Characterization of Sendai virus M protein mutants that can partially interfere with virus particle production

The family Paramyxoviridae is characterized by the negative, single-stranded nature of its RNA genome and by the bilipidic envelope carried by the virus particle. The virus particle, which is likely to adopt a spherical shape in solution (W. Chiu and D. Kolakofsky, personal communication), is nevertheless pleomorphic in size, possibly because it contains more than one genome unit. The functional genome is composed of the RNA, tightly wrapped by the N protein into a left-handed helical nucleocapsid structure, with which the L and P proteins, which form the RNA polymerase, are also associated. Structural and functional analysis has shown that, for Sendai virus (SeV), one turn of the helix encompasses thirteen N protein molecules and that each N protein is likely to contact six nucleotides (Egelman et al., 1989 ; Calain & Roux, 1993 ). The nucleocapsid is packaged into the virus particle delineated by a lipid bilayer derived from the cell plasma membrane, with which three viral proteins, M, HN and F, are associated.

F and HN are trans-membrane glycoproteins of types I and II, respectively. They protrude from the viral envelope as homotrimers and homotetramers, respectively, and form the spikes that decorate the virions, as seen by electron microscopy. HN contains the cell receptor- binding site (neuraminic acid), as well as the neuraminidase activity, involved in detachment of the virion. F exhibits the fusogenic activity required in the process of infection, when the viral envelope fuses with the cell plasma membrane to deliver the viral genome into the cytosol (for a recent review of the Paramyxoviridae cell cycle, see Lamb & Kolakofsky, 1996 ).

The M protein (matrix protein) is so named because it can maintain the compact structure of the virion even after dissolution of the envelope by detergent. This structure falls apart only after high-salt treatment. The SeV M protein contains 348 amino acids (Blumberg et al., 1984 ). It is highly basic (pI 10·17), is partially phosphorylated on Ser⁷⁰ (Sakaguchi et al. , 1997 ) and is acetylated on Ala² after cleavage of the initiating Met (Blumberg et al., 1984 ). It is predicted to contain little secondary structure, namely 11% α- helix, 19% ß-pleated sheet and 70% random coil (Giuffre et al. , 1982 ). It is found in infected cells and in virions in two isoforms defined by their reactivity for monoclonal antibodies (MAbs) (de Melo et al., 1992 ). The minor isoform forms a subset (~30%) characterized by the maturation of a native epitope. SeV M, and presumably the M proteins of all paramyxoviruses (Peeples, 1991 , and references therein), has been shown to associate specifically with membrane fractions in flotation gradients (Stricker et al., 1994 ). As is the case for M from other paramyxoviruses, SeV M is proposed to interact with at least one of the two viral glycoproteins (Yoshida et al., 1986 ; Sanderson et al., 1993 , 1994 , 1995 ). M also interacts with other M molecules, as evidenced by the formation of a crystalline array visualized by electron microscopy either in infected cells (Büechi & B ächi, 1982 ) or after purification (Hewitt, 1977 ; Hewitt & Nermut, 1977 ; Heggeness et al., 1982 ). Initially described as an aggregation, this interaction appears in fact to be a specific two-dimensional polymerization, as shown for vesicular stomatitis virus M (Gaudin et al., 1995 ). Finally, M interacts with the nucleocapsid as evidenced by MN cross- linking (Markwell & Fox, 1980 ) and by the co-capping of the nucleocapsid with the glycoproteins only in the presence of M (Yoshida et al., 1986 ). Because of these multiple interactions, M has been proposed as the `bandleader' in charge of virion assembly at the site of virus-particle budding (Peeples, 1991 ).

The assembly of virions of members of the Paramyxoviridae is generally proposed to take place at the plasma membrane, as the pre- budding structure is seen there before the newly formed virus particle pinches off. The two viral glycoproteins HN (or H) and F are expressed at the cell surface, anchored by their trans-membrane portion, with the large ectodomain on the outside and the short cytoplasmic tails on the inside (Nagai et al., 1975 ). At this stage, the two glycoproteins are thought to move freely in the lateral plane of the membrane (Nagai et al., 1976 ; Markwell & Fox, 1980 ). M forms a paracrystalline array at the inner face of the lipid bilayer (Bächi, 1980 ; Büechi & Bächi, 1982 ), which now concentrates the two glycoproteins via interactions with the cytoplasmic tail or the transmembrane portion of the glycoproteins. Concentration of these virus components is apparently achieved to the exclusion of the cellular surface proteins. The virally modified portion of the membrane now constitutes a recognition site for the viral nucleocapsid, which is found adhering to the M crystalline array (Büechi & Bächi, 1982 ). At this stage, addition of antibodies against the viral glycoproteins to the infected cells leads to a co-capping of all the virus components (Tyrrell & Ehrnst, 1979 ). This assembly model implies that M and the nucleocapsid travel to the plasma membrane before they can interact with the glycoproteins, which are transported via the default secretory pathway. Not much is known about the mechanism of transport of M and the nucleocapsids. The cytoskeleton may be involved here, as suggested by an affinity of M for actin (Giuffre et al. , 1982 ) and by the absence of co-capping of the nucleocapsid and M with the glycoproteins upon treatment with cytochalasin B. Also, the reduced transport of M to the cell surface upon monensin treatment has led to the proposal that M is transported to the plasma membrane via interactions with the glycoprotein cytoplasmic tail (Sanderson et al., 1994 ; for general reviews on the assembly and budding of paramyxoviruses and also on the M protein, see Dubois-Dalcq et al., 1984 ; Peeples, 1991 ; Ray et al., 1991 ).

With the availability of a reverse-genetic system for paramyxoviruses (reviewed in Conzelmann, 1996 ; Boyer & Haenni, 1994 ) came the possibility of creating site-directed mutants to analyse the structurefunction relationships of the viral proteins. For this purpose, we have developed a system that allows the expression of mutated viral proteins in the context of an otherwise normal SeV infection. This is based on the preparation of mixed SeV stocks containing, in addition to the non-defective (ND) viral genome, a defective genome that is derived from a natural, transcribing, defective-interfering (DI) genome, designed to replicate at a high level so as to express extra proteins very efficiently (Mottet et al., 1996 ). In this way, the SeV M protein, tagged with an influenza virus haemagglutinin (HA) protein epitope (HAM), was expressed during SeV infection along with wild-type M (M^wt). Next, mutant HAM proteins were generated that exhibited different properties with regard to their abilities to be incorporated into virus particles and to interfere with virion production (Mottet et al., 1996 ). The present study extends our mutant collection and characterizes some of them in more detail. Our original identification of an amphiphilic α-helix that was proposed to be responsible for M attachment to the membrane has been revised. Attempts to identify the step in assembly at which the mutant M proteins interfere with the normal process of assembly are presented.

SeV expression vector and construction of HAM mutants.
The cloning of the natural SeV transcribing DI RNA (E307) in pSP65 under the control of the T7 RNA polymerase promoter was described by Engelhorn et al. (1993) . Modification of this construct to generate an efficient SeV replication/transcription vector (Svec CB119) was described in detail by Mottet et al. (1996) . In the present work, this vector was modified by shortening the antigenomic promoter substitution (48 nucleotides in Svec CB119) to 36 nucleotides in the genomic promoter, creating Svec-36- CB119. This modification increases the efficiency of transcription from the vector without destroying its replication property. The various HAM mutants were obtained by appropriate modifications of the HA- M-CB119 plasmid, by using the fusion PCR methodology described previously (Mottet et al., 1996 ). The substitution at residue Ser⁷⁰ of the HAM mutants was introduced by preparing a fusion PCR product overlapping the Ser⁷⁰ codon and extending between two unique sites (SmaI/StuI) of the HAM gene as found in HA-M-CB119 (Mottet et al., 1996 ). At the Ser⁷⁰ position, the AGC codon was replaced by G(C/A)C, allowing the selection of clones carrying GCC (Ala) or GAC (Asp) substitutions. The SmaI/StuI fragments that originated from these clones were then sub-cloned into the corresponding HA-M³⁰-CB119, and the substituted HAM ^wt and HAM³⁰ genes were finally transferred into the Svec-36-CB119 vector (see above). The modified regions of the plasmids were sequenced extensively to confirm the various modifications.

Mixed SeV stocks.
Mixed SeV stocks refer to stocks containing DI genomes, supported in their replication and assembly by functions provided by the ND genome, in addition to the ND viral genome (Roux & Holland, 1979 ). In the present case, the DI genome consisted of a nucleocapsid containing the RNA vector (HA-M-CB119) carrying the open reading frame of the various M mutants. The methodology to generate these mixed SeV stocks has been described previously in detail (Calain & Roux, 1993 ; Mottet et al., 1996 ) and was not changed. The HAM virus stocks were characterized by the number of p.f.u. (in general ~10⁸ p.f.u./ml; Sugita et al., 1974 ) and by the ratio of DI vector to ND RNA upon infection (generally 20:1) estimated by Northern blot analysis as described previously (Calain & Roux, 1993 ). In cells infected with these mixed virus stocks, the HAM protein expressed from the vector RNA co-exists with M^wt expressed from the ND genome. In general, due to the high replication efficiency of the RNA vector, which carries a copy-back feature, the HAM protein was found in excess over M^wt.

Antibodies and protein analysis.
The anti-M protein antibody preparations used were a rabbit serum raised against SDS-denatured M protein (α-M_SDS) (Mottet et al., 1986 ; Tuffereau & Roux, 1988 ) and MAbs 383 and 376 obtained from Claes Örvell (Laboratory of Clinical Virology, Huddinge Hospital, Huddinge, Sweden). The anti-HA epitope MAb (12CA5; referred to here as α-HA) (Field et al., 1988 ) was obtained from Berkeley Antibody Co. A rabbit polyclonal serum raised against the whole purified virus (Rab-vir) was described by Mottet et al. (1986) . Protein-isotope labelling (TransLabel, New England Nuclear), preparation of cellular extracts for immunoprecipitation (IP), IP and SDSPAGE were performed as described previously (Mottet et al., 1996 ). Processing of the gels through PPO-DMSO enhancement or through Western blotting were exactly as described previously (Mottet et al., 1996 ), except that blots were developed with the ECL system (Amersham). Virus particles in clarified cell supernatants were pelleted through a 25% glycerolTNE cushion (10 mM TrisHCl, 1 mM EDTA, 50 mM NaCl) for 2 h at 50000 r.p.m. before SDSPAGE.

M protein co-precipitation.
Infected BHK cells were disrupted into Heggeness in vitro assembly buffer (Heggeness et al., 1982 ) (9·2 mM phosphate buffer, pH 7·2, 0·7% sodium deoxycholate, 2·14 mM Triton X-100, 2 mM PMSF and 1% aprotinin with or without 1 M NaCl) and incubated for 18 h at room temperature. After sonication for 10 s (Branson Sonic Sonifer B-12, lowest speed), cell extracts were spun for 45 s at 12000 r.p.m. in a microfuge. The supernatants were then processed for IP by using the anti-M MAb 383 or α-HA antibody. The immunoprecipitates were finally analysed by Western blotting and probed with MAb 383 (diluted 1:25000).

Flotation gradients.
Infected BHK cells were harvested 18 h post-infection and resuspended in 10% sucrose containing 2 mM PMSF and 1% aprotinin. After Dounce homogenization (20 strokes), the cell debris was pelleted (2 min, 1000 r.p.m., 4 °C) and the supernatants (1·2 ml), made to a final sucrose concentration of 70%, were loaded at the bottom of SW60 tubes (Beckman) and overlaid with 2·5 ml 65% sucrose and 0·8 ml 10% sucrose. After centrifugation to equilibrium (40000 r.p.m., 18 h, 4 °C), ten 500 µl fractions were collected, diluted to 1·5 ml with IP buffer and processed for IP with the Rab-vir rabbit serum. The immunoprecipitates were then analysed by Western blotting and probed with α-HA and MAb 383. Alternatively, the fractions were diluted to 4 ml with TE (20 mM TrisHCl, pH 7·5, 2 mM EDTA) and centrifuged for 2 h at 4 °C and 55000 r.p.m. in an SW60 rotor (Beckman) to pellet the nucleocapsids. These were resuspended in TE plus 0·2% Sarkosyl, extracted with phenolchloroform to isolate the nucleocapsid RNA and recovered by ethanol precipitation. The RNAs were then processed by agarose gel electrophoresis and Northern blotting as described previously (Mottet et al., 1990 ; Stricker et al., 1994 ). The probe was ³²P- labelled anti-M transcript synthesized in vitro from M-pGem4 (de Melo et al., 1992 ).

Immunofluorescent staining and confocal microscopy.
LLCMK2 cells were plated on poly-l-lysine-coated coverslips (coated with a 5 µg/ml solution for 15 min and dried under UV). After growth overnight, cells were infected and fixed 18 h later with 4% paraformaldehyde for 30 min at room temperature. After three washes with PBS lacking Ca²⁺ and Mg²⁺ (PBS ^-, 15 min at room temperature), cells were incubated with the primary antibodies, α-HA and α-M_SDS (diluted 1200 and 1100, respectively, in 3% BSA and 0·3% Triton X-100), for 2 h at room temperature. After three more washes with PBS^-, rhodamine-conjugated anti-mouse antibodies and FITC-conjugated anti-rabbit antibodies were added and the cells were incubated for 2 h at 4 °C in the dark. The procedure was completed by three final PBS^- washes and the coverslips were mounted in 20% glycerol plus 0·2 g/ml polyvinyl alcohol. Specimens were observed with a Zeiss confocal laser- scan fluorescent inverted microscope (LSM 410, Carl Zeiss) (for a detailed description of the confocal microscope see Tuchweber et al. , 1996 ) through an oil plan-neofluar 40x/1·3 objective. Images of 512x512 pixels, stored on an erasable optical disk (Sony), were imported into Micrograph^X Designer and printed at 1420 d.p.i. on glossy paper on an Epson Stylus Colour 800 printer.

Value of the α-helix prediction for amino acids 104119 for M protein function
In our previous study, a region encompassing amino acids 104119 of the SeV M protein was proposed to form an amphiphilic α-helix involved in binding of M protein to the bilipidic membrane. This conclusion was drawn from the observation that substitution of Val¹¹³, located on the hydrophobic face of the predicted α-helix, with a negatively charged Glu (Table 1 , mutants 11 and 30) resulted in decreased binding to membrane fractions, as measured by flotation in sucrose gradients. This phenotype was judged to be relevant because it correlated, at least in these mutants, with the absence of incorporation into virus particles when the protein was expressed in an SeV stock (Mottet et al., 1996 ). This interpretation, however, did not hold for mutant 21 (Thr¹¹² Lys), which exhibited binding affinity for the membrane that was even higher than that of wild-type, an increased binding that correlated with even better uptake into virus particles, despite the introduction of a positive charge. Other M mutants were then generated and expressed from the SeV DI vector in the context of a mixed virus infection, i.e. together with all the SeV proteins, including M^wt. Table 1 shows a summary of all the available mutants carrying substitutions in positions 112 or 113. The following observations can be made. (i) A positive or a negative charge in the putative hydrophobic face of the α-helix does not prevent HAM uptake into virus particles when introduced at position 112, i.e. replacing Thr¹¹² (mutants 20, 21 and 47). (ii) The replacement of Val¹¹³ with equivalent uncharged residues such as Leu or Ala (mutants 2 and 8) does not preserve HAM uptake. It therefore appears that the value of the predicted amphiphilic α-helix for M protein function, i.e. its binding to the membrane leading to incorporation into virions, cannot be verified experimentally.

Table 1. HAM mutants and their relevant properties

From the data in Table 1, it is noteworthy that the eleven HAM mutants that were not incorporated into virus particles had another residue in place of Val¹¹³ (from mutant 57 down), which suggests a correlation between the substitution per se and M dysfunction. Alignment of 13 M protein sequences from members of theParamyxoviridae demonstrates that SeV Val¹¹³ is part of a stretch of four very well-conserved residues, ValArg(Lys)Arg(Lys)Thr, in which Val is found naturally replaced only by Cys (in Newcastle disease virus; Fig. 1 ), a substitution accepted in SeV M (Table 1, mutant 51). It therefore appears that the data shown in Table 1 reflect the critical need for the integrity of an essential motif, the significance of which is unknown.

(45K): Fig. 1. Partial M protein sequence comparison. The M protein alignment in the region encompassing the putative α-helix (shaded box; residues 104119 in SeV) emphasizes the conservation of the tetrapeptide VRRT (in bold; residues 113116 in SeV) at an almost identical position in all the M proteins, as indicated by the residue numbering at the right. HPIV, Human parainfluenza virus; BPIV, bovine parainfluenza virus; RPV, rinderpest virus; DDV, dolphin distemper virus; CDV, canine distemper virus; NDV, Newcastle disease virus; SV41, simian parainfluenza virus 41; SV5, simian parainfluenza virus 5. Sequences were obtained from GenBank.

The M protein can be separated into two species by SDSPAGE, with the slower migrating species representing the phosphorylated form (Lamb & Choppin, 1977 ). The doublet patterns of the HAM derivatives have been described previously (Mottet et al. , 1996 ), with the peculiarity that only the fully competent HAM proteins exhibit this PAGE migration profile. In contrast, the HAM mutants that are excluded from virus particles migrate as single bands corresponding to the slower migrating form, the upper band (see Table 1 for a summary). Fig. 2 (a) illustrates this observation. HAM^wt and HAM²¹ clearly separate into doublets, as does M^wt, while HAM ⁵ and HAM³⁰ migrate as single upper bands. The migration profile of the HAM proteins is clearly tightly dependent on the residue at position 113. It was therefore relevant to investigate the reason for this observation. It is noteworthy that, although the doublet migration pattern was always consistent, it showed up only when PAGE resolution was optimal, which cannot always be achieved.

Table 1 ). MK: marker lane containing Mwt and HAMwt proteins obtained by immunoprecipitation. Bands above and below the M proteins correspond to the IgG heavy and light chains used during the IP step. (b) BHK cells were infected with mixed HAMwt or HAM30 SeV stocks, in which the expressed HAM proteins contained further substitutions at Ser70 as indicated (Ala or Asp). At 48 h post-infection, the virus particles in the supernatants were purified as indicated in Methods, analysed directly by PAGE, processed by Western blotting and probed with the anti-M MAb 383.

Val¹¹³ substitution results in a conformational change of the M protein
Firstly, the possibility that substitution of Val¹¹³ could induce complete phosphorylation of HAM was investigated. This would be consistent with exclusion from virus particles, since phosphorylated SeV M has been proposed to be incorporated poorly into virions (Lamb & Choppin, 1977 ). SeV M is phosphorylated only on Ser⁷⁰ (Sakaguchi et al., 1997 ), so this residue was mutated into Ala or Asp. Ala prevents phosphorylation and Asp introduces a negative charge at position 70, analogous to a phosphate residue. In HA-M^wt, the Ser⁷⁰ Ala substitution resulted in a single, lower-migrating band (Fig. 2b, lane 1), confirming the phosphorylated nature of the upper band. Also consistent was the single, upper- migrating band revealed by HAM^wt with a Ser⁷⁰ Asp substitution (Fig. 2b, lane 2). On the other hand, the Ser⁷⁰Ala substitution did not alter the migration profile of HAM³⁰ (not shown). Note that none of the HAM³⁰ mutants were present in virus particles (Fig. 2b, lanes 46), whether or not they carried a negative charge. Overall, these data support the conclusion that the exclusion of the HAM mutants from virions is unlikely to reflect phosphorylation at position 70. Attempts to dephosphorylate HAM³⁰ failed (not shown). Finally, in contrast to previous reports (Lamb & Choppin, 1977 ), phosphorylation of M (or the incorporation of a negative charge) at position 70 did not prevent protein incorporation into virions in the cells used here (Fig. 2b, lanes 2 and 3). This result therefore argues against the participation of phosphorylation in exclusion from virus particles.

Secondly, the possibility that the single-band PAGE migration could result from a conformational change induced by the Val¹¹³ substitution was investigated by monitoring the maturation of the minor native epitope on the M protein, which matures with time (de Melo et al., 1992 ; see Introduction). In contrast to the situation for M and HAM^wt, maturation of the native epitope of HAM³⁰ could not be demonstrated (not shown), a result that supports an alteration in the folding of the protein induced by the substitutions in HAM³⁰.

A drastic change in the tertiary structure of HAM³⁰ was confirmed by comparison of its tryptic digestion pattern with that of HAM^wt (Fig. 3 ). Here, only fragments including the conserved N terminus were detected since the Western blot was probed with α-HA, reacting with the HA epitope at the N terminus of M. The pattern appears different in the number of discrete bands, their relative position and intensity. The difference in the relative positions of some bands (for instance, X1/X2 and Y1/Y2) could be explained by the conservation (Y1) or the loss (Y2) of the substitution at position 113 being responsible for the PAGE migration change. However, this explanation would not hold for the difference between the full-length proteins and their respective larger digest fragments (X1 and Y1). Neither could the extra band (Y3) and the relative intensity of the others (X1/Y1 or X2/Y2) be explained in this way. In conclusion, the substitution at Val¹¹³ appears to induce a general conformational change in HAM that correlates with its dysfunction.

(51K): Fig. 3. Tryptic peptide analysis of HAMwt and HAM30. BHK cells infected with mixed HAM wt or HAM30 SeV stocks were disrupted in lysing buffer I and cytoplasmic extracts were submitted to acetylated trypsin digestion (10 µg/ml final concentration, 37 °C) as indicated. At the end of the incubation the samples were immediately boiled in the PAGE sample buffer and frozen until all the samples were collected. The samples were finally analysed by PAGE in parallel and processed through Western blotting and probed with the α-HA MAb. Numbers in white on the gels are computer-determined distances between the bands indicated.

From this point on, HAM³⁰ will be taken as the prototype for all the non-functional HAM proteins, since it exhibited the strongest interference with virion production (Mottet et al., 1996 ). Generalizations therefore assume that any mutant with a Val¹¹³ substitution will exhibit the same behaviour. We realize that this decision, although based on the partial characterization of other mutants, may be an oversimplification.

HAM³⁰ does not co-precipitate M, but binds to membrane fractions
M function has been defined as the ability to be incorporated into virions. This, however, implies a series of steps through which the protein has to progress before ending up in virus particles. Exclusion from virus particles may thus reflect a blockage of any one of the steps, the identification of which could be informative about the path followed.

Self-assembly is one of the steps through which M passes, and this was investigated by selective immunoprecipitation. Extracts of cells infected with mixed virus stocks expressing M^wt as well as HAM^wt or HAM³⁰ were immunoprecipitated with α-HA or α-M antibodies. The proteins selected were then analysed by Western blots probed with α-M antibody (Fig. 4 ). In a mixed infection expressing HAM^wt (Fig. 4a, lane 2), immunoprecipitation with α-HA resulted in the recovery of M^wt together with HAM^wt, reflecting an association between the two proteins. This association was quantitative, as shown by the equivalent recovery of M^wt with α-M (Fig. 4a , lane 1), which precipitates both proteins directly. Moreover, M^wt did not co-precipitate with HAM^wt in high salt (Fig. 4a, lane 4), conditions known to disassemble M^wt (Heggeness et al., 1982 ; de Melo et al., 1992 ). In contrast, no co-precipitation of M^wt with HAM³⁰ was apparent with the α-HA antibody (Fig. 4b , lane 2). This shows that HAM³⁰ did not associate with M^wt in the way that HAM^wt did, an indication that it has lost the ability to form at the least a homodimer.

(59K): Fig. 4. Co-precipitation of M and HAM. BHK cells infected with mixed HAMwt or HAM30 SeV stocks were harvested at 18 h post-infection and disrupted in Heggeness buffer, under conditions designed to protect (-NaCl) or to disrupt (+NaCl) possible protein associations (see Methods). Immunoprecipitation was performed by using either MAb 383 (α-M) or α-HA antibodies. Precipitated proteins were processed through Western blotting and probed with MAb 383, detecting all the different forms of M and HAM. hc-Ab and lc-Ab refer to the antibody heavy and light chains used in the IP step.

HAM³⁰ expressed alone from pGem4 exhibited reduced binding to membrane fractions (Mottet et al., 1996 ). In the context of a mixed virus infection, however, it was found associated with the membrane fractions to the same extent and equally as exclusively as HAM^wt or M^wt (not shown). Therefore, although HAM³⁰ has decreased M- binding ability, it retains its membrane binding ability in the presence of the other viral components.

HAM³⁰ reduces nucleocapsidmembrane association
We showed previously that more than 70% of the full-length (ND) nucleocapsids was found in association with the membrane fractions (Stricker et al., 1994 ). In contrast, defective (DI) nucleocapsids were found mainly (>>75%) in the free fractions. This result correlated with the efficiency with which each category of nucleocapsids was found in virus particles, so that the degree of membrane association was considered relevant to the process of virus assembly. The difference in association between the two populations of nucleocapsids was interpreted as reflecting competition for the M protein, known to be a limiting factor in virion production (Kingsbury, 1974 ). These earlier experiments were performed with mixed virus stocks composed of naturally derived DI genomes encoding no extra viral proteins. In Fig. 5 , the degree of ND and DI nucleocapsid association was analysed in mixed virus infections where the DI genomes encoded M proteins in excess, either functional (HAM^wt and HAM²¹) or non-functional (HAM³⁰ ). HAM²¹ was included here because it appeared to perform even better than HAM^wt (Mottet et al. , 1996 ). As before, the majority of ND nucleocapsids (88%) were membrane-bound in the absence of DI genomes (Fig. 5b , lanes 1 and 2). With HAM^wt expression, i.e. in the presence of DI genomes as well (Fig. 5b , lanes 3 and 4), the fraction of bound ND nucleocapsids appeared to be slightly reduced, possibly due to competition with DI genomes in a situation where M was less limiting, since it was expressed from both ND and DI genomes. DI nucleocapsids, however, were still mostly membrane-free (66%). With HAM²¹ expression, competition between ND and DI nucleocapsids appears to shift even more in favour of the latter (Fig. 5b , lanes 5 and 6). With HAM³⁰, however, the trend was altered (Fig. 5b, lanes 7 and 8). ND nucleocapsid binding still decreased, but now DI nucleocapsid binding was also drastically reduced, with the net result that a clear reduction of total nucleocapsid association with membrane fractions was observed, a trend that was verified in three separate experiments (not shown). This property of HAM³⁰ correlates with a decrease in total virus production noted before in mixed virus infections where non-functional M mutants were expressed (HAM ⁵, HAM¹¹, HAM³⁰; Mottet et al., 1996 ). It is unfortunate that ND genomes could not be visualized because of high replication interference by DI genomes, although the amounts could be scored over the background by phosphorimager scanning. Adding excess ND virus contributed to an increase in the ND:DI ratio, so that ND genomes were visible. Unfortunately, under these conditions, the M^wt:HAM ³⁰ ratio also increased such that the interference by HAM³⁰ on nucleocapsid binding was lost (not shown).

(46K): Fig. 5. Nucleocapsidmembrane association. Extracts from BHK cells infected with mixed HAMwt, HAM 21 or HAM30 SeV stocks were centrifuged in flotation gradients. The fractions collected (a) were pooled into free (Lower, L, fractions 15) or membrane-associated (Upper, U, fractions 610) material. (b) Nucleocapsids in the pooled fractions were recovered and the nucleocapsid RNA was processed through a Northern blot probed with an M riboprobe to evaluate the extent of nucleocapsid membrane association. The Northern blot was analysed by a phosphorimager (Molecular Dynamics) to obtain precise quantification of the signals in the U and L pools. ND RNAs were not always visible, but their radioactive signals were detected significantly over the background by the phosphorimager. DI RNAs expressed the HAM proteins.

HAM³⁰ is found poorly associated with the cell plasma membrane
The intracellular localization of the M proteins was investigated by immunofluorescent staining. Two antibody preparations were used. The first, α-M_SDS, is a rabbit serum raised against the SDS-denatured protein (Tuffereau & Roux, 1988 ), an antibody preparation that recognizes denatured, fixed or native SeV M of all kinds. The second is α-HA, which reacts with the HAM proteins only. Mixed-virus-infected cells were fixed and reacted first with the α-HA antibody. In a second step, the α-M_SDS preparation was added, before a final incubation with a mixture of rhodamine-coupled anti-mouse and FITC- coupled anti-rabbit antibodies (Fig. 6 ). The green staining (upper panels) shows the total M protein distribution, i.e. M^wt and HAM. In the HAM^wt-expressing cells, the green staining was rather diffuse, but nevertheless showed a sharp rim at the edges of the cells. In the HAM³⁰-expressing cells, the green staining appeared to be more central, growing fainter towards the cell edges. The red staining (middle panels) shows the contribution of the HAM proteins to the total M protein distribution. HAM^wt was clearly spread all over the cytoplasm, but contributed also to most of the sharp rim at the cell edges. In contrast, HAM³⁰ was mainly concentrated around the nucleus, showing that the faint green staining at the cell edge seen above was due to M^wt. Superimposition of the two staining patterns confirmed the presence of HAM^wt at the cell edges, where HAM³⁰ was not found. In conclusion, HAM³⁰ appears not to be efficiently transported to the plasma membrane, where virus particle budding takes place.

(61K): Fig. 6. Cellular localization of HAMwt and HAM30 proteins. LLCMK2 cells were infected on coverslips with mixed HAMwt or HAM30 SeV stocks. At 24 h post-infection, the cells were fixed and stained with α-MSDS and α-HA. The cells were then reacted with FITC-coupled anti-rabbit IgG and rhodamine-coupled anti-mouse IgG. Green fluorescence indicates both Mwt and HAM, while red fluorescence is specific for HAM.

In contrast to our earlier proposal (Mottet et al., 1996 ), amino acids 104119 of SeV M appear not to constitute an amphiphilic α-helix involved in membrane binding. Rather, it is the substitution of Val¹¹³, a residue that is part of a highly conserved VRRT tetrapeptide, which is responsible for the exclusion of the M mutants from virions. The valine substitution results in a conformational change that is likely to be responsible for the mutant protein phenotype. The involvement of phosphorylation in causing this phenotype was not supported experimentally, and this is consistent with the finding that the suppression of phosphorylation appears not to affect M protein function (Sakaguchi et al., 1997 ).

In the context of SeV infection, HAM³⁰ is shown here (i) to bind to total membrane fractions, (ii) not to co- precipitate M^wt, which indicates its inability to self- assemble, (iii) to reduce nucleocapsidmembrane association and (iv) not to be transported to the plasma membrane. These phenotypic traits can help to define the path that M follows en route to the virus particle. HAM³⁰ is able to attach to the total membrane fraction, with an apparent inability to form a homodimer. Attachment to the membrane could therefore be achieved by individual M proteins, which could then self-assemble once bound to membrane. When expressed alone from pGem plasmids, about 50 and 20%, respectively, of M ^wt and HAM³⁰ protein was bound to membrane (Mottet et al., 1996 ). For both proteins, this percentage increased to almost 100% in SeV infection (not shown). Co- operation with the other viral components could therefore drive M towards a more extensive association. Co-expression of F and M was reported to reinforce M membrane association and to promote its diffusion from a perinuclear location towards the cell edge (Sanderson et al., 1994 ). Although the participation of F in the binding of M to the membrane is a matter of debate (Stricker et al., 1994 ), the participation of other viral proteins in a more extensive membrane association and/or in the transport of M to the cell membrane is a valid assumption. As M can polymerize in vitro in the absence of other viral proteins in a concentration- dependent manner (Gaudin et al., 1995 ; Heggeness et al., 1982 ), attachment to the membrane may serve to increase the local M concentration, so as to prime self- association. If HAM³⁰, which cannot interact with M ^wt, binds efficiently to the membrane, it could interfere with self-assembly of M^wt, disturbing formation of the leaflet structure. This would in turn result in the observed decrease in virus budding (Mottet et al., 1996 ).

If this scheme is valid, and it does not contradict what has been proposed up to now, an intriguing observation comes from the ability of HAM³⁰ to interfere with nucleocapsid binding to the total membrane fraction (Fig. 5) even though HAM ³⁰ is not efficiently transported to the cell plasma membrane (Fig. 6). This suggests that the interaction of nucleocapsids with the membrane could take place in the perinuclear region, where HAM³⁰ is located (Fig. 6 ). The proposal that partial or even full assembly could take place on intra-cytoplasmic membranes is not new. Ö rvell & Grandien (1982) have already shown by immunofluorescent staining studies that the majority of SeV M never reaches the plasma membrane but rather remains associated with the ER membrane. The same observation has been made by Bellini et al. (cited in Dubois-Dalcq et al., 1984 ) for the measles virus M protein. Comparison of the budding of SeV particles during non-defective or mixed virus infections leads to the proposal that non-defective nucleocapsids could associate with M in perinuclear regions to promote effective transport of M (Stricker et al., 1994 ). Finally, budding of rhabdoviruses in the ER cisternae has been documented (Fekadu et al., 1982 ), an observation that could imply intra-cytoplasmic assembly of this category of viruses, which are normally shown to bud from the cell surface. As attractive as it looks, this model is, however, disproved by the great difficulty of detecting complete assembly structures anywhere but at the plasma membrane, and the present data are only suggestive of the possibility that viral assembly could take place internally.

One reason for the weak expression of the mutant phenotype certainly comes from the co-expression of M^wt and HAM³⁰ in the present system of mixed-virus infections. Indeed, attempts to recover a recombinant virus expressing only HAM³⁰ failed, in contrast to the successful recovery of recombinant SeV expressing HAM²¹ (not shown). The preparation of a cell line that conditionally expresses M^wt (as done by Tashiro et al., 1996 ) is presently being done, in order to allow recovery of recombinant SeV HAM ³⁰ so as to be able to follow the abortive assembly mediated by HAM³⁰ alone. If successful, this approach will allow us to use the trans-dominant negative M mutants in defining more clearly the step(s) at which assembly aborts and, hopefully, to identify the various steps of virus assembly.

The authors are indebted to Joseph Curran for careful reading of the manuscript and useful comments, to Marie-Luce Bochatton-Piallat and Nicolas Demaurex for an introduction to the use of the confocal microscope and to Claes Örvell (Huddinge, Sweden) for the kind gift of monoclonal antibodies. L.R. is a recipient of a grant from the Swiss National Foundation for Scientific Research.