Understanding the effect of magnesium ion concentration on the catalytic activity of ribonuclease H through computation: Does a third metal binding site modulate endonuclease activity?

Abstract

Ribonuclease H (RNase H) belongs to the nucleotidyl-transferase (NT) superfamily and hydrolyzes the phosphodiester linkage on the RNA strand of a DNA/RNA hybrid duplex. Due to its activity in HIV reverse transcription, it represents a promising target for anti-HIV drug design. While crystallographic data have located two ions in the catalytic site, there is ongoing debate concerning just how many metal ions bound at the active site are optimal for catalysis. Indeed, experiments have shown a dependency of the catalytic activity on the Mg²⁺ concentration. Moreover, in RNase H the glutamate residue E188 has been shown to be essential for full enzymatic activation regardless of the Mg²⁺ concentration. The catalytic center is known to contain two Mg²⁺ ions (Nowotny et al.) and E188 is not one of the primary metal ligands. Herein, classical molecular dynamics (MD) simulations are employed to study the metal-ligand coordination in RNase H at different concentration of Mg²⁺. Importantly, the presence of a third Mg²⁺ ion, bound to the second-shell ligand E188, is persistent feature of the MD simulations. Free energy calculations have identified two distinct conformations depending on the concentration of Mg²⁺. At standard concentration, a third Mg²⁺ is found in the catalytic pocket but it does not perturb the optimal RNase H active conformation. However, at higher concentration, the third Mg²⁺ ion heavily perturbs the nucleophilic water and thereby influences the catalytic efficiency of RNase H. In addition, the E188A mutant shows no ability to engage additional Mg²⁺ ions nearby the catalytic pocket. This finding likely explains the decrease in catalytic activity of E188A, and also supports the key role of E188 in localizing the third Mg²⁺ ion at the active site. Glutamate residues are commonly found surrounding the metal center in the endonuclease family, which suggests that this structural motif may be an important feature to enhance catalytic activity. The present MD calculations support the hypothesis that the RNase H can accommodate three divalent metal ions in its catalytic pocket, and provide an in-depth understanding of their dynamic role for catalysis.

INTRODUCTION

Ribonuclease (RNase) enzymes belong to the nucleotidyl-transferase (NT) superfamily, which catalyzes the hydrolysis of phosphodiester linkages that form the backbone of RNA strands. Various types of RNase have been identified and grouped into different categories based on the structures and the targeting substrates¹. Here we focus on the RNase H, one of the most important endoribonucleases. The structure and function of RNase H have been extensively studied and characterized through crystallographic and kinetic data^2–11. Specifically, RNase H is a non-sequence specific enzyme that cleaves the RNA strand of a DNA/RNA hybrid duplex. It has been found in many species, including viruses, bacteria and humans^4,12,13, and found to be an essential enzyme in some of them. For example, RNase H knockout mice die in embryo because of the malfunction of mitochondrial DNA synthesis¹². Malfunction of enzymes from the RNase superfamily could also result in the formation of tumors¹⁴. Moreover, RNase H is essential in HIV reverse transcriptase (RT)¹⁵, and therefore is a promising target for anti-HIV drug design^16–18. Thus, the development of ribonuclease inhibitors has drawn much attention and is an alternative strategy for anti-HIV therapies. To date, several compounds have been synthesized and showed promising results in drug discovery, such as the 2-hydroxyisoquinoline-1,3(2H,4H)dione and its derivatives^18–20.

RNase H is a metalloenzyme that consists of a novel α/β fold^3,21, with conserved carboxylate groups (i.e. DDE motif^22,23) in the catalytic center (Figure 1). This DDE motif is essential for metal binding and catalysis. In fact, the activity of RNase H is optimal when divalent metal ions are present^5,24, such as Mg²⁺ or Mn²⁺. Stereochemical^25,26 and computational²⁷ studies have shown that the reaction mechanism occurs via an in-line S_N2-like nucleophilic attack on the scissile phosphorus atom by a nucleophilic group, which is likely one hydroxyl ion in the final form, and produce 3′-hydroxyl and 5′-phosphate terminated products (Figure 2). Therefore, the function of the bound metal ions is both to activate the nucleophilic group by lowering its pK_a value, and to stabilize the formation of the transition state structure^28,29.

Figure 1. Crystallographic structure of Bacillus halodurans RNase H.

(A) Cartoon of the complex of RNase H and ds-RNA/DNA hybrid duplex. DNA and RNA are drawn in red and blue ribbons, respectively; orange spheres indicate the Mg²⁺ ions. (B) Close-up of the catalytic site, including the RNA strand, metal ligands and water molecules coordinated to the two crystallographic Mg²⁺ ions. The key residue E188 should be noted. The D192N inactive mutant is shown (PDB code: 1ZBL³).

Figure 2. Catalytic mechanism of RNase H as proposed by QM/MM calculations²⁷.

R: reactants state, the nucleophilic water (Wat_N) deprotonated in situ is represented in red, and the water H-bond network promoting the proton transfer to the phosphate group is shown in blue. TS₁-INT-TS₂: composite transition states and high-energy intermediate as calculated in previous studies.²⁷ P: product state of the cleaved RNA strand and its release from the metal site.

While divalent metal ions have been demonstrated to be essential for activity of the large family of riboendonucleases, their function and the exact number of bound ions in the active site^30–32 are still under debate. For example, Katayanagi et al. have proposed a single metal ion model according to the X-ray crystallographic analysis of E. coli RNase HI³³. In this model, the metal ion is bound to the RNA phosphodiester oxygen atom and three carboxyl groups. On the other hand, Smith and Pace³⁴ have performed kinetic analyses of RNase P and suggested that at least three metal ions are needed for optimal activity and are bound to the RNA phosphodiester group. Recently, Nowotny et al. have resolved a number of high-resolution crystal structures of RNase H from Bacillus halodurans (Bh) where two ions are situated in the catalytic pocket in complex with the substrate^3,4, thus suggesting that a bimetal site is responsible for the catalytic activity. In particular, two Mg²⁺ ions are jointly coordinated in the active site by one non-bridging oxygen atom of the scissile phosphate group of the RNA strand, four carboxylates (D71, E109, D132 and D192) and water molecules (Figure 1B). This configuration is in striking similarity to the two-metal-ion mechanism proposed by Steitz and Steitz²⁹, which suggests that one of the two metal ions activates the attacking nucleophile (a water molecule in RNase H) and the other stabilizes the formation of the transition state. This mechanism has also been suggested for other enzymes, such as Holliday junction resolvase, retroviral integrase, transposase, and RISC nuclease Argonaute^21,35–37. For the Bh RNase H enzyme, we have recently investigated the enzymatic mechanism by means of DFT-based quantum mechanics/molecular mechanics simulations²⁷ (Figure 2). Briefly, we showed how either one water molecule or one hydroxide ion can act as nucleophile, with free energy barriers of activation in good agreement with experimental data. In the case of a water molecule acting as a nucleophilic agent, water-mediated proton shuttles are involved in the water deprotonation. Importantly, we have observed the stabilization of a phosphorane intermediate during one of the possible reaction mechanisms (Figure 2). Also, the catalytic role of the two Mg²⁺ ions has been shown as critical in stabilizing the transition state.

Interestingly, it has been shown that the enzymatic activity of RNase H depends on the metal ionic concentration. In fact, experiments have demonstrated that high metal concentration is actually causing the inhibition of the activity of Bh RNase H: the activity is optimal at Mg²⁺ concentration of few mM, while it is inhibited at 50 mM concentration in the gel activity assay³. A detailed activity study has been done for human and mouse RNase H1 up to 80 mM Mg²⁺ concentration⁶. It showed that the highest activity is found at 20 mM and eventually decreased at higher concentration. In HIV RNase H, the optimal concentration is 8 mM, which is lower than that of other species³⁸. These results have suggested an attenuation effect ^39,40, where an additional ion can influence the catalytic activity, binding to the residues nearby the active site, usually acidic amino acids, and eventually impairing the catalytic reaction^39,41. For example, Marqusee et al.³⁹ suggested that an additional divalent metal ion could bind a carboxylate group (D134) close to the active site at high ionic concentration in E. coli RNase H. Consequently, a nearby histidine, which plays a role as proton shuttle during the catalysis, is neutralized thus losing its ability to act as proton transporter. In Bh RNase H, the same inhibitory effect was observed in the gel filtration binding assay experiment where the activity was reduced with concentrations up to 50 mM Mg²⁺. This unique attenuation character is not only found in RNase H, but also in other metalloenzymes. For instance, one crystal structure of binuclear zinc cluster of LpxC reveals that a second zinc ion plays a role as an inhibitor, diminishing the catalytic activity by engaging the side chains of E78 and H265, which are a general base and an electrostatic stabilizer in the enzymatic reaction⁴². In this scenario, while it is clear that divalent metal ions are essential for the catalytic function of ribonuclease enzymes, it is not clear yet how many ions are optimal for catalysis³¹.

In the case of Bh RNase H, Nowotny et al.³ have proposed that the E188, a glutamate residue located next to the active site, could recruit a third metal ion, which would help the catalytic activity through its structural and electrostatic role. In fact, E188 has shown high conformational flexibility in X-ray structures, being able to swing into and out from the active site. Also, when E188 is substituted with an alanine amino acid, RNase H has a reduced activity, which is however retained at high metal concentration³.

Here, we elucidate the nature of the catalytic metal binding site of RNase H and the role of E188 in response to different buffer concentrations. Towards this aim, we have performed a series of classical molecular dynamic (MD) simulations to study the metal-ligand coordination in Bh RNase H at different concentration of Mg²⁺. Importantly, we invariantly observe the presence of a third Mg²⁺ ion weakly bound in the proximity of the catalytic center, chelated by the second-shell metal ligand E188. We have characterized the free energy landscape associated with conformational switches of E188 and its coordination to a third Mg²⁺. Indeed, this conserved residue is able to bind the additional Mg²⁺ metal ion at all the considered ionic concentrations. Remarkably, the third Mg²⁺ ion does not perturb the enzymatic reactive state at standard conditions and likely tunes electrostatics for optimal catalysis. By chelating the third incoming ion, E188 prevents in fact a perturbation of the reactive position of the nucleophilic water molecule in the reactants state. Only when Mg²⁺ concentration increases the third Mg²⁺ ion manages to displace the nucleophilic water, likely affecting the catalytic efficiency of RNase H, as seen in experiments at high Mg²⁺ concentration.³ The adapting role of E188 is confirmed by the MD simulations of the E188A mutant, which agrees with the proposed attenuation hypothesis. As a result, the present computations not only support the hypothesis that the third Mg²⁺ ion is an important element in the catalytic pocket but also explain its dynamic role coupled to E188 for modulation of RNase H activity at different ionic concentrations.

RESULTS AND DISCUSSION

Conformation and dynamics of the active pocket at standard MgCl₂ concentration

A simulation at 25 mM MgCl₂ concentration was run to investigate the structural and dynamic features of wild-type RNase H at near-standard/optimal conditions for catalysis. The root mean square displacement (RMSD) value of the protein heavy atoms is 1.5 ± 0.2 Å compared to the reference crystal structure, which highlights the great stability of the enzyme framework. The DNA/RNA hybrid remained stable as well showing an average C₁′-C₁′ distance of 10.5 ± 0.3Å (Figures 1 and S1 in Supporting Information). The coordination shells of the two Mg²⁺ ions are well preserved during the simulation (Figures 1 and 3A). In detail, MgB coordinates to D71, E109, D132, the bridging (O₃) and non-bridging oxygen (O₁) atoms of the scissile phosphate of the RNA strand. The average distance of metal-carboxylate group is 1.90 ± 0.05 Å. The metal-RNA ligand distance is ~2.02 ± 0.08 Å and ~2.21 ± 0.14 Å for Mg-O₁ and Mg-O₃, respectively. The calculated binding distances are similar to the values observed in crystallographic magnesium-carboxylate complexes⁴³ and QM/MM calculations²⁷. The coordination geometry slightly deviates from a perfect octahedral configuration. In addition, the MgA coordinates to D71, D192, O₁ oxygen atom and three water molecules. Specifically, the average distance of metal-carboxylate group is 1.90 ± 0.04 Å, and is 1.98 ± 0.07 Å for Mg-O₁. These structural characteristics are similar to the MgB metal-ligand coordination. One of the water molecules behaves as a bridge, connecting MgA and D132. Another coordinated water molecule acts as the nucleophilic group (Wat_N) and hydrogen bonds to the nearby pro-Rp oxygen atom forming an in-line structure with the phosphate group (r_O-P ~ 3.06 ± 0.09 Å and ∠_O-P-O ~ 167 ± 5°) (Figures 1B and 2). Overall, the coordination of the metal-ligands for MgA is an ideal octahedron.

Figure 3. Active and inactive states as modulated by MgCl₂ concentration.

Two representative snapshots taken from the MD simulations represent the active state (A) and the inactive state (B) dominant, respectively, at low (25 mM) and high (500 mM) Mg²⁺ concentration. (A) Active state: the carboxylate group of E188 points out of the active site and Wat_N binds to MgA. (B) Inactive state: Wat_N coordinates to MgC and slightly loses contact to MgA, while the carboxylate group of E188 flips inward and points to the phosphodiester group. Metal-ligand coordination is indicated with dashed lines.

These structural determinants match well to the X-ray structures³ and previously reported MD and QM/MM calculations²⁷ on the reaction mechanism. The formation of a hydrogen bond with the pro-R_p oxygen atom and coordination with the metal ion are likely responsible for the pK_a reduction of the attacking water molecule and its increased nucleophilicity. The role of the metal ions and the nature of the nucleophile were investigated through DFT QM/MM simulations²⁷. Briefly, the two bound metal ions act cooperatively to facilitating nucleophile formation and stabilize both the transition state and leaving group. The nucleophile formation can be achieved in situ, after migration of one proton from the attacking water to the scissile phosphate in the transition state. This proton shuttle is mediated by surrounding solvation water molecules, which promotes the formation of a meta-stable phosphorane intermediate along the reaction²⁷ (Figure 2).

MgCl₂ concentration does not affect the conformation of the protein and ds-DNA/RNA hybrid duplex

A set of MD simulations were run using different MgCl₂ concentrations to test the response of the RNase H protein and the substrate DNA/RNA hybrid to different buffer conditions (from 25 mM to 500 mM). The relevant structural determinants of the active site are preserved regardless of counter-ion content (Table S2 in SI). Moreover, the RMSDs value of the protein heavy atoms in all the systems are similar to those in the X-ray structure (1.5 ± 0.1 Å, Figure S2 in SI), and the structure of the ds-DNA/RNA substrate bound to the enzyme does not change significantly (10.6 ± 0.4 Å, measuring the length of C1′ in each base pair, Figure S1 in SI). This indicates that the concentration of MgCl₂ does not affect the global conformation of the protein or the ds-DNA/RNA substrate, thus suggesting that the enzyme inhibition at high concentration of Mg²⁺ does not result from major conformational changes of the protein/substrate complex.

A third Mg²⁺ binding site is adjacent to the active pocket and E188

At standard MgCl₂ concentration, a third Mg²⁺ metal ion (MgC) is invariantly observed located in the close proximity of the active site (Figure 3). This metal ion neither replaces the original bound Mg²⁺ metal ions (i.e. MgA and MgB), nor significantly changes the coordination shell of the metal-ligands during the MD simulations. Interestingly, MgC binds to an area characterized by a high negative electrostatic potential: in particular, to E188, D192 and four water molecules, forming an octahedral coordination shell when concentration is 25 mM (Figure 3A). This third metal binding site is likely weaker than those for MgA and MgB due to the involvement of only two protein residues and to a larger solvent-accessible area. This can explain why it was never observed in any X-ray structure so far. In fact, B-factors for MgC estimated from MD trajectories at near-standard conditions and higher Mg²⁺ concentrations (see below) consistently indicate a much larger fluctuation (up to 30-fold, Table S4 in SI) compared to MgA and MgB. In general MgC shows high B-factor values (up to ~55 Ų, Table S4 in SI), which confirm its higher fluctuation.

E188 deserves special attention due to its potential effect on the RNase H activity. In fact, from the crystallographic data^3,5 E188 was found to have great conformational flexibility. Two E188 conformations were found in different crystal structures of mutants (D192N and D132N) under similar conditions. In D132N, the carboxylate group of E188 points toward the active site and is 3.9 Å away from MgA metal, and 5.5 Å from the phosphodiester group (PDB code: 1ZBI). In D192N, on the other hand, it swings out from the active site to distances of 7.4 Å and 9.8 Å away from the MgA and phosphodiester group, respectively (Figure 1). In our MD simulation of the wild-type conformation at 25 mM MgCl₂ concentration, the carboxylate group of E188 stabilizes in a conformation resembling the D192N mutant, which points away from the active site (Figure 3A). Stabilization is promoted by the presence of MgC: in fact in a MD simulation where additional bulk Mg²⁺ ions are absent (see SI), E188 keeps swinging into and away several times showing no preferential conformations in the ns timescale (Figures S3 in SI). This character might be important for the second-shell glutamate to bind divalent metals and to release the product after the chemical step⁵. Overall, this further indicates that the sidechain of E188 is very flexible, and the most stable conformation of E188 is induced by the electrostatic character of the surrounding environment.

The possible existence of three Mg²⁺ ion binding sites in the active pocket of endonucleases is also supported by several kinetic studies and crystallographic data^34,44,45. Pace et al. suggested that the third metal ion binding site is located around the scissile phosphate group, and this metal ion directly participate in the catalysis in their model of RNase P. However, the MgC binds to a second shell residue in our simulations and we did not observe a direct contact between the reactive species and the MgC. As a result, the function of MgC in RNase H seems not to be identical to that proposed in RNase P³⁴ and needs to be further investigated. Horton et al. suggested that the third Mg²⁺ ion would bind to Glu45 (second shell) and Aps74 (first shell) in EcoRV, and has mostly a structural role^44,45. Finally, Nowotny et al., based on the mutagenesis studies of RNase H complex³, suggested a possible binding site around E188 due to the strong electrostatic character of the carboxylate group. However, they were unable to detect such metal site with X-ray crystallography at any Mg²⁺ concentration, which could be explained by the high B-factors of the E188 carboxylate group. Thus, the present MD result supports the hypothesis of Nowotny et al., by showing the presence of a third metal ion bound to E188. Interestingly, a recent crystal structure of RNase III/ds-RNA complex presented by Ji et al. shows several Mg²⁺ ions around the bi-metal active site, which do not participate in the catalytic reaction. Remarkably, one of them is in a very similar position to MgC in our simulations, being bound to first shell and second shell glutamate residues⁴⁶ (see discussion below and Figure 7D).

Figure 7. Active sites of other members of the NT superfamily showing a second-shell glutamate residue.

(A) The active conformation from the present work on Bh RNase H; (B) E. coli RNase H X-ray structure (pdb code: 1g15); (C) prokaryotic DNA transposase (pdb code: 1mus); (D) Aa-RNase III/ds-RNA complex (pdb code: 2nug). HIV reverse transcriptase (pdb code: 1suq) also shows this motif (structure not shown). Magnesium ions are represented as orange spheres, manganese ions as green spheres. Divalent ions that superimpose to MgA, MgB, MgC in the present work are labeled specifically.

MgCl₂ concentration modulates the position of the third Mg²⁺ ion and the catalytic site architecture

The presence of MgC in the position adjacent to the catalytic site is confirmed by MD simulations in a broad range of Mg²⁺ concentrations, namely from 25 mM to 500 mM of MgCl₂. There are several metal ions bound to the phosphodiester groups of the ds-DNA/RNA strands in these systems. However, no additional Mg²⁺ ions are found within a sphere of ~12 Å radius around the active site at any MgCl₂ concentration (Figures 4A), apart from the MgA, MgB and MgC. Thus, regardless the MgCl₂ concentration, the third magnesium ion seems to have a preferential electrostatic binding pocket close to the active site, involving first (D192) and second (E188) shell metal ligands.

Figure 4. Mg²⁺ distribution at the active site of RNase H.

The number of coordinated Mg²⁺ ions (NC) around the C5 phosphodiester P atom is shown as extracted from the MD simulations of wild-type (A) and E188A mutated system (B) at different MgCl₂ concentration. NC(r) is derived from the integration of the P-Mg²⁺ radial distribution function. The localization of MgC in the active and inactive state is shown by the blue and red areas, respectively (Figure 3).

Importantly, two preferential binding modes are observed for MgC, which are regulated by Mg²⁺ concentration and determine the position of the flexible residue E188 (Figure 3). One stable conformation is that found at 25 mM, where E188 binds to the third ion, which however does not perturb the catalytic site, and likely leads to the optimal catalytic activity²⁷. This state is hereafter referred as active state, meaning that the nucleophilic Wat_N strongly binds to MgA (2.1 Å, Table S2 in SI) and represents a reactive Michaelis complex (Figure 3A). In the second conformation E188 is no longer able to protect the optimal catalytic structure, and the third ion disrupts the nucleophilic water orientation: this conformation is hereafter referred as inactive state, meaning that the nucleophilic group binds stronger to MgC than MgA (Figure 3B, Table S2 in SI).

The long solvent-exposed sidechain of E188 is able to chelate the third Mg²⁺ ion at normal concentration (i.e., 25 mM) and to maintain this third metal at a distance from the catalytic site, which is optimal for the catalytic mechanism. We propose that the presence of MgC can be important for the correct modulation of the catalytic reaction. At higher concentrations (e.g., 250, 500 mM), the third Mg²⁺ and its coordination shell is pushed closer to the catalytic site due to the higher electrostatic repulsion of the buffer (Figures 3B, 4A, and S7, Table S2 in SI). In this inactive conformation, MgC and MgA are closer to each other at a distance of ~4.5 Å. As a consequence, to compensate the strong charge-charge repulsion, MgA shifts slightly and moves toward MgB, producing frustration at the active site (MgA-MgB distance is ~3.4 Å, Table 2 in SI). The distance separating MgC and P is ~ 5.4 Å.

Most importantly, the MgA metal ion loses the coordination with the nucleophilic water (Wat_N) and forms a distorted octahedral geometry (Figure 3B). Wat_N moves away and coordinates to MgC while still forms hydrogen bond with pro-R_p oxygen. The distance to P becomes longer (~3.5 Å) and the value of the angle of Wat_N-P-O₃ becomes smaller (~153.0°). Moreover, the position of MgC does not allow for the insertion of an additional water molecule, which would repair the coordination shell of MgA, and act as a reactive nucleophile. All this leads to an unfavorable conformation of the nucleophilic group, which is distorted compared to the active reactant state (Figure 3B).

While the inactive conformation remains stable at high concentration (250 and 500 mM), it is also observed at lower concentrations. However, it is only transient at lower MgCl₂ concentration (from 25 mM to 100 mM), likely biased by the initial starting conformation, and structural changes rapidly occurring after few ns of MD lead to an equilibrated conformation in which the third Mg²⁺ ion does not significantly perturb the reactants state, as found in previous MD simulations²⁷ (Figure 3A).^**

The orientation of E188 is correlated to the position of MgC. In the active state, E188 carboxylate group mostly points away from the active site (Figure 3A). The dihedral angle (C_α-C_β-C_γ-C_δ) of the sidechain on E188 reflects the two states explored by free MD. The angle is ~ −90° and ~ −180° in the inactive state and in the active state, respectively (Figures 3 and Figure S3 in SI). It is worth mentioning that the dihedral angle of the active conformation for E188 is identical to the one in the D192N X-ray structure.

Overall, two conformations of the active site are observed, which are clearly related to position of MgC and its ligands E188 and D192. The MgB site retains its coordination geometry in both states, indicating that its coordination shell is very stable and is not affected by the appearance of an extra metal ion. On the other hand, the coordination geometry of the MgA site is correlated to the binding position of MgC and it can be largely affected along with the position of the nucleophilic Wat_N. From the conformational space sampled through free MD simulations, we can confidently state that at normal concentration of MgCl₂, the inactive state is not stable. However, at higher concentration of MgCl₂ the inactive state becomes more and more predominant, likely producing a significant perturbation of the optimal catalytic site architecture.

How MgCl₂ concentration and third Mg²⁺ position affect the catalytic mechanism

The MD simulations reveal two distinct conformations and yet it is not easy to determine the most stable structure without a very long MD trajectory. Hence, we performed free energy calculations using the ABF method⁴⁷ to study how the free energy surface is affected by MgCl₂ concentration.

A first reaction coordinate (RC₁) is chosen as the distance separating the P and MgC (Figure 5A). In other words, RC₁ describes the free energy surface (FES) related to the vicinity of MgC on the reactive center. Free energy profiles for systems at low and high concentration (i.e., 25 mM and 500 mM, respectively) are shown in Figure 5A. The minimum of the FES at 25 mM is at RC₁=7.1 Å, which shows identical structural determinants as observed in the active state at 25 mM of free MD runs. Another more shallow minimum is located at 5.2 Å < RC₁ < 5.7 Å, which is consistent with the position of the inactive state. Therefore, the ABF calculation clearly describes the two-state conformation behavior of MgC as found in the free MD simulations. The free energy of the inactive state is 5 kcal/mol higher than the active state, which implies an occupation ratio of 1:4400 for these conformations at 300 K. This also points to a large instability for the inactive state at low Mg²⁺ concentration. Importantly, at standard conditions MgC has a preferential pocket at the active site that likely serves to optimally tune the electrostatics of the reactants state. The FES shows a different character at 500 mM Mg²⁺ concentration. The global minimum is located at RC₁ = 6.5 Å, which is 0.6 Å closer to the active state than at low Mg²⁺ concentration. It is also broader, spanning from 6.0 Å to 7.0 Å, which could result from the loosely binding configuration of MgC with only one ligand (E188) at 500 mM Mg²⁺. A second minimum is found at 5.7 Å, which corresponds to the configuration of the inactive state. The free energy difference between these two conformations is less than 1.0 kcal/mol at high Mg²⁺ concentration, which indicates that the probability of finding these two conformations are almost equal at 300 K. As a result, the inactive state is more stable at higher Mg²⁺ concentration, pointing to a crucial role of the third metal for the inhibition of the chemical step as shown by experiments³.

Figure 5. Free energy landscape of the active site architecture at low and high MgCl₂ concentration.

The free energy surface at 25 mM (red) and 500 mM (blue) of MgCl₂ as a function of the MgC to phosphodiester group phosphorus atom (P) separation, RC₁ (A); the MgA-Wat_N separation, RC₂ (B); and E188 to phosphodiester group distance, RC₃ (C). The reaction coordinate RC₃ is chosen as the distance that separates the carboxylate group (C_δ) of E188 and the phosphorus atom (P). Insets represent graphically the reaction coordinates adopted in ABF calculations.

As shown above, the inactive state leads to an unfavorable conformation of the nucleophilic group, which is distorted compared to the active reactant state (Figure 3B). This suggests that the nucleophilic propensity of the Wat_N would likely be decreased, and explain the diminished catalytic activity at high concentration. Free energy calculations exploring the MgA-Wat_N reaction coordinate (RC₂) indicate, in fact, that Wat_N steadily binds to MgA in the active state (2.2 Å, Figure 5B). On the other hand, Wat_N prefers to bind to MgC in the inactive state at 500 mM (3.5 Å), and it takes ~6.7 kcal/mol to bring Wat_N closer to MgA (i.e. at 2.2 Å) in an optimal position for the nucleophilic attack as observed in previous QM/MM studies²⁷ (Figure 5B).

Moreover, E188 is playing a crucial role in regulating this mechanism: it chelates and screens MgC, modulating its binding position at different Mg²⁺ concentrations thanks to the flexibility of its sidechain (Figure 3). For example, the average distance between the E188 carboxylate group and P is 5.6 ± 0.2 Å in the inactive state, and becomes 8.5 ± 0.8 Å in the active state. A sampling of the free energy landscape associated with this distance at 25 mM and 500 mM MgCl₂ concentrations showed that the equilibrium distance of E188 is related to the ionic concentration (Figures 5C, S6 in SI). This degree of freedom is sampled using the reaction coordinate RC₃, which describes the FES related to the conformational change of the E188. At low concentration (25 mM), the energy minimum is located at ~ 8.8 Å (with second minimum at 6.0 Å), similar to the conformation of the active state (Figure 3A). It is worth pointing out that this conformation is also similar to the X-ray structure, in which D192 is mutated to N192 (Figure 1B). The energy barrier is very small, about 1.5 kcal/mol. The energy minimum shifts to 6.0 Å at 500 mM concentration. A shallow minimum was found at 8.0 Å. Moreover, the energy barrier increases to 3.0 kcal/mol. This suggests that the inactive state is more stable at high concentration of MgCl₂, in agreement with the previous free energy calculations (Figure 5A).

In summary, at normal MgCl₂ concentration E188 binds to the third Mg²⁺ ion, which is naturally attracted in the solvent-exposed pocket around the active metal center. The most favorable conformation obtained by the free energy exploration of the active site architecture is the one that preserves a reactive conformation of the nucleophilic water (Figure 3A). Conversely, at high concentration, the MgA coordination shell is distorted when MgC gets closer, and the nucleophile directly binds to MgC disrupting its optimal orientation for the nucleophilic attack (Figures 3B, and 5B).

Rationalization of the attenuation effect in RNase H

Both the free energy calculations and free MD simulations revealed the importance of E188 in the second shell of metal ligands, and suggested that the dynamics of this amino acid could be important for explaining the attenuation effect^39,40. Since RNase systems are extremely sensitive to metals and are inhibited by high Mg²⁺ concentration, many studies have been done to address this question. A possible explanation of the attenuation effect is the existence of multiple metal binding sites in the active site. Electrostatic potential calculation⁴⁸ shows that the area around D192 and E188 possesses a high negative electrostatic potential (Figure 6A), and thus indicates, consistently with previous calculations reported in this work, that E188 defines an additional metal binding site, besides the DDE motif in the RNase H. Therefore, the removal of this glutamate group can reduce the propensity of attracting the third metal ion, which is likely needed for optimal catalysis as emerged from our simulations.

Figure 6. Calculated electrostatic properties of the active site in RNase H.

Shown are electrostatic potentials (ESP) obtained using the Adaptive Poisson-Boltzmann Solver method with dielectric constants of 1.0 and 78.5 for the solute and solvent, respectively (color scale: +10 kT/e (blue) and −10 kT/e (red)). The RNA/DNA hybrid is represented as a yellow ribbon. The calculated ESP for the wild-type system (A) and for the E188A mutant (B) is mapped on the protein molecular surface. Position 188 in both the systems is indicated by residue labels; the orange sphere represents the location of the third Mg²⁺ ion.

This hypothesis was further tested by performing MD simulations of a mutated E188A system, which is known to show a reduced activity in vitro with respect to the wild-type.³ During MD simulations, the metal ions MgA and MgB maintain their coordination geometry very well and the overall structural determinants are always identical at different Mg²⁺ concentrations. These results are in agreement with what was observed in mutagenesis experiments of E188A, where activity is always retained at any buffer concentration³, despite showing a reduction compared to the wild-type enzyme. Strikingly, we did not observe in MD at any MgCl₂ concentrations a third Mg²⁺ bound to the third metal site found in the wild-type. In fact, apart from the two metal ions bound at the active site, no other Mg²⁺ ions were found within ~10 Å from P, based on the calculation of the radial distribution function (Figure 4B). In addition, the original negative electrostatic character calculated based on the Poisson-Boltzmann equation around E188 is not shown in the E188A system (Figure 6B). Thus, the E188A mutant does not produce a favorable third metal binding site regardless of the MgCl₂ concentration. Any structural feature along with the solvation structure around the metal active site is identical in the wild-type and E188A system apart the presence of the third Mg²⁺ metal. Importantly, the nucleophilic Wat_N is not perturbed at any concentration and always has the canonical conformation found in the active state of the reactants (Table S2 in SI). This strongly supports the hypothesis that the third Mg²⁺ metal is indeed important for the optimal tuning of the reaction: it is the only distinctive difference emerged from MD simulations that can explain the reduced catalytic activity shown by E188A mutants³. The removal of E188 not only could compromise the Mg²⁺ ion loading on the active site, but also perturbs the electrostatic field during turnover, leading to an unfavorable electrostatic environment for catalysis and the dissociation and release of the products⁵.

While E188 clearly shows its ability to bond to solvated divalent ions, the catalytic role of MgC is still not fully understood. Nowotny et al.³ proposed that E188 can bind to MgA and promote the product leaving. However, the observation of MgC in our MD simulations suggests that the role of MgA can be replaced by this third bound metal ion. An X-ray structure of the product state (pdb code: 2G8V) shows MgA still bound to the active site, suggesting that the products leave the active site without the dissociation of MgA. As a result, we propose that MgC could bind to the product after the catalysis and promote the dissociation of the hydrolyzed RNA strand. In this scenario, the third Mg²⁺ ion, at normal concentration (as found in MD simulations of wild-type), can be functional by enabling the efficient progress of the catalytic reaction. Without its presence (i.e. in E188A system) the reaction progresses anyway but with reduced activity. A more quantitative estimation of the catalytic contribution of the third ion upon change in concentration will be assessed by QM/MM simulations that are currently in progress.

A three-metal binding site can be a recurrent motif in the endonuclease family

The occurrence of a third and more mobile metal ion at the catalytic site, and coordinated by second-shell Glu/Asp residues, could in principle be functional not only in the Bacillus halodurans RNase H system, but also in other similar proteins. A search in the structure database showed that second-shell glutamate residues are found in several NT superfamily enzymes (Figure 7). For example, a glutamate group (E131) is found in the second shell of E. coli RNase H⁴⁰ (pdb code: 1G15, Figure 7B), which with D134 could define a similar third metal site as in Bh RNase H. Supporting this hypothesis there is the fact that also E. coli RNase H shows the attenuation effect^39,40. In prokaryotic DNA transposase⁴⁹ (Tn5, pdb code: 1MUS, Figure 7C) a glutamate group (E190) is found 6.8 Å away from the active site. A second-shell ligand glutamate residue E478 is also found in the mononuclear HIV reverse transcriptase⁵⁰ (pdb code: 1SUQ), and might have a similar role in recruiting additional metal ions at the active site. Recently, an Aa-RNase III/-RNA complex has been presented and for the first time more than two metal ions have been observed at the active site of an endonuclease⁴⁶. A total of five Mg²⁺ ions have been indeed resolved (pdb code: 2NUG, Figure 7D); importantly, apart the two main catalytic Mg²⁺ ions (MgA and MgB in Bh RNase H) that hold the phosphodiester group to be cleaved, one metal ion can be clearly spatially compared to MgC in RNase H, being directly coordinated to two glutamate residues, E40 and E37, which would correspond to D192 and E188 in Bh RNase H. The distance between this Mg²⁺ metal and P is ~ 7 Å, similar to the arrangement obtained from MD simulation for active state conformation. Although representing a product state for RNase III, this structure points to the existence of additional metal binding sites in the endonuclease catalytic pocket and, together with our present computational results for Bh RNase H, suggests a more complex metal modulation in endonuclease catalysis. In general, the presence of additional acidic amino acids in the second shell implies that they could have a role for binding metal ions to the catalytic site and for optimally tuning the catalytic pocket electrostatics.

CONCLUSIONS

RNase H has become a popular target because of its critical role in many biological processes, involved in the RNA degradation of ds-DNA/RNA hybrid duplex. Also, a recent crystallographic study of the influenza A polymerase⁵¹ reveals a similar structure of the active site, with a two-metal ion bound at the DDE motif. This indicates that a comprehensive study of the structural determinants of RNase family might be beneficial to drug discovery projects targeting other endonucleases, which share a similar two-divalent-metal catalytic mechanism. Numerous crystallographic and spectroscopic studies have been performed for RNase H and attempted to unravel the reaction mechanism. Experimental data and theoretical calculations suggest a two-divalent-metal-aided S_N2-like mechanism. However, the cause of the attenuation effect, by which the RNase H activity is perturbed at high concentration of divalent metal ions, is still unclear. In this article, we have presented an investigation of the structure of the RNase H active site using MD with empirical force fields. In addition, ABF calculations have been performed to determine the free energy landscape associated with conformational changes in the active site and to understand the role of conserved second shell metal ligands (E188) related to the concentration of solvated metal ions.

The present calculations indicate that in addition to the originally known two metal ions bound at the active site, there is an extra metal. At low Mg²⁺ concentration, this third metal ion binds to D192 and E188, and is ~7.0 Å away from the scissile phosphate group. Conversely, at high Mg²⁺ concentration, the stronger electrostatic influence of the buffer brings the third metal ion and its coordination sphere closer to the catalytic site. Importantly, this seems to lead to an inactive conformation of the catalytic site, where the nucleophilic water molecule is displaced from its optimal position for catalysis (Figure 3). The free energy calculations show that this inactive conformation is more stable at higher Mg²⁺ concentration (Figure 5), in which the nucleophilic water molecule is more distant from the scissile group. It has been shown that the free energy cost to bring the nucleophilic water close to the optimal reactive state is higher at higher concentration (Figure 5B), explaining the inhibitory effect caused by the high concentration of metal ions. Moreover, the mutated E188A RNase H system, which presents a weaker but conserved activity at any MgCl₂ concentration, shows that the third metal ion does not bind to the active site at any ionic concentration, pointing to a crucial catalytic role at standard conditions of the third Mg²⁺ metal coordinated to the acidic second-shell residue, E188.

The presence of glutamate residues in the second shell ligand layer in other endonuclease enzymes (Figure 7) might also suggest the functional conservation of this structural motif, which can optimally and flexibly modulate the RNA phosphodiester cleavage. Ultimately, these results could inspire the structure-based design of modulators/inhibitors that target the principal metal binding site along with the third site and its structural coupling to the acidic second-shell residue. Molecules could be designed in order to probe catalytic activity and/or inhibition with linkers connecting these two sites, which exploit the concentration-dependent distance observed between the metal sites. In alternative, the endonuclease activity could be modulated by engineering specific mutations affecting the third metal binding site. This could be the case for several endonucleases with second-shell carboxyl ligands that use a structural triad of divalent ions to favor metal uptake and electrostatic fine modulation of the catalytic turnover.

MATERIALS AND METHODS

Structural Models

A model of the Bacillus halodurans RNase H and ds-DNA/RNA 12-mer complex was generated based on the X-ray structure of Nowotny et al. (PDB entry code: 1ZBL, 2.2Å resolution, Figure 1)³, which is a D192N mutant enzyme with completely impaired activity. In this inactive enzyme, the active site architecture does not significantly change from the wild-type system. Thus, N192 is substituted by an aspartic acid to reproduce the wild-type sequence. Two bound metal ions found in the active site are conserved (MgA and MgB). For the MD simulations, a total of five enzyme-substrate adducts were generated with different MgCl₂ concentrations. Each system was immersed in a 65 Å × 73 Å × 76 Å rectangular water box. ~11,000 TIP3P water molecules were needed to solvate the system. Mg²⁺ and Cl⁻ ions were used as counter-ions to provide the target concentration in the five different MD runs. Specifically, in addition to the counter-ions needed to neutralize the system, extra solvated Mg²⁺ ions were added, namely 3, 22, 50 and 100, respectively. The corresponding MgCl₂ concentrations are estimated to be approximately 25, 120, 250 and 500 mM, respectively, where 25 mM is the condition closer to standard/optimal concentration for RNase H. In order to investigate the inhibitory effect of E188, a mutated system was generated by replacing the glutamate residue by an alanine. Various concentrations of MgCl₂ were used to prepare the E188A system (40, 120 and 500 mM of MgCl₂). Due to the limited simulation cell size and the necessity to neutralize the system in periodic boundary conditions, the computed Mg²⁺ concentration is influenced by the protein volume. In addition, the ratio between the solvated divalent metal ions and the enzyme in the simulation is much smaller than in the experiment setup, which is millimolar in Mg²⁺ vs. picomolar in the case of the RNase H enzyme⁶. Nevertheless, the results from the present study are likely to be indicative of a quantitative trend for relevant structural and dynamic properties of the active site when passing from low to very high concentration of Mg²⁺ ions.

Molecular Dynamics

The AMBER force field (parm99sb)⁵² was adopted for all standard residues and nucleic acids. The metal active site is treated with a flexible non-bonded approach based on the “atoms in molecules” partitioning scheme of the DFT-BLYP level electronic density of the active site, as explained in detail in ref.53. This allows one to account for the metal-ligand interactions and permits possible structural rearrangements at the active site during the MD simulations. Specifically, the Bader’s atomic charges were used for the two Mg²⁺ ion and their ligands in the active site (D71, E109, D132, D192 and phosphodiester group on the substrate RNA strand). Aqvist’s parameters for solvated Mg²⁺ ion were adopted for the remaining Mg²⁺ ions⁵⁴. More details on the MD setup procedure and the Bader atomic charges are reported in Supporting Information. All the simulations were performed with the NAMD package⁵⁵. A total of 20 ns of dynamics were performed for all the systems and the last 16 ns of trajectory were used for data analysis.

Free Energy Calculations

To investigate the effects due to the conformational changes of the active site after the binding of the third Mg²⁺ ion at different concentration, we performed unconstrained MD simulations in the NPT ensemble to compute the free energy surface (FES) derived from the potential of mean force (PMF), using the Adaptive Biasing Force (ABF) method⁴⁷. In brief, the PMF, or the derivative of the Gibbs free energy (G) along the defined reaction coordinate (RC, ξ), dG/dξ is computed as:

$$\frac{dG}{d\xi }=-{\langle \frac{d}{dt}\left[{\left(\sum _i\frac{1}{m_i}{\left(\frac{\partial \xi }{\partial x_i}\right)}^2\right)}^{-1}\frac{d\xi }{dt}\right]\rangle }_{\xi }$$

Here, m_i is the mass of atom i, x_i is the Cartesian coordinate of atom i, and t is time variable. The Gibbs free energy, in this case, is acquired from the integration of the PMF:

$$G(\xi )=-\int {\langle \frac{dG}{d\xi }\rangle }_{\xi }d\xi$$

A detailed description of this method can be found in ref. 55.

Three reaction coordinates for ABF calculations have been used to describe the FES associated with conformational changes of the active site upon third ion binding (Figure 5): 1) the distance RC₁ which separates the third bound Mg²⁺ ion (MgC) and the C5 phosphodiester phosphorus atom P. RC₁ describes the FES related to the effect of the vicinity of MgC to the reactive center of the active site. The boundary of RC₁ is set as 4.9 Å < RC₁ < 7.5 Å (Figures 5A). 2) RC₂ is defined as the distance between MgA and Wat_N, and explores the concentration-dependent nucleophilic character of Wat_N. The boundary of RC₂ is set to 1.5 Å < RC₂ < 4.5 Å (Figure 5B). 3). The distance RC₃ separating the carboxylate group (C_δ) of E188 and P atom (Figures 5C). RC₃ describes the FES related to the conformational change of the E188. For all calculations, the instantaneous values of the force were accrued in bins 0.2 Å wide. For each calculation, the trajectory of the system is monitored to make sure the system is well sampled along the reaction coordinate and that the variation of the computed free energy over a 2 ns windows is less than 0.1 kcal/mol (Figures S5 and S6 in SI).