Investigation of the Fuzzy Complex between RSV Nucleoprotein and Phosphoprotein to Optimize an Inhibition Assay by Fluorescence Polarization

The interaction between Respiratory Syncytial Virus phosphoprotein P and nucleoprotein N is essential for the formation of the holo RSV polymerase that carries out replication. In vitro screening of antivirals targeting the N-P protein interaction requires a molecular interaction model, ideally consisting of a complex between N protein and a short peptide corresponding to the C-terminal tail of the P protein. However, the flexibility of C-terminal P peptides as well as their phosphorylation status play a role in binding and may bias the outcome of an inhibition assay. We therefore investigated binding affinities and dynamics of this interaction by testing two N protein constructs and P peptides of different lengths and composition, using nuclear magnetic resonance and fluorescence polarization (FP). We show that, although the last C-terminal Phe241 residue is the main determinant for anchoring P to N, only longer peptides afford sub-micromolar affinity, despite increasing mobility towards the N-terminus. We investigated competitive binding by peptides and small compounds, including molecules used as fluorescent labels in FP. Based on these results, we draw optimized parameters for a robust RSV N-P inhibition assay and validated this assay with the M76 molecule, which displays antiviral properties, for further screening of chemical libraries.


Introduction
Respiratory Syncytial virus (RSV) is the most common pathogen for acute pediatric low respiratory tract infections (ALRI) and bronchiolitis [1]. RSV is an important cause of death in infants in developing countries and a substantial burden on healthcare systems and hospitals worldwide. In 2005, RSV led to~33 million RSV-associated ALRI episodes, 3 million hospitalizations and 55,000-199,000 deaths in children younger than 5 years [2,3]. RSV also represents a still underestimated risk of severe infection and mortality for immunocompromised and elderly persons [4].
There is still no licensed human RSV vaccine, even after more than 6 decades of research. The monoclonal antibody Palivizumab, which is not efficient against ongoing infection, has been administered for prophylaxis only to high-risk infants, due to high cost and limited efficacy [5][6][7]. Very recently Nirsevimab, a long-acting antibody for similar applications [8], was approved by the European Medicines Agency. More than 30 RSV prevention candidates are currently in clinical development [9,10]. The licensed therapeutic arsenal against RSV is even more limited. Ribavirin, a nucleoside analogue, has Since the N NTD -P CTD interaction is essential for the formation of the viral holo RNA polymerase complex and hence for RSV replication, we proposed that the RSV N NTD -P CTD protein-protein interaction (PPI) could be a new drug target for PPI inhibitors. We previously showed that P CTD could be displaced in vitro from N NTD by 1-benzyl pyrazole 3,5-dicarboxylate derivatives, which mimic the sidechain and C-terminal carboxylate of Phe 241 ( Figure 1C) [33]. After chemical modification to improve its membrane passage, the M76 molecule ( Figure 1D) was able to inhibit replication of recombinant fluorescent RSV in cells. Results from other groups reinforce this hypothesis. Hesperitin, a flavanone that reduced intracellular RSV replication [40], was also shown to bind to the P-binding pocket of N NTD and to compete with the 11-mer P 11 peptide [41]. The EDP-938 benzodiazepine, which demonstrated antiviral activity, elicited mutations on N close to the P-binding pocket, suggesting that it might be an N-P inhibitor [42].
Hesperitin and M76, which compete with P for N-binding in vitro and display antiviral activity [33,40,41], were identified from a limited set of molecules and different screening approaches: hesperitin was identified from a flavonoid series using an antiviral assay, whereas M76 was obtained from a subset of the ZINC database selected in silico with a stringent filter based on chemical structure. High-throughput screening (HTS) of chemical libraries would enable to more widely explore the potential of the RSV N NTD -P CTD PPI as an antiviral target. This requires a relevant model of the interaction combined to a robust HTS method. We have shown previously that the complex between N NTD and short C-terminal P peptides, derived from P Ctail , affords a minimal complex mimicking the interaction between the P protein and the RSV nucleocapsid [33]. Fluorescence anisotropy (FA) [43], or equivalently fluorescence polarization (FP) (see Materials and Methods), are compatible with HTS [44]. FP has already been applied in the case of the RSV N-P interaction to probe inhibition by hesperitin [41] and by P Ctail -derived peptides [45]. These FP measurements were performed under different experimental conditions, notably with different N protein constructs, different P peptides and different fluorescent labels. In this context, our aim was to obtain a deeper insight into the experimental requirements of an FP assay specific of the RSV N-P interaction. This led us to analyze the fuzzy complex formed between the N protein and the flexible P Ctail .

Phosphorylation of an RSV P Ctail -Derived Peptide Increases Affinity for RSV N NTD
We previously analyzed the RSV N NTD -P 12 complex using Nuclear Magnetic Resonance (NMR) [33]. As the isoelectric point of N NTD is 7.8, we used MES pH 6.5 buffer to ensure the stability of the protein. This pH is also suitable for amide-detected 2D NMR experiments. The buffer contained salt (250 mM NaCl) to further stabilize N NTD at protein concentrations required for NMR (50-100 µM). Under these experimental conditions we determined a K d of 54 µM.
Here, as P protein is constitutively phosphorylated on Ser 232 and Ser 237 residues in cells, we first investigated the impact of phosphorylation on the affinity of the N NTD -P Ctail complex. We performed an NMR titration experiment by adding increasing amounts of pP 11 peptide, where Ser 232 and Ser 237 residues were replaced by O-phosphoserines (pSer, Figure 1A), to a solution of 15 N-labeled N NTD , and measured 2D 1 H-15 N HSQC spectra ( Figure 2A). We worked in the same buffer as with P 12 . At pH 6.5, the pSer sidechains are expected to be mostly negatively charged [46]. Addition of pP 11 induced perturbations in the HSQC spectrum of N NTD , in particular chemical shift perturbations (CSPs). Saturation of this effect was reached after addition of~6 molar equivalents of peptide. At this peptide:protein ratio, the protein can be considered to be in a fully bound form. We measured combined 1 H and 15 N amide CSPs (Equation (1)) for N NTD signals in a fast chemical exchange regime, i.e., signals that shifted in the spectrum without broadening at intermediate titration points (e.g., Ile 129 in Figure 2A). Significant CSPs were located in and around the P-binding pocket, confirming that pP 11 binds to the previously determined P-binding epitope on N NTD ( Figure 2B,C) [33]. We fitted the NMR titration curves of these residues with a single site binding model to determine residue-specific K d s (Equation (2)), and calculated a mean K d of 17.0 ± 4.5 µM from 26 individual values. Compared to the unphosphorylated P 12 peptide, this represents a~3-fold gain in affinity, which confirms the role of negative charges in P Ctail for N-binding. Since measurements were made in a buffer with high ionic strength, the gain remains rather modest due to charge screening.
gime. These residues were also located in or in the close vicinity of the P-binding pocket, and thus report on the same binding mode as residues in fast exchange. From chemical shift differences measured for these residues we determined an exchange rate kex of ~800 s −1 (from Equation (3)) between free and bound NNTD, which indicates a fast dynamic process. Since kex (koff + kon × [NNTD]) is a combination of association (kon) and dissociation (koff) rates, the relative contributions of dissociation and association could not be assessed by this method.  11 peptide. The magnetic field was 14.1 T, and the temperature 20 • C. The color code corresponds to the pP 11 :N NTD molar ratio as indicated below the spectrum. (B) Bar diagram of combined amide chemical shift perturbations (∆δ HN ) measured for N NTD in the presence of 4.2 molar equivalents of pP 11 . Regions with significant CSPs are highlighted with a colored background: C-terminus of the αN1 helix (blue), β-hairpin (green), center of the αI2 helix (magenta), η1 loop (cyan), H 151 -loop (yellow). (C) Mapping of CSPs onto the X-ray structure of N NTD (PDB 4ucc): amide nitrogen atoms are represented as red spheres for residues with |∆δ HN | > mean + 2 SD, and orange spheres for |∆δ HN | > mean + SD. The color code is the same as in (B). The same view in surface representation highlights the P CTD binding pocket, indicated by a white arrow.
Overall, signals of bound N NTD were broader than those of unbound N NTD (Figure 2A). This can be explained by the increase of molecular weight, when the complex is formed. Line broadening at saturation was more marked for residues belonging to the P-binding site, suggesting exchange phenomena as additional sources of broadening. They may reflect local mobility of P in the complex and intermediate binding states. Some N NTD signals with large chemical shift differences between bound and unbound states (e.g., Y 135 in Figure 2A) were broadened at intermediate titration points. Signals were recovered when saturation was reached, indicating that they are in an intermediate exchange regime. These residues were also located in or in the close vicinity of the P-binding pocket, and thus report on the same binding mode as residues in fast exchange. From chemical shift differences measured for these residues we determined an exchange rate k ex of~800 s −1 (from Equation (3)) between free and bound N NTD , which indicates a fast dynamic process. Since k ex (k off + k on × [N NTD ]) is a combination of association (k on ) and dissociation (k off ) rates, the relative contributions of dissociation and association could not be assessed by this method.

Fluorescence Polarization Reveals a Potential Secondary Binding Site on RSV N NTD for Fluorescein-Labeled RSV P 11 Peptide
2D Protein-observed NMR is a powerful tool that gives detailed structural and thermodynamic insight, but requires large sample amounts and is time-consuming. Acquisition of a single titration point in Figure 2 took 15 min. It is therefore not adapted for medium to high-throughput assays. Fluorescence polarization (FP) is more time-effective. FP is routinely used in screening assays for binding inhibitors [43,44,47,48]. We thus tested if we could transpose the experimental NMR conditions to FP measurements. We first wondered if the N NTD construct, used for NMR, but also previously for isothermal titration calorimetry and surface plasmon resonance (SPR) experiments [33], was adapted for FP. As a general rule, fluorescence polarization increases with the apparent size of a fluorescent molecule. According to Perrin's equation (Equation (5)), the theoretical FP of fluorescein (fluorescence lifetime~4 ns) bound to N NTD (26.2 kDa) is~400 mFP. This is close to 500 mFP, the theoretical maximal FP value, and indicates that the size of N NTD is compatible with an FP assay.
We started by labeling the pP 11 peptide with N-terminally attached BODIPY FL, as described by Shapiro at al. [45]. However, we observed that the BODIPY FL-labeled peptide (BF 2 pP 11 ) was prone to degradation, either via peptide bond cleavage or BODIPY FL-defluorination, under slightly acidic conditions. We then produced F 11 , another labelled peptide P 11 , by changing the fluorescent label to fluorescein-5-thiourea (FTU) by coupling with fluorescein-5-isothiocyanate (FITC). F 11 proved to be stable. It is similar to the 6-carboxyfluorescein P peptide used by Sa et al. for competition with hesperitin [41]. To directly compare FP and NMR data, we made measurements in MES pH 6.5 buffer. Brij-35 detergent (0.01% w/v) was added below its critical micellar concentration (0.11% w/v) to prevent spurious binding of the peptide to the reading plates. To determine the binding affinity of F 11 peptide for N NTD , we measured FP for 200 nM F 11 by varying the concentration of N NTD , which ranged from 0.1 to 120 µM ( Figure 3A). We tested buffer with and without salt. In buffer without salt and in the presence of NNTD, F11 FP stabilized only after 30 min. Binding curves were obtained, but the FP data did not reach a plateau at high NNTD concentration. Instead, FP displayed a linear dependence with increasing NNTD:F11 molar ratios, suggesting a second binding mode with low affinity ( Figure 3B). We fitted the curves with a combination of a single binding site for the first mode and a linear contribution for the second mode. This yielded ΔFPmax = 175 mFP and Kd = 0.6 ± 0.1 µM. ΔFPmax was of the same order of magnitude in buffers with and without salt, indicating that the dynamic range of FP was not affected by salt. This suggests that the dynamics of the fluorescent Figure 3. Binding of F 11 to RSV N NTD and inhibition by M76 assayed by fluorescence polarization. The buffer was MES pH 6.5, with 250 mM NaCl or without salt. The concentration of the fluorescent probe was 200 nM. (A) Binding of N NTD to F 11 in buffer with salt: ∆FP = FP − FP min (the FP of free F 11 was FP min = 30 mFP) was plotted as a function of the N NTD :F 11 molar ratio. Error bars represent the difference between duplicate experiments. Data were fitted with a single site binding model (straight line) using K d and ∆FP max as parameters. R 2 indicates the quality of the fit. (B) Binding of N NTD to F 11 in salt-free buffer assayed by ∆FP. The area in the dotted contour box is expanded on the right side. Data were fitted with a single binding site model and an additional linear contribution. (C) Inhibition of N NTD (5 µm) binding to F 11  In the presence of salt, FP measurements were stable for incubation times up to 48 h. FP of free F 11 was 30-40 mFP. We fitted the binding curves with a single site binding model (Equation (10)). This yielded a maximal FP difference between bound and free F 11 peptide ∆FP max = 210-230 mFP and a dissociation constant K d = 90 ± 15 µM ( Figure 3A). Saturation of the binding curve could not be obtained within the applied N NTD concentration range, and this may induce an error on K d determination. However, the K d has the same order of magnitude as that determined by NMR for the P 12 peptide under similar buffer conditions (K d = 54 µM) [33]. The experimental FP max value was lower than the theoretical value (~400 mFP), suggesting that the fluorescent label displays some mobility with respect to the complex. As the label is attached to the N-terminus of the peptide and hence to the opposite side of the anchoring Phe 241 residue, this would be in line with a fuzzy N NTD -P CTD complex, where only the C-terminal Phe 241 residue is well bound, whereas the rest of the P peptide remains disordered and only loosely attached.
In buffer without salt and in the presence of N NTD , F 11 FP stabilized only after 30 min. Binding curves were obtained, but the FP data did not reach a plateau at high N NTD concentration. Instead, FP displayed a linear dependence with increasing N NTD :F 11 molar ratios, suggesting a second binding mode with low affinity ( Figure 3B). We fitted the curves with a combination of a single binding site for the first mode and a linear contribution for the second mode. This yielded ∆FP max = 175 mFP and K d = 0.6 ± 0.1 µM. ∆FP max was of the same order of magnitude in buffers with and without salt, indicating that the dynamic range of FP was not affected by salt. This suggests that the dynamics of the fluorescent probe were similar in these two conditions. In contrast, the affinity was significantly higher in the absence of salt, by 2 orders of magnitude, confirming the importance of electrostatic interactions for the N NTD -P 11 complex.
We previously showed by SPR that the M76 molecule ( Figure 1D) could compete with P for N NTD binding in vitro [33]. We therefore tested competition between M76 and F 11 by FP. Measurements were done in salt-free buffer to take advantage of the higher F 11 -N NTD affinity. The concentration of N NTD was set to 5 µM to benefit from a high range of FP values, comprised between FP min and~80% FP max [48]. With increasing M76 concentration (up to 100 µM), F 11 FP progressively decreased due to F 11 displacement from N NTD by M76 ( Figure 3C). However, even at high M76 concentration, FP did not reach the FP min value. Saturation occurred at a high FP of~105 mFP, suggesting that M76 was not able to fully displace F 11 . This corroborates the hypothesis of a secondary binding site on N NTD , for which M76 does not compete. We fitted the data with a Hill equation (Equation (12)), assuming that the maximal inhibition of F 11 binding to the canonical site by M76 was obtained at saturation. We obtained an IC 50 of 28 ± 5 µM. This was converted into an inhibition constant K i of 2.5 µM using the IC50-to-Ki server [49], assuming a competitive mechanism and following values: K d = 0.6 µM for F 11 , [N NTD ] = 5 µM, and [F 11 ] = 0.2 µM.

The Complex betweeen Fluorescein-Labeled C-Terminal P Peptides and Full-Length N Protein Provides a Robust Model of the RSV N-P Interaction for FP Measurements
As the FP data obtained with N NTD in salt-free buffer suggest the existence of a secondary binding site, we wanted to test if this was still the case with full-length RSV N protein. Wild-type N is produced in the form of ring shaped N-RNA complexes by overexpression in E. coli [32]. These rings mimic one turn of the helical nucleocapsid, which is the natural partner of P Ctail . We therefore chose to test N-RNA as an alternative N form. N-RNA complexes contain 10-11 protomers and 7 ribonucleotides per N protomer [32]. Due to the high molecular weight of N-RNA rings (~0.5 MDa), a high fluorescence polarization is expected.
We carried out FP measurements using the fluorescent F 11 peptide and N-RNA complex. As N-RNA is more stable at pH 8.0 than at pH 6.5, we worked in Tris buffer. No salt was added to promote high affinity. The FP binding data could be fitted with a single site binding model ( Figure 4A). We obtained ∆FP max = 125-130 mFP, which is less than determined with the N NTD construct, but again confirms the relative mobility of the fluorescent label. We determined a K d of 7.6 ± 0.9 µM. Surprisingly, the K d value obtained with N-RNA was one order of magnitude higher than with N NTD (0.6 µM), suggesting that binding to N NTD may not fully recapitulate binding to N-RNA.
plex. As N−RNA is more stable at pH 8.0 than at pH 6.5, we worked in Tris buffer. No salt was added to promote high affinity. The FP binding data could be fitted with a single site binding model ( Figure 4A). We obtained ΔFPmax = 125-130 mFP, which is less than determined with the NNTD construct, but again confirms the relative mobility of the fluorescent label. We determined a Kd of 7.6 ± 0.9 µM. Surprisingly, the Kd value obtained with N−RNA was one order of magnitude higher than with NNTD (0.6 µM), suggesting that binding to NNTD may not fully recapitulate binding to N−RNA. We next performed inhibition experiments for the F11−(N−RNA) complex by M76 ( Figure 4B). FP values decreased with increasing competitor concentration. At saturation, FP nearly reached the value of free F11, showing that the F11 peptide could be nearly completely displaced. Since the same concentration range was used for NNTD and N−RNA, poor solubility of M76 can be ruled out to explain the saturation observed before complete displacement of F11 in the presence NNTD ( Figure 3C). This reinforces the hypothesis of a second binding site on NNTD, outside the canonical P-binding site. We fitted an IC50 of 12.6 µM ( Figure 4B), which converts into a Ki of 2.4 µM (using Kd = 7.6 µM for F11, [N] = 10 µM and [F11] = 0.2 µM). The inhibition constant is similar to that obtained with NNTD in saltfree buffer, which suggests that M76 binds in a similar way to the P-binding site on NNTD and N−RNA.
M76 was dissolved in organic solvent such as ethanol or DMSO. This may affect the stability of the protein. We thus run a control experiment to evaluate the effect of solvent on FP measurements. We used the FTU-P7E peptide F7E ( Table 1). Addition of up to 10% ethanol did not significantly affect FP measurements ( Figure 4C). A slight decrease of FP was observed with 10% DMSO, but overall the FP assay is robust with respect to solvent. In summary, the N−RNA form appears to be suitable for FP measurements, as only specific binding of fluorescent P peptide and efficient inhibition by the M76 molecule were observed. Table 1. FTU-labeled RSV P peptides (Fn) designed for fluorescence polarization measurements with RSV N−RNA and binding parameters obtained from FP curve fits. ϕ designates the FTU label. Phosphomimetic peptides, where Ser232 and/or Ser237 were mutated into glutamates, are annotated with the letter E. FP measurements were done with 200 nM fluorescent probe in Tris pH 8.0 buffer We next performed inhibition experiments for the F 11 -(N-RNA) complex by M76 ( Figure 4B). FP values decreased with increasing competitor concentration. At saturation, FP nearly reached the value of free F 11 , showing that the F 11 peptide could be nearly completely displaced. Since the same concentration range was used for N NTD and N-RNA, poor solubility of M76 can be ruled out to explain the saturation observed before complete displacement of F 11 in the presence N NTD ( Figure 3C). This reinforces the hypothesis of a second binding site on N NTD , outside the canonical P-binding site. We fitted an IC 50 of 12.6 µM ( Figure 4B), which converts into a K i of 2.4 µM (using K d = 7.6 µM for F 11 , [N] = 10 µM and [F 11 ] = 0.2 µM). The inhibition constant is similar to that obtained with N NTD in salt-free buffer, which suggests that M76 binds in a similar way to the P-binding site on N NTD and N-RNA.
M76 was dissolved in organic solvent such as ethanol or DMSO. This may affect the stability of the protein. We thus run a control experiment to evaluate the effect of solvent on FP measurements. We used the FTU-P 7E peptide F 7E ( Table 1). Addition of up to 10% ethanol did not significantly affect FP measurements ( Figure 4C). A slight decrease of FP was observed with 10% DMSO, but overall the FP assay is robust with respect to solvent. In summary, the N-RNA form appears to be suitable for FP measurements, as only specific binding of fluorescent P peptide and efficient inhibition by the M76 molecule were observed. Table 1. FTU-labeled RSV P peptides (F n ) designed for fluorescence polarization measurements with RSV N-RNA and binding parameters obtained from FP curve fits. φ designates the FTU label. Phosphomimetic peptides, where Ser 232 and/or Ser 237 were mutated into glutamates, are annotated with the letter E. FP measurements were done with 200 nM fluorescent probe in Tris pH 8.0 buffer without salt. K d and ∆FP max were determined by fitting binding curves to a single site binding model or a combination with a linear contribution (lines marked with *).

Measurement of the Binding Affinity of RSV Fluorescein-P Peptides for RSV N-RNA and Influence of Peptide Length on Fluorescent Label Mobility
Fluorescent labels can only be attached to the N-terminus of the P peptides, since the C-terminus of P directly anchors onto the N protein via the Phe 241 residue ( Figure 1C). Since C-terminal P peptides are flexible [26], the fluorescent label displays relative mobility, as reflected by the moderate ∆FP max values reported in the previous paragraphs. This raises the question about how the P peptide length impacts binding affinity and FP max . We thus designed a series of eight P peptides with increasing length, from 3 to 11 amino acids, and labeled them with FTU (Table 1). Moreover, to investigate the effect of phosphorylation, we produced phosphomimetic peptides, where either Ser 237 alone or both Ser 232 and Ser 237 were replaced by glutamates.
We first measured binding of F n peptides (200 nM) to N-RNA in Tris buffer by FP ( Figure 5A,B). For F n peptides with n ≥ 5, saturation of fluorescence anisotropy could be reached within the concentration range used for N-RNA, i.e., with 40 µM of total N protein concentration. The fully bound state of F 3 and F 4 peptides could not be obtained. We fitted the data with ∆FP max and K d as parameters. We observed that ∆FP max decreased with the length of the peptide ( Table 1), indicating that the dynamics of the fluorophore increase with its distance to the Phe 241 anchor. Concomitantly, the affinity of F n peptides increased with peptide length: K d s decreased from 62 µM (F 3 ) to <1 µM (F 11SE ) ( Table 1), highlighting the contribution of residues other than the two C-terminal P residues for N binding. In the 11-mer peptide series, phosphomimetic substitutions increased the affinity of F 11 , 9-fold for F 11SE and 20-fold for F 11EE ( Figure 5B, Table 1). A closer inspection of the binding curves shows that F 11SE and F 11EE seem to display a second binding mode of weak affinity, which becomes visible once the canonical binding site has been saturated. For F 11SE this mode could be fitted with a linear contribution ( Figure 5B, Table 1).
To validate fluorescent peptide binding to N-RNA, we carried out competition experiments with unlabeled P peptides. We used 10 µM of N protein. We first probed competition of the F 11 peptide by three P peptides of different lengths: P 7E , P 9E and P 11EE ( Figure 5C). These phosphomimetic peptides were able to fully displace F 11 , with IC 50 values ranging from 22 µM for the longest P 11EE peptide to 180 µM for the shortest P 7E peptide ( Table 2). The trend of IC 50 suggests that the efficiency of P peptides in competing with F 11 increases with peptide length. This is in line with the increase of affinity observed with increasing length of fluorescent P peptides (Table 2). Inhibition constants, calculated with K d values determined for F 11 , ranged from 6 to 75 µM ( Table 2). The trend of K i was the same as that of IC 50 . It is noteworthy that K i s of unlabeled peptides were higher than the K d s of their FITC-labeled counterparts, suggesting that the fluorescent peptides may display higher affinities than the unlabeled ones. Table 2. Inhibition parameters obtained by FP for FTU-labeled RSV P peptides (F n ) in complex with RSV N-RNA. Samples were in Tris pH 8 buffer without salt. IC 50 values were determined from inhibition experiments with 10 µM N-RNA and 200 nM fluorescent probe. Fitting of the inhibition data was done with a Hill equation. K i was calculated from IC 50 and K d using the IC50-to-Ki server [49].   To validate fluorescent peptide binding to N−RNA, we carried out competition experiments with unlabeled P peptides. We used 10 µM of N protein. We first probed competition of the F11 peptide by three P peptides of different lengths: P7E, P9E and P11EE ( Figure  5C). These phosphomimetic peptides were able to fully displace F11, with IC50 values ranging from 22 µM for the longest P11EE peptide to 180 µM for the shortest P7E peptide ( Table  2). The trend of IC50 suggests that the efficiency of P peptides in competing with F11 increases with peptide length. This is in line with the increase of affinity observed with increasing length of fluorescent P peptides ( Table 2). Inhibition constants, calculated with Kd values determined for F11, ranged from 6 to 75 µM ( Table 2). The trend of Ki was the same as that of IC50. It is noteworthy that Kis of unlabeled peptides were higher than the Kds of their FITC-labeled counterparts, suggesting that the fluorescent peptides may display higher affinities than the unlabeled ones. Table 2. Inhibition parameters obtained by FP for FTU-labeled RSV P peptides (Fn) in complex with RSV N−RNA. Samples were in Tris pH 8 buffer without salt. IC50 values were determined from inhibition experiments with 10 µM N−RNA and 200 nM fluorescent probe. Fitting of the inhibition data was done with a Hill equation. Ki was calculated from IC50 and Kd using the IC50-to-Ki server [49]. We next made competition experiments between three fluorescent and unlabeled peptides with identical amino acid sequences: F 5E vs. P 5E , F 7E vs. P 7E and F 11 vs. P 11 ( Figure 5D, Table 2). The unlabeled peptides were able to fully displace their fluorescent counterparts. IC 50 and K i values decreased with increasing peptide length, underlying the importance of P peptide length for inhibition of the N-P complex. Comparison of P 11 with P 11EE for inhibition of F 11 binding (Table 2) showed that P 11 competed less efficiently than P 11EE with an IC 50 of 86 versus 22 µM (K i of 34 versus 6 µM). This confirms the role of additional negative charges for formation of the RSV N-P complex.
Finally, we probed inhibition by the M76 molecule of F 11 and F 7E in complex with N-RNA. M76 IC 50 and K i values were in the µM range, but still lower than those of any tested unlabeled P peptide, indicating that M76 is a more potent inhibitor than the peptides, including P 11EE . Moreover, M76 appears to be more potent to inhibit the N-F 11 complex than the N-F 7E complex (K i of 2.6 versus 8 µM). Although F 11 and F 7E cannot be directly compared because of their different lengths, this finding suggests that M76 is less efficient at competing with a phosphomimetic peptide. A rationale for this would be that M76 directly competes with Phe 241 , but not with Ser/Glu 237 that binds outside of the cavity.
It must be noted that measurements with N NTD and N-RNA were done at different pH. Fluorescein fluorescence depends on pH, since the phenols and carboxylic acid groups can be ionized in a pH-dependent manner [50]. At pH 6.5, fluorescein dianion and monoanion forms are in a 1:1 equilibrium, whereas at pH 8.0 the dianion is predominant [51]. At pH 6.5 fluorescence emission at 490 nm is thus decreased by 30-40% as compared to pH 8 [52]. Nevertheless, FP should not be affected by pH, since fluorescence polarization is a ratio of emitted fluorescence intensities (Equation (4)). However, lower fluorescence leads to a lower signal-over-noise ratio at pH 6.5. In conclusion, N-RNA in a pH 8.0 buffer appears to be better suited for an FP inhibition assay than N NTD in a pH 6.5 buffer.

Fluorescein Binds to the RSV P-Binding Site on RSV N Protein
Although unlabeled P peptides were able to displace their fluorescent counterparts from RSV N-RNA, we wondered if the difference between the K i of F n peptides and the K d of unlabeled peptides could originate in N-RNA binding of the fluorescein label. We therefore measured fluorescein FP in the presence of N-RNA. We obtained a binding curve ( Figure 6B). Complete saturation of fluorescein by N-RNA was not reached in the N concentration range used for the experiment. We tentatively extracted a K d of 91 µM (Table 3) from these binding data, assuming that FP max reaches the maximal theoretical value (500 mFP). The binding affinity was lower than that measured for F n peptides, but still significant. The contribution of fluorescein binding to FP could be non-negligible, especially for the shorter F 3 and F 4 peptides, for which we determined K d s of 62 and 35 µM, respectively (Table 1). However, the affinity of fluorescein was lower by at least of one order of magnitude as compared to peptides longer than F 5E . Hence, for these peptides, the fluorescent label is not expected to strongly compete with the peptide part. To test if fluorescein binds to the P-binding site on the N-RNA, we carried out a competition experiment with M76 ( Figure 6C). M76 was able to displace fluorescein rather efficiently, with an IC 50 of 27 µM and a K i of 13 µM ( fluorescein binds to the P-binding site on the N-RNA, we carried out a competition experiment with M76 ( Figure 6C). M76 was able to displace fluorescein rather efficiently, with an IC50 of 27 µM and a Ki of 13 µM (Table 3), indicating that fluorescein indeed targets the P-binding site of the N protein.   [49].  To validate binding of fluorescein to the P-binding site, we made an NMR titration experiment with 15 N-labeled N NTD , using experimental conditions similar to those of the titration of N NTD by the pP 11 peptide. Fluorescein induced CSPs located in the same region as those induced by pP 11 (Figure 7A), indicating that fluorescein targets the P-binding site of N NTD ( Figure 7B,C). To get atomic details about the N NTD -fluorescein complex, we run docking experiments using 2 docking softwares: MOE with GBVI/WSA dG scoring function [53] and SMINA with Vinardo scoring function [54]. Two different poses of fluorescein inserted into the P-binding pocket were predicted ( Figure 7D). Fluorescein was modeled in its open, fluorescent form. In all cases, one of the xanthene rings was inserted into the hydrophobic P-binding pocket of N NTD , and the benzoic acid was positioned similarly to the pyrazole ring of M76, forming salt bridges with R150 or R132 ( Figure 7D,E).
To avoid interference of the fluorophore with FP measurements, we sought to lower the affinity of fluorescein for RSV N protein, while retaining fluorescence property, by chemical modifications. We produced methyl and ethyl esters of the benzoic acid to change the charge and the size of the molecule ( Figure 6A). A series of FP measurements was made with different N-RNA concentrations. The two fluorescein derivatives also bound to N-RNA, but displayed 2-fold lower affinity ( Figure 6B, Table 3). M76 was able to displace these two variants ( Figure 6C). K i s were of the same order of magnitude at the K i of fluorescein (Table 3). These binding and inhibition experiments suggest that the behavior of the two fluorescein derivatives is close that of the original molecule. Indeed, fluorescein methyl ester adopted docking poses similar to those of fluorescein.

An FP Assay for the RSV N-P Interaction Using Full-Length N and BODIPY FL-Labeled P Peptides
The undesirable binding of fluorescein to N-RNA led us to reconsider the BODIPY FL as label for peptides, despite the preliminary mixed results obtained with this fluorophore. To get rid of the instability problems encountered with 4,4 difluoro BODIPY FL scaffold, we considered its 4,4 dicyano analogue reported as being more stable and with better photo-physical properties. Such derivatives were recently described as tags for bioconjugation [55,56], but, to our knowledge, not for an FP assay. While difluoro-4 and dicyano BODIPY FL 5 also bound to N-RNA, FP values were markedly lower than those of fluorescein at the same N-RNA concentrations ( Figure 6B), suggesting that BODIPY FL is less prone to bind. The K d for 4, obtained by fitting the ∆FP data with FP max = 500 mFP, was also higher than that of fluorescein (Table 3). Altogether, these results show that the nature of the fluorophore must be taken into account in an FP assay with RSV N and P proteins, and that BODIPY FL is less prone than fluorescein to affect FP measurements. Moreover, as compared to difluoro BODIPY FL-labeled peptides, dicyano BODIPY FLlabeled peptides were chemically stable (either under the acidic conditions used for their purification by HPLC or under the conditions of the FP assay). To assess the properties of BODIPY FL as a fluorescent label for P peptides, we produced a new series of eight dicyano BODIPY FL-labeled (BCN n ) peptides of variable length (Table 4) and tested them for N-RNA binding. To avoid interference of the fluorophore with FP measurements, we sought to lower the affinity of fluorescein for RSV N protein, while retaining fluorescence property, by chemical modifications. We produced methyl and ethyl esters of the benzoic acid to change the charge and the size of the molecule ( Figure 6A). A series of FP measurements was made with different N−RNA concentrations. The two fluorescein derivatives also bound to N−RNA, but displayed 2-fold lower affinity ( Figure 6B, Table 3). M76 was able to displace these two variants ( Figure 6C). Kis were of the same order of magnitude at the Ki of fluorescein (Table 3). These binding and inhibition experiments suggest that the behavior of the two fluorescein derivatives is close that of the original molecule. Indeed, fluorescein methyl ester adopted docking poses similar to those of fluorescein.

An FP Assay for the RSV N−P Interaction Using Full-Length N and BODIPY FL-Labeled P Peptides
The undesirable binding of fluorescein to N−RNA led us to reconsider the BODIPY FL as label for peptides, despite the preliminary mixed results obtained with this fluorophore. To get rid of the instability problems encountered with 4,4′ difluoro BODIPY FL  We first performed FP measurements in salt-free Tris pH 8 buffer ( Figure 8A). We measured binding curves using 200 nM of fluorescent peptides. Measurements were stable within a time span of 1 h. Saturation by N-RNA was reached for all peptides, except for the shortest one BCN 5E . Overall, the dynamic range of FP was more extended for BCN n peptides than for F n peptides, with ∆FP max of 270-430 mFP, as compared to 95-315 mFP for F n peptides (Table 4). Fitted K d s were similar to those of their equivalent F n peptides, suggesting that the fluorophore does not contribute to the affinity of P peptides longer than 5-mers. Overall, affinities increased with the length of the peptides. The BCN n peptide series exhibited a similar trend to that of the F n series, highlighting the binding contribution of residues beyond Asp 240 and Phe 241 . A gain of affinity was observed for BCN 11EE vs. BCN 11SE , showing that the phosphomimetic at position 232 contributes to binding. To evaluate the quality of FP measurements, we performed a Z' assay [57,58] with the BCN10EE peptide. An N−RNA concentration of 1 µM afforded a high dynamic range (ΔFPmax = 220 ± 13 mFP, with FPmin = 9 ± 3 mFP). We determined a Z' value of 0.78 from seven data points, which validates the BCN10EE probe. We further tested robustness with respect to solvent. FP values were not significantly affected by addition of up to 5% ethanol and 10% DMSO.
As compared to salt-free buffer, the ΔFPmax range in salty buffer (100 mM NaCl) was significantly reduced for the long BCN10EE, BCN11SE and BCN11EE peptides ( Figure 8B, Table  4). This is similar to what was observed for the F11 peptide. The fluorescent label of long P peptides thus appears to be more mobile under salty conditions, likely due to screening of electrostatic interactions involving acidic amino acids of P peptides. This was not observed for the shorter B6E and B8E peptides. However, it must be noted that the binding curves of BCN6E and BCN8E did not reach a plateau within the N−RNA concentration range used in our FP experiments, which may affect the fitted parameters. The affinity of BCN6E and BCN8E peptides decreased 5-fold in salty buffer, and that of BCN10EE 8-fold, as expected by screening of electrostatic interactions in a higher ionic strength buffer. Intriguingly, similar affinities were measured for BCN11EE and BCN11SE under both salt condi- To evaluate the quality of FP measurements, we performed a Z' assay [57,58] with the BCN 10EE peptide. An N-RNA concentration of 1 µM afforded a high dynamic range (∆FP max = 220 ± 13 mFP, with FP min = 9 ± 3 mFP). We determined a Z' value of 0.78 from seven data points, which validates the BCN 10EE probe. We further tested robustness with respect to solvent. FP values were not significantly affected by addition of up to 5% ethanol and 10% DMSO.
As compared to salt-free buffer, the ∆FP max range in salty buffer (100 mM NaCl) was significantly reduced for the long BCN 10EE , BCN 11SE and BCN 11EE peptides ( Figure 8B, Table 4). This is similar to what was observed for the F 11 peptide. The fluorescent label of long P peptides thus appears to be more mobile under salty conditions, likely due to screening of electrostatic interactions involving acidic amino acids of P peptides. This was not observed for the shorter B 6E and B 8E peptides. However, it must be noted that the binding curves of BCN 6E and BCN 8E did not reach a plateau within the N-RNA concentration range used in our FP experiments, which may affect the fitted parameters. The affinity of BCN 6E and BCN 8E peptides decreased 5-fold in salty buffer, and that of BCN 10EE 8-fold, as expected by screening of electrostatic interactions in a higher ionic strength buffer. Intriguingly, similar affinities were measured for BCN 11EE and BCN 11SE under both salt conditions, despite the phosphomimetic substitution in BCN 11EE .
To further investigate the complex between BCN n peptides and N-RNA, we made competition experiments between unlabeled peptides and fluorescent peptides of same length and/or charge, using the P 10EE /BCN 10EE and pP 11 /BCN 11EE pairs. The unlabeled peptides were able to fully displace their fluorescent equivalents ( Figure 8C), and the inhibition curves could be fitted with a Hill equation ( Figure 8D). The K i s of unlabeled peptides were in the µM range (Table 5). Table 5. Inhibition parameters obtained by FP for dicyano BODIPY FL-labeled RSV P peptides in complex with RSV N-RNA. Samples were in Tris pH 8 buffer. Fluorescent probe concentration was 200 nM. N concentration was between 1 and 20 µM as indicated. IC 50 and FP max were obtained by fitting the inhibition data with a Hill equation. K i was calculated using the IC50-to-Ki server [49]. To probe the potency of M76 as an inhibitor, we carried out competition experiments with the M76 molecule and several BCN n peptides. M76 was able to displace the fluorescent probes from N-RNA, and inhibition curves were obtained ( Figure 8E). Since we used different N concentrations to have similar dynamic ranges of FP for the series of fluorescent peptides, we cannot directly compare M76 IC 50 s. However, K i s may be compared. K i values were of the same order of magnitude, 0.5-1.8 µM, for peptides longer than B 6E ( Table 5), suggesting that the inhibition mechanism by M76 is the same for these BCN n peptides. Inhibition of BCN 5E yielded slightly higher K i values (3.6-4.8 µM), but the K d determined from a non-saturating binding curve might bias the result. M76 appeared to be more potent than the P 10EE peptide to displace BCN 10EE from N-RNA with a K i of 1.1 versus 2.9 µM (Table 5).

Discussion
Recognition of the RSV nucleocapsid by the RNA polymerase proceeds via a specific interaction between the essential RSV phosphoprotein polymerase cofactor P and the RSV nucleoprotein N in complex with genomic RNA. This interaction relies on the most Cterminal residue of P, Phe 241 , which inserts into a pocket at the surface of N. This view is supported by X-ray crystal structures of phenylalanine or short C-terminal P peptides in complex with the RSV N NTD construct, revealing the Phe 241 aromatic moiety as a main structural element that drives P binding thanks to well-defined interactions with N atoms delineating a binding pocket. Recently, Phe 241 was further confirmed to be the main determinant for P anchoring to N by in silico energetic analysis and mutational analysis of Phe 241 in vitro [59]. This view is also supported by in vitro binding and polymerase activity assays. While the aromatic Phe 241 Trp substitution maintained N binding, deletion of Phe 241 and substitution of Phe 241 by smaller amino acids like alanine and aspartate impaired N binding [31]. In vitro polymerase activity, as assessed with a minigenome, was completely abrogated by the Phe 241 Ala mutation [21]. However, even though Phe 241 acts as a linchpin, adjacent residues also significantly contribute to the RSV N-P complex. The Leu 238 Ala and the double Glu 239 Ala/Asp 240 Ala mutants reduced minigenome activity 2-fold [21]. Early investigations led to the conclusion that a tract of nine C-terminal residues was necessary and sufficient for N binding [31]. Later we showed that a 12-mer peptide displayed approximately the same affinity in vitro as the full P CTD domain, with K d of 30-55 µM in salty buffer [33]. Affinity is expected to be rather weak, due to the requirement for polymerase processivity to elongate the newly synthesized RNA. Structural analysis of the RSV phosphoprotein by solution NMR indicated that P Ctail remained unstructured in solution in the context of full-length P [26]. In contrast to other linear protein-binding motifs in P, like the binding sites for N 0 , M2-1 or L proteins, that transiently fold into α-helices and are stabilized in complex, P Ctail did not display any propensity for any secondary structure, neither α-helical nor extended [29]. This is a strong indication that P Ctail does not adopt any secondary structure in complex with N protein either. This implies that the entropic cost of binding likely remains moderate and that nearly all C-terminal amino acids of P may individually contribute to strengthen the N-P complex, even if they do not adopt a defined position.
The amino acid sequence of RSV N protein is rather conserved, and thus of particular interest for drug design [60,61]. The RSV N protein had already been identified before as a target for post-entry inhibitors. Notably the benzodiazepine RSV604, which reached phase II clinical trial before being discontinued, was proposed to bind to the N protein, since it elicited escape mutants on N [62,63]. Recently, EDP-938, another benzodiazepine in phase 2a clinical trial, was reported to target RSV N, also eliciting several mutations on N close to the P-binding pocket [42] leading to resistance. Based on the inhibitory potential of the M76 molecule we have suggested that the RSV P-binding pocket on N might be druggable [33]. This hypothesis was based on the assumption that a suitable inhibitor would be a direct competitor of the Asp 240 Phe 241 dipeptide. If the interaction surface of the N-P complex extends beyond the Asp 240 Phe 241 binding site and if up to 10 residues away from Phe 241 contribute to the interaction energy, this raises the question if targeting the P-binding site with a small molecule would be sufficient to inhibit the N-P complex and displace P Ctail and full-length P. In turn, this raises the question about how to design a meaningful assay to screen for inhibitors of the RSV N-P interaction. To address this question we sought for a more comprehensive view of the fuzzy complex formed between RSV N and C-terminal peptides to propose a relevant model of this interaction.
An inhibition assay by fluorescence polarization relies on the size difference between a free fluorescent probe and its complex with a target protein. We tested two forms of the nucleoprotein, the N NTD domain and the N-RNA complex. Our results suggest that the N NTD domain, although suitable in terms of size, induces binding biases, in particular one or more secondary interaction sites and binding of fluorescent labels to the P-binding pocket. In the N-RNA complex, which is structurally close to the nucleocapsid, the N protein surface is not fully accessible, which reduces unspecific binding, and even the P-binding site appears to be rather occluded ( Figure 1B,C). N-RNA thus appears to be more relevant than N NTD and the monomeric RNA-free N 0 -like N mutant used by Shapiro et al. [45]. Interestingly, Shapiro et al. had suggested a difference in binding affinity between BODIPY FL-labeled and unlabeled P 11 by comparing FP and SPR data, which hints at binding of the fluorescent label [45]. Here, we show that binding of BODIPY FL remains moderate as compared to fluorescein, and that the peptide part of BCN n peptides longer than 5-mers provides the main driving force for N-RNA binding. Steric hindrance in the N-RNA complex context could thus act as a filter to discriminate between specific and unspecific N binding.
Since C-terminal RSV P-peptides are highly negatively charged, electrostatic interactions likely play an important role for the affinity of the RSV N-P complex. We confirmed the role of electrostatic interactions by comparing binding affinities for several fluorescent P peptides in salty versus salt-free buffers. In salt-free buffer, sub-micromolar affinity could be achieved. This is an asset for an FP assay to eliminate low binding molecules and rank high-affinity inhibitors in an HTS inhibition assay. Shapiro et al. reported a 20-fold increase in affinity by phosphorylation of Ser 232 and Ser 237 residues in the P 11 peptide [45]. We observed similar effect on affinity by using phosphomimetic peptides, which exacerbate electrostatic interactions with N.
We previously assessed that the N-P complex was fuzzy, with the C-terminal Phe 241 serving as a main anchor. To illustrate the dynamics at the P-binding site, we built a 3D structural model of the N NTD -P 11 complex by docking the P 11 peptide onto N NTD , using the Haddock webserver ( Figure 9A-C). For this purpose, we used a structure of N NTD extracted from the N NTD -P 2 complex X-ray structure (PDB 4uc9) and generated structural models of P 11 using the PEP-FOLD3 server [64]. The structural ensemble of free P 11 was highly disordered, with the exception of the C-terminus that displays an incomplete α-helical turn involving residues Glu 239 -Phe 241 ( Figure 9D). The 9 N-terminal residues of P 11 (Asp 231 -Glu 239 ) were allowed to remain flexible during docking. Docked P 11 also displayed a high degree of disorder, and the α-helical turn partly unwound ( Figure 9C). Overall, docked P 11 structural models were more extended than free P 11 models, but it cannot be excluded that this is due to the difference of algorithms implemented in Haddock and PEP-FOLD3. We observed similar effect on affinity by using phosphomimetic peptides, which exacerbate electrostatic interactions with N. We previously assessed that the N−P complex was fuzzy, with the C-terminal Phe241 serving as a main anchor. To illustrate the dynamics at the P-binding site, we built a 3D structural model of the NNTD−P11 complex by docking the P11 peptide onto NNTD, using the Haddock webserver ( Figure 9A-C). For this purpose, we used a structure of NNTD extracted from the NNTD−P2 complex X-ray structure (PDB 4uc9) and generated structural models of P11 using the PEP-FOLD3 server [64]. The structural ensemble of free P11 was highly disordered, with the exception of the C-terminus that displays an incomplete αhelical turn involving residues Glu239-Phe241 ( Figure 9D). The 9 N-terminal residues of P11 (Asp231-Glu239) were allowed to remain flexible during docking. Docked P11 also displayed a high degree of disorder, and the α-helical turn partly unwound ( Figure 9C). Overall, docked P11 structural models were more extended than free P11 models, but it cannot be excluded that this is due to the difference of algorithms implemented in Haddock and PEP-FOLD3. The electrostatic surface potential of NNTD does not reveal a unique positively charged surface area, but rather a number of anchoring points afforded by positively charged residues. P11 displays neither defined conformation nor position, but seems to adopt a preferential orientation towards a positively charged patch located near the exit the cavity occupied by Phe241 ( Figure 9A). This structural model provides a rationale for the overall increase of binding affinity observed with increasing peptide length up to 11 amino acids. Each amino acid contributes to binding affinity, either by electrostatic or van der Waals interactions. It also explains the concomitant decrease of FP in fluorescently labeled P pep-  [65], based on the crystal structure of the N NTD -P 2 complex (PDB 4uc9) for N NTD and on a P 11 model generated with PEP-FOLD3 [64]. 10 models were superimposed. The docked P 11 peptides are in yellow cartoon. (A) Two views of N NTD rotated by 90 • show the electrostatic surface potential calculated with Delphi software [66], with a scale ranging from −10 (red) to +10 kcal.mol −1 (blue). (B) The backbone of N NTD is shown in cartoon representation. (C) The P 11 starting structure used for Haddock docking, in magenta, was aligned with the docked structures represented in other colors. The view is rotated, as compared to (A,B). (D) 20 P 11 models generated with PEP-FOLD3 were structurally aligned along the three C-terminal residues Glu 239 -Phe 241 . The model used for Haddock docking is in magenta, and the other models in colors unrelated to those used in (C).
The electrostatic surface potential of N NTD does not reveal a unique positively charged surface area, but rather a number of anchoring points afforded by positively charged residues. P 11 displays neither defined conformation nor position, but seems to adopt a preferential orientation towards a positively charged patch located near the exit the cavity occupied by Phe 241 (Figure 9A). This structural model provides a rationale for the overall increase of binding affinity observed with increasing peptide length up to 11 amino acids. Each amino acid contributes to binding affinity, either by electrostatic or van der Waals interactions. It also explains the concomitant decrease of FP in fluorescently labeled P peptides. As the label is attached to the opposite N-terminal side of the peptide, the fluorescent label experiences increased motional freedom, when the length of the tether increases. As the majority of C-terminal P peptide residues is charged, it is not surprising that a decrease in FP is observed in buffers with higher ionic strength.
As compared to difluoro BODIPY FL-labeled peptides, dicyano BODIPY FL-labeled peptides were more stable. The FP dynamic range was more extended for BCN n peptides than for F n peptides: This is an advantage for an FP assay. An FP assay relies on a compromise between peptide length and mobility of the fluorescent label. In this respect, 10-mer or 11-mer BCN n peptides, which bind with sub-micromolar affinity in salt-free buffer and offer an extended dynamic FP range, appear to be suitable. Finally, M76 proved to be a suitable benchmark to screen for inhibitors, as it is able to fully displace even longer fluorescent P peptides.
HPLC analyses were performed on Hypersil C18 column (120 Å, 5 µm, 150 × 4.6 mm) using Waters Alliance 2690 separation module equipped with a Single Quadrupole Detector 2 (ESI quadrupole mass spectrometer), an ELS detector (Waters 2420) and a photodiode array detector (Waters 996). Reversed-phase ultra performance liquid chromatography-mass spectrometry (RP-UPLC-MS) analyses were performed on Waters equipment consisting of an ACQUITY UPLC H-Class separation module, photodiode array detector (eLambda detector), and a ESI triple quadrupole mass spectrometer (TQ detector). The analytical column used was the ACQUITY UPLC BEH C18 column (130 Å, 1.7 µm, 2.1 mm × 50 mm) operating at 0.6 mL·min −1 with linear gradient programs in 2.5 min run time (classical program: 5 to 100% of B in 2.5 min). UV monitoring was performed most of the time between 200 and 500 nm and was extracted at 214 nm. Solvent A consisted of H 2 O containing 0.1% (v/v) formic acid and solvent B was CH 3 CN containing 0.1% (v/v) FA. RP-HPLC purifications were performed on Waters equipment consisting of a 2545 quaternary pump, a photodiode array detector (Waters 2998), and an injector collector (Waters 2767). The preparative column, XBridge BEH C18 column (130 Å, 5 µm, 30 mm × 150 mm) was operated at 30 mL·min −1 with linear gradient programs in 15 min run time. The semi-preparative column, XBridge BEH C18 column (130 Å, 5 µm, 10 mm × 150 mm) was operated at 5 mL·min −1 with linear gradient programs in 15 min run time. Solvent C consisted of H 2 O containing 0.1% (v/v) TFA and solvent D consisted of CH 3 CN containing 9.9% (v/v) H 2 O and 0.1% TFA. Classical focused programs on preparative or semi-preparative column: 10 or 20% (v/v) slope of D in 15 min. Water was of Milli-Q quality and was obtained after filtration of distilled water through a Milli-Q cartridge system. CH 3 CN, FA, and TFA were of spectroscopic grade. High resolution mass spectra (HRMS-ESI) were obtained on a LCT Waters XE mass spectrometer equipped with an electrospray ionisation source. NMR spectra were performed on Bruker Avance spectrometers operating at 699 MHz for 1 H NMR, 176 MHz for 13 C NMR, 282 MHz for 19 F and 160 MHz for 11 B NMR experiments. The chemical shifts are reported in ppm relative to tetramethylsilane with the solvent resonance as the internal standard. Multiplicities were given as: s (singlet); d (doublets); t (triplets); q (quadruplets) m (multiplets). Coupling constants are reported as a J value in Hz.

Synthesis of Fluorescent Molecules
The fluorescein methyl ester 2 and ethyl ester 3 were synthetized in 73% and 40% yield from fluorescein according to Lu et al. [67] and C. Y. Ng et al. [68], respectively. Difluoro BODIPY FL 4 was synthesized according to Gießler et al. [69]. The dicyano BODIPY FL 5 was synthesized in two steps from the difluoro BODIPY FL methyl ester (Scheme 1). The latter was obtained according to K. Gießler et al. [69].

Peptide Synthesis
The BODIPY FL-labeled P peptide BF2pP11 was synthesized from purchased pP11 peptide and BODIPY FL as described in [45].  (7): BF 3 ·OEt 2 (34.0 µL, 0.27 mmol) was added to a cooled solution of difluoro BODIPY FL derivative 6 (415 mg, 1.36 mmol) in anhydrous CH 2 Cl 2 (70 mL) at 0 • C. The mixture was then stirred at 25 • C for 10 min and TMSCN (851 µL, 6.80 mmol) was added. The reaction mixture was stirred at room temperature for 2 h. A saturated aqueous NaHCO 3 solution was added. The organic phase was washed with water and was dried over Na 2 SO 4 . The solvent was evaporated under reduced pressure to give the titled product as a red solid (0.43 g, 98%). 1 (5): Concentrated HCl (16 mL) was added to a solution of methyl ester 7 (0.20 g, 0.62 mmol) in a mixture of THF (40 mL) and water (28 mL). The solution was stirred for 12 h at room temperature. Water (50 mL) and CH 2 CL 2 (50 mL) were then added. The aqueous phase was extracted with CH 2 Cl 2 (3 × 50 mL) and the combined organic phases were washed with brine and were dried over Na 2 SO 4 . The solvent was evaporated under reduced pressure to give the titled product as a red solid (0.17 g, 90%). M.p.: 211-232 • C. 1
Assemblies of protected peptides were carried out manually on 2-Chlorotritylchloride resin (theoretical loading of 1.6 mmol/g-0.2 g) using the Fmoc/tBu strategy in a polypropylene reaction vessel fitted with polyethylene frits. After swelling of the resin for 15 min in CH 2 Cl 2 , the first amino acid was loaded through nucleophilic substitution by shaking the resin in a solution of Fmoc-Phe-OH (1 mmol/g) and DIPEA (1.5 equiv) in anhydrous CH 2 Cl 2 (10 mL/g resin) for 30 min at room temperature. The un-reacted sites were then capped by shaking the resin in a mixture of MeOH/DIPEA/CH 2 Cl 2 (2/1/17 v/v/v; 10.0 mL/g resin) for 10 min (repeated two times). The resin was then washed two times with DMF and three times with CH 2 Cl 2 (10.0 mL/g resin) for 30 s. The overall process yielded a resin loading of around 0.65 mmol/g [determined from the dosage of the Fmoc released under the conditions of procedure A]. The number of reagent equivalents used in the following procedures is calculated from this resin loading. This step is followed by Fmoc removal according to procedure A. The elongation of the peptide was performed by cycle repetition of peptide coupling and Fmoc removal according to procedures B and A, respectively. Coupling of dicyano BODIPY FL, if required, was performed on the resin at the end of the peptide elongation on the amino terminus of the peptide according to procedure C. Cleavage of the peptide from the resin was carried out according to procedure D.

Procedure A: Fmoc Removal
N-α-Fmoc protecting groups were removed by shaking the resin in piperidine/DMF solution (1/4, v/v; 10.0 mL/g resin) for 5 min at room temperature. The process was repeated two times for 10 min, and the completeness of deprotection was monitored by ultraviolet absorption measurement of the filtrate at 299 nm. The resin was then washed five times with DMF and once with CH 2 Cl 2 (10.0 mL/g resin) for 30 s.

Procedure B: Coupling Steps
Coupling reactions were performed by shaking with a solution of N-α-Fmoc-protected amino acid (2 equiv), PyBOP (2 equiv) and DIPEA (4 equiv) in DMF (10.0 mL/g resin) for 30 min at room temperature. The resin was then washed three times with DMF and once with CH 2 Cl 2 (10.0 mL/g resin) for 30 s. Completeness of the coupling was controlled by TNBS and KAISER tests.

Procedure C: Coupling of Dicyano BODIPY FL
The resin was shaken with a solution of dicyano BODIPY FL 5 (1 equiv), DIPEA (4-8 equiv depending of the number of peptide acidic functions) and PyBOP (2 equiv) at room temperature in DMF (10.0 mL/g resin). The resin was then washed 14 times with DMF and 8 times with CH 2 Cl 2 (10.0 mL/g resin) for 30 s.

Procedure D: Resin Cleavage and Protecting Groups Removal
The resin was first washed five times with CH 2 Cl 2 (10.0 mL/g resin) before shaking in a mixture of TFA/TIS/H 2 O MQ (95/2.5/2.5 v/v/v; 10.0 mL/g resin) for 2 h at room temperature and then filtrated. The resin was rinsed once with the same acidic mixture of TFA/TIS/H 2 O MQ (95/2.5/2.5 v/v/v) for 1 min. The recovered solutions were concentrated under reduced pressure, and white solid was obtained by precipitation, triturating, and washing three times with diethyl ether. These crude peptides and crude dicyano BODIPY FL peptides, obtained as TFA salts, were purified by RP-HPLC on preparative or on semi-preparative column, respectively, using focused gradients, and freeze dried before their analysis by RP-UPLC-MS. Due to reported instability of FTU-labelled peptides under acidic conditions [70], the coupling of FITC with synthesized P 3 , P 4 , P 5E , P 7E , P 9E , P 11SE , P 11EE peptides and commercial pP 11 and P 11SS peptides was performed in solution. FITC (2 equiv) was added to a solution of the peptides (1 equiv) and DIPEA (9 equiv) in anhydrous DMF (0.06 mL/mg peptide). Solutions were stirred for 30 min to 1h depending on the peptide. Solutions were then purified by HPLC on semi-preparative column using focused gradients and freeze dried. UPLC traces and analysis for F n peptides are given in Figures S12-S20.

Bacterial Expression and Purification of RSV N Protein
Recombinant RSV N protein was produced in E. coli BL21(DE3) bacteria (Novagen, Madison, WI). The N-terminal domain N NTD (residues 31-252), containing a C-terminal 6x histidine tag, was overexpressed using the pET-N(31-252) plasmid and purified as described previously [21]. In a final step N NTD was dialyzed into MES 20 mM pH 6.5, NaCl 250 mM buffer [33]. Full-length N was produced by co-expression with the Cterminal domain of RSV P protein in E. coli BL21(DE3) co-transformed with pET-N and pGEX-P(161-241) plasmids [21]. Cultures were grown at 37 • C for 8 h in Luria-Bertani (LB) medium supplemented with 50 µg/mL kanamycin and 100 µg/mL ampicillin. An equivalent volume of LB was then added, and protein expression was induced with 80 µg/mL isopropyl-β-D-thio-galactoside (IPTG) for 15 h at 28 • C. Bacteria were harvested by centrifugation. The bacterial pellet was resuspended in lysis buffer (20 mM Tris-HCl pH 8.5, 150 mM NaCl, 1 mM EDTA, 2 mM dithiothreitol, 0.2% Triton X-100, 1 mg/mL lysozyme). Complete protease inhibitor cocktail (Roche, Mannheim, Germany) was added, and the suspension was incubated for 1 h on ice, sonicated and centrifuged at 4 • C for 30 min at 10,000× g. Glutathione-Sepharose 4B beads (GE Healthcare, Vélizy-Villacoublay, France) were added to the clarified supernatant and incubated at 4 • C for 3 h. The beads were washed twice in lysis buffer and three times in 20 mM Tris-HCl pH 8.5, 150 mM NaCl buffer. GST was cleaved from the N-P(161-241) complex by treating the beads with thrombin (Novagen) for 16 h at 20 • C. The supernatant was then loaded onto a Superdex 200 16/30 column (GE Healthcare) and eluted in 20 mM Tris-HCl pH 8.5, 150 mM NaCl buffer. Finally, the fractions containing the N protein in the form of an N-RNA complex were pooled and concentrated up to 2 mg/mL (45 µM of N protein). The protein was subsequently dialyzed into 20 mM Tris pH 8.0 buffer containing 0.01% Brij-35. The concentration of N protein was determined with a BCA assay (Thermo Fisher, Illkirch-Graffenstaden, France), calibrated with BSA.

NMR Measurements
NMR titration experiments with 15 N-N NTD were performed in 20 mM MES pH 6.5, 250 mM NaCl buffer. pP 11 peptide was solubilized at 1 mg/mL in MQ water by addition of 1 M NaOH until the pH became neutral. pP 11 aliquots were lyophilized. For the titration experiment, pP 11 was added stepwise to a 50 µM 15 N-N NTD solution, using 0.2-6.2 molar equivalents. NMR data were acquired on a 14.1 T (600 MHz 1 H frequency) Bruker Avance III NMR spectrometer equipped with a cryogenic TCI probe. A standard 2D 1 H-15 N HSQC spectrum was recorded for each titration point.
For titration with fluorescein, 0.5-8 molar equivalents of fluorescein were added from a 10 mM stock solution in water at pH 6.5 to 55 µM 15 N-N NTD . NMR measurements were performed on a 16.4 T (700 MHz 1 H frequency) Bruker NMR spectrometer equipped with a NEO console and a cryoTXO probe. All samples contained 7.5% 2 H 2 O to lock the spectrometer frequency and 100 µM DSS to reference 1 H chemical shifts. The temperature was set to 20 • C. NMR data were processed with TopSpin 4.0 (Bruker Biospin, Wissembourg, France) and analyzed with CcpNmr Analysis 2.4 software [71]. Backbone chemical shift assignment of N NTD was done previously [33].
Combined amide 1 H and 15 N chemical shift perturbations ∆δ HN were calculated with a scaling factor of 1/10 for 15 N, which corresponds to the ratio of 15 N and 1 H gyromagnetic ratios (Equation (1)): dissociation constant of the complex with N NTD . δ free and δ bound (in ppm) are the chemical shifts of free and bound N NTD .
For residues in intermediate exchange regime, the exchange rate between free and bound protein states k ex (s −1 ) was estimated according to Equation (3), where BF is the Larmor frequency (MHz) of the observed nucleus.

Fluorescence Polarization Measurements
Fluorescence polarization measurements were carried out on a Paradigm Detection Platform (Beckman Coulter, Brea, CA, USA) operating with a microplate reader, a fluorescein detection cartridge (excitation range 485/20 nm, emission range 535/25 nm for both parallel and perpendicular components) and the SpectraMax Pro 6.1 software (Molecular Devices, San Jose, CA, USA). Samples were placed in polystyrene black flat bottom 96-well microplates (Cellstar, Greiner bio-one, Frickenhausen, Germany) with 20 µL final volume in each well. If not stated otherwise, the final concentration of fluorescent ligands was 200 nM. The buffer was either 20 mM MES pH 6.5, salt-free or with 250 mM NaCl, or 20 mM Tris pH 8.0 salt-free or with 100 mM NaCl, supplemented with 0.01% Brij-35. Mixing was done using the corresponding plate reader option. The temperature was set to 20 • C. Data were recorded in top reading mode. Measurements were done in triplicate and after an incubation time of 30 min, if not indicated otherwise. Fluorescence polarization (FP) in mFP units was determined from the parallel (I ) and perpendicular (I ⊥ ) components of the emitted light according to Equation (4) [43]. A grating factor G of 1.2 was determined using 1 nM fluorescein in 1 mM NaOH and assuming a theoretical FP of 27 mFP [48]. Background signal was measured from wells containing only buffer, and the mean value was subtracted from each I and I ⊥ component. FP (mP) = 1000 × I − G * I ⊥ / I + G × I ⊥ Theoretical fluorescence polarization values of a fluorophore bound to N protein were calculated from the fluorescence lifetime τ and the molecular weight MW using Perrin's equation (Equation (5)). A is the fluorescence anisotropy related to FP by Equation (6) [43], A 0 the intrinsic anisotropy (0.4), T the temperature in Kelvin, k the Planck constant, η the dynamic viscosity, and V h the hydrated volume. V h was calculated according to Equation (7), where N is the Avogadro number, v 2 the partial specific volume of the protein (7.34 × 10 −7 m 3 /g), δ the degree of hydration (0.75), and v 1 0 the specific volume of water (10 −6 m 3 /g). We used A 0 , v 2 , δ and v 1 0 values reported for bovine serum albumine [72].
We assumed that the measured fluorescence polarization FP is a linear combination of the FP of the fluorescent probe in the absence of N protein (FP min ) and the FP of Nbound fluorescent probe (FP max ), weighted by the molar fraction. The molar fraction x = [P] bound /[P] 0 can then be expressed as a ratio of FP differences (Equation (8)).
x = (FP − FP min )/(FP max − FP min ) = ∆FP/∆FP max (8) To determine the K d value of a complex between a fluorescent probe and N protein, a series of samples with constant probe concentration and varying N concentration was measured. FP binding curves were fitted with a single binding site model and a 1:1 stoichiometry. The molar fraction of bound fluorescent probe was expressed as a function of the ratio of protein versus probe concentrations, r = [N] 0 /[P] 0 , according to Equation (9). Binding curves were fitted in Origin 7 software according to Equation (10), obtained by combining Equations (8) and (9).
To assess inhibition of a complex between a fluorescent probe and N protein by peptides or the M76 molecule, the concentration of N was set to work with an FP range comprised between FP min and~80% FP max . Peptides were added from 1 or 2 mM stock solution in water at neutral pH. M76 was added from 10 or 20 mM stock solutions in ethanol or in DMSO. The percentage of inhibition (% i ) was calculated as the relative change in FP (Equation (11)). % i was fitted as a function of inhibitor concentration with a Hill equation, where IC 50 is the inhibitor concentration resulting in 50% inhibition (Equation (12)). Inhibition data were fitted with Origin 7, and in parallel with the Excel solver module. All data were compatible with an absence of cooperativity. FP max and FP min were fitted, and the maximal inhibition % i,max was set to 100%.
A Z'-factor for FP measurements without test compound [58] was calculated from 7 data points for the condition 200 nM B 10EE and 1 µM N-RNA, using Equation (14), where σ + and σare the standard deviations of positive (fluorescent peptide bound to N protein) and negative (free fluorescent peptide) controls, respectively, and µ + and µtheir mean values.

Complex Modelling with Haddock
Docking of the P 11 peptide onto N NTD was performed with the Haddock 2.4 software [65] on the WeNMR server [73]. A structure of P 11 was built with the PEP-FOLD3 server [64] and allowed to be flexible (residues 1-9) during docking. The C-terminal Asp 10 and Phe 11 residues (equivalent to Asp 240 and Phe 241 ) were declared as active residues. The structure of N NTD was extracted from the X-ray structure of the N NTD -P 2 complex (PDB 4uc9). N residues 46, 50, 53, 128, 131, 132, 135, 145, and 151, lining the P-binding pocket, were declared as active residues. Passive residues were automatically defined around active residues. Default scoring parameters for protein-protein complexes were used, except for Evdw 1, which was increased from 0.01 to 1. 1000 initial structures were generated. 200 final structures were refined in water and clustered according to the RMSD criterion. Haddock clustered 156 structures in 12 clusters. Cluster 1 with the best haddock score (−96.4 ± 4.9) contained 63 structures, i.e., 40% of clustered structures.

Docking of Small Compounds
Fluorescein and fluorescein methyl ester were docked on the P-binding pocket of N NTD (PDB 4ucc) using Smina software and MOE [53]. For Smina, Vinardo [54] scoring function was selected with an exhaustiveness of 8. The pose with the best scoring value was retained. For MOE, the placement method used was Triangle Matcher with London dG score and the number of poses was set to 30. These poses were then refined with GBVI/WSA dG score with rigid receptor, and the number of output poses was set to 5. The pose with the best GBVI/WSA dG score was retained.