Interferon-inducible mouse Mx1 protein that confers resistance to influenza virus is GTPase.

The murine Mx1 protein is an interferon-inducible nuclear protein and confers resistance to influenza virus infection even though the resistance mechanism is yet unclear. The Mx1 protein contains a tripartite GTP-binding domain consisting of GXXXXGKS, DXXG, and T/NKXD motifs. In the GTPase gene superfamily such as p21ras protein, signal-transducing G protein, and translation elongation factor, the GTPase activity plays a key role in each protein function. Here we show that GTPase activity is indeed associated with the intact Mx1 protein purified from Escherichia coli expressing Mx1 cDNA. Amino acid substitution within the GTP-binding motif led to significant reduction in the GTPase activity. Yeast vacuolar protein sorting (VPS1) protein and the rat microtubule-associated mechanochemical enzyme dynamin were found to be homologous to Mx1 not only in the tripartite GTP-binding motif, but also in the amino-terminal region of approximately 300 amino acids in length. The function of Mx1 is discussed in comparison with these proteins.

T h e murine Mxl protein is an interferon-inducible nuclear protein present in inbred mouse strains resistant to influenza infection, but not in strains sensitive t o this virus (1). Mxl cDNA was cloned from A2G mice resistant to influenza virus (2). Sequence analysis indicated that the Mxl protein consists of 631 amino acids, >30% of which are charged, and contains at the carboxyl terminus a nuclear location signal that is essential for translocation of the Mx protein into nuclei (3). T h e Mxl gene was mapped to mouse chromosome 16 and consists of 14 exons spread over >55 kilobase pairs of DNA (4). Two Mxalleles were identified one lacking three exons and the other containing a nonsense mutation that leaves a potential coding capacity for the amino-terminal half of the Mxl protein ( 5 ) . Expression of the Mxl protein in influenzasensitive cells and animals lacking this protein is sufficient to promote an antiviral state (2,6). Moreover, microinjection of an anti-Mxl antibody into Mx+ cells effectively neutralizes the antiviral state of interferon-treated cells (7). However, the molecular mechanism by which it confers this resistance is still far from being understood. * This work was supported by grants-in-aid from the Ministry of Education, Science, and Culture of Japan (to M. N., K. N., and A. I.) and a grant for the Biodesign Research Program from RIKEN (to K. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$$ To whom correspondence should be addressed.
Mx-like proteins and genes have been found in a number of vertebrates. For instance, two Mx-like proteins have been identified in human. Human Mx proteins are, however, prese n t in cytoplasm. Moreover, human MxA produces a broader range of antiviral state than murine Mxl, preventing both orthomyxovirus and rhabdovirus infections (8), whereas the human MxB protein shows no detectable inhibition of viral infection. All the Mx proteins so far sequenced contain the three consensus sequence elements typically found in GTPbinding proteins (8,9). In this report, we show that the Mxl protein is indeed a GTP-binding protein with GTPase activity by analyzing both t h e authentic Mxl protein and a variant Mxl protein that contains a point mutation in one of the GTP-binding domains. The function of the Mxl protein is discussed in comparison with other GTPase proteins.

MATERIALS AND METHODS
Antibodies against Mxl Protein and Immunoblot Analysis-Two kinds of rabbit hyperimmune serum containing anti-Mxl antibodies (anti-XNO1 or anti-XC04) were prepared by immunizing animals with synthetic peptides corresponding to either the amino-terminal region (amino acids 79-89) or the carboxyl-terminal region (amino acids 617-631) of the Mxl protein. For immunoblot analysis, proteins were separated by electrophoresis on 7.5% polyacrylamide gel in the presence of 0.1% sodium dodecyl sulfate (SDS)' and electrophoretically transferred to a nitrocellulose filter (Schleicher & Schuell). The filter was exposed first to either the anti-XNO1 or anti-XCO4 antibodies and then to horseradish peroxidase-conjugated second antibodies against the rabbit p chain.
Purification of Mxl Protein-Escherichia coli DH5 containing pTrpAl2Mx was grown in 20 liters of M9 minimal medium at 30 "C (10). Cells were suspended in 80 ml of buffer A (20 mM HEPES/ NaOH (pH 7.6 at 25 "C), 1.5 mM MgC12, 0.2 mM EDTA, 0.1 mM dithiothreitol, and 0.1 mM phenylmethylsulfonyl fluoride) containing 200 mM KC1 and disrupted with sonication. The extract was centrifuged at 12,000 rpm for 30 min and then at 30,000 rpm for 2.5 h. The supernatant was dialyzed against buffer B (20 mM HEPES/NaOH (pH 7.6 at 25 "C), 0.1 mM EDTA, 10% glycerol, 0.1 mM dithiothreitol, and 0.1 mM phenylmethylsulfonyl fluoride) containing 50 mM KC1 and subjected to chromatography on a phosphocellulose column (1.5 X 17 cm) previously equilibrated with buffer B containing 50 mM KCl. Proteins were fractionated with a 50-700 mM KC1 linear gradient at a flow rate of 23 ml/h. The Mxl protein fractions eluting around 270 mM KC1 were pooled and dialyzed against buffer C (buffer A plus 10% glycerol) containing 50 mM KCl. The dialyzed sample was subjected to chromatography on a DEAE-TOYOPEARL column (1.5 X 5.7 cm) previously equilibrated with buffer C containing 50 mM KCl. Proteins were fractionated with a 50-700 mM KC1 linear gradient at a flow rate of 30 ml/h. The Mxl protein fractions eluting around 255 mM KC1 were pooled and fractionated by gel filtration through a Protein-Pak 300 column previously equilibrated with buffer C containing 200 mM KCl. The pooled Mx fraction (5.9 mg) was The abbreviations used are: SDS, sodium dodecyl sulfate; HEPES, N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid; HPLC, high pressure liquid chromatography.
dialyzed against buffer C containing 50 mM KC1 and subjected to SP-TOYOPEARL column chromatography. The column (0.7 X 6.5 cm) was eluted using a linear 50-700 mM KC1 gradient a t a flow rate of 16.8 ml/h. The Mxl protein was eluted around 260 mM KCI.
Assay of GTPase Activity-The assay of GTPase activity was performed a t 37 "C for 60 min in 25 pl of reaction mixture containing 50 mM Tris-HC1 (pH 8.0), 5 mM MgC12, 0.1 mM dithiothreitol, 130 p~ GTP, and 13 nM [LY-~*P]GTP (1 pCi, 3000 Ci/mmol). Products of the reaction were resolved by chromatography on polyethyleneiminecellulose plates in 1.6 M LiC1.
Construction and Purification of Mutant Mxl Protein-Expression vector pHR3 was constructed by replacing the 375-base pair EcoRI-BamHI fragment of plasmid pHR148 (11) with the 21-base pair EcoRI-BamHI fragment of M13mp18. For site-directed mutagenesis of Mxl cDNA, an oligonucleotide with the sequence d(AACAGAG ATCTTCCCAGA) was used as primer for DNA synthesis. In the resulting mutant Mx gene, Ser a t position 50 of the Mxl protein was substituted with Ile. Insertion of the mutagenized Mxl cDNA into pHR3 resulted in pS50I. The mutant Mxl protein was purified as described above with slight modifications. In brief, the pellet after high-speed centrifugation of cell extracts was dissolved in buffer B containing 50 mM KC1 and loaded onto a phosphocellulose column. After washing with a 5-column volume of buffer B containing 50 mM KC1, proteins were eluted with 700 mM KC1 in buffer B. The eluted fraction was dialyzed against buffer C containing 50 mM KC1 and directly applied to a Protein-Pak G-DEAE HPLC column (8.2 X 75 mm) equilibrated with the same buffer. The column was washed with 11.9 ml of buffer C, and then proteins were eluted with 39.9 ml of a 50-700 mM KC1 linear gradient in buffer C at a flow rate of 0.7 ml/ min. The Mxl protein was eluted around 490 mM KC1.
Binding of [LY-~*P]GTP to Mx1 Protein-Samples were placed in a 96-well tissue culture plate on ice and irradiated with a UV source (Mineralight lamp UV254) a t 18 watts a t a distance of 5 cm for 30 min. Proteins were separated by electrophoresis on 7.5% polyacrylamide gel in the presence of 0.1% SDS. Labeled proteins were visualized by autoradiography of the gel. Radioactivity of the GTP binding to the Mxl protein was measured using a Fuji Bio-Image analyzer.

RESULTS
Purification of Mxl Protein-To prepare the intact murine Mxl protein, plasmid pTrpAl2Mx was constructed by placing the Mxl cDNA clone (10) downstream of the E. coli trp promoter in the expression vector plasmid pHR3 (11). A polypeptide with an apparent molecular mass of 72 kDa was produced in E. coli strain DH5 harboring the plasmid pTrpA12Mx but not the vector plasmid. For identification of the Mxl protein, two kinds of rabbit hyperimmune serum were prepared by immunizing animals with snythetic peptides corresponding to either the amino-or carboxyl-terminal region of the Mxl protein. The resulting sera specifically reacted with the 72-kDa Mx protein synthesized in Mx-positive A2G mouse cells treated with interferon-alp (Fig. 1). This protein was not detected in extracts of A2G mouse cells not treated with interferon or in extracts of Mx-negative NIH 3T3 cells either with or without interferon treatment. Both antisera also immunoprecipitated the Mxl protein synthesized in vitro by translating transcripts of Mxl cDNA (data not shown).
The 72-kDa polypeptide in the Mxl cDNA-expressing E. coli was identified as the Mxl protein by Western blotting using the Mxl-specific antibodies (see  (8,9). Using the purified Mx protein, we examined whether GTPase activity is associated with Mxl or not. After incubation with [a-32P]GTP, reaction products were analyzed by thin-layer chromatography on polyethyleneimine-cellulose using either 1.0 or 1.6 M LiCl. As illustrated in Fig. 2' 2, [CU-~~PIGTP was converted to [cY-~'P]GDP, and this GTPase activity co-chromatographed with the Mxl protein. Under the same reaction conditions except that [cx-~~PIGTP was replaced with one of the other 32P-labeled nucleoside triphosphates, the hydrolysis of ATP was -37% the activity of GTP hydrolysis, but that of CTP and UTP was below the detection level (data not shown). Thus, the Mxl protein possesses the substrate preference for GTP.
In the case of p21ra*, GXXXXGKS and DXXG motifs constitute the phosphate-binding loop, and the T/NKXD motif forms the guanine nucleotide-binding site (12). In this case, the serine residue in the GXXXXGKS motif within the phosphate-binding domain is essential for binding of GTP in coordination with Mg2+ (12,13). To test that the corresponding serine residue in the Mxl protein is also involved in its GTPase activity, a mutant Mxl protein was prepared in which the serine residue at position 50 was substituted with isoleucine. For this purpose, the mutagenized Mxl cDNA was synthesized in vitro and inserted into pHR3 to produce plasmid pS50I. Extracts of E. coli transformed with pHR3, pTrpAlBMx, or pS50I were prepared and subjected to fractionation by a modified method of Mxl protein purification. Since the Mxl protein was found to form aggregates, it was The GTPase assay was performed at 37 "C for 60 min as described under "Materials and Methods" using aliquots (5 pl each) of the indicated SP-TOYOPEARL column chromatography fractions. Products of the reaction were resolved by chromatography on a polyethyleneimine-cellulose plate in 1.6 M LiCl. The first lane contained no fraction. The positions of AMP, CMP, GMP, UMP, GDP, and GTP were identified using unlabeled reference nucleotides. recovered as precipitates by centrifugation for 2.5 h at 30,000 rpm. The pellet was dissolved and absorbed onto phosphocellulose, and a one-step eluate at 0.7 M KC1 was fractionated by Protein-Pak G-DEAE HPLC using a 50-700 mM KC1 linear gradient. The recovery of the Mxl protein by this modified method was better than that by the aforementioned method using the supernatant as starting material (see "Materials and Methods"), yielding 0.45 mg of pure Mxl protein starting with 2 g of E. coli cells in a typical experiment. When the same amount of E. coli cells was used for purification, a similar yield was reproducibly obtained by this modified method. For instance, Fig. 3A shows a gel pattern of the purified preparations of wild-type and mutant Mx proteins analyzed using the same volumes of the equivalent column fraction. Both the wild-type and mutant Mx protein fractions gave a single major band with similar staining intensity. The mutant Mx protein was identified by immunoprecipitation analysis using anti-Mx antibodies (data not shown). The control fraction from extracts of E. coli harboring the parent expression vector pHR3 did not contain any cross-reacting proteins (Fig. 3A).
Next, we compared the GTPase activity among these three fractions. As shown in Fig. 3B, the GTPase activity of the or pS50I (lane 4 ) or the corresponding fraction from extracts of cells containing pHR3 (lane 2). Lane I , 2 pg each of @-galactosidase (130 kDa), bovine serum albumin (69 kDa), and ovalbumin (46 kDa). Each sample (12 pl) was incubated in 50 pl of the GTPase reaction mixture without cold GTP for 15 min a t 30 "C and then exposed to UV light. mutant Mxl protein was only one-fourth the level of the wildtype Mxl protein, whereas virtually no GTPase activity was detected for the same volume of the equivalent column fraction from cells without Mxl cDNA. These observations indicate that GTPase is an intrinsic activity associated with the Mxl protein and the GXXXXGKS motif plays a role in this GTPase activity. The calculated turnover number for the wild-type Mx protein is 5.8 mol of GTP hydrolyzed per min/ mol of protein at 37 "C.
Detection of GTP Binding Activity of Mxl Protein-To demonstrate direct interaction between the Mxl protein and GTP, we attempted UV photoaffinity labeling (14). When the wild-type Mxl protein was exposed to UV light in the presence of [cY-~*P]GTP, a single band of labeled protein was detected after gel electrophoresis (Fig. 4). No proteins were labeled without UV irradiation (data not shown). Likewise, no band appeared when several unrelated proteins (@-galactosidase, bovine serum albumin, and ovalbumin) were tested by this treatment. The Mxl protein cross-linked with GTP migrated a little slower than the untreated Mxl protein. This band was, however, not detected for the fraction derived from an extract of cells containing pHR3. These observations indicate that the Mxl protein physically interacts with GTP. The level of GTP cross-linked to the wild-type Mxl protein was 4.2fold higher than that cross-linked to the mutant Mxl protein. This ratio of GTP binding is in good agreement with the ratio of GTPase activity. Thus, we concluded that the amino acid change from Ser to Ile at position 50 of the Mxl protein resulted in a decrease in the GTPase activity due to reduction in GTP binding affinity.

DISCUSSION
In the family of GTP-binding proteins including signal transducing G protein and elongation and initiation factors involved in protein synthesis, the GTPase activity plays a regulatory role in each protein function. More recently, low molecular mass GTP-binding proteins related to the ras protooncogene have been implicated in playing a role(s) in regulating membrane vesicle budding and fusion involved in transport from the endoplasmic reticulum through the Golgi complex as well as in processes of endocytosis and exocytosis. ras-related GTP-binding proteins also appear to be involved in cytoskeletal polymerization. Furthermore, secretory protein transport across the endoplasmic reticulum is mediated by the receptor of the signal recognition particle, which has been shown to be a GTP-binding protein. When compared to typical enzymatic reactions, the rates at which these GTPbinding proteins convert GTP to GDP are relatively slow, ranging from 12 to 250 mmol/min/mol of protein at 37 "C (15). The rate of GTP hydrolysis by the Mx protein is also slow (-6 mol/min/mol of protein). Thus, the Mxl protein may act as a molecular switch like these GTPase family proteins.
The GTPase family proteins are interconvertible among three conformational states, i.e. GDP-bound, empty, and GTP-bound, by a cyclic reaction of binding and hydrolysis of GTP (16). For many GTPases, the cycle of GTP binding and hydrolysis takes place involving two regulatory components: a guanine nucleotide release protein and a GTPase-activating protein. Similar regulatory proteins may therefore be involved in the control of antiviral activity of the Mxl protein. Alternatively, if the Mxl protein is categorized into another GTPase family such as tubulin, the Mxl protein itself could be reversibly assembled by the medium of GTP hydrolysis. The assembled form of the Mxl protein may be active in the establishment of the antiviral state.
The mouse Mxl protein is located in the nucleus (2), whereas the human MxA protein is a cytoplasmic protein.
The mouse Mxl protein confers selective resistance to influenza virus, whereas the human MxA protein can inhibit both influenza virus and vesicular stomatitis virus. Nevertheless, the amino-terminal regions of the Mx protein family, including murine (Mxl and Mx2), rat (Mxl to Mx3) (17), human (MxA and MxB), and fish Mx proteins, show a high degree of sequence conservation (18). The conserved regions contain a tripartite GTP-binding domain, suggesting that all the Mx family proteins have GTPase activity and that this GTPase activity is essential for the antiviral activity.
Moreover, the Mx protein appears to have an important cellular function(s) because Mx homolog proteins have been identified in wide varieties of animal species including monkey, pig, cattle, sheep, horse, dog, cat, and rabbit (19) and because the Mx homolog protein has been shown to be essential for the viability of yeast. Along this line, it is worthwhile to note that the murine Mxl protein is homologous to not only the Mx family proteins, but also to two other cellular proteins. One is the product of yeast vacuolar protein sorting (VPS1) protein (20). Yeast cells contain a prominent vacuole that is considered to be equivalent to mammalian lysosome because of its low pH. Vacuolar protein sorting (VPS) encodes a protein required for sorting of newly synthesized vacuolar proteins from secretory proteins during their transport through the yeast secretory pathway. The other is the rat microtubule-associated mechanochemical enzyme dynamin (21). Microtubules are involved in various steps of intracellular motility. Cytoplasmic microtubules are composed of tubulin and a diverse set of microtubule-associated proteins. Dynamin is one of the microtubule-associated proteins, but is distinct from dynein and kinesin. Dynamin forms periodic cross-bridges between microtubules and converts them into highly organized bundles. The amino-terminal region of -300 amino acids is highly conserved among Mxl, vacuolar protein sorting gene 1, and dynamin (21). The tripartite GTP-binding domain exists in this conserved region, suggesting that both vacuolar protein sorting (VPS1) protein and dynamin are also GTPases. In fact, a low level of GTPase activity, i.e. -1 mol/ min/mol of protein (cited in Ref. 22), was detected for dynamin in addition to ATPase activity. This GTPase activity is markedly stimulated by microtubules. Along this line, it would be worthwhile to test possible effects of microtubules on the GTPase activity of the Mxl protein.
The Mx protein may therefore be involved in either protein sorting or intracellular motility. In the case of the human MxA protein, it is a cytoplasmic protein and inhibits growth of both influenza virus and vesicular stomatitis virus. Furthermore, it is noteworthy that vesicular stomatitis virus, Newcastle disease virus, and Sendai virus require cytoplasmic microtubule-associated proteins, cytoskeleton, and tubulin, respectively, for transcription and/or replication (23, 24). Likewise, the murine Mxl protein may interact with a nuclear structure such as the nuclear matrix. If so, it is possible that the growth of influenza virus may depend on the nuclear structure.
In the case of p21'"", the level of GTPase activity is closely related to its transformation activity (24). To understand whether the GTPase activity of the Mxl protein is indeed important for its antiviral activity, a systematic comparison is being made between the degree of influenza virus resistance and the level of GTPase activity for mutant Mxl proteins using both cultured cell and transgenic animal systems. As a further extension of this line of study, antiviral animals could be established by introducing mutant Mx genes that carry either accelerated or decelerated levels of the GTPase activity.