Evidence for Ubiquitin-Regulated Nuclear and Subnuclear Trafficking among Paramyxovirinae Matrix Proteins

The paramyxovirus matrix (M) protein is a molecular scaffold required for viral morphogenesis and budding at the plasma membrane. Transient nuclear residence of some M proteins hints at non-structural roles. However, little is known regarding the mechanisms that regulate the nuclear sojourn. Previously, we found that the nuclear-cytoplasmic trafficking of Nipah virus M (NiV-M) is a prerequisite for budding, and is regulated by a bipartite nuclear localization signal (NLSbp), a leucine-rich nuclear export signal (NES), and monoubiquitination of the K258 residue within the NLSbp itself (NLSbp-lysine). To define whether the sequence determinants of nuclear trafficking identified in NiV-M are common among other Paramyxovirinae M proteins, we generated the homologous NES and NLSbp-lysine mutations in M proteins from the five major Paramyxovirinae genera. Using quantitative 3D confocal microscopy, we determined that the NES and NLSbp-lysine are required for the efficient nuclear export of the M proteins of Nipah virus, Hendra virus, Sendai virus, and Mumps virus. Pharmacological depletion of free ubiquitin or mutation of the conserved NLSbp-lysine to an arginine, which inhibits M ubiquitination, also results in nuclear and nucleolar retention of these M proteins. Recombinant Sendai virus (rSeV-eGFP) bearing the NES or NLSbp-lysine M mutants rescued at similar efficiencies to wild type. However, foci of cells expressing the M mutants displayed marked fusogenicity in contrast to wild type, and infection did not spread. Recombinant Mumps virus (rMuV-eGFP) bearing the homologous mutations showed similar defects in viral morphogenesis. Finally, shotgun proteomics experiments indicated that the interactomes of Paramyxovirinae M proteins are significantly enriched for components of the nuclear pore complex, nuclear transport receptors, and nucleolar proteins. We then synthesize our functional and proteomics data to propose a working model for the ubiquitin-regulated nuclear-cytoplasmic trafficking of cognate paramyxovirus M proteins that show a consistent nuclear trafficking phenotype.


Introduction
Paramyxoviruses include pathogens of global medical and agricultural concern. These viruses occupy broad ecological niches infecting a wide range of hosts including mammals, reptiles, birds and fish, and they cause diverse outcomes ranging from asymptomatic infection to lethal disease. Measles virus (MeV), mumps virus (MuV), the human parainfluenza viruses (hPIVs), respiratory syncytial virus (RSV), and human metapneumoviruses remain significant causes of human morbidity and mortality [1]. Animal pathogens, such as Newcastle disease virus (NDV) and the recently eradicated Rinderpest virus [2], have caused significant rates of lethal disease in birds and cattle, respectively. The newly emergent zoonotic paramyxoviruses Nipah virus (NiV) and Hendra virus (HeV) are among the most deadly known pathogens, showing case-fatality rates in excess of 70% in humans, and are classified as biosafety level 4 pathogens due to the absence of vaccines or therapeutics approved for human use [3][4][5][6].
Paramyxoviruses are released as enveloped virions from the host cell plasma membrane. Virions are~150-300 nm in diameter and are spherical, pleomorphic or filamentous in shape, depending on the virus and the producer cell-type. The non-segmented, single-strand, negative-sense RNA genomes of paramyxoviruses consist of six principal genes: nucleocapsid (N), phosphoprotein (P), matrix (M), fusion (F) and attachment (HN, H or G) glycoproteins, and polymerase (L) [1, 5,7]. The attachment and fusion glycoproteins mediate binding to sialic acid moieties or to specific protein receptors on the cell surface and the fusion of the viral envelope with the host cell plasma membrane [8][9][10]. Within the virion, the ribonucleoprotein (RNP) consists of the RNA-dependent RNA polymerase complex formed by P and L associated with the N-encapsidated RNA genome. L is required for viral RNA synthesis during viral replication [1,5].

Results
Nuclear export of some Paramyxovirinae matrix proteins is regulated by the ubiquitin-proteasome system Since the nuclear-cytoplasmic trafficking of the Nipah virus matrix protein (NiV-M) is regulated by its monoubiquitination [39], we wondered whether the ubiquitin-proteasome system similarly regulates the nuclear sojourn of other Paramyxovirinae M proteins. We cloned 3X-Flag-and GFP-tagged-M from prototypical members of the five Paramyxoviri-  [55]. HA-UbK0 functions as a ubiquitin (Ub) chain terminator or as monoubiquitin because all lysines have been mutated to arginines. We used this construct to visualize discrete ubiquitin bands and to determine if matrix proteins can be monoubiquitinated since this posttranslational modification can regulate the function of proteins without promoting proteasome-dependent protein degradation [56]. Cell lysates were subjected to anti-Flag immunoprecipitation (IP) and immunoblots were simultaneously probed with anti-HA and anti-Flag antibodies. As shown in Fig. 1A, for all the M proteins the majority of M is unmodified (M 0 ) at steady state. However, a detectable minority of M (M 1 ) is size-shifted by the molecular weight of at least one ubiquitin monomer (Ub,~8.5 kDa) (Fig. 1A, merge). These results indicate that all 3X-Flag-tagged M proteins investigated are ubiquitin substrates.
We have found that proteasome inhibition results in nuclear retention of NiV-M in transfected and in NiV-infected cells (S1 Fig.) [39]. Proteasome inhibition stabilizes polyubiquitinated proteins and depletes the cellular levels of free ubiquitin available for conjugation [57][58][59][60][61][62]. To determine whether ubiquitination is involved in the nuclear export of the other Paramyxovirinae M proteins, we treated GFP-M-expressing HeLa cells with the proteasome inhibitor MG132 (Fig. 1B-G). We used quantitative 3D confocal microscopy to characterize the subcellular localization of M. The cells were counterstained with DAPI to visualize nuclei, and with fluorescent phalloidin to visualize the entire cell, and the proportion of nuclear M was determined computationally as described in Materials and Methods. As with GFP-NiV-M, ubiquitin depletion via proteasome inhibition resulted in significant nuclear retention of GFP-tagged HeV-M, SeV-M and MuV-M ( Fig. 1B-E) [39]. We further confirmed biochemically that MG132 reduces the direct conjugation of ubiquitin to 3X-Flag-tagged NiV-M, HeV-M, SeV-M, MuV-M, MeV-M and NDV-M by co-IP of each and HA-UbK0, as described above, with quantification of immunoblot band integrated intensities as described in Materials and Methods (S2 Fig.). The ubiquitination of the various M proteins were differentially sensitive to proteasome inhibition with NDV-M and MeV-M being the least and most sensitive to proteasome inhibition, respectively (S2E-F Fig.). For NDV-M (Avulavirus) and MeV-M (Morbillivirus), reduction of ubiquitin conjugation did not result in a nuclear retention phenotype under the conditions and cell type examined (Fig. 1F-G), suggesting there is no strict correlation between the degree of matrix ubiquitination per se and M nuclear localization. In contrast, the ubiquitin-proteasome system appears to regulate the nuclearcytoplasmic trafficking of the Henipavirus  Nuclear export of some Paramyxovirinae matrix proteins is regulated by a putative NES and a lysine within the NLS bp We have shown that the nuclear-export of NiV-M is regulated by a leucine-rich nuclear export signal (NES) as well as by the K258 lysine residue located within the second basic patch of the bipartite nuclear localization signal (NLS bp ; Fig. 2A, blue residues) [39]. A K258A mutation partially disrupts the NLS bp and decreases nuclear localization of NiV-M, while a K258R mutation is unexpectedly retained in the nucleus despite having intact NES sequences and preservation of the positive charge necessary for NLSbp function. However, both mutants are impaired for ubiquitination [39]. Sequence alignment of M proteins indicates that a lysine is present in the homologously aligned position across the Paramyxovirinae genera. Thus, we hypothesized that this residue might be conserved for regulation of M ubiquitin-dependent nuclear export ( Fig. 2A, bold and underlined blue residues). To interrogate our hypothesis, we mutated the NLS bp -lysine to an arginine in all M proteins studied and analyzed their subcellular localization by quantitative 3D confocal microcopy as described above (Fig. 2B-G, quantified in Fig. 2H). Since NDV-M contains another lysine adjacent to this position we mutated both ( Fig. 2A, bold and underlined blue residues). A lysine to arginine mutation is expected to preserve the nuclear import function of the putative NLS bp , but prevents posttranslational modification at that position. As a comparison for nuclear retention, we also mutated the leucines that correspond to the NES of NiV-M within all M proteins ( Fig. 2A, bold and underlined blue residues) [39].
Mutation of the NES of GFP-NiV-M (M L106A L107A ) resulted in a significant increase in nuclear localization of the protein, which confirms our previous findings (Fig. 2B, 2H) [39]. Similarly, but to varying degrees, GFP-tagged HeV-M L106A 107A , SeV-M L102A L103A , and MuV-M L106A also exhibited significantly increased nuclear retention compared to their respective wild type (WT) proteins ( Fig. 2C-2E, 2H). In contrast, GFP-NDV-M L103A L106A had an apparent nuclear exclusion phenotype (Fig. 2F, 2H) contrary to expectations, while the nuclear localization of GFP-MeV-M L90A 191A was not significantly different than WT (Fig. 2G, 2H), indicating that these motifs are either redundant or non-functional in NDV-M and MeV-M.
Mutation of the NLS bp -lysine resulted in significantly enhanced nuclear localization of GFP-tagged NiV-M K258R , HeV-M K258R , SeV-M K254R , MuV-M K261R , and NDV-M K259R K260R , but not MeV-M K240R (Fig. 2B-2G). These phenotypic differences are quantified in Fig. 2H. Note that the spread in the degree of nuclear localization for any given M mutant also emphasizes the need to score a sufficient number of cells by computationally defined volumetric criteria (see Materials and Methods) in order to obtain robust statistics from inherently variable cell biological data. Thus, each data point in Fig. 2H (as in Fig. 1B-G) represents reconstructed volumetric data from a single cell, acquired from~20-30 confocal optical Z-stacks (at 0.3-0.5 μm/ step) per cell. In contrast to the NLS bp -lysine-to-arginine mutations, our attempt to disrupt the NLS bp consensus sequence through alanine substitutions in the second patch of basic residues (bp2) resulted in diffuse cytoplasmic localization of GFP-tagged NiV-M, HeV-M, SeV-M, and MuV-M (S3 Fig.). However, similar mutations did not appear to disrupt the localization of MeV-M or NDV-M (S3 Fig.) indicating that this motif does not function as the NLS bp in MeV-Extended Focus (maximum intensity projection) view of 3D confocal micrographs of HeLa cells transfected with GFP-tagged (B) NiV-M, (C) HeV-M, (D) SeV-M, (E) MuV-M, (F) NDV-M, or (G) MeV-M. At 16 h posttransfection, cells were treated with 50 μM MG132/0.5% DMSO or 0.5% DMSO for 8h. Cells were counterstained with DAPI to visualize nuclear DNA, blue, and fluorescent phalloidin to visualize the F-actin cytoskeleton, red. Scale bar 10 μm. In the corresponding graphs, the nuclear M fluorescence per cell was quantified from 3D-reconstructed confocal micrographs. *p<0.05; ****p<0.0001; NS, not significant by Student's t-test. Ubiquitination of Nipah, Hendra, Sendai and Mumps matrix proteins is dynamically regulated by an NLS bp lysine Since the proteasome inhibitor MG132 inhibits the nuclear export of GFP-tagged NiV-M, HeV-M, SeV-M, and MuV-M (Fig. 1), we wanted to test whether the NLS bp -lysine that regulates their nuclear export (Fig. 2B-E) also regulates their ubiquitination. In Fig. 3, we first assessed the ability of 3X-Flag-tagged NiV-M K258R , HeV-M K258R , SeV-M K254R, MuV-M K261R , MeV-M K240R and NDV-M K259R K260R to be ubiquitinated biochemically, via co-IP of 3X-Flagtagged-M with HA-UbK0, as described above (Fig. 3A-D). Although NDV-M and MeV-M did not exhibit a ubiquitin-dependent nuclear trafficking phenotype ( Fig. 1F-G), we included NDV-M K259R K260R and MeV-M K240R in this experiment since the ubiquitin conjugation of WT NDV-M and MeV-M was sensitive to MG132 inhibition, albeit to varying degrees (S2E-F Fig.). We controlled for protein abundance by normalizing the integrated intensity of the Ub band by the integrated intensity of the total M (Ub/M 0 +M 1 ). Using this measure, 3X-Flagtagged NiV-M K258R (Fig. 3A), HeV-M K258R (Fig. 3B) and MeV-M K240R (Fig. 3E) exhibited the greatest reduction in relative monoubiquitination compared to the WT proteins (>70%), while 3X-Flag-tagged SeV-M K254R (Fig. 3C) and MuV-M K261R (Fig. 3D) showed only a modest to mild impairment in monoubiquitination (36% and~15% reduction, respectively). 3X-Flagtagged NDV-M K259R K260R did not display reduced ubiquitination (Fig. 3F). Residual ubiquitination of the matrix mutants indicates that other lysines within the 3X-Flag-tagged M proteins are also targets of ubiquitin conjugation. Table 1  Ubiquitination is a dynamic process determined in part by the rates of conjugation versus de-conjugation, but our co-IP and immunoblot analysis of HA-UbK0-modified 3X-Flagtagged M is a steady-state assay. It is possible that this assay for ubiquitinated M might not efficiently detect subtle differences that arise from such dynamic processes since i) ubiquitinated 3X-Flag-tagged M proteins are of low stoichiometry relative to unmodified native protein, ii) 3X-Flag-tagged M proteins might be mono-and/or polyubiquitinated on multiple lysines, further obscuring a contribution of any single lysine to the sum total ubiquitination, iii) ubiquitination is reversible, and iv) HA-UbK0 must compete with endogenous ubiquitin. In an attempt to overcome these issues, and to further assess whether NiV-M K258R , HeV-M K258R, SeV-M K254R and MuV-M K261R are impaired for ubiquitination, we developed a bimolecular fluorescence conjugation (BiFC) ubiquitination assay in which ubiquitin-conjugation of M produces an irreversible fluorescence signal (S4 Fig.) [63,64]. We fused Ub and M to split Nand C-terminal fragments of the fluorescent protein Venus, VN173 and VC155, respectively. Covalent conjugation of Ub to M brings the spilt Venus fragments into close proximity and allows the two otherwise non-fluorescent Venus fragments to reconstitute a functional fluorophore [65,66]. This complemented Venus will remain associated with M (via VC155-M) even if the VN173-Ub moiety is subsequently cleaved from M by a deubiquitinating enzyme (DUB), preserving an atemporal record of ubiquitin conjugation (S4A Fig.). Analysis of total cellular fluorescence showed that the Ub-M BiFC signal was decreased by almost 80% for the Henipavirus-M K258R mutants, confirming the significant role of K258 in ubiquitination of NiV-M (Fig. 3A, Fig. 4A) and HeV-M (Fig. 3B, Fig. 4B). In addition, we determined that the Ub-M BiFC signals for SeV-M K254R (Fig. 4C) and MuV-M K261R (Fig. 4D) were also significantly decreased in BiFC signal by nearly 70% compared to the WT proteins.   . In sum, for the cognate paramyxovirus matrix proteins that exhibit a consistent nuclear trafficking phenotype that is both ubiquitin-and motif-dependent, our data supports a model where proper matrix ubiquitination is required for efficient nucleolar exit and/or preventing retention in the nucleolus.

Recombinant Sendai virus bearing matrix nuclear export mutants are defective for viral morphogenesis
We previously determined that NiV-M L106A 107A and NiV-M K258R are defective at budding virus like particles (VLPs) [39]. We wanted to compare the effects of the corresponding mutations in 3X-Flag-tagged SeV-M or MuV-M, however these proteins have a poor budding efficiency that is less than 10% of 3X-Flag-tagged Henipavirus-M proteins (Fig. 7A). This may be due to the presence of the 3X-Flag-tag or to the fact that SeV-M and MuV-M do not efficiently bud VLPs without the support of other viral proteins [41][42][43]. To overcome these technical difficulties and to study these mutations in a biologically relevant context, we engineered SeV-M L102A L103A and SeV-M K254R into a recombinant T7-driven, GFP-expressing Sendai virus genome (rSeV-eGFP) that can be rescued as live virus via the cotransfection of support plasmids expressing N, P, L (comprising the necessary replication complex) and a codon-optimized T7 polymerase. This highly efficient reverse genetics system allows us to quantify the number of rescue events directly in transfected producer cells at early time-points (see Materials and Methods). At two days post-transfection, GFP-positive cells (rescue events) could be observed by epifluorescence and quantified by FACS analysis. As a control for background GFP expression in the absence of virus production, we found that cotransfection of WT rSeV-eGFP and T7 polymerase without the N, P and L support plasmids resulted in no GFP-positive cells. We determined that rSeV-eGFP-M L102A L103A , and rSeV-eGFP-M K254R rescued at similar if not higher efficiencies than rSeV-eGFP-M WT (Fig. 7B). However, only rSeV-eGFP-M WT produced infectious viral titers (~10 7 I.U./ml) at day 6 post-rescue, while the mutants did not produce detectible infectious virus (<10 I.U./ml) (Fig. 7C).
To determine the nature of the defect in viral replication, we counterstained the viral rescue cells with anti-SeV-M or anti-SeV-F antibodies and analyzed them by 3D confocal microscopy. By day 6 post-rescue of rSeV-eGFP-M WT , infection has spread to all cells without evidence of cell-cell fusion (Fig. 7D). It is known that SeV replication in cell culture does not result in cellcell fusion [45,47,68], an unusual phenotype as most paramyxovirus infections result in extensive cell-cell fusion (e.g. see S1 Fig. for NiV). Interestingly, although the rSeV-eGFP-M L102A L103A and rSeV-eGFP-M K254R rescue cells did not produce infectious virus, the GFP-positive rescue cells did initiate the formation of large foci of fused cells (Fig. 7D, second and third horizontal panels). Virus-cell and cell-cell fusion require the presence of F and HN [9,10], and we confirmed that SeV-F is expressed on rSeV-eGFP-M L102A L103A and rSeV-eGFP-M K254R foci (Fig. 7E). Recombinant MuV-eGFP genomes engineered with M nuclear export mutants were also unable to efficiently spread beyond the fused cells formed at sites of rescue (S5 Fig.). These data indicate that proper M nuclear-cytoplasmic trafficking is necessary for viral morphogenesis.
To determine the nuclear localization of SeV-M, we counterstained rSeV-eGFP rescue cells with DAPI to visualize nuclei and anti-fibrillarin antibodies to visualize nucleoli. SeV-M WT was primarily extranuclear at the cell periphery (Fig. 7D, 7F). SeV-M L102A L103A did not have an obvious nuclear localization in the viral context, although intracellular inclusions were apparent, suggesting that this mutant nonetheless had an altered localization (Fig. 7F). SeV-M K254R , on the other hand, was strongly nuclear and enriched within the nucleoli (Fig. 7D, 7F). These results are consistent with the previous transient transfection experiments in which SeV-M K254R was also more strongly localized to the nucleus than SeV-M L102A L103A (Fig. 7H). Thus, Proteomics analysis supports a model of regulated nuclear transport of Paramyxovirinae matrix proteins that involves a critical nucleolar transit phase Having characterized determinants of Paramyxovirinae M nuclear-cytoplasmic trafficking encoded within some M proteins, we turned to identifying potential cellular regulators of this process. We generated inducible 3X-Flag-M-expressing stable HEK 293 cell lines to efficiently copurify M-interacting proteins and analyzed their composition using multidimensional Analyses of Paramyxovirinae Matrix Nuclear Trafficking protein identification technology (MudPIT) as described in Materials and Methods (S1-S4 Tables). We opted to determine the protein interactomes of NiV-M, HeV-M, SeV-M and NDV-M since these are the Paramyxovirinae M proteins with confirmed nuclear trafficking during live virus infection (Fig. 6, Fig. 7, S10 Fig.) [39,48,[51][52][53][54]69], and because these proteins cover the widest range of sequence homology to NiV-M:~90% amino acid identity for HeV-M, 37% amino acid identity for SeV-M, and~20% amino acid identity for NDV-M. S6 Fig. shows our experimental schema and stringent filtering that resulted in our list of putative M protein interactors detailed below and listed in S1-S4 Tables. Nonspecific interactions in MudPIT analyses tend to be independent of the bait of interest. Rather, they are background contaminants related to the cell type and the affinity purification scheme [70]. To remove background contaminants, our putative M interactomes represent only those proteins identified in the sample purifications that are absent in 3 independent negative-control purifications using lysates from the parental/isogenic Flp-In T-REx-293 cells, irrespective of relative abundances (S6 Fig., Worksheet 1 in S1-S4 Tables). As an independent confirmation of stringency, comparison of the putative NiV-M, HeV-M, SeV-M, and NDV-M interactomes to 21 relevant control experiments in the mass spectrometry contaminant repository, CRAPome, revealed relatively few additional proteins that are common sources of contamination (S6 Fig., Worksheet 2 in S1-S4 Tables) [70]. These were primarily actins, tubulins, histones and ribosomal proteins, which are also the most common contaminants across the entire CRAPome (Worksheet 2 in S1-S4 Tables) [70]. During manuscript revisions, another group published the identification of~130 HeV-M-interacting proteins using affinity-purification, in-gel digestion and mass spectrometric identification, and further characterized AP3B1 as a Henipavirus M interactor that regulates VLP production [71]. A majority of their proteins either went undetected by our global analyses or were excluded as background contaminants because they were present in the control purifications. For the purpose of comparison, the proteins in their paper that are also present in our HeV-M and/or NiV-M interactomes, excluding some ribosomal proteins, are KRI1, RFC1, FAM120A, SMC1A, SART3, UPF1, Nat10, Smarca5, UTP14A, POP1, ZC3HAV1, RAD18, AP3D1, USP7, Tat-SF1, SKIV2L2, PARP-1, and Importin-7 (Worksheet 1 in S1 and S2 Tables). Other than PARP-1, these proteins were unlikely to be present as contaminants in the 21 relevant CRAPome control experiments. We noted that the E3 ubiquitin ligase RAD18 was not present as a contaminant in any of the 21 CRAPome control experiments and was the least likely to be encountered in the entire CRAPome database; it was found in Analyses of Paramyxovirinae Matrix Nuclear Trafficking only 4 of 411 experiments with an average of only 1.3 spectra per experiment, whereas we measured 11 unique spectra (22.6% coverage) in the NiV-M affinity purification [70]. We confirmed the interaction of NiV-M with RAD18 by co-IP and immunoblot analysis (S7A Fig.), indicating that a combination of experimental and computational approaches to background contaminant subtraction can facilitate the identification and characterization of bona fide protein-protein interactions (S6 Fig.). Other ubiquitin ligases identified in our proteomics experiments (UBE2O and Cullin ring ligases) were also confirmed for copurification with M proteins by co-IP and immunoblot analysis (S7B-C Fig.).
Confident that our putative M interactomes were largely reflective of true protein-protein interactions, we further analyzed the interactomes bioinformatically and biochemically (Fig. 8,  S7-S9 Fig.). Comparisons of our putative NiV-M, HeV-M, SeV-M, and NDV-M interactomes to one another revealed significant overlap; over 60% of the proteins found in any single interactome were also found in the interactomes of one or more of the other three (Fig. 8A). Furthermore, we identified 178 proteins common to all M interactomes, the majority of which are not present in the 21 historical control experiments (Worksheet 3 in S4 Table). This common set of proteins represents 24-48% of all the proteins in any single viral M interactome (Fig. 8A, S1-S4 Tables). Interestingly, proteins associated with the nuclear pore complex were significantly enriched within individual M interactomes as well as the subset of common interacting proteins (Fig. 8B,-log 10 (p-value)>10). These include nuclear pore complex components (RanBP2, Nup37, Nup93, Nup107, Nup155, Nup205, Sec13, Seh1), nuclear transport receptors (NTRs) required for nuclear import of proteins (α/β-importins), nuclear export of proteins (Exp1/CRM1, Exp2), nuclear export of dsRNA/dsRNA-binding proteins (Exp5), nuclear export of tRNA (Exportin-T), nuclear export of mRNA (Rae1), and a regulator of the RanGTP/ GDP cycle that modulates the association/dissociation of cargo with NTRs (RanGAP1) (Fig. 8C, Worksheet 1 in S1-S4 Tables). All of these proteins were unlikely background contaminants (Worksheet 2 in S1-S4 Tables) and we confirmed that NiV-M interacts with αimportins and Exp1/CRM1 (S8 Fig.). Thus Paramyxovirinae M proteins interact with a highly interconnected network of proteins necessary for transport of NLS and NES containing cargo proteins across the nuclear pore.
Our cell biological and proteomic findings are synthesized into a working model for ubiquitin-regulated nuclear-cytoplasmic trafficking of M proteins shown in Fig. 8D. Consistent with the nucleolar transit phase exhibited by the Paramyxovirinae M proteins under study, the M interactomes also revealed a significant enrichment of resident or transient nucleolar proteins (Fig. 8E and 8F). One of these, the RNA polymerase I transcription factor UBF-1 (UBTF) had abundant spectral counts in the SeV-M and NiV-M interactomes but was essentially absent in the matched CRAPome control experiments (S1 Table and S3 Table). We determined that UBF-1 sequesters NiV-M in the nucleus and inhibits NiV-M budding when overexpressed (S9 Fig.). Thus, our proteomic and functional data indicate that M proteins interact with an array of nuclear and nucleolar proteins, at least some of which can modulate M nuclearcytoplasmic trafficking.

Discussion
Whether or not they replicate in the nucleus, many viruses are known to target, modify, and hijack nuclear components and nuclear functions to promote the infectious life cycle. It is generally thought that paramyxoviruses replicate in the cytosol without a nuclear stage. However, it is becoming increasingly clear that nuclear trafficking of M is shared by a number of paramyxoviruses. It was previously observed that SeV-M, NDV-M and RSV-M traffic through the nucleus [49,50,69] and a functional bipartite nuclear localization signal (NLS bp ) has been  (Fig. 8D, S3 Fig.) [39,48]. NLSs specify translocation through the nuclear pore through high-affinity interactions with importins, which in turn interact with cognate nuclear pore components on the cytoplasmic side [72]. In our proteomic analyses we identified numerous importins, exportins and nuclear pore complex components as common candidate interactors of NiV-M, HeV-M, SeV-M and NDV-M (Fig. 8, S8 Fig., S1-S4 Tables). Thus, the size of Paramyxovirinae M proteins (>40 kDa), the presence of a functional NLS bp within M proteins, and the interaction with nuclear transport receptors are strong evidence that the nuclear localization of M proteins is an active and regulated transport process. That the putative M interactomes show such a strong enrichment of proteins involved in nuclear-cytoplasmic transport also suggests that M proteins may antagonize the nuclear-cytoplasmic trafficking of host proteins and RNA to facilitate viral replication [73].
In addition to the NLS bp , we show that a leucine-rich NES sequence is functionally conserved within NiV-M, HeV-M, SeV-M and MuV-M (Fig. 2, Fig. 5, Fig. 8D). We note that mutation of the corresponding region in NDV-M resulted in decreased nuclear localization. However, this sequence is not as well conserved as in the other M proteins (Fig. 2A) and a recent study identified other functional NES motifs within different regions of NDV-M [52]. Thus, Paramyxovirinae M proteins appear to have both shared and unique determinants of nuclear-cytoplasmic trafficking depending on their evolutionary heritage. We also acknowledge that our experimental system utilizing human cells may not fully recapitulate the regulation of NDV-M trafficking since NDV is an avian virus, while NiV, HeV, SeV, MuV and MeV are mammalian viruses.
Nuclear export of NiV-M, HeV-M, SeV-M and MuV-M is also regulated by a lysine within the second basic patch of the NLS bp . Mutating this lysine to an arginine results in decreased ubiquitination and a nuclear retention phenotype that is phenocopied by pharmacological depletion of free ubiquitin with a proteasome inhibitor (Fig. 1B-E, Fig. 2B-E, Fig. 5, Fig. 6). We hypothesize that mutation of this lysine prevents its ubiquitination. Alternately, mutation of this lysine may prevent the ubiquitination of a nearby lysine within the NLS bp by preventing the interaction of M with a ubiquitin ligase. Whether ubiquitination regulates MeV-M or NDV-M function(s) remains indeterminate. The subcellular localization phenotypes of the NES and the various NLS bp mutants for NDV-M and MeV-M are not congruent (compare S3 Fig. and Fig. 2F-G). Furthermore, NDV-M and MeV-M also exhibit divergent sensitivity with regards to ubiquitin conjugation (S2E-F Fig.), and neither were found to be sensitive to MG132 induced nuclear retention (Fig. 1F-G). Altogether, our results suggest that the degree of the nuclear trafficking phenotypes of transfected M proteins mutated at the putative NES, the putative NLS bp , or the homologously aligned lysine within the NLS bp , are strongest for the Henipavirus M proteins, moderate for SeV-M and MuV-M, variable for NDV-M, and inconclusive/absent for MeV-M under the cell type and conditions examined (Table 1). Although we and nuclear transport receptors within the NiV-M, HeV-M, SeV-M and NDV-M interactomes using Cytoscape with the GeneMANIA plugin. (D) A model for matrix nuclear-cytoplasmic trafficking. Import through the nuclear pore complex is mediated by the interaction between an importin and the NLS bp of M proteins. Nuclear export is mediated by the interaction between an exportin and the NES of M proteins. Ubiquitination of some matrix proteins by as yet unknown E3 ubiquitin-ligases regulates nuclear export and/or nuclear reimport by masking the NLS bp .  [74].
How could NLS ubiquitination regulate nuclear trafficking? Protein import into the nucleus is regulated by the affinity of importins for cargo NLSs, which can be modulated by intermolecular or intramolecular masking of the NLS itself [75,76]. For example, ubiquitination of the NLS of p53 by MDM2 has been shown to block p53 nuclear import by preventing the binding of importin-α3 [77]. It was recently shown that the ubiquitin-conjugating enzyme UBE2O multi-monoubiquitinates tumor suppressor BAP1 on its NLS bp to promote cytoplasmic localization. It was further determined that UBE2O specifically binds and ubiquitinates a number of similar bipartite NLSs within nuclear trafficking proteins known to regulate RNA processing, transcription, DNA replication, and chromatin remodeling [78]. We hypothesize that ubiquitination of M on a lysine within the NLS bp itself prevents importin binding as a means to prevent nuclear re-entry once the protein has completed its nuclear sojourn (Fig. 8D, S8A Fig.). Given this model for ubiquitin-dependent nuclear-cytoplasmic trafficking, it is possible that M utilizes a nuclear-resident E3 ubiquitin ligase [79]. Our proteomic analyses identified a number of candidate ubiquitin ligases that interact with M proteins including UBE2O, which was found within the NiV-M and NDV-M interactomes (S7B Fig., S1 Table, S4 Table). Further study of these ubiquitin ligases will help resolve the spatiotemporal dynamics of M ubiquitination vis-à-vis nuclear trafficking (S7 Fig.).
Nuclear-cytoplasmic trafficking is a prerequisite for M budding and viral morphogenesis. We previously showed that NiV-M mutants defective in either nuclear import or nuclear export were also defective at budding VLPs [39]. Alternately, overexpression of a nuclear NiV-M-interacting protein UBF-1 can sequester NiV-M in the nucleus and inhibits efficient budding of VLPs (S9 Fig.). Other viruses have been reported to interact with UBF-1 and utilize or modify UBF-1 function. For example, UBF-1 is inactivated during Poliovirus infection as part of a viral strategy to inhibit host cell transcription globally [80]. The DNA viruses Adenovirus and HSV-1 co-opt UBF-1 into viral DNA replication centers, and it has been hypothesized that UBF-1 is used as a cofactor in viral DNA replication [81][82][83]. Finally, the SV40 large T antigen and the HCV NS5A protein stimulate RNA Pol I transcriptional activity and enhance rRNA synthesis by hyperphosphorylation of UBF-1 [84,85]. Such rRNA transcriptional activation is thought to contribute to the cell transformation caused by these tumorigenic viruses. For NiV, even though NiV-M clearly interacts with overexpressed UBF-1, and is colocalized with UBF-1 in the nucleus, it is unclear whether NiV utilizes UBF-1 to benefit viral replication under physiological expression levels and conditions (S9 Fig.). Nonetheless, the inhibitory effects of nuclear UBF-1 on VLP budding supports the model that functional trafficking to the plasma membrane requires properly regulated nuclear import and export (Fig. 8D).
Here, we also engineered M nuclear export mutants into recombinant SeV. rSeV-eGFP-M L102A L103A or rSeV-eGFP-M K254R were completely attenuated for production of infectious virus, but formed large foci of fused cells at sites of viral rescue. Moreover, nuclear localization of SeV-M K254R , the ubiquitination mutant, was observed in both virus rescue and transient transfection experiments (Fig. 2D, Fig. 5C, Fig. 7F). It is known that mutations that abrogate the interaction of M with the glycoproteins, including M deletion, can increase cell-cell fusion in SeV and MeV, while mutations that enhance their interaction can decrease cell-cell fusion [33,34,[45][46][47]68]. Since M and F proteins were expressed in the foci of fused cells, our results indicate that rSeV-eGFP-M L102A L103A and rSeV-eGFP-M K254R are defective in proper assembly of viral components at the plasma membrane rather than in expression of viral components necessary for budding per se. The link between viral replication and M ubiquitin-dependent nuclear-cytoplasmic trafficking may explain why proteasome inhibitors that deplete free cellular pools of ubiquitin have been found to inhibit SeV and NiV replication [39,86].
Beyond regulating nuclear import/export itself, we previously found that ubiquitination of NiV-M is necessary for membrane targeting and budding [39]. It is possible that the ubiquitination of M proteins promotes recognition by cellular factors such as ESCRT complexes known to mediate transport and budding of many enveloped viruses [56,87,88], especially in light of known sequence motifs in PIV5-M, SeV-M and MuV-M that can bind ESCRT complex components [27,43,89]. The status of M ubiquitination may also regulate the interactions of M with cellular factors inside the nucleus and within subnuclear compartments such as the nucleolus. A number of cellular proteins become enriched in the nucleolus upon proteasome inhibition, including p53 [90][91][92][93][94][95][96][97][98][99]. Similarly, pharmacological or genetic inhibition of NiV-M, HeV-M, SeV-M, and MuV-M ubiquitination sequesters these proteins in the nucleolus (Fig. 5,  Fig. 6), and nucleolar localization of SeV-M K254R was also confirmed in the context of rSeV-eGFP rescue (Fig. 7F). Nucleolar localization of M proteins is a natural feature of the nuclear sojourn of some M proteins, with M enriched at nucleoli during the early stage of live NiV and NDV infections (S10 Fig.) [69,100]. Although M proteins do not have an evident nucleolar localization signal (NoLS) that would have predicted this observation [101], our proteomics experiments suggest that Paramyxovirinae M proteins can interact with a number of nucleolar hub proteins ( Fig. 8E and F, S1-S4 Tables) [102].
Most, if not all, viral families interact with the nucleolus, often to usurp cellular functions and promote viral replication [103][104][105], as the nucleolus is a dynamic structure involved in a vast array of biological functions beyond ribosome biogenesis, including tRNA and mRNA processing and export from the nucleus, cell cycle regulation, and response to cellular stress. Additionally, there is growing recognition that NLS containing viral proteins target the nuclear pore complex to alter the export of macromolecules and mRNA, thereby counteracting antiviral responses and promoting viral gene expression at the expense of host gene expression [73]. For example, influenza NS1 is a multifunctional protein known to translocate to the nucleolus and to the nuclear pore where it inhibits host mRNA export factors resulting in impaired immune responses and enhanced viral virulence [106]. Vesicular stomatitis virus M also inhibits mRNA nuclear export through interaction with nuclear pore components [73,107]. Further, RSV-M is shuttled to the host cell nucleus where it inhibits host gene expression and induces cell cycle arrest, indicating that paramyxovirus M proteins also antagonize nuclear functions [108,109]. We hypothesize that ubiquitin-dependent nuclear and subnuclear trafficking of some Paramyxovirinae M proteins is part of a viral strategy to promote viral replication. Therefore, the study of M interactions with the nucleolus and the nuclear pore complex represents an opportunity to gain new insights into the cell biology of the nucleus and to identify novel antiviral targets.

Cell culture and transfection
HeLa, Vero, and HEK 293T cells were maintained at 37°C in a 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% 100X penicillin/streptomycin solution (Gibco/Life Technologies, Gaithersburg, MD). For confocal microscopy imaging, cells were seeded on 22 mm #1.5 coverglass coated with Collagen Type I (BD Biosciences, San Jose, California). Cells were transfected using Lipofectamine LTX per the manufacturer's instructions (Invitrogen/Life Technologies). 3X-Flag-M Flp-In T-REx-293 cell lines, generated as described below, were maintained at 37°C in a 5% CO 2 atmosphere in DMEM supplemented with 10% dialyzed FBS and 1% 100X penicillin/streptomycin solution. 3X-Flag-M protein expression was induced and immunoprecipitated as described below.
The Flp-In T-REx system (Invitrogen) was used to generate doxycycline-inducible 3X-Flag-M cell lines. Codon-optimized 3X-Flag-tagged NiV-M, HeV-M, SeV-M, and NDV-M were inserted within the KpnI and XhoI sites of pcDNA5/FRT/TO. The constructs and pOG44 were cotransfected into Flp-In T-REx-293 cells, and stable cell lines were selected with hygromycin and blasticidin according to the manufacturer's instructions.
The recombinant Sendai virus (rSeV) anti-genome RGV0, a Fushimi strain construct with F1-R strain mutations in F and M, and helper plasmids encoding SeV N, P and L were the kind gift of Dr. Nancy McQueen and are described in [67]. The encoded virus has the ability to replicate in mammalian cells without the addition of trypsin. We further modified the rSeV anti-genome construct by inserting an eGFP reporter flanked at the 3' end by a unique NotI site between the N and P genes. A hammerhead ribozyme sequence was inserted between the optimal T7 promoter and the start of the anti-genome. Mutations were introduced into the SeV-M ORF by OE-PCR using primers containing the desired mutations, followed by insertion into NotI and AfeI sites in the parental rSeV-eGFP construct.
The full-length construct encoding recombinant Mumps virus (rMuV) anti-genome of the Jeryl Lynn 5 (JL5) vaccine strain and helper plasmids encoding MuV-JL5 N, P and L proteins were a kind gift from Dr. W. Paul Duprex and are described in [118]. We modified the rMuV anti-genome construct by inserting an eGFP reporter between the NP and P genes. A hammerhead ribozyme sequence was inserted between the optimal T7 promoter and the start of the anti-genome. Mutations were introduced into the MuV-M ORF by OE-PCR using primers containing the desired mutations, followed by insertion into SalI and SbfI sites in the parental rMuV construct.

Multidimensional protein identification technology (MudPIT) analysis of matrix interactomes
3X-Flag-M Flp-In T-REx-293 cell lines were grown to~80% confluency and induced for protein expression with 100 ng/mL doxycycline for 24h. Cells were washed three times in dPBS and lysed in 100 mM Tris-HCL pH 8, 150 mM NaCL, 5 mM EDTA, 5% glycerol, 0.1% NP40, complete protease cocktail (Roche), PhosSTOP (Roche) and 25 mM N-ethylmaleimide. Cell lysate was clarified by centrifugation at >15,000×g for 15 min at 4°C and incubated with lysis buffer-equilibrated anti-Flag M2 affinity gel (Sigma-Aldrich, St. Louis, MO) for 2 hours at 4°C. The affinity gel was extensively washed with lysis buffer and then with elution buffer consisting of 100 mM Tris-HCL pH 8, 150 mM NaCL, 5 mM EDTA, and 5% glycerol. Bound proteins were eluted from the affinity gel with elution buffer containing 3X-Flag peptide (Sigma-Aldrich), were precipitated with trichloroacetic acid, washed with acetone twice, dried, and stored at -20°C until further processing.
Protein samples were resuspended in 8M urea in 100 mM Tris pH 8.5, reduced, alkylated and digested by the sequential addition of lys-C and trypsin proteases as previously described [119]. The digested peptide solution was fractionated online using strong-cation exchange and reverse phase chromatography and eluted directly into an LTQ-Orbitrap mass spectrometer (Thermofisher) [119,120]. MS/MS spectra were collected and subsequently analyzed using the ProLuCID and DTASelect algorithms [121,122]. Database searches were performed against a human database containing the relevant paramyxovirus M protein sequence. Protein and peptide identifications were further filtered with a false positive rate of less than 5% as estimated by a decoy database strategy [123]. Normalized spectral abundance factor (NSAF) values were calculated as described [124]. Proteins were considered candidate M-interacting proteins if they were identified in the relevant affinity purification but not present in 3 independent control purifications using lysates from the parental Flp-In T-REx-293 cells. Analysis of other potential background contaminants was performed using CRAPome [70]. Venn diagrams were generated using jvenn [125]. Gene-annotation enrichment analysis was performed using DAVID Bioinformatics Resources 6.7 [126,127]. Physical and predicted protein interaction networks were visualized using the GeneMANIA plugin for Cytoscape 3.1 [128,129].

Immunoprecipitations and immunoblot analysis of matrix ubiquitination
Transfected HEK 293T cells were washed once in dPBS and lysed in 100 mM Tris-HCL pH 8, 150 mM NaCL, 5 mM EDTA, 5% glycerol, 0.1% NP40, complete protease cocktail (Roche) and 25 mM N-ethylmaleimide. The cell extract was clarified by centrifugation at >15,000×g for 15 min at 4°C before incubation overnight at 4°C with lysis buffer-equilibrated anti-Flag M2 affinity gel (Sigma-Aldrich, St. Louis, MO). The affinity gel was extensively washed with lysis buffer and then with elution buffer consisting of 100 mM Tris-HCL pH 8, 150 mM NaCL, 5 mM EDTA, and 5% glycerol. Bound proteins were eluted from the affinity gel with elution buffer containing 3X-Flag peptide (Sigma-Aldrich), were subjected to SDS-PAGE and transferred to immobilon-FL PVDF membrane (EMD Millipore, Billerica, MA). To analyze M ubiquitination by immunoblot, HEK 293T cells were cotransfected with 3X-Flag-M and HA-UbK0 for 24h and subjected to immunoprecipitation as described above. Membranes were simultaneously probed with mouse anti-Flag M2 primary antibodies (Sigma-Aldrich) and rabbit anti-HA primary antibodies (Novus Biologicals, Littleton, CO) followed by anti-mouse-680 and anti-rabbit-800 secondary antibodies (LI-COR, Lincoln, Nebraska) and imaged on an Odyssey infrared scanner (LI-COR) according to the manufacturer's instructions. To quantify relative ubiquitination, the background subtracted integrated fluorescence intensities of the monoubiquitin bands (Ub) normalized to total M (M 0 +M 1 ) was determined using LI-COR Odyssey software.

Depletion of cellular free ubiquitin
For 3D confocal microscopy analysis, transfected HeLa cells were treated with 50 μM MG132 or 0.5% DMSO at 16 h post-transfection for 8 hours, then fixed and processed for quantitative image analysis as described below. For immunoblot analysis of M ubiquitination during ubiquitin depletion, transfected HEK 293T cells were treated with 10 μM MG132 or 0.1% DMSO at 18 h post-transfection for 6 hours, and 3X-Flag-M was immunoprecipitated as described above.
Quantification of virus-like particle budding VLP budding assays were performed as described in [39]. Briefly, precleared supernatants from 3X-Flag-M transfected HEK 293T were ultracentrifuged through a 20% (w/v) sucrose at 36,000 rpm for 2 h at 4°C (AH-650 rotor, Thermo Scientific). VLP pellets and cells were resuspended in lysis buffer and subjected to SDS-PAGE and anti-Flag immunoblotting. Relative integrated intensity of VLP/cell lysate bands were quantified and normalized relative to the budding of 3X-Flag-NiV-M.

Nipah virus infection and recombinant virus rescue
HeLa cells were infected with Nipah virus under biosafety level 4 (BSL-4) conditions as described in [39]. For rescue of WT or mutant rSeV-eGFP, 2X10 6 HEK 293T cells were transfected with recombinant plasmid encoding the anti-genome (4 μg) along with the cognate accessory plasmids encoding SeV NP (1.44 μg), P (0.77 μg), and L (0.07 μg), and a codon optimized T7 RNA polymerase (4 μg) using Lipofectamine LTX (8.9 μL) and Plus Reagent (5.5 μL), according to manufacturer's instructions. Cells were harvested for FACS analysis at 48 hours post-transfection (the earliest time point when GFP-positive cells can be observed by epifluorescence microscopy, yet when supernatant titer is still not detectable) to quantify rescue efficiency. The number of GFP-positive cells (rescue events) was determined from 500,000 cells analyzed with a FACSCalibur Flow Cytometer (BD Biosciences) and FlowJo software (TreeStar Inc., Ashland, OR). Cells plated on coverslips were fixed at day 6 post-transfection for analysis of rescued virus infection by 3D confocal microscopy. Supernatant was collected from rescue cells at day 6 post-transfection for quantification of viral titers. Briefly, supernatant stored at -80°C was thawed on ice and serial diluted 2-fold in serum-free DMEM. 100 μL of each dilution was used to infect~60,000 Vero cells in a 24-well plate for 1 hour. After 1 h, 500 μL of DMEM 10% FBS was added to each well and the cells were incubated at 37°C. Cells were harvested for FACS analysis at 24 h post infection and titers were calculated based on percent infection in the linear range of supernatant dilutions. For rescue of WT or mutant rMuV, 4X10 5 BSR-T7 cells were transfected with recombinant plasmid encoding the anti-genome (5 μg) along with the cognate accessory plasmids encoding MuV NP (0.3 μg), P (0.1 μg), and L (0.2 μg), and a codon optimized T7 RNA polymerase (2 μg) using Lipofectamine LTX (18.75 μL) and Plus Reagent (7.5 μL), according to manufacturer's instructions.

Quantitative analysis of 3D confocal micrographs
Random 40X fields were imaged using acquisition settings ensuring no under-saturated or over-saturated pixel intensities. Volocity 5.5 software was used for quantitative analysis of 3D confocal images. To determine the quantity of nuclear M, the nuclear compartment was defined with the find objects function within the DAPI-fluorescence channel. Holes in objects (DNA-absent regions such as nucleoli) were filled, and fluorescent objects smaller than nuclei were excluded. The entire cell body was defined by drawing a region of interest (ROI) encompassing all F-actin staining. The sum of voxel intensities in the GFP channel was measured within these defined sets. The average voxel fluorescence of untransfected cells was used for background subtraction. To determine the quantity of nucleolar M, the nucleolus was defined with the find objects function within the Fibrillarin-stained-fluorescence channel. Nucleolar objects were grouped within their respective nuclei, defined as above based on the DAPIfluorescence channel. To quantify fluorescence from bimolecular fluorescence complementation images, a ROI was drawn around cells fluorescent in the YFP channel. The average voxel fluorescence of untransfected cells was used for background subtraction.

Statistical analysis
For analysis of M nuclear localization, p-values were generated with a Student's t test when analyzing two sample groups. To analyze three or more sample groups, p-values were generated by ANOVA with Bonferroni correction for multiple comparisons. To analyze BiFC experiments, p-values were generated using a Mann-Whitney test. All graphs and statistical analyses were generated with Prism 6 (GraphPad Software, La Jolla, CA).