Oligomerization and GTP-binding requirements of MxA for viral target recognition and antiviral activity against influenza a virus

: The IFN-induced human myxovirus resistance protein A (MxA) exhibits a broad antiviral activity against many viruses, including influenza A virus (IAV). MxA belongs to the family of dynamin-like GTPases and assembles in vitro into dimers, tetramers, and oligomeric ring-like structures. The molecular mechanism of action remains to be elucidated. Furthermore, it is not clear whether MxA exerts its antiviral activity in a monomeric and/or multimeric form. Using a set of MxA mutants that form complexes with defined stoichiometry, we observed that, in the presence of guanosine 5’-O-(thiotriphosphate), purified MxA disassembled into tetramers and dimers. Dimeric forms did not further disassemble into monomers. Infection experiments revealed that besides wild-type MxA, dimeric and monomeric variants of MxA also efficiently restricted IAV at a replication step after primary transcription. Moreover, only dimeric MxA was able to form stable complexes with the nucleoprotein (NP) of IAV. MxA interacted with NP independently of other viral components. Interestingly, the dimeric form of MxA was able to efficiently bind to NP from several MxA-sensitive strains but interacted much more weakly with NP from the MxA-resistant PR8 strain derived from the H1N1 1918 lineage. Taken together, these data suggest that, during infection, a fraction of MxA disassembles into dimers that bind to NP synthesized following primary transcription in the cytoplasm, thereby preventing viral replication. The IFN-induced human myxovirus resistance protein A (MxA) exhibits a broad antiviral activity against many viruses, including influenza A virus (IAV). MxA belongs to the family of dynamin-like GTPases and assembles in vitro into dimers, tetramers, and oligomeric ring-like structures. The molecular mechanism of action remains to be elucidated. Furthermore, it is not clear whether MxA exerts its antiviral activity in a monomeric and/or multimeric form. Using a set of MxA mutants that form complexes with defined stoichiometry, we observed that, in the presence of guanosine 5 (cid:2) - O -(thiotriphosphate), purified MxA disassembled into tetramers and dimers. Dimeric forms did not further disassemble into monomers. Infection experiments revealed that besides wild-type MxA, dimeric and monomeric variants of MxA also efficiently restricted IAV at a replication step after primary transcription. Moreover, only dimeric MxA was able to form stable complexes with the nucleoprotein (NP) of IAV. MxA interacted with NP independently of other viral components. Interestingly, the dimeric form of MxA was able to efficiently bind to NP from several MxA-sensitive strains but interacted much more weakly with NP from the MxA-resist-ant PR8 strain derived from the H1N1 1918 lineage. Taken together, these data suggest that, during infection, a fraction of MxA disassembles into dimers that bind to NP synthesized following primary 13,000 (cid:3) g for 20 min at 4 °C. As described for blue native PAGE (32), lysates were dialyzed against 20 m M Tris-HCl (pH 6.8), 10% glycerol, 0.1% CHAPS, and 0.5 m M DTT at 4 °C for 4 h using Slide-A-Lyzer MINI dialysis devices with a 10,000 cutoff (Thermo Scientific). Dialyzed samples were centrifuged at 13,000 (cid:3) g and 4 °C for 20 min before loading on a precast 4–16% TGX TM gradient gel (Bio-Rad). Using 25 m M Tris-HCl (pH 8.3), 192 m M glycine, 0.5 m M DTT, and 0.1% CHAPS as running buffer, the gel was run at 4 °C with a constant current of 25 mA for 4 h. The gel was incubated in 25 m M Tris-HCl (pH 8.3), 192 m M glycine, and 0.1% SDS for 10 min before Western blotting. Blots were stained for MxA (rabbit anti-Mx1, Novus Biologicals, catalog no. H00004599_D01P).

The interferon system plays a pivotal role in the defense against viral infection. Interferons type I and III induce the expression of more than 300 proteins, many of them exhibiting intrinsic antiviral activity (1)(2)(3). The human myxovirus resistance protein A (MxA) 3 protein is induced exclusively by type I and type III interferons and restricts the replication of a large variety of RNA and DNA viruses (reviewed in Ref. 4). Mx proteins are members of the dynamin superfamily of large GTPases. They contain an amino-terminal globular GTPase (G) domain and a carboxy-terminal stalk domain (5,6). The stalk domain harbors the GTPase effector domain and additional structural elements, including the L4 loop required for antiviral activity and target specificity (7)(8)(9). The G and stalk domains are connected via the bundle signaling element (BSE), which strongly resembles the BSE responsible for intramolecular signaling in dynamin (10). Mx proteins exhibit a high intrinsic rate of GTP hydrolysis, although the affinity to GTP is weak (11). Moreover, the human MxA protein forms dimers and tetramers and has the capacity to form higher oligomeric ringlike structures through a series of intra-and intermolecular interactions (5,6,12,13). Introduction of mutations into the interfaces of the intermolecular interaction sites or hinge regions of MxA prevents assembly into higher-order multimeric forms. These mutations lead to the formation of MxA into tetramers, dimers, and monomers (5,12).
So far, little is known about the molecular mechanism of action of the human MxA protein. Previous studies have demonstrated that GTP binding is required for its efficient antiviral activity against vesicular stomatitis virus and IAV (6,14). In the case of IAV, several previous studies have demonstrated that ectopic expression of the nucleocapsid protein (NP) or the polymerase subunit 2 (PB2) of the virus partially abrogated the antiviral activity of mouse Mx1 (15)(16)(17). Moreover, human IAV strains are partially resistant to the activity of MxA, whereas non-adapted avian IAV strains such as H5N1 are highly sensitive to MxA. Recent findings indicate that resistance of human IAV strains to the activity of MxA is determined by NP (18). Adaptive mutations in a surface-exposed cluster of amino acids have been identified to counteract the restriction by MxA (19,20). So far, there is only little evidence for a physical interaction between NP of IAV and MxA (21). We have shown recently that MxA binds to UAP56, a DEAD-box helicase. UAP56 is also an interaction partner of NP and is required for efficient replication of IAV and evasion of the IFN response (22)(23)(24). Therefore, UAP56 may recruit MxA and NP to form a complex involving all three proteins.
To date, it is not clear whether the assembly of MxA into higher-ordered oligomeric structures is required to confer antiviral activity. Several previous studies have revealed that a monomeric mutant of MxA (L612K) still exhibited a pronounced antiviral activity against IAV and Thogotovirus and vesicular stomatitis virus (6,(25)(26)(27). On the basis of these findings, we previously proposed a model where newly synthesized MxA would assemble into higher-ordered oligomeric, preactivated structures and would disassemble into antivirally active monomers in response to a viral trigger in infected cells (6). On the other hand, on the basis of the recently solved crystal structure of MxA, Gao et al. (5,12) proposed a model where MxA would assemble into ring-like structures around incoming viral ribonucleoprotein (vRNP) complexes. These ring-like structures would then interfere with nuclear import and/or primary transcription. However, previous studies with stable MxA-expressing cell lines have clearly demonstrated that MxA inhibits an as yet undefined cytoplasmic step of the IAV life cycle that follows primary transcription (28).
To clarify this issue, we first evaluated in vitro the influence of GTP binding on the assembly state of MxA and assessed various interface and hinge mutants with defined monomeric or multimeric forms for their antiviral activity and capacity to interact with NP of IAV. The data indicate that, in the presence of GTP␥S, recombinant MxA disassembles from higher-ordered oligomeric structures into tetramers and dimers. Contrary to previous reports by Gao et al. (5,12), dimeric and monomeric forms of MxA also exerted a pronounced antiviral activity against avian IAV. Furthermore, only MxA in its dimeric form was able to form stable complexes with NP deriving from MxA-sensitive IAV strains independent of other viral components.

Experimental Procedures
Cells and Viruses-Madin-Darby canine kidney cells, African green monkey kidney (Vero) cells, HeLa cells, and HEK293T cells (all from the ATCC, LGC Standards) were cultured in growth medium consisting of Dulbecco's modified Eagle's medium supplemented with 10% FCS, 1ϫ Glutamax, and 1ϫ penicillin/streptomycin (Invitrogen). Cells transiently expressing human MxA (wild type or variants) were generated by transfecting the cells with pcDNA3.1(ϩ)neo constructs coding for MxA wild-type or mutant using JetPRIME (Polyplus Transfection) according to the protocol of the manufacturer.
GTPase Activity Analysis of Recombinant Human MxA-The GTPase activity of purified recombinant MxA was determined by incubating 0.01, 0.1, 1, 10, and 100 g/ml purified recombinant MxA sample with 10 M GTP and 0.1 M AMP-PNP at 37°C for 1 h. The sample was analyzed using the Transcreener GDP assay (Belbrooks Lab) according to the manual of the manufacturer. The GTPase activity was normalized to 100% GDP conversion by incubating the reaction mixture with 10 M GDP instead of GTP.
Size Exclusion Chromatography and Multi-angle Light Scattering (SEC-MALS)-Purified recombinant human MxA samples were dialyzed against 50 mM Hepes (pH 7.5), 200 mM KCl, 5 mM MgCl 2 , and 0.75 mM DTT overnight. Then 1 mg/ml purified recombinant human MxA samples were incubated with either 1 mM GTP␥S or buffer at 37°C for 1 h prior to analysis by SEC-MALS. To determine the exact molar mass of each recombinant human MxA construct, SEC-MALS measurements were performed as follows. 100 l of the sample was loaded on a 24-ml Superdex 100 column (GE Healthcare Biosciences) with a flow rate of 0.5 ml/min and 50 mM Hepes (pH 7.5), 200 mM KCl, 5 mM MgCl 2 , and 0.75 mM DTT as running buffer. The chromatograph was calculated according to the UV absorbance (280 nm). The SEC was followed by miniDAWN three-angle light-scattering detector (Wyatt Technology) and Optilab rEX differential refractometer (Wyatt Technology) analyses. The calculation of the molar mass was determined by the scattered light (Rayleigh ratio) measured by the MALS detector. The data were recorded and analyzed using the ASTRA V software (Wyatt Technology).
Split-GFP Complementation Assay-The cloning of the vectors pCDNA3.1-GFP(158 -238) and pCDNA3.1-GFP(1-157) has been described before (30). To clone the MxA wild type and variants in either of these vectors, the open reading frames of those constructs were amplified using PCR and subsequently introduced into the multiple cloning site located at the C terminus of the GFP fragment. HeLa cells were transfected with these constructs using Viafect (Promega) according to the protocol of the manufacturer. 24 h post-transfection, cells were incubated at 27°C for 2 h before fixation with 3% paraformaldehyde. After permeabilization with 0.5% Triton X-100, cells were stained with a mouse monoclonal MxA antibody (HB143 supernatant, in-house). Samples were mounted in DAPI containing mounting medium (Fluoromount, Southern Biotech) and consequently analyzed by confocal laser-scanning microscopy (CLSM Leica SP5, Leica Microsystems) using the LAS AF software (Leica).
RT qPCR-HEK293T cells transiently expressing MxA (wild type or variants) were grown in 6-well dishes and pretreated with 10 M cycloheximide or solvent for 1 h at 37°C before infection with either pH1N1-NP(H5N1) or PR8 at m.o.i. 5 for 6 h. Total RNA was purified as follows. Cells were lysed in 500 l of TRIzol reagent (Invitrogen). 100 l of chloroform was added to the lysates, mixed, and centrifuged at 13, 000 ϫ g for 15 min at 4°C. After centrifugation, 600 l of the aqueous RNAcontaining phase was mixed with 600 l of 70% EtOH. The resulting solution was transferred onto RNeasy spin columns (Qiagen), and further RNA extraction steps were performed according to the RNeasy Mini Kit protocol (Qiagen). cDNA was synthesized using SuperScript III reverse transcriptase (Life Technologies) according to the protocol of the manufacturer, using oligo(dT)15 (Promega) as a primer. PB2 and GAPDH transcripts were detected in an EvaGreen (Biotum)-based qPCR reaction (denaturation, 95°C and 15 s; annealing and extension, 60°C and 60 s). The primers for PB2 were as follows: forward, 5-GCAGAACCCAACAGAAGAGC; reverse, 5-GCT-GTTGCTCTTCTCCCAAC. The primers for GAPDH were as follows: forward, 5-CTGGCGTCTTCACCACCATGG; reverse, 5-CATCACGCCACAGTTTCCCGG. The -fold change (2 Ϫ⌬⌬Ct ) was calculated according to Pfaffl (34). Expression levels were analyzed using Western blotting. Blots were immunostained using the following antibodies: monoclonal mouse anti-MxA (35), monospecific rabbit anti-Mx1 (36), polyclonal rabbit anti-NP, 4 and polyclonal rabbit anti-GAPDH (Santa Cruz Biotechnology, catalog no. sc-25778).
vRNP Import Assay-Vero cells were transfected with either pcDNA3.1(ϩ)neo MxA or empty vector using Viafect (Promega). 24 h post-transfection, cells were pretreated with 10 M cycloheximide or solvent for 1 h at 37°C. Then cells were infected with FPV at m.o.i. 10 for 4 h in the presence of cycloheximide or solvent. Cells were fixed with 3% paraformaldehyde and permeabilized using 0.5% Triton X-100. Cells were stained using an affinity-purified polyclonal MxA antibody (derived from rabbit serum against recombinant His-MxA) and a monoclonal NP antibody (catalog no. HB65, ATCC). Samples were mounted in DAPI containing mounting medium (Fluoromount, Southern Biotech) and consequently analyzed by confocal laser-scanning microscopy (CLSM Leica SP5, Leica Microsystems) using LAS AF software (Leica). Images were assessed for nuclear import of vRNPs by counting MxA-expressing cells with a nuclear NP signal. Additionally, the NP mean nuclear signal intensity of MxA-expressing cells was analyzed using ImageJ software. Cells not expressing MxA were used as a negative control for both analyses.

Results
Oligomeric MxA Disassembles upon GTP Binding into Tetramers and Dimers-To investigate whether MxA has to first adopt a defined oligomeric or monomeric form to exert antiviral activity against IAV, we made use of several mutants of MxA that form complexes with a defined stoichiometry (Fig.  1A) (5,12). First, we assessed whether GTP binding has an influence on the oligomerization of recombinant wild-type MxA and MxA interface mutants (Fig. 1A) using SEC-MALS analysis. For this purpose, recombinant His-tagged MxA proteins were expressed in bacteria and subsequently affinity-purified using nickel-agarose columns. The purification was monitored by SDS-PAGE and Coomassie staining (Fig. 1B). The activity of the purified proteins was then tested by measuring their GTPase activity in dependence of their concentration (Fig. 1C). With the exception of the G domain mutant MxA(T103A), wild-type MxA as well as the MxA mutants forming complexes with defined stoichiometry exhibited GTPase activity in a very similar manner, as described in previous reports (5,12). Subsequently, the oligomeric state of the purified MxA protein complexes was assessed by SEC-MALS analysis in the presence or absence of the non-hydrolysable GTP analog GTP␥S (Fig. 1D). In the presence of 1 mM GTP␥S, wild-type MxA disassembled from higher-ordered oligomeric structures to tetramers and, to a small extent, also to dimers. The hinge 1 mutant MxA(R640A) shifted from a tetrameric to a dimeric state. By contrast, in the presence of GTP␥S, the stalk interface 1 and 2 mutants, dimeric MxA(L617D), and monomeric MxA(M527D) did not alter their stoichiometries (Fig.  1D). These data clearly demonstrate that GTP binding induced purified MxA to disassemble from higher oligomeric structures to tetramers and dimers. Similar data have been obtained previously for dynamin. Under low-salt conditions and in the presence of GTP␥S, preassembled dynamin is destabilized, forming lower oligomeric structures (37). As expected, the addition of GTP␥S did not lead to a stable dimerization of the monomeric MxA(M527D) via the G domain.
Oligomeric MxA Accumulates in the Perinuclear Region of Mammalian Cells-To assess whether the oligomerization of MxA occurs at defined subcellular sites, we carried out the split-GFP complementation assay as described previously using HeLa cells and concomitantly immunostained MxA with a specific antibody (30). As expected, pairwise co-transfection of constructs encoding split-GFP-MxA fusion proteins of wild-type MxA, MxA(R640A), and MxA(L617D) led to pronounced GFP signals indicative of the formation of dimers or higher-ordered oligomers, whereas co-transfection of the constructs encoding the split-GFP-MxA fusion proteins of the monomeric variant MxA(M527D) did not exhibit any GFP signal. Wild-type MxA, MxA(R640A), and MxA(L617D) accumulated in a pronounced punctuate pattern in the perinuclear region ( Fig. 2A). Parallel immunostaining of the split-GFP-MxA fusion proteins with an MxA-specific antibody revealed an identical distribution of the MxA and GFP signals. This observation indicates that a very high percentage of the fusion proteins is engaged in oligomeric structures. The MxA-specific staining of split-GFP-MxA (M527D) fusion protein revealed a diffuse, predominantly perinuclear localization ( Fig. 2A). Immunostaining of an untagged version of wild-type MxA and MxA mutants expressed in Vero cells revealed a more dispersed distribution of MxA in the cytoplasm. Because split-GFP complexes have to be stabilized by incubation at 27°C for 2 h to yield a detectable signal, we also incubated the untagged version of MxA under these conditions before fixation. This led to a pronounced accumulation of MxA in dot-like structures ( Fig. 2A). Therefore, it is conceivable that the observed punctuate pattern of the split-GFP-MxA fusion proteins is, at least in part, the result of incubation at 27°C. As reported previously, no significant MxA immunostaining or GFP-MxA signal was detected in the nuclei ( Fig. 2A) (6, 30).
In the next step, we assessed which stoichiometry wild-type MxA and its interface or hinge 1 variants adopt in mammalian cells using gel electrophoresis under non-denaturing conditions. For this purpose, Vero cells were transfected with expression vectors encoding wild-type MxA, MxA(R640A), MxA (L617D), MxA(M527D), and the GTPase-deficient MxA(T103A). After 24 h, the cells were lysed under non-denaturing conditions, and the lysates were subjected to non-denaturing PAGE and Western blot analysis using an MxA-specific antibody (Fig.  2B). The results revealed that most of the wild-type MxA ran at a molecular mass of ϳ300 kDa, indicating that it predominantly formed tetramers. Interestingly, MxA(T103A) deficient for GTPbinding and hydrolysis also ran as a tetramer, suggesting that, in mammalian cells, MxA assumes a tetrameric form independent of GTP binding. The hinge 1 mutant MxA(R640A), however, which exhibited a molecular mass of ϳ140 kDa, appeared to assemble into dimers. This contrasted with our observation made with purified MxA(R640A), which assembled primarily into tetramers in the absence of GTP␥S (Fig. 1D). As expected, the interface variants MxA(L617D) and MxA(M527D) accu-mulated as dimers and monomers, respectively (Fig. 2B). MxA from lysates of IFN-induced A549 cells also separated as tetramers, indicating that the stoichiometry of ectopically expressed MxA protein complexes is comparable with the endogenous protein (Fig. 2B). Incubation of extracts of MxAexpressing cells with 1 mM GTP or GTP␥S for 1 h at 4°C before separation on denaturing PA gels had no influence on the stoichiometry of MxA. Incubation at higher temperatures (25°C or 37°C) led to a rapid degradation of MxA despite the presence of a high concentration of protease inhibitors (data not shown). Therefore, when expressed in mammalian cells, wild-type MxA, and in particular also the hinge 1 mutant MxA(R640A), adopted a similar stoichiometry as purified recombinant MxA incubated with GTP␥S.
MxA Interface and Hinge 1 Mutants Exert Antiviral Activity against IAV in a Dose-dependent Manner-In the next step, we evaluated whether the MxA interface and hinge 1 variants restricted IAV infection. For this purpose, we transfected HEK293T cells with plasmids encoding wild-type MxA, the interface variants MxA(L617D) (dimer) and MxA(M527D) (monomer), and the hinge 1 variant MxA(R640A) (dimer/tetramer) as well as the inactive MxA(T103A) and infected them with the MxA-sensitive IAV strain pH1N1-NP (H5N1) at various concentrations (18). The virus titer was determined by measuring the TCID 50 per milliliter directly on the cells expressing wild-type MxA or MxA mutants, as outlined under "Experimental Procedures." Wild-type MxA as well as all interface and hinge 1 mutants exhibited a pronounced inhibition of IAV growth (ϳ90%) compared with cells expressing the inactive mutant MxA(T103A) (Fig. 3A). MxA expression control of the transfected cells, depicted in Fig. 3B, showed that the antivirally active variants of MxA were not expressed at higher levels than the positive and negative controls, wild-type MxA and MxA(T103A), respectively. In addition, we tested the antiviral activity of MxA complexes with defined stoichiometry using the plasmid-based IAV minimal replicon reconstitution assay of the A/Thailand/1/2004 (Kan-1) strain (Fig. 3C) (18). To that end, we co-transfected plasmids encoding the components of the viral replicase, the virus-dependent luciferase reporter gene, and increasing amounts of MxA-encoding plasmids (200, 400, and 800 ng of plasmid). The data clearly showed that the tested MxA stalk interface 1 and 2 mutants as well as the BSE hinge 1 variant strongly restricted the amplification of the luciferase reporter gene in an expression level-dependent manner, whereas MxA(T103A) had no effect, even at high expression levels (Fig. 3, C and D). As reported previously (8,12), the stalk interface 3 mutant MxA(R408A) and MxA lacking the L4 loop MxA(⌬L4) showed no antiviral activity, even at very high expression levels (data not shown). Most likely, mutation of MxA at residue Arg-408 not only prevents higher-order oligomerization but also disturbs a region that is essential for its antiviral function, and the L4 loop appears to be important for the interaction with the NP of IAV (8,9,12). Taken together, the data clearly demonstrate that dimeric as well as monomeric forms of MxA exhibited a pronounced antiviral activity, confirming that MxA assembly into higher-ordered oligomers is not a prerequisite for restricting IAV replication.
MxA Inhibits IAV Replication at a Step After Primary Transcription-We have shown previously that MxA inhibits replication of avian IAV in mouse cells at a cytoplasmic step following primary transcription of viral mRNAs (28). To verify that MxA also inhibits IAV replication in human cells at a step occurring after primary transcription, we infected cycloheximide-treated or untreated HEK293T cells expressing wild-type MxA, MxA(R640A), MxA(L617D), MxA(M527D), MxA(T103A), wild-type mouse Mx1 (mMx1), or the inactive mMx1(K49A) for 6 h with pH1N1-NP(H5N1) and performed RT qPCR analyses from isolated RNA with PB2-specific primers (Fig. 4A). mMx1 served as a positive control for the inhibition of primary transcription (Fig. 4A) (28). The data clearly show that, in cycloheximide-treated cells, wild-type MxA and MxA mutants forming complexes with defined stoichiometry did not affect primary transcription of IAV, whereas mouse Mx1, known to inhibit primary transcription, reduced the level of PB2 mRNA to less than 30%. As expected, the inactive Mx1(K49A) mutant exhibited no effect on the primary transcription of IAV. The activity of cycloheximide was verified in the same experiment by Western blotting (Fig. 4B). In the absence of cycloheximide, all antivirally active variants of MxA as well as mMx1 reduced the levels of PB2 mRNA to ϳ50% of the inactive controls MxA(T103A) and mMx1(K49A).
Next we monitored the nuclear import of IAV vRNPs into the nucleus of infected cells in the presence or absence of MxA expression. To make sure that no other IFN-induced effector protein was interfering with the assay, we utilized Vero cells lacking IFN type I genes. We infected Vero cells ectopically expressing wild-type MxA in the presence of cycloheximide with m.o.i. 10 of fowl plague virus (FPV, H7N7) and analyzed the nuclear import of vRNP by immunofluorescence microscopy using antibodies specific for IAV NP and MxA. FPV was chosen because of its high sensitivity to MxA (36) and efficient translocation of incoming vRNP into the nucleus. 5 As shown in Fig. 4, C-E, we were not able to detect any retention of the vRNPs in the cytoplasm when MxA was present. Images were assessed for nuclear import of vRNPs by counting MxA-expressing cells with a nuclear NP signal (Fig. 4D) or measuring the NP mean nuclear signal intensity of MxA-expressing cells (Fig. 4E). Cells transfected with empty vector were used as a negative control (Fig. 4, D and E). Taken together, these data demonstrate that human MxA does not interfere with the nuclear import of vRNPs and that it inhibits IAV replication at a step that follows primary transcription.
Overexpression of NP Interferes with the Antiviral Function of Human MxA-We and others have shown previously that overexpression of PB2 and NP interfered with the antiviral function of mouse Mx1 (15)(16)(17). Therefore, we next wanted to test whether the activity of MxA could be overcome by increased expression of NP and/or PB2. For this purpose, we 5 P. E. Nigg and J. Pavlovic, unpublished data. made use of the IAV minimal replicon reconstitution assay in HEK293T cells either expressing wild-type MxA or the GTP binding-deficient mutant MxA(T103A). The level of wild-type MxA and MxA(T103A) expression was first adjusted to a level yielding ϳ50% to 70% inhibition of IAV polymerase activity by wild-type MxA compared with MxA(T103A) (Fig. 5, A and B). As a control, we employed mouse wild-type Mx1 and its inactive mutant Mx1(K49A) (Fig. 5, C and D). To assess whether NP or PB2 interferes with the MxA-mediated inhibition, gradually increasing amounts of plasmids encoding either PB2 (10 -320 ng) or NP (50 -1600 ng) were added to constant amounts of plasmids encoding MxA or MxA(T103A). To assess the interference of NP or PB2 with the activity of human MxA or mouse Mx1, the values (normalized for transfection efficiency) obtained for MxA or Mx1 were divided by the corresponding values for MxA(T103A) or Mx1(K49A) and expressed as percentage inhibition relative to the inactive mutants. For MxA, the results revealed that only increasing amounts of NP led to interference with the MxA activity, whereas increasing amounts of PB2 had no effect (Fig. 5, A and B). By contrast, overexpression of NP as well as PB2 interfered with the activity of mouse Mx1 (Fig. 5, C and D). Therefore, at least in this exper-  imental setup, MxA appeared to target only NP, whereas Mx1 appeared to functionally interact with both NP and PB2, as described previously. The increased values obtained in samples with a high concentration of NP independent of MxA or Mx1 are the result of enhanced replication efficiency of the mini replicon (Fig. 5, B and D) in the presence of a high concentration of NP.
Dimeric Variants of MxA Interact with NP Independent of Other Viral Proteins-Despite the fact that, so far, experimental evidence for a physical interaction of human MxA with NP is lacking, the available data strongly suggest a functional relationship between these two proteins (18,19,33) (Fig. 5B). On the basis of our observations that GTP binding led to a disassembly of MxA from higher-ordered oligomers to tetramers and dimers that efficiently blocked IAV replication, we wanted to test whether MxA variants forming complexes with defined stoichiometry would exhibit a higher affinity to NP than wildtype MxA. For this purpose, we carried out co-immunoprecipitation assays. We transfected Vero cells with plasmids encoding wild-type MxA, MxA(R640A), MxA(L617D), MxA(M527D), or MxA(T103A) and subsequently infected them for 6 h with the IAV strains pH1N1-NP(H5N1) or rSC35M, both highly sensitive to the activity of MxA. Immunoprecipitation of NP revealed that both dimer-forming variants, the hinge 1 mutant MxA(R640A), as well as the interface mutant MxA(L617D) efficiently co-precipitated with NP. Wild-type MxA, the monomeric variant MxA(M527D), and the GTP binding-deficient mutant MxA(T103A) exhibited no or only a barely detectable interaction with NP (Fig. 6A). Therefore, these data strongly support our model that oligomeric MxA has first to be converted into dimers to efficiently bind to NP. Because NP exists in a vRNP-bound form as well as in an unbound form in infected cells, we assessed in a next step whether MxA is able to interact with ectopically expressed NP in the absence of any additional viral component. Therefore, we co-transfected plasmids encoding wild-type MxA, MxA(R640A), MxA(L617D), MxA(M527D), or MxA(T103A) with plasmids encoding NP either derived from a highly Mx-sensitive avian IAV strain (Kan-1) or two MxA-resistant strains (A/Hamburg/4/2009 (pH1N1) and A/PR/8/34 strain (PR8)). The NP of the MxAsensitive strain efficiently co-precipitated with the two dimeric variants of MxA (Fig. 6B), corroborating the findings observed with IAV infection (Fig. 6A). We also observed co-immunoprecipitation of dimeric forms of MxA with NP of the MxA-resistant strains p(H1N1) and PR8, although to a lesser extent than with the NP of Kan-1 (Fig. 6B).

Discussion
The human MxA protein restricts the replication of a broad variety of negative-and positive-stranded RNA viruses as well as DNA viruses. Besides their different genomes, these viruses replicate at different subcellular locations, employing different replication strategies. Therefore, MxA either interferes with viral replication via a general, rather nonspecific mechanism affecting many different viruses, or MxA has developed different molecular mechanisms to block virus replication. Although, for some viruses, such as IAV, Thogotovirus, and La Crosse virus, the replication steps and/or viral targets inhibited by MxA have been identified, for most MxA-restricted viruses the mode of action or the viral targets remain undetermined.
Using murine 3T3 cells, we have shown previously that MxA inhibits IAV replication at a step occurring after primary transcription. Synthesis of primary transcripts of the polymerase subunits, hemaglutinin, neuraminidase, and NP are unaffected by MxA, whereas, surprisingly, the production of primary transcripts of NS1 and M1 is increased 2-to 3-fold. The primary transcripts are exported efficiently and have the capacity to be translated in vitro (28). However, when translocated to the nucleus by means of an N-terminal nuclear localization sequence, MxA efficiently blocks primary transcription of IAV (7). We now observed very similar results in human cells. Ectopic expression of MxA restricted IAV replication without affecting the nuclear import of incoming vRNPs (Fig. 4). In addition, our RT qPCR data obtained from mRNA isolated from HEK293T MxA-expressing cells infected with IAV revealed that, in the presence of MxA, primary transcription of PB2 mRNA is not blocked (Fig. 4A). Xiao et al. (39) have reported recently that MxA interferes with IAV replication at two distinct stages, a step prior to primary transcription together with an unknown IFN-induced protein and a step following primary transcription. In our study, we focused only on the second step and analyzed the antiviral function of MxA in the absence of other IFN-induced effector proteins known to inhibit IAV entry and uncoating, such as IFITMs (40).
Mx proteins have the capacity to assemble into oligomeric structures of various complexities. In vitro, MxA protein assembles into tetramers (12) and has the capacity to assume higher-ordered oligomeric structures by forming rings or helical filaments (41,42). Moreover, on the basis of the crystal structure, Gao et al. (5,12) have proposed that MxA assembles into ring-like structures. The question whether monomeric . NP was immunoprecipitated (IP) from lysates using a mouse monoclonal NP-specific antibody, and the resulting complexes were subjected to Western blot analyses using polyclonal rabbit antibodies specific for MxA and NP. In addition, input samples (whole cell lysates, WCL) also were monitored for expression of MxA, NP, and GAPDH (loading control) by Western blot analysis. B, Vero cells were co-transfected with plasmids encoding the indicated MxA variants together with plasmids encoding NP from the indicated IAV strains. Immunoprecipitation of NP and Western blot analyses of immunoprecipitated proteins and whole cell lysates were performed as described in A. Immunostaining of the Western blots was performed with antibodies against MxA, NP, or GAPDH. The Western blots shown are representative of three independent experiments. and/or oligomeric forms of MxA exert their antiviral activity against IAV and other viruses remains an unsolved issue. Several previous studies have demonstrated that the monomeric mutant MxA(L612K) efficiently restricts IAV, Thogotovirus, and vesicular stomatitis virus (6,25,27). However, Gao et al. (5,12) have reported recently that dimeric and monomeric interface and hinge mutants of MxA failed to exhibit antiviral activity against IAV and La Crosse virus using a minimal replicon reconstitution assay for IAV or an N sequestration assay for La Crosse virus, respectively. On the basis of their findings, Gao et al. (5) have proposed a model in which MxA would assemble into oligomeric rings around incoming vRNPs of IAV. The major caveat of this model, however, is the fact that MxA inhibits IAV replication after the nuclear import of viral RNPs and after primary transcription of viral mRNAs has taken place ( Fig.  4) (7, 28). Therefore, we re-evaluated the antiviral activity of the dimeric and monomeric variants of MxA (plasmids provided by Georg Kochs) by performing infection experiments with IAV. Intriguingly, dimeric MxA variants with mutations in the stalk interface 1 or BSE hinge region as well as the monomeric interface 2 mutant efficiently restricted IAV, indicating that MxA does not necessarily have to assemble into higher oligomeric structures to exert antiviral activity against IAV. On the basis of the same minimal replicon reconstitution assay we used in this study (Fig. 3C), Gao et al. (5,12) have shown previously that the MxA variants forming complexes with defined stoichiometry, MxA(R640A), MxA(L617D), and MxA(M527D), do not restrict IAV replication. However, they adjusted the amount of MxA expression to a level where wild-type MxA only reduced the activity of the IAV polymerase complex to 70 -80% of the negative control. In fact, we obtained very similar results at low expression levels of MxA proteins (Fig. 3C).
The question of whether GTP promotes oligomerization or disassembly of MxA oligomers remains an issue of debate. Clearly, oligomerization of purified MxA is highly salt-dependent (41). Here we observed that, in the absence of GTP under medium salt conditions (200 mM KCl), purified recombinant MxA formed higher-order oligomers that, upon addition of GTP␥S, disassembled primarily into tetramers and, to a lower extent, also into dimers. Similarly, Kochs et al. (41) have demonstrated previously that, under high-salt conditions (300 mM NaCl), MxA exists predominantly in a lower oligomeric, nonsedimentable form. Addition of GTP␥S does not increase the sedimentable fraction of MxA, indicating that GTP␥S does not mediate oligomerization. By contrast, Gao et al. (5,12) have reported recently that, under the same experimental conditions, the non-sedimentable fraction of MxA converts into a sedimentable oligomeric structure. The authors explained their observation with the generation of an additional nucleotide-dependent interphase in the G domain (5). However, recent evidence indicates that GTP-dependent G domain dimerization of MxA is a transient process that cannot be stabilized by GTP␥S (13). Under low-salt conditions, MxA forms predominantly condensed oligomeric and sedimentable structures that, in the presence of GTP␥S, rearrange into spiral or stack-like forms (41). In the case of dynamin, preassembled sedimentable oligomers convert into lower oligomeric, non-sedimentable structures in the presence of GTP␥S (37).
In vitro, the hinge 1 mutant MxA(R640A) predominantly formed tetramers that converted into dimers in the presence of GTP␥S (Fig. 1B). Gao et al. (5) have purified recombinant MxA(R640A) predominantly in the form of dimers, even in the absence of GTP␥S, but in the presence of substantially higher salt concentrations (400 mM) . Therefore, it is conceivable that, for MxA(R640A) like for wild-type MxA, the stoichiometry of these recombinant proteins is not only a function of GTP but also strongly dependent on salt concentration.
Non-denaturing PAGE analyses of MxA expressed endogenously or ectopically in human cells revealed that wild-type MxA accumulated primarily in a tetrameric form. The reason why MxA(R640A), in our hands, runs predominantly as a tetramer when affinity-purified but as a dimer when analyzed in lysates of mammalian cells is not entirely clear, but it might be due to the presence of GTP and/or accessory proteins in the cellular context. Taken together, wild-type MxA appears to exist primarily as tetramer in the cytoplasm of infected cells and, at least in vitro, has the capacity to disassemble into dimers in a GTP-dependent fashion. The fact that the dimeric forms MxA(R640A) and MxA(L617D) exert antiviral activity strongly suggests that the GTP-dependent formation of dimers is important for the function of MxA. In this context, it is interesting to note that human MxB (Mx2), shown recently to restrict human immunodeficiency virus, appears to act as a dimer, albeit in a GTP-independent manner (43)(44)(45).
The observation that predominantly dimeric forms of MxA have the capacity form stable complexes with NP in infected cells lends further support to our hypothesis that dimers play a pivotal role for the activity of MxA. The fact that MxA dimers also formed a complex with ectopically expressed NP from a highly Mx-sensitive avian H5N1 strain in the absence of any other viral component demonstrates that MxA can interact with NP that is not bound within the vRNP. Wild-type MxA did not form a complex with NP at detectable levels, even when the cells were infected with a highly MxA-sensitive virus (Fig. 6A). A possible explanation for this observation may be that, following infection with IAV, only small amounts of MxA are disassembled into dimers to interact with NP. The major part of MxA could then remain in higher-ordered oligomeric structures. Alternatively, it may be possible that the high affinity of the MxA dimer to NP is only required for the initial recognition of NP. After the initial interaction, MxA could convert again into higher oligomeric structures exhibiting a low affinity for NP. In vitro NP is known to form oligomers even in the absence of RNA (46). Therefore, it may be possible that MxA reoligomerizes around newly synthesized NP oligomers, thereby inhibiting its function. In the case of the antivirally active monomeric form of MxA, MxA(M527D), it is conceivable that a lowaffinity interaction to NP is sufficient for antiviral activity in certain assays. Alternatively, the transient G domain dimerization stimulated by GTP binding (13) could also result in a transiently increased affinity for NP that may be sufficient to disrupt binding of NP to UAP56.
Testing the interaction of NPs from two distinct MxA-resistant strains revealed that NP from the 2009 pandemic H1N1 strain (pH1N1) was still capable of interacting with MxA, whereas NP from the PR8 strain (a derivative of the 1918 H1N1 strain) showed reduced complex formation. Interestingly, these twoIAVstrainsdevelopedtheirMxAresistancephenotypeindependently of each other (19) and, therefore, introduced distinct adaptive mutations into their NP coding sequence. These findings led us to the conclusion that there are at least two mechanisms for IAV to evade restriction by MxA. For the PR8 strain, this entails minimizing the affinity to MxA, whereas the pH1N1 strain employed a different strategy, possibly by reducing the affinity to an MxA-associated protein such as UAP56 (30).
Whether MxA interacts directly with NP or via the MxAassociated UAP56 remains to be determined. Of course, it is possible that MxA restricts the replication of viruses other than IAV by oligomerizing into ring-like structures around nucleocapsids, "strangling" them in a dynamin-like fashion by GTPdependent mechanical constriction (13). However, on the basis of our data presented here, we propose a new model in which, in response to IFN induction, MxA is synthesized and assembled into higher oligomeric structures, possibly partially associated with membranes of the smooth endoplasmic reticulum (38). These oligomeric structures may represent a preactive form of MxA. In infected cells, a still undefined viral component would then trigger a GTP-driven disassembly of oligomeric MxA into dimers, allowing the association with newly synthesized NP translated from primary transcripts. Therefore, sequestration of NP from replication sites in the nucleus may represent the long sought after mechanism of MxA action against IAV.