|Home | About | Journals | Submit | Contact Us | Français|
Canine distemper virus (CDV) causes in dogs a severe systemic infection, with a high frequency of demyelinating encephalitis. Among the six genes transcribed by CDV, the P gene encodes the polymerase cofactor protein (P) as well as two additional nonstructural proteins, C and V; of these V was shown to act as a virulence factor. We investigated the molecular mechanisms by which the P gene products of the neurovirulent CDV A75/17 strain disrupt type I interferon (IFN-α/β)-induced signaling that results in the establishment of the antiviral state. Using recombinant knockout A75/17 viruses, the V protein was identified as the main antagonist of IFN-α/β-mediated signaling. Importantly, immunofluorescence analysis illustrated that the inhibition of IFN-α/β-mediated signaling correlated with impaired STAT1/STAT2 nuclear import, whereas the phosphorylation state of these proteins was not affected. Coimmunoprecipitation assays identified the N-terminal region of V (VNT) responsible for STAT1 targeting, which correlated with its ability to inhibit the activity of the IFN-α/β-mediated antiviral state. Conversely, while the C-terminal domain of V (VCT) could not function autonomously, when fused to VNT it optimally interacted with STAT2 and subsequently efficiently suppressed the IFN-α/β-mediated signaling pathway. The latter result was further supported by a single mutation at position 110 within the VNT domain of CDV V protein, resulting in a mutant that lost STAT1 binding while retaining a partial STAT2 association. Taken together, our results identified the CDV VNT and VCT as two essential modules that complement each other to interfere with the antiviral state induced by IFN-α/β-mediated signaling. Hence, our experiments reveal a novel mechanism of IFN-α/β evasion among the morbilliviruses.
Virulent canine distemper virus (CDV) causes a severe systemic infection in dogs that is characterized by high fever, diarrhea, and pneumonia. Large-scale immunosuppression is a hallmark of infection, and some virus strains additionally invade the central nervous system to cause chronic demyelinating encephalitis. The molecular mechanisms differentiating virulent from attenuated strains are poorly understood. However, the fact that dogs can be protected from infection with virulent CDV by vaccination with attenuated strains suggests that reliable induction of adaptive immunity is possible, provided that the critical early stage of infection is successfully mastered by the host. During the early stage of infection, host defense depends on the innate immune system, which is also responsible for generating signals that activate the adaptive immune response (27). The interferons of type I (IFN-I, e.g., IFN-α/β) are a critical element of the innate immune defense against viruses (13, 36, 41). Virtually all nucleated cells are capable of sensing viral infection by receptors such as Rig-I, MDA-5, or Toll-like receptor-3 (16). Activation of these receptors initiates a signal cascade that results in transcription, translation, and release from the cells of IFN-α/β. This part of the IFN defense is referred to as the induction stage. IFN action is initiated by the binding of IFN to type I IFN receptors that activates the receptor-associated tyrosine kinases JAK1 and Tyk2, which, in turn, phosphorylate the signal transducers and activators of transcription (STATs) (21, 41). Subsequently, the activated STAT1 and STAT2 together with IFN regulatory factor 9 (IRF9) form a complex, the IFN-stimulated gene factor 3 (ISGF3), which, once translocated to the nucleus, binds the IFN-stimulated response element (ISRE) sequence (39, 45). This initiates the expression of well over 100 proteins which are responsible for the antiviral effect of IFN (36). In recent years, gene products targeting specific steps of IFN induction or action have been found in virtually all viruses studied, indicating the crucial role of IFN evasion in any successful interaction of viruses with their hosts.
CDV, a Morbillivirus of the Paramyxoviridae, contains a nonsegmented, single-stranded, negative-sense RNA genome. The genome consists of six genes expressing the structural nucleocapsid (N), matrix (M), fusion (F), hemagglutinin (H), and large (L) proteins and the phospho-protein (P) (20). The P and the L proteins together form the RNA polymerase. In addition to the P protein, the nonstructural C and V proteins are also expressed from the P gene (20). Recently, it has been documented that a virulent CDV strain (5804P) genetically modified to inactivate V was attenuated in ferrets, whereas a C-defective CDV was fully immunosuppressive (47). These findings demonstrate that wild-type (wt) CDV suppressed IFN induction, but the issue of whether additional modulation of IFN-mediated signaling sustains viral attenuation remains to be determined. In addition, recent work done with C- and V-deficient recombinant measles virus (MV) and rinderpest virus (RPV) indicated that V and, to some extent, P contribute to the final control of the IFN-α/β-mediated signaling pathway. However, to analyze the functions of the P gene products, these recombinant viruses were based on the genetic background of vaccine strains (8, 25). Nevertheless, the paramyxovirus V protein has been identified as the main inhibitor of the IFN-induced antiviral state though various molecular mechanisms were unraveled (12, 14, 15). Expression of the CDV V protein (CDV-V) depends on the insertion of a nontemplated guanine nucleotide at a precise location, called an “editing site,” which generates an mRNA which differs from that of P by one or two nucleotides. This produces an mRNA with an altered open reading frame (ORF) downstream of the editing site, and, thus, due to this specific mechanism, the N-terminal domain of P and V are identical, whereas their C-terminal domains are unique. The C-terminal domain of V (VCT) is known to contain a conserved cysteine-rich region (31, 43), and recent X-ray studies confirmed that VCT folds into a zinc finger conformation (22).
In this study we investigated the role of the P gene products of the highly virulent CDV A75/17 strain in counteracting the IFN-α/β-mediated signaling pathway. Importantly, this strain was isolated from a naturally infected dog and subsequently kept amplified only in dogs, where it has been reported to maintain its virulence (6). Therefore, this virus has never been adapted to any cell lines. However, the generation of recombinant virus stocks (rA75/17) with sufficient titers to work with requires two to three passages in Vero cells expressing signaling lymphocyte activation molecule (Vero-SLAM) after virus rescue from primary full-length cDNA-transfected cells (37). Because these limited amplification steps might already select viral variants, the entire genome of rA75/17 was compared by direct sequencing to that of the parental A75/17 strain and showed no nucleotide differences (37), thereby validating the unique opportunity to investigate the molecular mechanisms of virus-host cell interactions based on a demyelinating morbillivirus strain. Hence, recombinant A75/17 viruses and expression plasmids were generated to investigate the role of the P gene products in mediating IFN evasion. Infection and transfection experiments were performed in Vero cells stably expressing the SLAM receptor for CDV. IFN production from Vero-SLAM cells is defective, and thus they not only provide an optimal tool to exclusively study IFN signaling independently of IFN induction but also support very efficient CDV A75/17 replication.
Our results demonstrate that the V protein was the main viral factor responsible for disrupting the IFN-α/β-mediated signaling pathway. The latter inhibition was neither due to STAT1 or STAT2 degradation nor to an impairment of their phosphorylation states upon IFN-α/β treatment. Rather, the CDV-V protein efficiently associated with both STAT1 and STAT2, which correlated with a complete inhibition of the nuclear import of both transcription factors. Furthermore, transient expression experiments of engineered V proteins identified both the N-terminal and the C-terminal domains as two interdependent modules necessary to exhibit optimal IFN evasion.
Vero-SLAM cells (kindly provided by V. von Messling, INRS-Institute Armand-Frappier, University of Quebec, Laval, Quebec, Canada), MDCK-SLAM cells, and Bsr-T7 cells(stably expressing T7 RNA polymerase) (3) were grown in Dulbecco's modified Eagle medium (Invitrogen) supplemented with 10% fetal calf serum (FCS), penicillin, and streptomycin. The selection of Vero cells expressing the SLAM receptor was maintained by adding zeocin (Invitrogen). All cells were kept at 37°C in the presence of 5% CO2. The transfection and all the infection experiments were performed in Vero-SLAM cells, which are easily infected by CDV. The infection experiments were performed with recombinant viruses based on the wild-type CDV A75/17 strain.
In Western blotting the viral proteins were detected by a rabbit anti-P, anti-V, and anti-C antibody and by monoclonal mouse anti-N antibody D110 (2). The first two antibodies were formerly produced in our laboratory. Briefly, the P-specific and the V-specific domains (both C-terminal) were produced in bacteria, purified, and injected in rabbits to produce P and V antisera. The anti-C antibody was produced against a mix of two synthetic peptides (nh2-RSAASETKPATQARRMEPQACRK-cooh and nh2-RQSSPLKMTSNQDLE-cooh) corresponding to amino acids (aa) 24 to 46 and 85 to 99 of the C protein (Primm srl; Custom Antibodies, Milano, Italy). The cellular STAT proteins were detected by anti-STAT1 (Cell Signaling), anti-STAT2 (A-7; Santa Cruz Biotechnology), anti-phospho-STAT1 (Tyr 701; 58D6; Cell Signaling), and anti-phospho-STAT2 (Tyr 690; Cell Signaling) antibodies. For immunofluorescence staining the anti-N antibody D110, STAT1α p91 (C-24; Santa Cruz Biotechnology), STAT2 (C-20; Santa Cruz Biotechnology), and anti-phospho-STAT1 (Tyr 701; 58D6; Cell Signaling) were used in combination with Alexa Fluor 555 and 488 (Invitrogen) anti-mouse and anti-rabbit antibodies, respectively. The coimmunoprecipitations (co-IPs) were performed with either anti-STAT1 p84/p91 (C-136; Santa Cruz Biotechnology) antibody or anti-STAT2 (A-7; Santa Cruz Biotechnology). In selected co-IP experiments, the anti-hemagglutinin (HA) tag monoclonal antibody (MAb) coupled with Sepharose beads (3F10; Roche) was used. Anti-β-actin (Sigma) served as a loading control. The anti-HA tag monoclonal antibody 16B12 (Covance) was used for immunoblotting and immunofluorescence analyses.
The recombinant viruses are based on a neurovirulent wild-type isolate, CDV A75/17. The mutations to knock out either the C (Cko) or the V (Vko) protein or both proteins (CVko) were performed on a shuttle vector containing the sequences of the N, P, and L genes and produced from a full-length cDNA clone (pFL-A75/17, ). The mutations were obtained by applying a QuikChange Site-Directed Mutagenesis Kit (Stratagene). After digestion with NotI and PmeI, the M, F, and H genes of CDV A75/17 and a red fluorescence marker gene (tD-tomato; kindly offered by D. Garcin, University of Geneva, Switzerland) (23) placed between the first two genes were cloned into the shuttle vectors, which contained the mutated P genes, using T4 DNA ligase (New England BioLabs). To rescue the recombinant viruses we proceeded as previously described (33). Briefly, Bsr-T7 cells were cotransfected with plasmids containing the relevant full-length cDNA and helper plasmids coding for the N, P, and L proteins of CDV. After 3 days the transfected Bsr-T7 cells were cocultured with Vero-SLAM cells. When a cytopathic effect had developed, the cells were lysed by two freeze-thaw cycles, and the viruses (referred to as rA75/17red, rA75/17red Cko, rA75/17red Vko, and rA75/17red CVko) were stored at −80°C. Finally, the recombinant viruses were amplified in two passages and titrated by a plaque assay in Vero-SLAM cells, starting at a multiplicity of infection (MOI) of 0.02. The virus titers were determined by counting red fluorescent single cells or syncytia as infectious particles per ml.
To generate the expression plasmids encoding the P, V, and C proteins (called pCI-P, pCI-V, and pCI-C, respectively), the three relevant sequences of the rA75/17red cDNA P gene were amplified by using an Expand High Fidelity Plus PCR System (Roche). The PCR products were digested with RsrII and cloned in a pCI mammalian expression vector (Promega) by applying the T4 DNA ligase (New England BioLabs). To silence the C gene, the same mutations as for the knockout viruses were performed on the plasmids pCI-P and pCI-V, and the additionally required G was introduced at the editing site in plasmid pCI-V by site-directed mutagenesis (Stratagene). To generate pCI-RFP (where RFP is red fluorescent protein), the marker gene was amplified from the cDNA clone by PCR and cloned into the pCI expression vector previously digested with the identical restriction sites as described above. Then, a small linker peptide (SGGSGGTG) and the HA-tagged peptide (YPYDVPDYA) were added by PCR technology in frame to the C-terminal domain of the RFP ORF (pCI-RFP-Linker-HA). Next, pCI-RFP-HA-Vwt (where Vwt is the wt, or full-length, V protein), pCI-RFP-HA-VNT, and pCI-RFP-HA-VCT plasmids were generated by PCR amplification of the different V domains from the pCI-V vector and subsequently cloned into the pCI-RFP-Linker-HA-cleaved plasmid. Finally, the single substitution Y110D in V was generated by site-directed mutagenesis (Stratagene), thus providing the pCI-Vwt Y110D or pCI-RFP-HA-Vwt Y110D expression plasmids. HA-tagged versions of the following plasmids were produced by PCR technology: pCI-HA-Vwt, pCI-HA-P, pCI-HA-C, pCI-HA-VNT, pCI-HA-VCT, and pCI-HA-Vwt Y110D. The green fluorescent protein (GFP) gene was amplified by PCR technology from a plasmid pgA75/17-V cDNA clone (33) and subsequently digested and ligated into the pCI-cleaved plasmid, thus generating pCI-GFP.
Vero-SLAM cells were grown in 24-well plates and transfected the following day with pISRE-Luc (encoding an IFN-inducible firefly luciferase), pTK-RL (coding for a constitutively expressed Renilla luciferase as a transfection control; both plasmids were kindly provided by D. Garcin, University of Geneva, Switzerland), and pCI-P, -V, -C (or the derivative RFP constructs), the empty pCI, or the control plasmid pCI-GFP using Lipofectamine 2000 (Invitrogen) and Opti-MEM (Invitrogen). The next day the cells were treated (or left untreated) with 1,000 IU/ml universal IFN-α/β (IFN type I; PBL) for 6 h. Then the cells were lysed, and the luciferase activity was measured by applying a dual-luciferase reporter assay system (Promega) according to the manufacturer's recommendation. The luminescence signals of the firefly and the Renilla luciferase were measured with a TD-20/20 Luminometer (Promega), and their ratio was called relative luciferase activity, with the ratio of the empty vector pCI set to 1. For MDA5 signaling assays, cells were transfected with a FLAG-tagged MDA5 construct, pβ-IFN-fl-lucter (both vectors kindly provided by D. Garcin, University of Geneva), and pTK-RL as well as with an RFP-expressing plasmid or one of the different V protein-expressing plasmids. After 24 h of transfection, the cells were stimulated with 1.5 μg of poly(I:C)/ml (Sigma) by transfection with Fugene HD (Roche). The cells were harvested after 15 h and assayed for firefly and Renilla luciferase activity as described above. The ratio between the two luciferase activities obtained from vector-transfected cells (pCI-RFP) in the presence of poly(I:C) was set to 100%.
Vero-SLAM or Vero cells were grown on cover slides in six-well plates, and the following day cells were either infected with the recombinant viruses (rA75/17red, rA75/17red Cko, rA75/17red Vko, and rA75/17red CVko) at an MOI of 0.02 or transfected with the desired expression plasmids. At 24 h postinfection or transfection, cells were treated (or left untreated) with 1,000 IU/ml IFN-α/β for 30 min at 37°C and then fixed with phosphate-buffered saline (PBS) containing 4% paraformaldehyde (PFA). After permeabilization with PFA and 0.1% Triton X-100, the cells were washed in PBS. Blocking of nonspecific binding sites and incubation with the first antibody were performed in PBS supplemented with 2% FCS for 1 h at room temperature and for 1.5 h at 37°C, respectively. Then, the slides were washed in PBS and incubated for 1 h at room temperature with the second antibody diluted in PBS. The cells were washed again, and the nuclei were stained with TOTO3 (Invitrogen) according to the manufacturer's protocol. Pictures were taken with a confocal microscope (Olympus).
Immunoblotting was performed as previously described (32). Briefly, Vero-SLAM cells were cultured in six-well plates and infected (with the various recombinant CDVs [rCDVs]) at an MOI of 0.02 1 day later or transfected with the different expression vectors. At 2 days postinfection or posttransfection, the cells were washed with cold PBS and lysed at 4°C in radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl [500 mM NaCl to extract the C protein], 1% NP-40, 0.5% Na-deoxycholate, 0.1% sodium dodecyl sulfate]SDS], 50 mM Tris-Cl, pH 7.4) containing Halt Protease Inhibitor Cocktail (Socochim) and Phosphatase Inhibitor Cocktail Set II (Merck) to protect phosphorylated STAT proteins. After centrifugation, Laemmli sample buffer (Bio-Rad) supplemented with dl-dithiothreitol (Fluka) was added to the samples, and they were boiled for 5 min and separated by SDS-polyacrylamide gel electrophoresis. Afterwards, the proteins were transferred to nitrocellulose membranes (Hybond-ECL; Amersham Biosciences), which subsequently were blocked in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) and 5% nonfat dry milk (or bovine serum albumin for detection of the C protein) for 1 h. The membranes were incubated in blocking buffer containing the first antibody at 4°C overnight. Then the membranes were washed three times in TBS-T and incubated with the horseradish peroxidase-conjugated second antibody for 1 h at room temperature. After the membranes were washed three times, ECL substrate (Amersham Biosciences) was used, and chemiluminescence signals were detected with an LAS-3000 camera (Fujifilm). Relative protein amounts were calculated with AIDA Image Analyzer software (Raytest Schweiz AG, Wetzikon, Switzerland).
Vero-SLAM cells were seeded in six-well plates, and the next day cells were infected at an MOI of 0.02 with rA75/17red or rA75/17red Vko or left uninfected. In parallel experiments different expression plasmids were transfected. At 2 days postinfection or 1 day posttransfection, the cells were treated (or left untreated) with 1,000 IU/ml IFN-α/β for 30 min at 37°C and then washed with cold PBS and incubated on ice with 50 mM Tris, 150 mM NaCl, 2 mM EDTA, and 1% NP-40 (pH = 7.5) complemented with Halt Protease Inhibitor Cocktail (Socochim) for 45 min until complete lysis. After a 10-min centrifugation at 5,000 × g at 4°C, aliquots for total protein analyses were separated, and the remaining supernatants were incubated for 3 h with the first antibody (anti-STAT1, anti-STAT2, anti-V, or anti-HA) or without antibody, followed by the addition of protein G-Sepharose beads at 4°C overnight. After a 1-min centrifugation at 5,000 × g at 4°C, the pellets were washed four times with the same buffer used earlier and finally dissolved directly in Laemmli sample buffer for Western blotting performed as described above.
MDCK cells were transduced with pRRL lentivirus vectors at an MOI of 5. Subsequently, a highly SLAM-expressing clone was selected by limiting dilution and was used for further experiments.
The lentivirus vector pRRL as been described elsewhere (7) (kindly provided by Patrick Salomon, University of Geneva). Stock of lentivirus vectors was generated in 293T/17 cells as previously described (7).
Standard reverse transcription-PCR (RT-PCR) techniques were employed to amplify and clone the canine IFN-β (cIFN-β) cDNA into the pCI vector (primers are available upon request). To express cIFN-β, 293T cells were transfected with pCI-cIFN-β for 3 days. Then, the supernatant was harvested, filtered through a 0.45-μm-pore-size filter, and concentrated using 10-kDa size exclusion Centricon columns (Millipore).
In order to investigate any potential role of the CDV-P gene products in modulating the IFN-α/β-mediated signaling pathway, we first generated expression vectors encoding the P, V, and C proteins. In addition, specific nucleotides in the P and V genes were mutated to close the open reading frame (ORF) of the C protein without affecting the ORFs of P and V (Fig. (Fig.1B,1B, top). Western blot analysis confirmed that all proteins were efficiently expressed and migrated according to their expected molecular weights in SDS-PAGE (Fig. (Fig.1D).1D). Next, a dual luciferase assay was performed to assess the role of each single protein in controlling IFN-α/β-induced signaling. As expected, after treatment with IFN-α/β, the relative luciferase activity clearly increased in cells transiently transfected with both an empty control plasmid (pCI) and with a plasmid expressing the irrelevant control green fluorescent protein (GFP) (Fig. (Fig.1C).1C). This observation confirmed the bioactivity of the recombinant human IFN in Vero-SLAM cells. In contrast, the activation of the ISRE promoter was strikingly inhibited in cells transfected with the V expression plasmid. Interestingly, while we did not observe any effect at all in cells expressing the C protein, we noticed a partial inhibition in cells transiently expressing the P protein (Fig. (Fig.1C).1C). Taken together, these results demonstrated that the CDV-V protein potently inhibited the activation of the ISRE promoter, whereas the P protein could only slightly control the IFN-α/β-mediated signaling pathway.
We then investigated at which step the CDV-V protein was able to affect the IFN-α/β-mediated signaling pathway. However, conclusions drawn from single-protein overexpression experiments may differ from results obtained with the same protein in the context of a viral infection (25). In order to overcome this problem, we generated recombinant C and/or V knockout viruses based on the highly virulent CDV strain A75/17 (Fig. (Fig.1A).1A). Mutations ablating the C ORF were identical to those applied in the P and V expression vectors. Furthermore, specific nucleotide substitutions were performed at the P gene editing site, which preserved the P reading frame while silencing the production of the V mRNA (Fig. (Fig.1B,1B, bottom). All viruses were successfully rescued from cDNA, and subsequently the expression of the P gene products in cells infected with rA75/17red, rA75/17red Cko, rA75/17red Vko, and rA75/17red CVko was confirmed by immunoblot analysis. As shown in Fig. Fig.1E1E all recombinant viruses expressed the expected pattern of P, V, and C proteins. Growth kinetics indicated that the absence of the C and V proteins scarcely affected the efficiency of viral replication in Vero-SLAM cells since all the recombinant viruses showed very similar growth curves (Fig. (Fig.1F1F).
Different molecular mechanisms were documented for several viruses among the family Paramyxoviridae in order to evade the IFN-α/β-induced antiviral state. For instance, the V proteins of rubulaviruses are responsible for a direct degradation of the pool of STAT molecules in the cytoplasm, whereas the V proteins of MV (depending on the viral strain) and RPV inhibit the phosphorylation of STAT1 and STAT2 upon IFN treatment (5, 8, 25, 30, 42). In line with these results, we next examined by immunoblotting whether the CDV-V-mediated inhibition of the IFN-α/β-mediated signaling pathway was due to either STAT degradation or inhibition of STAT phosphorylation. Figure Figure2A2A documents that after 2 days of infection in Vero-SLAM cells, all rCDVs mediated a typical cytopathic effect, with the formation of syncytia involving about 80% of the cells. Even though only about 20% of the cells remained noninfected, the expression and stability of both STAT1 and STAT2 were not modulated in infected cells compared to mock-treated cells (Fig. (Fig.2B).2B). Furthermore, none of the recombinant viruses was able to block the phosphorylation of STAT1 and STAT2. Indeed, immunoblotting revealed no major differences in the amount of phospho-STAT molecules detected in cells infected with any of the recombinant viruses compared to mock-infected, IFN-α/β-treated, cells (Fig. (Fig.2B).2B). Taken together, these results indicate that the aforementioned role of V in inhibiting an IFN-α/β-dependent response is due neither to the degradation of STAT1 and STAT2 nor to the inhibition of their phosphorylation.
Upon IFN-α/β treatment, accumulation of STAT1 and STAT2 in the cytoplasm rather than in the nucleus was reported to occur in cells infected with measles and rinderpest morbilliviruses (MV and RPV, respectively) (8, 25, 28). Thus, taking advantage of our newly generated recombinant viruses, we next assessed by immunofluorescence analysis whether the nuclear accumulation of STAT1 and STAT2 was modulated by these viruses. In mock-infected Vero-SLAM cells, STAT1 staining exhibited the anticipated shift between cytoplasmic to mainly nuclear localization upon treatment with IFN-α/β (Fig. (Fig.3A,3A, mock). The identical phenotype was observed in cells infected with rA75/17red Vko and rA75/17red CVko. Strikingly, however, cells infected with the rA75/17red and rA75/17red Cko viruses (the two viruses expressing the V protein) exhibited a clear accumulation of STAT1 in the cytoplasm (Fig. (Fig.3A).3A). Identical results were observed for STAT2 under corresponding conditions (Fig. (Fig.3B).3B). Next, to verify whether STAT1 nuclear accumulation was dependent on IFN addition, STAT1 cellular localization was assessed in infected but non-IFN-treated cells. In the absence of IFN treatment, rA75/17red Vko and rA75/17red CVko (data not shown) could not mediate STAT1 nuclear translocation (Fig. (Fig.3C).3C). Rather, STAT1 was readily detected in the cytoplasm of rA75/17red-infected but non-IFN-treated cells, thereby validating the notion that STAT1 is not degraded in rA75/17red-infected cells. The identical phenotypes were observed when the cellular localization of STAT2 was assessed (Fig. (Fig.3C).3C). These observations are in agreement with the results found in our transient transfection assay and confirmed a key regulating role of the CDV-V protein in inhibiting IFN-α/β-dependent activation of innate immunity.
Upon IFN-α/β treatment, STAT molecules are first phosphorylated, then translocate into the nucleus, and finally relocate to the cytoplasm through their nuclear export signal (NES) (1, 24). Hence, CDV-V-dependent STAT cytoplasmic accumulation may result either from nuclear import inhibition or from an accelerated nuclear export mechanism. To discriminate between these two possibilities, the Crm1-dependent nuclear export inhibitor leptomycin B (LMB) was used. Immunofluorescence analysis was then performed to assess the cellular localization of STAT1 in cells infected with rA75/17red or rA75/17red Vko. As a control, we used the simian virus 5 (SV5) V protein ([SV5-V] known to be targeted to the nucleus ) fused at its N-terminal region with an NES-tagged green fluorescent protein (NES-GFP-SV5-V). As expected, upon LMB treatment, the fluorescent fusion protein remained accumulated mainly in the nucleus while it was clearly relocated in the cytoplasm in the absence of the drug (Fig. (Fig.4A).4A). Importantly, in IFN- and LMB-treated cells, STAT1 was found to be strongly accumulated in the nucleus in recombinant V knockout virus-infected cells. Conversely, in cells infected with the wild-type rA75/17red virus, STAT1 could not be detected in nuclei (Fig. (Fig.4B).4B). Together, these results indicate that cytoplasmic accumulation of STAT1 in cells infected with CDVs expressing the V protein is caused by nuclear import inhibition rather than by an accelerated export mechanism.
Coimmunoprecipitation (co-IP) from infected cells was next performed to assess whether the A75/17 CDV-V protein may influence STAT nuclear accumulation by binding to endogenous STAT1 and/or STAT2. Thus, STAT1 from lysates of infected and noninfected cells was immunoprecipitated with an anti-STAT1 monoclonal antibody (MAb), followed by immunoblotting using an anti-CDV-V polyclonal antibody. The identical strategy was employed using an anti-STAT2 MAb to determine whether V can associate with endogenous STAT2. An anti-HA MAb immunoprecipitation or immunoprecipitation in the absence of antibody was performed in parallel to validate the co-IP assay. Indeed, in rA75/17red-infected cells, V could be efficiently copurified after STAT1 and STAT2 immunoprecipitation (Fig. (Fig.5).5). Moreover, these interactions were formed independently of the activation of the signaling pathway since V was coprecipitated in both the presence and the absence of IFN-α/β treatment (data not shown). As expected, V was not coprecipitated in rA75/17red Vko-infected cells or when the anti-HA MAb or no MAb was used for immunoprecipitation (Fig. (Fig.5).5). Western blotting performed with cell lysates taken prior to immunoprecipitation revealed the expected pattern of V expression. Indeed, V was produced by rA75/17red but not by rA75/17red Vko (Fig. (Fig.5).5). Finally, Western blotting using anti-STAT1 and anti-STAT2 antibodies demonstrated that under all conditions both endogenous transcription factors were expressed in very similar amounts (Fig. (Fig.5).5). These results strongly suggest that the A75/17 CDV-V protein controls IFN-α/β-induced signaling by efficiently forming a complex with STAT1 and STAT2.
To initiate the mapping of the CDV-V-dependent STAT-interacting domains, various HA-tagged expression vectors were engineered (Fig. (Fig.6A).6A). A first construct, composed of the shared N-terminal domain between P and V, was produced (VNT; 240 aa) (Fig. (Fig.6A).6A). The second encompassed the cysteine-rich C-terminal domain of V (VCT; 78 aa) (Fig. (Fig.6A).6A). Next, a single substitution (Y110D) was introduced into the N-terminal region of the full-length HA-tagged V protein (308 aa) (Fig. (Fig.6A).6A). This tyrosine is, indeed, highly conserved among morbilliviruses and has been shown for MV-V to be responsible for specific STAT1 binding (4, 5, 8, 28, 35). Interestingly, sequence comparison between several wild-type CDV strains and the large plaque-forming variant of the Onderstepoort (OP) vaccine strain of CDV revealed a substitution at that precise location (Y110D) (Fig. (Fig.6B).6B). Finally, we also fused the HA tag sequence to the N-terminal part of the P gene construct to allow immunoprecipitation under similar conditions (Fig. (Fig.6A6A).
In order to verify the activity of the HA-tagged full-length constructs and to determine the role of each individual V subdomain in inhibiting IFN-α/β-mediated signaling, ISRE luciferase reporter gene assays were performed from IFN-treated Vero-SLAM cells transfected with the various expression plasmids. Two different plasmid concentrations were used for transfection in order to control for the amount of protein expressed. As expected, HA-Vwt strongly suppressed IFN-α/β-mediated signaling in a concentration-dependent manner, whereas HA-P caused a partial inhibition at its higher concentration only (Fig. (Fig.6C).6C). Similarly, the HA-VNT and the HA-Vwt Y110D mutants showed partial inhibition as well when expressed at high concentrations. Conversely, the HA-VCT construct alone was not sufficient to control IFN-α/β-mediated signaling.
To confirm the expression of the various mutants, immunoblot analysis from total cell extract of transfected Vero-SLAM cells was undertaken using an anti-HA MAb to detect the various HA-tagged proteins. Figure Figure7A7A (bottom panel) documents that all proteins, except HA-VCT, were correctly expressed and migrated according to their expected molecular weights in the SDS polyacrylamide gel. It is possible that the extremely small size of HA-VCT affected proper expression and/or stability. Thus, to overcome these putative defects, Vwt, Vwt Y110D, VNT, and VCT were fused to RFP. In addition, these fusion proteins were designed to contain both a small linker peptide and the HA tag sequence to retain both protein functionalities and to facilitate detection and immunoprecipitation, respectively (Fig. (Fig.6D).6D). Indeed, using this strategy, all engineered proteins were properly expressed, as demonstrated by immunoblot analysis (Fig. (Fig.7B,7B, bottom panel). ISRE luciferase reporter gene assays were performed in order to assess the ability of these fusion proteins to control IFN-α/β-mediated signaling. Figure Figure6E6E illustrates that all proteins modulated the IFN-induced activity to the same extent as the identical nonfused V mutant proteins (Fig. (Fig.6C).6C). These results indicate that the CDV-VCT module alone lacks the capacity to control IFN-α/β-induced activity, whereas VNT was able to function as an autonomous module, albeit to a limited extent compared to Vwt.
To verify the proper folding of the C-terminal region of V in HA-VCT- and RFP-HA-VCT-expressing cells, their ability to disrupt signaling by the RNA helicase protein MDA5 was investigated. Indeed, it has been previously reported that the measles virus VCT domain was sufficient to suppress the MDA5-mediated signaling pathway (35). The results shown in Fig. Fig.7C7C indicate that HA-VCT presumably does not fold into a biologically active conformation since this domain was not able to control the MDA5-dependent signaling pathway to the IFN-β promoter reporter gene. In contrast, RFP-HA-VCT was fully able to ablate MDA5-mediated signaling, as were Vwt and Vwt Y110D, but not VNT, with or without fusion to RFP (Fig. (Fig.7C).7C). We thus concluded that VCT alone was very likely misfolded and thus not functional but that it could fold into an active conformation if fused to an irrelevant, stabilizing protein. Importantly, this result furthermore confirms the selective incapacity of CDV-VCT to control IFN-α/β-induced activity on its own.
A coimmunoprecipitation assay was used to investigate the ability of the different V domain mutants to bind the endogenous STAT1 and STAT2 proteins (in the absence of IFN treatment). The anti-HA 3F10 monoclonal antibody was used to immunoprecipitate the different V constructs, followed by immunoblot analysis using anti-STAT1 or anti-STAT2 antibodies for detection. Figure 7A and B document that, independently of whether Vwt was fused to the RFP, both STAT1 and STAT2 could efficiently be copurified (Fig. 7A and B, upper panels). Interestingly, although VNT retained some interaction with STAT1, it displayed significantly reduced binding to STAT2, whereas VCT, even when fused to RFP, showed no binding activity at all to either STAT molecule. The latter results are in agreement with the data obtained in the ISRE luciferase reporter gene assay. Moreover, while the single point mutant Vwt Y110D almost completely lost the capacity to associate with STAT1, we detected a slight interaction with STAT2 (Fig. 7A and B, upper panels). Finally, the ability of the P and C proteins in binding STAT1 and STAT2 was also investigated (HA-tagged versions were constructed). The C protein was unable to bind both STAT1 and STAT2, whereas the P protein weakly interacted with STAT1 only (Fig. (Fig.7A).7A). Immunoblot analysis of STAT1 and STAT2 prior to immunoprecipitation revealed that both transcription factors were expressed to very similar amounts under all conditions. These results suggest that VCT, when fused to VNT, confers the capacity of Vwt to enhance STAT1 binding and to associate with STAT2.
We next assessed whether the differential binding capacities of the V mutants correlated with the nuclear import inhibition mechanism described above. To this purpose, Vero-SLAM or Vero cells were transfected with the various expression plasmids and treated with IFN-α/β, and STAT1/STAT2 cellular localization was subsequently investigated by immunofluorescence analysis (Fig. (Fig.88 and and9,9, for STAT1 and STAT2, respectively). Clearly, while HA-Vwt efficiently inhibited STAT1 and STAT2 nuclear import, HA-P partially retained STAT1 but not STAT2 in the cytoplasm (Fig. (Fig.8A8A and and9A).9A). Interestingly, the extent of STAT1 nuclear import inhibition seemed to correlate with the level of P expression in transfected cells (Fig. (Fig.8A).8A). Conversely, HA-C did not at all suppress IFN-α/β-induced STAT1 and STAT2 nuclear accumulation (Fig. (Fig.8A8A and and9A),9A), confirming the data obtained in the co-IP assay (Fig. (Fig.77).
In order to assess the ability of VNT, VCT, and Vwt Y110D to modulate STAT1 and STAT2 nuclear translocation, the identical experiments were performed using the RFP fusion proteins. Here again, only RFP-HA-Vwt efficiently inhibited both STAT1 and STAT2 nuclear import (Fig. (Fig.8B8B and and9B).9B). Intriguingly, whereas RFP-HA-VNT was able to efficiently retain STAT1 in the cytoplasm, STAT2 mainly localized in the nucleus in the presence of the same construct though these experiments were performed in the presence of IFN. Remarkably, the exact reverse correlation was observed in RFP-HA-Vwt Y110D-transfected Vero-SLAM cells. In this case, STAT1 efficiently translocated to the nucleus in the presence of the V mutant, whereas STAT2 exhibited a clear cytoplasmic accumulation. In contrast, and in agreement with the co-IP assay, RFP-HA-VCT did not function autonomously since both STAT molecules were located in the nucleus in IFN-α/β-treated cells (Fig. (Fig.8B8B and and9B).9B). These results demonstrate that VNT and VCT need to be fused to efficiently control nuclear import of both STAT molecules.
We next confirmed the notion that the phosphorylation state of STAT1 is not affected by the CDV-V proteins. To this aim, immunofluorescence analyses using an anti-phospho-STAT1 MAb were performed in cells transfected with RFP-HA-Vwt, -VNT, -VCT, and -Vwt Y110D and treated with type I IFN. As expected, nuclear translocation of phospho-STAT1 was observed in IFN-treated and RFP-transfected cells. Importantly, in RFP-HA-Vwt- and RFP-HA-VNT-transfected cells, phospho-STAT1 accumulated in the cytoplasm to a greater extent than in nontransfected cells of the same area (Fig. 10A). In contrast, phospho-STAT1 was clearly detected in the nucleus of RFP-HA-VCT- and RFP-HA-Vwt Y110D-transfected cells (Fig. 10A).
The identical experiments were repeated in rCDV-infected cells. Hence, in rA75/17red- and rA75/17red Cko-infected cells, phospho-STAT1 could be detected in the cytoplasm, whereas nuclear staining was observed in rA75/17red Vko- and rA75/17red CVko-infected cells (Fig. 10B). We noticed less staining of phospho-STAT1 in the cytoplasm of infected cells than in transfected cells, probably as a result of the formation of large syncytia by the different recombinant viruses. Taken together, the above data clearly validate the notion that the CDV-V protein inhibits STAT1 nuclear import without affecting its phosphorylation state both in transfected and infected cells.
Taken together, the above results suggest that V is essential in counteracting IFN-α/β-dependent signaling although the P protein was able to exert partial control. To verify whether the different phenotypes described above correlate with differences in growth kinetics, Vero-SLAM cells were infected with the various recombinant CDVs (Fig. (Fig.11).11). Next, cells were treated (or left untreated) with IFN-α/β at 3, 12, 24, and 36 h postinfection. Finally, at 48 h postinfection, virus titers of cell-associated viruses were determined by limiting dilution assay. Interestingly, all viruses had reduced viral titers compared to those obtained in IFN-untreated cells (Fig. (Fig.9).9). This is probably because we used an MOI of 0.02 and treated the cells with IFN as early as 3 h postinfection. Thus, most of the uninfected cells had probably established an antiviral state before being infected, in turn affecting proper viral growth even in the case of V-expressing viruses. Nevertheless, viruses lacking V expression had approximately 10 times less progeny virus production, confirming that V is crucial for counteracting the IFN-α/β-mediated antiviral state. It is important to note that the concentration of IFN used in these experiments (1,000 units/ml) was 10 to 50 times higher than the minimal concentration required to completely inhibit a VSV-induced cytopathic effect in Vero-SLAM cells (data not shown). Nevertheless, albeit not to the same extent, all viruses were able to grow, which suggested that a viral component(s) in addition to V may provide partial control of innate immunity.
To confirm the findings that CDV-V controls IFN-induced STAT1 nuclear import not only in Vero but also in canine cells, we sought to determine the effect of CDV-V in MDCK cells. Indeed, these cells are functional in both IFN induction and IFN action (data not shown). First, a cell line stably expressing the universal morbillivirus receptor CD150/SLAM was produced in order to allow the virulent virus to replicate. Second, since none of the tested recombinant IFN (rIFN) molecules was functional in these cells (human, bovine, feline, and universal rIFN), we cloned the canine IFN-β obtained from primary canine keratinocytes. Finally, the recombinant canine IFN-β was produced in 293T cells. Figure Figure1212 illustrates that upon addition of cIFN-β, STAT1 was, indeed, readily translocated into the nucleus of the engineered MDCK-SLAM cells.
As expected, in cells infected with rA75/17red (Fig. 12A) and rA75/17red Cko viruses (data not shown), STAT1 nuclear import was inhibited, whereas in rA75/17red Vko-infected cells (Fig. 12A and B) and rA75/17red CVko-infected cells (data not shown), STAT1 was located in the nucleus. The identical phenotypes were observed for STAT2 (data not shown). Thus, the main conclusion obtained from Vero-SLAM cells correlated with results in canine cells, which are characterized by an intact IFN response system. In addition, we performed titration experiments at 2 days postinfection in the presence and absence of canine IFN-β in rA75/17red- and rA75/17red Vko-infected cells. The Vko virus grew less efficiently than rA75/17red (Fig. 12C) even in the absence of IFN, presumably because the Vko virus is deficient in suppressing both the IFN action and induction signaling pathways. Nevertheless, IFN treatment reduced the replication of the Vko virus about 2 orders of magnitude, whereas the V-expressing virus was reduced only about 10-fold.
It has recently been reported that V knockout CDV (based on the 5804P virulent strain) was attenuated in infected ferrets, which was associated, at least in part, with inhibition of IFN-α/β induction in peripheral blood mononuclear cells (PBMCs) (47). We now show that the V protein of the highly virulent CDV A75/17 strain also counteracts IFN action by additionally disrupting the IFN-α/β-dependent signaling. Importantly, this does not seem to be valid in only Vero-SLAM cells as preliminary experiments in canine MDCK-SLAM cells provided strong evidence that the effects observed in Vero cells are also active in canine cells. Detailed molecular analysis enabled us to demonstrate that CDV-V specifically ablated the nuclear import of STAT1 and STAT2 without affecting their activated phosphorylation states. Furthermore, inhibition of IFN-α/β-dependent signaling correlated with the capacity of the V protein to efficiently interact with both STAT molecules. Finally, we identified both the N-terminal and the C-terminal regions of V as playing a synergistic role in IFN evasion.
Initial attempts to map the domains of the V protein that interact with STAT1 and STAT2 revealed that the N-terminal region of V was able to function as an autonomous domain interfering with IFN-α/β-induced signaling. Importantly, co-IP experiments indicated that VNT retained association with STAT1 but failed to copurify STAT2. Since the full-length V protein (Vwt) efficiently coprecipitated both STAT molecules, this suggests that VNT is very likely responsible for STAT1 interaction, whereas VCT is necessary to target STAT2. In agreement with this notion, the single-amino acid mutant Vwt Y110D retained slight STAT2 binding but almost completely lost STAT1 interaction. Nevertheless, we cannot exclude the possibility that VCT, when fused to VNT, determines a specific conformational state of VNT that confers the capacity of the N-terminal region to target both STAT molecules. Recent work done with MV is consistent with the former hypothesis since the N-terminal domain of MV-V was assigned to STAT1 binding, whereas STAT2 has been discovered to be the main target of the MV-VCT module (4, 5, 35). The main difference that we observed in this study between both morbillivirus V protein functionalities is that the CDV-VCT domain expressed alone could not disrupt both the IFN-α/β-mediated and the MDA5-mediated signaling pathways, thus suggesting improper folding and/or protein degradation. Remarkably, when stabilized by an irrelevant protein, the VCT domain selectively suppressed MDA5-mediated signaling but not signaling induced by IFN-α/β. A sequence alignment of the MV- and CDV-VCT domains shows only about 50% amino acid identity, which may explain the different functions (Fig. (Fig.13).13). It may be possible that VCT, when fused to RFP but not in the wt protein, adopts a conformational state that remains functional in inhibiting the MDA5-mediated signaling pathway but loses its intrinsic ability to bind STAT1 and/or STAT2 to suppress the IFN-α/β-mediated signaling. Alternatively, VCT may fold similarly when fused to the RFP or when fused to VNT. In this case, the CDV-VCT domain may be responsible for (i) disrupting the MDA5-mediated signaling pathway and (ii) conferring proper folding to the VNT domain, which consequently will efficiently engage STAT1 and STAT2 to control IFN-α/β-mediated signaling. Since Vwt elicited enhanced binding avidity to STAT1 compared to VNT alone, this indeed suggests that VNT's conformational state is modulated by the presence of VCT. Taking these observations together, we propose that CDV-VNT and -VCT are two interdependent modules that function synergistically to allow proper folding of the full-length V protein. In turn, Vwt gains the ability to efficiently interact with STAT1 and STAT2, which offers optimal conditions to prevent nuclear import of the two STAT molecules and, consequently, to control IFN-α/β-mediated signaling. These results contrast with those obtained with the henipavirus family, where the C-terminal region of V was totally dispensable for STAT binding and IFN evasion (38, 40). The requirement of both the N- and C-terminal domains of V to evade innate immunity is consistent with the studies of the rubulavirus family, though in that case both domains of V were necessary to mediate STAT1 (9, 19, 22, 30, 49), STAT2 (17, 18, 26, 29, 30, 50), or STAT3 (44) proteasomal degradation rather than STAT nuclear import inhibition without affecting their phosphorylation states.
Results obtained in ISRE luciferase reporter gene assays clearly demonstrated the capacity of both VNT and Vwt Y110D to interfere with IFN activity although to a limited extent compared to Vwt. Remarkably, immunofluorescence analysis of STAT1 and STAT2 intracellular localizations in the presence of IFN and in the presence of these two V mutants revealed an unanticipated reverse localization of both transcription factors. Indeed, while VNT specifically disrupted STAT1 (but not STAT2) nuclear import, Vwt Y110D impaired STAT2 (but not STAT1) nuclear localization. These results seem to be in contradiction with the current model of the IFN-α/β-induced signaling pathway, which states that phosphorylated STAT1 and STAT2 dimerize upon IFN treatment prior to being translocated to the nucleus. We therefore hypothesize that the association of VNT with STAT1 and the association of Vwt Y110D with STAT2 compete for the dimerization of phosphorylated STAT1 and STAT2. Subsequently, in the presence of VNT and IFN, phosphorylated STAT1 accumulates in the cytoplasm, whereas phosphorylated STAT2 would translocate and accumulate in the nucleus. Conversely, in the presence of Vwt Y110D, which exhibits STAT1 binding impairment but efficient STAT2 association, the reverse phenotype for STAT cellular localization is observed. This hypothesis is in excellent agreement with our results demonstrating that CDV-V does not affect STAT1 and STAT2 phosphorylation, but further experiments are required to consolidate this model.
In addition to V, the CDV-P protein (sharing the identical N-terminal region with V) was found to retain weak interaction with STAT1, which correlated with partial suppression of IFN-α/β-mediated signaling. The biological relevance of this limited interaction with STAT1 was supported by the fact that although the growth of all rCDVs was reduced, they were not completely inhibited in the presence of IFN. This effect of the CDV-P is consistent with observations made with P proteins of other negative-strand RNA viruses, e.g., MV, RPV, Nipah virus, and rabies virus (8, 10, 25, 46). Nevertheless, our results obtained in the co-IP assay indicated that P bound STAT1 rather inefficiently, suggesting that a high concentration of P may be required for effective inhibition of IFN signaling. Supporting this notion, we noticed that at late time points after initial infection (48 h postinfection) all knockout viruses were ultimately able to disrupt nuclear import of STATs (data not shown). It is noteworthy that a differential effect of MV-P in IFN evasion was recently documented. Indeed, results suggested that the origin of the virus strain determined the extent of the P functionality (11). Thus, further investigations are required to demonstrate whether the extent by which the CDV-P protein counteracts the IFN-α/β response is also regulated by the origin of the strain and/or by the host cell environment.
The mechanism by which the A75/17 CDV-V protein inhibits the IFN-α/β-mediated response differs from that of other morbilliviruses. To our knowledge, all studies performed with MV-V, with the exception of that by Palosaari et al. (28), have reported an inhibition of the phosphorylation of STAT1 (5, 42, 48) and STAT2 (4, 8, 42). The reasons for the differences between CDV and MV remain unclear. In addition to genuine biological differences between these two morbilliviruses, the origin and passaging histories of the strains used to study evasion from IFN action may be a factor. Indeed, we studied a highly virulent viral strain not adapted to cultured cells, whereas the strain of MV was attenuated (8) or persistently infected cells were investigated (48). There are also similarities, however, between CDV and other morbilliviruses. Consistent with the findings in CDV, MV-V was shown to be more potent in inhibiting IFN-α/β signaling than the C protein, and this observation was equally true for all strains independent of their virulence (11). Similarly, using recombinant knockout viruses based on a vaccine strain, RPV-V was shown to block IFN-mediated phosphorylation of STAT1 and STAT2 without causing the degradation of these proteins (25).
Taken together, our results shed light on a unique molecular mechanism by which a highly virulent CDV strain interferes with IFN action by disrupting signaling for the synthesis of antiviral proteins. While the mechanism of CDV virulence is likely to be complex and may involve different host cells and interactions between cells, disruption of the IFN defense may deprive the host of an early mechanism known to limit viral replication and spread at a critical stage of infection. Understanding the precise mechanisms of IFN evasion may also lead to the rational design of vaccines that produce an optimally balanced stimulation of the innate immune system, which is known to be essential for activation of an effective adaptive immune response (27).
We thank D. Garcin for offering the red fluorescent marker protein, the luciferase plasmids pISRE-Luc, pβ-IFN-fl-lucter, and pTK-RL, and the NES-GFP-SV5-V and FLAG-tagged MDA5-expressing plasmids, and we thank V. von Messling for providing the Vero-SLAM cells. We are grateful to Ruth Parham for linguistic improvement of the manuscript.
Published ahead of print on 28 April 2010.