Structure and Functional Analysis of the RNA- and Viral Phosphoprotein-Binding Domain of Respiratory Syncytial Virus M2-1 Protein

Respiratory syncytial virus (RSV) protein M2-1 functions as an essential transcriptional cofactor of the viral RNA-dependent RNA polymerase (RdRp) complex by increasing polymerase processivity. M2-1 is a modular RNA binding protein that also interacts with the viral phosphoprotein P, another component of the RdRp complex. These binding properties are related to the core region of M2-1 encompassing residues S58 to K177. Here we report the NMR structure of the RSV M2-158–177 core domain, which is structurally homologous to the C-terminal domain of Ebola virus VP30, a transcription co-factor sharing functional similarity with M2-1. The partial overlap of RNA and P interaction surfaces on M2-158–177, as determined by NMR, rationalizes the previously observed competitive behavior of RNA versus P. Using site-directed mutagenesis, we identified eight residues located on these surfaces that are critical for an efficient transcription activity of the RdRp complex. Single mutations of these residues disrupted specifically either P or RNA binding to M2-1 in vitro. M2-1 recruitment to cytoplasmic inclusion bodies, which are regarded as sites of viral RNA synthesis, was impaired by mutations affecting only binding to P, but not to RNA, suggesting that M2-1 is associated to the holonucleocapsid by interacting with P. These results reveal that RNA and P binding to M2-1 can be uncoupled and that both are critical for the transcriptional antitermination function of M2-1.


Introduction
Human respiratory syncytial virus (RSV), a pneumovirus of the Paramyxoviridae family in the Mononegavirales order, is an important respiratory pathogen and the major cause of bronchiolitis and pneumonia in children [1]. Bovine RSV on the other hand represents an important economic issue due to the high morbidity and mortality of infected calves [2]. Whereas current efforts are mainly focused on the development of safe RSV vaccines for infants, the development of antiviral drugs specifically targeting viral-specific functions such as the RSV RNA-dependent RNA polymerase complex (RdRp) represents a promising alternative for treatment.
Four of the 11 proteins (the nucleoprotein N, the phosphoprotein P, M2-1 and the large polymerase subunit L), encoded by the RSV single-stranded negative-sense genomic RNA, are associated with the viral genome to form the holonucleocapsid [3]. The genomic RNA of RSV is maintained as a nuclease-resistant N-RNA ribonucleoprotein complex, which acts as a template for the RdRp that is responsible for both replication and transcription of the genome. Whereas the highly processive replicase generates a complete positive-sense RNA, which acts in turn as a template for genomic RNA synthesis, the transcriptase produces ten different subgenomic capped and polyadenylated mRNAs. Transcription proceeds by a sequential stop-and re-start mechanism in which the polymerase responds to cis-acting signals present in intergenic regions [4]. Transcription is (re)initiated at a highly conserved 9-10 nucleotide transcription promoter (gene start, GS) signal. Semiconserved 12-13 nucleotide gene ends (GE) signal for polyadenylation and release of the nascent mRNA [5]. The polymerase has a propensity to dissociate from the N-RNA template, but cannot reinitiate at a downstream gene in case of premature termination [6], which leads to a decreasing transcription gradient from the 39 to the 59 end of the genome.
For all known pneumoviruses, RdRp driven transcription depends on M2-1. RSV M2-1 is a transcription antitermination factor that is important for the efficient synthesis of full-length mRNAs [7] as well as for the synthesis of polycistronic readthrough mRNAs [4,6,8,9]. The latter activity is thought to facilitate polymerase access to promoter-distal regions of the genome, and hence transcription of all genes [3]. It was shown that the M2-1 protein reduced termination at all gene junctions, but that the efficiency in the presence of M2-1 varied at the different gene junctions [6,9]. However, mechanisms by which M2-1 prevents the polymerase from terminating transcription remain to be clarified. There are at least three different scenarios. (i) M2-1 could bind to the nascent mRNA transcript to facilitate transcription elongation, perhaps by preventing the mRNA from re-hybridizing to the template, or forming secondary structures that might destabilize the transcription complex. This hypothesis is sustained by the finding that RSV mRNA are co-precipitated with M2-1 from RSV infected cells [10]. (ii) The polymerase processivity enhancing effect of M2-1 could be due to an increase of the affinity of the polymerase for the genomic RNA template in a sequence non-specific manner. (iii) M2-1 could recognize GE sequences either on the nascent mRNA or on the RNA template and prevent the release of the polymerase complex from its template, favoring transcription re-initiation at the downstream GS sequences.
In RSV-infected cells, M2-1 co-localizes with the other RdRp components in inclusion bodies (IBs) [11], which are regarded as centers of RNA synthesis [12]. The basic 194-residue RSV M2-1 has been shown to be an RNA binding protein, but specificity of M2-1 RNA binding has been debated. It was reported that M2-1 was able to bind to long RNAs without sequence specificity and with an apparent K d of 30 nM, and that it bound specifically to short (80 nucleotides) but not long (700 nucleotides) RNAs containing the positive-sense antigenomic leader sequence with an apparent K d of 90 nM [13]. Elsewhere it was demonstrated that M2-1 interacts more specifically with viral mRNAs not containing the leader sequence during infection [10]. In addition M2-1 interacts with P in vitro [14], competitively to RNA [15]. M2-1 is a modular protein that contains four domains. M2-1 forms tetramers in solution [15,16], and the oligomerization domain was mapped to the region 33-62 [15]. The N-terminal region (residues 1-30) contains a putative zinc binding domain with a Cys 3 His motif, which is essential for the function of M2-1, but whose exact role is still unknown [17,18,19]. The RNA and P binding properties are related to the central part (or core domain) of the molecule (residues 53-177) [13,15]. The C-terminal tail is predicted to be unstructured.
Here we have addressed the molecular basis of the interaction between RSV M2-1 and its partners. We have investigated the structural aspects of P and RNA binding to the core domain of M2-1 by NMR spectroscopy and report the solution structure of RSV M2-1  , which shows structural homology with the Cterminal domain (CTD) of the Ebola virus (EBOV) VP30 protein.
In this a-helical domain we have identified residues that contribute to two adjacent, partially overlapping contact surfaces, with P and RNA respectively. We show that mutations of several of these residues specifically disrupt the M2-1:P or M2-1:RNA interaction in vitro and have a drastic effect on intracellular co-localization of full-length M2-1 with P as well as on the function of M2-1 as a transcription co-factor.
RNA binds to the main basic cluster of M2-1  Incubation of M2-1  with yeast RNA in a ,1:1 molar ratio resulted in simultaneous shifting and broadening of several 1 H-15 N cross peaks in the 1 H-15 N HSQC spectrum of M2-1 58-177 (see Figure S1). Treatment with RNAse A reversed these effects. This experiment confirmed the RNA binding ability of M2-1 58-177 in vitro. The observed fast to intermediate exchange regime was an indication for a weak interaction. To get rid of the broadening contribution by transversal relaxation of large RNAs twice the molecular weight of M2-1 58-177 , we investigated the RNA:M2-1 58-177 interaction by NMR by using short synthetic 10-15 nucleotide negative-sense (genomic) RNAs containing selected transcriptional signals [3,21] as well as their complementary positive-sense sequences. We tested the 39 polymerase entry site (leader), the U-rich region upstream of the first GS signal, the GS, and the F and SH gene ends (GE_F and GE_SH). Sequences are detailed in Table 1. The oligonucleotides were designed to minimize self-association or formation of secondary structure. Only the leader-neg (59-CGCAUUUUUUCGCGU-39) and long U-rich (59-CCCAUUUUUUUGGUU-39) sequences were predicted to form hairpins with a poly-U loop and a stem of two or three G-C and A-U pairs with negative free energies. Calculations were carried out on the mfold web server [22]. The absence of selfassociation was assessed by 1 H NMR using 200 mM oligonucleotide solutions in water, except for leader-pos and GE_F-pos, for

Author Summary
Premature termination of transcription by the RNAdependent RNA polymerase (RdRp) complex of respiratory syncytial virus (RSV) is prevented by the M2-1 protein. This transcription factor interacts with both RNA and viral phosphoprotein P, the main RdRp cofactor, through a specific ''core'' domain. Using NMR, we solved the 3D structure of this domain and characterized the surface residues involved in P-and RNA-binding. Based on these data, we designed point mutations impairing binding to either RNA or P. We studied the functional implications of these mutations for transcription and co-localization of full-length M2-1 in cytoplasmic inclusion bodies, where viral transcription likely occurs. We found that the RNAand P-binding surfaces are in close proximity, accounting for competitive binding in vitro. Our results suggest that binding to both RNA and to P is necessary for transcriptional antitermination by M2-1, but that the role of the interaction with P is primarily to recruit M2-1 to the RdRp complex. Finally we show that the M2-1 core domain is homologous to the C-terminal domain of Ebola virus VP30 despite low sequence identity, solidifying the relationship between these two proteins and transcriptional regulation strategies shared by viruses belonging to the Filoviridae family and the Pneumovirinae subfamily.
which two broad imino proton signals were observed in the guanosine and uridine regions respectively, corresponding to partial formation of two G-C and U-A base pairs. Finally formation of an RNA duplex with the two complementary strands GE_F-pos and GE_F-neg was observed by NMR with sharp signals corresponding to base-paired uridine imino protons.
Oligonucleotide binding was followed by chemical shift perturbation experiments in 1 H-15 N HSQC and 1 H-13 C HSQC experiments, as exemplified for the short polyA sequence in Figure 2 (panels A to C). The chemical shift variation profile in the 1 H-15 N HSQC spectrum of short polyA is shown in Figure 2D. Chemical shift perturbation profiles were similar for all RNAs and consistent with those observed for yeast RNA. A complete set of profiles for all oligonucleotides is given in Figure S2. For all RNAs the largest chemical shift changes were observed for residues belonging to the main basic cluster, i.e. K92-V97 (a2), L149-L152 (a5) and D155-K159 (a6), as illustrated in Figure 2E. These residues contribute to forming a continuous positively charged surface, which is consistent with an RNA binding surface ( Figure 2F). Chemical shift variations in 1 H-13 C HSQC experiments were observed for residues in the same region, solvent exposed methyl side chains in the hinge between helices a5 and a6 as well as R151-H d protons being affected by RNA binding (see Figure 2B and 2C).
Short A-rich RSV RNA sequences preferentially bind to M2-1  The fast to moderately fast exchange regime in both 1 H-15 N and 1 H-13 C HSQCs allowed to extract apparent dissociation  constants for a binding model with a 1:1 stoichiometry. The fitted chemical shift variation curves for individual residues in a fast exchange regime are given in Figure S3. The values of the apparent dissociation constants (K d ) are recapitulated in Table 1. They range from 2.5 mM to .600 mM and fall into two groups. Except for a short U-rich RNA sequence, K d s for U-rich negativesense RNAs are in the 75-125 mM range. UGA2, an RSVunrelated stem-loop RNA comprising a U-stretch and five base pairs, used as a control, binds with similar affinity (K d = 60 mM). A second control was carried out with a short single stranded DNA equivalent to the GE_F-neg RNA, showing that the same DNA sequence (250 mM) binds with slightly less affinity than the RNA sequence (85 mM). Apparent K d s for A-rich positive-sense RNA sequences (leader-pos, GE_SH-pos, GE_F-pos and short A-rich) are in the 2.5-22 mM range. A last experiment with the double-stranded F gene end (GE_F-ds, 4.5 mM) shows that the affinity is further increased as compared to the affinities of single-stranded RNAs. Taken together, these results indicate that M2-1  binds to RNA with lower affinity than reported for full-length M2-1 [10,13]. Importantly they show that the presence of A-rich stretches increases binding affinity as compared to U-rich stretches and that RNA base-pairing might also play a role. As shown above, helices a2, a5 and a6 form a uniform RNA binding surface. However, residues L74 and G75 in helix a1, which are located on the opposite negatively charged side of the protein, are also affected by RNA binding. For all RNAs, K d s determined for residues in helix a1 were always the same as those measured for residues of the main binding site. This strongly suggests that perturbations in a1 on the one hand and in a2, a5 and a6 on the other hand are related to the same binding event.  Table 1) are indicated with red spheres. Side chains of R151 and K92, for which RNA induces 13 Cd-1 Hd and 13 Ce-1 He chemical shift variations, are in stick representation. The surface formed by all these residues is colored in red. P153, for which no information is available from 1 H- 15  Moreover as the hypothetical second contact surface with RNA would be very distant from the main binding site, we exclude the possibility of a second binding site at a1. However RNA binding to a2 could be transmitted to a1 by slight alterations of helix-helix packing.
The P binding region of M2-1 58-177 is adjacent to the RNA binding site To compare RNA and P affinities for M2-1 58-177 , we performed isothermal titration calorimetry (ITC) experiments using this truncated form of M2-1. Results showed that the phosphoprotein P binds to M2-1 58-177 with a stoichiometry of 1:1 and a K d of ,3 mM ( Figure S4). We investigated this interaction further by NMR for atomic details. Perturbations induced by tetrameric fulllength P in 1 H-15 N and 1 H-13 C HSQC spectra of M2-1 58-177 were monitored. Due to the size of the complex and the unfavorable exchange regime, addition of 0,5 molar equivalents of P was accompanied by extensive overall line broadening in the 1 H-15 N HSQC spectra, except for the unstructured N-and C-termini (see Figure S5A). At lower P concentrations, the majority of crosspeaks remained detectable. Residues T130-L165 exhibited larger line broadening, suggesting that P binds to the a4/a5/a6 region. Transferred cross-saturation (TCS) experiments [23] were carried out with 2 H 15 N-M2-1 58-177 to more specifically identify residues involved in P binding ( Figure 3, panels A to D). Saturation of the methyl protons of P resulted in reduction of M2-1 cross peak intensities in the regions V127-S137 and L152-T164 (helices a4 and a6, and a5/a6 hinge). They form a nearly uniform surface on M2-1 ( Figure 3D). Since P residues 100-120 were reported to be critical for M2-1 binding [14,24], we also monitored intensity variations in 1 H-15 N HSQC spectra in the presence of a truncated form of P, P 100-166 , comprising this region as well as the oligomerization domain of P [25,26]. The results are shown in Figure 3 (panels E to H). Line broadening was enhanced for helices a4 and a6 ( Figures 3F and 3G), which is consistent with the results obtained with full-length P. In summary, P binds to a region proximal to the RNA anchoring surface. Helices a6 and a5 were sensitive to both P and RNA binding in our NMR experiments, and could thus contribute to both interaction sites, while a4 and a2 appear to be involved specifically in P binding and RNA binding, respectively.
Effects of surface mutations in the P and RNA binding regions on M2-1 controlled transcription probed by a minigenome Based on these results, we designed single residue mutations of the full-length M2-1 protein targeting the binding sites of P and RNA. Solvent-exposed residues were first substituted by Ala. The effect of these mutations on transcription antitermination by M2-1, which was previously shown to be the same for RSV-specific and heterologous sequences [6], was assessed using an RSV dicistronic subgenomic replicon, pM/Luc. It contains the authentic M/SH gene junction, and the Luc reporter gene downstream of the gene start sequence present in this gene junction. The expression of the Luc gene in this system is absolutely dependent on the presence of a functional M2-1 [15]. The pM/Luc plasmid was co-transfected in BHK-21 BSRT7/5 cells expressing T7 RNA polymerase together with p-b-gal, pL, pP, PN, and pM2-1. Luciferase activities were determined and normalized based on bgalactosidase expression [5,8,15,27,28]. Except for L148A and N163A, most of the Ala substitutions had only little effect on transcription, as assessed by Luc expression ( Figure 4A). As the Pand RNA-binding surface identified by NMR is highly positively charged, residues were substituted by Asp to emphasize the electrostatic effect of mutations. When substituted by Asp, K92 (in helix a2), a residue involved in RNA but not in P binding according to the NMR results, and R126 and T130 (in helix a4), two residues involved in P but not in RNA binding, appeared to be critical for transcription. A similar effect on transcription was observed for four other residues (K150, R151, T160 and N163, in helices a5 and a6), that belong to the dual RNA/P binding surface. In contrast, Ala and Asp mutants of K158, which is also located in the RNA/P binding region, still displayed more than 80% activity as compared to the WT ( Figure 4A). As a control, we further verified by NMR that the deactivating mutations did not disrupt the protein fold ( Figure S6).

Interaction of M2-1 surface mutants with P-N complexes in cellulo
Co-expression of P and N proteins (in the absence of other viral proteins) in cells induces the formation of cytoplasmic inclusion bodies containing P-N complexes, as observed in RSV infected cells [11]. When co-expressed with P and N, WT M2-1 also localizes preferentially in these IBs as seen in Figure 4D [11]. We analyzed the intracellular localization of the M2-1 mutants in cells co-transfected with expression vectors for P, N and M2-1 by fluorescence microscopy (Figure 4C and 4D). Contrary to WT M2-1, the mutants R126D, T130D, L148A, T160D and N163D were excluded from the IBs and spread all throughout the cytoplasm. In contrast K92D, K150D and R151D retained their localization to cytoplasmic IBs. These mutants were expressed at comparable levels, as determined by Western blotting (Figure 4B), and were correctly folded when purified from E. coli and analyzed by NMR ( Figure S6).

In vitro binding of M2-1 mutants to P and RNA
To verify that the residues identified using NMR and the minigenome assay are critical for RNA-and/or P-binding, the in vitro RNA and P binding capacities of eight M2-1 mutants selected by the Luc assay were investigated. As M2-1 did not migrate in native agarose gel, it was not possible to obtain electrophoretic mobility shift assays (EMSA) with the GST-free forms. We thus used M2-1 fused to GST for the in vitro binding assays with RNA and P. For RNA binding assays, we used either full-length (tetrameric) or truncated 58-177 (monomeric) forms of M2-1 fused to GST. GST-M2-1 58-177 , incubated with tRNA, was analyzed by EMSA. Formation of GST-M2-1 58-177 :RNA complexes was only impaired by the K92D, K150D and R151D mutations, which did not prevent M2-1 association with IBs ( Figure 5A and 5B). The effect of single mutations of the full-length form on in vitro RNAbinding was smaller than with truncated M2-1  , probably owing to the higher avidity of tetramers for RNA compared to monomers, and the Lys/Arg repetitions that limit the effect of single substitutions (data not shown). In vitro M2-1:P interactions were assessed by GST pull-down assays. As shown in Figure 5C and 5D, a$50% decrease in P binding was observed for the mutants R126D, T130D, L148A, T160D and N163D, which were excluded from IBs, but could still bind RNA. The two separate binding surfaces for RNA and P, determined from these 8 residues, are illustrated in Figure 5E. Together with the cellular localization experiments, these results show that there is a strong correlation between the reduced capacity of M2-1 to pull-down P in vitro and its exclusion from IBs. Conversely, since residues specifically involved in RNA binding appear not to be crucial for recruitment of M2-1 to IBs, the M2-1:RNA interaction is not required for this process.

Similarity between pneumovirus M2-1 and filovirus VP30 transcription co-factors
In contrast to other members of the Mononegavirales order for which transcription is driven by only three proteins (N/NP, P and L), pneumoviruses and filoviruses encode a fourth transcription cofactor (M2-1 and VP30 respectively). Parallels have been drawn between VP30 and M2-1 on the basis of their functions and domain organization [17,29]. Both were shown to be necessary for efficient transcription in reconstituted minigenome systems [7,30]. In addition, although VP30 and M2-1 were shown to be dispensable for RNA replication [30,31,32], both are required to rescue recombinant EBOV or Marburg virus and RSV [33,34,35]. M2-1 and VP30 contain an N-terminal Cys 3 His 1 motif that does not bind RNA directly, but that is essential for VP30 RNA binding [15,17,36,37]. In the case of M2-1 this motif is indispensable for function, but its exact role is still unknown [17,18,19]. Like M2-1, VP30 also contains an oligomerization domain downstream of the Cys 3 His 1 motif, which is necessary for its function during transcription [15,36]. But, despite functional similarities, the sequence identity between the two proteins is very low and it was not possible to know that they were evolutionary related.
Here we show that the globular domains M2-1 58-177 and EBOV VP30 CTD [38] are structurally homologous and that they display the same a-helix bundle fold. The structures were aligned on the DALI server [39] which yielded a Z-score of 5.7 and an rmsd of 3.9 Å for 92 aligned residues and 9% sequence identity (see Figure 6). The core helices a1, a2, a5 and a6 align well, with a slight shift of a1 M2-1 with respect to a1 VP30 . The a3-a4 hairpins are skewed relative to each other. Moreover M2-1 58-177 and VP30 CTD contain several identical hydrophobic residues in helices a5 and a6 (indicated in Figure 6), which are involved in stabilizing inter-helical contacts and appear to be semi-conserved among pneumovirus/metapneumovirus M2-1 and filovirus VP30 protein sequences ( Figure S7). Although they mainly contribute to interhelix packing, it is noteworthy that one of them (L148) is also critical for M2-1 transcription antitermination. VP30 CTD contains an additional 7th C-terminal helix, which has no counterpart in M2-1. It stabilizes the crystallographic VP30 CTD dimer by interaction with a1. Since full-length M2-1 forms tetramers with the oligomerization domain just upstream of the core domain, its monomeric state in isolation suggests that the core domains may at best be loosely associated with each other in the tetramer, in the absence of partner molecules, as schemed in Figure 1C.
The overall structural match rationalizes the relationship between M2-1 and VP30. Importantly both proteins associate with the nucleocapsid by means of their globular core domains [10,14,38,40]. Whereas the VP30 CTD :nucleocapsid interaction is mediated by the EBOV nucleoprotein but not by RNA [41], the M2-1:nucleocapsid interaction has been proposed to be mediated by RNA [10] and by P [14]. RNA binding to VP30 CTD has not been observed [38], in contrast to what we report for M2-1  . As no direct M2-1:N complex was evidenced in vitro [10,14], the correlation found between M2-1 localization to cytoplasmic IBs, containing N and P, and the capacity of M2-1 to bind P in vitro indicates that binding of M2-1 to P drives recruitment of M2-1 to the holonucleocapsid.
M2-1 58-177 binds to mRNA transcripts rather than to genomic RNA M2-1 was first described as an antitermination factor, preventing cessation of chain elongation and release of the nascent mRNA [7]. Later M2-1 was reported to inhibit transcription termination at the GE signals to produce polycistronic readthrough mRNAs [6,8,9], depending on GE sequences. The semi-conserved 12-13 nucleotide GEs fall into three groups with respect to this property: the first (NS1/NS2, NS2/N, M2/L, and L/trailer) contains sequences inefficient for transcription termination and is insensitive to M2-1; the second (N/P, P/M, M/SH, SH/G, G/F) is very efficient in transcription termination but not sensitive to M2-1; finally (F/M2) is highly sensitive to M2-1 [9]. M2-1 did not direct readthrough at the leader-NS1 junction [6]. Thus, it was suggested that M2-1 would not only prevent inappropriate intragenic termination, but also allow the polymerase to access to promoter-distal regions of the genome and to transcribe downstream genes [9]. Cuesta et al. reported that renatured M2-1 bound preferentially to a short positive-sense leader RNA in vitro [13], while Cartee and Wertz co-immunoprecipitated RSV mRNA with M2-1 from infected cells treated with Actinomycin D [10]. These observations suggested that M2-1 could modulate transcription termination by recognizing specific viral RNA sequences, either on transcribed mRNAs or on the genomic RNA template. Contrary to transcribed viral mRNA, genomic RNA is encapsidated. However if the model of RNA synthesis by the vesicular stomatitis virus, a prototype of nonsegmented negative-strand RNA viruses, could be transposed to RSV, genomic RNA would become accessible as the nucleocapsid is being locally disassembled by the transcription complex [42,43]. By comparing the affinities of M2-1 58-177 for short 10-15 nucleotide RSV RNA sequences, we found  (2) or in the presence of P (+), washed, and analyzed by SDS-PAGE. P was also run alone (lane P). (D) Coomassie bluestained gels were scanned and M2-1:P binding was quantified using ImageJ software and corrected for nonspecific binding to GST. Errors were estimated to 65%. (E) The surfaces formed by the 8 residues, for which mutants were analyzed, are indicated on the M2-1 structure according to their binding partner: in red (RNA binding) and blue (P binding). doi:10.1371/journal.ppat.1002734.g005 that M2-1 58-177 bound preferentially to positive-sense RNA sequences such as the 12 first bases of the positive-sense leader (K d = 2.5 mM), positive-sense GEs (11 and 13 mM) and an A-rich sequence located on the positive-sense leader (22 mM). Affinities of the negative-sense signal sequences were systematically lower by one order of magnitude or more than those of their complementary sequences. But there was no difference between the SH and F gene ends, which are respectively insensitive and highly sensitive to M2-1. The 39 and 59 extremities of the leader and the GS displayed a similar behavior to that of GEs of same polarity, indicating that M2-1  does not discriminate between transcription signals of same polarity. Sequence specificity of RSV transcription antitermination thus appears not to be linked to the M2-1 core domain. These results are also consistent with the observation that M2-1 is not required for initiation of RNA transcription [7]. Our results show that M2-1 58-177 displays a preference for purine-rich and especially A-rich RNAs found in positive-sense RSV RNAs over pyrimidine-rich sequences containing U-stretches found in negative-sense RNAs. Thus, they reinforce the hypothesis that M2-1 binds preferentially to positive sense RNA transcripts rather than to the template, in agreement with previous reports [10,13]. An additional finding is that M2-1 binds to the double-stranded F gene end with similar or better affinity (4.5 mM) than to the positive-sense sequence alone (11 mM). This implies that M2-1 could bind to the nascent mRNA transcript, either still bound to the template or released from it. In both cases, in order to facilitate transcription elongation, M2-1 could prevent formation of mRNA secondary structures that might destabilize the transcription complex, in analogy to the function of N protein that binds to the nascent RNA during replication.

Functional implications of the interaction of M2-1 with P
Using in vitro experiments, we have previously shown that RNA and P bind to M2-1  in a competitive way [15]. Here we have characterized two distinct binding surfaces for P and RNA by NMR ( Figure 5E). Their edges are partially overlapping. The P binding epitope, which is buried between helix a4 and the hinge between a5 and a6 ( Figure 3D and 3H), might be occluded by RNA binding in the vicinity, and conversely RNA binding might be hampered by bound P due to steric hindrance. Both binding sites are located on the positively charged face of M2-1  , underlining the importance of electrostatic interactions for association with RNA and with P. This hypothesis is also in agreement with the predicted negatively charged surface in computer models of the M2-1 binding region of the P tetramer, spanning residues 100-120 of P [44]. Redundancy of positive charges in the P and RNA binding region certainly accounts for the weak effect on Luc expression observed for Ala mutants of residues in this region in the minigenome assay, as compared to Asp substitution, which introduced opposite charges. Electrostatic interactions could be further emphasized in the M2-1 tetramer if the core domains were arranged to provide an extended binding surface consisting of two adjacent positively charged clusters. In addition to charge effects, specific hydrophobic interactions may also contribute, as suggested for the interaction between P and M2-1 in [14]. This could apply to RNA as well. Although the range of RNA affinities for M2-1 58-177 (micromolar to millimolar) is an indicator for non-specific binding mediated by the RNA phosphate backbone, corroborated by binding of a DNA sequence equivalent to GE_F-neg, the difference between A-rich and U-rich RNA affinities indicates that the nature of the bases also comes into play.
The large M2-1  :RNA binding surface determined by NMR coincides with the main positively charged cluster of M2-1   (Figure 1B and 2F), which is well conserved among Pneumovirinae ( Figure S7). We confirmed by mutagenesis that the three basic residues K92, K150 and R151, included in this epitope, are crucial for in vitro RNA binding to M2-1 58-177 and for transcription enhancement in vivo by full-length M2-1. However they did not prevent association with cytoplasmic IBs or P binding. In contrast, mutants R126D, T130D, L148A, T160D and N163D, for which M2-1-controlled transcription was impaired and which still bound RNA in vitro, had lost their ability to bind P in vitro and did not co-localize with N-P complexes in cytoplasmic IBs. These residues are not well conserved in Pneumovirinae M2-1 proteins ( Figure S7A), but this would be consistent with the sequence variability of the M2-1 binding region of P(100-120), which has co-evolved with M2-1. Altogether the NMR and mutagenesis results provide a rationale for the competitive binding previously observed between P and RNA [15]. They highlight the role of P for the recruitment of M2-1 to cytoplasmic IBs that contain N and P and where viral RNA synthesis takes place, by analogy to Rhabdoviridae [12].
Possible cooperative effects in full-length tetrameric M2-1, involving the N-terminal Cys 3 His motif and interactions between core domains, may induce an increased affinity for both P and RNA as well as sequence specific recognition of RNA, with respect to the core domain of M2-1. However the overall lower affinities of RNA for M2-1  , as compared to P, suggest that M2-1 binds P Figure 6. Structural alignment of RSV M2-1  and EBOV VP30 CTD . Cartoon representations of aligned RSV M2-1  (residues L74 to T172) and EBOV VP30 CTD (pdb 2I8B, [38]). Disordered N-and C-termini are not shown. Structural alignment with M2-1 58-177 was generated by the Dali server [39] (Z-score = 5.7; rmsd = 3.9 Å ; 92 aligned residues; 9% sequence identity). VP30 CTD helix a7, which has no counterpart in M2-1 58-177 , is not represented. Identical hydrophobic residues in helices a5 and a6 are represented with sticks for M2-1 58-177 and VP30 CTD . doi:10.1371/journal.ppat.1002734.g006 preferentially in the absence of viral RNA so that M2-1 would be recruited as an RNA-free M2-1:P complex to the IBs. This hypothesis is in agreement with our fluorescence microscopy observations, since WT M2-1 and mutants that do not bind RNA are found likewise in IBs containing N and P ( Figure 4C and 4D). Indeed, higher affinity to RNA would result in sequestration of M2-1 in the cytoplasm, where cellular RNA is highly accessible, and M2-1 would not be recruited to the IBs. The relatively low K d s are in the same range as those measured for the Measles virus P-N interaction [45], this intrinsically weak association being probably required for movement of the viral polymerase during RNA transcription.
In summary, our results indicate that not only P:M2-1, but also RNA:M2-1 interactions are required for efficient transcription activation by M2-1. Association with P is strictly required for recruitment to the viral RNA synthesis site. As suggested by the proximity of the binding sites, it is likely that the P:M2-1 interaction is displaced in favor of other interactions, RNA:M2-1 interactions in particular, in the context of the holonucleocapsid. The higher affinity of M2-1 58-177 for the 59 end of the positivesense Leader RNA (2.5 mM), which is in the same range as the affinity for P (3 mM), suggests that M2-1 could be loaded onto the polymerase during transcription initiation. It can further be speculated that an L:M2-1 interaction also takes place when the P:M2-1 interaction breaks down, thus altering the sensitivity of the polymerase to transcription termination signals. A similar L:VP30 interaction was recently reported for EBOV [46]. This hypothesis is currently under investigation for RSV.

Minigenome assay
BSRT7/5 cells stably expressing T7 RNA polymerase [47] were maintained in EMEM supplemented with 10% FCS/L-glutamine/ penicillin-streptomycin solution. The cells were grown in an incubator at 37uC under 5% CO 2 . pN, pP, pM2-1 and pL plasmids coding for HRSV (strain Long) N, P, M2-1 and L proteins respectively, under the control of the T7 promoter, have been described previously [15,48]. An encephalomyocarditis virus internal ribosome entry site (IRES) sequence was placed between the T7 promoter and the inserted ORF to enhance protein expression in BSR/T7-5 cells, as previously described [48,49]. The pM/Luc subgenomic replicon was derived from the pM/SH subgenomic replicon [8] and has been described previously [15]. It contains two transcription units, the second encoding the firefly luciferase (Luc) gene under the control of the M/SH intragenic sequence. The expression of the Luc gene in this system is absolutely dependent on the presence of M2-1 [15]. BSRT7/5 cells were transfected with pN, pP, pL, pM2-1, p-b-Gal coding for beta-galactosidase under the control of the Rous sarcoma virus promoter (Promega) and pM/Luc, where luciferase expression is controlled by the RSV M/SH intergenic region [15]. Luciferase activity was determined in triplicate 24 h post-transfection as previously described [15]. Cells were lysed in luciferase lysis buffer (30 mM Tris, pH 7.9, 10 mM MgCl 2 , 1 mM dithiothreitol [DTT], 1% [vol/vol] Triton X-100, and 15% [vol/vol] glycerol), and luciferase activities were evaluated twice for each cell lysate with an Anthos Lucy 3 luminometer (Bio Advance).

Analysis of in vitro RNA and P binding by M2-1 and M2-1 mutants
GST-M2-1 58-177 fusion proteins (WT or mutants, final concentration 100 mM) were eluted by GSH and incubated with yeast tRNA (Sigma, final concentration 50 mM) in a final volume of 10 ml. Complexes were resolved by 1.5% agarose gel electrophoresis in 16 Tris-Glycine buffer at 4uC, stained with ethidium bromide and amido black. GST pull-down of purified recombinant P by full-length GST-M2-1 fusion proteins (WT and mutants) was performed by incubating 10 ml aliquots of a 50% slurry of Glutathione-Sepharose 4B beads (GE Healthcare) containing ,25 mM GST-M2-1 in PBS with a 3-fold molar excess of P for 1 h at 20uC under agitation. Beads were washed extensively with PBS, boiled in 25 ml Laemmli buffer and analyzed by SDS-PAGE and Coomassie blue staining. Bands were quantified using ImageJ software. Affinity between P and M2-1 was determined by isothermal titration calorimetry.

Isothermal titration calorimetry assays
Raw ITC data were processed and quantitative analysis of the P-M2-1  interaction was performed on a MicroCal ITC200 microcalorimeter (Microcal, Northampton, MA). Samples were dialyzed against 16 PBS for 15 h. The experiments were carried out at 20uC. The P concentration in the microcalorimeter cell (1.4 mL) was of 33 mM. In total, 20 injections of 2 mL of M2-1  solution (concentration 335 mM) were carried out at 180 s intervals, with stirring at 1000 rpm. The experimental data were fitted to a theoretical titration curve with software supplied by MicroCal (ORIGIN). This software generates titration curves based on the relationship between the heat generated by each injection and DH (enthalpy change in kcal/mol), K d (dissociation constant), n (number of binding sites), the total protein concentration and free and total ligand concentrations. Gene Assembler Plus using phenoxyacetyl b-RNA phosphoramidites and Universal Support resin (Glen Research) with a protocol adapted for RNA from DNA solid-phase synthesis [51]. The products were HPLC-purified (Beckman) by anionic chromatography on a DEAE-Sepharose column (Pharmacia) in 10 mM phosphate buffer pH 6.8 and eluated with a linear gradient from 0.55 to 1 M NaCl in 80 min. The oligonucleotides were extensively dialyzed in water. If necessary the pH was adjusted to 7 with 0.5 M NaOH and the concentration determined using the A260 value. Aliquots were then lyophilized and stored at 220uC. The oligonucleotide sequences are given in Table 1. Folding properties of RNAs were estimated by using the mfold web server [22].

NMR structure calculation
M2-1 58-177 structures were calculated with CYANA 2.1 [55], using distance restraints obtained from 15 N-and 13 C-NOESY-HSQC spectra and backbone torsion angles generated with TALOS. They were refined in Xplor-NIH [56] using 1 D HN residual dipolar couplings measured in a stretched polyacrylamide gel and a hexanol/C12E5 phase. Structure statistics are summarized in Table S1. More details are provided in Text S1. Graphic representations were performed with PyMOL [57].
NMR M2-1 58-177 chemical shift and cross-peak intensity variation experiments with RNA and phosphoprotein Chemical shift and cross-peak intensity perturbations experiments were carried out with yeast RNA (Roche) and short synthesized oligonucleotides (see Table 1). 0.25-20 equivalents of lyophilized RNA were added stepwise to 50-60 mM 15 N-, 13 C-or 15 N 13 C-labeled M2-1  . 1 H-15 N and 1 H-13 C HSQC spectra were recorded at 293 K at each step, at magnetic fields of 14.1 T and/or 22.3 T.
Interaction with phosphoprotein was probed by cross-peak intensity perturbations in 1 H-15 N and 1 H-13 C HSQC experiments, with full-length P (P 1-241 ) and truncated P 100-166 . Samples with P 1-241 were prepared by mixing concentrated solutions of P 1-241 (30 mM) and 15   r .
The apparent dissociation constant K d was obtained by fitting 1 H or 15 N chemical shift differences at each titration point with two parameters in a binding model with 1:1 stoichiometry and with a user-defined function in Origin 7 software as follows:

Transferred cross-saturation experiments
Samples were prepared with a 30 mM solution of P 1-241 centrifuged at 15 krpm in a TLA55 rotor (Beckman) for 15 min and exchanged into D 2 O buffer with biospin columns according to the manufacturer's instructions (Biorad). Transferred cross-saturation experiments [23] were conducted with 150 mM 2 H 15 N labeled M2-1  in 91% D 2 O in the presence of 15 mM (10%) unlabeled P 1-241 protein. Broadband proton saturation (2 s) was achieved with a 1.8 kHz Wurst pulse centered at 0.8 ppm. Spectra were recorded in interleaved fashion with and without saturation, with 1.5 s recycling delay, at 14.1 T and 293 K. Intensity ratios were determined based on experiments with and without saturation. Control experiments were carried out without P to account for spin diffusion and effects of residual aliphatic protons in M2-1  .

Data deposition
Atomic coordinates and structural constraints have been deposited in the Protein Data Bank (PDB accession code 2L9J).