Side chain dynamics of carboxyl and carbonyl groups in the catalytic function of Escherichia coli ribonuclease H

Abstract

Many proteins use Asx and Glx (x = n, p, or u) side chains as key functional groups in enzymatic catalysis and molecular recognition. In this study, NMR spin relaxation experiments and molecular dynamics (MD) simulations are used to measure the dynamics of the side chain amide and carboxyl groups, ¹³C^γ/δ, in Escherichia coli ribonuclease HI (RNase H). Model-free analysis shows that the catalytic residues in RNase H are pre-organized on ps-ns timescales via a network of electrostatic interactions. However, chemical exchange line broadening shows that these residues display significant conformational dynamics on μs – ms timescales upon binding of Mg²⁺ ions. Two groups of catalytic residues exhibit differential linebroadening, implicating distinct reorganizational processes upon binding of metal ions. These results support the “mobile metal ion” hypothesis, which was inferred from structural studies of RNase H.

Understanding protein dynamics is critical for elucidating the molecular mechanisms of many biological processes, including enzyme catalysis and protein - protein interactions.^1–3 Recent developments in nuclear magnetic resonance (NMR) spin relaxation methods have enabled the investigation of biomolecular dynamics in atomic detail.^4–7 Nonetheless, despite the importance of Asx and Glx residues in enzyme catalysis and substrate binding processes, surprisingly few studies on the side chain dynamics of these residues have been reported.^8–10 Here, we present a ¹³C^γ/δ relaxation investigation of the side chain amide and carboxyl groups in Asx/Glx residues of E. coli RNase H, by using recently developed NMR relaxation methods.⁸ To gain more mechanistic insights, we have compared our experimental findings with the results of molecular dynamics (MD) simulations. RNase H cleaves the RNA strand of a RNA/DNA hybrid (Figure 1).¹¹ The active site includes four acidic residues (D10, E48, D70, and D134) that require Mg²⁺ ions for catalytic activity.^12–14 Experimental and simulated side chain generalized order parameters, S², which represents the mobility of the C-C′ bond vector in the side chain amide and carboxyl groups of Asx/Glx residues, including the catalytic residues, provide direct evidence of pre-organization of the catalytic residues on ps – ns timescales. However, these pre-organized catalytic residues show significant conformational dynamics on μs – ms timescales, which might be required for reorganizing the catalytic groups upon sequential binding of metal ions.

Figure 1.

(A) Ribbon representation of RNase H (PDB ID 2RN2). All Asx/Glx residues are shown as stick models and color coded based on the value of S²_MD. Key catalytic residues are labeled. (B) Structure of RNA/DNA hybrid substrate from the human RNase H complex (PDB ID 2KQ9) is superimposed onto E. coli protein (PDB ID 2RN2). The most rigid side chains are shown as space-filling models. DNA and RNA strands are shown in green and cyan, respectively.

Residue-specific values of S² were determined using the model-free formalism applied to longitudinal (R₁) and cross-correlated transverse (η_xy) relaxation rate constants of the ¹³C^γ/δ nuclei.⁸ The chemical shift anisotropy (CSA) tensors for carbonyl and carboxyl ¹³C^γ/δ are highly asymmetric and depend upon the local environment.¹⁵ For carbonyl ¹³C, we used the average backbone carbonyl ¹³C′ CSA values determined for GB1 using solid state NMR spectroscopy: δ_xx = 240.87 ppm, δ_yy = 196.62 ppm, and δ_zz = 93.5 ppm.¹⁵ For carboxyl ¹³C, δ_xx = 242 ppm, δ_yy = 191 ppm, and δ_zz = 105 ppm were assumed.¹⁶ The model-free analysis was also repeated with extreme CSA values of the carbonyl and carboxyl ¹³C′ to test the stability of the results (see Supporting Information, SI). We were able to analyze relaxation rates for 22 of the 34 Asx/Glx residues. Many of the excluded residues from the analysis are well-exposed to the solvent in the structure and showed very slow development of the cross peaks through cross-correlated relaxation; additional details of experiments and the model-free analysis are described in SI.

To complement experimental studies with direct observations of side chain dynamics, we performed MD simulations of E. coli RNase H initiated from the crystal structure (PDB ID 2RN2) protonated in accordance with experimental pK_a measurements¹³ to reflect pH values of 5.5 and 8.0. The protein was described with the Amber99SB force field,¹⁷ solvated in TIP3P water, and neutralized with Cl⁻ ions. Simulations of 100 ns duration were performed using Desmond Academic release 3 or source release 2.4.2.1.¹⁸ Experimental (S²_NMR) and MD-derived (S²_MD) side chain order parameters showed significant correlation (R² = 0.72 and r.m.s.d. = 0.11) (Figure 2A), allowing us to use S²_MD to infer the mechanistic details of dynamics in the ps-ns timescale.

Figure 2.

(A) Comparison of order parameters between the experiment (S²_NMR) and simulation (S²_MD). (B) Correlation plot between S²_MD and crystallographic B-factor for side chains (PDB ID 2RN2). The B-factors are the sum of side chain heavy atoms (C, N, and O). (C) S²_MD vs. relative solvent accessibility (RSA) of the side chain calculated from the crystal structure of RNase H (2RN2). RSA was calculated using GetArea.²⁵ (D) Correlation plot between RSA and ΔpK_a. The ΔpK_a for D148 and D102 were estimated by assuming pK_a = 2.0.¹³ Corresponding graphs with S²_NMR are shown in SI.

Order parameters of Asx/Glx side chains showed a broad spectrum of dynamics and were uncorrelated with backbone N-H bond mobility (see SI). Values of S²_MD correlate reasonably well with crystallographic B-factors of the side chains, despite the differences in the nature and timescale of the motions detected by each parameter (Figure 2B). On the contrary, S²_MD are poorly correlated with relative solvent accessibility (RSA) of the side chains (Figure 2C). Recent analysis of order parameters for K side chain amino groups showed strong correlations with both crystallography B-factors and RSA of side chain amino groups.¹⁹ This difference suggests that the side chains of Asx/Glx interact more strongly with other parts of the protein because of a shorter side chain length relative to K. Notably, perturbations of pK_a, relative to intrinsic values, do not linearly correlate with solvent accessibility of side chains for Asp or Glu (Figure 2D). With the exception of residue D70, pK_a’s are close to intrinsic values for residues with RSA > 50%, but become highly heterogeneous for larger degrees of burial. The unusual perturbations of pKa for D10 and D70 reflect coupling arising from close spatial proximity in the active site.¹³

Overall, MD simulations showed that the side chains of D (S²_MD,ave = 0.87 ± 0.08) and N (S²_MD,ave = 0.85 ± 0.08) residues are the most rigid, followed by E (S²_MD,ave = 0.64 ± 0.21) and Q (S²_MD,ave = 0.46 ± 0.21) residues. S²_MD values were used to obtain non-biased statistics because experimental data did not include S²_NMR values of highly flexible residues. Nonetheless, the same pattern was observed with the experimental S²_NMR values. A previous study on the side chain dynamics of calbindin D_9k showed similar statistics for each amino acid type.⁸ It is possible that this effect is primarily due to side chain length; however, functional group rigidity is not necessary correlated to overall side chain entropy.²⁰ Our data provide additional experimental evidence that the side chain of D is usually more rigid than that of E and support the hypothesis that the higher rigidity of the D side chain favors this residue in enzyme active sites.²¹

Rigid side chains are mainly located in the active site of RNase H, which binds to the RNA strand, and the interface with the DNA strand of the RNA/DNA hybrid substrate (Figure 1B). Experimental (simulation) S² values of ¹³C^γ/δ for D10, E48, D70 and D134 are 0.67 ± 0.08 (0.85 ± 0.01), 0.72 ± 0.11 (0.76 ± 0.01), 0.75 ± 0.09 (0.78 ± 0.03), and 0.64 ± 0.04 (0.81 ± 0.01), respectively. Our NMR experiments were performed at pD = 6.2, which is close to the pK_a of D10 (6.1).¹³ Therefore, experimental S² values of the catalytic residues represent the average mobility of D10 in protonated and deprotonated states. High rigidity of the catalytic residues is also suggested by low crystallographic B-factors determined under conditions where D10 would be depro-tonated (pH > 8.8; PDB ID 2RN2).²² Because we calculated S²_MD values with D10 in the protonated state and other acidic residues in the deprotonated state, we further tested if D10 remains rigid at high pH by running another simulation with deprotonated D10. During the simulation, we observed that deprotonated D10 forms a water-mediated hydrogen bond with D134 and remains highly rigid (S²_MD = 0.847 ± 0.002). Discrepancies in S² values for active site residues may reflect inability of standard molecular mechanics forces fields to model changes in protonation state and not incomplete conformational sampling alone. The crystal structure of Mg²⁺-bound RNaseH (PDB ID, 1RDD) shows relatively minor conformational change of the catalytic residues, relative to the apo state (PDB ID 2RN2)²³(see SI). Therefore, our data indicate that the catalytic residues in RNaseH are pre-organized to minimize entropy penalties upon coordinating Mg²⁺ for catalysis.²⁴

The net free energy change, ΔG, from electrostatic interactions of the acidic residues, D and E, can be calculated from the pK_a shift relative to intrinsic pK_a (Figure 3A). The free energy is the sum of three terms: ΔG = ΔH_bind − T(ΔS_sol + ΔS_conf), in which ΔH_bind is the enthalpy change, and ΔS_sol and ΔS_conf are the entropy changes associated with solvent and protein conformational changes, respectively. The contribution of −TΔS_conf to the free energy can be estimated from ΔS².^26,27 Figure 3B presents – TΔS_conf, which was calculated as shown by Yang and Kay.²⁶ The horizontal distance between each data point and the calculated −TΔS_conf line is an estimate of ΔH_bind − TΔS_sol provided by salt bridge formation (Figure 3B). This plot shows that the magnitudes of the local electrostatic interactions are highly heterogeneous.^28,29 One group of residues has −0.5 kcal/mol ≤ ΔG ≤ 0 kcal/mol and the entropic penalty −TΔS_conf nearly offsets the favorable contributions from ΔH_bind − TΔS_sol. A second set of residues has ΔG ≤ − 1 kcal/mol with large favorable contributions from ΔH_bind − TΔS_sol that offset the entropic penalty, despite the high rigidity of side chains in this group. Residue D10 is the only residue in the third group, having a positive value of ΔG, again reflecting interactions with D70 in the active site. These data indicate that the measurement of the side chain entropy of acidic residues can be critical for understanding the role of electrostatic interactions in biological processes. Although not considered in this analysis, −TΔS_conf of the interacting partner molecule should also be determined for better estimation of ΔH_bind − TΔS_sol. We tested whether the differential burial of charged residues generates the heterogeneity in ΔH_bind − TΔS_sol. Interestingly, solvent accessibility of salt bridges only weakly correlates with ΔH_bind − TΔS_sol (R² = 0.12) (Figure 3B), indicating that the geometry of the salt bridge may be more critical for its stability.²⁹ Similar approaches using backbone and methyl-bearing side chains highlighted the role of conformational entropy in ligand binding.^6,26,27,30 Because individual free energy changes upon salt bridge formation can be directly measured using ΔpK_a, acidic side chain order parameters enable the site-specific analysis of energetic contribution of conformational dynamics. This is an additional benefit compared to the analysis using order parameters for backbone or methyl-bearing side chains.

Figure 3.

(A) Plot of S²_MD vs. ΔG associated with the changes in pK_a of each acidic residue. Solid line represents −TΔS_conf.²⁶ For calculation of −TΔS_conf, the order parameter was assumed to change from 0.05 to each S²_MD value (i.e. ΔS²_MD = S²_MD – 0.05). (B) Calculated ΔH_bind − TΔS_sol (bars) and relative solvent accessibility (closed circles) for each acidic residue. Uncertainties for the reported pKa values are not available so the calculated ΔH_bind − TΔS_sol do not include uncertainties. Plots with the experimental order parameters are shown in SI.

The ¹³C^γ/δ order parameters become extremely heterogeneous as side chains becomes more exposed to the solvent (RSA > 35%). Although buried residues (RSA < 20 %) tend to have high S² values, many partially or fully solvent-exposed residues also have high S² values (Figure 2D). For example, E57 and R106 form a solvent-exposed salt bridge, which results in high rigidity for the side chains of both residues. By contrast, D102 and R46 form a completely buried salt bridge and also showed high rigidity for both interacting side chains. This indicates that solvent-exposed salt-bridges and hydrogen bonds can restrict the motion of the C-C′ bond as effectively as completely buried salt-bridges. The partially solvent-exposed D134 (RSA = 44.5 %) forms a salt-bridge with R138, resulting in high order parameters for both residues. The D134-R138 salt bridge is strong enough to offset 1.13 kcal mol⁻¹ entropic penalty from −TΔS_conf, giving ΔG ≈ 0. When the enzyme-substrate (ES) complex is formed, this salt bridge breaks and D134 coordinates a metal ion for catalysis and R138 interacts with the phosphodiester bond of the RNA/DNA substrate to stabilize the ES complex. Therefore, the entropic cost of forming the ES complex may be minimized by preordering these two side chains. This example illustrates the insights provided by side chain order parameters into the molecular origins of the specificity and binding affinity of proteins.

The RNase H family is found in nearly all organisms, from viruses to human. We investigated if side chains with high S² values are evolutionary conserved (see SI). Interestingly, the most conserved Asx/Glx residues are D10, N16, N45, E48, D70, N130, E131, and D134, all of which have high S² values (> 0.8). The rigidity of N45 and N130 is particularly notable because these residues interact with the substrate in the H. sapiens complex and are located within 4Å of the active-site catalytic residues. However, not all residues with high values of S² are well conserved. For example, E6 forms a partially buried salt bridge with R27 in E. coli RNase H. Replacement of E6 by a hydrophobic residue, V in human RNase H, is accompanied by replacement of R27 by a hydrophobic residue, also V in human RNase H (see SI). Diverse functional roles of buried salt bridges have been described in other contexts.^31–33 Waldburger et al. reported that replacement of a salt bridge network by hydrophobic residues increases the stability of the Arc repressor, but changes binding cooperativity.³² RNase H exists as a separate entity in E. coli, whereas it is part of multi-domain protein in many other organisms including human. Thus, dissection of the role of buried electrostatic interactions should be considered in the context of full-length proteins, and not just isolated domains.

Despite many structural and biochemical studies, the details of Mg²⁺ binding and coordination in the active site of RNase H are not completely understood.^12,34,35 Figure 4A shows the effects on ¹³C^γ/δ line widths from chemical exchange on the μs – ms time-scale associated with Mg²⁺ binding processes determined from R_1ρ data acquired with and without Mg²⁺. Catalytic residuesin the active site, except E48, showed significant R_ex = R_1ρ^Mg − R_1ρ^apo. To determine if the observed ¹³C^γ/δ dynamics represent the same motional process, we plotted R_ex versus squared chemical shift differences, Δω², between apo and Mg²⁺-bound RNase H. Assuming a fast exchange process, residues with non-zero R_ex are expected to belong to a single correlation line between R_ex and Δω² if the residues experience the same conformational exchange process. The pK_a of some residues are changed upon binding of Mg²⁺; however, the pKa shifts do not affect R_ex (see SI). Interestingly, we found two distinct groups of residues showing differential dynamics in response to Mg²⁺ binding (Figure 4B and 4C). Group-1 consists of N44, E57, and D102, and group-2 consists of D10, E48, D70 and D134. E48 has large uncertainty because of low signal-to-noise ratio in Mg²⁺-bound state. We included D10 in group-2 because R_ex was too large to be measured and this residue has the largest chemical shift difference (Δω² = 45.0 × 10⁵ s⁻²) between apo and Mg-bound RNase H, consistent with the proposed role of D10 as a major metal ion-coordinating residue.

Figure 4.

(A) Contribution of μs − ms timescale dynamics, R_ex (= R_1ρ^Mg − R_1ρ^apo, upon binding of Mg²⁺ to the active site. The D10 peak was too broadened to measure. The spin-lock RF amplitude was ω_SL/2π = 1.1 kHz (see SI). (B) Correlation plot between R_ex and the squared chemical shift differences, Δω², between apo and Mg²⁺-bound states. The residues belong to group-1 and -2 are labeled in blue and red, respectively. (C) Ribbon representation of the active site of RNase H (PDB ID 1RDD). Group-1 and -2 residues are color coded according to panel B. Only active site residues are shown for clarity. The Mg²⁺ ion is shown as a cyan sphere.

RNase H has been proposed to coordinate two Mg²⁺ ions for catalysis,^14,34 although only a single Mg²⁺ ion was found in the active site in the absence of substrate.²³ According to the substrate-bound structure of human RNase H (the two Mg²⁺ ion model), the A-site consists of D10 and D134, while the B-site consists of D10, D70, and E48.¹⁴ Proper binding of the substrate is essential for promoting the coordination of two Mg²⁺ ions in the active site.¹⁴ Moreover, the position of metal ion in the active site may be is highly heterogeneous (see SI) and mobile depending on the catalytic step.^12,14,35 The difference in μs - ms timescale dynamics between group-1 and -2 may indicate that the binding of the first Mg²⁺ can trigger conformational dynamics of the residues required for coordinating the second Mg²⁺ ion and substrate. Complex interplay among active site residues upon binding of metal ions was also proposed based on structural analysis of Mn²⁺-bound RNase H.¹² However, our current data do not distinguish whether the observed μ – ms timescale dynamics of the group-1 residues is induced by the binding of the first Mg²⁺ ion to the group-2, or whether metal ion binding changes the chemical environment in a way that hidden, but already present, dynamics become observable. In summary, although the catalytic residues of RNase H are highly rigid in the ps – ns timescale, they undergo significant conformational dynamics in the μs – ms timescale upon metal ion binding, which may reflect reorganization of the active site.^24,36