Abstract
Summary Double-stranded RNA-binding domains (dsRBDs) are involved in a variety of biological functions via recognition and processing of dsRNAs. Though the primary substrate of the dsRBDs are dsRNAs with A-form helical geometry; they are known to interact with structurally diverse dsRNAs. Here, we have employed two model dsRBDs – TAR-RNA binding protein and Adenosine deaminase that acts on RNA – to understand the role of intrinsic protein dynamics in RNA binding. We have performed a detailed characterization of the residue-level dynamics by NMR spectroscopy for the two dsRBDs. While the dynamics profiles at the ps-ns timescale of the two dsRBDs were found to be different, a striking similarity was observed in the μs-ms timescale dynamics for both the dsRBDs. Motions at fast μs timescale (kex > 50000 s−1) were found to be present not only in the RNA-binding residues but also in some allosteric residues of the dsRBDs. We propose that this intrinsic μs timescale dynamics in two distinct dsRBDs allows them to undergo conformational rearrangement that may aid dsRBDs to target substrate dsRNA from the pool of structurally different RNAs in cellular environment.
Introduction
Protein-RNA interactions play a significant role in the day-to-day cellular functions. The interactions between double-stranded RNA (dsRNA) and dsRNA-binding domains (dsRBDs) are involved in complex cellular processes like RNA splicing (Fu et al., 2016), RNA editing (Stephens, Haudenschild and Beal, 2004), RNA maturation (Ha and Kim, 2014), RNA transport across cellular membranes (Wang et al., 2011), etc.; and hence play a central role in regulating and executing major cellular pathways governing life. Structurally, dsRBDs contain α1-β1-β2-β3-α2 fold spanning a length of about 65-70 amino acid (Bycroft et al., 1995; Kharrat et al., 1995; Masliah, Barraud and Allain, 2013). Three regions of the dsRBD namely, 1) middle of helix α1, 2) loop between β1 and β2, and 3) N-terminal residues of the helix α2 are involved in the interactions with the backbone of the dsRNA at its minor-major-minor grove spanning a length of ~12 bp. Most dsRBDs are known to interact with dsRNAs having A-form helical structure independent of the sequence (Masliah, Barraud and Allain, 2013; Tants et al., 2017). However, the target dsRNAs often have structural defects like internal mismatches/loops and bulges perturbing the A-form helical shape of the RNA. Such defects also lead to change in the spread of the minor-major-minor groves of the dsRNA (each represented by a unique structure defined by Euler angles (Bailor, Sun and Al-Hashimi, 2010)), thereby affecting the interaction between the dsRBD and dsRNA. Acevedo et al. have shown that interaction between dsRBD2 of TRBP does not lead to a significant conformational change in the substrate dsRNA (Acevedo et al., 2016). Thus, to effectively and efficiently target a dsRNA from the wide pool of structurally different dsRNAs in the cellular matrix, dsRBDs must adapt itself to accommodate the conformational heterogeneity of the target dsRNAs.
The structural adaptations required in the proteins to perform a particular function involve conformational exchange processes characterized by motions at multiple timescales, including μs-ms timescale (Kleckner and Foster, 2011; Kamba, Nagata and Katahira, 2015; Chakrabarti et al., 2017; Jie, Löhr and Barbar, 2017). These structural adaptations involve domain motions (Beach et al., 2005; Kovrigin and Loria, 2006; Fraser et al., 2009), catalytic interactions (Eisenmesser et al., 2002; Boehr et al., 2006), allosteric effects (Popovych et al., 2006; Tzeng and Kalodimos, 2009), etc. Koh et al. have shown involvement of dynamics at the dsRNA-dsRBD interface where the diffusion of dsRBDs of TRBP along pre-miRNA favor its cleavage to the miRNA:miRNA* duplex by Dicer – an RNase III enzyme (Koh et al., 2013). Heber et al. have recently compared binding affinities and binding stabilities of various dsRBDs in a multi-domain protein for RNA-recognition in Drosophila Staufen protein (Heber et al., 2019). Fareh et al. have shown that the recognition of target dsRNA by TRBP involves dual binding modes to discriminate between the target dsRNA and other cellular dsRNAs to which protein is exposed (Fareh et al., 2016).
NMR spectroscopy can provide detailed insights into such dynamic processes at an atomic resolution (Ishima and Torchia, 2000; Kempf and Loria, 2002; Loria, Berlow and Watt, 2008; Gaspari and Perczel, 2010; Kleckner and Foster, 2011). Literature available till date discusses dynamics studies by NMR spectroscopy on the dsRBDs that have been restricted to the characterization of fast (ps-ns) timescale motions in these proteins and a presence of Rex calculated using model-free analysis of such motions (Nanduri et al., 2000; Wostenberg, Quarles and Showalter, 2010; Wostenberg et al., 2012; Hartman et al., 2013). Although Chiliveri et al. have detected slow μs-ms motions (using CPMG relaxation dispersion experiments) in the α1 helix of dsRBD1 of DRB4 protein of A. thaliana (Chiliveri et al., 2017); they only have described Rex values in their study and the detailed characterization of dynamics parameters including pB (population of excited state), Δω (chemical shift difference between ground and excited state), and kex (exchange rate between ground and excited state) remain to be performed. Apart from this study, there is no literature thus far that reports on the comprehensive characterization of μs-ms timescale dynamics in a dsRBD.
The current study attempts to characterize detailed dynamics of two model dsRBDs – namely dsRBD1 of human TAR RNA Binding Protein (TRBP2) and Drosophila Adenosine Deaminase Acting on RNA (dADAR) protein – at multiple timescales using NMR spectroscopy with nuclear spin relaxation and relaxation dispersion experiments. TRBP, first identified as a protein involved in an interaction with HIV TAR RNA (Gatignol et al., 1991), is also known to assist Dicer (an RNase III endonuclease) to process pre-miRNAs to generate the mature microRNAs (Daniels et al., 2009; Noland, Ma and Doudna, 2011; Lee and Doudna, 2012). Being a part of the canonical microRNA biogenesis pathway, TRBP is exposed to a wide array of structurally different dsRNAs. In addition to canonical α1-β1-β2-β3-α2 secondary fold, an additional N-terminal helix α0 was recently detected at the N terminus of dsRBD1 of TRBP2 (isoform1 of TRBP) (Masliah et al., 2018; Paithankar et al., 2018), the function of which remains to be understood. The second model protein used for the current investigation is the ADAR protein, known to catalyze the conversion of Adenosine to Inosine (A → I), which is critical for regulating gene expression (Bass, 2007; Barraud et al., 2012; Savva, Rieder and Reenan, 2012). Human ADAR2 protein is reported to interact with dsRNAs in a sequence-specific manner (Stephens, Haudenschild and Beal, 2004; Stefl et al., 2010), while Drosophila ADAR-dsRBD1 is reported to have less sequence-specificity (Barraud et al., 2012). While both the proteins contain multiple dsRBDs, this study focusses on the dsRBD1 of these proteins to allow understanding of the intrinsic dynamics of single dsRBD (keeping any other possible domain-domain interaction aside). We observe that the two dsRBDs have a different dynamic profile at ps-ns timescale, while their dynamics profile at fast μs timescale is similar and is present not only in the RNA-binding regions but also in few allosteric residues of the dsRBDs. Being a common feature observed in two dsRBDs from two different proteins from two different species, we propose that the slow timescale motions are an intrinsic feature of the dsRBDs, that allow them to access the ensemble of conformations, required to interact with topologically different RNAs in the cellular environment.
Results
Conformational plasticity of the two dsRBDs
To investigate the conformational plasticity, if present, in the dsRBDs, CD-based melting studies were carried out. The temperature-dependent far-UV CD spectra for TRBP2-dsRBD1 showed that the protein conformation starts to open at 25°C (where all the measurements are carried out in this study) (Figure 1). The unfolding event was characterized by an enthalpy change (ΔH) of 41.11±2.32 kcal/mol and corresponding TM of ~44°C obtained by fitting of the CD data with Gibbs-Helmholtz equation (see Methods section). The temperature-induced unfolding of dADAR-dsRBD1 using far-UV CD showed a very similar profile with enthalpy change (ΔH) of 33.58±1.57 kcal/mol and corresponding TM of ~44°C (Figure 1). Though the thermodynamics parameters, obtained from Gibbs-Helmholtz equation, cannot be correctly interpreted due to irreversible nature of the unfolding transition (Greenfield, 2006), the melting profiles with low TM values (similar values were obtained from the derivative plot dθ/dT vs Temperature, Figure 1E,F) hinted towards low thermal stability of the proteins at 25°C.
TRBP2-dsRBD1 and dADAR-dsRBD1 show differential dynamics at the ps-ns timescale
To investigate the timescale of motions present in the proteins, NMR spin relaxation data (R1, R2 and [1H]-15N NOE) was analysed for 69 non-overlapping peaks out of 96 assigned peaks for TRBP2-dsRBD1 (Paithankar et al., 2018) and for 58 such peaks out of 74 assigned peaks for dADAR-dsRBD1 (Barraud et al., 2012) (Figure S1). The analysis of the data using the extended model-free approach (Lipari and Szabo, 1982b, 1982a; Clore et al., 1990) allowed the extraction of dynamic parameters at the ps-ns timescale.
The order parameter S2 plotted against the residue number for TRBP2-dsRBD1 have been depicted in Figure 2A,B. The theoretical values of S2 range from 0 to 1, where 0 indicates high flexibility and 1 indicates a rigid residue (Lipari and Szabo, 1982b, 1982a). The plot showed large variations of S2 values throughout the protein backbone. The mean S2 for the core RNA- binding region (31-95 aa) was found to be 0.61; which suggests the presence of fast motions at ps-ns timescale in the structured regions of the protein. The residues in the α0 helix (19-25 aa) of the protein showed higher S2 than the N-terminal unstructured region (1-16 aa) and lower S2 than the dsRBD core indicating the structured yet dynamic stretch of the residues in this helix. Further, the S2 values in the stretches near the RNA-binding region showed a higher amplitude of motions compared to the overall motions of the dsRBD core. Around 78% of the analyzed residues of TRBP2-dsRBD1 showed the best fit to a ‘model-free’ model with motions in two timescales. The fast internal motions of the timescale of tens of picoseconds reflected random thermal motions, which were observed at the terminal and in the loop regions in the protein. Slower motions at nanosecond timescale (1-4 ns), but faster than the global tumbling time of the protein (7.64 ns), were present throughout the protein backbone.
The residues best fit with a model containing exchange term (Rex) indicated the presence of motions at μs-ms timescale, which involved conformational exchange processes (Figure 3A). The residues having large Rex were mainly observed in helix α1 and α2 near the RNA-binding region, and the loop regions of the protein which correlated with the lower S2. Interestingly, a pair of residues in helix α2 (A88 and E89) showed high S2 values and Rex at the same time. The S2 values calculated from model-free analysis of nuclear spin relaxation data for dADAR-dsRBD1 showed higher rigidity (when compared with that of TRBP2-dsRBD1) in the structured region of the protein with an average S2 of 0.80 (Figure 2C,D). The only residues that showed more flexibility with lower than average S2 value were the loop regions. The residues in the protein backbone showed motions at ns timescale (~2 ns) with fast motions at ps timescale in loop residues. Interestingly, many residues near RNA-binding region (T65, L72, E81, S82, T84, F92, I94, Q106, G107, E116, L122, R123, and S124) showed the presence of μs-ms timescale motions (similar to what was observed in TRBP2-dsRBD1) as comprehended from Rex terms (Figure 3B) calculated by the extended model-free analysis. The global tumbling time for the dADAR-dsRBD1 was estimated to be a 9.39 ns from the model-free analysis.
Conformational exchange in the μs-ms timescale in dsRBDs
Driven by the presence of Rex along the length of the two dsRBDs, dynamics at μs-ms timescale was tested using Carr-Purcell-Meiboom-Gill relaxation dispersion (CPMG-RD) experiments (Loria, Berlow and Watt, 2008; Kleckner and Foster, 2011). However, R2eff was found to be invariant to the CPMG frequencies (Figure S2 and Figure S3) used to create the dispersion suggesting that the dsRBDs are not conformationally exchanging at the timescale (~0.3 to 10 ms (Kleckner and Foster, 2011)) sensitive to CPMG-RD experiments. Next, dynamics at the μs-ms timescale was tested using Heteronuclear Adiabatic Relaxation Dispersion (HARD) (Mangia et al., 2010; Traaseth et al., 2012; Chao and Byrd, 2016) NMR experiment that monitors NMR spin relaxation in a rotating frame in the presence of hyperbolic secant (HSn, where n = stretching factor) adiabatic pulses. HARD NMR experiment allows to explore the conformational exchange processes occurring on the 10 μs to 10 ms timescales. The plot of relaxation rates R1ρ and R2ρ against residue numbers showed that as the applied field strength increases from HS1 to HS8, the R1ρ rates increased and the R2ρ rates decreased for both the dsRBDs (Figure 4). The dispersion in the relaxation rates observed points to the dynamic processes that can be probed by this method. Assuming two-state exchange between a ground state and an excited state, fitting of the R1, R1ρ, and R2ρ rates to the solution of the Bloch-McConnel equations by using geometric approximation approach (Chao and Byrd, 2016) allowed to extract the residue-specific dynamic parameters (the conformational exchange rate, kex, the equilibrium populations of the two states, pA or pB, and the chemical shift difference, Δω).
Since there is a partial overlap in the timescale of motions that can be probed by CPMG-RD and HARD NMR experiments, we simulated the R2eff rates for CPMG relaxation dispersion using the dynamics parameters obtained from HARD NMR experiments to verify the absence of dispersion in R2eff rates as observed in CPMG-RD. The simulations for R2eff values for CPMG relaxation dispersion experiment were performed using the following equation (Kleckner and Foster, 2011): where, Ф = PA * PB * Δω2 with pA, pB, Δω, and kex values used in simulation were obtained from HARD NMR data analysis. νCPMG is varied in the simulations as per the range used to record CPMG-RD data.
The simulation of the R2eff rates for CPMG relaxation dispersion with equation 6 with the dynamics parameters extracted from HARD data showed no dispersion in the R2eff relaxation rates.
Careful analysis of the kex and the pB values for TRBP2-dsRBD1, depicted on the protein structure (Figure 5A), showed that the residues in the core RNA-binding region exhibited large kex (>50000 s−1) with a large excited-state population (Figure S4). The L35 and Q36 residues (lie in the middle of the α1), the residue E54 (in the β1-β2 loop), and the residue A82 (lies towards the N-terminal of the helix α2) displayed large kex values. A few of the non RNA- binding residues (A25, N26, N61, F64, T67, and V68 – although not directly interacting with dsRNA, but are present in purlieu of RNA-binding residues) also showed kex > 50000 s−1. The observed exchange in these residues might have originated to maintain the overall tertiary fold of the protein during the conformational exchange happening in the RNA-binding residues. The residues in the middle of β strands, which are involved in maintaining the tertiary structure of the protein through a network of interactions (hydrophobic interactions or charge-charge interactions or H-bonding interaction) and the residues in loop/terminal regions displayed a conformational exchange occurring at the rate of 50000 s−1> kex >8000 s−1. These residues included (M1, T10, G13, C14, L51, V66, C73, H94, L95, A104) and about 10-40% of the population of these residues existed in excited states. Rest of the residues showed populations of <10% and are exchanging at rates <5000 s−1. Along with the wide range of kex and the pB values, wide dispersion (5-1100 Hz) in the corresponding Δω values (between ground state and excited state) in 15N indicated the availability of a large conformational space which the protein may populate allowing target dsRNAs to select a conformation for efficient binding (Figure S4).
Despite differences in the primary sequence of the two dsRBDs, a strikingly similar behaviour of kex was observed in dADAR-dsRBD1 (Figure 5 and Figure S4). Residues lying in the core RNA-binding region of the dsRBD, including residues in the middle of α1 (V66, A67, M68, N70), residues in β1-β2 loop (T84, A89, L91), and residues towards N-terminal of α2 (R114, I115, E116, A117) were all found to have kex > 50000 s−1. Similar to TRBP2-dsRBD1, a few of the non-RNA binding residues (L76, T93, I94, S95, V96, Q101, and L104) also showed kex > 50000 s−1 in the region suggesting allosteric role (L76, I94, and L104 directly interact with RNA-binding residues) for the proposed conformational adaptation by dsRBD. Other residues in the non-RNA-binding region had kex between 8000 s−1 to 50000 s−1 and were found to lie in either loops or in the middle of β strands and the fraying end of helix α2. Rest of the residues showed populations of <10% and are exchanging at rates <5000 s−1. A similar chemical shift dispersion was observed in the Δω values in 15N for dADAR-dsRBD1 (Figure S4) as observed for TRBP2-dsRBD1.
Discussion
In order to understand the role of protein dynamics in interaction with the topologically broad range of dsRNAs, two dsRBDs – TRBP2-dsRBD1 and dADAR-dsRBD1 – were tested for the presence of dynamics. The sequence identity between the two dsRBDs was found to be 33.3% and the similarity of 23.6% with the canonical tertiary fold common to all dsRBDs. Despite the large differences in the primary sequence, both the proteins showed similar overall flexibility as evidenced by the similar CD melting profile with a TM of ~44°C (Figure 1). Fraction-folded, as calculated from the CD data for TRBP2-dsRBD1 and dADAR-dsRBD1, showed that ~95% of the population of the dsRBDs exist in folded state at 25 °C, where all NMR experiments are recorded, suggesting presence of large conformational entropy.
Interestingly, the analysis of nuclear spin relaxation experiment by extended model-free approach showed lower order parameter values (S2<0.6) in the structured region of the protein in addition to the low values in the loop and terminal areas for TRBP2-dsRBD1. This supported the internal flexibility in an otherwise globular protein. Order parameters for dADAR-dsRBD1, however, showed routinely observed profile with lower values (S2<0.4) in the loop and terminal regions and higher values (S2>0.8) in the structured regions of the protein. The global tumbling time for the two proteins as calculated from the model-free analysis was higher (7.64 ns for TRBP2-dsRBD1 and 9.39 ns for dADAR-dsRBD1) than that observed routinely for a ~11-12 kDa protein. A higher global rotational correlation time is attributed to the presence of conformational heterogeneity, the presence of an oligomeric species, and the presence of large amount of unstructured regions in the two proteins. The presence of an oligomeric species affecting the correlation time is ruled out by size-exclusion chromatography coupled with multiple angle light scattering study of the proteins. While dADAR-dsRBD1 (rotational correlation time of 9.39 s) showed a pure monomeric species, TRBP2-dsRBD1 (rotational correlation time of 7.64 ns) showed only a minor population of tetramer (Figure S5). Secondary structure analysis of the TRBP2-dsRBD1 shows that it has 45% residues in the unstructured and 55% residues in the structured regions. A similar analysis showed 59% residues in unstructured and 41% residues in structured regions of dADAR-dsRBD1. Thus, the presence of unstructured regions in the two dsRBDs could be one of the reasons for relatively higher correlation times. The presence of slow motions at μs-ms timescale suggested by model-free analysis is further tested using heteronuclear adiabatic relaxation dispersion experiment.
Analysis of HARD NMR data showed large dispersion profiles in both R1ρ and R2ρ rates when measured as a function of adiabatic pulse stretching factor, suggesting the presence of slow-dynamics at μs-ms timescale (Figure 4). The presence of strikingly similar dispersion profile in relaxation rates in both the dsRBDs indicated parallel dynamics at μs-ms timescale. Residues in the core RNA-binding region were found to have kex > 50000 s−1 to excited state(s) and are characterized by a large change in 15N chemical shift values (Δω) and significant populations in excited state(s) (Figure 5 and Figure S4). Significant excited state population with a dispersion of values of Δω from 5-1100 Hz is suggestive of large conformational entropy present in the dsRBDs. In addition to the faster μs timescale motions in the RNA-binding regions, conformational exchange with slower exchange rates (kex < 5000 s−1) was found all along the protein backbone in both the proteins, suggesting a possibility of conformational rearrangement all along the protein chain. Although CD probed equilibrium unfolding/folding ratio as a function of temperature, and HARD relaxation dispersion experiments measured dynamics in a tight window of μs-ms timescale motions, intriguingly, they both pointed toward the presence of conformational entropy in both the dsRBDs in the experimental conditions used. Interestingly, despite the significant differences in the primary sequence of the two dsRBDs, the motions at μs-ms timescale are highly similar in RNA-binding residues and in a few allosteric regions. This suggests that the analogous dynamics observed by relaxation dispersion method in two dsRBDs are not affected by primary sequence and are intrinsic to the protein’s secondary/tertiary structure. Additionally, higher kex values observed in non-RNA-binding regions manifests in the allosteric role being played by these residues to maintain the structural integrity in the conformational exchange process (Figure 5 and Figure S4). Since the exchange is present all along the protein backbone and the sign of Δω cannot be determined by the HARD experiment (Chao and Byrd, 2016) we could not calculate the structure of the excited state(s).
Here, for the first time, we report the detailed characterization of the conformational dynamics of the dsRBDs, which suggests that the presence of similar μs-ms timescale motions in two dsRBDs. Further, it hints towards these motions as a general feature of dsRBDs. These conformational dynamics lead to the creation of a conformational pool for the dsRBDs that might allow it to interact with target dsRNAs in the cellular environment. This may further allow processing of the accurate substrate dsRNAs by the catalytic protein partners of dsRBDs. Identification and characterization of the dynamics at the μs-ms timescale shows conformational adaptability of the dsRBDs and is a step towards in understanding the mechanism of non-specific interaction between dsRBD and target dsRNAs.
Conclusions
The observations show the presence of conformational plasticity in two dsRBDs irrespective of differences in their primary sequences. The effect of different primary sequences was reflected in the ps-ns timescale dynamics of the two proteins. However, the similar Rex rates measured by the extended model-free approach and the μs-ms timescale dynamics measured by the NMR relaxation dispersion method are present all along the backbone of the protein in addition to the faster μs timescale motions in the RNA-binding region. This conformational plasticity and the μs-ms timescale dynamics in two dsRBDs might be responsible for the adaptation required by the dsRBDs to target conformationally distinct dsRNAs present in the cellular pool.
Author Contributions
J.C. and H.P. conceived the approaches to dynamically characterize the model dsRBDs and wrote the paper. H.P. performed all the experiments and the data analysis in active discussions with J.C.
Declaration of interest
The authors declare no competing interest.
Supplementary Information
Supplementary Information (SI) available.
Methods Details
Protein Expression and Purification
The plasmids containing the gene for TRBP2-dsRBD1 and dADAR-dsRBD1 were kindly provided by Prof. Jennifer Doudna (Noland, Ma and Doudna, 2011; Noland and Doudna, 2013) and Prof. Frederic Allain (Barraud et al., 2012) respectively. The recombinant proteins – TRBP2-dsRBD1 and dADAR-dsRBD1 – were expressed and purified as described elsewhere (Barraud et al., 2012; Paithankar et al., 2018). For 15N labeling, the protein was prepared by growing the cells in the M9 media containing 15NH4Cl (Cambridge Isotope Laboratories) as a sole source of nitrogen.
Circular Dichroism Spectroscopy
All the Circular dichroism (CD) experiments were carried out on Jasco J-815 CD spectropolarimeter using a quartz rectangular cuvette of 2 mm path length. The bandwidth was set at 1 nm and the data was acquired with a scanning speed of 200 nm/min with an averaging time of 1 sec. The protein samples were exchanged with Buffer A (10 mM Sodium phosphate, 100 mM NaCl, 1 mM EDTA and 1 mM DTT) pH 6.4. The CD data was acquired on 20 μM of protein in the far-UV region from 200 to 250 nm in steps of 1 nm measured in temperature range from 10 to 80 °C at steps of 5 °C. An average of 3 scans was taken at each temperature. Protein was equilibrated for at least 10 min at a given temperature before starting the measurement. All CD spectra were baseline corrected by buffer.
The CD data thus obtained were smoothened using a five-point averaging method. The fraction folded (α) at each temperature was calculated as (Greenfield, 2006): where, Sobs, SU, SF are the molar ellipticity at a given temperature, for the completely unfolded state and for the folded state, respectively. The thermodynamic parameters of the unfolding event were calculated by fitting CD data to Gibbs-Helmholtz equations (Greenfield, 2006): where, ΔH describes the enthalpy change, ΔCP the heat capacity change, TM is the temperature at mid-point of the unfolding transition, ΔG is the free energy of unfolding at Temperature T (Kelvin), K is the constant of folding at temperature T, and R is the gas constant (1.98 cal K−1mol−1).
Size Exclusion Chromatography – Multiple Angle Light Scattering
SEC-MALS analysis of the purified proteins was performed on Superdex 75 10/300 GL column (GE Healthcare) using an Agilent HPLC system equipped with the 18-angle light scattering detector (Wyatt Dawn HELIOS II) and a refractive index detector (Wyatt Optilab T-rEX). The system was calibrated with a 100 μl injection volume of 30 μM Bovine Serum Albumin solution (ThermoScientific). 100 μl protein sample was injected (in duplicates) at a concentration of 459 μM and 1.1 mM for TRBP2-dsRBD1 and dADAR-dsRBD1, respectively. The molecular weights of the peaks were determined using Zimm model in ASTRA software version 6.1.7.17 (Wyatt Technologies).
NMR spectroscopy
All the NMR experiments were carried out on 1) Ascend™ Bruker 600 MHz NMR spectrometer equipped with quadruple-resonance (1H/15N/13C/31P) 5 mm Cryoprobe, X, Y, Z-gradients, and dual receiver operating; and 2) Ascend™ Bruker 750 MHz NMR spectrometer equipped with TXI probe, Z-gradient, and deuterium decoupling. All the NMR experiments were recorded at 25 °C. All the NMR spectra were processed using NMRPipe/NMRDraw (Delaglio et al., 1995) and were analyzed using SPARKY (Lee, Tonelli and Markley, 2015). Representative 1H-15N-HSQC spectra for TRBP2-dsRBD1 and dADAR-dsRBD1 have been depicted in Figure S6.
Nuclear spin relaxation experiments were measured at two field strengths (an in-house 600 MHz NMR spectrometer and 750 MHz NMR spectrometer located at IIT Bombay) on a 1.8 mM 15N-TRBP2-dsRBD1 sample and 0.69 mM 15N-dADAR-dsRBD1 sample in Shigemi tubes. For TRBP2-dsRBD1, 15N transverse relaxation rates (R2) were measured with eight CPMG delays of 17, 34*, 51, 68, 85, 102, 136*, and 170 ms. 15N longitudinal relaxation rates (R1) were measured with ten inversion recovery delays of 10, 30, 50*, 100, 200, 300, 450, 600, 750*, and 900 ms. For dADAR-dsRBD1, seven CPMG delays of 17, 34, 68*, 102, 136, 170, and 204* ms and eight inversion recovery delays of 10, 30, 50*, 100, 200, 350, 500*, and 750 ms were used to measure R2 and R1 relaxation rates, respectively. Delays marked with an asterisk in both the experiments were measured in duplicate for estimation of errors in relaxation rates. Steady-state [1H]-15N heteronuclear nOe measurements were carried out with a 1H saturation time of 3 s and a relaxation delay of 2 s for both the proteins. For the experiment without 1H saturation, the relaxation delay of 5 s was used. All the nuclear spin relaxation experiments were measured in an interleaved fashion and with randomized order of delays.
15N relaxation dispersion experiments were measured using constant time CPMG (Carr-Purcell-Meiboom-Gill) experiments (Tollinger et al., 2001) at two static magnetic fields (600 MHz and 750 MHz). For TRBP2-dsRBD1, each relaxation dispersion profile was composed of 9 points with vcpmg values of 25, 50*, 100, 200, 350, 500, 650*, 800, and 1000 Hz at constant relaxation time, Trelax, of 40 ms. For dADAR-dsRBD1, relaxation dispersion profile was recorded with νcpmg of 50, 100, 150*, 200, 300, 450, 600*, 800 and 1000 Hz with Trelax of 20 ms. Duplicate points were measured for error analysis and have been marked with an asterisk.
Heteronuclear Adiabatic Relaxation Dispersion (HARD) experiments were performed at 600 MHz NMR spectrometer. A composite adiabatic pulse of 16 ms containing four hyperbolic secant family pulses of 4 ms each was used for spin-locking the 15N magnetization in both R*1ρ and R2ρ experiments, as described earlier (Mangia et al., 2010; Traaseth et al., 2012). Relaxation dispersion was created with adiabatic hyperbolic secant pulses with different stretching factors (n=1,2,4,6,8). As the stretching factor was increased from 1 to 8, the effective spin-lock field strength increased. Relaxation delays for adiabatic R1ρ and R2ρ experiments were varied by varying the number of composite pulses applied during evolution. The relaxation delays used were 0, 16, 32, 48 and 64 ms corresponding to the number of composite adiabatic pulses of 0, 1, 2, 3, 4. R1 experiments were acquired in the same way as R1ρ and R2ρ experiments without using the adiabatic pulse during evolution. Relaxation delays used for the R1 experiment were 16, 48, 96, 192, 320, and 480 ms. An inter-scan delay of 2.5 s and 3 s was used for TRBP2-dsRBD1 and dADAR-dsRBD1, respectively in all the above relaxation experiments.
NMR relaxation Data Analysis
Relaxation rates in R1/R2/R1ρ/R2ρ experiments were calculated by fitting the intensity data against relaxation delays to mono-exponential decays in Mathematica (Spyracopoulos, 2006). Errors in the relaxation rates were calculated as fit errors using a combination of duplicate delay and Monte Carlo simulations. [1H]-15N nOe values were obtained as a ratio of the intensity of respective peaks of the spectra recorded with and without saturation. Errors in nOe values were obtained by propagating the errors from RMSD values of baseline noise as obtained from Sparky.
Analysis of the 15N-relaxation data (R1, R2, [1H]-15N nOe) recorded at two magnetic fields was done using the extended model-free formalism (Lipari and Szabo, 1982c, 1982a; Clore et al., 1990) with the graphical user interface provided by the Relax v4.0.3 software (d’Auvergne, 2006; Bieri, d’Auvergne and Gooley, 2011). The 1H-15N dipole-dipole interaction and CSA values (−172 ppm) were included in calculations as per the default protocol. The distance between N and H was set as 1.02 Å. First, the local τm diffusion model was optimized with no global diffusion parameter defined. Diffusion tensor values for various diffusion models – sphere, spheroid, and ellipsoid – were calculated using the program quadric diffusion (Lee et al., 1997) for data recorded at both the magnetic fields. Solution structure of TRBP2-dsRBD1 (Paithankar et al., 2018) and dADAR-dsRBD1 (Barraud et al., 2012) was used to optimize the diffusion tensor parameters. These diffusion tensor parameters were then used to calculate the model-free parameters from the ten model-free models and the best model for each residue was selected based on Akaike’s Information Criteria. After the selection of these local models, diffusion tensor was further optimized with fixed local models. This was repeated until all the model-free parameters converged. The global diffusion model was then selected based on the chi-squared values obtained from optimized diffusion models. The error analysis for the selected model (global and local) was done using the Monte Carlo method.
For CPMG relaxation dispersion experiments, effective transverse relaxation rates at each CPMG frequency were extracted using the following equation (Loria, Berlow and Watt, 2008): where, I0 is the intensity of a peak in reference spectrum and Icpmg is the intensity of the respective peak in the spectrum with CPMG field applied at a frequency vcpmg. The R2eff rates thus obtained were plotted against the CPMG frequency vcpmg.
For HARD experiments, the relaxation rates (R1, R1ρ, and R2ρ) along with rotational correlation time calculated from the model-free analysis was used to fit the data to the time-dependent Bloch-McConnell equation assuming a two-site exchange model. The solution to the equation by using a geometric approximation method allowed to extract the dynamic parameters and the relaxation rates during the chemical exchange process as described in detail elsewhere (Chao and Byrd, 2016).
Acknowledgements
Authors thank Prof. Jennifer Doudna (University of California, Berkeley) for the TRBP plasmids and Prof. Frederic H.T. Allain (ETH Zurich) for ADAR-plasmid. Authors also acknowledge High Field NMR facility at IISER-Pune (co-funded by DST-FIST and IISER Pune) and the High-Field NMR facility at IIT Bombay. This work was supported by funding from Indian Institute of Science, Education and Research, Pune; Department of Biotechnology, Govt. of India [No. BT/PR24185/BRB/10/1605/2017]; and extramural funding from the Science and Engineering Research Board (SERB), Govt. of India [EMR/2015/001966]. We acknowledge Dr. Radha Chauhan, Ms. Sangeeta Niranjan, and Ms. Bhawana Burdak for the help with measurement and analysis of the SEC-MALS data at NCCS, Pune. Authors also acknowledge Prof. Gianluigi Veglia (University of Minnesota, Minnesota) and Dr. Fa-An Chao (National Cancer Institute, Maryland) for active discussions while setting up HARD experiments and data analysis. Authors thank Dr. Arnab Mukherjee and Dr. T.S. Mahesh at IISER Pune for helpful discussion as Research Advisory Committee members for HP. Authors also thank Dr. Shilpy Sharma (Savitribai Phule Pune University, Pune) for active discussions during the data analysis and writing of the manuscript. HP is thankful to IISER Pune for the fellowship.