Paramyxoviruses spread through cell monolayers or through an organ by two mechanisms, successive rounds of virus infection and cell-cell fusion. Thus, infected cell surfaces produce virus particles as well as fusing with adjacent cells. In order to begin to understand the relationships between these two processes in NDV-infected cells, the protein requirements for assembly and release of virus-like particles were characterized. Avian cells, expressing the viral NP, M, HN, and F proteins, released VLPs nearly as efficiently as virus. These released particles had protein ratios similar to those of infectious virus, and their densities were homogeneous and only slightly less than that of authentic virus. The efficiencies of release of VLPs produced in three other paramyxovirus systems were 10% (SV), 34% (SV5), and 70% (SV) (42
). All previously described paramyxovirus systems have utilized human 293T cells for expression of viral proteins. We have found that NDV VLP release from avian cells, the natural host cell for NDV, had an efficiency of nearly 84%, but the efficiency of NDV VLP release from 293T or COS-7 cells was approximately 50% (unpublished observations). Therefore, differences between the efficiencies of VLP formation in the NDV system reported here and the efficiencies reported for other systems may be due to a cell type-dependent effect. Thus, NDV VLP assembly in avian cells represents an ideal system for exploring protein requirements for assembly and release of virus particles.
Using avian cells, we found that the NDV M protein, and only M protein, was sufficient for particle release. Expression of M protein alone resulted in release of M protein-containing particles with an efficiency comparable to that observed when all four proteins were expressed, suggesting that no other protein is required for efficient release. Particles released from cells expressing NDV M protein alone, however, were very heterogeneous with respect to density. While the reasons for this finding are unclear, it is possible that budding of M protein particles occurred indiscriminately from different cell membranes with differing densities. Alternatively, it is possible that particles contained different lipid-to-protein ratios due to variable oligomerization of the M protein. M proteins of other negative-stranded RNA viruses are reported to form oligomeric structures (11
). Particles formed from monomer M protein may have a higher lipid-to-protein ratio than particles formed from M protein in an oligomeric state. While M proteins of SV and human parainfluenza virus type 1 were also shown to be sufficient for particle release (4
), the SV5 M protein was not sufficient (42
NDV M protein was also necessary for particle release. In the absence of M protein expression, no other viral protein or combination of proteins resulted in significant particle release. By contrast, previous studies by two different groups reported that SV F protein exhibited an autonomous exocytosis activity demonstrated by the release of vesicles containing only the F protein (47
), although the level of release was very low. We found that cells expressing the NDV F protein alone did not release F protein-containing material, results similar to those reported by Schmitt et al. for the SV-5 system (42
). We did observe a trace amount of very light-density material that contained HN protein when this protein was expressed alone, but no HN protein was released when it was coexpressed with combinations of NP and F-K115Q.
Although all studies agreed upon the central role played by the M protein in virus release, specific interactions of other viral proteins with M protein required for the assembly of complete VLPs are still poorly understood. To define these interactions required for NDV assembly, we used three approaches. First, we determined the requirements for efficient incorporation of NP, F, and HN proteins into particles by expressing all combinations of these proteins with M protein. Second, the protein interactions in particles formed with all combinations of three and four proteins were defined by coimmunoprecipitation. Last, the colocalization of cell surface HN and F proteins with M protein when expressed in different combinations with M and NP proteins were characterized.
Pairwise expression of NP, HN, or F protein with M protein resulted in only trace amounts of NP, HN, or F protein incorporated into M-containing particles. In addition, expression of NP, F, or HN protein with M protein did not change the heterogeneous density of M protein-containing particles. In contrast, coexpression of M protein with two other proteins significantly increased the incorporation of NP, HN, or F protein into particles. The released particles had more-homogenous density, similar to that of particles containing all four proteins, a result that suggested that necessary and specific interactions between the three proteins resulted in both efficient incorporation of NP or glycoproteins and more-ordered particles. Furthermore, coexpression of two proteins with M protein also significantly increased the colocalization of M protein with either HN or F protein in the plasma membrane, indicating increased interactions with M protein.
To define these protein-protein interactions, particles formed with different combinations of three and four proteins were solubilized with nonionic detergent and proteins precipitated with cocktails of monospecific antibodies for M, HN, or F protein. Each antibody cocktail precipitated all proteins from VLPs formed with M, HN, F, and NP, although the efficiency of precipitation for each protein varied with the antibody specificity. These results are consistent with a network of interactions between all four proteins, such that precipitation of one resulted in the precipitation of the other three proteins but with efficiencies that varied, determined by how directly a protein was linked to the precipitated protein.
The protein-protein interactions were more clearly defined by immunoprecipitation of proteins from particles formed with all combinations of three proteins. These results show a specific interaction between HN and M proteins, between NP and M protein, and between F protein and NP (diagramed in Fig. ). There is no evidence for a direct interaction between F protein and M protein. There is likely a weak interaction between F and HN proteins, since anti-F protein antibodies precipitated HN protein from particles containing M, HN, and F proteins. In addition, since there is no interaction between F and M proteins, incorporation of F protein into these particles must be accomplished by interactions with HN protein. Our results cannot rule out an interaction between HN protein and NP.
FIG. 9. Protein-protein interactions in particles. Inset shows viral protein-protein interactions detected by coimmunoprecipitation of proteins in particles. Also shown are interactions proposed to result in assembly of particles formed by coexpression of all (more ...)
Thus, when all four proteins are coexpressed, NP and HN protein are incorporated into VLPs by a direct interaction with M protein (Fig. ). F protein is likely incorporated indirectly due to interactions with NP and HN protein. An ordered complex of the four proteins is supported by the dramatic colocalization of M protein with F protein and M protein with HN protein in the plasma membrane when all four proteins are coexpressed.
However, when only F is expressed with M protein, F protein is likely not significantly incorporated into particles because there is no direct interaction between the two proteins (Fig. ). Supporting this conclusion is the observation that there was no colocalization of F and M proteins in the plasma membrane in these cells.
In spite of direct associations of M with NP, there was little NP protein incorporation into particles when NP and M proteins were coexpressed in the pairwise combination. Previous reports that show that the M protein of Sendai virus is recruited in the cytoplasm by the viral nucleocapsid (46
). Perhaps NP causes the retargeting of M protein to this compartment. Indeed, coexpression of M protein with NP resulted in a 2.5-fold suppression of M protein-containing particle release, a result also consistent with retention of M protein in cells by NP protein.
Although precipitations of particles formed with M, HN, and F protein indicated a direct interaction of HN protein with M protein, there were only low levels of incorporation of HN protein into particles when HN and M proteins were coexpressed in a pairwise combination. Furthermore, there was little colocalization of the two proteins in the plasma membrane. Perhaps in the absence of other proteins, HN and M proteins are never localized in the same regions of the cell, preventing their association. It is also possible that the conformation of the HN protein transmembrane or cytoplasmic tail may be different in the absence of expression of F protein or NP protein, inhibiting association of HN protein with M protein. The reason for the 50% reduction of M protein particles upon coexpression of HN protein with M protein is unclear but has been previously reported for the Sendai virus system (47
Particles formed with NP, M, and F proteins are likely due to interactions between M and NP and interactions between F and NP (Fig. ). F protein may relocate NP to the plasma membrane, drawing M to specific domains containing F protein. Indeed, addition of NP increases the colocalization of M protein with F protein in the plasma membrane. Particles formed with NP, M, and HN proteins likely form due to interactions of both HN protein and NP with M protein (Fig. ). Expression of NP with HN and M proteins certainly increases the colocalization of M and HN proteins in the plasma membrane. Perhaps NP-M protein interactions alter the conformation of M, facilitating its interaction with HN protein. Indeed, surface HN protein in the presence of NP appears more punctuate along the cell edges.
The network of interactions proposed here could account for the conclusions of Schmitt et al. that the cytoplasmic domains (CT) of the HN and F proteins have redundant functions (42
). The CT domain of the F protein may target NP-M complexes to the plasma membrane by interactions with NP protein, while the HN protein CT domain targets these complexes by virtue of direct interactions with M protein.
The proposed interaction of M protein and NP is supported by studies of Stricker et al. of Sendai virus (46
). Interaction of HN protein with M protein is consistent with numerous studies suggesting an interaction of M protein with viral glycoproteins in paramyxovirus-infected cells or in cells transfected with paramyxovirus cDNAs (1
). Indeed, it has been reported that the respiratory syncytial virus G protein specifically interacts with M protein (13
). However, there are no previous reports of a direct interaction between F protein and NP. It is possible that interactions between viral proteins vary within paramyxoviruses, and the requirements for formation of VLPs may depend upon the distribution of late domains on the viral proteins.
The results presented here are consistent with the proposal that the NDV M protein buds indiscriminately from different cellular membranes in the absence of other viral proteins. When both glycoproteins are present in the plasma membrane, the M protein association with the plasma membrane may stabilize. NP association with F and M protein may also further stabilize and organize the network of interactions within the assembling particle.
In summary, we have established a VLP production system for NDV. We also showed that the M protein is sufficient and is required for NDV particle budding. Moreover, there are specific protein-protein interactions in VLPs involved in the ordered assembly of particles. Interactions identified between M and HN or F and NP may play a role in targeting M and NP into assembly sites in the plasma membrane.