New Insights into Structural Disorder in Human Respiratory Syncytial Virus Phosphoprotein and Implications for Binding of Protein Partners*

Phosphoprotein is the main cofactor of the viral RNA polymerase of Mononegavirales. It is involved in multiple interactions that are essential for the polymerase function. Most prominently it positions the polymerase complex onto the nucleocapsid, but also acts as a chaperone for the nucleoprotein. Mononegavirales phosphoproteins lack sequence conservation, but contain all large disordered regions. We show here that N- and C-terminal intrinsically disordered regions account for 80% of the phosphoprotein of the respiratory syncytial virus. But these regions display marked dynamic heterogeneity. Whereas almost stable helices are formed C terminally to the oligomerization domain, extremely transient helices are present in the N-terminal region. They all mediate internal long-range contacts in this non-globular protein. Transient secondary elements together with fully disordered regions also provide protein binding sites recognized by the respiratory syncytial virus nucleoprotein and compatible with weak interactions required for the processivity of the polymerase.

infected by the age of two, requiring hospitalization in ϳ5% cases (3). Elderly and immunocompromised adults are also at increased risk. No efficient treatment is presently available for hRSV (4), and vaccination is challenging due to complex immunogenicity (5). The search for hRSV antiviral drugs directed toward specific viral functions is therefore still ongoing (6).
The hRSV RNA-dependent RNA complex (RdRp) constitutes a virus-specific target with specific protein-protein interactions that have not all been investigated in detail (7). It uses the nonsegmented single-stranded negative sense RNA genome of hRSV as a template. In infected cells, the viral RdRp is found in specific inclusion bodies (8), which have been shown to be transcription and replication centers for other Mononegavirales, e.g. rabies (9) and vesicular stomatitis viruses (10). The apo RdRp complex is composed a minima of the large catalytic subunit (L) and its essential cofactor, the phosphoprotein (P) (11,12). The P protein plays a central role in the RdRp by interacting with all main RdRp components. During transcription and replication it tethers the L protein to the nucleocapsid (NC), consisting of the genomic RNA packaged by the nucleoprotein (N), by direct interaction with N (13)(14)(15)(16). hRSV P also binds to the transcription antitermination factor M2-1 (17)(18)(19). Phosphorylation of P has been proposed to regulate these interactions, although it is not essential for replication (20 -22). P also acts as a chaperone for neo-synthesized N by forming an N 0 ⅐P complex that preserves N in a monomeric and RNA-free state (23). We have shown previously that formation of hRSV NC⅐P and N 0 ⅐P complexes proceeds via two distinct binding sites on P (14,24).
Bioinformatic and biochemical investigations have established that hRSV P is tetrameric and contains large disordered N-and C-terminal regions (25)(26)(27). Fragment Y* (Table 1) was described as a minimal oligomerization domain (OD) with predicted helical coiled-coil structure (28). However, a clear picture of the overall structure of P is still lacking, mainly because of its structural disorder. Our aim was to get a deeper insight into the structural plasticity of P and to explore the role of transiently ordered regions for interactions with hRSV RdRp proteins, here with N, by using NMR spectroscopy.

Extent of Intrinsically Disordered Regions in hRSV
Phosphoprotein-To probe the structural organization of hRSV phosphoprotein at the single residue level by NMR, we first used full-length P protein (P FL ). The two-dimensional 1 H- 15 N HSQC spectrum of P FL exhibits sharp amide signals with narrow 1 H chemical shift dispersion in the 7.5-8.5 ppm range (Fig.  1). This is the signature of intrinsically disordered proteins (IDPs) and regions (IDRs) (29). Experimental conditions were adjusted for the detection of IDP amide protons, low temperature (288 K) and acidic pH (6.5), to reduce the contribution of water exchange to line widths. Only 60% (140 of 229) of expected P FL amide signals were observed. 40% of amide signals were too broad to be detected and correspond to protein regions undergoing dynamic processes and conformational exchange at the s-ms time scale. They comprise the OD (fragment Y*, ϳ40 residues), but also extensive additional regions (ϳ50 residues).
Due to the structural heterogeneity of P, we resorted to protein fragments to delineate domains. Several fragments had been produced before (13,24,25,30). Constructs are detailed in Table 1. In particular, P NDϩOD and P ODϩCD correspond to the N-and C-terminal domains with OD. P ND and P CD are their counterparts without OD. The 1 H-15 N HSQC spectra of P fragments exhibit sharp lines and narrow 1 H chemical shift dispersion, similarly to P FL (Fig. 1). All signals superimpose well, showing that the fragments are representative of the corresponding domains in P FL . For instance, overlay of P NDϩOD and P ODϩCD spectra reproduces the spectrum of P FL (Fig. 1). Comparison of P ND and P NDϩOD indicates that the OD signals are missing for P NDϩOD . Remarkably, fragment P CD displays more signals than P ODϩCD , revealing that the C-terminal domain of P contains residues that are not completely disordered when attached to the OD (Fig. 1).
Sequential assignment of backbone chemical shifts was carried out separately for all fragments. The signals of the 120 N-terminal residues and 40 C-terminal residues could be observed for all constructs, including P FL , indicating that they form two independent N-and C-terminal IDRs in P. The signals of the Asp 125 -Thr 160 region, equivalent to fragment Y* (Table 1), were missing for all constructs containing the OD. Residues Ser 161 -Glu 204 were also missing in the spectra of P ODϩCD and P FL , but were present in the spectrum of P CD .
Determination of Transient Secondary Structure Elements in hRSV Phosphoprotein-On closer inspection, many NMR signals display local heterogeneity in intensity and line width. Taking advantage of the sensitivity of 13 C backbone and 1 H␣ chemical shifts to protein dihedral angles, we determined residue-specific secondary structure propensities (SSPs) in P ND , P NDϩOD , P CD , P ⌬OD , and P FL , using Talosϩ (31) (Fig. 2A). Outside the Ser 161 -Glu 204 region no significant secondary structure was detected. This confirms that the N and C termini of P are fully disordered. Still, weak ␣-helical propensity is observed in the N-terminal IDR for residues Asp 12 -Ile 24 and Phe 98 -Lys 103 . The latter define two transient helices ␣ N1 and ␣ N2 (Fig. 2B).
In contrast, residues Leu 173 -Lys 205 , which can only be observed in P fragments devoid of the OD, display high ␣-hel-ical propensity and define two helices, ␣ C1 (Leu 173 -Met 187 ) and ␣ C2 (Asn 189 -Lys 205 ), with up to 70 and 95% propensity, respectively (Fig. 2, A and B). Because SSPs depend on the model used to extract them, we also extracted SSPs with ␦2D (32). We obtained lower ␣-helical propensities than with Talosϩ (up to 20% for ␣ C1 and 70% for ␣ C2 ), indicating that these helices are also transient, with ␣ C2 being almost stable. The ␣ C1/2 region does not induce oligomerization on its own, because P ⌬OD and P CD display the same line widths as P ND (Fig. 1). ␣ C1 and ␣ C2 thus form a second C-terminal IDR with high ␣-helical propensity, in addition to the fully disordered C terminus. This domain likely accounts for the thermal transition observed in P at physiological temperature (27), indicating that these helices do not tightly associate in the P tetramer. Signal broadening of ␣ C1/2 in tetrameric P fragments may be explained by the increased molecular size, which affects overall dynamics in solution, or by interactions between these transient helices and possibly with the OD.
Investigation of the Dynamics of hRSV Phosphoprotein by 15 N Nuclear Relaxation-We measured 15 N relaxation parameters for a tetrameric (P NDϩOD ) and three monomeric (P ND , P CD , and P ⌬OD ) fragments of P (Fig. 2C) to analyze the dynamic behavior of the different regions of P. P IDRs display overall homogeneous, negative or near zero heteronuclear 1 H-15 N nuclear Overhauser effects (NOEs), indicative of large amplitude backbone fluctuations on a ps-ns time scale. The ␣ C1/2 region displays higher NOE values (0 -0.5), suggesting that it is more ordered. This is also the case for the transient ␣ N1 helix. 15 N transverse relaxation rates (R 2 ) are more heterogeneous along the sequence of P, but consistent among P fragments. R 2 values are much higher for regions with ␣-helical SSP as compared with completely random regions, denoting differential tumbling in solution and conformational exchange between disordered and ordered conformational states, on a s-ms time scale. A fifth region (Asn 78 -Ser 86 ) without clear SSP displays similar behavior (Fig. 2C), suggesting that exchange broadening also arises from internal interactions. 15 N longitudinal relaxation rates (R 1 ), which are not sensitive to exchange, are nearly uniform along the sequence of P, but underline the structural singularity of ␣ C1 . Relaxation parameters of P IDRs are globally independent of the length of the fragments, suggesting that their motions are not significantly restricted and that they are not stably associated with any part of the protein.
Detection of Long-range Contacts in hRSV P by Paramagnetic Relaxation Enhancement-Next, we used paramagnetic relaxation enhancement (PRE) to investigate the spatial organization of P. Line broadening due to PRE in a ϳ15 Å radius around a paramagnetic spin label can be used to measure long-range distances by NMR in globular proteins, but also to probe longrange contacts in IDPs (33,34). The sequence of P FL does not contain any cysteine. We therefore introduced cysteines by mutating individual residues distributed along the sequence of P FL and labeled them with IAP free radical (35). The ␣ C1/2 region remained undetectable in the 1 H-15 N HSQC spectra of all Cys mutants in their diamagnetic state, similarly to wild type P FL . We therefore concluded that the mutations did not impact P oligomerization. Hence PREs could only be measured outside the OD and ␣ C1/2 region.
The PRE profiles (Fig. 3) are all rather broad and consistent with diffuse contacts mediated by large highly flexible regions. Gradually decreasing PREs are observed for up to 40 residues on each side of a spin label in fully disordered regions of P (positions 23, 99, and 237 mutants, but also between this region and the OD, using the S143C and S156C mutants, and even the C-terminal ␣-helical IDR using E179C and E193C. Only the C terminus does not appear to be involved in any specific contact, because strong PREs are only observed for proximal spin labels. The absence of PREs between the C-and N-terminal IDRs moreover provides evidence for a parallel organization of the hRSV P tetramer. Interactions of hRSV P with Nucleocapsid Analogs-Because transiently structured regions of IDPs are potential molecular recognition elements (36,37), we carried out NMR interaction experiments to investigate the impact of transient structures within P on hRSV nucleoprotein binding. We first tested N in the form of N-RNA rings, which mimic the hRSV nucleocapsid (38). As the nucleocapsid binding domain of P (P NCBD ) had been assigned to its 9 C-terminal residues (13), we worked with 15 N-labeled P CD instead of P FL . In the presence of N-RNA rings, the signals of the eight last residues are completely broadened out in the 1 H-15 N HSQC spectrum of P CD (Fig. 4, A and B), confirming that they are involved in P binding to the NC. An overall 15% intensity loss seems to arise from increased viscosity, as assessed by a control experiment with excess BSA (Fig. 4B).
Direct observation of N-bound P residues by 1 H-15 N HSQC could not be achieved due to the large size of N-RNA rings. We therefore proceeded with the monomeric N-terminal domain of N (N NTD ), which was shown to be relevant for NC⅐P binding (14,39). By titrating N NTD into 15 N-P CD , linear chemical shift perturbations (CSPs) were observed for P NCBD (e.g. Asn 234 and Ser 237 in Fig. 4C). A titration end point was reached with 5 eq of N NTD (Fig. 4C). By fitting the CSP data with a two-site fast exchange model, a K d of 25-50 M was determined, in agreement with a K d of 30 M previously determined for the N NTD ⅐P CD complex by isothermal titration calorimetry (39).
Notably, CSPs and line broadening occurred concomitantly, pointing to moderately fast exchange between free and N NTD -bound P. However, at the titration end point lines are broader than expected for a 35-kDa complex (Fig. 4C). Surprisingly, no signal was recovered for the very C-terminal Phe 241 residue, which in X-ray structures appears to be the main structural determinant for NC⅐P complex formation, by tightly inserting into a hydrophobic pocket at the surface of N NTD (39). Our results thus suggest the possibility of additional binding modes corresponding to weaker interactions. Indeed, the large line broadening observed for ␣ C1 (Fig. 4D) would be in favor of a secondary binding site on ␣ C1 . This effect is also observed with the P 161-229 fragment deleted of P NCBD , although attenuated ( Fig. 4D), suggesting that binding to ␣ C1 could be promoted by binding to P NCBD .
Investigation of the hRSV N 0 ⅐P Binding Mode-In a last part we investigated the N 0 ⅐P binding mode by using N mono , a monomeric N mutant impaired for RNA binding (24). N mono leads to line broadening at the ␣ N1 site in P NDϩOD (Fig. 4, E and F), in agreement with previous results that showed that the N-terminal P 40 peptide was able to pull down N mono (24). NMR interaction experiments with N mono and P FL show that N mono is competent for both NC⅐P and N 0 ⅐P binding modes, via P NCBD and ␣ N1 , respectively (Fig. 4F). Because both sites in P FL can be occupied with only 1 eq of N mono , either none of the complexes is very tight or the two sites are not mutually exclusive.
Unexpectedly, a third region, ␣ N2 , was perturbed (Fig. 4, E and F). As previous experiments with the P 60 -126 fragment showed that this region does not bind to N mono (24), line broadening of ␣ N2 signals would not be explained by direct binding to N. However, we showed by PRE that ␣ N2 transiently associates with ␣ N1 . Formation of an N 0 ⅐␣ N1 complex could displace the equilibrium between free and ␣ N1 -bound ␣ N2 .
As a control, we performed additional interaction experiments with 15 N-labeled N-terminal P fragments and N NTD . Surprisingly, we observed perturbations (Fig. 5, A and B). However, the intensity ratio patterns are different from N mono . N mono perturbed a large region (Met 1 -Ser 30 ), whereas N NTD affects only a few residues around Lys 25 (Figs. 4F and 5B), suggesting a difference in binding. We carried out complementary experiments by measuring spectra of 15 N-N NTD in the presence of P 40 . Line broadening was induced in different N NTD regions (Fig. 5, C and D) delineating a contiguous surface on N (Fig. 5E). This surface is on the inside of the nucleocapsid as opposed to the binding site of the C terminus of P in the NC⅐P complex (Fig.  5E). It partly overlaps with the interaction surface of the N-terminal arm of the adjacent protomer in N-RNA rings, but is shifted with respect to the interaction surface recently published for the human Metapneumovirus (hMPV) N 0 ⅐P complex, where the N-terminal P 1-28 peptide obstructs the binding sites of the N-and most prominently C-terminal arms of adjacent N protomers (40).  15 N HSQC spectra of P FL , P NDϩOD , P ODϩCD , P ND , P CD , and P ⌬OD were acquired under identical experimental conditions (50 -100 M concentration, 288 K temperature, 14.1 T magnetic field). A different color was used for each construct to plot contours and show assignments, with full-length P (P FL ) in black. For all deletion mutants, the spectrum is superimposed onto that of P FL for comparison. Amide resonance assignments are indicated for P NDϩOD , P ODϩCD , P ND , and P CD (residue number and amino acid type in single-letter code). Asn and Gln NH 2 side chain signals are not individually assigned. For P ND the inset shows assignments of the crowded central region of the spectrum.

Discussion
Domain Organization in hRSV P-hRSV phosphoprotein is characterized by extensive structural disorder that hampers high resolution structural characterization. The only structured part appears to be the tetramerization domain, which has been investigated by bioinformatic tools (26), resistance to protease digestion (28,41), and deletion series (13,25,42). Although fragment Y*, which exhibits high stability and homogeneity (25,27,28), has been acknowledged as the core of the tetramerization domain, the OD of hRSV P is often represented by fragment X (Table 1), longer then Y* by 15 residues at its N terminus. Predicted structural models of fragment X present this stretch as a second short coiled-coil domain (28,42). We show here that this stretch has no significant SSP and is highly dynamic in solution, at least in the absence of protein partners. This is also in agreement with the observation that P ND was not able to form oligomers (25).
In contrast to Mononegavirales N proteins, which share a conserved fold (43), P proteins have largely diverged, hRSV P having the shortest sequence (44). However, several regions of P are highly conserved among Pneumoviridae, more particularly the OD (Fig. 6A). Its structure was solved by X-ray diffraction for hMPV P (45). It displays a coiled-coil helical conformation for a region equivalent to the hRSV fragment Y*, indicating that a short OD is specific of this family. The protomers are arranged in a parallel orientation, consistently with the PRE  results for hRSV P, and not as a dimer of anti-parallel dimers like in Mumps virus P (46,47).
Finally, we were able to identify a C-terminal domain with high ␣-helical propensity, whose dynamics are distinct from those of the OD and the fully disordered C terminus. The presence of transient C-terminal helices was also proposed for hMPV, based on small angle X-ray scattering data (45), showing that they constitute another hallmark of Pneumoviridae. But whereas ␣ C1 is partly conserved, ␣ C2 appears to be specific of the Orthopneumovirus genus (Fig. 6A). A tentative structural model of P in its most disordered state, summarizing the structural information determined by NMR, is given in Fig. 6B.
Functional Relevance of Transiently Structured Regions in hRSV P-The central role of the phosphoprotein in hRSV replication is associated to its role as a hub inside the RdRp complex, mediating interactions with both viral and cellular proteins (18,24,48). We show here that two regions identified before as binding regions for N mono (24) and M2-1 (17,18), ␣ N1 and ␣ N2 , respectively, display weak ␣-helical propensity. These transient helices might fold completely upon binding, as related for other protein-protein interactions (36,37). Such a hypothesis would be supported by the large number (ϳ30) of N-terminal P residues involved in the interaction with N mono (Fig. 4). Our definition of ␣ N1 matches with the second helix formed by hMPV P 1-28 in the X-ray structure of the N 0 ⅐P complex (40). The sequence of ␣ N1 is rather well conserved among Pneumoviridae (Fig. 6A), suggesting that ␣ N1 constitutes a molecular recognition element for N 0 .
Contradictory data are available for the interaction properties of the ␣-helical C-terminal domain of P. Immunoprecipitation assays showed that deletion of residues 160 -180 impaired N binding in bRSV P (49), but not in hRSV P (50), despite high sequence conservation of N and P between hRSV and bRSV. This internal region partly overlaps with ␣ C1 and was also perturbed in NMR interaction experiments with N NTD (Fig. 4), which prompted us to propose it as a secondary N binding site. Moreover, temperature-sensitive mutations of P were reported in the same region (50). G172S and E176G do not support replication at 37°C, which was linked to decreased interaction between N and P. The triple mutant R174A/E175A/ E176A proved non-functional in a minigenome assay, but was still able to bind to N. Finally, a recombinant E176G RSV virus reverted to Asp 176 (50). These data underline the functional importance of residue Glu 176 , which contributes to a negative cluster with Glu 175 and Glu 179 , exacerbated in the structural context of the ␣ C1 helix (Fig. 6B). If we link these results with the lability of ␣ C1 , it appears that structural modulation of this helix could have a direct impact on RSV replication. In contrast,

JOURNAL OF BIOLOGICAL CHEMISTRY 2125
Leu 198 -Asn 217 has been reported to be a "negative N-binding region," the deletion of which enhanced N binding (16). This region partly overlaps with ␣ C2 and was not affected by N NTD in our experiments (Fig. 4). Disordered Regions in hRSV P Mediate Diffuse as Well as Specific Interactions-In light of the X-ray structure of the hMPV N 0 ⅐P complex (40), the hRSV P 40 /N NTD interaction does not seem to be relevant for the N 0 ⅐P complex. However, it may shed light on the binding properties of ␣ N1. The interaction surfaces on N NTD and N 0 are proximal and connected by the interaction surface with the N-terminal arm of an adjacent protomer in N-RNA rings (Fig. 5E). The P 40 /N NTD interaction might therefore correspond to an intermediate state on the binding/folding pathway of ␣ N1 , whereby hydrophobic interactions play a role, as shown by the hydrophobic face in fully formed ␣ N1 (Fig. 6B), involved in N 0 binding (40). ␣ N1 appears to be a sticky helix that is able to mediate external (N/P) as well as internal interactions, the latter being favored by the tetrameric organization of P.
More generally, we identified transient internal long-range contacts in hRSV by PRE, mainly mediated by regions with SSPs. These seemingly unspecific interactions may play several roles. Compaction of the structure of P can be achieved, preventing unspecific interactions with other proteins in the cell. The interplay between ␣ N1 and ␣ N2 , which both expose hydrophobic faces when stabilized (Fig. 6B), suggests that upon bind-    ing to one site, another site could become competent for binding. Plasticity of the structure of P can also help fulfilling the requirement of simultaneous binding at the C and N terminus to a same protein partner, e.g. to N. At the same time diffuse interactions may play a role by retaining RdRp relevant proteins in hRSV inclusion bodies (8).
The hRSV L binding site was recently reported to span residues Pro 218 -Glu 239 (Fig. 6B), and it was proposed that this region might fold into a helix (30). Under our experimental conditions this region did not display any significant SSP, conformational exchange, or internal contacts, similarly to P NCBD . Analogy with P NCBD suggests that this region might not fold upon binding. Except for Phe 241 , P NCBD is disordered in its bound form, as shown by X-ray crystal structures of N NTD complexed to C-terminal P peptides (39). Although Phe 241 is essen-tial for NC⅐P binding, fuzzy electrostatic interactions mediated by acidic residues as well as phosphorylation of serines in P NCBD significantly contribute to the affinity of P (14,39,51). Our NMR data suggest that even Phe 241 , the linchpin for P binding to the NC, could explore different binding sites. Interactions that come into play in the NC⅐P complex are thus based on disorder and on a balance between recognition and moderate affinity, required for the processivity of the polymerase. This could be a more general scheme for P interactions with other RdRp components.

Experimental Procedures
Plasmids-Plasmids for expression of recombinant hRSV proteins in E. coli were described previously for N terminally GST-fused hRSV phosphoprotein and P fragments listed in     (13,24,25,30) and C terminally His-tagged hRSV nucleoprotein (13), N NTD (N residues 31-252) (14), and the K170A/R185A N mono mutant (24). Expression and Purification of Proteins-All proteins were expressed in E. coli BL21(DE3). 15 N-and 15 N, 13 C-labeled P protein samples for NMR experiments were produced in minimum M9 medium supplemented with 1 g liter Ϫ1 15 NH 4 Cl (Eurisotop), 4 or 3 g liter Ϫ1 unlabeled or 13 C-labeled glucose (Eurisotop), and 100 g ml Ϫ1 of ampicillin, using a protocol adapted from cultures in rich medium (30). Bacteria were disrupted (Constant Systems Ltd) in 50 mM Tris, pH 7.8, 60 mM NaCl, 1 mM EDTA, 2 mM ␤-mercaptoethanol, 0.2% Triton X-100 lysis buffer. After clarification by ultracentrifugation the supernatant was mixed with 2 ml of glutathione-Sepharose beads (GE Healthcare) per liter of culture and incubated for 15 h at 4°C. The resin beads were then washed with thrombin cleavage buffer (20 mM Tris, pH 8.4, 150 mM NaCl, 2 mM ␤-mercaptoethanol, 2.5 mM CaCl 2 ) before addition of 5 units of biotinylated thrombin (Novagen). The beads were incubated for 16 h at 4°C. Thrombin was removed with streptavidin resin (Novagen) according to the manufacturer's instructions. Purification of N protein was carried out as described previously for N mono (24), N NTD (39), and N-RNA rings (13). His tags were not removed. The quality of protein samples was assessed by SDS-PAGE. Samples were subsequently dialyzed into NMR buffer (20 mM sodium phosphate, pH 6.5, 100 mM NaCl) and concentrated to 50 -300 M on Amicon Ultra centrifugal filters (10 kDa cut-off, Merck-Millipore).

. M K F A P E V E N K K E E L K H R S F P S E K P L A G I P N T A T H V . . T K Y N M P P L R F K L P S P R V A A N L T E P S A P P T T P P P T P P Q N K E E Q P K S . . . D V D I E T M
Paramagnetic Spin Labeling-Individual residues of hRSV P, preferentially serines (Ser 23 , Ser 99 , Ser 143 , Ser 156 , Glu 179 , Glu 193 , Ser 237 ), were mutated into cysteines using the Quik-Change mutagenesis kit (Stratagene). Mutant proteins were expressed and purified like wild type. Buffers contained 5 mM dithiothreitol (DTT) as a reducing agent. Protein samples were completely reduced by addition of another 5 mM DTT at room temperature for 2 h. DTT was then removed by passing twice through a Biospin desalting column (Bio-Rad) equilibrated in 50 mM Tris, pH 8.0, 200 mM NaCl. Protein samples were reacted overnight in the dark at 15°C with 10 molar eq of 3-(2-iodoacetamido)-PROXYL radical (IAP, Sigma) in a 45 mM solution in ethanol. Unreacted IAP was removed by applying the samples three times onto Biospin desalting columns equilibrated in NMR buffer.
NMR Spectroscopy-NMR measurements were carried out on a Bruker Avance III spectrometer at a magnetic field of 14.1 T (600 MHz 1 H frequency) equipped with a cryogenic TCI probe. The magnetic field was locked with 7% 2 H 2 O. The temperature was 288 K if not indicated otherwise. Spectra were processed with Topspin 3.1 (Bruker Biospin) and analyzed with CCPNMR 2.2 (52).
NMR Interaction Experiments-NMR interaction experiments were carried out at a magnetic field of 14.1 T by acquiring 1 H-15 N HSQC spectra of 15 N-labeled P constructs in a 30 -100 M concentration range in the presence of 0.25 to 10 molar eq of unlabeled N protein in the form of N NTD , N mono , or N-RNA rings. Samples were prepared by mixing concentrated protein solutions. Line broadening was analyzed by measuring the intensity ratios of amide signals between the spectra with and without N protein in CCPNMR 2.2. Dissociation constants for the P CD ⅐N NTD complex were determined for each perturbed residue by assuming a two-site fast exchange model with a 1:1 stoichiometry and by fitting 1  Complementary experiments were performed with 15 N-N NTD (50 M in 20 mM MES, pH 6.5, 250 mM NaCl buffer) at 293 K, by adding lyophilized P 40 with 1:1 to 12:1 ratios (solubility limit). 15 N Relaxation Measurements-15 N relaxation data were recorded at a magnetic field of 14.1 T and a temperature of 288 K. R 1 and R 2 relaxation rates were measured for 50 -200 M 15 N-labeled P ND , P NDϩOD , P CD , and P ⌬OD with a pseudo-threedimensional version recorded in an interleaved manner with a recycling delay of 4 s. Relaxation delays were 5, 50, 100, 200(*2), 400, 600, 800, 1200, and 2000 ms for R 1 measurements and 34, 68, 136, 204(*2), 271, 339, 407, 543, and 814 ms for R 2 . Heteronuclear 1 H-15 N NOEs were measured by recording two interleaved spectra with on-and off-resonance 1 H saturation during the recycling delay. Peak intensities were extracted in CCPNMR 2.2. Relaxation curves were fitted to a monoexponential model and errors estimated from covariance matrix analysis in CCPNMR 2.2.
PRE Measurements-PREs were determined as the ratios between 1 H-15 N HSQC peak intensities in the paramagnetic and diamagnetic state (I para /I dia ). Measurements were carried out at 14.1 T and 288 K. The diamagnetic state was obtained by incubating the spin-labeled sample with 10 molar eq of ascorbic acid (Sigma) from a 45 mM solution at pH 6.5 for 3-4 h at 303 K.