AACT (SI Materials and Methods). Two independent, reverse AACT labeling experiments were carried out, each with H and L labeling of virus (thus we studied a total of four independent virus samples). In the second experiment, virus and mock-infected supernatant fluid concentrates were mixed before, rather than following, isopycnic ultracentrifugation.

Fig. 1.

Stable isotope labeling of eHAV. (A) Distribution of HAV RNA (log₁₀ scale) in fractions collected from iodixanol gradients loaded with eHAV concentrated from supernatant fluids of infected Huh-7 cell cultures maintained in media containing H [¹³C] (Left) or L [¹²C] (Right) amino acids in experiment 1 and centrifuged to equilibrium. The red brackets indicate fractions that were pooled with fractions of similar density from gradients loaded with supernatant fluids from noninfected cells maintained in the opposing media and subjected to mass spectrometry. ○, HAV RNA; GEs (genome equivalents) per 20 µL; ♦, density (g/cm³). (B) Distribution of isotopic enrichment in peptides identified in a search against the human UniProt database in the four independently ¹³C-labeled (Left) and ¹²C-labeled (Right) eHAV samples submitted for mass spectrometry in experiments 1 (Top) and 2 (Bottom). In experiment 1, infected and uninfected cell-culture harvests were mixed following density gradient fractionation, whereas these samples were mixed before fractionation in experiment 2. The fold enrichment of HAV-encoded peptides is noted in each panel.

A total of 638 proteins were identified by the presence of at least one unique peptide in at least one of the four samples when screened against the human UniProt database and HAV polyprotein sequence. Of these, 294 were present in both H and L virus samples. Multiple peptides were derived from the four structural proteins of HAV (Fig. S1). One peptide spanned the VP4–VP2 cleavage (QGIFQTVGSGLDHILSLA^DIEEEQMIQSVDR), suggesting that some empty capsids may have been present in the eHAV preparations. Another peptide spanned the VP1pX cleavage site (APLNSNAMLSTES^MMSR) (10), consistent with the lack of VP1pX processing noted previously in eHAV (2). Overall, 53.7% of the sequence of the structural HAV proteins was represented by peptides, whereas no peptides were identified from nonstructural proteins that comprise a larger proportion of the viral polyprotein (Fig. S1). This indicates that the packaging of HAV capsids within the membranes of eHAV vesicles is highly selective, and excludes virally encoded components of the replication complex.

Fig. S1.

HAV peptide sequence coverage. The polyprotein sequence of HM175/p16 virus (GenBank accession no. KP879217.1) is shown with capsid proteins (VP4 to pX) (Top) and nonstructural proteins (2B to 3D^pol) (Bottom). Amino acid residues shaded in green represent those in peptides identified by mass spectrometry in any of the four independently labeled eHAV samples. Residues shaded in magenta represent individual peptides spanning the VP4–VP2 maturation cleavage and VP1–pX cleavage sites. Overall peptide coverage was 53.7% for capsid (structural) proteins and 0% for nonstructural proteins.

Reflecting robust metabolic labeling, peptides derived from HAV structural proteins were enriched in the H (¹³C) or L (¹²C) isotope of their culture medium of origin by 8.4- to 47-fold in the four samples (mean 26.9-fold ± 7.9 SEM) (Fig. 1B). However, when all peptides were ranked according to the intensity of H versus L isotope, significant skewing was evident, with a predominance of peptides enriched for L isotope in all four samples (Fig. 1B and Table S1): Overall, 42.3% of the peptides were enriched >12-fold for L isotope, versus only 4.6% of peptides for H isotope (P = 0.029 by two-sided Mann–Whitney test). This likely reflects the presence of bovine peptides derived from FCS added to media to maintain cell-culture integrity, as these were not subject to metabolic labeling and thus would be L-enriched. Consistent with this, peptides representing 53% of the identified protein groups shared sequence in common with bovine peptides, rendering their species of origin indeterminate. A total of 292 unique bovine proteins were also identified. These had significantly lower automated H/L ratios than peptides with unique human sequence (P < 0.0001 by ANOVA) or peptides of indeterminate origin (P = 0.0003) (Fig. S2). However, there was no difference between the H/L ratios of peptides of indeterminate versus definitive human origin (P = 0.543), suggesting that most “indeterminate” proteins were actually of human (Huh-7.5 cell) origin.

Table S1.

Proteins identified in four independently labeled samples

		Isotopic enrichment, no. of proteins (n)
Experiment	≥1 unique peptide, n	>2-fold	>3-fold	>6-fold	HAV peptide enrichment, fold enrichment
Virus heavy no. 1	337	106	63	24	26.5
Virus heavy no. 2	440	32	19	8	8.4
Virus light no. 1	232	197	194	186	25.4
Virus light no. 2	270	234	211	139	47.3

Fig. S2.

Distribution of automated MaxQuant normalized H/L ratios for human and bovine protein groups identified by quantifiable peptides (pooled results from four virus samples in two independent experiments). From the top down: all proteins (n = 460), uniquely human (n = 154), indeterminate species of origin (n = 172), and uniquely bovine (n = 134). The dashed horizontal line in each panel represents the median H/L ratio for all proteins. Arrows indicate the point at which the distribution crosses the median H/L ratio value of all proteins.

Peptides representing 105 host proteins (either definitively human or of indeterminate origin) were classified as ≥2-fold–enriched for the cognate eHAV isotope based on automated AACT quantitation of peptides from at least two of the four virus samples (Fig. S3 and Table S1). Peptides from 26 proteins were ≥2-fold–enriched in at least three samples (Fig. 2). The cellular components with which these proteins associate are similar for both sets of proteins (Fig. 3). Over 90% of the proteins in either set have been identified previously in exosomes, and 13 of the 26 proteins identified in at least three virus samples rank among the top 120 exosome-associated proteins in the Vesiclepedia database (www.microvesicles.org) (11). Both sets of proteins were highly enriched for lysosome-associated proteins (P < 0.001), with LAMP1, a classic lysosomal membrane marker, >2-fold–enriched in all four virus samples (geometric mean 8.2-fold).

Fig. S3.

Venn diagram showing the distribution of viral and host cell proteins (n = 106) found to have ≥2-fold isotope enrichment in at least two of the four independently labeled eHAV samples submitted for mass spectrometry: one ¹³C-labeled (heavy) and one ¹²C-labeled (light) eHAV sample from each of two independent experiments. In experiment 1, eHAV particles produced in cells grown in the two different media were purified in iodixanol gradients before mixing, whereas in experiment 2 virus harvests were mixed before iodixanol gradient centrifugation (see SI Materials and Methods for details). The total number of proteins with ≥2-fold enrichment is shown below the label for each sample.

Fig. 2.

Sequence coverage and geometric mean fold isotope enrichment of HAV and 26 nonviral proteins with ≥2-fold isotope enrichment (o) in at least three of four independently labeled eHAV samples submitted for mass spectrometry. *Peptides of indeterminate human versus bovine origin (main text). ^†ALIX: PDCD6IP.

Fig. 3.

Cellular components mapped by FunRich (39) to proteins with ≥2-fold isotope enrichment in either ≥2 or ≥3 of 4 independently labeled eHAV samples. *−Log₁₀(P value) for enrichment versus UniProt database.

We focused further on the 26 proteins that were classified as ≥2-fold–enriched in at least three samples (Fig. 2). Whereas 11 were of indeterminate origin, peptide sequences of the remaining 15 were uniquely human. Sequence coverage varied from a low of 11.3% (NDFIP2; NEDD4 family-interacting protein 2) to a high of 80.1% (PRDX1; peroxiredoxin 1) (mean 39.7 ± 20.2 SD) (Fig. 2). Five of the 26 proteins are functionally associated with ESCRT-III, including ALIX (also known as PDCD6IP), shown previously to be essential for eHAV biogenesis (2), CHMP1A, CHMP1B, CHMP4B, and IST1, whereas SDCBP (syntenin) plays a key role in syndecan-ALIX–mediated exosome biogenesis (12). Two other proteins, RAB5C and RAB7A, are Rab GTPases that function in endosomal trafficking. Manual inspection of spectra for selected proteins (LAMP1, CHMP1A, CHMP4B, RAB5C, RAB7A, and SDCBP) confirmed quantifiable amounts of peptides and high H/L ratios. This array of proteins and their strong endolysosomal association suggest that eHAV egress involves an MVB-like, endosomal export pathway. Importantly, no peptides derived from the autophagosome membrane marker LC3 were identified in any of the purified virus samples.

Several proteins shown in Fig. 2, including CD9, its binding partner IGSF8 (Ig superfamily member 8), DPP4 (dipeptidyl peptidase 4, CD26), and EPCAM (epithelial cell-adhesion molecule), are expressed on the exterior surface of cell membranes and are thus of interest as proteins potentially displayed on the surface of eHAV vesicles. Manual inspection of the spectra confirmed specific isotopic enrichment of DPP4, CD9, and EPCAM peptides. Interestingly, DPP4 localizes to the apical membrane of polarized hepatic cells (13, 14), whereas our recent work shows that eHAV is secreted across the apical, canalicular membrane of infected hepatocytes into the biliary tract of infected Ifnar1^−/− or Mavs^−/− mice (14). Apical membrane locations have also been inferred for CD9 and EPCAM in some cell types (15, 16). Proteins expressed exclusively on the apical membrane would not be expected a priori in eHAV vesicles with an endosomal origin. However, DPP4, CD9, and EPCAM have each been identified in exosomes from a variety of cell types (17–19) (Vesiclepedia database), likely reflecting ambiguity in their membrane associations. Indeed, despite its use as a marker of apical membrane polarity, DPP4 is a well-established lysosomal membrane component (20).

We adopted an immunoaffinity approach to physically confirm the association of selected host proteins with eHAV vesicles, quantifying HAV RNA by RT-PCR in virus immunoprecipitated with specific antibodies from gradient fractions containing eHAV. Prior treatment of the eHAV fraction with the detergent Nonidet P-40 resulted in a large (∼100-fold) increase in HAV RNA precipitated by an anti-capsid monoclonal antibody (K24F2), consistent with the capsid being occluded by membranes disrupted by the detergent (2) (Fig. 4A). Large amounts of virus were also precipitated by antibody to ALIX and DPP4, confirming their association with eHAV. Prior detergent treatment resulted in significant increases in the amount of virus precipitated with anti-ALIX, consistent with previous work suggesting that ALIX interacts with the capsid via two tandem YPX₃L “late domains” in the VP2 capsid protein (2). In contrast, Nonidet P-40 treatment resulted in marked reductions in the amount of viral RNA precipitated by antibody to DPP4, indicating that DPP4 is primarily membrane-associated (Fig. 4A). In addition to its dipeptidyl exopeptidase activity, DPP4 has substantial receptor-binding activities, as it interacts with insulin-like growth factor II receptor (IGFIIR) as well as the chemokine receptor CXCR4 (21). Antibody to DPP4 failed to neutralize the infectivity of eHAV virions (Fig. S4), suggesting that this receptor-binding activity is not essential for eHAV entry. Similar negative neutralization results were obtained with antibodies to the tetraspanins CD9, CD63, and CD81 (Fig. S4). Further studies will thus be needed to ascertain how infectious eHAV virions attach to and gain entry into cells.

Fig. 4.

Immunoprecipitation of eHAV. (A) HAV RNA precipitated from gradient-purified eHAV preparations (mean ± SD) by anti-capsid (K24F2), anti-ALIX, or anti-DPP4 antibodies before (solid bars) and after (striped bars) Nonidet P-40 treatment. IgG, nonspecific antibody. (B) Antibody-mediated pull-downs of eHAV in the absence of detergent treatment. Results shown are mean percentage ± SD of the amount of HAV RNA precipitated by monoclonal K24F2 anti-HAV following Nonidet P-40 treatment. (C) eHAV precipitated by selected antibodies from two acute-phase infection human sera: (Left) A424 and (Right) A595. IgG, control Ig. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.5 for comparisons with control IgG by ANOVA with Dunn’s or Sidak’s tests for multiple comparisons; asterisks in parentheses indicate P values for comparisons between Nonidet P-40 treatment and no Nonidet P-40 by unpaired t test. n = 3 to 9 independent RT-PCR assays.

Fig. S4.

Lack of neutralization of eHAV infectivity by antibodies specific for eHAV surface-associated proteins by infrared immunofocus reduction assay. (A) Percent HAV antigen expression in cell cultures infected with antibody–eHAV mixtures as the percent of antigen expressed by cells infected with eHAV mixed with isotype control Ig. Data are mean ± range, n = 2 independent experiments. (B) Representative infrared fluorescence scans of HAV immunofoci developing in cell sheets infected with HM175/p16 eHAV–antibody mixtures. Shown (Top) is a control culture infected with virus mixed with the cognate isotype Ig control: rabbit for anti-DPP4, and murine for anti-CD9, anti-CD63, and anti-CD81.

Next, we determined the ability of an expanded panel of antibodies to precipitate eHAV in the absence of detergent treatment, scoring precipitates relative to the amount of viral RNA pulled down by anti-capsid antibody (K24F2) in the presence of detergent (Fig. 4B). Consistently, anti-DPP4 pulled down as much HAV RNA in the absence of detergent as K24F2 did following detergent treatment. Antibodies to CD9, NDFIP2, and histone H4 (HIST1H4A) also precipitated substantial quantities of eHAV, consistent with the exposure of these proteins on the eHAV surface. Lesser amounts of virus (<1%) were precipitated by antibody to LAMP1, or the common exosome components CD63 (isotopically enriched in two virus samples; Table S2) or CD81 (which was present in two of four virus samples but not isotopically enriched) (17). In addition to anti-ALIX, antibodies to several other ESCRT-III–associated proteins not identified in the proteomics screen (CHMP4A, CHMP7, and VPS4B) precipitated large amounts of eHAV (Fig. 4B). CHMP4A and CHMP4B (which was identified in the screen; Fig. 2) are known to interact with ALIX and CHMP7 (22, 23), and our previous RNAi depletion studies demonstrated that VPS4B is essential for eHAV biogenesis (2). Notably, however, antibody to TSG101, an ESCRT-I–associated protein that unlike ALIX and VPS4B is not required for eHAV biogenesis (2), failed to precipitate eHAV.

Table S2.

Proteins with ≥2-fold isotopic enrichment in at least two of four independent eHAV samples

Four of four samples	Three of four samples	Two of four samples
DPP4	ALDH1A1	ACTG1	LDHA
HIST1H2BN	ATP1B1	AHCY	LGALS3
LAMP1	CD9	AHCYL2	MARCKSL1
	CHMP1A	ALDOA	MEP1A
	CHMP1B	ANPEP	MITD1
	CHMP4B	ANXA2	NME1
	EEF2	ANXA4	NPC1
	ENO1	ARL8B	P4HB
	EPCAM	ATP1A1	PCBP1
	H2AFV	ATP1B3	PFN1
	HIST1H4A	CA4	PKM
	HSP90AB1	CALR	PLD3
	IGSF8	CD63	PSAT1
	IST1	CFL1	PSMB6
	NDFIP2	CHMP2A	RAB1A
	PDCD6IP	CKB	RAB6B
	PPIA	CLDN1	RAC1
	PRDX1	CLIC1	RAP1B
	RAB5C	COL1A1	RHOA
	RAB7A	COL2A1	SERINC1
	SDCBP	CTNNA1	SFT2D2
	SFN	EEF1A1P5	SLC12A2
	VCP	EEF1G	SLC16A7
		EGFR	SLC1A5
		EIF4A1	SLC2A1
		FABP1	SLC39A1
		FABP5	SLC3A2
		FAM151A	SLC4A1
		FASN	SLC7A5
		GAPDH	SPTAN1
		GSTP1	STOM
		H3F3B	TMEM106B
		HIST2H2AC	TPI1
		HIST2H3PS2	TUBA1C
		HSP90AA1	TUBB2A
		HSPA5	TUBB4B
		HSPA6	TXN
		HSPD1	VIM
		IDH1	YWHAE
		ITGB1	YWHAZ
		JUP

To confirm that proteins associated with eHAV produced in cell culture are also present in eHAV in vivo, we immunoprecipitated virus from human serum samples collected from subjects in the earliest stages of acute hepatitis A infection during a nationwide outbreak of the disease in Korea (24). These are rare specimens, and only available in small volumes. Our previous work shows that such sera contain only eHAV without detectable quantities of nonenveloped virus (2). Consistent with this, the ability of anti-capsid antibody to precipitate virus (HAV RNA) from two acute-phase human sera was significantly increased by detergent treatment (Fig. 4C). Antibodies to ALIX and DPP4 also precipitated substantial amounts of eHAV from human sera, confirming their presence in eHAV virions in vivo. Lesser amounts of eHAV were precipitated by antibodies to CD9, histone H4, and HSP90.

We assessed the contribution of selected eHAV-associated proteins to biogenesis of quasi-enveloped virions by monitoring the release of eHAV from infected cells transfected with host mRNA-specific siRNA (Table S3). High-grade depletion of ALIX, CHMP2A, CHMP2B, and IST1 mRNAs resulted in significant reductions in the egress of virus into cell-culture supernatant fluids without appreciably affecting the levels of intracellular viral RNA (Fig. 5 and Fig. S5). These results suggest that these ESCRT-III–associated proteins may have essential functions in eHAV biogenesis. Although CHMP1A and CHMP1B knockdowns were much less efficient (Fig. S5A), for uncertain reasons partial CHMP1B knockdown reproducibly enhanced eHAV release. Density gradient analysis, a more specific measure of eHAV release than simple supernatant HAV RNA quantitation, confirmed significant reductions in eHAV egress following knockdown of CHMP2A (63% reduction) or combined knockdown of both CHMP2A and CHMP2B (75% reduction) (Fig. 5B). Knockdown of CHMP2B alone had substantially less effect (18% reduction), whereas combined CHMP1A and CHMP1B knockdown resulted in a 68% increase in supernatant eHAV. In contrast, high-grade depletion of DPP4 and RAB7A mRNA had no discernible effect on eHAV egress (Fig. 5A).

Table S3.

SMARTpool siRNA sequences

Gene target (GE Dharmacon no.)	siRNA 1, 5′→3′	siRNA 2, 5′→3′	siRNA 3, 5′→3′	siRNA 4, 5′→3′
ALIX (L-004233-00)	CAGAUCUGCUUGACAUUUA	UCGAGACGCUCCUGAGAUA	GCGUAUGGCCAGUAUAAUA	GUACCUCAGUCUAUAUUGA
DPP4 (L-004181-00)	CACUCUAACUGAUUACUUA	UAGCAUAUGCCCAAUUUAA	CAAGUUGAGUACCUCCUUA	UAUAGUAGCUAGCUUUGAU
RAB7A (L-010388-00)	CUAGAUAGCUGGAGAGAUG	AAACGGAGGUGGAGCUGUA	GAUGGUGGAUGACAGGCUA	GGGAAGACAUCACUCAUGA
CHMP1A (L-010339-00)	GGACGUCCAUACAUCGGUG	CAGAAGGUCUCCUCAGUGA	GAGUGCUGCUCCUGUCAUU	AGAGUGUGCCCGUGUGUAU
CHMP1B (L-004698-00)	CAAAUUCGAGCACCAGUUU	GUCGAUGGCUGGUGUGGUU	CCUUCGGGAUCAAGUGUGA	CAUGGAAGUUGCGAGGAUA
CHMP2A (L-020247-01)	CAAAAGACUUGGUGCGCAC	UGAUGGAGUUUGAGCGGCA	CAGAUGAGCUGUCGAACCU	GCGCAAGUUUGUAUUGAUG
CHMP2B (L-004700-01)	ACUUGUGCAUCUACGGAAA	GGCUAUAAUCAGAGAUCGA	UGAAGAGAUUGAACGGCAA	AAUCGAGAGUUACGAGGUA
IST1 (L-020977-01)	GCAAAGAAUUACAACGUAC	CAGUUGAAGGCCUCGCUUA	GCAAAUACGCCUUUCUCAU	CCAAGUAUAGCAAGGAAUA
SDCBP (L-008270-00)	GGAGAGAAGAUUACCAUGA	GCAUUUGACUCUUAAGAUU	GACCAAGUACUUCAGAUCA	GGAUGGUCUUAGAAUAUUU

Fig. 5.

RNAi analysis of potential DPP4-, RAB7A-, and ESCRT-III–associated host protein function in eHAV biogenesis. (A) RT-PCR detection of virus released into supernatant fluids of persistently infected cell cultures between 72 and 96 h after transfection with host mRNA-specific or control, nontargeting siRNA pools. Results represent the percentage HAV RNA relative to that in supernatants from cultures transfected with nontargeting siRNA, and are shown as mean ± range from two independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by two-way ANOVA with Holm–Sidak’s test for multiple comparisons. (B) eHAV present in fractions of iodixanol gradients loaded with 96-h supernatants from cell cultures transfected with the indicated host mRNA-specific siRNAs. Nontargeting siRNA control results are shown as bars. Also shown are the density traces (dashed lines) from each gradient. eHAV was quantified by computing the area under the curve (AUC) for peaks in the RNA profiles.

Fig. S5.

siRNA knockdown of host gene expression. (A) Impact of host mRNA-targeting siRNAs on intracellular HAV RNA abundance 96 h post transfection. Results shown are the mean ± range of duplicate RT-PCR assays from a representative experiment, and are presented as genome equivalents per cell-culture well. siRNAs were transfected into cells as single SMARTpools (GE Dharmacon) of four siRNAs, or in some cases as pairs of SMARTpools (CHMP1A and CHIMP1B, and CHIMP2A and CHIMP2B). (B) Knockdown efficiencies achieved with siRNAs targeting host mRNAs. Results shown represent the mean ± range of replicate assays from a representative experiment, and are presented as the percent of mRNA present in cells transfected with nontargeting siRNA.

SI Materials and Methods

Antibodies.

We used the following antibodies: anti-HAV capsid (K24F2; CSL) (36); anti-CD63 (BD556019; BD); anti-CD81 (sc23962; Santa Cruz); anti-CHMP7 (sc66206; Santa Cruz); anti-CHMP2B (ab33174; Abcam); anti-CHMP4A (sc67229; Santa Cruz); anti-LAMP1 (ab24170; Abcam); anti-VPS4B (ab86303; Abcam); anti-TSG101 (ab83; Abcam); anti-ALIX (sc53540; Santa Cruz); anti-CD9 (CBL162; Millipore); anti-NDFIP2 (HPA009160; Sigma); anti-LC3 (2775; Cell Signaling); anti-DPP4 (67138; Cell Signaling); anti-histone H4 (2592; Cell Signaling); anti-HSP90 (4877; Cell Signaling); and anti-EPCAM (VU1D9; Cell Signaling). “JC” is a postconvalescent human plasma sample collected several months following symptomatic acute hepatitis A (2).

Sample Preparation for Mass Spectrometry.

The AACT labeled eHAV and control samples, “L” (light; ¹²C) or “H” (heavy; ¹³C), were mixed either following (experiment 1) or prior to (experiment 2) iodixanol gradient centrifugation as described in Results. Mixing the differentially labeled samples following iodixanol gradient purification (experiment 1) resulted in materials of slightly different buoyant density being examined by mass spectrometry due to unavoidable experimental variation in gradient structure/fractionation. On the other hand, mixing the samples before iodixanol gradient centrifugation (experiment 2) ensured that the differentially labeled samples examined by mass spectrometry had precisely the same buoyant density but resulted in some loss of material due to technical issues related to the harvesting of virus. Significantly, 91 of the 107 proteins (∼85%) listed in Table S2 were identified in samples prepared by both methods (Fig. S3). For in-gel digests, the mixture was separated by SDS/PAGE and stained with SYPRO Ruby. The gel was cut into slices and each slice divided into small pieces. The samples were subjected to reduction with DTT and alkylation with iodoacetamide followed by tryptic digest. Peptides were extracted and dried, and then desalted on C18 tips. For in-solution digests, the protein mixture was precipitated with sodium deoxycholate and trichloroacetic acid. Protein precipitates were resolubilized in 8 M urea, reduced, and alkylated. The solution was diluted fourfold with 25 mM Tris⋅HCl (pH 8.0), 1 mM CaCl₂, and then digested with trypsin at room temperature overnight. The digested products were acidified with TFA and desalted.

Tandem Mass Spectrometry for Peptide Sequencing and Protein Identification.

Peptides were resuspended in aqueous 0.1% formic acid. Sample analysis was performed via reversed-phase LC-MS/MS using a nano-LC system (EASY-nLC 1000) coupled to a Q Exactive mass spectrometer (Thermo Scientific). Analytical separation of all tryptic peptides was achieved with a linear gradient of 2 to 30% buffer B over 90 min for in-gel digest peptides or 120 min for in-solution digest samples at a 250 nL/min flow rate, where buffer A was aqueous 0.1% formic acid and buffer B was acetonitrile with 0.1% formic acid. LC-MS experiments were carried out in a data-dependent mode with full MS (externally calibrated to a mass accuracy of <5 ppm and a resolution of 70,000 at m/z 200) followed by high-energy collision-activated dissociation-MS/MS of the top 20 most intense ions. High-energy collision-activated dissociation-MS/MS was used to dissociate peptides at a normalized collision energy of 27 eV in the presence of nitrogen bath gas atoms.

Peptide and Protein AACT Quantitation.

Mass spectra were processed and peptides were identified using the Andromeda search engine and MaxQuant software version 1.5.3.2 (Max Planck Institute). Protein database searches were carried out against UniProt human and bovine sequence databases and human HAV, genotype IB HM175/p16 (A0A0F7Q1W1). Peptide inference was made with a false discovery rate (FDR) of 1%, and peptides were assigned to proteins with a protein FDR of 5%. A precursor ion mass tolerance of 20 ppm was used for the first search that allowed for m/z retention time recalibration of precursor ions that were then subjected to a main search using a precursor ion mass tolerance of 6 ppm and a product ion mass tolerance 0.5 Da. Search parameters included up to two missed cleavages at Lys/Arg on the sequence, oxidation of methionine, and protein N-terminal acetylation as a dynamic modification. Carbamidomethylation of cysteine residues was considered as a static modification. Peptide identifications were reported by filtering of reverse and contaminant entries and assigning to their leading razor protein.

The enrichment of proteins with H or L isotope was estimated from the H/L ratio of proteins, which was derived from a comparison of the extracted ion chromatogram peak areas of all quantifiable matched light versus heavy peptides (7, 8). Ambiguous H/L ratios were resolved by manual inspection of the raw mass spectra. Proteins with ≥2-fold enrichment of the cognate isotope in a minimum of two independently labeled eHAV samples were considered to be potentially associated with eHAV, whereas those with ≥2-fold enrichment in three or four independently labeled samples were considered highly likely to be eHAV-associated. The cellular components and biological pathways with which these proteins are associated were assessed using open-source FunRich 2.1.2 functional enrichment software (www.funrich.org) and UniProt and Vesiclepedia databases (11, 39).

Immunoprecipitation of eHAV.

Fifteen to 50 µL of peak iodixanol gradient fractions containing eHAV was diluted to 500 µL with Hanks’ basic salt solution and mixed with antibody (2 µL, or 1:250 final dilution), and then incubated in a 4 °C rotator overnight. Protein G Sepharose 4 Fast Flow (GE Healthcare) suspended in PBS (50 µL) was added, and the resulting solution was incubated in a 4 °C rotator for 2 to 4 h. The Sepharose slurry was applied to a snap cap spin column (Thermo Fisher) and washed five times with 500 µL cold PBS, with the final wash volume being 140 µL. RNA was extracted from the final wash using the QIAamp Viral RNA Mini Kit (Qiagen), and HAV RNA was quantified by RT-PCR using the Express SYBR GreenER protocol and reagents (Invitrogen) in a Bio-Rad CFX96 Real-Time System with primers as described previously (37).

siRNA-Mediated Host Protein Depletion.

HM175/p16.2-infected Huh-7.5 cells seeded on 12-well plates were transfected with 25 nM gene-specific SMARTpool siRNAs (GE Dharmacon; Table S3) using RNAiMAX transfection reagent (Thermo Fisher) according to the manufacturers’ instructions. Cells were refed with fresh medium at 72 h, and cultures were harvested at 96 h post siRNA transfection. Supernatant fluids were centrifuged at 1,000 × g for 20 min to remove cell debris, and HAV RNA was extracted using the QIAmp Viral RNA Kit (Qiagen) and quantified using a two-step qRT-PCR procedure as described previously (2). RNA was extracted from cell lysates with the RNeasy Kit (Qiagen) and cDNA was synthesized with oligo (dT)₂₀ followed by RNaseH digestion. Host mRNA target abundance was determined using gene-specific primers (Table S4) and normalized to glyceraldehyde-3-phosphate dehydrogenase levels; knockdown efficiency was calculated as the percent mRNA expression relative to nontargeting siRNA samples. Additionally, for selected siRNAs, supernatant fluids collected at 96 h were mixed with 40% stock iodixanol solution (OptiPrep) to a final iodixanol concentration of 24%, 800 µL of which was loaded onto a two-step 32 to 40% iodixanol (100 µL each) cushion and centrifuged in a TLS-55 rotor at 50,000 rpm for 20 h at 4 °C in a Beckman TL-100 ultracentrifuge. Approximately 10 fractions were collected from the top of the resulting density gradient. The density of each fraction was determined with a refractometer (Mettler Toledo; 30GS). HAV RNA was extracted from fractions with the QIAamp Viral RNA Isolation Kit (Qiagen) and quantified in a SYBR Green One-Step Real-Time qRT-PCR (Bio-Rad) assay against a synthetic RNA standard curve as described previously (2) using primers targeting HAV 5′ untranslated RNA. eHAV was quantified by computing the AUC from gradient profiles using Prism 6 for Mac OS X software, version 6.0h (GraphPad). Results are expressed as the percentage of eHAV present in gradients loaded with cell-culture supernatant from the nontargeting siRNA control.

Table S4.

Gene-specific RT-PCR primers

Gene	Forward primer, 5′→3′	Reverse primer, 5′→3′
ALIX	GGAAGAGCAGGCCCAGTA	CGAAAGCATCCTTCCAGGT
DPP4	TGAACTGGGCCACTTACCTT	TTGCAATTCGTTTGTTGTCC
RAB7A	CATCCTGGGAGATTCTGGAG	GTAGAAGGCCACACCGAGAG
CHMP1A	GCGTCCCTTCTCCCTCTC	CTTCGTTCTTCTTGCGGATG
CHMP1B	GACGATGGGCAAGGTGAC	TCATCGTGTCTTCCATTTGC
CHMP2A	AGAAGATGGCCAAGCAAGG	GCCATCGAGTTGTTGGACTT
CHMP2B	GGGCGCAGTCTTTAACCAT	CCAATCTTGGCCATTTTCTT
IST1	GTTGGTGGACCTGATGGAAC	GGGGAGTTGCTGGTATCTGA
SDCBP	GCCTCATTGGTTGGTCTGAG	CATGGTAATCGTCCGTTCAA

eHAV Neutralization Assay.

Antibody-mediated neutralization of eHAV derived from supernatant fluids of HM175/p16 virus-infected cells was assessed as follows in a modified infrared fluorescence immunofocus assay (38). Antibody (1.5 µg) was mixed with 200 µL of a suspension of eHAV collected from peak fractions of a preformed five-step iodixanol gradient prepared as described previously (2), and incubated at 37 °C for 1 h before inoculation onto a 70% confluent FRhK-4 cell monolayer. Following a 2-h incubation at 37 °C, the inoculum was removed and the cells were washed three times with PBS and overlaid with medium containing 1% methylcellulose and 2% FBS. Following a 14-d incubation at 37 °C in a 5% CO₂ atmosphere, the overlay was removed and the cells were fixed with 4% paraformaldehyde and permeabilized with 0.5% Triton X-100, blocked with 1× Odyssey Blocking Buffer (LI-COR Biosciences) in PBS, and incubated for 2 h at room temperature with human anti-HAV IgG conjugated to IRDye 800CW (LI-COR Biosciences). Cell sheets were then extensively washed before visualization of replication foci and quantitation of infrared fluorescence with an Odyssey Infrared Imaging System (LI-COR Biosciences).

Discussion

Recent studies have brought increasing recognition to the fact that multiple members of the Picornaviridae are released from cells enclosed in membranous vesicles in the absence of cell lysis (2–6). This mode of cellular egress has been suggested to play an important role in disease pathogenesis by masking viral antigens from the immune system (2) or increasing the number of viral genomes delivered to single infected cells (6). However, many questions remain concerning the mechanisms underlying the release of virus in extracellular vesicles. Mammalian cells shed a variety of types of extracellular vesicles that are highly heterogeneous in size, density, and protein and nucleic acid composition and differ in their mechanisms of biogenesis (9, 17). The term “exosome” is generally used to describe a specific subset of these extracellular vesicles that are ∼40 to 150 nm in diameter and formed by inward budding of endosomal membranes in an ESCRT-dependent process that results in MVBs containing multiple intraluminal vesicles (hence the alternative term “multivesicular endosomes”) (17, 25). MVBs generally transport their cargo to lysosomes for destruction, but some MVBs disgorge their intraluminal vesicles as exosomes following fusion with the plasma membrane. The specific signals controlling this dichotomy of vesicle movement are not well-understood.

We identified proteins associated with eHAV vesicles using a combination of quantitative proteomics and immunoprecipitation with specific antibodies. Importantly, the use of amino acid mass tagging with stable isotopes (7, 8) allowed us to unambiguously identify proteins associated with eHAV versus other extracellular vesicles with size and density equivalent to eHAV in iodixanol gradients. Over 90% of the proteins that we found to be associated with eHAV have been reported previously in exosomes (Fig. 3) (11). Moreover, eHAV-associated proteins were highly enriched for those that normally localize to the endolysosomal system of cells (Fig. 3). This includes DPP4, which, although an integral apical transmembrane protein, is also associated with lysosomes (20). No array of proteins has been shown to be completely specific for exosomes versus other types of extracellular vesicles formed by outward budding at the plasma membrane rather than inward budding into MVBs (17). Moreover, the protein composition of exosomes may vary with the type of cell from which they are produced. However, the presence of the tetraspanin CD9 and the strong enrichment of endolysosomal proteins in eHAV vesicles are strongly suggestive of an endosomal exosome-like origin for eHAV (9). We also found eHAV to be enriched for RAB5C and RAB7A (Fig. 2), and isoforms of both of these RAB GTPases are known to play prominent roles in endosomal trafficking and exosome biogenesis (12, 26).

Several other conclusions can be drawn from the proteomics data. First, the packaging of HAV capsids into membranous vesicles is not random but instead results from a highly selective and specific sorting process. Despite high peptide coverage for each of the four structural proteins present in the HAV capsid (53.7% overall; Fig. S1), no peptides were identified from the larger, nonstructural segment of the viral polyprotein in any of the four virus samples studied by mass spectrometry. This indicates that nonstructural HAV proteins are not copackaged with capsids in eHAV vesicles, although they are likely present at high local abundance within the cellular compartment(s) in which the viral RNA replicates and the capsid assembles. ALIX–VP2 capsid protein interactions likely play a major role in this sorting process, although other, yet-to-be-determined sorting mechanisms may well be operative (2, 3).

The quantitative proteomics data also point to the engagement of multiple ESCRT-III–associated proteins in eHAV egress. ALIX, CHMP1A, CHMP1B, CHMP4B, and IST1 were all >2-fold–enriched in three of the four virus samples studied by mass spectrometry (Fig. 2), whereas the presence of CHMP4A, CHMP7, and VPS4B was indicated by immunoprecipitation of gradient-purified eHAV with specific antibodies (Fig. 4A). The close association of these ESCRT-III–associated proteins with eHAV is consistent with prior studies showing that RNAi-mediated knockdown of either ALIX or VPS4B effectively ablates eHAV release from cells (2). Similar knockdown experiments confirm functional roles for additional ESCRT-III–associated proteins identified in the proteomics screen, including in particular CHMP2A (Fig. 5).

Whereas ESCRT-I and ESCRT-II proteins are important for exosome biogenesis (17, 27), none were associated with eHAV. TSG101, an ESCRT-I–associated protein that is typically present in exosomes derived from MVBs (9), was not isotopically enriched in any of the four eHAV samples subjected to mass spectrometry. Previous RNAi studies also suggest that neither TSG101 nor the ESCRT-0–associated protein HRS is essential for eHAV release (2). Thus, whereas ESCRT-III proteins are required for membrane scission in eHAV biogenesis, there is as yet no evidence that other ESCRTs play a role in eHAV cargo selection. This contrasts with current paradigms for MVB formation and exosome biogenesis (17, 25). HAV may thus usurp only part of, and not the entire, exosome biogenesis pathway to gain egress from infected cells.

Poliovirus and other enteroviruses appear to egress nonlytically from infected cells in autophagosome-derived vesicles containing LC3-II (4, 5, 28). These enteroviral vesicles are much larger than eHAV, and contain a much greater number of viral capsids than eHAV vesicles (2, 6). Their size resembles extracellular vesicles and autophagous blebs shed from the plasma membrane, and is distinct from MVB-derived exosomes that are generally much smaller (<150 nm) and similar to eHAV (2, 17). Nonetheless, it is not possible to fully exclude a role for autophagy-mediated secretion (29, 30) in the biogenesis of quasi-enveloped eHAV vesicles. Purified eHAV preparations contained no detectable peptides derived from LC3, which in its lipidated LC3-II form is associated with autophagosome membranes. However, LC3 appears to be difficult to detect by mass spectrometry in autophagosome-derived extracellular vesicles (31, 32). Nonetheless, eHAV vesicles were not precipitated by anti-LC3 antibody (Fig. 4B), and we have shown previously that RNAi-mediated depletion of beclin 1 has no effect on eHAV release (2). Autophagy-related release of poliovirus is inhibited by beclin-1 knockdown (6) but not by spautin 1, an autophagy inhibitor that blocks deubiquitinases targeting beclin 1 (33, 34). Further studies are needed to determine the extent to which eHAV biogenesis may involve beclin 1-independent aspects of secretory autophagy.

Hepatitis E virus, which is phylogenetically distinct from the picornaviruses and classified within its own family, the Hepeviridae, and Bluetongue virus, a member of the Orbivirus genus in the Reoviridae family, are nonenveloped viruses that also gain egress from cells via nonlytic processes dependent upon ESCRT-associated proteins (reviewed in ref. 3). Unlike HAV, these viruses are dependent upon TSG101 for their egress. We anticipate that additional nonenveloped viruses will be shown to be released nonlytically in extracellular vesicles, and that the responsible mechanisms may prove to be as heterogeneous and varied as the different types of extracellular vesicles released from eukaryotic cells (17).

Materials and Methods

Cells and Plasmids.

Huh-7.5 human hepatoma cells (35) and FRhK-4 fetal rhesus kidney cells (36) were cultured in DMEM supplemented with 10% FBS, except as otherwise noted. HM175/p16 virus, a low-passage, noncytopathic, cell culture-adapted variant of the HM175 strain of HAV, has been described previously (37, 38). An infectious molecular clone of its genome, pHM175p16.2 (GenBank accession no. KP879217.1), was assembled, under control of the T7 promoter, from PCR-amplified cDNA fragments using standard methods.

Human Serum Samples.

Human serum samples were collected from individuals involved in a nationwide outbreak of hepatitis A in Korea with the approval of the institutional review board of Seoul National University Bundang Hospital (2). Deidentified serum samples were graciously provided by Sook-Hyang Jeong. The UNC Office of Human Research Ethics determined this research did not require Institutional Review Board approval.

Quantitative AACT Proteomics Analysis of Quasi-enveloped eHAV Virions.

RNA transcripts were prepared from pHM175p16.2 DNA and transfected into Huh-7.5 cells by electroporation as previously described (2). After culturing for 2 wk in DMEM with 10% FBS, the cells were passaged five times (1:10 split) in media containing [¹³C₆]Lys/[¹³C₆]Arg (H media) or [¹²C₆]Lys/[¹²C₆]Arg (L media) (Thermo Fisher) in parallel with nontransfected cells. Approximately 5 × 10⁶ cells were subcultured in 850-cm² roller bottles (Corning) with media supplemented with 25 mM Hepes and exosome-depleted FBS (Thermo Fisher). Roller bottles were placed on an Argos FlexiRoll set at 0.01 rpm in a 35.5 °C incubator with no CO₂. After 5 d, the virus-containing media were removed for storage at 4 °C, the cells were refed, and the roller bottles were cultured for an additional 5 d. Following a second and final collection, media from transfected (infected) and nontransfected (noninfected) cells were clarified by centrifugation twice at 10,000 × g for 30 min, thereby removing large extracellular vesicles. eHAV virions remaining in the media were then concentrated by centrifugation at 100,000 × g in a Sorvall Ultra 80 ultracentrifuge with a SureSpin 630 rotor (Thermo Scientific) and 36-mL Ultra-Clear tubes (Beckman Coulter). Pellets were resuspended in 100 µL of PBS, layered on top of a preformed preparative iodixanol gradient, and centrifuged to equilibrium as previously described (2). Fractions containing virus were identified by extraction of RNA followed by HAV-specific RT-PCR (2), mixed with fractions of equivalent density from a similar gradient loaded with concentrated supernatant fluids from noninfected cells maintained in the opposing H or L medium, and subjected to either in-gel or in-solution tryptic digestion followed by LC-MS/MS for peptide sequencing and protein identification. Mass spectra were processed and peptides were identified using the Andromeda search engine and MaxQuant software version 1.5.3.2 (Max Planck Institute) against UniProt human and bovine sequence databases and the UniProt HM175/p16 virus polyprotein sequence (A0A0F7Q1W1). The enrichment of proteins with H (¹³C) or L (¹²C) isotope was estimated from the H/L ratio of peptides, which was derived from a comparison of the extracted ion chromatogram peak areas of all quantifiable matched light ([¹²C]-enriched) versus heavy ([¹³C]-enriched) peptides (7, 8). The cellular components and biological pathways with which these proteins are associated were assessed using open-source FunRich 2.1.2 functional enrichment software and the UniProt and Vesiclepedia databases (11, 39). Complete details are provided in SI Materials and Methods.

Statistics.

Statistical significance was assessed by ANOVA or unpaired t tests as described in Results. Calculations were carried out with Prism 6 for Mac OS X software, version 6.0h (GraphPad). A P value <0.05 was considered significant.