Membrane Association of Influenza Virus Matrix Protein Does Not Require Specific Hydrophobic Domains or the Viral Glycoproteins

VIROLOGY

220, 37–45 (1996) 0283

ARTICLE NO.

Membrane Association of Influenza Virus Matrix Protein Does Not Require Specific Hydrophobic Domains or the Viral Glycoproteins EVELYNE KRETZSCHMAR, MATTHEW BUI, and JOHN K. ROSE1 Department of Pathology, Cell Biology and Biology, Yale University School of Medicine, 310 Cedar Street, New Haven, Connecticut 06510 Received January 25, 1996; accepted April 2, 1996 The matrix protein of influenza virus is a major structural component of the virion which is generally believed to bridge between the membrane envelope and the ribonucleocapsid core. To investigate the interaction of M1 with cellular membranes in the absence of other influenza proteins, we examined its distribution by subcellular fractionation after expression in HeLa cells. Approximately 81 to 88% of M1 protein, expressed without other viral proteins, was soluble, whereas the remaining 12 to 19% was tightly associated with membranes. Conditions known to release peripherally associated membrane proteins did not detach M1 proteins from isolated membranes, suggesting that the fraction of M1 bound to membranes behaves as an integral protein. Coexpression of M1 with hemagglutinin or neuraminidase did not alter the extent of membrane association of M1 protein, indicating that there is no strong interaction between M1 and the cytoplasmic tails of the viral glycoproteins. Additional attempts were made to identify membrane binding domains in M1 protein. Mutants constructed with mutations in the four hydrophobic regions thought to be responsible for membrane association still exhibited the same levels of membrane association as that observed with wild-type matrix protein. Therefore, specific hydrophobic domains are apparently not required for membrane binding. q 1996 Academic Press, Inc.

INTRODUCTION

bilayer surrounding an inner nucleocapsid core. The viruses acquire their lipid bilayer as they bud from the host cell plasma membrane at areas enriched in viral envelope proteins. In the case of influenza A viruses there are four different viral membrane proteins: the hemagglutinin (HA) protein which recognizes neuraminic acid containing receptors at the cell surface; the enzyme neuraminidase (NA) which destroys cellular receptors; a small transmembrane protein, M2; and the most abundant viral protein, M1. In the infected cell, M1 is produced as a cytosolic protein and is incorporated into virions almost immediately after synthesis (Hay and Skehel, 1975). In the virion this protein is localized as an electrondense layer underneath the lipid bilayer. This position allows M1 to interact with the membrane lipids and the cytoplasmic tails of the viral glycoproteins hemagglutinin and neuraminidase as well as with the viral ribonucleoprotein (RNP) and with the viral NS2 protein (Yasuda et al., 1993). Distinct domains of the M1 protein such as RNA binding sites (Ye et al., 1987, 1989) and membrane binding domains (Gregoriades, 1980) may allow the association with vRNPs or membranes, respectively. Therefore, M1 might act as an adapter molecule and possibly initiate the process of viral assembly. However, there are many questions about the molecular mechanisms that govern viral assembly that have yet to be answered. So far, no evidence has been obtained for a direct interaction between M1 and the viral glycoproteins hemagglutinin and neuraminidase. The nature of the M1 protein

The assembly of components of enveloped viruses at the plasma membrane and their budding from the membrane have not been defined at the molecular level of protein–protein interactions. It has been suggested that there are specific interactions involving viral tails and the nucleocapsid which together assist the formation of a budding virion (reviewed in Compans and Choppin, 1975, and Dubois-Dalcq et al., 1984). With the alpha virus Sindbis, evidence has been obtained for interactions between the cytoplasmic tail of the viral glycoproteins and the nucleocapsids (Lopez et al., 1994), and for Semliki Forest virus, it has been shown that the coexpression of the nucleocapsid and spike proteins are sufficient to allow budding of the virus to occur (Suomalainen et al., 1992). However, other enveloped RNA viruses, such as Rhabdoviridae, Orthomyxoviridae, and Paramyxoviridae, present a more complicated picture in terms of the interactions that coordinate viral assembly, because they possess a matrix protein localized inside the lipid bilayer. For these viruses, the sequence of events and the molecular mechanisms that govern viral assembly require further analysis. Influenza viruses, members of the Orthomyxoviridae, are negative-stranded RNA viruses with an outer lipid

1 To whom correspondence and reprint requests should be addressed. Fax: (203) 785-7467. E-mail: [email protected].

0042-6822/96 $18.00

37

AID

VY 7936

/

6a17$$$141

04-30-96 00:53:57

Copyright q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.

viras

AP: Virology

38

KRETZSCHMAR, BUI, AND ROSE

interaction with lipid and protein components of membranes is also not clear (reviewed in Lamb, 1989). Membrane interactions were examined in vivo for the matrix protein of the enveloped, negative-stranded RNA rhabdovirus vesicular stomatitis virus (VSV) (Chong and Rose, 1993, 1994). These studies showed that a substantial fraction of matrix protein was tightly associated with the plasma membrane and was chemically or conformationally different from soluble matrix protein. Based on these findings the authors suggested a model for the assembly of VSV in which a fraction of newly synthesized matrix protein remains soluble in the cytoplasm, where it binds to and facilitates assembly of nucleocapsids. Another fraction of the matrix protein binds tightly to the cellular plasma membrane. Condensed nucleocapsids would then bind to regions of the membrane containing the matrix protein. It is possible that the matrix protein M1 of influenza virus functions in a manner similar to that proposed for the matrix protein of VSV. We wanted to determine whether the influenza A virus is an example of a second family of enveloped negative-stranded RNA viruses whose matrix protein is able to associate with host cell membranes. Furthermore we wanted to characterize the nature of M1’s membrane binding, define potential membrane-binding domains by site-directed mutagenesis in vivo, and to determine if the viral glycoproteins influenced binding of M1 to membranes. MATERIALS AND METHODS Viruses and antibodies Plaque-purified influenza WSN (H1N1) was used to infect embryonated chicken eggs as described earlier (Doms et al., 1985). To purify virions from the allantoic fluid, the virus was pelleted by centrifugation for 90 min at 35,000 rpm in a Beckman Ti45 rotor and resuspended in 10% sucrose. The virus was then passaged through successively smaller needles from 18 to 26 G and loaded onto a discontinuous sucrose gradient of 30 and 60% (wt/vol) sucrose in MNT (20 mM MES: 30 mM Tris, 100 mM NaCl, pH 7.5). The virus bands were collected, UV inactivated, and used to immunize rabbits. The specificity of the antibody obtained was confirmed by immunoprecipitating radioactively labeled MDCK cells infected with influenza WSN virus. The recombinant vaccinia virus vTF7-3 (Fuerst et al., 1986) was prepared as previously described (Whitt et al., 1989). Plasmid constructions Plasmid DNAs encoding the matrix, hemagglutinin, or neuraminidase proteins of influenza WSN (generously provided by Dr. Peter Palese, Mount Sinai Hospital, New York) were cloned into the XbaI site, HindIII/XbaI sites,

AID

VY 7936

/

6a17$$$141

04-30-96 00:53:57

or into the EcoRI/HindIII sites of pBS(KS/) Bluescript, respectively. To generate mutants of the matrix protein gene, site-directed mutagenesis was performed using the Sculptur in vitro mutagenesis system (Amersham). This technique is based upon the method described by Nakamaye and Eckstein (1986) in which the mutated strand contains thionucleotides. This strand is protected against restriction cleavage with NciI and can be selected after exonuclease digestion. The following oligonucleotides were used to generate mutants 1–7 (Fig. 4): 5*GATATTGAAAGATGAGTGGTGGAACCGAGGGCGAAACGGGCGGTGGCTCTATCGTCCCGTC3* for mutant 1, the oligonucleotide 5*CCTCTGACTAAGGGGGGTGGAGGAGGTGGGGGCACGCTCACCGTGCCTAG for mutant 2, the oligonucleotide 5*GCCAGTTGTATGGGCGGCGGAGGCAACAGGATGGGGGCTG3* for mutant 3 and the oligonucleotide 5*GGGCTGTGACCACTGAAGGGGGAGGTGGCGGGGGATGCGCAACCTGTGAACAGA to generate mutant 4. For mutants 1 to 4, the underlined nucleotides replaced the existing ones with glycine (see Fig. 4). The oligonucleotide 5*GGATTTGTGTTCACGAAGACCGTGCCTAG was used to generate mutant 5 and the oligonucleotide 5*GGGGCCAAAGAAATAGCAAAGAGTTATTCTGCTGGTGCAAAGGCCAGTTGTATGGG to generate mutant 6. The underlined nucleotides in mutants 5 and 6 replaced leucine residues with charged amino acids (Fig. 4). A 41 mer (5*CATTCCATGGGGCCAAAGAAGTATGCGCAACCTGTGAACAG) consisting of 20 nucleotides at the 3* (underlined) and 22 nucleotides at the 5* end of the desired mutation was used to delete the sequence encoding amino acids 115–146 and was designated mutant 7. The mutant genes were cloned into pBS(KS/) Bluescript where they can be expressed under control of the T7 promoter. The presence of the mutations was verified by sequencing. In addition to the described mutants, double mutants combining mutations with previously designed mutants were constructed: mutant 1 / 3, mutant 1 / 4, mutant 1 / 7, and mutant 5 / 6. This has been achieved for mutants 1 / 3, 1 / 4, and 1 / 7 by ligating the mutated regions together. In this case, a KpnI–AflII fragment containing partial 5* sequences of mutant 1 was ligated into the KpnI–AflII sites of the plasmids containing mutant 3, 4, and 7 genes, respectively. Mutant 5 / 6 was constructed by using the two oligonucleotides used to generate mutants 5 and 6 simultaneously for the site-directed mutagenesis reaction. A plasmid for engineering of a cell line expressing the wildtype matrix protein was constructed by cloning the gene into the XbaI site of the vector pCB6 (Hunziker et al., 1991). The plasmid was designated pCB6-M1. The coding region in this construct was expressed from the CMV promoter. Cells MDCK cells and HeLa cells were maintained in DMEM supplemented with 7 or 5% fetal bovine serum, respec-

viras

AP: Virology

MEMBRANE ASSOCIATION OF INFLUENZA VIRUS M1

tively. Monolayers of MDCK cells were transfected with pCB6-M1 via calcium–phosphate precipitation (Wigler et al., 1979). G418 (0.5 mg/ml)-resistant colonies of cells were isolated and screened for expression of M1 protein by immunofluorescence, immunoprecipitation, and Western blot. The transformed cell line was designated MDCK-M1.

107) by the equilibrium sedimentation method described above. The visible membrane band at the 10/65% sucrose interface was removed by side puncturing the centrifuge tube. Membranes were diluted to 10 ml with 10 mM Tris–HCl (pH 7.4) and pelleted by centrifugation at 35,000 rpm for 4 hr at 47 using a Beckman SW41 rotor. The membrane pellet was resuspended in 300 ml of 10 mM Tris–HCl (pH 7.4) and kept on ice.

Expression, radiolabeling, and immunoprecipitation of proteins HeLa cells (106 cells per 3.5-cm plate) were infected with vTF7-3 (Fuerst et al., 1986) at a multiplicity of infection of 10. After 30 min, the plasmids encoding wild-type or mutant matrix proteins, hemagglutinin, or neuraminidase were transfected using TransfectACE (Rose et al., 1991) in 1 ml DMEM. Plasmid amounts were 5 mg for plasmids expressed individually or 1.5 mg of M1 and 3.5 mg of hemagglutinin or neuraminidase for coexpression experiments. After 3 hr, the cells were supplemented with 5% fetal bovine serum and incubated for an additional hour. Cells were then labeled for 1 hr with [35S]methionine, followed by a 1-hr chase. For labeling of MDCK-M1 cells, the cells were induced 17 hr with 10 mM sodium butyrate in complete culture media to upregulate M1 expression and labeled with [35S]methionine for 1 hr before further processing.

39

RESULTS M1 protein associates with cellular membranes in the absence of other influenza proteins

Total cellular membranes were prepared from transfected and metabolically labeled HeLa cells (1.5 1

To obtain a high level of M1 protein expression in the absence of other influenza proteins, we used a transient expression system in which HeLa cells were first infected with a recombinant vaccinia virus (vTF7-3) encoding the bacteriophage T7 polymerase (Fuerst et al., 1986). Cells were then transfected with a plasmid encoding M1 (influenza A, WSN) under the control of the T7 promoter. Four and one-half hours postinfection, cells were labeled for 60 min with [35S]methionine, followed by a 1-hr chase. When cell lysates were immunoprecipitated with antiinfluenza serum and analyzed by SDS–PAGE, a 27-kDa protein was detected. The expressed protein comigrated with M1 protein from solubilized wild-type influenza A WSN virions (data not shown). To examine the distribution of M1 protein in transfected HeLa cells, the cell lysates were centrifuged on equilibrium density gradients (Bergmann and Fusco, 1988). This method is used to determine the membrane association of proteins. For this assay, total cell lysates were prepared by disrupting cells with a Dounce homogenizer. The postnuclear supernatants were adjusted to 80% sucrose, placed at the bottom of a centrifuge tube, and overlaid with sucrose in steps of decreasing density (65 and 10%). Following centrifugation, membranes and membrane-associated proteins band at the 10/65% interface, whereas soluble proteins remain at the bottom of the gradient. When the M1 protein was expressed in HeLa cells and analyzed for membrane association by this fractionation scheme, the majority of M1 protein remained with cytosolic proteins at the bottom of the gradient, while 12 to 19% of M1 colocalized with membranes at the 10/65% sucrose interface (Fig. 1A). In order to determine if M1 associates with membranes in a time-dependent manner, a pulse-chase experiment labeling M1 for 15 min with [35S]methionine or chasing for an additional 45 min was performed. The amount of membrane-bound M1 was slightly but probably not significantly lower directly after the chase, indicating that membrane association occurs very rapidly (data not shown).

AID

viras

Isopycnic gradient centrifugation and immunoprecipitation Transfected and metabolically labeled HeLa cells (2 1 106) or metabolically labeled MDCK-M1 cells (2 1 106) were harvested and fractionated essentially as described by Bergmann and Fusco (1988), with modifications. Plates were first rinsed with PBS and then scraped into an ice-cold 10% (wt/wt) sucrose homogenization buffer containing 10 mM Tris–hydrochloride (Tris–HCl, pH 7.4), 1 mM EDTA, and 100 kallikrein units of aprotinin per milliliter. Cells were disrupted with 25 strokes of a Dounce homogenizer on ice. Nuclei and debris were removed from the cell lysate by centrifugation at 4000 rpm for 4 min at 47. The resulting supernatant was made to 80% (wt/vol) sucrose, placed at the bottom of a Beckman SW41 centrifuge tube, and overlaid with 5 ml 65% (wt/vol) and 2.5 ml 10% (wt/vol) sucrose. The step gradient was then centrifuged to equilibrium at 35,000 rpm for 18 hr at 47. Fractions were collected from the top, diluted with detergent solution, and immunoprecipitated with rabbit anti-influenza serum as previously described (Rose and Bergmann, 1983). Labeled proteins were analyzed by SDS–PAGE and fluorography. Preparation of cellular membranes

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology

40

KRETZSCHMAR, BUI, AND ROSE

M1 associates with membranes in the presence of hemagglutinin and neuraminidase It was then of interest to determine if the coexpression of either hemagglutinin or neuraminidase with M1 would alter the fraction of M1 associated with membranes. In this case, separate plasmids encoding the hemagglutinin or neuraminidase and M1 proteins were transfected together into HeLa cells. Coexpression was verified by immunofluorescence microscopy at 6 hr posttransfection (data not shown). The membrane association of cells coexpressing M1 and hemagglutinin, or M1 and neuraminidase, was studied by subjecting the transfected cell lysates to isopycnic gradient analysis. Samples were fractionated and immunoprecipitated. After separation by SDS–PAGE, the amount of M1 in the membrane fraction and in the soluble fractions was determined by densitometry (Figs. 1B and 1C). Again, approximately 15% of M1 was membrane-bound. Thus, coexpression of M1 with hemagglutinin or neuraminidase did not alter the extent of membrane association of M1 protein. In the above experiments, the interaction of the M1 protein with membranes could have been an artifact of vaccinia virus infection. To examine membrane association of M1 in the absence of vaccinia virus, we used a MDCK cell line expressing M1 constitutively. When M1 protein was metabolically labeled after induction with sodium butyrate and its distribution was examined in this cell line, 15% of M1 was found to be bound to membranes (Fig. 2A). M1 was not expressed to visible levels in the uninduced control (Fig. 2B). Thus, the membrane binding of M1 was not specific to vaccinia-infected cells but was a property of the matrix protein itself. M1 protein interacts stably with cellular membranes

FIG. 1. Analysis of membrane-associated matrix protein M1. HeLa cells were first infected with vTF7-3 and then transfected with plasmid DNA encoding the M1, hemagglutinin, or neuraminidase protein. Four and one-half hours postinfection, the cells were pulse-labeled with [35S]methionine for 60 min, chased for 1 hr, and then subsequently homogenized with a Dounce homogenizer, brought to 80% sucrose, and overlaid with 65 and 10% sucrose layers. The step gradient was centrifuged to equilibrium and then fractionated from the top of the gradient. The fractions were immunoprecipitated with rabbit anti-influenza serum, analyzed by SDS–PAGE, and visualized by fluorography. A shows a gradient from cells expressing influenza M1 protein individually. B shows a gradient from cells coexpressing hemagglutinin and M1 proteins. C shows a gradient from cells coexpressing neuraminidase and M1 proteins. Fractions are numbered from the top to the bottom. Vaccinia virus (v.v.) background proteins are indicated. Quantitation of the amount of labeled protein in each fraction was carried out by scanning densitometry of autoradiographs.

Although the M1 protein structure has not been determined, analysis of the primary amino acid sequence of M1 does not reveal any stretches of hydrophobic residues sufficient to act as a transmembrane domain. Thus, M1 might be expected to behave as a peripheral membrane protein. In order to test if M1 is a peripheral protein, membrane-bound M1 protein was subjected to treatments which would characteristically remove peripheral membrane proteins, such as high pH, changes in ionic strength, or chelation of divalent cations. For these experiments, total cellular membranes containing associated [35S]methionine-labeled M1 protein from transfected HeLa cells were prepared by the sucrose flotation method. The visible membrane band at the 65/10% interface was removed by side puncture of the centrifuge tube, dialyzed, pelleted, and then treated with 2 M KCl, 50 mM EDTA, or carbonate buffer (pH 11). High-salt extraction (2 M KCl) is expected to shield charges and weaken ionic interactions which bind peripheral proteins to membranes either directly or indi-

AID

viras

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology

MEMBRANE ASSOCIATION OF INFLUENZA VIRUS M1

FIG. 2. Distribution of M1 in MDCK cells expressing M1 constitutively. To examine the distribution of M1 in MDCK-M1 cells, M1 expression was induced for 17 hr with sodium butyrate. The cells were then labeled with [35S]methionine for 1 hr and subjected to isopycnic gradient analysis. After fractionation and immunoprecipitation with a polyclonal antiinfluenza serum, samples were separated by SDS–PAGE and visualized by fluorography. A shows an influenza virus from cells expressing M1 after induction with sodium butyrate; B, without induction with sodium butyrate. In this case, M1 is not expressed to visible levels. The bands above M1 represent background MDCK bands.

41

workers suggested that the lipid binding domain is located exclusively within the 80 amino acids of the amino terminus. However, both suggestions were based on in vitro studies of cyanogen bromide (Gregoriades and Frangione, 1981) or formic acid and V8 protease (Ye et al., 1987) fragments of M1 that interacted with liposomes. Because discrepancies exist in the reported location of the lipid binding domains, based on in vitro binding of protein fragments, we wanted to examine the lipid binding domains in vivo. To examine these domains, we initially changed the hydrophobicity of the amino acids in the two areas of M1 suggested by Gregoriades and Frangione (1981). If these regions are indeed responsible for lipid binding, then membrane association would be expected to decrease or disappear with the mutant proteins. The hydrophobicity of these regions was altered by substituting the underlined hydrophobic amino acids with glycines (Fig. 4, mutants 2 and 3). Another set of mutants was created by substituting neutral amino acids (Leu 66, Leu 117, Leu 124) with charged amino acids (Lys) (Fig. 4, mutants 5 and 6). Furthermore, a mutant was created by deleting amino acids 115–146 (Fig. 4, mutant 7). The mutated M1 constructs were transiently expressed in HeLa cells. The binding of the mutant proteins to the plasma membrane was examined by immunofluorescence. Although this is not a quantitative method, the binding of the mutants was indistinguishable from wild-type M1 (data not shown). When the mem-

rectly through other membrane proteins. Membrane association mediated by divalent cation bridge formation can be disrupted by the addition of EDTA. Vesicles can also be transformed into membrane sheets when treated with carbonate buffer, which consequently should release soluble or peripheral proteins trapped in vesicles (Fujiki et al., 1982). To determine whether any of these conditions disrupted M1 association with membranes, the samples were subjected to sucrose flotation analysis. None of these conditions released M1 from the membrane fraction (Fig. 3). Thus, although the M1 protein has no transmembrane domain in its primary sequence, the fraction of M1 bound to membranes behaves as an integral protein.

Studies from other groups have demonstrated the interaction of M1 with lipids by reconstituting lipid and purified M1 protein (Bucher et al., 1980; Gregoriades, 1980; Gregoriades and Frangione, 1981; Ye et al., 1987). Gregoriades and Frangione assigned the lipid binding site to two regions of the M1 protein (amino acids 62– 68 and 114–133). In contrast to this finding, Ye and co-

FIG. 3. Stability of influenza M1 protein association with membranes. Total cellular membranes containing associated M1 protein from transfected and [35S]methionine-labeled HeLa cells were prepared by equilibrium density gradients. Membranes were extracted with carbonate buffer (pH 11) for 30 min at 07 or treated with 2 M KCl, 10 mM EDTA, 10 mM Tris–HCl (pH 7.4), or with 50 mM EDTA, Tris–HCl (pH 7.4) for 1 hr at 257. Control membranes were left untreated. Samples were brought to 80% sucrose, and membranes were reisolated in a second round of isopycnic gradient analysis. Fractions were collected, diluted with detergent solution, immunoprecipitated with a polyclonal serum raised against influenza virions, and analyzed by SDS–PAGE. Fractions are numbered from the top to the bottom.

AID

viras

Site-directed mutagenesis of functional domains of the matrix protein

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology

42

KRETZSCHMAR, BUI, AND ROSE

brane-binding capacity of these proteins was examined by isopycnic gradient analysis, all of the mutants still showed membrane binding to the same level as wildtype M1. Two of them are shown as an example (Figs. 5B and 5C). Reexamination of the hydrophobicity plots (Fig. 4) revealed two additional regions which might be responsible for the membrane binding (amino acids 3–12 and 142–147). Therefore, additional M1 mutant proteins were constructed with substitutions in these two new potential lipid-binding domains (Fig. 4, mutants 1 and 4). These mutants also showed the wild-type level of membrane binding (Figs. 5A and 5D). Furthermore, double mutants were constructed combining mutations in these two new potential lipid-binding domains with previously designed mutants: mutants 1 / 3, 1 / 4, 1 / 7, and 5 / 6. All these mutants were transiently expressed in HeLa cells and showed a distribution pattern by immunofluorescence similar to that of the wild-type M1 protein. The mutants were subsequently analyzed by sucrose flotation analysis to examine membrane binding. The results showed that all mutants exhibited the same level of membrane association as that observed with the wild-type matrix protein (data not shown). Mutant matrix proteins interact stably with membranes Although these mutants still associated with cellular membranes, the stability of the membrane binding might be altered. Therefore, HeLa cell membranes containing the mutated proteins were prepared and subjected to high-salt or high-pH treatments. The results for those mutants examined (mutants 1–7) revealed that M1 was still bound tightly to membranes. As an example, the results for two mutants are shown (Fig. 6). DISCUSSION We show here that approximately 15% of the matrix protein M1 of influenza virus WSN binds to cellular membranes in vivo when expressed in the absence or presence of the viral glycoproteins. This fraction of membrane-bound matrix protein was also observed in MDCK cells expressing the matrix protein constitutively. Therefore, the ability of a substantial fraction of M1 protein to bind tightly to membranes is a special feature of the

FIG. 4. Hydrophobicity of M1 proteins with mutations in potential lipid binding domains. Mutants of the M1 protein were constructed by substituting the underlined amino acids with glycine (mutants 1–4), replacing leucine residues with charged amino acids (mutants 5 and 6), or deleting hydrophobic regions C and D (mutant 7). The blackened areas in the hydrophobicity plots represent the hydrophobicity associated with amino acids that were removed or replaced. Hydrophobicity of wild-type M1 protein is shown as well.

AID

VY 7936

/

6a17$$$141

04-30-96 00:53:57

viras

AP: Virology

MEMBRANE ASSOCIATION OF INFLUENZA VIRUS M1

43

FIG. 5. Analysis of membrane-associated mutant matrix protein M1. Mutant M1 proteins were transiently expressed in HeLa cells, metabolically labeled with [35S]methionine, and subjected to sucrose flotation analysis. Fractions were diluted with detergent solution, immunoprecipitated with rabbit anti-influenza serum, and analyzed by SDS–PAGE. A through D show gradients from cells expressing mutant M1 proteins 1, 2, 3, or 4, respectively. Fractions are numbered from the top to the bottom. Quantitation of the amount of labeled protein in each fraction was carried out by scanning densitometry of autoradiographs.

matrix protein M1. Others have already reported M protein binding to artificial phospholipid vesicles (Bucher et al., 1980; Gregoriades and Frangione, 1981). Although a previous study involving analysis of influenza-infected cells suggested that M1 proteins could associate with cellular membranes (Hay, 1974), the question of a requirement for the viral glycoproteins hemagglutinin and neuraminidase remained open. It has been suggested that once M1 and HA and/or NA proteins are localized to the plasma membrane in infected cells, a subsequent M1–HA/NA interaction may induce the clustering of HA and NA proteins and specific host cell glycoproteins into areas active in virus assembly (Choppin et al., 1972). Although this seems plausible, we did not observe enhancement of M1 protein binding to membranes containing HA or NA proteins although the interactions could be weak and unstable to gradient sedimentation. A further possibility might be that M1 needs to be assembled with a nucleocapsid complex in order to associate with the cytoplasmic domain of the glycoproteins. However, recent reports that influenza virus particles containing HA or NA without their cytoplasmic tail could be obtained (Jin et al., 1994; Garcia-Sastre and Palese, 1995) suggested that the interactions of M1 with the cytoplasmic tails may be dispensable for virus assembly. Interestingly, the matrix proteins of a second family of enveloped RNA viruses, the rhabdovirus VSV, and of Sendai virus, a member of a third family, the paramyxoviruses, have also been shown not to increase membrane binding after coexpression of the viral glyco-

protein (Chong and Rose, 1993; Stricker et al., 1994). However, Sanderson et al. (1993) reported that Sendai virus M protein associates with membranes only in the presence of the viral glycoprotein. The reason for this discrepancy is unclear. We also found that the membrane-bound fraction of M1 shows the characteristics of an integral membrane protein rather than those of a peripheral protein. There is no evidence that M1 is covalently modified with lipids, which might explain its strong membrane binding. The stable membrane association demonstrated here suggests that part of M1 may extend into the lipid bilayer, rendering it resistant to conditions that characteristically remove peripherally associated membrane proteins. A membrane-penetrating, cross-linking reagent has labeled M1 proteins in intact virions (Gregoriades and Frangione, 1981), and there is evidence obtained from in vitro studies of cyanogen bromide and formic acid fragments that the M1 protein is at least partly embedded in the viral envelope, probably like VSV matrix protein as an amphipathic helix (Lenard and Vanderoef, 1990; Mancarella and Lenard, 1981; Wiener et al., 1985). Such proteins are believed to introduce strain into the otherwise fluid lipid bilayer and induce lateral curvature, which can promote protein–protein interactions as well as cause membrane rigidity (Gennis, 1989). The other studies described here were designed to identify membrane binding domains in M1 protein. Identification of these domains might provide a clearer understanding of the mechanism by which matrix proteins in

AID

viras

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology

44

KRETZSCHMAR, BUI, AND ROSE

demonstrated membrane binding. Since mutant M1 proteins still bound stably to cellular membranes, the question of the mechanism of binding still remains. A possible explanation for our results is that membrane association does not involve the hydrophobic domains targeted for analysis. Other, amphipathic regions of M1 may fold in such a way that they bind tightly to membranes. Alternatively, the membrane interaction could involve several of the hydrophobic domains, and the remaining domains even in the double mutants might be sufficient for binding. It is clearly difficult to carry out meaningful mutagenesis studies in the absence of a three-dimensional structure of M1 protein. The availability of a structure of M1 will likely be required to guide any future studies. ACKNOWLEDGMENTS We thank all members of the Rose lab, especially Linda Buonocore, for suggestions and encourangement during the course of this work. We also thank Elisa Konieczko and Gary Whittaker for helpul suggestions on the manuscript. This work was supported by a fellowship to E.K. from the Deutsche Forschungsgemeinschaft, to M.B. from the Life & Health Insurance Medical Research Fund, and by Grants AI24345 and GM-07205 from the National Institutes of Health.

REFERENCES

general interact with the lipid bilayer. In vitro binding studies (Gregoriades and Frangione, 1981) suggested that the hydrophobic regions interacting with membranes might be regions B and C (see Fig. 4). However, our mutants with decreased hydrophobicity in these regions (mutants 2, 3, 5, and 6) or completely deleted region C (mutant 7) had unaltered membrane binding capacity. Therefore we conclude that none of the domains examined are essential for the association of M1 protein with membranes in vivo. Furthermore, in the mutants tested so far, none of these regions are required to achieve normal stability of membrane association, as seen with wild-type M1 protein. Even mutants covering hydrophobic regions A and D (mutants 1, 4, and 7) and double mutants eliminating pairs of hydrophobic regions still

Bergmann, J. E., and Fusco, P. J. (1988). The M protein of vesicular virus associates specifically with the basolateral membranes of polarized epithelial cells independently of the G protein. J. Cell Biol. 107, 1707– 1715. Bucher, D. J., Kharitonenkov, I. G., Zakomirdin, J. A., Grigoriev, V. B., Klimenko, S. M., and Davis, J. F. (1980). Incorporation of influenza virus M-protein into liposomes. J. Virol. 36, 586–590. Chong, L. D., and Rose, J. K. (1993). Membrane association of functional vesicular stomatitis virus matrix protein in vivo. J. Virol. 67, 407–414. Chong, L. D., and Rose, J. K. (1994). Interactions of normal and mutant vesicular stomatitis virus matrix proteins with the plasma membrane and nucleocapsids. J. Virol. 68, 441–447. Choppin, P. W., Compans, R. W., Scheid, A., McSharry, J. J., and Lazarowitz, S. G. (1972). Structure and assembly of viral membranes. In ‘‘Membrane Research ’’ (C. F. Fox, Ed.), pp. 163–186. Academic Press, New York. Compans, R. W., and Choppin, P. W. (1975). Reproduction of myxoviruses. In ‘‘Comprehensive Virology’’ (H. Fraenkel-Conrat and R. R. Wagner, Eds.), pp. 179–252. Plenum, New York. Doms, R. W., Helenius, A., and White, J. (1985). Membrane fusion activity of the influenza virus hemagglutinin. The low pH-induced conformational change. J. Biol. Chem. 260, 2973–2981. Dubois-Dalcq, M., Holmes, K. V., and Rentier, B. (1984). ‘‘Assembly of Enveloped RNA Viruses,’’ pp. 1–20. Springer-Verlag, Vienna. Fuerst, T. R., Niles, E. G., Studier, F. W., and Moss, B. (1986). Eukaryotic transient-expression system based on recombinant vaccinia virus that synthesizes bacteriophage T7 RNA polymerase. Proc. Natl. Acad. Sci. USA 83, 8122–8126. Fujiki, Y., Hubbard, A. L., Fowler, S., and Lazarow. (1982). Isolation of intracellular membranes by means of sodium carbonate treatment: Application to endoplasmic reticulum. J. Cell Biol. 93, 97–102. Garcia-Sastre, A., and Palese, P. (1995). The cytoplasmic tail of the neuraminidase protein of influenza A virus does not play an important role in the packaging of this protein into viral envelopes. Virus Res. 37, 37–47.

AID

viras

FIG. 6. Stable interaction of mutant M1 proteins with cellular membranes. Mutants 3 and 4 of the M1 protein (see Fig. 4) were tested for their stability of membrane binding. Therefore, total cellular membranes containing [35S]methionine-labeled mutant M1 protein from transfected HeLa cells were prepared by equilibrium density gradients. Membranes were treated with carbonate buffer (pH 11) for 30 min at 07 or with 2 M KCl for 1 hr at 257. Control samples were left untreated. Samples were brought to 80% sucrose, and membranes were subjected to a second round of sucrose flotation analysis. Fractions were collected, diluted with detergent solution, immunoprecititated with antiinfluenza serum, and analyzed by SDS–PAGE. Fractions are numbered from the top to the bottom. A shows gradients from cells expressing mutant 3 and B a gradient from cells expressing mutant 4.

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology

MEMBRANE ASSOCIATION OF INFLUENZA VIRUS M1

45

Gennis, R. B. (1989). Characterization and structural properties of membrane proteins. In ‘‘Biomembranes’’ (C. R. Cantor, Ed.), pp. 85–137. Springer-Verlag, New York. Gregoriades, A. (1980). Interaction of influenza M protein with viral lipid and phosphatidylcholine vesicles. J. Virol. 36, 470–479. Gregoriades, A., and Frangione, B. (1981). Insertion of influenza M protein into the viral lipid bilayer and localization of site of insertion. J. Virol. 40, 323–328. Hay, A. J. (1974). Studies on the formation of the influenza virus envelope. Virology 60, 398–418. Hay, A. J., and Skehel, J. J. (1975). Studies on the synthesis of influenza virus proteins. In ‘‘Negative Strand Viruses’’ (B. W. J. Mahy and R. D. Barry, Eds.), pp. 635–655. Academic Press, London. Hunziker, W., Harter, C., Matter, K., and Mellman, I. (1991). Basolateral sorting in MDCK cells requires a distinct cytoplasmic domain determinant. Cell 66, 907–920. Jin, H., Leser, G. P., and Lamb, R. A. (1994). The influenza virus hemagglutinin cytoplasmic tail is not essential for virus assembly or infectivity. EMBO J. 13, 5504–5515. Lamb, R. A. (1989). Genes and proteins of the influenza viruses. In ‘‘The Influenza Viruses’’ (R. M. Krug, Ed.), pp. 1–87. Plenum, New York. Lenard, J., and Vanderoef, R. (1990). Localization of the membraneassociated region of vesicular stomatitis virus M protein at the N terminus, using the hydrophobic, photoreactive probe 125I-TID. J. Virol. 64, 3486–3491. Lopez, S., Yao, J. S., Kuhn, R. J., Strauss, E. G., and Strauss, J. H. (1994). Nucleocapsid–glycoprotein interactions required for assembly of alphaviruses. J. Virol. 68, 1316–1323. Mancarella, D. A., and Lenard, J. (1981). Interactions of wild type and mutant M protein of vesicular stomatitis virus with viral nucleocapsid and envelope in intact virions. Evidence from [125I]azide labeling and specific cross-linking. Biochemistry 20, 6872–6877. Nakamye, K., and Eckstein, F. (1986). Inhibition of restriction endonuclease NciI cleavage by phophorothioate groups and its application to oligonucleotide-directed mutagenesis. Nucleic Acids Res. 14, 9679– 9698.

Rose, J. K., and Bergmann, J. E. (1983). Altered cytoplasmic domains affect introcellular transport of the vesicular stomatitis virus glycoprotein. Cell 34, 513–524. Rose, J. K., Buonocore, L., and Whitt, M. A. (1991). A new cationic liposome reagent mediating nearly quantitative transfection of animal cells. BioTechniques 10, 520–525. Sanderson, C. M., McQueen, N. L., and Nayak, D. P. (1993). Sendai virus assembly: M protein binds to viral glycoproteins in transit through the secretory pathway. J. Virol. 67, 651–663. Stricker, R., Mottet, G., and Roux, L. (1994). The Sendai virus matrix protein appears to be recruited in the cytoplasm by the viral nucleocapsid to function in viral assembly and budding. J. Gen. Virol. 75, 1031–1042. Suomalainen, M., Liljestrom, P., and Garoff, H. (1992). Spike protein– nucleocapsid interactions drive the budding of alphaviruses. J. Virol. 66, 4737–4747. Whitt, M. A., Chong, L., and Rose, J. K. (1989). Glycoprotein cytoplasmic domain sequences required for rescue of a vesicular stomatitis virus glycoprotein mutant. J. Virol. 63, 3569–3578. Wiener, J. R., Pal, Y., Barenholz, Y., and Wagner, R. R. (1985). Effect of the vesicular stomatitis virus matrix protein on the lateral organization of lipid bilayers containing phosphatidylglycerol: Use of fluorecent phospholipid analogues. Biochemistry 24, 2162–2170. Wigler, M., Pellicer, A., Silverstein, S., Axel, R., Urlaub, G., and Chasin, L. (1979). DNA-mediated transfer of the adenine phosphoribosyltransferase locus into mammalian cells. Proc. Natl. Acad. Sci. USA 76, 1373–1376. Yasuda, J., Nakada, S., Kato, A., Toyoda, T., and Ishihama, A. (1993). Molecular assembly of influenza virus: association of the NS2 protein with virion matrix. Virology 196, 249–255. Ye, Z., Baylor, N. W., and Wagner, R. R. (1989). Transcription-inhibition and RNA-binding domains of influenza A virus matrix protein mapped with anti-idiotypic antibodies and synthetic peptides. J. Virol. 63, 3586–3594. Ye, Z. P., Pal, R., Fox, J. W., and Wagner, R. R. (1987). Functional and antigenic domains of the matrix (M1) protein of influenza A virus. J. Virol. 61, 239–246.

AID

viras

VY 7936

/

6a17$$$141

04-30-96 00:53:57

AP: Virology