Influenza A virus uses the aggresome processing machinery for host cell entry

See allHide authors and affiliations

Science  24 Oct 2014:
Vol. 346, Issue 6208, pp. 473-477
DOI: 10.1126/science.1257037

Flu mimics damaged proteins during entry

Viruses are master manipulators. The early stages of how flu viruses enter cells are very well understood, but Banerjee et al. describe a new wrinkle (see the Perspective by Rajsbaum and García-Sastre). It seems that the virus carries with it into the cell ubiquitin: a molecule involved in marking proteins for destruction. The virus then exploits host cell machinery involved in recognizing and dealing with damaged proteins to uncoat its own RNA genome, ready to continue its path toward successful infection.

Science, this issue p. 473; see also p. 427


During cell entry, capsids of incoming influenza A viruses (IAVs) must be uncoated before viral ribonucleoproteins (vRNPs) can enter the nucleus for replication. After hemagglutinin-mediated membrane fusion in late endocytic vacuoles, the vRNPs and the matrix proteins dissociate from each other and disperse within the cytosol. Here, we found that for capsid disassembly, IAV takes advantage of the host cell’s aggresome formation and disassembly machinery. The capsids mimicked misfolded protein aggregates by carrying unanchored ubiquitin chains that activated a histone deacetylase 6 (HDAC6)–dependent pathway. The ubiquitin-binding domain was essential for recruitment of HDAC6 to viral fusion sites and for efficient uncoating and infection. That other components of the aggresome processing machinery, including dynein, dynactin, and myosin II, were also required suggested that physical forces generated by microtubule- and actin-associated motors are essential for IAV entry.

Influenza A virus (IAV) is an enveloped virus of great medical impact. With the risk of an influenza pandemic growing, it is increasingly important to understand virus-host interactions in detail and to develop new antiviral strategies (1). IAV has a single-stranded, negative-sense RNA genome divided between eight subgenomic RNA molecules. These are individually packaged into helical viral ribonucleoproteins (vRNPs) that contain a viral polymerase complex and multiple copies of the nucleoprotein, NP. In the virus, the vRNPs form a stable, capsid-like, supramacromolecular complex with the matrix protein, M1, which forms a shell around the vRNPs (2). During IAV entry into a host cell (Fig. 1A, top), the uncoating process is initiated in early endosomes where the mildly acidic pH triggers the opening of the M2 channel in the viral envelope leading to influx of protons and potassium ions (35). A conformational change occurs that renders the capsid uncoating-competent. When the pH drops further in late endosomes (LEs), the membrane fusion function of the hemagglutinin (HA) is activated, and the capsid is transferred to the cytosol (6, 7). The M1 disperses, and the vRNPs are imported into the nucleus through nuclear pore complexes (811).

Fig. 1 IAV requires the Ub-binding function of HDAC6 for capsid uncoating.

(A) (Top) IAV entry steps. The virus particle is endocytosed into early endosomes that undergo dynein-dependent retrograde transport toward the microtubule-organizing center (MTOC). The intraluminal pH drops due to vacuolar adenosine triphosphatase (v-ATPase) activity, inducing HA conformational change and fusion. The capsid disassembles in an HDAC6-dependent manner, followed by vRNP release into the cytosol and import into the nucleus. (Bottom) Quantification of IAV entry assays. Endocytosis: endocytosed IAV particles were detected at 30 min after virus uptake. Dynasore (Dyn) was used at 80 μM to block dynamin-dependent endocytosis. HA acidification: Acidified HA was detected at 1 hour after uptake. Endosome acidification was blocked by 50 nM bafilomycin A1 (BafA1). Fusion: SP-DiOC18(3)/R18–labeled particles were allowed to enter cells for 1.5 hours, and dequenched SP-DiOC18(3) was quantified. Uncoating: M1 uncoating was quantified 2.5 hours after virus uptake. vRNP import: Nuclear NP signal was quantified 4 hours after virus uptake. All assays were performed in the presence of 1 mM cycloheximide to block new synthesis of viral proteins. (B) HDAC6 domain organization. (C) Uncoating in WT and HDAC6 (WTr, HDm, and ZnFm) MEFs. (D) After 2.5 hours of virus uptake into WT, HDAC6 KO, and HDAC6 (ZnFm) MEFs, M1 (green) and LAMP1-positive (red) endosomes were imaged by confocal microscopy. BafA1 was used to block uncoating. Insets show enlarged images of representative endosomes (arrowhead). All data are represented as mean ± SD. ns, P > 0.05; *P < 0.05; **P < 0.01. Scale bars, 10 μm.

While analyzing the role of HDAC8 and HDAC3 in endosome maturation and IAV penetration (12), we noticed that another histone deacetylase, HDAC6, was also required for infection. A cytosolic enzyme responsible for deacetylation of tubulin and some other substrates, it also serves as a key component in ubiquitin (Ub)–dependent aggresome formation and disassembly (13, 14). We found that HDAC6 depletion by RNA interference (RNAi) in A549 cells (a human bronchial epithelial cell line) reduced infection by half (fig. S1, A and H). In HDAC6 knock-out (KO) mouse embryonic fibroblasts (MEFs), infection was reduced to 30%, and viral titers to 48%, compared with wild-type (WT) MEFs (fig. S1, B, C, and H). When WT and HDAC6 KO mice (15) were infected intratracheally with IAV strains PR8 and X31 in HDAC6 KO mice, lung viral titers were reduced to 48% and 31%, respectively, at day 5 compared with controls (fig. S2, A and B). The antiviral immune responses were comparable when tested in PR8-infected mice (fig. S2, C to E).

When the stepwise IAV entry was analyzed in HDAC6 KO MEFs using quantitative assays (8), no significant difference was observed with WT MEFs in endocytic uptake, acid-induced HA conversion, and fusion (Fig. 1A and fig. S1G). However, capsid uncoating and nuclear import of vRNPs were reduced to 22 and 17%, respectively (Fig. 1A). In HDAC6-depleted A549 cells, uncoating was reduced to 21% compared with controls, with no effect on HA acidification or fusion (fig. S1, D to F). It was apparent that HDAC6 played a role in the release of viral capsids from the cytosolic surface of endosomes, the dissociation of M1 from vRNPs, and the dispersion of capsid components in the cytosol.

To confirm that the HDAC6 requirement was postfusion, we induced fusion of the virus directly with the plasma membrane (PM), a process that allows delivery of viral capsids into the cytosol without endocytosis (fig. S3A) (6, 7). Although fusion efficiency was comparable, the M1 was diffusely distributed through the cytoplasm in WT MEFs but remained punctate in HDAC6 KO MEFs (fig. S3, B and C). That uncoating was reduced to 32% in the HDAC6 KO MEFs and to 47% in HDAC6-depleted A549 cells (fig. S3, D and E) compared with control cells confirmed that HDAC6 was required after fusion.

To determine which of the functions of HDAC6 were required for uncoating, we used HDAC6 KO MEFs that had been rescued either with WT HDAC6 (WTr) or with HDAC6s with point mutations either in the two deacetylase domains (HDm) (16) or in the zinc-finger domain (ZnFm) (17, 18). The zinc-finger ubiquitin binding domain (ZnF-UBP) is C-terminal and binds Ub. It plays a critical role in aggresome formation and disassembly (13, 14) (Fig. 1B). Whereas most Ub-binding proteins interact with the hydrophobic core of Ub, the ZnF-UBP domain of HDAC6 binds to the C-terminal Gly-Gly residues (13, 19). This means that HDAC6 only binds to Ub and poly-Ub chains that are not coupled to proteins. The point mutations H1094A and H1098A in the HDAC6 (ZnFm) cells block Ub binding to the ZnF-UBP domain (20, 21).

All three cell lines exhibited normal HA acidification during IAV entry (fig. S4A). However, in the HDAC6 (ZnFm) cells, uncoating, nuclear import of vRNPs, and infection were reduced to 15, 8, and 17%, respectively (Fig. 1C and fig. S4, B and C). Instead of being dispersed as in WT, M1 was confined to LAMP1-positive vacuoles (Fig. 1D). Uncoating and infection were also blocked after virus fusion with PM (fig. S4, D and E). A point mutation W1182A in the human HDAC6 ZnF-UBP disrupts binding of free Ub chains (13). We generated a mutant cell line HDAC6 (ZnFm-W1116A) with a mutation in the corresponding mouse HDAC6 ZnF-UBP. Uncoating was reduced to 21% in these cells (fig. S4F). However, in the HDAC6 (HDm) cells, which were devoid of HDAC6 deacetylase activity, uncoating, nuclear import, and infection were normal (Fig. 1C and fig. S4, B and C). An inhibitor of the deacetylase activity of HDAC6, tubastatin A (22), had no effect on uncoating (fig. S4G). Thus, the ZnF-UBP domain of HDAC6 was critical for IAV uncoating and infection, whereas the deacetylase activity was not.

Previous mass spectroscopy studies have shown that Ub is present in purified IAV (23). Our biochemical analysis confirmed this and showed that virions contained mono- and poly Ub, some of which were unanchored di-, tri-, and tetra-Ub (Fig. 2A). The poly-Ub chains were sensitive to proteinase K digestion only after solubilization of the envelope using Triton X-100, indicating that they were protected (Fig. 2A). In an in vitro pull-down assay, the HDAC6 ZnF-UBP domain associated with Ub molecules present in the virion (Fig. 2, C and D). These were verified as free Ub and poly Ub by immunoblotting with a C-terminal–specific monoclonal antibody (Fig. 2D). Super-resolution microscopy of purified IAV using the same antibody showed free Ub staining in more than half of virus particles (Fig. 2B and fig. S5). The free Ub signal was detected within a boundary defined by the HA signal, indicating that it was inside the particle (Fig. 2B and fig. S5). Viral NP and M1 were also pulled down by HDAC6, but the interaction appeared to be independent of the ZnF-UBP domain (Fig. 2D). When IAV was allowed to enter WT and HDAC6 KO MEFs for 2.5 hours, NP and M1 coimmunoprecipitated with HDAC6 (fig. S4H). In addition, it was observed using indirect immunofluorescence that after ammonium chloride (NH4Cl) washout to allow penetration of capsids from LAMP1-positive endosomes, the distribution of HDAC6 changed (Fig. 2E). Instead of a diffuse cytosolic staining, HDAC6 now associated with M1-containing vacuoles, where it gave a distinct rim-staining. Association with late endosomal vacuoles was transient and was not observed in the HDAC6 (ZnFm) cells (fig. S6B).

Fig. 2 IAV packages free Ub chains that recruit HDAC6 to LE fusion sites via ZnF-UBP.

(A) Western blot analyses of Ub, NP, and M1 after proteinase K treatment of purified IAV virions in the presence or absence of Triton X-100. (B) Super-resolution microscopy image of purified IAV. Virus particles were bound on the coverslip, followed by incubation in pH 5.5 medium at 37°C for 30 min, and fixed. After permeabilization, the particles were stained with antibodies to C terminus Ub (C-Ub) and HA, followed by appropriate secondary antibody labeled with Alexa Fluor. Images were surface-reconstructed with the Imaris software. Scale bar, 100 nm. (C) Schematic representation of different HDAC6 deletion mutants. (D) Pull-down assay of purified, His-tagged HDAC6 full-length, and deletion mutants after incubation with purified IAV lysates. Western blotting was done to detect HDAC6, His, NP, M1, Ub, and C-terminus Ub. (E) HDAC6 recruitment to M1-positive endosomes after viral fusion. IAV endocytic uptake was allowed for 30 min followed by treatment with 20 mM NH4Cl for 1 hour. After washout of the drug to allow viral fusion and penetration, cells were fixed, permeabilized, and stained with antibodies to HDAC6 and M1. Before (left) and 30 min after (right) washout. Scale bar, 10 μm.

Via its interaction with dynein and dynactin, HDAC6 acts as an adaptor that mediates retrograde transport of misfolded protein aggregates along microtubules (MTs) to aggresomes (14). Consistent with a similar role for HDAC6 in IAV infection, we found that depletion of dynactin 2 (the p50 subunit of the dynactin complex) by RNAi and inhibition of dynein by ciliobrevin D (CilioD) (24) reduced uncoating after PM fusion to 51 and 62%, respectively (Fig. 3A). In MEFs in which the HDAC6 lacked the dynein-binding domain (14) (fig. S7, A and B) and thus failed to bind dynein (Fig. 3B), uncoating was reduced to 58% compared with WT MEFs (Dbm) (Fig. 3A). Similar loss of uncoating capacity was observed when IAV entered HDAC6 (Dbm) MEFs via the normal endocytic pathway (fig. S7, C and D). Thus, HDAC6 binding to dynein and dynactin promoted IAV uncoating.

Fig. 3 Dynein/dynactin, myosin II, MTs, and actin promote IAV uncoating.

(A) Quantification of PM-fused uncoating in A549 cells treated with AllStars negative (ASN) and dynactin 2 small interfering RNAs (siRNAs), A549 cells pretreated with 100 μM of ciliobrevin D (CilioD) 1 hour before and during the assay to block dynein motor activity, and WT and HDAC6 (Dbm) MEFs. Depletion of dynactin 2 was confirmed by Western blotting. Actin was used as a loading control. (B) Coimmunoprecipitation with antibody to HDAC6 using WT, HDAC6 KO, and HDAC6 (Dbm) MEFs treated with 5 μM MG132 for 6 hours. Western blotting was done to detect HDAC6, dynein, and proliferating cell nuclear antigen (PCNA). Immunoglobulin G (IgG) served as a loading control. (C) Quantification of PM-fused uncoating in A549 cells treated with ASN, myosin 9 and 10 siRNAs, and A549 cells pretreated with 25 μM ML-9 and 100 μM blebbistatin (Blebb) 1 hour before and during the assay. Depletion of myosin 9 and 10 was confirmed by Western blotting. Actin was used as a loading control. (D) Coimmunoprecipitation using antibody to HDAC6 in WT and HDAC6 KO MEFs infected with IAV for 2.5 hours. Western blotting was done to detect HDAC6, myosin 10, actin, and PCNA. IgG served as a loading control. (E) Quantification of PM-fused uncoating in A549 cells treated with 10 μM cytochalasin D (CytoD), 30 nM nocodazole (Noc), or both drugs 1 hour before and during the assay. Relevant negative and positive controls were included in each experiment. (F) A549 cells were treated with 100 μM CilioD, 25 μM ML-9, 30 nM Noc, 10 μM CytoD, 5 μM MG132, 40 μM Importazole (Ipz), and 20 mM NH4Cl, as indicated in the experimental scheme. M1 uncoating was quantified by fluorescence-activated cell sorting. All data are represented as mean ± SD. ns, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

Because dynein inhibition alone did not inhibit IAV uncoating completely, we investigated a possible additional role for the actomyosin system. It is known that in aggresomes, free Ub chains are generated by deubiquitination of poly-ubiquitinated proteins. This activates HDAC6 binding and stimulates type II nonmuscle myosin 10 (IIB), which promotes aggresome deaggregation, and clearance (13). Using RNAi, we found that uncoating after PM fusion was decreased in myosin 10–depleted but not in myosin 9–depleted A549 cells (to 42% of control) (Fig. 3C). Inhibition of type II myosins with ML-9 and blebbistatin reduced uncoating to 38 and 48%, respectively (Fig. 3C). HDAC6 was, moreover, found to interact both with myosin 10 and actin (Fig. 3D). When actin filaments were depolymerized with cytochalasin D (CytoD) and MTs by nocodazole (Noc), uncoating after PM fusion dropped to 17 and 64%, respectively. Uncoating was, however, blocked completely when both drugs were used together (Fig. 3E).

Finally, we established that dynein, myosin II, MTs, and actin were also involved in uncoating when IAV entered by endocytosis. After virus uptake for 30 min in A549 cells, acid-activated penetration by fusion was prevented by addition of NH4Cl. Different drugs were now added, the NH4Cl was washed out to allow penetration, and the effect of the added drugs on uncoating was analyzed after 60 min. The results showed that inhibitors of dynein, myosin II, actin, and MT assembly all reduced uncoating significantly (Fig. 3F). Inhibition of proteasome activity by MG132, or importin-β–mediated nuclear import by importazole (Ipz) (25), had no significant effect (Fig. 3F).

Taken together, our results demonstrated that key components of the aggresome formation and disassembly machinery promote IAV uncoating. Among them, HDAC6 is recruited to LEs by newly penetrated capsids by binding with capsid-associated Ubs via its ZnF-UBP domain. It serves as a linker molecule to cytoskeletal motors that generate opposing physical forces (26) to break apart the capsid (fig.S8). This cytoskeleton motor-assisted uncoating program offers potential targets such as myosin 10, dynein, and HDAC6 itself for new antiviral strategies.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (2729)

References and Notes

  1. Acknowledgments: We thank many current and former members of the Helenius laboratory for helpful discussions, T. Heger for providing us the infection counter algorithm, and ScopeM of ETH Zurich for assistance in microscopy. This work was supported by grants to A.H. from ETH Zurich, the European Research Council, and the Swiss National Science Foundation (SNSF); to C.S. and M.K. from the SNSF and NIH; and to Y.M. and P.M. from the Novartis Research Foundation. The findings of this work are filed under patent no. 14172323.9-1464 “New treatment against influenza virus.”
View Abstract

Stay Connected to Science

Navigate This Article