To define the SARS-CoV genes essential for viral assembly, mammalian expression vectors encoding different viral genes were analyzed. These genes included the S, M, N, and E proteins (Fig. ) synthesized from the predicted amino acid sequence by reverse translation, with codon usage typical of human cells. The DNA sequence of each gene was confirmed, and expression was confirmed by Western blot analysis or in vitro transcription-translation (Fig. ). To evaluate the contribution of these gene products to viral assembly, the expression of the SARS-CoV S, M, N, and E genes was analyzed in different combinations in transfected 293 cells. Coexpression of all four gene products resulted in the assembly of electron-dense structures that resembled pseudoparticles of ~100 nm (Fig. ), characteristic of the SARS coronavirus. The absence of positive-stranded viral genomic RNA, protease, or the viral polymerase indicated that they were not essential for the formation of the SARS-CoV core particles, though the central lucency raised the possibility that RNA was taken up nonspecifically by the pseudoparticles (Fig. , right panel).
FIG. 1. SARS-CoV gene organization and analysis of viral assembly by transmission EM. (A) Schematic representation of the SARS-CoV genome. The genes encoding viral proteins are drawn to scale. Bars denote the coding regions, and lines indicate the noncoding sequences. (more ...)
To examine the minimum requirements for pseudoparticle formation, different combinations of these viral genes were analyzed systematically by cotransfection. Any combination of genes that expressed M and N, with or without S and/or E, was able to generate intracellular virus-like particles (VLPs), and the pseudoparticles did not form in the absence of the M and N proteins (Table and Fig. ). No single viral gene was able to support the formation of viral capsids within these cells. These structures were localized primarily to endosomal vesicles and were seen most prominently in juxtaposition to the nucleus (Fig. ), which is typical of many coronaviruses. They appeared similar in size and morphology to replication-competent virus (18
). Though cotransfection of the N and M proteins allowed detection of viral pseudoparticles intracellularly, no budding virus or corona-like structures were visible in these transfected cells. The addition of the S glycoprotein, however, allowed for budding and formation of a corona-like structure (Fig. ), indicating that S is likely to be important not only for viral fusion but also for maturation and egress from cells.
Summary of SARS-CoV capsid formation after cotransfection of viral genes by scanning transmission EMa
FIG. 2. Analysis by transmission EM of the contribution of different viral gene products to viral assembly. Nucleocapsid formation by cotransfection of the indicated combinations of expression vectors encoding the specified viral gene products. Transfections (more ...)
FIG. 3. Formation of coronavirus-like particle by inclusion of S glycoprotein expression vector. Electron micrograph of virus particles in 293T cells transfected, as described in the note to Table , with S, M, and N. High-power view of VLPs forming (more ...)
To characterize these synthetic SARS-CoV particles further, we performed buoyant density gradient sedimentation analysis. Because the capsid core was found in higher quantities within cells, lysates were prepared from transfected cells that were frozen and thawed three times. When fractions from a gradient of clarified cell lysates were analyzed by Western blotting with human immune serum, the peak of viral protein expression, composed primarily of N and S proteins (Fig. ), was detected at a density of 1.18 g/ml (Fig. ), comparable to the buoyant density described for other coronaviruses (24
). The specificity of this gradient was further confirmed by transfection with N alone, which was distributed evenly across the gradient and did not show a buoyant density similar to the pseudoparticles (Fig. ).
FIG. 4. Release of assembled capsids from transfected 293T cells and buoyant density gradient analysis. (A) Western blotting and gradient sedimentation analysis of assembled capsids from clarified cell lysates transfected with S, M, and N (top) or transfected (more ...)
To determine the specific domains required for SARS-CoV VLP assembly, deletion mutants of the M protein were prepared. A membrane-spanning structure was predicted by the SOSUI computer program (http://sosui.proteome.bio.tuat.ac.jp
), and three recognizable domains were identified: an NH2
-terminal 12-amino-acid extracellular domain, a central region with three membrane spanning segments, and a COOH-terminal cytoplasmic domain (Fig. ). These domains also exist in IBV and MHV and are essential for viral assembly (5
). Three mutants were therefore created with either a deletion of the putative extracellular domain (MΔN12
), excision of the transmembrane domains (MΔTM), or removal of the putative cytoplasmic domain (MΔC). In vitro transcription-translation showed that these mutants expressed proteins of the expected size (Fig. ). Coexpression of N with mutant M genes that lacked either the transmembrane or the cytoplasmic domain abolished its ability to form pseudoparticles (Fig. ). In contrast, deletion of the NH2
-terminal putative extracellular region had no effect on particle formation (Fig. ), indicating that both transmembrane and cytoplasmic domains of M are required for viral particle formation.
FIG. 5. Expression and interaction of M and N gene products in vitro and model of the interactions leading to the formation of SARS-CoV. (A) Schematic representation of M expression vector and mutants. The ability of the indicated mutants to form nucleocapsids (more ...)
The interaction between these viral proteins was analyzed further to establish the biochemical basis for virus particle formation. The major structural proteins were synthesized by transcription with T7 RNA polymerase and translation with rabbit reticulocyte lysates in vitro, as were mutant forms of M (Fig. ). Canine pancreatic microsomal membranes were added to improve the expression of S, E, M, and the three M mutants and to model the membrane interactions. The full-length M protein was able to associate with the COOH-terminal His-tagged N protein when they were coincubated at 30°C for 30 min and were then pulled down with polyclonal anti-His antibody (Fig. , lane 7). In fact, the soluble COOH-terminal region of M bound more avidly to N than to the full-length protein, suggesting that this interaction was specific and localized to the cytoplasmic domain of the protein (Fig. , lane 9), which may be less well exposed during protein translation of the complete protein in vitro. In contrast, we were unable to detect an interaction of the S glycoprotein with N, though it did bind to M and E (Fig. , lane 17 versus lanes 18 and 19). A control membrane-binding protein, gp145, did not bind to S, N, M, or E under the same conditions (Fig. , lanes 16, 20, 21, and 22). Together, these findings suggest that M plays a pivotal role in nucleocapsid assembly through its ability to interact with N through its COOH-terminal cytoplasmic domain and with S through other regions (Fig. ). It thus serves as a critical bridge between essential internal and external components of the virus.
FIG. 6. Proposed schema of interactions between S, M, and N and their roles in nucleocapsid formation and viral assembly. The newly synthesized M protein is anchored in intracellular membrane through its membrane-spanning region. The COOH-terminal domain of M (more ...)