Disruption of innate immunity due to mitochondrialtargeting of a picornaviral protease precursorYan Yang†, Yuqiong Liang, Lin Qu, Zeming Chen, MinKyung Yi, Kui Li, and Stanley M. Lemon‡

Center for Hepatitis Research, Institute for Human Infections and Immunity, Department of Microbiology and Immunology, University of Texas MedicalBranch, Galveston, TX 77555-1019

Edited by Peter Palese, Mount Sinai School of Medicine, New York, NY, and approved March 12, 2007 (received for review December 22, 2006)

Mitochondrial antiviral signaling protein (MAVS) is an essentialcomponent of virus-activated signaling pathways that induce pro-tective IFN responses. Its localization to the outer mitochondrialmembrane suggests an important yet unexplained role for mito-chondria in innate immunity. Here, we show that hepatitis A virus(HAV), a hepatotropic picornavirus, ablates type 1 IFN responses bytargeting the 3ABC precursor of its 3Cpro cysteine protease tomitochondria where it colocalizes with and cleaves MAVS, therebydisrupting activation of IRF3 through the MDA5 pathway. The3ABC cleavage of MAVS requires both the protease activity of 3Cpro

and a transmembrane domain in 3A that directs 3ABC to mito-chondria. Lacking this domain, mature 3Cpro protease is incapableof MAVS proteolysis. HAV thus disrupts host signaling by a mech-anism that parallels that of the serine NS3/4A protease of hepatitisC virus, but differs in its use of a stable, catalytically activepolyprotein processing intermediate. The unique requirement formitochondrial localization of 3ABC underscores the importance ofmitochondria to host control of virus infections within the liver.

hepatitis virus � interferon regulatory factor 3 � interferon-beta �melanoma differentiation associated gene 5 �mitochondrial antiviral signaling protein

Mammalian cells have evolved complex and effective mech-anisms to sense and defend against invading viruses.

Toll-like receptor 3 (TLR3), retinoic acid-inducible gene I(RIG-I), and melanoma differentiation-associated gene 5(MDA5) are pathogen-associated pattern recognition receptorsthat sense the presence of RNA viruses and stimulate signalingpathways that lead to induction of an antiviral state (1). Theengagement of RIG-I and MDA5, caspase-recruitment domain(CARD)-containing DExD/H RNA helicases (2–4), by viralRNA leads to complex formation with mitochondrial antiviralsignaling protein (MAVS, also known as IPS-1, VISA, or Cardif)(5–8). MAVS is a unique adaptor protein that is localized to theouter mitochondrial membrane through a C-terminal transmem-brane domain (5). For unknown reasons, this membrane asso-ciation is crucial to its ability to signal to downstream kinases,IKK� and Tank-binding kinase 1 (TBK1), responsible for thephosphorylation and activation of IFN regulatory factor 3(IRF3). The phosphorylation of IRF3, a constitutively expressedlatent cytoplasmic transcription factor, leads to its dimerizationand relocalization to the nucleus where it induces IFN-� tran-scription in association with NF-�B and p300/CBP (9).

IRF3 is central to type 1 IFN responses, and many viruses haveevolved mechanisms that disrupt its activation. Hepatitis C virus(HCV), a hepatotropic human flavivirus, expresses a serineprotease, NS3/4A, that disrupts the virus activation of IRF3 byproteolytically targeting MAVS, preventing signaling to IRF3from the RIG-I receptor (7, 10, 11). This block at a proximalpoint in the signaling pathway also inhibits activation of NK-�B.NS3/4A also cleaves the TLR3 adaptor protein, TRIF, disruptingdsRNA signaling to IRF3 and NF-�B through this pathway aswell (12). However, RIG-I appears to be the major pathogenrecognition receptor for intracellular HCV RNA (13). Thedisruption of RIG-I signaling by HCV may attenuate host innate

responses and contribute to the unique capacity of HCV toestablish persistent infections (14).

Hepatitis A virus (HAV) provides a striking contrast to HCVin terms of its natural history. Also a positive-strand RNA virus,but classified within the family Picornaviridae rather than theFlaviviridae (15), HAV, like HCV, has strong tropism for thehuman hepatocyte. However, HAV never establishes long-termpersistent infection and is always eliminated by host defenses.Nonetheless, recent reports indicate that HAV also is capable ofblocking RIG-I mediated IRF3 activation and IFN-� expression(16, 17). We show here that HAV infection strongly down-regulates the expression of MAVS, ablating signaling throughthe MDA5 pathway. In a remarkable parallel to HCV, a stableintermediate product of HAV polyprotein processing, 3ABC,targets MAVS for proteolysis. MAVS cleavage requires a trans-membrane (TM) domain in 3A that directs 3ABC to mitochon-dria, a surprising feature of the 3ABC protease that is uniqueamong picornaviruses.

ResultsVirus Activation of IRF3 Is Blocked by Autonomous Replication of aSubgenomic HAV RNA. Recent reports suggest that HAV infectionblocks induction of IFN-� synthesis through the RIG-I pathwayby preventing virus activation of IRF3 (16, 17). To confirm this,we studied fetal rhesus kidney (FRhK-4) cells infected with a cellculture-adapted variant of HAV, HM175/18f (18). We foundthat HAV infection blocked the induction of IFN-� and ISG56promoters that normally accompanies infection with Sendaivirus (SenV), a well characterized stimulator of the RIG-Isignaling pathway (2) [supporting information (SI) Fig. 6A].Although previous studies suggest that HAV infection activatesNF-�B-dependent promoters (17), we found no stimulation ofthe basal activity of the PRD-II promoter (an NF-�B-dependentelement of the IFN-� promoter). Instead, we observed completeinhibition of SenV-activation of this promoter (SI Fig. 6B).Because HAV infection did not interfere with SenV proteinexpression (SI Fig. 6D), we conclude that HAV disrupts RIG-Isignaling at a proximal point in the pathway, before its bifurca-tion to IRF3 and NF-�B.

To better understand this interference with IFN signaling, we

Author contributions: Y.Y., Y.L., and L.Q. contributed equally to this work; Y.Y., Y.L., L.Q.,Z.C., M.Y., K.L., and S.M.L. designed research; Y.Y., Y.L., L.Q., Z.C., and M.Y. performedresearch; Y.Y., Y.L., L.Q., Z.C., M.Y., K.L., and S.M.L. analyzed data; and Y.Y., Y.L., K.L., andS.M.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: RIG-I, retinoic acid-inducible gene I; CARD, caspase-recruitment domain;MAVS, mitochondrial antiviral signaling protein; MDA5, melenoma differentation associ-ated gene 5; HCV, hepatitis C virus; HAV, hepatitis A virus; TM, transmembrane; SenV,Sendai virus.

†Present address: University of Texas M.D. Anderson Cancer Center, Houston, TX 77030.

‡To whom correspondence should be addressed. E-mail: smlemon@utmb.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0611506104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0611506104 PNAS � April 24, 2007 � vol. 104 � no. 17 � 7253–7258

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0

studied a stable cell line containing an autonomously replicatingsubgenomic HAV RNA. Although we previously reported tran-sient replication of subgenomic HAV replicons in Huh7 humanhepatoma cells (19), stable HAV replicon-containing cell lineshave not been described. The HAV-Bla replicon contains anin-frame substitution of most of the P1 (capsid protein-coding)sequence of HM175/18f virus with sequence encoding blasticidin(Fig. 1A). Blasticidin treatment of transfected Huh7 cells se-lected a single, stable cell line (HAV-Bla) that both expressesHAV antigen (Fig. 1B Left) and contains HAV RNA (SI Fig. 7).To confirm that this subgenomic RNA replicates autonomouslyin HAV-Bla cells, and to provide a clonally related control cellline, we treated the cells with IFN-� in the absence of blasticidin,thereby eliminating the replicon RNA. The cured cells (Bla-C)no longer expressed HAV antigen (Fig. 1B Right) or detectableviral RNA, and none survived when blasticidin was added backto the culture media.

SenV-induced activation of the IFN-� promoter was pro-foundly blocked in the HAV-Bla replicon cells compared withtheir cured Bla-C progeny (Fig. 1C). We also observed a markedreduction in SenV-induced dimerization of IRF3 (Fig. 1D),suggesting that activation of IRF3 is compromised by the HAVreplicon. SenV infection also failed to cause nuclear transloca-tion of IRF3 in HAV-Bla cells, whereas this was readily apparentin the cured Bla-C cells (Fig. 1E Lower). The induction of IFNstimulated gene-15 (ISG15) synthesis was also inhibited (SI Fig.6E), and although SenV infection typically induces only lowlevels of detectable phosphoserine-396 IRF3 in normal Huh7cells (20), this was substantially (albeit, not completely) elimi-nated in the replicon cells (SI Fig. 6E, compare lanes 2 and 4).Importantly, immunoblots indicated that both the cured andreplicon-containing cell lines were equally permissive for SenVreplication (Fig. 1D Lower). These results confirm that HAVreplication and protein expression blocks activation of IRF3through the RIG-I pathway.

RIG-I recognizes 5� triphosphates present on some viralRNAs (21, 22), but not HAV RNA, which is covalently linked toa small peptide, 3B (VPg), at its 5� terminus (23). MDA5 is thuslikely to be more important in sensing HAV RNA; it recognizesRNA from other picornaviruses (24, 25). To determine whetherMDA5 signaling is disrupted by HAV infection, we ectopicallyexpressed constitutively active MDA5 and RIG-I mutants rep-resenting the N-terminal CARD-like domain of each molecule(MDA5-N and N-RIG, respectively). Both mutants activated theIFN-� promoter in cured Bla-C cells, whereas this response wassubstantially blocked in HAV-Bla cells (Fig. 1F Left). Theseresults confirm that HAV blocks MDA5 as well as RIG-Isignaling. The IFN-� response to overexpression of MAVS wasalso reduced, yet not ablated, whereas expression of the kinaseIKK� activated the IFN-� promoter equally in both cell lines(Fig. 1F). Collectively, these results indicate that HAV disruptssignaling between MDA5/RIG-I and the downstream kinases,most likely at the level of MAVS which is an essential adaptorfor both RIG-I and MDA5 (5–7). Because MAVS is required forRIG-I signaling to both IRF3 and NF-�B (5), this finding isconsistent with the block in NF-�B activation observed inHAV-infected cells (SI Fig. 6). We also observed profoundsuppression of SenV-induced IL-6 transcription in HAV-Blacells (SI Fig. 8).

HAV Disrupts MDA5 Signaling by Posttranscriptional Down-Regulation of MAVS Expression. To better understand how HAVmight block MDA5 signaling, we analyzed MAVS expression inthe replicon cells. Remarkably, immunoblots demonstratednearly complete absence of MAVS expression in these cellscompared with the cured Bla-C cells or parental Huh7 cells (Fig.2A). Confocal microscopy confirmed a marked loss of endoge-nous MAVS in HAV-Bla versus Bla-C cells (Fig. 2B). Moreover,

we found a striking reduction in the abundance of MAVS inHAV-infected FRhK-4 cells compared with adjacent nonin-fected cells (Fig. 2C). Immunoblots revealed that MAVS has agreater molecular mass in uninfected FRhK-4 cells than Huh7cells (suggesting a species-specific, rhesus vs. human difference),but HAV infection of either cell type eliminated MAVS expres-

Fig. 1. HAV blocks activation of IRF3 at the level of MAVS. (A) Organizationof the HAV genome (Upper) and the subgenomic HAV-Bla replicon in whichmost of the P1 sequence of HM175/18f is replaced with blasticidin sequence(Lower, shaded box, Bla). (B) Detection of HAV antigen by immunofluores-cence labeling with human anti-HAV IgG in HAV-Bla replicon cells (Left) andIFN-cured Bla-C cells (Right). Nuclei were visualized with DAPI. (C) Reporterassays showing that SenV-induced activation of the IFN-� promoter is blockedin HAV-Bla replicon cells compared with cured Bla-C cells. (D) Immunoblotanalysis of IRF3 in HAV-Bla and Bla-C cell extracts prepared 16 h after SenVinfection vs. mock infection, separated by native PAGE. The arrowhead indi-cates slowly migrating IRF3 dimers induced by SenV infection in cured Bla-Ccells; minimal dimerization was present in similarly infected HAV-Bla repliconcells. (Lower) Equivalent expression of SenV proteins in both cell types. (E)Cellular localization of IRF3 in HAV-Bla (Left) and Bla-C (Right) cells after mock(Upper) versus SenV (Lower) challenge. (F) IFN-� promoter assays of HAV-Blaand Bla-C cells expressing MDA5-N and N-RIG, MAVS, and IKK�. To the rightare immunoblots showing the expression levels of these proteins. A uniquerapidly migrating MAVS-reactive protein in HAV-Bla cells is indicated by (*).GAPDH served as a loading control.

7254 � www.pnas.org�cgi�doi�10.1073�pnas.0611506104 Yang et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0

sion (Fig. 2D). Northern blots indicated no differences in theabundance of MAVS-specific mRNA in the HAV-Bla vs. curedBla-C cells (Fig. 2E). Also, when we cultured HAV-Bla cells inthe presence of MG-115, a potent inhibitor of the proteasome,there was no increase in MAVS abundance (data not shown). Weconclude from these results that HAV infection causes profoundposttranscriptional down-regulation of MAVS expression by aproteasome-independent mechanism.

The HAV 3ABC Protease Precursor Down-Regulates MAVS. Primarycleavage of the HAV polyprotein occurs between 2A and 2B,and is catalyzed by 3Cpro, a cysteine protease which is the onlyprotease expressed by HAV (Fig. 1 A) (26, 27). 3Cpro subse-quently directs other processing events within the polyprotein,except for cleavage at VP1/2A, which is catalyzed by an unknowncellular protease, and VP4/VP2, which occurs late in virionmaturation. Thus, there are seven distinct nonstructural proteinsand several intermediate precursors expressed by the HAV-Blareplicon (Fig. 3A) (28). To determine which might down-regulate MAVS, we cotransfected Huh7 cells with vectorsexpressing an N-terminal GFP-MAVS fusion protein and N-terminally HA-tagged HAV proteins shown in Fig. 3A. Areproducible reduction in GFP-MAVS abundance was observed

only with coexpression of 3ABC or 3ABCD (Fig. 3B Top), andin both cases this was accompanied by appearance of a lowermass (�95 kDa) protein reactive with anti-GFP (asterisk in Fig.3B). In contrast, there was no reduction in the abundance of GFPwhen it was coexpressed with 3ABC or 3ABCD, nor any changein the ratio of GFP-MAVS to GFP abundance when both werecoexpressed with 3AB or 3BC (data not shown). Because 3ABCis known to be a stable, catalytically active precursor of 3Cpro

(28), we suspected that the 95-kDa protein might result fromproteolysis within the MAVS sequence. Immunoblots demon-strated autoprocessing of 3ABCD to 3ABC (Fig. 3B Middle),confirming both the catalytic activity of the 3Cpro sequence inthese constructs and the stability of 3ABC. Importantly, therewas no evidence for MAVS cleavage in cells expressing themature 3Cpro protease (Fig. 3B Top), even though 3Cpro wasefficiently expressed (Fig. 3B Middle, lane 10). In contrast, 3CDwas not detected, possibly due to a ubiquitination signal within3Dpol (29), preventing any conclusions about its ability to cleaveMAVS. Parallel expression of the HCV NS3/4A protease re-sulted in a slight reduction in the mass of GFP-MAVS, asexpected for cleavage near its C terminus (7, 10).

Confocal microscopy showed GFP-MAVS colocalized with

Fig. 2. MAVS is down-regulated posttranscriptionally by HAV. (A) Immuno-blot showing endogenous MAVS (Top), HAV protein (Middle), and GAPDH(Bottom) in extracts from HAV-Bla replicon cells, normal Huh7, and curedBla-C cells. (B) Confocal microscopic images of HAV-Bla (Left) and Bla-C (Right)cells labeled with human anti-HAV (green) and rabbit anti-MAVS (red). (C)Down-regulation of MAVS in FRhK-4 cells infected with HM175/18f virus atlow MOI (�0.01) 3 days before labeling with anti-HAV (green) and anti-MAVS(red). Two HAV-infected cells are evident, both of which show marked sup-pression of MAVS expression. (D) Immunoblots showing MAVS, a �97-kDaHAV protein, and actin loading controls in extracts of FRhK-4 (lanes 1 and 2)and Huh7 (lane 3 and 4) cells prepared 3 days after mock (lanes 2 and 4) or highMOI HAV infection (MOI � 2, lanes 1 and 3). (E) Northern blots of MAVS andactin mRNAs and HAV RNA.

Fig. 3. HAV 3ABC, but not 3Cpro, down-regulates MAVS. (A) Schematicshowing HAV proteins and processing intermediates. (B) (Top) Immunoblotsshowing GFP-MAVS labeled with anti-GFP in extracts of Huh7 cells co-transfected with vectors expressing GFP-MAVS and N-terminally HA-taggedHAV proteins (lanes 3–12), HCV NS3/4A (lane 1), or empty vector (lane 2). Arapidly migrating GFP-MAVS species (*) was detected only in cells expressing3ABC and 3ABCD (lanes 5 and 6). (Middle) Immunoblots of ectopically ex-pressed HAV proteins, detected with anti-HA. (Bottom) GAPDH loading con-trols. (C) (Left) Immunoblots showing endogenous MAVS (Upper) and GAPDHloading controls (Lower) in extracts of Huh7 cells expressing C-terminallyFlag-tagged HAV proteins (lanes 3–6), HCV NS3/4A (lane 1), or empty vector(lane 2). (Right) Immunoblot showing expression levels of Flag-tagged 3ABC,3Cpro, 3A, and 3BC. The arrowhead marks the 3Cpro processing product of3ABC.

Yang et al. PNAS � April 24, 2007 � vol. 104 � no. 17 � 7255

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0

MitoTracker, a dye taken up specifically by mitochondria (SI Fig.9 Left). This finding is consistent with the mitochondrial local-ization of MAVS (5). However, in cells expressing 3ABC (butnot 3Cpro), the subcellular distribution of GFP-MAVS wasaltered to a diffuse, cytoplasmic pattern, indicating disruption ofits mitochondrial association (SI Fig. 9 Right).

To determine whether 3ABC also targets endogenous MAVS,we transfected Huh7 cells with vectors expressing C-terminallyFlag-tagged 3ABC and other HAV proteins. 3ABC expressioncaused a severe reduction in the abundance of endogenousMAVS (Fig. 3C Left), confirming that 3ABC is responsible fordown-regulation of MAVS. In contrast, there was no reductionin MAVS abundance associated with ectopic expression of 3Cpro,3AB, or a mixture of 3A and 3BC. Despite this, immunoblottingconfirmed abundant expression of 3Cpro (Fig. 3C Right). A verylow abundance of 3Cpro was also detected in cells expressing3ABC (Fig. 3C, lane 1, arrow), again confirming the activity ofthe cysteine protease as well as the stability of 3ABC (28).

MAVS Is Targeted for Proteolysis by the Cysteine Protease Activity of3ABC. To confirm that MAVS is degraded by 3ABC-mediatedproteolysis, we constructed a 3ABC mutant in which the activesite nucleophile, 3C-Cys-172, was substituted with Ala (3ABC–C172A). This mutation is known to eliminate 3Cpro proteaseactivity (30). The absence of autoprotease activity in 3ABC-C172A stabilized the expression product, resulting in a greaterabundance compared with wild-type 3ABC in transfected cells(Fig. 4A Right). Nonetheless, in contrast to 3ABC, there was nocleavage of GFP-MAVS, nor any cleavage of endogenousMAVS in cells expressing 3ABC-C172A (Fig. 4A and data notshown). These results confirm that 3ABC is responsible forMAVS proteolysis. The products of MAVS cleavage appear tobe unstable and rapidly degraded, as we observed no consistentevidence of lower mass proteins with MAVS antigenicity in cellsexpressing 3ABC.

We next examined the sequence of MAVS for a potential3Cpro cleavage site. With the exception of the 3A/3B cleavage,

each of the 3Cpro cleavage sites within the HAV polyproteinpossess the consensus sequence (L/V/I)x(T/S)Q2x (31). Such asequence exists at Gln-428 of MAVS (Fig. 4B), 80 residuesupstream of the HCV NS3/4A cleavage site, Cys-508 (Fig. 4D).The P-side residues at this potential cleavage site, LASQ, aresimilar to the 2A/2B and 2C/3A cleavage sites, LFSQ and LWSQ,respectively. To determine whether MAVS is cleaved by 3ABCat Gln-428, we substituted Gln-428 with Ala in a C-terminallyFlag-tagged MAVS expression vector (MAVS–Q428A). We alsomade a second mutant, MAVS–E463A, eliminating a Glu–Glydipeptide sequence with weak homology to the 3A/3B cleavagesite (Fig. 4B). Although we observed cleavage of wild-typeMAVS and the E463A mutant in transfected HAV-Bla repliconcells, this did not occur with the Q428A mutant (Fig. 4C). Asexpected, none of these proteins was cleaved in the cured Bla-Ccells (lanes 4–6). Confirming these results, ectopic expression ofMAVS–Q428A led to equivalent levels of IFN-� promoteractivation in HAV-Bla and Bla-C cells (SI Fig. 10C), whereas theresponse to wild-type MAVS was partially blocked in replicon-containing cells (Fig. 1F). We conclude from these results thatthe 3C cysteine protease activity of 3ABC directs proteolysis ofMAVS at Gln-428 (Fig. 4D).

Mitochondrial Targeting of 3ABC Is Essential for MAVS Proteolysis.3ABC contains a hydrophobic transmembrane (TM) domain in3A (amino acid residues 39–59, Fig. 5A) (32). We hypothesizedthat the association of 3ABC with membranes via this TMdomain might be required for the cleavage of MAVS. To testthis, we created an in-frame deletion within the 3ABC expres-sion vector, removing the TM domain (Fig. 5B, 3ABC–�TM).Ectopically expressed 3ABC-�TM failed to cause cleavage ofGFP-MAVS, in contrast to a similar 3B-deletion mutant (Fig.5C). Because cotransfection of vectors expressing 3A and 3BCdid not result in cleavage of endogenous MAVS (Fig. 3C Left,lane 4), 3A must be expressed in cis with 3Cpro to cause thecleavage. Together, these data suggest that membrane associa-tion, determined by the TM domain in 3A, is essential for 3ABCcleavage of MAVS.

The TM domains of picornaviral 3A proteins anchor the viralreplicase to intracellular membranes on which viral RNA syn-thesis occurs (see Discussion). Although mitochondrial mem-branes have not been implicated in picornaviral replication,there is limited homology between the HAV 3A TM domain andthe C-terminal TM domain of MAVS which anchors MAVSspecifically to the outer mitochondrial membrane (Fig. 5A) (5).Surprisingly, 3ABC co-localized strongly with MitoTracker intransfected cells, suggesting a specific mitochondrial localization(Fig. 5D). 3AC also colocalized with MitoTracker, but not3ABC–�TM or the mature protease, 3Cpro, which showed adiffuse staining pattern consistent with the absence of mem-brane binding. Cell fractionation studies confirmed these results.3ABC, 3ABC–C172A (not shown), and 3A were found exclu-sively in a crude mitochondrial cell fraction (P5 pellet), coseg-regating with the 39-kDa subunit of cytochrome C oxidase I(CI-39) (Fig. 5E, lanes 1 and 7). These HAV proteins were notdetectable in an S15 fraction enriched for ER membranes (Fig.5E, lanes 3 and 9). 3ABC-�TM was not detected in eitherfraction (Fig. 5E, lanes 4 and 6).

Further studies confirmed these results. Coexpression of3ABC and GFP-MAVS led to a diffuse staining pattern for GFP,due to cleavage and release of GFP-MAVS from the mitochon-drial membrane, and mitochondrial localization of 3ABC (SIFig. 10A). Such cells demonstrated only partial co-localization ofthe proteins. In contrast, coexpression of 3ABC-�TM andGFP-MAVS, led to a diffuse staining pattern for the mutated3ABC and largely unaltered mitochondrial localization of GFP-MAVS. Coexpression of the catalytically inactive 3ABC-C172Amutant (which retains the 3A TM domain) led to strong colo-

Fig. 4. The cysteine protease activity of 3ABC is responsible for MAVScleavage. (A) Immunoblots of GFP-MAVS (Top), 3ABC (Middle), and GAPDH(Bottom) in extracts of Huh7 cells expressing GFP-MAVS and either 3ABC or3ABC-C172A; a cleavage product was detected only with wild-type 3ABC (lane1, asterisk). (B) Alignment of the sequences surrounding 3Cpro-catalyzed cleav-ages in the HAV polyprotein and possible cleavage sites, Q428 and E463, inMAVS. Underlined residues in MAVS sequences are those found at the sameposition relative to any site of scission in the polyprotein. (C) Immunoblotshowing MAVS and actin loading controls in HAV-Bla (lanes 1–3) and Bla-C(lanes 4–6) cells expressing wild-type (lanes 1 and 4), Q428A (lanes 2 and 5), orE463A (lanes 3 and 6) MAVS-Flag mutants. Cleavage was eliminated by theQ428A mutation. (D) Schematic of MAVS showing the location of the CARD-like, proline-rich domains, and C-terminal TM domains, and the positions ofHAV 3ABC cleavage at Q428 and HCV NS3/4A cleavage at C508.

7256 � www.pnas.org�cgi�doi�10.1073�pnas.0611506104 Yang et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0

calization of the protease precursor and GFP-MAVS, in apattern consistent with the localization of both proteins to themitochondrial membrane. Consistent with these results, ectopicexpression of 3ABC, but not 3ABC-�TM, 3ABC-C172A, or3Cpro, effectively blocked SenV activation of the IRF3-dependent PRD-III-I promoter in FRhK-4 cells (SI Fig. 10B).

DiscussionBoth HAV and HCV are positive-strand RNA viruses that sharetropism for the liver in humans, despite important differences ingenome structure and replication strategy. Viral RNAs pro-duced during the replication of these viruses are also likely to besensed differently by the host. The 5� nucleotides of bothpositive- and negative-strand HAV RNA are linked covalently to3B (otherwise known as VPg) (23), and thus unlikely to be

recognized by RIG-I (21, 22). Presumably, their presence isdetected by MDA5, a related CARD containing helicase thatsenses the RNA of other picornaviruses (24, 25). In contrast,HCV appears to activate IRF3 by engaging RIG-I (13). None-theless, both RIG-I and MDA5 share MAVS as a commonadaptor, the integrity and mitochondrial location of which areessential to activation of both IRF3 and NF-�B (5). The targetingof MAVS by both the cysteine protease of HAV and the serineprotease of HCV thus represents a remarkable example ofconvergent virus evolution, and provides strong albeit indirectevidence for the importance of MAVS to host control of virusinfections in the liver.

Although some evidence suggests that disruption of RIG-Isignaling could limit the ability of cells to restrict HCV replica-tion (13, 14), its importance in the pathogenesis of hepatitis C iscontroversial (33). Unlike HCV, HAV is incapable of establish-ing long-term persistent infections. Thus, our results show thatvirus-mediated proteolysis of MAVS is not, by itself, sufficientfor persistent infection with a positive-strand RNA virus. HAVinfection is nonetheless clinically silent for 3–5 weeks afterinfection, despite extensive replication within the liver (34).3ABC-mediated proteolysis of MAVS may contribute to thisclinically quiescent phase of the infection. It may also facilitatethe ability of the virus to establish persistent noncytopathicinfections in cultured cells, a feature that typifies HAV despiteits sensitivity to IFN (SI Fig. 7).

Despite similarity in the mechanisms by which HAV and HCVdisrupt signaling through the MDA5/RIG-I pathways, there aremarked contrasts. In the case of HCV and the distantly relatedflavivirus, GB virus B, it is the fully mature, assembled NS3/4Aprotease complex that cleaves MAVS (7, 35), whereas for HAVit is a stable polyprotein processing intermediate, 3ABC, thatdirects its proteolysis (Fig. 3C). Unlike 3ABC, the matureprotease, 3Cpro, is unable to cleave MAVS when expressed incells. 3ABC is directed to the mitochondrial membrane where itcolocalizes with MAVS and attacks it proteolytically. Thissubcellular localization, not described previously for any picor-navirus, depends on the TM domain of 3A, which has homologywith the TM domain of MAVS (Fig. 5A). Deletion of the TMdomain in 3ABC resulted in the loss of MAVS cleavage (Fig.5C), as did a substitution of the active site cysteine of 3Cpro withalanine (Fig. 4A). Thus, the destruction of MAVS requires boththe mitochondrial targeting properties of 3A and the cysteineprotease activity of 3Cpro. The HCV NS3/4A protease alsocolocalizes with MAVS (10), but is not directed specifically tomitochondrial membranes.

The specific mitochondrial localization of 3ABC is surprising,as mitochondrial membranes have not been implicated in picor-naviral replication. The 2C protein of poliovirus, a well studiedpicornavirus, is responsible for reorganizing intracellular mem-branes that support RNA replication; evidence suggests theseare derived from the ER and the cellular anterograde membranetraffic system (36, 37). Although not as well studied, HAV 2Calso induces rearrangement of intracellular membranes, whichhave been presumed to be ER-derived and the site of RNAsynthesis (38). Among other functions, picornaviral 3A proteinsfulfill a central role in cis assembly of the replicase complex, asthe TM domain of 3A tethers 3AB and associated viral andcellular proteins to the membrane (39, 40). Thus, the mitochon-dria-specific targeting of 3ABC suggests the possibility thatHAV RNA replication may occur on mitochondrial membranes,and not membranes of the ER as long suspected. Althoughrequiring investigation, this would not be without precedentamong positive-strand RNA viruses. Nodavirus RNA replicationoccurs on outer mitochondrial membranes in insect cells (41). Itcould explain mitochondrial abnormalities observed in earlyultrastructural studies of acute hepatitis A in chimpanzees; afinding not present in acute hepatitis B (42).

Fig. 5. The 3A TM domain targets 3ABC to mitochondria and is essential for3ABC cleavage of MAVS. (A) Schematic of the HAV 3A polypeptide showinglocation and sequence of the TM domain aligned with that of MAVS (residuesin red). (B) Schematic of C-terminally Flag-tagged HAV P3 expression con-structs. (C) Immunoblots of GFP-MAVS (Top), 3ABC expression products (Mid-dle), and GAPDH (Bottom) in extracts of cotransfected Huh7 cells. Asteriskmarks GFP-MAVS cleavage products (lanes 1 and 3 only). (D) Laser-scanningconfocal microscopy images of transfected Huh7 cells showing cellular local-ization of C-terminally Flag-tagged HAV P3 expression products (Top), mito-chondria (MitoTracker) (Middle), and merged images (Bottom) with DAPIlocalization of nuclei. To compensate for quantitative differences in HAVprotein expression, the green channel gain was increased for 3ABC–�TM anddecreased for 3ABC–C172A. (E) Ectopically expressed GFP-MAVS and HA-tagged 3ABC and 3A, but not 3ABC–�TM, associate with mitochondria. Par-tially purified cell fractions were prepared from transfected Huh7 cells ac-cording to the scheme shown to the left, as described (5). CI-39, complex I39-kDa subunit (mitochondria); PDI, protein disulfide isomerase (endoplasmicreticulum).

Yang et al. PNAS � April 24, 2007 � vol. 104 � no. 17 � 7257

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0

Our results and those of others indicate that the accessoryfunctions of 3A vary among different picornaviruses. The 3Aprotein of enteroviruses is directed to ER membranes, asdescribed above, and contains an N-terminal domain that dis-rupts ER-to-Golgi traffic (43, 44). This function is not essentialfor RNA replication, and may contribute to immune evasion byreducing secretion of IFN-� and proinflammatory cytokines, aswell as by limiting movement of class I molecules to the plasmamembrane (43). As we have shown here, the 3A protein of HAValso contributes to immune evasion, but through a very differentmechanism.

Materials and MethodsCells and Viruses. Fetal rhesus monkey kidney cells (FRhK-4),human embryonic kidney (HEK) 293, Huh7 human hepatomacells, and Huh7-derived HAV-Bla, and Bla-C cells were culturedin DMEM with 10% FBS. The HAV-Bla subgenomic repliconcell line and viruses used in these studies are described in detailin SI Text.

Plasmids and Promoter Reporter Assays. See SI Text.

Confocal Imaging. Cells were cultured on Labtek chamber slides(Nunc) and fixed in 4% paraformaldehyde in PBS for 30 min.Cells were permeabilized with Triton X-100 (0.2%) for 15 minand blocked with 10% normal goat serum at room temperaturefor 1 h. Slides were incubated with appropriate dilutions of

primary antibodies for 1 h at room temperature: rabbit anti-IRF3(Michael David, University of California at San Diego, La Jolla,CA), human polyclonal anti-HAV IgG, rabbit anti-MAVS (Zhi-jian Chen, University of Texas Southwestern Medical Center), orrabbit anti-Flag (Sigma). Slides were washed and incubated foran additional hr with appropriate secondary antibodies. Mito-chondria were labeled with MitoTracker Red CMXRos (Invitro-gen). Slides were counterstained with DAPI and mounted withVectashield mounting medium (Vector Laboratories). Slideswere sealed and examined with a Zeiss LSM-510 laser scanningconfocal microscope within the Infectious Disease and Toxicol-ogy Optical Imaging Core, University of Texas Medical Branch.

Northern Blot Analysis and Real-Time Quantitative RT-PCR. See SIText.

Subcellular Fractionation. The preparation of subcellular fractionsfor determining the localization of 3ABC was carried out asdescribed (5).

We are grateful to Eugene Knutson for assistance with confocal mi-croscopy and Zihong Chen and Francis Bodola for technical assistance.This work was supported by National Institutes of Health GrantsU19-AI40035 (to S.M.L.), R21-DA018054 (to K.L.), and R21-AI063451(to M.Y.), and the John Mitchell Hemophilia of Georgia Liver ScholarAward of the American Liver Foundation (to K.L.). Y.Y. was recipientof a McLaughlin Postdoctoral Fellowship in Infection and Immunity;K.L. is the Cain Foundation Investigator in Innate Immunity.

1. Meylan E, Tschopp J (2006) Mol Cell 22:561–569.2. Yoneyama M, Kikuchi M, Natsukawa T, Shinobu N, Imaizumi T, Miyagishi M,

Taira K, Akira S, Fujita T (2004) Nat Immunol 5:730–737.3. Yoneyama M, Kikuchi M, Matsumoto K, Imaizumi T, Miyagishi M, Taira K,

Foy E, Loo YM, Gale M, Jr, Akira S, et al. (2005) J Immunol 175:2851–2858.4. Andrejeva J, Childs KS, Young DF, Carlos TS, Stock N, Goodbourn S, Randall

RE (2004) Proc Natl Acad Sci USA 101:17264–17269.5. Seth RB, Sun L, Ea CK, Chen ZJ (2005) Cell 122:669–682.6. Kawai T, Takahashi K, Sato S, Coban C, Kumar H, Kato H, Ishii KJ, Takeuchi

O, Akira S (2005) Nat Immunol 6:981–988.7. Meylan E, Curran J, Hofmann K, Moradpour D, Binder M, Bartenschlager R,

Tschopp J (2005) Nature 437:1167–1172.8. Xu LG, Wang YY, Han KJ, Li LY, Zhai Z, Shu HB (2005) Mol Cell 19:727–740.9. Honda K, Takaoka A, Taniguchi T (2006) Immunity 25:349–360.

10. Li XD, Sun L, Seth RB, Pineda G, Chen ZJ (2005) Proc Natl Acad Sci USA102:17717–17722.

11. Loo YM, Owen DM, Li K, Erickson AL, Johnson CL, Fish P, Carney DS, WangT, Ishida H, Yoneyama M, et al. (2006) Proc Natl Acad Sci USA 103:6001–6006.

12. Li K, Foy E, Ferreon JC, Nakamura M, Ferreon ACM, Ikeda M, Ray SC, GaleM, Jr, Lemon SM (2005) Proc Natl Acad Sci USA 102:2992–2997.

13. Sumpter R, Jr., Loo MY, Foy E, Li K, Yoneyama M, Fujita T, Lemon SM, GaleMJ, Jr (2005) J Virol 79:2689–2699.

14. Gale M, Jr, Foy EM (2005) Nature 436:939–945.15. Martin A, Lemon SM (2002) in Hepatitis Viruses, ed Ou J-H (Kluwer Academic,

Norwell, MA), pp 23–50.16. Brack K, Berk I, Magulski T, Lederer J, Dotzauer A, Vallbracht A (2002) J Virol

76:11920–11930.17. Fensterl V, Grotheer D, Berk I, Schlemminger S, Vallbracht A, Dotzauer A

(2005) J Virol 79:10968–10977.18. Lemon SM, Murphy PC, Shields PA, Ping L-H, Feinstone SM, Cromeans T,

Jansen RW (1991) J Virol 65:2056–2065.19. Yi M, Bodola F, Lemon SM (2002) Virology 304:197–210.20. Li K, Chen Z, Kato N, Gale M, Jr, Lemon SM (2005) J Biol Chem 280:16739–16747.21. Pichlmair A, Schulz O, Tan CP, Naslund TI, Liljestrom P, Weber F, Reiser M,

Reis e Sousa C (2006) Science 314:999–1001.22. Hornung V, Ellegast J, Kim S, Brzozka K, Jung A, Kato H, Poeck H, Akira S,

Conzelmann KK, Schlee M, et al. (2006) Science 314:994–997.

23. Weitz M, Baroudy BM, Maloy WL, Ticehurst JR, Purcell RH (1986) J Virol60:124–130.

24. Gitlin L, Barchet W, Gilfillan S, Cella M, Beutler B, Flavell RA, Diamond MS,Colonna M (2006) Proc Natl Acad Sci USA 103:8459–8464.

25. Kato H, Takeuchi O, Sato S, Yoneyama M, Yamamoto M, Matsui K, UematsuS, Jung A, Kawai T, Ishii KJ, et al. (2006) Nature 441:101–105.

26. Schultheiss T, Kusov YY, Gauss-Muller V (1994) Virology 198:275–281.27. Martin A, Benichou D, Chao SF, Cohen LM, Lemon SM (1999) J Virol

73:6220–6227.28. Probst C, Jecht M, Gauss-Muller V (1998) J Virol 72:8013–8020.29. Losick VP, Schlax PE, Emmons RA, Lawson TG (2003) Virology 309:306–319.30. Bergmann EM, Mosimann SC, Chernaia MM, Malcolm BA, James MNG

(1997) J Virol 71:2436–2448.31. Seipelt J, Guarne, A., Bergmann E, James MC, Sommergruber W, Fita I, Skern

T (1999) Virus Res 62:159–168.32. Beneduce F, Ciervo A, Kusov Y, Gauss-Muller V, Morace G (1999) Virology

264:410–421.33. Wieland SF, Chisari FV (2005) J Virol 79:9369–9380.34. Lemon SM, Binn LN, Marchwicki R, Murphy PC, Ping L-H, Jansen RW, Asher

LVS, Stapleton JT, Taylor DG, LeDuc JW (1990) J Infect Dis 161:7–13.35. Chen Z, Benureau Y, Rijnbrand R, Yi J, Wang T, Warter L, Lanford RE,

Weinman SA, Lemon SM, Martin A, et al. (2007) J Virol 81:964–976.36. Egger D, Bienz K (2005) J Gen Virol 86:707–718.37. Rust RC, Landmann L, Gosert R, Tang BL, Hong W, Hauri HP, Egger D,

Bienz K (2001) J Virol 75:9808–9818.38. Teterina NL, Bienz K, Egger D, Gorbalenya AE, Ehrenfeld E (1997) Virology

237:66–77.39. Xiang W, Cuconati A, Paul AV, Cao X, Wimmer E (1995) RNA 1:892–904.40. Suhy DA, Giddings TH, Jr, Kirkegaard K (2000) J Virol 74:8953–8965.41. Miller DJ, Schwartz MD, Ahlquist P (2001) J Virol 75:11664–11676.42. Schaffner F, Dienstag JL, Purcell RH, Popper H (1977) Arch Pathol Lab Med

101:113–117.43. Choe SS, Dodd DA, Kirkegaard K (2005) Virology 337:18–29.44. Wessels E, Duijsings D, Lanke KH, van Dooren SH, Jackson CL, Melchers WJ,

van Kuppeveld FJ (2006) J Virol 80:11852–11860.

7258 � www.pnas.org�cgi�doi�10.1073�pnas.0611506104 Yang et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0