|Home | About | Journals | Submit | Contact Us | Français|
Pestiviruses prevent alpha/beta interferon (IFN-α/β) production by promoting proteasomal degradation of interferon regulatory factor 3 (IRF3) by means of the viral Npro nonstructural protein. Npro is also an autoprotease, and its amino-terminal coding sequence is involved in translation initiation. We previously showed with classical swine fever virus (CSFV) that deletion of the entire Npro gene resulted in attenuation in pigs. In order to elaborate on the role of the Npro-mediated IRF3 degradation in classical swine fever pathogenesis, we searched for minimal amino acid substitutions in Npro that would specifically abrogate this function. Our mutational analyses showed that degradation of IRF3 and autoprotease activity are two independent but structurally overlapping functions of Npro. We describe two mutations in Npro that eliminate Npro-mediated IRF3 degradation without affecting the autoprotease activity. We also show that the conserved standard sequence at these particular positions is essential for Npro to interact with IRF3. Surprisingly, when these two mutations are introduced independently in the backbones of highly and moderately virulent CSFV, the resulting viruses are not attenuated, or are only partially attenuated, in 8- to 10-week-old pigs. This contrasts with the fact that these mutant viruses have lost the capacity to degrade IRF3 and to prevent IFN-α/β induction in porcine cell lines and monocyte-derived dendritic cells. Taken together, these results demonstrate that contrary to previous assumptions and to the case for other viral systems, impairment of IRF3-dependent IFN-α/β induction is not a prerequisite for CSFV virulence.
Interference with the innate immune response, particularly the alpha/beta interferon (IFN-α/β) system, has been considered a premise for the successful establishment of infection by an ever-increasing number of viruses (for selected recent reviews, see references 13, 16, and 34), including pestiviruses (1, 2, 7, 8, 33, 37, 43, 44). The strategies adopted by viruses to interfere with the primary immune response are extremely diverse. These strategies include suppression of IFN-α/β production, downregulation of IFN-α/β signaling, and inhibition of IFN-α/β-induced antiviral proteins. Most reports relate the in vivo consequences of a deletion, or inactivation of the gene or gene product interfering with innate immune defenses, to virus attenuation, irrespective of the pathway targeted by the virus (as exemplified in references 50 and 54).
Pestiviruses comprise classical swine fever virus (CSFV), a highly contagious pathogen of pigs (32); bovine viral diarrhea virus (BVDV), affecting cattle; and border disease virus, whose natural host is sheep. These animal viruses belong to the family Flaviviridae along with the flaviviruses and hepatitis C virus, which represent important human pathogens (11). CSFV is an enveloped virus carrying a single-stranded 12.3-kb RNA genome of positive polarity. The proteins encoded by the single open reading frame, flanked by 5′ and 3′ nontranslated regions (NTR), comprise four structural and eight nonstructural proteins (23). The first protein of the open reading frame is Npro, an autoprotease that is present in all pestiviruses but not in the related flaviviruses and hepaciviruses. The Npro protease autocatalytically generates its own carboxy terminus (38, 47, 52).
It is well documented that pestiviruses subvert innate immune defenses by preventing IFN-α/β induction, a property attributed to the viral Npro protein (5, 9, 14, 17, 19, 22, 35, 37, 45). Several recent reports have shown that the Npros of both CSFV and BVDV interfere with IFN-β synthesis by inducing proteasomal degradation of IFN regulatory factor 3 (IRF3) (5, 9, 17, 22, 45). CSFV mutants lacking the complete Npro gene (ΔNpro CSFV) display growth characteristics similar to those of the respective wild-type viruses in porcine cell lines lacking the capacity for IFN-α/β synthesis. In contrast, replication of these Npro mutants is impaired in IFN-α/β-producing cells such as porcine macrophages (37). Importantly, ΔNpro CSFV mutants are attenuated in pigs (25, 49). Furthermore, replacement of the Npro gene in a highly virulent CSFV by the corresponding gene of an avirulent vaccine strain resulted in a virus that was highly virulent (25). This indicated that the varying virulence observed between individual strains of CSFV is not dependent on Npro. All these observations led to the hypothesis that Npro may have key functions in the pathogenesis of classical swine fever (CSF), by interfering with the IFN-α/β system rather than performing an essential function in replication per se (37, 49). However, the in vivo experiments with ΔNpro CSFV did not determine if the attenuation of highly virulent CSFV after Npro deletion was due solely to the loss of the Npro-mediated block of IFN-α/β induction. The possibility remained that the lack of additional functions of Npro, such as the protease function or a suboptimal RNA structure of the internal ribosomal entry site (IRES) due to the fusion of the 5′ NTR to the core (C) gene, was influential.
Therefore, the present study was aimed at specifically defining the role of the Npro-mediated block of IRF3-dependent IFN-α/β induction in the pathogenesis of CSF. To this end, we searched for minimal amino acid changes that would inactivate the Npro-mediated IRF3 degradation without affecting other potential functions of the Npro protein or of the corresponding genomic region. Having demonstrated that IRF3 degradation was independent of the proteolytic activity of Npro, we took advantage of rare CSFV strains reported to induce IFN-α/β in cell culture. Most pestiviruses will not induce IFN-α/β secretion in cell culture; in addition, they also inhibit the induction of IFN-α/β by strong viral inducers such as Newcastle disease virus (NDV) (35, 37). This was initially described as the “exaltation of NDV” (END) phenomenon, based on the observation that NDV replicates more efficiently in pestivirus-infected cells than in cells devoid of pestivirus infection (20, 21). The aforementioned rare CSFV strains that induce IFN-α/β in cell culture do not exhibit this END phenotype. Such strains were designated END negative (END−) CSFV (39, 41, 46). Accordingly, we focused on amino acid residues occurring uniquely in the END− CSFV strains GPE−, Ames-END−, and ALD-END− to identify amino acid substitutions that would specifically eliminate the capacity of Npro to degrade IRF3. Subsequently, we analyzed the effect of these mutations in CSFV infections in 8- to 10-week-old pigs to determine if this particular function of Npro played a role in CSFV virulence.
The swine kidney cell lines SK-6 (kindly provided by M. Pensaert, Faculty of Veterinary Medicine, Ghent, Belgium) and PK-15 (American Type Culture Collection [ATCC], Manassas, VA) were propagated in Earle's minimal essential medium (EMEM) and in Dulbecco's minimal essential medium supplemented with 7% horse serum (SVA, Hatunaholm, Sweden), respectively. HEK 293T cells (ATCC) were propagated in EMEM supplemented with 7% fetal bovine serum (Biochrom AG, Switzerland). The MDBK-t2 cell line was kindly provided by Martin D. Fray (Institute of Animal Health, Compton, Newbury, Berkshire, United Kingdom) and maintained in EMEM supplemented with 7% fetal bovine serum and 10 μg/ml blasticidin (Invitrogen).
CSFV strains vA187-1, vA187-ΔNpro, vA187-ΔNpro-Ub and vEy-37 were derived from the full-length cDNA clones pA187-1 (36), pA187-ΔNpro (37), pA187-ΔNpro-Ub (49), and pEy-37 (26), respectively. The vA187-1- and vEy-37-derived viruses with single (vA187-H5Y, vA187-L32S, vA187-K61N, vA187-C112R, vA187-D136N, vEy-C112R, and vEy-D136N) or multiple (vA187-H5Y-L32S-D136N) amino acid mutations in Npro were generated from the corresponding mutant cDNA plasmids obtained by PCR-based site-directed mutagenesis with oligonucleotide primers containing the respective mutations and PfuUltra DNA polymerase (Stratagene), using previously described techniques (24). The subscript numbers refer to the amino acid position in the CSFV polyprotein. The bicistronic virus vA187-UbNpro-EI-C was rescued from pA187-UbNpro-EI-C, an infectious cDNA clone containing downstream of the viral 5′ NTR a cistron composed of the murine ubiquitin (Ub) gene fused to the Npro gene followed by a stop codon, the encephalomyocarditis virus (EMCV) IRES, the C gene, and the remainder of the vA187-1 sequence. 5′ and 3′ truncations of the Npro gene were introduced in the latter cDNA clone by PCR-based site-directed mutagenesis as described above, and the respective mutant bicistronic viruses were derived accordingly. The CSFV strain GPE− was obtained from the Community Reference Laboratory for CSF, Institute of Virology, School of Veterinary Medicine, Hannover, Germany. The CSFV strain Ames-END− was isolated from the Ames strain by the reverse plaque formation method (40, 41). The recombinant viruses vA187-Npro(GPE−) and vEy-Npro(GPE−) were generated from pA187-Npro(GPE−) and pEy-Npro(GPE−), obtained by replacement of the Npro gene in pA187-1 and pEy-37, respectively, with the corresponding sequence of the GPE−strain. All cDNA-derived viruses were rescued essentially as described elsewhere (30). Briefly, plasmid constructs were linearized at the SrfI site located at the end of the viral genomic cDNA sequence, and RNA was obtained by runoff transcription using the MEGAscript T7 kit (Ambion). After DNase I treatment and purification on S-400 HR Sephadex columns (GE Healthcare), RNA was quantified with a spectrophotometer and used to electroporate SK-6 cells.
Plasmids pA187-1 (36) and pEy-37 (26), carrying the full-length genome of the Alfort/187 (GenBank accession number X87939) and Eystrup (GenBank accession number AF326963) strains, respectively, served as a basis for the construction of all mutant viruses described above. The nucleotide and amino acid numbers included in the present study refer to the sequences of the respective strains. All mutant constructs were generated using standard DNA cloning techniques. Mutagenesis was performed by PCR with PfuUltra DNA polymerase (Stratagene). Plasmids pCI-Flag-Npro and pCI-Npro-C*-EGFP are based on the pCI expression vector (Promega). pCI-Flag-Npro encodes downstream of the AUG start codon the Flag epitope DYKDDDDK fused in frame to the Npro gene of the Alfort/187 strain lacking the start codon. pCI-Npro-C*-EGFP contains the Npro gene of the Alfort/187 strain followed by the 5′-terminal 243 nucleotides of the C gene, a sequence encoding a GGSGGS linker, and the enhanced green fluorescent protein (EGFP) gene derived from pEGFP-N1 (Clontech). In the Npro genes of plasmids pCI-Flag-Npro and pCI-Npro-C*-EGFP, the single amino acid substitutions H5Y, E22V, L32S, H49L, K61N, C69S, C69A, C112R, C112K, C112S, C112D, C112A, D136N, D136E, D136Q, D136R, D136V; the amino acid deletions at position 112 (ΔC112) and 136 (ΔD136); and the 5′-terminal deletions resulting in the lack of 19, 22, and 27 amino acids at the amino terminus of Npro (NΔ19, NΔ22, and NΔ27) were obtained by PCR-based site-directed mutagenesis as described above. The Npro genes of strains GPE− and Ames-END− were amplified from RNAs of the respective viruses by reverse transcription-PCR using appropriate primers located in the 5′ NTR and C gene of the viral genome and cloned in pCR4-TOPO (Invitrogen) for GPE− Npro and in pGEM-T Easy (Promega) for Ames-END− Npro. The respective Npro genes were then subcloned in pCI downstream of the Flag epitope in analogy to pCI-Flag-Npro as described above. The S32L and N136D mutations in GPE− Npro and the R112C mutation in Ames-END− Npro were constructed as described above. For cytomegalovirus promoter-driven transcription of partially nontranslatable Npro RNA, nucleotide 195 of the Npro gene was deleted in pCI-Flag-Npro, resulting in full-length Npro RNA expressing only amino acids 2 to 74 of Npro fused to the Flag tag. Plasmid pCI-Ub-Core was described before (35). For the mammalian two-hybrid assays, the porcine IRF3 gene (5) was introduced between the SgfI and PmeI sites of pFN10A(ACT) (Promega) in frame with the VP16 gene. The Npro gene of the Alfort/187 strain and the mutants thereof encoding the C69A, C112R, and D136N substitutions were cloned into the SgfI and PmeI sites of pFN11A(BIND) (Promega) in frame with the yeast GAL4 gene. To this end, the respective genes were flanked with the sequences of the SgfI and PmeI restriction sites using standard PCR techniques as described above. The reporter plasmid p125Luc expressing firefly luciferase under the control of the IFN-β promoter (53) was a gift of Takashi Fujita, and phRL-SV40 constitutively expressing Renilla luciferase was obtained from Promega. Plasmids were propagated in Escherichia coli XL-1 blue cells (Stratagene) and purified using the NucleoBond plasmid DNA purification system (Macherey-Nagel). All newly synthesized and cloned DNA fragments were verified by nucleotide sequencing using the Thermo Sequenase DYEnamic direct cycle sequencing kit (GE Healthcare) and the Global IR2 DNA sequencing system with e-Seq and AlignIR software (LI-COR Biosciences).
The E2-specific monoclonal antibody (MAb) HC/TC26 (15) was kindly provided by I. Greiser-Wilke (Hannover Veterinary School, Hannover, Germany). The rabbit polyclonal antisera against recombinant affinity-purified CSFV Npro (35), CSFV C, and porcine IRF3 (5) were previously described. The anti-FLAG M2 MAb was purchased from Sigma. The MAb JL-8 (Clontech) was used for the detection of EGFP. Secondary antibodies Alexa Fluor 680 goat anti-rabbit immunoglobulin G (IgG) and IRDye 800 goat anti-mouse IgG were from Molecular Probes Inc., and Rockland Inc., respectively. The horseradish peroxidase-conjugated rabbit anti-mouse IgG was from Dako.
The virus titer, expressed in 50% tissue culture infective doses (TCID50)/ml, was determined by end point dilution on SK-6 cells and immunoperoxidase staining using MAb HC/TC26 as described earlier (29).
Proteins were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed by Western blotting essentially as previously described (35). Briefly, cells were lysed in a hypotonic buffer containing 20 mM morpholinepropanesulfonic acid, 10 mM NaCl, 1.5 mM MgCl2, 1% Triton X-100 (pH 6.5), and 1 μl protease inhibitor cocktail (Sigma) per 106 cells to be lysed. After clarification by centrifugation at 15,000 × g for 10 min at 4°C, equal amounts of total protein were separated by SDS-PAGE under reducing conditions and transferred onto a 0.2-μm nitrocellulose membrane (Bio-Rad) using a Trans-Blot semidry transfer device (Bio-Rad). Membranes were blocked for 1 h at room temperature or overnight at 4°C with Odyssey blocking reagent (LI-COR Biosciences). For immunodetection, the membranes were incubated at 37°C with the respective primary and secondary antibodies for 30 min each in Odyssey blocking reagent. Between each incubation step, membranes were washed with phosphate-buffered saline supplemented with 0.1% Tween 20. After two final wash steps in the absence of detergent, the signal was acquired using the Odyssey infrared imaging system (LI-COR Biosciences).
Protein interactions were assayed with the CheckMate/Flexi Vector mammalian two-hybrid system (Promega). HEK 293T cells seeded at a density of 105 cells per well in a 24-well plate were transfected with a mixture of 100 ng/well each of pFN10A(ACT)- and pFN11A(BIND)-derived plasmid or control vector and 50 ng/well of the pGL4.31(luc2P/GAL4UAS/Hygro) reporter plasmid (Promega). The DNA transfections were performed with FuGENE 6 transfection reagent (Roche) according to the manufacturer's protocol with a 5:1 (vol/wt) ratio of FuGENE 6 to DNA. Cells were incubated for 48 h at 37°C prior to extraction and quantification of firefly and Renilla luciferase using the dual luciferase reporter assay system (Promega) and a Centro LB 960 luminometer with two automated injectors (Berthold Technologies). The individual firefly luciferase activities were normalized to the respective Renilla luciferase activity. The relative induction of firefly luciferase activity was then calculated based on the activity obtained with the empty vectors set to 1. Expression of the respective fusion proteins was analyzed by SDS-PAGE and Western blotting using the CSFV Npro- and porcine IRF3-specific rabbit antisera described above.
The reporter assay for IFN-β promoter activation was carried out as described before (35). Plasmid p125Luc served as a reporter for the inducible IFN-β promoter activity, and plasmid phRL-SV40 (Promega) was used for internal normalization. Briefly, HEK 293T cells were transfected with the expression and reporter plasmids using FuGENE 6 (Roche) as described above and challenged with poly(IC) transfection 24 h later. At 28 h after poly(IC) treatment, firefly and Renilla luciferase were quantified using the dual luciferase reporter assay system (Promega) and a Lumat LB 9507 or a Centro LB 960 luminometer with two automated injectors (Berthold Technologies), and the firefly luciferase values were normalized as described above.
The bioactivity of porcine IFN-α/β was assayed as previously described (35, 37) using the Mx/CAT reporter gene assay developed for the quantification of bovine IFN-α/β (12) and kindly provided by Martin D. Fray (Institute of Animal Health, Compton, Newbury, Berkshire, United Kingdom). To this end, the culture supernatants and the porcine serum samples were added to the MDBK-t2 cells at appropriate dilutions allowing measurement in the linear range of the assay. Recombinant porcine IFN-α produced in HEK 293-EBNA cells (3) was used as a standard. Chloramphenicol acetyltransferase (CAT) was quantified using a CAT enzyme-linked immunosorbent assay (Roche). Infectious CSFV present in the samples was neutralized with polyclonal pig anti-CSFV serum, and poly(IC) was removed by RNase treatment as described previously (35).
Specific-pathogen-free (SPF) pigs, 8 to 10 weeks old and with a body weight of approximately 30 to 40 kg, obtained from the breeding unit at the Institute of Virology and Immunoprophylaxis were infected via the oronasal route with a defined dose of virus (expressed in TCID50 based on virus titration on SK-6 cells). Virus used for inoculation was diluted in 10 ml EMEM, and one half of the dose each was administered intranasally and orally. All inocula were titrated on SK-6 cells immediately before and after use to confirm the inoculation dose. For each virus to be tested, groups of three or four littermates were kept in separate isolation units. Body temperature and clinical score were monitored daily according to a defined scoring system as described previously (28). At daily intervals, whole blood was collected for serum preparation and EDTA-blood for white blood cell count and for preparation of leukocytes.
For BVDV, it was shown that nearly the entire Npro, especially structures of the amino-terminal region, was critical for prevention of IFN-α/β induction (9, 14, 17). In order to determine if this was similar for the Npro of CSFV, we constructed viruses expressing amino- and carboxy-terminally truncated Npro. In the natural situation, Npro generates the amino terminus of the C protein. Thus, the autoprotease function of Npro is a prerequisite for the synthesis of viable virus. For analyzing Npro deletion mutants independently of the autoprotease function of Npro, we inserted an IRES sequence of the EMCV (EI) between the Npro and the C genes, to obtain the bicistronic virus vA187-UbNpro-EI-C (Fig. (Fig.1A).1A). The mouse Ub gene was inserted upstream of the Npro gene to avoid variations in the translation efficiency due to diverse nucleotide sequences when the Npro gene was truncated at its 5′ end. As expected, following infection of PK-15 cells, no IFN-α/β bioactivity was measured in the culture supernatant, whether monocistronic vA187-1 or bicistronic vA187-UbNpro-EI-C virus carrying the entire Npro gene was employed. This shows that Npro can block IFN-α/β induction in the bicistronic context (Fig. (Fig.1B).1B). Deletion of up to 19 amino acids at the amino terminus of Npro did not abolish this function. However, truncation of at least 22 amino acids at the amino terminus or of 33 amino acids at the carboxy terminus of Npro resulted in viruses that induced IFN-α/β similarly to viruses with deletion of the complete Npro (vA187-ΔNpro and vA187-ΔNpro-Ub). This demonstrates that the functional domains of Npro relating to inhibition of IFN-α/β induction are located downstream of the glycine residue at position 19 (G19).
Residue E22 is involved in the proteolytic activity of Npro (38). Considering that deletion of the amino-terminal 22 amino acids of Npro abolishes the Npro-mediated inhibition of IFN-α/β induction, this may suggest a requirement for a functional Npro protease. For BVDV, there is good evidence that the proteolytic activity of Npro is not required for the inhibition of IFN-α/β induction (9, 14, 17). In order to determine if this was also true for the Npro of CSFV, we constructed plasmids for eukaryotic expression of a fusion polypeptide consisting of Npro, carboxy-terminally truncated C, a flexible amino acid linker, and EGFP (Npro-C*-EGFP) (Fig. (Fig.2A).2A). The three residues E22, H49, and C69, considered to form the active site of the autoprotease (38), were mutated individually or in combination. Based on the results obtained with the bicistronic viruses (Fig. (Fig.1),1), we also included expression constructs lacking codons for 19, 22, and 27 amino acids at the amino terminus (Fig. (Fig.2A).2A). The polyprotein processing between Npro and C of the parent and mutant Npro-C*-EGFP was monitored in transient-expression experiments in HEK 293T (Fig. (Fig.2B)2B) and SK-6 (data not shown) cells. In parallel, the same expression constructs were analyzed for their effect on poly(IC)-mediated IFN-β induction in HEK 293T cells. After mutagenesis of amino acids E22, H49, and C69, cleavage was either partially (E22V and C69S) or completely (H49L, C69A, E22V-H49L, and E22V-C69S) abolished. Also, the deletion of 22 and 27, but not of 19, amino acids at the amino terminus of Npro was sufficient to prevent its proteolytic activity (Fig. (Fig.2B).2B). In agreement with the results obtained with the bicistronic viruses (Fig. (Fig.1B),1B), Npro lacking the 19 amino-terminal amino acids prevented poly(IC)-mediated activation of the IFN-β promoter, whereas constructs lacking 22 and 27 amino acids had lost this function (Fig. (Fig.2B2B).
These results indicate that amino acids 20 to 22 are crucial for the interference with IFN-β induction. Indeed, the E22V substitution generated an Npro that had completely lost its capacity to prevent IFN-β promoter activation. In contrast, the protease activity was still partially maintained (Fig. (Fig.2B).2B). When C69 of the catalytic domain of the Npro protease was mutated to serine or to alanine, the Npro-mediated inhibition of the IFN-β promoter activation was not affected. With C69S, the protease was still partially active, while with C69A, there was no detectable proteolytic activity remaining (Fig. (Fig.2B).2B). These data demonstrate that the absence of protease activity does not necessarily abolish the inhibition of the IFN-β promoter induction. Conversely, as described above, the E22V mutation that partially maintains protease activity can fully abolish the inhibition of the IFN-β promoter activity. These results were confirmed with bicistronic viruses carrying the respective mutations of the proteolytic domains (data not shown). Altogether, based on the results of this study, the autoprotease activity is not required for Npro of CSFV to block IFN-α/β induction.
The work described above demonstrated that the Npro-mediated block of IFN-α/β induction is independent of an active protease. Consequently, we sought to identify single residues influencing Npro interference with the IFN-α/β induction pathway, independently of the proteolytic function of Npro. For this, we took advantage of the END− CSFV described to induce IFN-α/β in various cell culture systems. Figure Figure33 shows an alignment of the Npro amino acid sequences of the END− strains GPE−, ALD-END−, and Ames-END− and of a selection of END+ CSFV strains. Special attention was given to the amino acids unique to these END− strains. The alignment revealed H5Y, L32S, K61N, C112R, and D136N as candidate mutations that might abolish the Npro-mediated IRF3 degradation and consequently confer the IFN-α/β induction phenotype to CSFV.
To determine whether the residues identified above were critical for the Npro-mediated IRF3 degradation, we constructed CSFV mutants with multiple or single amino acid substitutions in the Npro of the vA187-1 backbone virus (Fig. (Fig.4A).4A). Following infection of PK-15 cells with the respective mutant and parent viruses, IRF3 degradation was analyzed and IFN-α/β bioactivity measured. For this work, the original GPE− strain was amplified on SK-6 and PK-15 cells. Interestingly, the steady-state titer of GPE− virus recovered from SK-6 cells was approximately 102 TCID50/ml higher than the titer obtained with PK-15 cells. As expected from its END− phenotype, the GPE− virus induced IFN-α/β in PK-15 cells but not in SK-6 cells (data not shown). The latter do not produce IFN-α/β upon stimulation with double-stranded RNA (37). Infection of PK-15 cells with vA187-ΔNpro and with the chimeric virus vA187-Npro(GPE−) led to IFN-α/β induction, along with a lack of IRF3 degradation; this contrasted with infection by the vA187-1 parent virus (Fig. (Fig.4B).4B). These results showed that the Npro gene of GPE− was sufficient to confer the GPE− phenotype to vA187-1. We also observed IFN-α/β induction and lack of IRF3 degradation in PK-15 cells following infection with a chimeric vA187-1 virus carrying the Npro gene of the Ames-END− strain (data not shown).
We then systematically introduced the candidate amino acid substitutions found in the GPE−, ALD-END−, and Ames-END− strains (Fig. (Fig.33 and and4A),4A), in combination or as single mutations, into the Npro gene of the vA187-1 backbone (Fig. (Fig.4A).4A). These mutant viruses were then analyzed for their capacity to induce IFN-α/β and degrade IRF3 in PK-15 cells. With this approach, we identified two single-amino-acid substitutions, namely, a cysteine-to-arginine substitution at position 112 (C112R) and an aspartic acid-to-asparagine substitution at position 136 (D136N), that are individually responsible for the phenotypic switch from END+ to END− in the three END− viruses analyzed. Each of these two mutations abolished the capacity of Npro to degrade IRF3 (Fig. (Fig.4B)4B) but did not alter the replication characteristics of the virus in SK-6 cells (Fig. (Fig.4C4C).
Having demonstrated that C112 and D136 are key residues of Npro for mediating IRF3 degradation and thus preventing IFN-β induction, we next analyzed the amino acid requirement at these two positions. Flag-tagged Npro mutants were analyzed in the transient IFN-β promoter reporter assay for their capacity to prevent poly(IC)-mediated IFN-β induction. At position 112 of Npro of the vA187-1 strain, mutation of the cysteine to a basic, acidic, or hydrophobic amino acid abolished the Npro function. A serine at this position partially retained the Npro function. Conversely, replacement of the arginine of Npro of the Ames-END− strain with a cysteine (R112C) restored the capacity of Npro to prevent IFN-β promoter activation (Fig. (Fig.5A).5A). This suggests a strict requirement for a cysteine residue at position 112 of Npro for its function in prevention of IFN-β induction. At position 136, the aspartic acid could be replaced with a glutamic acid without any loss of function. All other nonconservative substitutions, in particular the substitution with an arginine, abolished the Npro-mediated inhibition of IFN-β promoter activation; this demonstrated the requirement for a negatively charged amino acid at this position (Fig. (Fig.5B).5B). Interestingly, the two substitutions S32L and N136D together were required for conferring to Npro of CSFV GPE− the capacity to block poly(IC)-mediated induction of the IFN-β promoter (data not shown).
Negatively charged residues and cysteines are likely to be involved in protein-protein interactions. In a previous report, we have shown that Npro interacts with IRF3 (5). We therefore asked whether the residues C112 and D136 were involved in this interaction. To this end, we performed a mammalian two-hybrid assay with porcine IRF3 and the C112R and the D136N mutants of Npro. Coexpression of Npro fused to the yeast GAL4 DNA-binding domain and of porcine IRF3 fused to the VP16 activation domain resulted in 300-fold induction of the GAL4 promoter-driven luciferase expression, compared to the empty vectors, confirming the interaction of Npro with IRF3 (Fig. (Fig.6).6). Both the C112R and the D136N mutations completely abolished this interaction, with only a twofold induction. The C69A mutation, which does not alter the Npro-mediated IRF3 degradation (data not shown) and block of IFN-β induction (Fig. (Fig.2),2), does not abrogate the interaction of Npro with IRF3. There were comparable protein expression levels of IRF3 and parent and mutant Npro fusion proteins in parallel cellular extracts (Fig. (Fig.6).6). These data strongly indicate a crucial role of the two residues C112 and D136 and of the carboxy-terminal region of CSFV Npro in the interaction of Npro with IRF3.
Inactivation of a viral gene function that interferes with the IFN system is expected to result in attenuation. Indeed, deletion of the entire Npro gene from the highly virulent vEy-37 strain resulted in an avirulent virus (25). However, a complete gene deletion may affect different functions. In order to specifically investigate the role of the Npro-mediated IRF3 degradation in pathogenesis of CSF, we studied the pathogenicity in pigs of vEy-37-derived viruses carrying minimal mutations in Npro specifically affecting Npro-mediated IRF3 degradation. First, we introduced the three GPE−-specific substitutions H5Y, L32S, and D136N or the D136N mutation alone in the vEy-37 backbone, resulting in vEy-Npro(GPE−) and vEy-D136N, respectively. In order to exclude the accidental introduction of any attenuating mutation elsewhere in the genome during cloning, we reconstructed the parent vEy-37 virus by introducing the wild-type Npro sequence in the vEy-D136N backbone. As expected, vEy-Npro(GPE−) and vEy-D136N induced IFN-α/β in PK-15 cells and did not induce IRF3 degradation, contrasting with the parent vEy-37 (data not shown).
Groups of four 8-week-old SPF pigs were infected via the oronasal route with 105 TCID50/pig of mutant or parent virus. Infection with vEy-37 resulted in the typical clinical picture of acute CSF with severe symptoms and a biphasic body temperature curve (Fig. 7A and B, right). The pigs were euthanized when the clinical score was between 18 and 20. Surprisingly, pigs infected with the mutant viruses developed severe disease, with a slight delay of 1 day compared to wild-type virus infection before the onset of clinical signs. One animal infected with vEy-Npro(GPE−) (pig 988), as well as the four animals infected with vEy-D136N, died unexpectedly between days 7 and 8 postinfection, although their clinical status (clinical scores between 13 and 16) was not critical (Fig. 7A and B, left and middle). Necropsy of the latter animals, as well as of pig 990 [vEy-Npro(GPE−) group] (Fig. (Fig.7,7, left) euthanized on day 9, revealed marked hemoabdomen and extensive ecchymotic and petechial hemorrhages within various organs, including tonsils, stomach, kidneys, urinary bladder, intestine, and muscles. Two pigs infected with vEy-Npro(GPE−) spontaneously recovered and were free of any symptoms by day 12 postinfection. Sequence analysis of the Npro gene of the virus rescued from organs of animals that had died showed that the mutations had not reverted in any of the pigs (data not shown). Early after infection, similar levels of viremia were observed in the three groups (Fig. (Fig.7C).7C). Interestingly, the highest values of IFN-α/β bioactivity were measured in the sera of the animals infected with the Npro mutant viruses (Fig. (Fig.7D7D).
In order to confirm these unexpected observations and to determine whether the exacerbation of disease was associated specifically with the presence of the D136N mutation, we repeated the experiment with the inclusion of a vEy-37-derived mutant carrying the C112R substitution (vEy-C112R) instead of the vEy-Npro(GPE−) virus. The pigs were infected at 10 weeks of age with the same dose of virus as in the previous experiment. Again, only a very little difference was observed between the three groups in terms of clinical score, body temperature, viremia, and IFN-α/β bioactivity in the serum (Fig. (Fig.8).8). The overall picture was reminiscent of the first experiment (compare Fig. Fig.77 and and8).8). Two of the four pigs infected with vEy-C112R reached clinical scores of 18 and 20 at day 6 postinfection and were euthanized (Fig. (Fig.8A,8A, left). All the remaining animals showed severe symptoms 1 day later and were euthanized for animal welfare reasons. Based on the clinical score and body temperature, it is likely that two of the pigs infected with vEy-C112R may have survived the infection (Fig. 8A and B, left). It is noteworthy that the highest IFN-α/β bioactivity was again measured in the sera of animals infected with the Npro mutant viruses, with values between 5,000 and 6,000 U/ml serum (Fig. (Fig.8D,8D, left and middle). Taken together, the data from the two experiments demonstrate that the Npro-mediated degradation of IRF3 is not a prerequisite for a highly virulent CSFV to express its virulence.
The unexpected results with vEy-37-derived Npro mutants in vivo suggest that a highly virulent CSFV does not rely on counteracting the IFN system to replicate in vivo and cause severe disease. In order to better understand the role of Npro in vivo, we sought to verify the findings with a less virulent virus. Accordingly, a third experiment with C112R and D136N mutants of the moderately virulent vA187-1 virus was performed. In the SK-6 cell line, which is incapable of IFN-α/β production, the vA187-C112R, the vA187-D136N, and the vA187-1 viruses were indistinguishable with respect to their virus replication characteristics, despite the absence of IRF3 degradation with the mutant viruses (Fig. 4B and C). Three groups of three 9-week-old SPF pigs were infected with 106 TCID50/pig of the respective virus via the oronasal route. Figure Figure99 shows clinical scores, body temperatures, viremia, and serum IFN-α/β levels monitored as in the previous two experiments. Between days 7 and 11 postinfection, the vA187-1-infected pigs were euthanized when they reached a clinical score of above 18 (Fig. (Fig.9A,9A, right). The pigs infected with the vA187-C112R and the vA187-D136N mutant viruses reached maximal clinical scores of 7 and 11, respectively, and completely recovered within 12 to 13 days (Fig. (Fig.9A,9A, left and middle). The body temperatures reflected the milder symptoms observed with the mutant viruses (Fig. (Fig.9B).9B). All pigs infected with the mutant viruses could eventually eliminate the virus, whereas the vA187-1-infected animals had virus titers of up to 105 TCID50/ml serum at the time of euthanasia (Fig. (Fig.9C).9C). Interestingly, the overall IFN-α/β levels in the serum were approximately four to five times lower with all three viruses than the corresponding IFN-α/β bioactivities measured with the vEy-37-derived viruses (Fig. (Fig.9D9D and and8D8D [note the different scales of the y axes]); the highest values in the third experiment were observed with vA187-1 virus (Fig. (Fig.9D,9D, right). This set of data indicates that the effect of the loss of Npro-mediated IRF3 degradation becomes more apparent with a less virulent virus but results in only a moderate attenuation.
Npro of pestiviruses has at least two functions: an autoprotease activity for cotranslational cleavage from the nascent polyprotein and an antagonistic effect on the IFN-α/β induction pathway by targeting IRF3 for proteasomal degradation. All reports of in vivo experiments with pestiviruses in which nearly the entire Npro gene was deleted show that the lack of Npro results in considerable or total attenuation (25, 27, 49). It was therefore plausible to hypothesize that the degradation of IRF3 allows pestiviruses to colonize their host efficiently and express their pathogenic potential. In order to verify this hypothesis, we searched for minimal amino acid substitutions that would specifically abrogate the Npro-mediated block of IFN-α/β induction, to elaborate on the role of this Npro function in vivo.
The Npro autoprotease was identified as a novel type of cysteine protease with some similarity to subtilisin-like proteases (38, 47, 52). Site-directed mutagenesis and translation in cell-free lysates revealed that amino acid residues E22, H49, and C69 of Npro were essential for proteolysis (38). Our observation that the deletion of 22 or more amino acids at the amino terminus inactivates the Npro protease is in agreement with these previous studies. Interestingly, the E22V and C69S mutations only partially reduced the processing activity of Npro. This is in disagreement with previously published data showing complete inactivation of the protease with these two substitutions in Npro of BVDV (9, 17) and CSFV (38). This discrepancy may be due in part to the use of rabbit reticulocyte lysate for expression in certain studies (17, 38).
With respect to the control of IFN-α/β induction by CSFV, it was not clear if the protease activity of Npro was involved. An initial study with BVDV suggested that the proteolytic activity of Npro would be required for suppressing IRF3-mediated responses (19). However, subsequent observations showed that the proteasome inhibitors MG132 and epoxomicin prevented IRF3 degradation in the presence of Npro expression by both BVDV and CSFV (5, 9, 17, 45). This would indicate that the loss of IRF3 was more likely due to proteasomal rather than Npro-mediated proteolysis. With Npro harboring mutations that completely inactivate its proteolytic activity, it became clear for both BVDV (9, 14, 17) and CSFV (this study) that Npro counteracts IFN-α/β production independently of an active protease. This is particularly obvious in the present study, where we monitored Npro-mediated protein processing and IFN-β promoter induction using the same expression constructs in parallel transfections. Mutation of glutamic acid at position 22 to valine (E22V) both reduces the proteolytic activity of Npro and abolishes the capacity of Npro to block IFN-β induction. This suggests that the autoprotease domain and the IRF3-degrading domain structurally overlap, sharing common essential residues. It was also proposed that the BVDV Npro-mediated block of IFN-α/β induction involved the amino-terminal (14) and carboxy-terminal (9, 17) regions of Npro. For CSFV, we have now shown that the amino terminus is dispensable for blocking IFN-α/β induction; Npro functions in the absence of the 19 amino-terminal amino acids, both in a viral context and in transient assays.
In order to pursue our search for amino acid residues of Npro that are essential for Npro-mediated degradation of IRF3 and inhibition of IFN-α/β induction, we focused on amino acids unique to END− CSFV strains. The END− strain GPE− has been used as a live attenuated vaccine in Japan. It was derived from the virulent strain ALD by serial passage in guinea pig kidney cells. During this procedure, the virus was attenuated, additionally losing its capacity to prevent IFN-α/β induction. The same phenomenon was observed with the END− CSFV strains Ames-END− and ALD-END− (39, 41). All of these END− viruses have a functional autoprotease, which led us to hypothesize that residues of Npro found solely in these strains might be critical for the autoprotease-independent function of Npro in the IFN induction pathway. After introduction of these individual changes in a standard CSFV backbone, we did indeed identify two single amino acids critical for Npro-mediated IRF3 degradation: C112 and D136.
We proposed a model for CSF pathogenesis in which CSFV counteracts IFN-α/β induction by depleting IRF3 in the primary target cells, allowing the virus to establish a productive infection. In a second stage, the virus infects plasmacytoid dendritic cells that can produce IFN-α independently of IRF3 (18); this results in overproduction of IFN-α, orchestrating the immunopathological effects typical of CSF (4, 5, 48). The in vivo data in the present report with CSFV mutants specifically lacking the capacity to degrade IRF3 demonstrate that the situation is clearly more complicated. In particular, the pertinence of Npro-mediated IRF3 degradation to disease development during the acute phase of CSF is limited. The loss of Npro-mediated IRF3 degradation did not result in attenuation of highly virulent CSFV, while only a moderate attenuation was observed with a less virulent strain. Certainly, it is important to note that all experiments were performed with 8- to 10-week-old pigs, and we cannot discount the possibility that Npro-mediated attenuation might be more prominent in older animals. Nevertheless, in contrast to our original hypothesis based on observations with other virus systems, our data now indicate that Npro-mediated IRF3 degradation has a minor contribution in terms of permitting virus replication and disease development during the acute phase of CSF. We therefore speculate that Npro plays a more important role in the sense of promoting a longer duration of virus infection in its target cells and in the host. This might be important also for virus persistence in the wild boar, the natural CSFV reservoir. In domestic pigs, the IRF3-degrading function of Npro would certainly play a role with moderately virulent or low-virulent strains that tend to induce more chronic disease. Chronic CSFV infections can last up to 100 days (51). Npro might allow the virus to persist in a particular cell type or organ, for example, in vascular endothelial cells (6), lymph nodes, spleen, and intestine, in particular the gut-associated lymphoid tissue (10, 42). It is also possible that Npro is important for establishing persistent infections in offspring of animals infected during pregnancy. While the importance of this pathway of infection is not clear for CSF, it represents a key characteristic of BVDV infections in cattle. Persistence of noncytopathogenic BVDV in the bovine fetus was associated with failure of the virus to induce IFN-α/β (8). A recent report suggests that Npro and Erns of BVDV are involved in persistence after transplacental infection (27). Such studies with the present mutants generated for BVDV are required to clarify this point, in particular whether both the RNase function of Erns and the IRF3 degradation mediated by Npro are required for the establishment of persistent infections.
Both the present work and previous studies (48) have noted an early, strong, systemic IFN-α response following CSFV infection. The fact that the viral titers continue to rise despite this response indicates that the contribution of the IFN-α/β system in limiting virus replication during the acute phase of CSF is incomplete. This also suggests that CSFV possesses an inherent resistance to the antiviral effect of IFN-α/β, which is independent of the function of Npro described in the present work. Our observations relate to the data of Schweizer and coworkers showing resistance of the pestivirus BVDV to IFN-α/β in vitro once the infection is established, while the virus does not apparently target the IFN-α/β signaling pathway (43).
The disease exacerbation observed with the vEy-37 Npro mutant viruses, which induced particularly high levels of IFN-α/β, is interesting with respect to the previously proposed role of aberrant IFN-α/β responses in CSF immunopathogenesis (48). This also supports the idea that IFN-α/β responses contribute more to disease development than to protection with virulent strains of CSFV. However, additional factors such as proinflammatory cytokines may well contribute to CSF pathogenesis (31), because the highest levels of IFN-α did not necessarily correlate with the most severe clinical scores or death.
Taken together, the results of the present study show that specific abrogation of a viral function that efficiently prevents the induction of IFN-α/β in non-plasmacytoid dendritic cells does not necessarily result in attenuation. These data shed new light not only on the role of the IFN-α/β system during the early phase of CSFV infection in vivo but also on the role of Npro in the biology of pestiviruses.
This work was supported in part by Swiss National Science Foundation grants 3100A0-102066 and 3100A0-116608 and in part by Swiss Federal Veterinary Office grant 1.03.04. C.G.L. was supported by a Students Training in Advanced Research summer fellowship, sponsored by NIH/NCRR, to the School of Veterinary Medicine, University of California, Davis, CA.
We thank C. Griot for continuous support and Kenneth C. McCullough for critical reading of the manuscript. We also thank Alexandra Meindl, Sandra Blome, and Irene Greiser-Wilke (Community Reference Laboratory for CSF, School of Veterinary Medicine, Hannover, Germany) for the GPE− strain, Martin D. Fray (Institute of Animal Health, Compton, Newbury, Berkshire, United Kingdom) for the Mx/CAT reporter gene assay, and Takashi Fujita (Kyoto University, Japan) for the reporter plasmid p125Luc.
Published ahead of print on 5 November 2008.