Targeting of the Influenza A Virus Polymerase PB1-PB2 Interface Indicates Strain-Specific Assembly Differences

Abstract

Assembly of the heterotrimeric influenza virus polymerase complex from the individual subunits PB1, PA, and PB2 is a prerequisite for viral replication. The conserved protein-protein interaction sites have been suggested as potential drug targets. To characterize the PB1-PB2 interface, we fused the PB1-binding domain of PB2 to green fluorescent protein (PB2_1-37-GFP) and determined its competitive inhibitory effect on the polymerase activity of influenza A virus strains. Coexpression of PB2_1-37-GFP in a polymerase reconstitution system led to substantial inhibition of the polymerase of A/WSN/33 (H1N1). Surprisingly, polymerases of other strains, including A/SC35M (H7N7), A/Puerto Rico/8/34 (H1N1), A/Hamburg/4/2009 (H1N1), and A/Thailand/1(KAN-1)/2004 (H5N1), showed various degrees of resistance. Individual exchange of polymerase subunits and the nucleoprotein between the sensitive WSN polymerase and the insensitive SC35M polymerase mapped the resistance to both PB1 and PA of SC35M polymerase. While PB2_1-37-GFP bound equally well to the PB1 subunits of both virus strains, PB1-PA dimers of SC35M polymerase showed impaired binding compared to PB1-PA dimers of WSN polymerase. The use of PA_SC35M/WSN chimeras revealed that the reduced affinity of the SC35M PB1-PA dimer was mediated by the N-terminal 277 amino acids of PA. Based on these observations, we speculate that the PB1-PA dimer formation of resistant polymerases shields the PB2_1-37 binding site, whereas sensitive polymerases allow this interaction, suggesting different assembly strategies of the trimeric polymerase complex between different influenza A virus strains.

INTRODUCTION

Despite existing vaccines and antiviral drugs, influenza annually claims 250,000 to 500,000 lives worldwide (http://www.ncbi.nlm.nih.gov/genomes/FLU/FLU.html). Since the current vaccines against influenza viruses offer only incomplete protection, antiviral drugs are greatly needed. Whereas the M2 ion channel inhibitors amantadine and rimantadine have been rendered obsolete by the unprecedented spread of resistant virus strains (45), the neuraminidase inhibitors oseltamivir and zanamivir represent the gold standard for both prophylaxis and treatment of influenza (26). However, the recent emergence of oseltamivir-resistant viruses (8, 25) and the constant threat posed by newly emerging influenza virus strains from animal reservoirs call for new strategies to inhibit influenza viruses. Many antivirals are active site inhibitors, but the emergence of resistant viruses is a serious drawback, as exemplified by the human immunodeficiency virus reverse transcriptase or protease inhibitors (4) and the influenza virus neuraminidase inhibitors (8, 25). Therefore, protein-protein interaction inhibitors could expand the limited target spectrum of antiviral drugs.

The influenza virus polymerase, consisting of the three subunits PB1, PA, and PB2, is a major virulence factor (15, 38). Whereas PB1 harbors the polymerase active site, thus sharing sequence homology with both cellular polymerases and the key motifs of other RNA-dependent RNA polymerases (27, 34), PB2 contains a cap-binding domain (2, 13, 18) and PA contains an endonuclease domain (9, 44). These features of individual polymerase subunits and the assembly to a trimeric polymerase complex are required for viral genome transcription and replication (7), which take place in the host cell nucleus. Complex formation between PB1 and PA has been described previously (30, 32, 33), as well as that between PB1 and PB2 (1, 10, 40). However, only a weak transient interaction has been proposed for PB2 and PA (21). As suggested recently (22), PB1 and PA translocate to the nucleus upon dimer formation and subsequently bind to PB2. This is compatible with the observations that PB1 accumulates efficiently in the host cell nucleus only in the presence of PA (14) and that PB1-PA dimers can associate with PB2 in vitro (6). PB1 and PA interact via the N-terminal 12 amino acids of PB1 and the C-terminal domain of PA (30, 32, 33). This interaction domain has already been established as a promising antiviral target structure (16, 20, 30, 43).

Controversial reports exist about the domains of PB1 and PB2 required for complex formation (17, 30, 35, 36, 40). However, a crystal structure of a PB1_C (residues 678 to 757)-PB2_N (residues 1 to 37) complex was determined recently (40). In this case, the PB1-binding domain of PB2 (PB2_1-37) consists of three α-helices, of which only α₁, at the extreme N terminus of PB2, mediates contact with PB1, indicating a small binding site with a flat and extended interface which might be suitable for peptide or peptidomimetic inhibitors. Mutations at the interface inhibit RNA synthesis (40), suggesting that compounds that can dissociate the PB1-PB2 complex are potential influenza A virus inhibitors. Furthermore, high sequence conservation of the contact amino acids among influenza A viruses (40) indicates a broad-spectrum antiviral target. However, it remains to be addressed experimentally whether the PB1-PB2 interaction site can be targeted by assembly inhibitors. We recently showed that assembly inhibition by a peptide derived from the PA interaction domain of PB1 (PB1_1-25) can interfere with viral replication of several influenza A viruses (16, 43). Furthermore, peptide screenings revealed amino acid substitutions which enhanced the affinity of PB1 for the PA protein and led to a more efficient inhibition of viral growth (42, 43). To address whether the PB1-PB2 interaction site provides a similar potential for assembly inhibitors, we fused the PB1-binding domain of PB2 (PB2_1-37) to green fluorescent protein (PB2_1-37-GFP) to facilitate stability and detection and tested this fusion construct for inhibition of influenza A virus polymerase activity in polymerase reconstitution assays.

Here we provide evidence that PB2_1-37-GFP inhibits the activity of the polymerase of A/WSN/33, whereas polymerases of other influenza A virus strains, including A/SC35M (SC35M), remain unaffected. Biochemical studies revealed that PB2_1-37-GFP could bind efficiently to the PB1 subunits of both viral strains, whereas the affinity of the PB1-PA dimer of SC35M polymerase for PB2_1-37-GFP was impaired compared to that of the PB1-PA dimer of WSN polymerase. The use of PA_WSN/SC35M chimeras mapped the resistance to the N-terminal domain of PA of SC35M. Based on these data, we speculate that PB1-PA dimer formation of the resistant SC35M polymerase leads to a conformational change of the PB2-binding domain of PB1, thereby preventing the binding of PB2_1-37-GFP.

MATERIALS AND METHODS

Plasmid construction.

The pCAGGS PB2_1-37-GFP expression plasmid was obtained by PCR amplification of nucleotides (nt) 1 to 111 in the A/WSN/33 PB2 gene, using a forward primer containing an EcoRI restriction site and a reverse primer containing a NotI restriction site. This PCR fragment was digested with the indicated restriction enzymes and ligated into a previously described digested pCAGGS PB1_1-25-GFP vector (16), yielding pCAGGS PB2_1-37-GFP, with a 3-alanine linker between the PB2 and GFP coding sequences. pCAGGS PB2_1-23- GFP was obtained by annealing of two DNA oligonucleotides coding for the PB2_1-23 sequence flanked by sequences identical to digested EcoRI and NotI restriction sites and was ligated into the digested pCAGGS PB1_1-25-GFP vector. pCAGGS PB2_1-12-GFP was generated by the same strategy as that for pCAGGS PB2_1-23-GFP, using oligonucleotides coding for the PB2_1-12 sequence. pCAGGS PB2_1-37-GFP mutants were generated by site-directed mutagenesis, using forward primers containing an EcoRI restriction site and reverse primers containing a NotI restriction site.

To generate A/Thailand/1(KAN-1)/2004 (H5N1) (KAN-1) expression plasmids, the previously described KAN-1 pHW2000 vectors (39) were used for PCR amplification of the open reading frames (ORFs), and the PCR products were digested with NotI and XhoI (PB2, PB1, and NP) or NotI and NheI (PA) and subsequently cloned into the vector pCAGGS (B. Mänz, L. Brunotte, and M. Schwemmle, submitted for publication). The expression plasmids for A/WSN/33 (H1N1) (43), A/SC35M (H7N7) (15), A/Puerto Rico/8/34 (H1N1) (11), and A/Hamburg/4/2009 (H1N1) (HH4) (46) have been described elsewhere.

The plasmids pCAGGS PA_1-277WSN,_278-716SC35M, coding for the N-terminal 277 amino acids of PA_WSN and amino acids 278 to 716 of PA_SC35M, and pCAGGS PA_1-277SC35M,_278-716WSN, coding for the N-terminal 277 amino acids of PA_SC35M and amino acids 278 to 716 of PA_WSN, were generated by assembly PCR, using forward and reverse primers flanking the PA ORF and containing a NotI and an XhoI restriction site, respectively. Internal primers were designed as reverse complements harboring approximately 10 to 15 nt of the PA_SC35M and PA_WSN coding sequences at the fusion site.

Reconstitution of influenza virus polymerase activity.

293T cells were transiently transfected with a plasmid mixture containing influenza A or influenza B virus-derived PB1, PB2, PA, and NP expression plasmids, a polymerase I (Pol I)-driven plasmid transcribing an influenza A or influenza B virus-like RNA coding for the reporter protein firefly luciferase to monitor viral polymerase activity, and expression plasmids coding for the indicated GFP fusion proteins. Both minigenome RNAs were flanked by noncoding sequences of segment 8 of influenza A virus and influenza B virus, respectively. The transfection mixture also contained a plasmid constitutively expressing Renilla luciferase, which served to normalize variations in transfection efficiency. The reporter activity was determined at 24 h posttransfection and normalized using a dual-luciferase assay system (Promega). As a control, PX-GFP, which harbors the X-binding domain of the P protein of Borna disease virus (16), was used. The activity observed with transfection reaction mixtures containing PX-GFP was set to 100%. Using this approach, the relative polymerase activities of A/WSN/33 (H1N1), B/Yamagata/73, A/SC35M (H7N7), A/Puerto Rico/8/34 (H1N1), A/Hamburg/4/2009 (H1N1), and A/Thailand/1 (KAN1)/2004 (H5N1) were determined. To compare polymerase activities between the different strains, the data are displayed as fold induction, obtained by dividing the relative firefly luciferase activity by the negative-control activity.

Peptides.

PB2_1-37-biotin was obtained from Pepscan Presto, The Netherlands. PB1_{1-25 scrambled}-biotin was used as a control peptide and was described previously (43).

Immunoprecipitation experiments.

293T cells were transfected with the indicated plasmids in 6-well plates by use of Lipofectamine (Invitrogen). Cells were incubated at 24 h posttransfection with lysis buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 0.5 mM EDTA, 0.5% NP-40, 1% protease inhibitor mix G [Serva, Heidelberg, Germany], 1 mM dithiothreitol [DTT]) for 15 min on ice. After centrifugation at 13,000 rpm at 4°C, the supernatant was incubated with hemagglutinin (HA)- or Flag-specific antibodies coupled to agarose beads (Sigma) for 1 h at 4°C. After three washes with 1 ml of washing buffer (lysis buffer without protease inhibitor mix), bound material was eluted under denaturing conditions, separated in SDS-PAGE gels, and transferred to polyvinylidene difluoride (PVDF) membranes. Where indicated, mixing of cell extract prior to immunoprecipitation was performed on ice for 1 h. Viral polymerase subunits, GFP fusion proteins, and actin (cell extract only) were detected with antibodies directed against the HA (Covance, Berkeley, CA), Flag (Sigma-Aldrich), or GFP (Santa Cruz Biotechnology) tag, against PB1 (kindly provided by Paul Digard), or against actin (Sigma-Aldrich).

Enzyme-linked immunosorbent assay (ELISA) to determine binding of PB1 or PB1-PA dimers to PB2_1-37 in vitro.

Streptavidin-coated microwell plates were incubated with saturating concentrations of biotinylated PB2_1-37 peptide, washed, and subsequently incubated at room temperature with HA-tagged PB1. To obtain PB1-HA or PB1–PA-HA dimers, 293T cells were seeded into 94-mm dishes, transfected with the respective plasmids, and treated with lysis buffer at 24 h posttransfection as previously described (43). After washing of the microwell plates, the wells were incubated with an HA-specific primary antibody (MMS-101R; Covance), followed by three washes and incubation with a peroxidase-coupled secondary antibody (Jackson ImmunoResearch, Newmarket, United Kingdom). After the final wash step, ABTS substrate (ready-to-use solution; Sigma) was added, and the optical density was determined at 405 nm.

RESULTS

PB2_1-37-GFP binds specifically to PB1 and inhibits the polymerase of WSN.

To address the question of whether the PB1-PB2 interaction site can be verified as an antiviral target experimentally, we coexpressed the highly conserved PB1-binding domain of PB2 derived from an influenza A virus PB2 consensus sequence (Fig. 1 A) as a GFP fusion construct (PB2_1-37-GFP) in a WSN polymerase activity reconstitution assay. As a control, the activity of PX-GFP, which harbors the X-binding domain of the P protein of Borna disease virus (16), was set to 100%. As shown in Fig. 1B and C, the WSN polymerase could be inhibited by PB2_1-37-GFP, to about 20% of the control level, in a concentration-dependent manner. The degree of inhibition was similar to the previously described inhibition by PB1_1-25-GFP (Fig. 1B) (16, 43). Since the PB1-binding domains of PB2 between influenza A and B viruses show a virus-type-specific conservation and little sequence identity in the first 37 amino acids (Fig. 1A), we speculated that PB2_1-37-GFP might fail to inhibit the polymerase activity of an influenza B virus strain. Indeed, the polymerase activity of B/Yamagata/73 was not affected by PB2_1-37-GFP (Fig. 1D). Consistent with this observation, PB2_1-37-GFP failed to bind to PB1 of B/Yamagata/73 (Fig. 1E). Next, we wanted to map the minimal inhibitory domain of PB2_1-37. Since PB2_1-37 folds into 3 discrete α-helices (40), expression plasmids were generated to code for truncated versions of PB2_1-37-GFP, designated PB2_1-23-GFP and PB2_1-12-GFP, containing either α₁ and α₂ or only the α₁ helix. As shown in Fig. 1F, both truncated versions of PB2_1-37-GFP maintained their inhibitory effect on the WSN polymerase, albeit to a slightly lesser extent. This degree of inhibition correlated with binding of the GFP fusion constructs to PB1-HA in a coimmunoprecipitation assay (Fig. 1G), where both PB2_1-23-GFP and PB2_1-12-GFP maintained their binding activity toward PB1. To further characterize the inhibition by PB2_1-37-GFP and to identify amino acid positions that are either crucial or dispensable for its inhibitory effect, we tested PB2_1-37-GFP mutants harboring single alanine substitutions of amino acids 2 to 12 within the PB1-binding domain. All mutants were equally well expressed (Fig. 2 A). However, only substitution of leucine at position 7 and methionine at position 11 reduced the inhibitory effect on the polymerase (Fig. 2B), correlating with reduced binding in a coimmunoprecipitation assay (Fig. 2C and D). In contrast, mutation of arginine to alanine at position 8 rendered PB2_1-37-GFP more active (Fig. 2B). To address the relevance of these mutations in the context of the full-length PB2 protein, the L7A (PB2_L7A), R8A (PB2_R8A), and M11A (PB2_M11A) mutations were introduced into PB2_WSN and tested in the polymerase reconstitution assay. As shown in Fig. 2E, these mutant proteins supported polymerase activity.

Fig. 1.

PB2_1-37-GFP specifically inhibits A/WSN/33 but not B/Yamagata/73 polymerase. (A) Amino acid sequence alignment of influenza A (FluAV) and influenza B (FluBV) virus consensus sequences of PB2 residues 1 to 37. Identical amino acid residues are shaded in gray. The consensus sequences of FluAV and FluBV were derived from the available full-length PB2 sequences of human isolates provided in the NCBI influenza database. (B) Activity of A/WSN/33 polymerase in the presence of PB2_1-37- or PB1_1-25-GFP. 293T cells were transiently transfected with a plasmid mixture containing PB1, PB2, PA, and NP expression plasmids, a Pol I expression plasmid expressing an influenza virus-like RNA coding for the reporter protein firefly luciferase to monitor viral polymerase activity, and an expression plasmid (1 μg) coding for the indicated GFP fusion proteins. The transfection mixture also contained a plasmid constitutively expressing Renilla luciferase, which served to normalize variations in transfection efficiency. The activity observed with transfection reaction mixtures containing PX-GFP was set to 100%. The omission of PB1 in the transfection mixture served as a negative control. (C) Activity of A/WSN/33 polymerase in the presence of the indicated GFP fusion proteins. The experiments were carried out as described for panel B, using increasing amounts of transfected GFP expression plasmids relative to the PX-GFP control, which was set to 100% for 50 ng PX-GFP expression plasmid. (D) Activity of B/Yamagata/73 polymerase in the presence of PB2_1-37. Experiments were carried out as described for panel B, using expression plasmids for NP, PB1, PB2, and PA derived from B/Yamagata/73 and a Pol I expression plasmid expressing an influenza virus-like RNA with flanking viral sequences of influenza B virus. (E) Binding of PB2_1-37-GFP to HA-tagged PB1 (PB1-HA) of B/Yamagata/73. The indicated proteins were expressed in HEK293T cells, and binding of the GFP fusion proteins to PB1-HA was analyzed by coimmunoprecipitation (IP) using anti-HA-specific agarose beads and subsequent immunoblotting. Precipitated material was analyzed for the presence of either PB1-HA or PB2_1-37-GFP by using the indicated antibodies against the HA tag (αHA) or GFP (αGFP). (F) Polymerase inhibitory activity of PB2_1-37-GFP, PB2_1-23-GFP, and PB2_1-12-GFP fusion proteins in an A/WSN/33 polymerase reconstitution assay. (G) Complex formation of the indicated GFP fusion proteins and HA-tagged PB1 of A/WSN/33. Binding was analyzed as described for panel E. Note that the differences in migration of the indicated GFP fusion constructs do not correspond to their lengths but are probably due to differences in folding. Error bars represent standard deviations for at least three independent experiments.

Fig. 2.

Alanine scanning mutagenesis of amino acid residues 2 to 12 in PB2_1-37-GFP reveals critical amino acid positions. (A) Analysis of expression levels of PB2_1-37-GFP-derived alanine mutants in HEK293T cells by Western blot analysis. (B) Polymerase activity of WSN in the presence of the indicated PB2_1-37-derived GFP fusion proteins. The activity observed with transfection reaction mixtures containing PX-GFP was set to 100%. (C) Complex formation of the indicated PB2_1-37-GFP-derived alanine mutants with PB1 of WSN transiently expressed in HEK293T cells was analyzed by coimmunoprecipitation (IP) as described in the legend to Fig. 1E. (D) Quantification of the GFP signals observed in panel C. (E) Activity of A/WSN/33 polymerase in the presence of the indicated PB2 mutants. The top panels show expression levels of the indicated HA-tagged PB2 mutants in HEK293T cells by Western blot analysis. Error bars represent standard deviations for at least three independent experiments.

A panel of influenza A virus polymerases show various degrees of sensitivity to PB2_1-37-GFP.

The high amino acid sequence conservation among influenza A viruses at the PB1-PB2 binding site strongly suggested that all influenza A virus polymerases should be sensitive to inhibition by PB2_1-37-GFP. To verify this hypothesis, we tested a panel of influenza A virus polymerases from various strains, including A/SC35M (H7N7) (SC35M), A/PR/8/34 (H1N1) (PR8), A/Hamburg/4/2009 (H1N1) (HH4), and A/Thailand/1(KAN-1)/2004 (H5N1) (KAN-1), for inhibition by PB2_1-37-GFP. To our surprise, only the WSN polymerase activity was reduced, to about 20% of the PX-GFP control level (Fig. 3 A). In contrast, the polymerases of SC35M and HH4 even displayed slight increases in activity in the presence of PB2_1-37-GFP (Fig. 3B and E), whereas the polymerases of KAN-1 (Fig. 3C) and PR8 (Fig. 3D) showed an intermediate phenotype of inhibition, to approximately 50% and 70%, respectively. In order to rule out the possibility that sensitivity or resistance to PB2_1-37-GFP is caused solely by differences in the activities of different influenza A virus polymerases, the absolute reporter values in the absence of PB2_1-37- GFP were compared (Fig. 3F). This revealed the lowest polymerase activity for the sensitive WSN polymerase. However, one of the most active polymerases, namely, the KAN-1 polymerase (Fig. 3F), was partially inhibited by PB2_1-37-GFP (Fig. 3C), suggesting that sensitivity to PB2_1-37-GFP is independent of polymerase activity.

Fig. 3.

Inhibitory effect of PB2_1-37-GFP on influenza A virus polymerases derived from different strains. The polymerase inhibitory activity of a PB2_1-37-derived GFP fusion protein was determined in A/WSN/33 (WSN) (A), A/SC35M (SC35M) (B), A/Thailand/1(KAN-1)/2004 (KAN-1) (C), A/PR/8/34 (PR8) (D), and A/Hamburg/4/09 (HH4) (E) polymerase reconstitution assays. The activity observed with transfection reaction mixtures containing PX-GFP was set to 100% in panels A to E. Student's t test was performed to assess whether the values obtained with PB2_1-37-GFP differed from the control values obtained with PX-GFP. *, P < 0.01. (F) Comparison of absolute polymerase activities of A/WSN/33, SC35M, KAN-1, PR8, and HH4 in the presence of 1 μg PX-GFP expression plasmid. Fold induction was calculated by dividing the firefly luciferase reporter signal of the positive control by the signal of the negative control, where PB1 was omitted from the transfection mixture. Error bars represent standard deviations for at least three independent experiments.

Resistance of SC35M polymerase is independent of polymerase activity and is not affected by a WSN-specific amino acid residue within PB2_1-37-GFP.

To further investigate why only the WSN polymerase could be inhibited substantially by PB2_1-37-GFP, despite high sequence conservation of the target structure, we proceeded to compare this sensitive polymerase to the highly resistant polymerase of SC35M. We first wanted to rule out that the inhibitory effect of PB2_1-37-GFP was caused by differences in polymerase activity. We therefore compared the activities of the WSN and SC35M polymerases in the presence of the control (PX-GFP) or PB2_1-37-GFP at different time points posttransfection (Fig. 4 A and B). If the high polymerase activity of SC35M would simply overrun the inhibitory effect of PB2_1-37-GFP, we would expect to see inhibition of SC35M polymerase at early time points posttransfection. However, in contrast to the polymerase of WSN (Fig. 4A), the SC35M polymerase was not inhibited by PB2_1-37-GFP at any of the measured time points (Fig. 4B), indicating that the strain-dependent difference in polymerase activity was most likely not responsible for the resistance to PB2_1-37-GFP.

Fig. 4.

Resistance of SC35M to inhibition by PB2_1-37-GFP is independent of polymerase activity and is not affected by a strain-specific amino acid residue within PB2_1-37-GFP. (A and B) Activities of WSN and SC35M polymerases at the indicated time points posttransfection for expression constructs (1 μg) coding for PB2_1-37-GFP or PX-GFP. The polymerase activity in the presence of PX-GFP at 24 h posttransfection was set to 100%. (C) Alignment of the PB2_1-37 sequences of the indicated strains. (D to H) Polymerase activity was determined at 24 h posttransfection for either the WSN or SC35M polymerase in the presence of PB2_1-37N9D-GFP (D and E), the PB2-binding domain of PB1 fused to GFP (PB1_678-757-GFP) (F and G), or PB2_1-37R8A-GFP (H). The polymerase activities observed in the presence of PX-GFP were set to 100%. Error bars represent standard deviations for at least three independent experiments.

A striking difference between WSN and SC35M is the amino acid substitution N to D at position 9 of PB2 (Fig. 4C), which might cause differences in susceptibility to PB2_1-37-GFP. We therefore introduced this mutation into PB2_1-37-GFP and tested for inhibition of polymerase activity in the two influenza A virus strains. However, whereas the polymerase of WSN was inhibited by PB2_1-37N9D-GFP (Fig. 4D), the SC35M polymerase remained resistant (Fig. 4E), excluding a role of this amino acid substitution in the strain-specific inhibition by PB2_1-37-GFP.

Next, we tested whether the PB2-binding domain PB1_678-757, derived from PB1 of A/WSN/33 fused to GFP (PB1_678-757-GFP), might be active against all influenza A virus polymerases. As shown in Fig. 4F, the polymerase of WSN was substantially inhibited in the presence of PB1_678-757-GFP. In contrast, the activity of the SC35M polymerase increased to ca. 150% in the presence of PB1_678-757-GFP (Fig. 4G), consistent with an increase in activity of SC35M polymerase in the presence of PB2_1-37-GFP (Fig. 3B). The R8A mutation in PB2_1-37-GFP (PB2_1-37R8A-GFP), which resulted in increased inhibition of the WSN polymerase activity (Fig. 2B), did not potentiate the activity against the resistant SC35M polymerase (Fig. 4H). In conclusion, the resistance of SC35M to PB2_1-37-GFP did not depend on polymerase activity, a strain-specific amino acid substitution in PB2, or the interaction domain (PB1- or PB2-binding domain) used for inhibition.

Resistance to PB2_1-37-GFP is mediated by either PB1_SC35M or PA_SC35M.

In order to identify the viral proteins determining the resistance to PB2_1-37-GFP, we exchanged all viral ribonucleoprotein (vRNP) components (NP, PB1, PB2, and PA) of the sensitive WSN polymerase with their SC35M counterparts and determined the polymerase activity in the presence of PB2_1-37-GFP. With the exception of PB2_SC35M, which reduced the polymerase activity to background levels, reconstitution of the WSN polymerase with NP_SC35M or PA_SC35M resulted in similar polymerase activities to that observed with the WSN polymerase, whereas an increased activity was determined with PB1_SC35M (Fig. 5 A). Although PB2_SC35M rendered the WSN polymerase inactive, the trimeric polymerase complex was formed with PA_WSN and PB1_WSN (Fig. 5B). However, PB2_WSN supported the polymerase activity of SC35M (Fig. 5C). The incompatibility of WSN and SC35M PB2 might be caused either by impaired polymerase assembly or by the presence of PB2 627E in SC35M, which drastically reduces polymerase activity when introduced into PB2_WSN (23). Since we could rule out the assembly defect and PB2_WSN was functional in the SC35M minireplicon, we assumed that the avian signature glutamic acid 627 in PB2_SC35M caused the observed nonfunctionality. Consequently, we could not assess the role of PB2_SC35M in the resistance phenotype. However, exchange of the remaining RNP components yielded insights into the resistance phenotype of SC35M. Importantly, while NP_SC35M did not alter the sensitive phenotype of the WSN polymerase in the presence of PB2_1-37-GFP (Fig. 5D), exchange of the polymerase subunits PB1 (Fig. 5E) and PA (Fig. 5F) with their SC35M counterparts turned the polymerase resistant to inhibition by PB2_1-37-GFP. Intriguingly, PA_SC35M conferred resistance of the WSN polymerase to PB2_1-37-GFP, although no direct interaction with PA was observed (Fig. 6 H).

Fig. 5.

PB1 and PA of SC35M render the WSN polymerase resistant to inhibition by PB2_1-37-GFP. (A) Absolute polymerase activities of reconstituted WSN polymerases after substitution of the indicated SC35M vRNP components. (B) Formation of the trimeric polymerase complex between WSN PA-PB1 and PB2_SC35M. The indicated proteins were expressed in HEK293T cells, and binding of PB2_SC35M to PA-PB1 was analyzed by coimmunoprecipitation (IP) using Flag-specific agarose beads and subsequent immunoblotting. Precipitated material was analyzed for the presence of either PA-Flag or PB2-HA, using the indicated antibodies against the Flag tag (αFlag) or HA tag (αHA). (C) Polymerase activity of SC35M or SC35M reconstituted with the PB2 subunit of WSN [SC35M (PB2_WSN)]. Student's t test was performed to assess whether the values with PB2_1-37-GFP differed from the control values obtained with PX-GFP. *, P < 0.01. (D to F) Absolute polymerase activities of SC35M vRNP components harboring WSN PB2. The polymerase activities observed in the presence of PX-GFP were set to 100%. Error bars represent standard deviations for at least three independent experiments.

Fig. 6.

Resistance to PB2_1-37-GFP correlates with reduced binding to PB1-PA dimers. (A to C) Binding of HA-tagged PB1 or PB1 coexpressed with HA-tagged PA of WSN or SC35M in HEK293T cell extracts to an immobilized PB2_1-37 peptide was determined by ELISA. Binding of PB1 or PB1-PA-HA dimers to a control peptide was subtracted from the signal obtained with the PB2_1-37 peptide and expressed as relative binding. Error bars represent standard deviations for at least three independent experiments. (D and E) Western blots of the PB1-HA- and PB1-PA-containing cell extracts used for panels A to C. (F) Complex formation of HA-tagged PA and PB1 derived from either WSN or SC35M. The indicated proteins were expressed in HEK293T cells, and binding of the GFP fusion proteins was analyzed by coimmunoprecipitation (IP) as described in the legend to Fig. 1E. (G and H) Complex formation of PB2_1-37-GFP with PB1 or PB1-PA dimers of WSN (G) or SC35M (H) was analyzed by coimmunoprecipitation using anti-HA agarose, with subsequent immunoblotting. Note that complex formation of PX-GFP with either PB1 or PB1-PA dimers was not observed. (I) Complex formation of PB2_1-37-GFP with PB1 or PB1-PA dimers of WSN or SC35M by coimmunoprecipitation (IP) as described for panels G and H, after incubation of cell extracts harboring either PB2_1-37-GFP, PB1, or PB1-PA dimers. (J) Quantification of the GFP signals observed in panel I. Error bars represent standard deviations for three independent experiments.

Resistance to PB2_1-37-GFP correlates with impaired PB1-PA dimer binding.

We speculated that the actual target structure of PB2_1-37-GFP should be the PB1-PA dimer instead of the PB1 monomer, since it was shown that preformed PB1-PA dimers associate preferentially with PB2 (6). Furthermore, the observation that PA_SC35M confers resistance to PB2_1-37-GFP (Fig. 5F) suggests that the PA-PB1 dimer, not PB1 alone, is decisive for sensitivity or resistance. We therefore hypothesized that resistance to PB2_1-37-GFP is mediated by a failure to bind to the PB1_SC35M–PA_SC35M dimer. In order to study the interaction between PB2_1-37 and either PB1 or the PB1-PA dimer in a minimal setting, we established an ELISA-based binding assay. In this assay, biotinylated PB2_1-37 was bound to neutravidin-coated 96-well plates and probed for binding of HA-tagged PB1 or PB1–PA-HA expressed in HEK293T cells. The amount of bound PB1-HA or PB1–PA-HA was determined using antibodies specifically recognizing the HA tag. As shown in Fig. 6A, the HA-tagged PB1 subunits of both WSN and SC35M, designated PB1_WSN-HA and PB1_SC35M-HA, respectively, bound to the immobilized PB2_1-37 peptides in a dose-dependent manner. PA-HA, which was used as a specificity control, did not bind to PB2_1-37 (data not shown). We further determined binding of either PB1_WSN–PA_WSN-HA or PB1_SC35M–PA_SC35M-HA dimers coexpressed in HEK293T cells to PB2_1-37. Since PA-HA alone failed to bind to PB2_1-37, signals obtained in the ELISA by use of HA-specific antibodies reflected specific binding of PB1–PA-HA dimers but not PB1 monomers. As shown in Fig. 6B, PB1_WSN–PA_WSN-HA dimers associated efficiently with PB2_1-37, in a dose-dependent manner. Furthermore, binding of these dimers to PB2_1-37 was slightly more efficient than that of the PB1_WSN-HA monomer (Fig. 6A) at comparable expression levels (Fig. 6D). Thus, the PB1-PA dimer seems to be the bona fide target structure for PB2_1-37-GFP. In contrast, PB1_SC35M–PA_SC35M-HA failed to bind efficiently to PB2_1-37 (Fig. 6B). Since both PB1_SC35M and PA_SC35M mediated resistance to PB2_1-37-GFP in the WSN ribonucleoprotein reconstitution assay, we expected that PB1_WSN–PA_SC35M-HA and PB1_SC35M–PA_WSN-HA dimers would also fail to bind to PB2_1-37 because they led to resistance of the polymerase in the subunit swap experiment (Fig. 5E and F). Indeed, these mixed dimers showed reduced binding to PB2_1-37 (Fig. 6C). Since all PB1-PA combinations showed similar expression levels as analyzed by Western blotting (Fig. 6E), and since PB1_WSN–PA_SC35M-HA and PB1_SC35M–PA_WSN-HA associated efficiently into dimers as judged by coimmunoprecipitation studies (Fig. 6F), we concluded that the resistance phenotype of either the SC35M polymerase or a WSN polymerase harboring PB1_SC35M or PA_SC35M was caused by impaired binding to PB2_1-37-GFP. To confirm this finding in a cellular system, we tested for PB2_1-37-GFP binding to the indicated PB1-PA-HA dimers in coimmunoprecipitation experiments (Fig. 6G and H). These studies confirmed that PB1_WSN-HA as well as PB1_WSN–PA_WSN-HA efficiently forms complexes with PB2_1-37-GFP (Fig. 6G). However, whereas PB1_SC35M-HA bound to PB2_1-37-GFP, we observed a reduced binding of PB2_1-37-GFP by the PB1_SC35M–PA_SC35M-HA dimer (Fig. 6H). We repeatedly observed decreased and various levels of PB2_1-37-GFP in the cell extract used for immunoprecipitation. To accurately quantify the binding efficiencies, we expressed PB1-HA, PB1-PA-HA, and PB2_1-37-GFP individually and used those cell extracts for a pulldown assay. Consistent with the results observed after simultaneous expression of all components (Fig. 6G and H), binding of PB2_1-37-GFP was less efficient with the PA-PB1 dimer of SC35M than with that of WSN (Fig. 6I and J). Based on these results, we hypothesized that PB1_SC35M–PA_SC35M dimer formation results in a conformational change of PB1_SC35M, thereby shielding the C-terminal PB1 domain from interaction with PB2_1-37-GFP. Interestingly, PA_SC35M was sufficient to confer resistance to the WSN polymerase (Fig. 5F). To address the question of which domain of PA_SC35M was responsible for this “closed” dimer conformation, defined here as not accessible for PB2_1-37-GFP, we designed PA_WSN/SC35M chimeras and tested them for inhibition by PB2_1-37-GFP. Crystallization (9, 20, 29, 44) and partial proteolytic digestion (19) of PA revealed an N-terminal domain (PA_1-277) harboring the endonuclease function, a linker region, and a C-terminal domain (PA_278-716) which mediates binding to the extreme N terminus of PB1. We therefore designed a PA chimera harboring the N-terminal domain of the WSN polymerase and the C-terminal domain of the SC35M polymerase, designated PA_1-277WSN, _278-716SC35M, or vice versa, designated PA_1-277SC35M, _278-716WSN (Fig. 7 A). These PA chimeras supported WSN polymerase activity, although to different extents, as depicted in Fig. 7B. However, PA_1-277SC35M,_278-716WSN was still able to confer resistance in the WSN polymerase reconstitution assay (Fig. 7C), albeit to a slightly decreased extent compared to PA_SC35M (Fig. 5F). In contrast, PA_1-277WSN,_278-716SC35M was sensitive to inhibition by PB2_1-37-GFP (Fig. 7D). Consistent with the resistance phenotype mediated by PA_1-277SC35M,_278-716WSN, we observed reduced binding of PB2_1-37-GFP to PB1_WSN-PA_1-277SC35M, _278-716WSN dimers in coimmunoprecipitation experiments using extracts of cells either transiently transfected with the polymerase subunits and PB2_1-37-GFP (Fig. 7E) or after combining cell extracts harboring either PB2_1-37-GFP or PA-PB1 dimers (Fig. 7F and G). In summary, these results mapped the resistance to PB2_1-37-GFP to the N-terminal domain of the PA protein.

Fig. 7.

Resistance to PB2_1-37-GFP can be mapped to the N-terminal 277 amino acids of PA. (A) WSN- and SC35M-derived PA chimeras used for the experiments in panels B to D. (B) Absolute polymerase activities of reconstituted WSN polymerases, using the PA chimeras depicted in panel A. (C and D) Polymerase activities of reconstituted A/WSN/33 polymerase, using the PA chimeras depicted in panel A in the presence of PB2_1-37-GFP. The polymerase activities observed in the presence of PX-GFP were set to 100%. Error bars represent standard deviations for at least three independent experiments. (E) Complex formation of PB2_1-37-GFP with the indicated PB1-PA dimers. The experiments were carried out essentially as described in the legend to Fig. 5H. (F) Complex formation of PB2_1-37-GFP, as described for panel E, after incubation of cell extracts harboring PB2_1-37-GFP and the indicated PB1-PA dimers. (G) Quantification of the GFP signals observed in panel F. Error bars represent standard deviations for three independent experiments.

DISCUSSION

Since the resolution of the crystal structure of the PB1-PB2 interface raised the exciting opportunity of designing peptide or peptidomimetic assembly inhibitors targeting this interaction site, the objective of this study was to address the feasibility of such an approach experimentally. Several lines of evidence have previously suggested the PB1-PB2 interaction as a valuable protein-protein interaction target. Earlier studies used a more extended N-terminal domain of PB2 (PB2_1-269) comprising the PB1- and NP-binding domains, which competed with the formation of a PB1-PB2 and PB2-NP complex, thereby inhibiting viral gene expression (31, 35). However, since the PB1-PB2 interface has been mapped mainly by use of deletion mutants, resulting in contradictory results and an inexact interaction map, previous attempts to use short peptide inhibitors derived from the PB2-binding domain of PB1 (PB1_715-740) to interfere with this interaction failed (16). PB1_715-740 used in the earlier study did not contain the full three-helical bundle that mediates the interaction. Here we showed that both the PB2 interaction domain of PB1 fused to GFP (PB1_678-757-GFP), containing the complete three-helical bundle, and the PB1 interaction domain of PB2 (PB2_1-37-GFP) are able to inhibit the polymerase activity of WSN. Although in vitro studies suggested that a peptide containing solely the first α-helix of PB2 would not fold into its three-dimensional structure and would thereby fail to bind to PB1 (40), the PB2_1-12-GFP fusion construct retained its inhibitory activity and its capacity to bind to PB1, presumably by stabilization of folding by the fusion to GFP. Thus, considering only the inhibition of the WSN polymerase, this short, 12-amino-acid sequence would be a valuable starting point for the development of antiviral peptides or peptidomimetics, which in this case would be favored over small-molecule inhibitors due to the flat and extended fold of the targeted interaction site (40). However, despite high sequence conservation on the amino acid level, PB2_1-37-GFP failed to inhibit a panel of other influenza A virus polymerases, suggesting that it might be difficult to develop an efficient broad-spectrum inhibitor targeting the PB1-PB2 interaction site.

The resistance of most of the tested influenza A virus polymerases to inhibition by PB2_1-37-GFP could be explained by different hypotheses, two of which we discuss here in more detail. First, these results might indicate that the PB1-PB2 interaction via the PB1_678-757 and PB2_1-37 domains does not play a crucial role in polymerase assembly for all influenza A virus strains. Consequently, an alternative PB1-PB2 interaction site would be essential for assembly of these polymerases, and the high conservation of the PB1_678-757-PB2_1-37 interaction site would have resulted from a different function. Second, the PB1_678-757 and PB2_1-37 protein-binding domains mediate the assembly of PB2 with PB1-PA dimers, but some PB1-PA dimers shield the PB2_1-37 interaction site in the cytoplasm to prevent premature formation of a trimeric polymerase complex in the cytoplasm.

The first hypothesis is supported by the failure of PB1_SC35M-PA_SC35M dimers to bind to PB2_1-37. An alternative PB1-binding site in PB2, which has been mapped to the C-terminal part of PB2 (35), might be responsible for the interaction with PB1-PA dimers of the resistant influenza A virus strains. Interestingly, in the case of SC35M, we even observed an increase in polymerase activity in the presence of PB2_1-37-GFP, potentially indicating a regulatory effect of the PB1-PB2_1-37 interaction on polymerase activity. This is in line with the previously described PB1 mutations F669A and I750D, which resulted in decreased PB1-PB2 binding but increased enzymatic activity of the polymerase (40). The proposed regulatory function might be due to a PB2-NP interaction site overlapping with the PB1-PB2 interaction site at PB2_1-37. As described by others (35), binding of PB1 to the N-terminal PB1-binding site of PB2 abrogates the association with NP. Blocking the PB1-PB2 interaction by binding of PB2_1-37-GFP to PB1 might therefore result in an increase in PB2-NP binding, which might shift some polymerase complexes from the transcriptase to the replicase mode (37). The interdependency of the replication and transcription processes might explain the increase in polymerase activity observed for SC35M in the presence of PB2_1-37-GFP compared to the PX-GFP control. Since deletion of the N-terminal 12 or 37 amino acids of PB2 results in an enzymatically inactive polymerase for both WSN and SC35M (40; data not shown), this regulatory function is most likely crucial for polymerase activity of both sensitive and resistant viruses.

However, our results could also be explained by the second hypothesis, that PB1-PA dimers shield the PB2_1-37 interaction by conformational changes, thus preventing premature formation of a trimeric polymerase complex in the cytoplasm. Because comparison between the resistant SC35M polymerase and the sensitive WSN polymerase determined the N-terminal domain of the PA protein to be a major molecular determinant for resistance or sensitivity to inhibition by PB2_1-37-GFP, we speculate that formation of the PB1_SC35M-PA_SC35M dimer induces a conformational change which results in a “closed” conformation in which PB1_678-757 is rendered inaccessible for PB2_1-37, thereby leading to resistance. The occurrence of conformational changes is not unlikely, as the influenza virus polymerase complex has been described to be highly dynamic, adopting different conformations (5, 41). Although the interaction between the extreme N terminus of PB1 and the C-terminal domain of the PA protein has been established as a crucial high-affinity interaction, this interaction is most likely only the first contact. Upon initial binding, PB1 and PA presumably interact over a more extended interaction site that might even include the linker domain between the C-terminal and N-terminal domains of PA and render PA more stable than in the monomeric state, as shown recently by others (19). Furthermore, this extended interaction is underscored by a biochemical study in which the accessibility of surface lysines in five different regions of PB1 was changed upon association with PA (12). In this light, it remains surprising, however explainable, that the N-terminal domain of PA can induce a conformational change of PB1 that hinders the accommodation of PB2_1-37-GFP. Although we favor the model of a conformational change in PB1 exerted by PA, the obtained data do not distinguish this hypothesis from two alternative, though not mutually exclusive, possibilities. First, a direct interaction between PB2 and PA might be the decisive factor for resistance, since PA_1-256, which mediates resistance in our system, has also been shown to interact transiently with PB2 (21). Consequently, PA itself could block the interaction between PB1 and PB2. Second, the PB1-PA dimer might recruit host factors which block the association with PB2 in the cytoplasm.

Independent of the precise mode of resistance of some influenza A virus polymerases to inhibition by PB2_1-37-GFP, we speculate that these polymerases prevent association of PB1-PA dimers with PB2 by shielding the PB2-binding site on PB1 to avoid premature formation of the active trimeric polymerase complex in the cytoplasm. PB1-PA dimers that do not bind to PB2 are transported to the nucleus, where they might adopt an open conformation again. This open conformation might be chaperoned by the associated host factor Hsp90 (3, 24, 28), allowing efficient assembly with PB2 via the PB2_1-37 domain or alternative PB1-PB2-binding sites.