Molecular Requirements for Self-Interaction of the Respiratory Syncytial Virus Matrix Protein in Living Mammalian Cells

Respiratory syncytial virus (RSV) is an important human pathogen, which infects respiratory tract epithelial cells causing bronchiolitis and pneumonia in children and the elderly. Recent studies have linked RSV matrix (M) ability to self-interaction and viral budding. However, RSV M has been crystalized both as a monomer and a dimer, and no formal proof exists to date that it forms dimers in cells. Here, by using a combination of confocal laser scanning microscopy and bioluminescent resonant energy transfer applied to differently tagged deletion mutants of RSV M, we show that the protein can self-interact in living mammalian cells and that both the N and C-terminus of the protein are strictly required for the process, consistent with the reported dimeric crystal structure.


Introduction
Respiratory syncytial virus (RSV) is the major cause of lower respiratory tract disease in infants and young children [1][2][3], responsible for one-third of deaths resulting from acute lower respiratory infection in the first year of life [4][5][6]. RSV also causes severe respiratory tract disease in immunosuppressed and older adults, leading to substantial annual mortality [7]. There are no vaccines or antiviral drugs that effectively target RSV despite decades of research [8]. Deeper understanding of the molecular mechanisms that underlie RSV assembly could pave the way to the identification of new vaccine/antiviral targets.
RSV is an enveloped virus with a non-segmented negative sense RNA genome and belongs to the Orthopneumovirus genus of the Pneumoviridae family [9]. The RSV genome is tightly encapsidated within the nucleocapsid, which is composed of nucleocapsid protein N, the RNA polymerase L and its cofactor phosphoprotein P, as well as the M2-1 protein. External to the nucleocapsid is a layer of matrix (M) protein which acts as a bridge between the nucleocapsid and the lipid bilayer envelope. Embedded in the envelope are the fusion (F), large (G) and small hydrophobic (SH) glycoproteins. M2-2, NS1 and NS2 proteins are not found in the virion in any significant amount but have important roles in the RSV replication cycle [10][11][12][13][14][15].
For BRET experiments, cells were trypsinized and 1 × 10 5 HEK293-T cells were seeded onto 24-well plates 1 day before transfection [39]. Each well was transfected with a total of 500 ng of plasmid DNA and 2 µL of lipofectamine 2000 (Thermofisher). BRET saturation experiments were performed transfecting cells with 0.5 ng of RLuc-M (1-256) and increasing amounts (0-450 ng) of YFP-M (1-256). Total DNA amount was normalized to 500 ng total with plasmid pCDNA3.1 (Thermofisher, Waltham, MA, USA). Importantly, no signs of cell toxicity were observed upon transfection of all M expression plasmids.

Microscopy/CLSM/Image Analysis
Subcellular localization of fluorescently tagged fusion proteins was visualized 24 h and 48 h after transfection using an inverted epi-fluorescent microscope (Leica, Wetzlar, Germany) equipped with a 40× objective, essentially as described previously [40]. 48 h after transfection, cells were fixed with 4% paraformaldehyde 15 min at room temperature (RT), before being mounted onto glass coverslips with FluoromountG (Southern Biotech, Birmingham, AL, USA). When required nuclei where counterstained with DRAQ5 (Life Technologies, Carlsbad, CA, USA, 1:1000). Samples were processed by confocal laser scanning microscopy (CLSM) using a Leica TCT-SP2 system, equipped with a Planapo fluor 63× oil immersion objective (Leica). The Fn/c values were determined using the NIH ImageJ 1.62 public domain software, from single cell measurements for each of the nuclear (Fn) and cytoplasmic (Fc) fluorescence, subsequent to the subtraction of fluorescence due to autofluorescence/background as described previously [41]. Co-localization analysis was performed using the coloc2 plugin. Data were plotted and analyzed using Prism 6 (GraphPad) software (La Jolla, CA, USA).

Bioluminescence Resonance Energy Transfer (BRET) Assays
BRET experiments were performed as described in [35]. Briefly, 293T cells were transfected in 24-well plates with appropriate amounts of BRET donor expressing plasmids. For each construct, the donor (RLuc) expressing plasmid was transfected either in the absence or in the presence of the relative acceptor (YFP) expressing plasmid to allow calculation of background BRET signal. 48 h post transfection, culture media was removed from wells and cells were very gently washed with 1 mL of PBS, before being resuspended with 290 µL of fresh PBS. Cells were further resuspended and 90 µL of mixture were transferred to a black bottomed 96-well plate (Costar ® , Washington, DC, USA, product number 3916) well in triplicate, and signals acquired using a spectrometer compatible with BRET measurements (VICTOR X2 Multilabel Plate Reader, PerkinElmer, Waltham, MA, USA). Fluorescent signal (YFPnet) relative to YFP fluorescent emission were acquired using a fluorimetric excitation filter (band pass 485 ± 14 nm) and a fluorimetric emission filter (band pass 535 ± 25 nm). Luminometric readings were performed at 5', 15', 30', 45' and 60' after addition of the substrate (native Coelenterazine or Coelenterazine-h, depending on the assay, 5 µM PJK, Kleinblittersdorf, Germany). Data were acquired for 1 s/well, using a luminometric 535 ± 25 nm emission filter (YFP signal) and a luminometric 460 ± 25 nm emission filter (RLuc signal). Before reading, the plate was shaken for 1 s at normal speed and with double orbit. After background subtraction using values relative to mock transfected cells, the data obtained were used to calculate the BRET signal, defined as the ratio between the YFP and RLuc signals calculated for a specific BRET pair, according to the formula: Similarly, the BRET ratio, defined as the difference between the BRET value relative to a BRET pair and the BRET value relative to the BRET donor alone, was calculated according to the formula: BRET saturation curves were then calculated using the GraphPad Prism software by plotting each individual BRET ratio value to the YFPnet/RLuc signal, and interpolating such values using the one-site binding hyperbola function of GraphPad Prism. Specific BRET pairs generate logarithmic shaped curves and reach a plateau. This allowed calculation of BRETmax (Bmax) and BRET 50 (B 50 ) values, indicative of maximum energy transfer and relative affinity of the BRET pair tested.

Visualization of RSV M crystal structures
PDB file 42V3 was downloaded from the protein data bank website and Molecular graphics and analyses were performed with the UCSF Chimera package [42]. Chimera is developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIGMS P41-GM103311).

Deletion of N-and C-Terminal Portions of RSV M Affects Protein Subcellular Localization
We aimed to investigate whether M exists as a dimer/oligomer within cells, and which protein domains are involved in dimerization. We initially analyzed the subcellular localization of several RSV M deletion mutants as expressed in Mammalian cells when C-terminally fused to YFP and CFP.     were quantitatively analyzed using software ImageJ to calculate the Fn/c ratio relative to each fusion protein, as described in Material and Method section. The mean ± SD relative to at least 75 cells from 2 independent experiments is shown. The dotted line represents Fn/c of 1, corresponding to an even distribution between the nucleus and the cytoplasm.

Deletion of N-and C-Terminal Portions of RSV M Affects M's Ability to Colocalize with Full-Length Protein
We decided to investigate whether the elements involved in dimer formation in vitro [27] were required for M self-interaction in cells. To this end, we expressed CFP-M (1-256) in the presence or in the absence of the M deletion mutants as fused to YFP and investigated the ability of each fusion protein to co-localize, as well as to reciprocally affect each other's subcellular localization. As expected, expression of CFP-M (1-256) resulted in a mainly cytosolic protein, occasionally showing a punctate pattern within the cytosol (Fn/c = 0.28 ± 12). Our data indicate that co-expression between CFP-M (1-256) and all YFP-M mutants tested in this study did not affect their reciprocal subcellular localization, as compared to when expressed individually (see Figure 3A-C).
in the absence of the M deletion mutants as fused to YFP and investigated the ability of each fusion protein to co-localize, as well as to reciprocally affect each other's subcellular localization. As expected, expression of CFP-M (1-256) resulted in a mainly cytosolic protein, occasionally showing a punctate pattern within the cytosol (Fn/c = 0.28 ± 12). Our data indicate that co-expression between CFP-M (1-256) and all YFP-M mutants tested in this study did not affect their reciprocal subcellular localization, as compared to when expressed individually (see Figure 3A-C).  Figure 3D).

RSV M Can Self-Interact in Living Cells
Our results suggest that deletion of M C-or N-terminal domain affects its ability to self-interact in a cellular context. However, they do not prove that the full-length protein is able to self-interact or that the C-terminal of the protein is able to interact with the full-length protein. Indeed, the co-localization observed between CFP and YFP tagged versions of full-length M may simply reflect the fact that tagging M with such spontaneously fluorescent proteins does not affect its subcellular localization, so that both CFP-and YFP-tagged version of M localize in the same area of the cell. We directly addressed this issue by bioluminescent energy resonant energy transfer (BRET) assays. To this end, BRET saturation experiments were performed by transfecting HEK293-T with a fixed amount of BRET DONOR plasmid RLuc-M (1-256; 0.5 ng) in the presence of increasing amounts of BRET ACCEPTOR plasmid YFP-M (1:256; 0-450 ng). As a positive control and a reference for data normalization, a fusion protein between RLuc and YFP (RLuc-YFP) was also expressed, and RLuc and YFP were individually co-expressed as a negative control ( Figure 4A).  At 48 h post transfection, YFP fluorescent and BRET signals were acquired in living cells, and BRET ratios calculated, as described in the Materials and Methods section. Our results indicate that the RLuc-YFP fusion generated a BRET ratio of 0.34 ± 0.02 ( Figure 4A), while the RLuc and YFP protein, generated a BRET ratio of 0. Importantly, co-expression of RLuc-M (1-256) and YFP-M (1-256) generated a BRET ratio which rapidly increased with the ratio between YFP-M (1-256) and RLuc-M (1-256) expression levels ( Figure 4A), and which quickly reached saturation. Data fitting allowed us to calculate the Bmax, corresponding to the maximal BRET ratio obtainable for the BRET pair (0.43

Discussion
The data presented in the current study shows that the RSV M protein can self-interact when expressed in living mammalian cells. Our study confirms in cell culture the self-interaction of M, previously shown by a number of studies reporting its ability to form homodimers and higher order oligomers in vitro [19,28,29], highlighting the physiological relevance of the in vitro observations.
Our discovery that this interaction in a cellular context requires both the N and C terminal domains of the protein is consistent with a recently reported head to tail dimeric structure of M, whereby the N-terminal domain of one subunit interacts with the C-terminal domain of the other subunit [27], and is in contrast with initially resolved monomeric M structure [26].
When transiently expressed in mammalian cells as YFP-M fusions, full-length M and its deletion mutants localized to the expected cellular compartment, depending on the presence or the absence of M NES and NLS, thus confirming and validating our previous work defining the nuclear transport motifs of M (see Figure 2; [43,44]). Importantly, YFP-M (1-200), lacking two leucine residues belonging to M NES (residues 194-206), localized significantly more in the nucleus than FL YFP-M. As expected, YFP-M (110-183), lacking the NES but bearing the NLS was present equally within the nucleus and the cytoplasm as has been shown previously, while YFP-M (183-256) that has the NES but lacks the NLS, was cytoplasmic [43].
Subcellular localization and co-localization analysis upon co-expression of CFP-M (1-256) with YFP-M (1-256) or its deletion mutants suggested that CFP-M (1-256) does not likely interact with any of the deletion mutants, with the possible exception of YFP-M (183-256), as indicated by the drop in co-localization (see Figure 3D). Furthermore, the subcellular localization of CFP-M (1-256) and YFP-M (110-183) was not affected upon co-expression, with the former remaining mainly cytosolic and the latter equally distributing between the cytoplasm and the nucleus (see Figure 3A-C). In contrast, YFP-M (183-256) partially co-localized with CFP-M (1-256), thus implying potential interaction between the two proteins. However, the co-localization observed likely reflects their presence in the same location and not necessarily an interaction (see below). Interestingly, YFP-M (1-200) formed cytoplasmic inclusion bodies (IBs) that look very similar to those formed by the CFP-M (1-256), yet when the two proteins were co-expressed, they did not co-localize. Expression of YFP-and CFP-fused deletion mutants also suggested that M (1-200) and M (183-256) may be able to self-interact as they formed IBs when expressed in living cells.
BRET analysis in living cells clearly showed that full-length M is very effective in forming dimers in living cells (see Figure 4A). However, none of the deletion mutants analyzed in our study were able to self-interact (see Figure 4B). This suggests that the observed IBs are formed due to aggregation that may be brought about by misfolding of the proteins. However, massive misfolding of the M mutants tested in our study is unlikely, since similar deletion mutants (containing M aas 1-144, 114-256 and 1-110), still interact with viral nucleocapsids to similar levels as the full-length protein [22]. Furthermore, RSV M (110-183) has been shown to inhibit host cell transcription to similar levels as compared to the full-length protein (Ghildyal et al.,unpublished observations [45]). Our data is consistent with the M dimer being formed by the head to tail interaction of the subunits. In addition, none of the deletion mutants were able to dimerize with the full-length M (see Figure 4). This finding is consistent with the structure of the M dimer, which has a very large interface [27].
The dimerization interface comprises residues 63 to 68, 92 to 105, 129 to 134, 144, 163, 225 to 235, while residue T205 likely modulates M oligomerization in a phosphorylation dependent manner [27]. Our data shows that all the residues are needed to form a stable dimer, consistent with previous work, which demonstrated that mutation of single residues in the context of the full-length protein has an observable effect on the filament formation and dimerization/oligomerization behavior of M [27]. Clearly, there are complex interactions with several residues in the dimerization interface with each having a specific role in stabilizing the structure. Since point mutations destabilizing RSV M self-interaction in vitro also negatively affected viral budding, it is reasonable to consider M dimerization as an attractive potential target for the development of antiviral agents. In this context, the BRET-based assay described here to monitor M self-interaction might provide a valuable tool for screening of compounds interfering with M self-interaction, or for the validation of hits identified by other methods [35,[46][47][48].