The ¹⁵N, ¹³C-labeled T. maritima RNase P protein sample was prepared following the previous protocol (Zeng et al., 2018). A pGEX4Ta vector containing the rnpA gene from T. maritima and an N-terminal glutathione S-transferase (GST) fusion protein was transformed and expressed in the BL21 Gold Escherichia coli cell strain and cultured in M9 minimal media at 303 K supplemented with ¹⁵N ammonium chloride and/or ¹³C glucose. After lysis sonication in the presence of cOmplete™ protease inhibitor cocktail (Roche), lysate was separated by centrifugation (28,000 g), filtered, and treated with 800 NIH units of thrombin per 40 ml to remove the GST tag from RNase P protein. A denaturation−renaturation purification strategy was applied (Paul et al., 2001; Krivenko et al., 2002; Buck et al., 2005a; Buck et al., 2005b; Reiter et al., 2012). The thrombin-treated lysate was combined with denaturation buffer (50 mM Tris-HCl pH7.5, 4 mM EDTA, 8 M urea) to a final concentration of 5 M urea, and was subject to 15S cation exchange chromatography. Fractions containing the denatured protein were subsequently dialyzed against refolding buffer (50 mM Tris-HCl pH 7.5, 0.2 mM EDTA, 1 M NaCl) for 1–2 days, upon which the sample was diluted and subject to a second 15S cation exchange column under the identical condition excluding urea. Purified ¹⁵N, ¹³C-labeled RNase P protein fractions were collected, concentrated, and dialyzed against a buffer of 50 mM Tris-HCl pH 7.5, 0.2 mM EDTA.

All NMR experiments were conducted using either low (153 μM) or high (466 μM) ¹⁵N, ¹³C RNase P protein concentrations in 20 mM sodium phosphate pH 6.0, 80 mM NaCl, 50 μM 4, 4-dimethyl-4-silapentane-sulfonate (DSS) at 318 K with 10% (v/v) D₂O in a 3 mm Norell^® select series NMR tube. NMR spectra were acquired on Bruker Avance-III 600 and 800 MHz and 900 MHz spectrometers, equipped with cryogenic probes, as well as Varian VNMRS 600 and 800 MHz spectrometer equipped with a cryogenic probe. NMR spectra were processed with Topspin 3.5.7 (Bruker Inc.) and NMRPipe (Delaglio et al., 1995), and analyzed with NMRViewJ (One Moon Scientific, Inc.) and CARA (Keller 2004). ¹H chemical shifts were referenced with respect to internal DSS, and ¹³C and ¹⁵N chemical shifts were referenced indirectly using nuclei-specific gyromagnetic ratios (Wishart et al., 1995).

NMR chemical shift assignments of the protein backbone were determined on separate samples with protein concentrations of 466 and 153 μM, respectively. At the higher protein concentration, backbone assignments were derived from ¹H, ¹⁵N-heteronuclear single quantum coherence (HSQC) and a set of traditional triple-resonance experiments, including HNCO, HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, and HN(CA)CO. In addition, backbone and side-chain assignments were carried out using a three-dimensional ¹⁵N-NOESY-HSQC experiment (120 ms mixing time). At the lower protein concentration, backbone assignments were confirmed with ¹H, ¹⁵N-HSQC, HNCO, HNCA, and HNCACB, comparing and referencing to the assignments of the higher concentration sample. The chemical shift assignments for the 466 μM and the 153 μM sample were deposited to the biological magnetic resonance bank (BMRB) as accession numbers 27307 and 51021, respectively. To investigate the temperature dependence of the backbone assignments of P protein (153 μM), a set of ¹H, ¹⁵N-HSQC spectra were acquired with identical parameters at 35 (308), 45 (318), 55 (328), and 65°C (338 K). Additional parameters and data processing, including window function, base-line correction, and linear prediction were identical for all ¹H, ¹⁵N-HSQC spectra.

Two ¹³C/¹⁵N-labeled protein samples with concentrations of 153 and 466 μM were prepared for high-resolution DOSY (diffusion-ordered NMR spectroscopy) experiments. 2D DOSY-¹H, ¹⁵N-HSQC experiments were performed on a Varian VNMRS 800 MHz spectrometer with a diffusion delay (Δ) of 70 ms and gradient width (δ) of 3 ms at 318 K. Six different gradient strengths (0.9, 9.7, 18.6, 27.4, 36.2, and 45.1 Gauss/cm) were used to measure diffusion rate of the sample with higher concentration; while 8 gradient strengths (0.88, 7.24, 13.6, 20.0, 26.3, 32.7, 39.1, and 45.4 Gauss/cm) were applied to the low concentration sample. Each experiment was acquired with a data matrix of 168 (t₁, ¹⁵N) × 2048 (t₂, ¹H) complex points and 64 scans. NMRPipe and NMRViewJ were used for data processing and analysis, respectively. The exact centers of the cross peaks for the selected resonances were identified and the fitting of the results and error were propagated in the analysis of signal decay curves. Reference diffusion coefficients for the RNase P monomer, dimer, and tetramer species were calculated from the high-resolution crystal structure (PDB: 1NZ0) using HYDROPRO and HYDRONMR (García de la Torre et al., 2000; Ortega et al., 2011). Molecule A, A/C, and A-D of PDB 1NZ0 were used for simulations of the monomer, dimer, and tetramer, respectively.

Relaxation parameters were measured at 318 K of the 466 μM P protein on a Bruker Avance-III 800 MHz spectrometer. For R₁, five relaxation time points were collected at 100, 200, 400, 600, and 1,000 ms. For R₂, five relaxation time points were collected at 17, 34, 51, 85, and 119 ms. For R₁ and R₂ measurements, a recycle delay of 1 s was used between transitions. Errors were estimated by duplicate measurements using the shortest and longest relaxation time. For heteronuclear NOE measurements, the steady-state ¹H saturation time was 5 s, a recycle delay of 5 s was implemented in the reference experiment, and both reference and NOE measurements were repeated in duplicate. All spectra of relaxation measurements were collected with the data matrix of 300 (t₁, ¹⁵N) × 1,024 (t₂, ¹H) points and 32 scans. NMRPipe was used for data processing and data were evaluated using the rate analysis module of NMRViewJ (Johnson and Blevins 1994; Delaglio et al., 1995). To estimate rotational correlation time from R₁ and R₂ values, calculation were made by performing Eq. 1 (Kay et al., 1989): τ c ≈ 1 4 πν N 6 R 2 R 1 − 7 (1) where ν_N is the ¹⁵N resonance frequency in Hz.

Two sets of NMR chemical shifts were extracted from the high and low protein concentration samples (BMRB #27307 and #51021) and models were generated using chemical-shift (CS-) Rosetta predictive modeling as described elsewhere (Shen et al., 2008; Shen and Bax 2013). The identical number of chemical shift assignments from the “low” and “high” concentration samples were imported to the CS-Rosetta web server to assess predictive modeling results in a direct, comparative manner. NMR chemical shift information was the only experimental data incorporated into model prediction calculations. A total of 3,000 structural models were generated for each set (low and high P protein concentration) and the top 10 best scoring models were selected. Visualization and analysis of predictive models were performed in Molmol (Koradi et al., 1996).

NMR titrations investigated the binding process between the RNase P protein and the ptRNA 5’ leader. A 7-mer 5’ leader sequence of ptRNA (AAGGCGU) was purchased from Dharmacon Co. with purity >95%. Both protein and ptRNA leader were pre-equilibrated in the identical NMR buffer as described above. During titrations, the leader RNA was gradually added with increased molar ratios of 0.2:1, 0.4:1, 1:1, and 2:1 to the 153 μM protein sample. At each molar ratio, a ¹H, ¹⁵N-HSQC spectrum was acquired at 318 K. All data were processed by NMRpipe and analyzed by NMRviewJ. The analysis of amide chemical shift changes (Δδ) were calculated following Eq. 2 (Schumann et al., 2007; Collier et al., 2014): Δδ ( ppm ) = Δδ ( 1 H ) 2 + ( 0.152 Δδ ( 15 N ) ) 2 (2) Δδ s of the peaks at the highest leader RNA concentration (5’ leader/protein molar ratio = 2:1) were used for chemical shift perturbation (CSP) analysis. Dissociation constants (K_D) for each shifting peak were calculated by fitting the data to Eq. 3: Δ δ ( ppm ) = Δ max ( K D + [ L ] + [ P ] ) − ( K D + [ L ] + [ P ] ) 2 − 4   [ P ] [ L ]   2 [ P ] (3) where ∆δ(ppm) is the amide chemical shift changes, Δ_max the maximum amide chemical shift changes, [P] and [L] the protein and ligand concentrations, respectively. Only defined monomeric chemical shift assignments were used for the K_D calculation. However, the CSP analysis included monomer-only peaks and identified monomer-dimer assigned peaks. In the 153 μM P protein sample, dimer-only peaks undergo chemical shift changes upon RNA binding but were too weak and unreliable to be included in the analysis.

HADDOCK was used to predict the structure of the RNase P protein-5’ ptRNA leader complex, based on the CSP results of NMR titrations (Dominguez et al., 2003). Certain residues which have ∆δs larger than ∆δ_average + STD_∆δ were set as active residues, while amides having ∆δs between ∆δ_average and ∆δ_average + STD_∆δ were set as passive residues. The initial structure of the P protein was the representative structure from the CS-Rosetta calculated ensemble that incorporated the NMR peak assignments at low protein concentration. The initial model of the 7-mer ptRNA leader was generated in Coot (Emsley and Cowtan 2004). The docking calculations were performed on the HADDOCK web server; generating 127 predicted models using standard HADDOCK settings.

The functional T. maritima ribonuclease P holoenzyme structure contains single RNA (ribozyme) and protein components, yet the large 110 kDa P RNA-only crystallizes as a dimer and the small 14 kDa P protein-only crystal structure (PDB-1NZ0) contains two dimers within its asymmetric unit (Figure 1) (Paul et al., 2001; Kazantsev et al., 2003; Torres-Larios et al., 2005; Reiter et al., 2012). The P protein dimer interface in the crystal structure obscures the 5’ leader RNA binding track (Kazantsev et al., 2003) and this is problematic in characterizing the 5’ leader pre-tRNA substrate-protein interaction. Preliminary NMR experiments reveal that highly purified T. maritima RNase P protein, hence termed P protein, persists as a dimer at 25°C but is a heterogeneous species in solution. Thus, oligomerization is not merely an artifact of crystallization. To define the NMR assignments of an RNA binding competent P protein, ¹H, ¹⁵N-HSQC spectra at various temperatures and on samples containing protein concentrations in the range of 153–466 μM were collected and analyzed. A series of triple-resonance experiments at both high (466 μM) and low (153 μM) protein concentrations were also collected and analyzed at 45°C. At P protein concentrations of 153 and 466 μM, there are always 2 conformations in solution (Figure 2, Supplementary Figure S1). The analysis of various concentration and temperature spectra enabled the identification and assignment of two sets of amide peaks that correspond to two conformations in solution (BMRB #27307 and #51021).

When comparing P protein low (153 μM) and high (466 μM) concentrations at 45°C, we observed overall peak intensity changes between the two conformations depending upon the protein concentrations (Figure 2). This suggests a concentration-dependent shift in the equilibrium of the two species in solution. Between 35 to 65°C, similar temperature dependent chemical shifts were observed for both samples but no significant changes occurred in the relative peak intensities for the two conformations (Supplementary Figure S2). This suggests that both conformations are stable over a wide temperature range and that the oligomeric equilibrium of the P protein is concentration dependent and temperature-independent in the range of 35–65°C.

We report high confidence backbone assignments at high (96.5% amides, 99.1% for all C_α and C_β, and 97.4% C’ resonances) and low P protein concentrations (93.0% amides, 90.6% for C_α, 70.7% for C_β, and 87.2% C’ resonances). Nearly 50% of all residues have different chemical shift assignments in two conformations of the ¹H, ¹⁵N-HSQC, thus facilitating the accurate analysis of the distinct conformations of the P protein (Supplementary Table S1, BMRB #27307 and #51021).

To define oligomerization status of the distinct conformations, 2D-DOSY spectra were acquired at low (153 μM) and high (466 μM) P protein concentrations (Table 1). Analysis was performed on 48 residues that contained distinct chemical shifts for the two conformations in the ¹H, ¹⁵N-HSQC spectra. Peak intensities of the dominant signal at low or high P protein concentrations were identified and selected for signal decay curve analyses from DOSY experiments. The average diffusion coefficients at 45°C were determined from the analysis of the selected peak intensities against gradient field strengths (Table 1). Experimental diffusion coefficient values of the low (153 μM) concentration conformer (D_low = (2.17 ± 0.36)*10^–10 m²/s) and the high (466 μM) concentration conformer (D_high = (1.87 ± 0.40)*10^–10 m²/s) are consistent with HYDROPRO and HYDRONMR software-simulated diffusion coefficients of a monomer (D ∼ 1.96 *10^–10 m²/s) and dimer (D ∼ 1.54 *10^–10 m²/s) P protein conformation (Table 1). The calculated, reference diffusion coefficients from HYDROPRO and HYDRONMR were derived from the high-resolution crystal structure PDB-1NZ0. A value difference of 0.3 for the experimental diffusion data (D ∼ 2.17 vs. 1.87) compared to the 0.42 value difference in the calculated values (D ∼ 1.96 vs. 1.54) reflect the fact that both the low and high concentration samples contain mixed monomer/dimer populations. While there is no evidence of higher order oligomerization, the monomeric form predominates at low protein concentration and the dimeric form primarily occurs at higher concentrations. The resolution and calculated error in the NMR-derived diffusion coefficients likely reflect the mixed populations present at both concentrations. Nonetheless, this comparative diffusion coefficient analysis identifies that the assigned resonances in the low (153 μM) and high (466 μM) concentration P protein samples correspond primarily to the monomeric and dimeric species, respectively.

Verification of the identified monomer/dimer species was also determined through the analysis of the R₂/R₁ relaxation data. Despite heterogeneous sample conditions, 48 chemical shifts were extracted from two sets of residues in the spectra. One set of analyzed peaks was representative of the dominant conformer at the high concentration (466 μM) and the other set of peaks was representative of a minor conformer that is also present at 466 μM P protein. The observed minor conformer chemical shifts at 466 μM are identical to the dominant conformer at 153 μM P protein concentration. R₁ and R₂ rates were determined for the selected residues of the minor and dominant P protein conformers and their average R₂/R₁ ratios identified as 5.95 and 17.04, respectively, (Figures 3A–C). R₂ values were considerably higher for peaks arising from the dominant conformer in comparison to peaks from the minor conformer (Figures 3A,C). In addition, rotational correlation times (τ_c) (Kay et al., 1989) were determined from R₂/R₁ ratio values, revealing τ_c,low and τ_c,high values of approximately 5.3 and 9.7 ns, respectively. These parameters support a monomer–dimer equilibrium model, where the monomer conformer (τ_{c, monomer} ∼ 5.3 ns) predominates at ∼150 μM and the dimer species (τ_{c, dimer} ∼9.7 ns) predominates at higher P protein concentrations. The estimated τ_c values are substantially smaller than a typical 14 or 28 kDa idealized spherical protein at 25°C, but are consistent with the estimated empirical range of correlation times at 45°C (http://nickanthis.com/tools/tau). Thus, analysis of two independent NMR methodologies (DOSY and NMR relaxation) both validate that monomer and dimer P protein assembles coexist in solution. The monomeric conformation predominates at 153 μM P protein concentrations whereas the dimer species predominates at higher (466 μM P protein) concentrations.

The two sets of backbone NMR assignments (HN, C_α, and C’) based upon BMRB entries #27307 and #51021 were input into CS-Rosetta web server to generate models of the monomer and dimer forms of the P protein in solution. To inspect the structural differences caused only by chemical shifts, the same numbers of chemical shifts from exactly the same assigned resonances were implemented for CS-Rosetta calculations. The ensemble models converge well and reflect the dimeric and monomeric forms of the P protein in solution (Figures 4A,B, respectively). All ensembles include the 10 lowest-energy models and both sets of predicted structures contain near identical backbone folds that are consistent with the crystal structure (PDB: 1NZ0). However, whereas the backbone RMSD of residues 24–117 of the monomer and dimer species compared to PDB-1NZ0 are 0.97 Å and 1.19 Å, respectively, both models show a higher degree of flexibility at the N-terminus (residues 1–23). Both monomeric and dimeric conformers exhibit a lack of structural convergence within the first 23 residues at N terminal of the P protein, with a higher degree of structural variability observed in the monomeric-derived ensemble (Figure 4B).

A comparison of NOEs within the N-terminus (residues 8–12, 18–22) and a defined α-helical region (residues 59–68) reflect the extent or lack of structural convergence observed in CS-Rosetta models (Supplementary Figure S3). Poorly ordered regions at the N-terminus (residues 8–12) reveal only a few HN-HN (i,i+1) NOEs, a partially structured region that contains α-helical character (residues 18–22, α₁ helix) reveals sequential (i, i+1) and some medium range NOEs, and a highly ordered region (residues 59–68, α₃ helix) shows HN-HN (i, i+2), H_α-HN (i, i+3), and H_α-HN (i ,i+4) medium and long range NOE connectivities. The representative NOEs identified in ¹⁵N NOESY-HSQC spectrum highlight the lack of structural convergence in the N-terminus observed in the CS-Rosetta model.

Thus, comparative modeling and analysis of NOE data reveal that the N-terminal P protein is more flexible than other regions of the P protein. The CS-Rosetta data also suggest that the N-terminal region of the monomer conformation is more flexible than the dimer conformer despite using the exact same number of chemical shifts from the same assigned resonances. The alpha1 helix (α1) extends from residue 14–23 in the dimeric assigned CS-Rosetta ensemble with a backbone RMSD of 0.38 Å relative to PDB 1NZ0 (Figure 4B). This implies that the dimerization interface in solution is similar to the X-ray structure (PDB 1NZ0) (Figure 1B, Supplementary Figures S4, S5). In contrast, the monomer assigned CS-Rosetta models reveals a truncated α1 helix (residues 17–23) that appears intrinsically disordered (RMSD ∼1.56 Å) relative to 1NZ0. Taken together, this NMR-based data supports a structural mechanism whereby the N-terminal region (residues 1–12) becomes stabilized by the P RNA in the RNase P holoenzyme crystal structure (Reiter et al., 2010) and the transient structure of the N-terminal helix (residues 13–23) helps to optimally align the 5’ leader ssRNA substrate region (Niranjanakumari et al., 1998; Zahler et al., 2003; Buck et al., 2005a; Buck et al., 2005b; Sun et al., 2006; Niranjanakumari et al., 2007; Koutmou et al., 2010b; Reiter et al., 2012; Lin et al., 2016; Niland et al., 2017).

To understand the mechanism of RNase P holoenzyme activation and ptRNA cleavage, it is essential to study the structure and binding interaction between RNase P protein and 5’ ptRNA leader. We chose to work with RNase P from Thermotoga maritima because extensive structure, biochemical analyses, and small molecule screening studies have been performed on the RNase P holoenzyme (Paul et al., 2001; Buck et al., 2005a; Torres-Larios et al., 2005; Reiter et al., 2010; Reiter et al., 2012; Madrigal-Carrillo et al., 2019). We determined that the functionally relevant P protein monomer predominates at low (153 μM) concentrations, enabling NMR titrations to probe the binding interactions of the P protein and 5’ leader RNA. Chemical shift perturbation (CSP) analysis was performed when the 5’ leader RNA (5′- AAGGCGU-3′) was titrated into the P protein to a ratio of 2:1. Higher RNA:protein titration ratios were attempted >4:1 yet were followed by rapid sample precipitation, prohibiting the accurate collection of reliable chemical shift information. Nonetheless, CSP analysis of the protein-RNA titration was monitored via ¹H, ¹⁵N-HSQC experiments, focusing on 57 assignments attributed to the monomer, 28 peaks corresponding to both monomer and dimer conformers, and 8 resonances attributed solely to the dimer species (Supplementary Figure S6). Dimer-identified peaks are minor at the 153 μM P protein concentrations (Figure 2); yet detectable minor chemical shift differences (Δδ) upon ligand binding were observed for both monomer and dimer species (Supplementary Figure S7). Specifically, residues Q28 and F82 represent dimer peaks that were shifted upon RNA titration at 153 μM P protein concentrations. Both Q28 and F82 HN resonances are structurally accessible to bind RNA in the monomer and dimer conformers yet only Q28 exists at a functionally relevant RNA binding interface. While it is possible that further RNA addition may shift the equilibrium from a dimer species towards the monomeric conformation, it is unclear from the available NMR data whether RNA binding substantially alters the dimer-monomer conformational equilibrium.

Nonetheless, the magnitude of changes in peak positions induced by RNA binding correlate to weak protein-RNA interactions that occur in fast exchange on the NMR timescale (Figure 5A). The dissociation constant (K_D) was determined by examining the individual change in peak positions (Δδs) against ligand concentrations for residues attributed to only the monomer P protein. 16 data sets were fitted and then combined for a global fitting K_D determination, yielding an overall apparent K_D ^global of 47.2 ± 14.0 μM (Figure 5B). Significant chemical shifts were observed exclusively in the identified monomer assignments (11 residues, S4, R8, G24, L27, V34, V49, R73, ⁸²F, V85, I87, and G117), where significant shifts imply Δδ ≥ ∆δ_average + STD_∆δ (Figure 5C). Nearly all mapped protein residues undergo fast exchange kinetics upon RNA binding except for amide residues of S26 and V99, which are selectively line broadened and undergo millisecond exchange kinetics. The observed different binding modes may be due to the 5’ leader RNA interactions with the monomer and a heterogeneous monomer-dimer species, thus complicating interpretation. For this reason, only residues that have defined monomeric chemical shift assignments and undergoing fast exchange were included in the dissociation constant calculation. The distributed RNA binding P protein residues were mapped on one of the CS-Rosetta derived model and, as expected, residues with the largest changes in peak positions are located across the β-sheet binding cleft (Figure 5D). These NMR titration experiments validate the 5’ leader-binding interface previously identified in the T. maritima RNase P holoenzyme-tRNA complex (Reiter et al., 2010).

Based on CSP results, HADDOCK was performed to generate an ensemble of the P protein monomer-5’ pre-tRNA leader interaction in solution (Figures 5D,E). Chemical shift perturbation results were included as ambiguous restraints between the individual P protein residues of the 5’ leader RNA. Monomeric P protein residues exhibiting peak changes located near the central β-sheet cleft were selected for docking, consistent with T. maritima RNase P structure and biochemical studies (Reiter et al., 2010; Reiter et al., 2012). Models generated were solely dependent upon CSP data and emphasize the flexibility of the P protein in accommodating a broad array of 5’ leader RNA nucleotides.