Magnesium induced structural reorganization in the active site of adenylate kinase

Abstract

Phosphoryl transfer is a fundamental reaction in cellular signaling and metabolism that requires Mg²⁺ as an essential cofactor. While the primary function of Mg²⁺ is electrostatic activation of substrates, such as ATP, the full spectrum of catalytic mechanisms exerted by Mg²⁺ is not known. In this study, we integrate structural biology methods, molecular dynamic (MD) simulations, phylogeny, and enzymology assays to provide molecular insights into Mg²⁺-dependent structural reorganization in the active site of the metabolic enzyme adenylate kinase. Our results demonstrate that Mg²⁺ induces a conformational rearrangement of the substrates (ATP and ADP), resulting in a 30° adjustment of the angle essential for reversible phosphoryl transfer, thereby optimizing it for catalysis. MD simulations revealed transitions between conformational substates that link the fluctuation of the angle to large-scale enzyme dynamics. The findings contribute detailed insight into Mg²⁺ activation of enzymes and may be relevant for reversible and irreversible phosphoryl transfer reactions.

Magnesium optimizes the angle defining the nucleophilic attack of substrates in phosphoryl transfer catalyzed by adenylate kinase.

INTRODUCTION

Phosphoryl group transfer plays vital roles in biology and is a key aspect of, for instance, signal transduction and regulation of eukaryotic cells (1), adenosine triphosphate (ATP) production by ATP synthase (2), and energy homeostasis (3). In cells, ATP is bound to Mg²⁺ to form the biologically active state (4), and it is estimated that more than 90% of cellular ATP exists as a complex with Mg²⁺ (5). The importance of Mg²⁺ as a metal cofactor of enzymes can be attributed to its high abundance in the cell and favorable physicochemical properties such as redox inertness and high charge density due to a small ionic radius (6). This high charge density allows Mg²⁺ to function as a strong Lewis acid. In particular, when complexed with a phosphate group, it attracts electrons from the phosphorus atom, increasing its electrophilicity for catalysis [so-called electrostatic catalysis (7)]. In general, Mg²⁺ adopts a tetragonal bipyramidal/octahedral geometry both in its complex with water as well as with ATP, with a distinct preference for oxygen donor ligands and a typical coordination distance to oxygen of about 2.0 Å (8). Direct interactions between the side chains of proteins with Mg²⁺ are called inner-sphere interactions, whereas indirect interactions mediated by water molecules are referred to as outer-sphere interactions (9). While there exists considerable knowledge regarding the structural biology and activation of Mg²⁺–nucleoside triphosphate complexes in enzymes (10–13), our understanding of the role of Mg²⁺ during a complete enzymatic reaction cycle remains limited.

Phosphoryl transfer catalyzed by protein kinases is inherently irreversible and requires the action of phosphatases for the reverse reaction (14). In contrast, phosphoryl transfer between nucleosides can be fully reversible. This is indeed the case with the adenylate kinase (AK) reaction (15), with reversible phosphorylation of adenosine monophosphate (AMP) by ATP to produce two adenosine diphosphate (ADP) molecules (16). Escherichia coli AK, hereafter referred to as AK_eco, has emerged as one of the principal model systems for studies of the linkages between structure (15, 17, 18), dynamics (19–22), stability (23–26), and catalysis (16, 27). Important factors that make AK_eco suitable for these studies include its relatively small size (24 kDa), which enables molecular dynamic (MD) approaches (28–35), its high solubility (36), high-quality nuclear magnetic resonance (NMR) spectra of both backbone (25, 37, 38) and arginine side chain resonances (22), and its ease of crystallizing in various functional states (12, 39, 40). In addition and of key importance is that AK enzymes undergo large-scale conformational changes in response to binding of both ATP and AMP to generate the closed and catalytically active state (15, 17). These conformational changes are to a substantial extent mediated by binding energy directed toward the adenosine groups of both ATP (41) and AMP (42, 43).While the majority of these studies have been performed with AK_eco, AKs from organisms representing all domains of life have also been investigated (12, 44–48). Two of the above-mentioned studies have provided valuable insight into the roles of Mg²⁺ during catalytic turnover. In a study of AK isolated from Odinarchaeota, Mg²⁺ was found to increase the apparent binding affinity of the substrates (45), while in a study of AK from Aquifex aeolicus, Mg²⁺ was found to increase the opening rate of the active site lids of the enzyme (12).

Structural knowledge of enzymes in both the reactant and product states (RS and PS) would provide valuable insights into the mechanisms underlying enzymatic catalysis. However, such structures are notoriously difficult to obtain, since the catalytic turnover produces heterogeneous mixtures of both reactants and products. To further elucidate the role of Mg²⁺ ions, we have here determined the crystallographic structures of wild-type AK_eco during the complete reaction cycle, i.e., for both the RS (ATP and AMP bound) and the PS (two ADPs bound) in complex with the Mg²⁺ cofactor. The structures provide unprecedented detail of the phosphoryl transfer reaction including an atomic description of the optimal angle, named the reaction angle, for the nucleophilic attack, and Mg²⁺ coordination in RS and PS together with a map of catalytic side chain interactions in both states. By integrating the structural studies with NMR and EPR spectroscopy, MD simulations, protein engineering, and activity assays, valuable insights were gained into the active site dynamics underlying enzymatic turnover and the contribution of water-mediated coordination of Mg²⁺ to catalysis. Together, the study advances our knowledge of how the Mg²⁺ cofactor acts to increase the rate of reversible phosphoryl transfer reactions in general and specifically for the commonly used model system AK.

RESULTS AND DISCUSSION

X-ray structures of the RS and PS of AK_eco in complex with Mg²⁺

In a previous study, we demonstrated that it is possible to use the slow and AK_eco-mediated hydrolysis of the alarmone Ap4A to obtain a crystallographic structure of the enzyme in complex with two product ADP molecules [Protein Data Bank (PDB) ID: 8CRG] (49). On the basis of this finding, we hypothesized that crystallization of AK_eco in the presence of both Ap4A and Mg²⁺ could enable determination of structures of catalytically competent states, which can exist either as bound to ATP and AMP or to two ADPs, i.e., the RS and PS, respectively (Fig. 1A). Screening of x-ray structures determined from AK_eco crystals grown with Ap4A and Mg²⁺ resulted in the successful determination of both the RS and PS structures in complex with the Mg²⁺ cofactor (Fig. 1, B and C; fig. S1; and table S1). The RS and PS structures have multiple copies of the enzyme in their asymmetric units, and there exists a variability regarding the presence of Mg²⁺ for individual chains in the asymmetric unit. These changes reflect the heterogeneity that can be expected in structural analysis during turnover conditions. In the RS, there are two chains in the asymmetric unit where chain A has ATP/AMP and Mg²⁺ bound, whereas chain B has a catalytically unproductive substrate composition consisting of ADP/AMP and Mg²⁺ (fig. S1C). For the PS, chains A and B contain ADP/ADP and Mg²⁺, whereas chains C and D contain ADP/ADP but lack bound Mg²⁺. Chains A in both the RS and PS are identified as catalytically competent states (table S2), and therefore, the structural analysis described below are focused on chains A.

Fig. 1. Reactant and product AK_eco state structures in complex with Mg²⁺.

(A) The AK reaction is shown together with the definition of RS and PS. (B) Structure of the RS (ATP, AMP) in complex with Mg²⁺ (green sphere) (chain A, PDB ID: 8RJ6, this work). (C) Structure of the PS (ADP, ADP) in complex with Mg²⁺ (green sphere) (chain A, PDB ID: 8RJ4, this work). In both (B) and (C), the ATPlid is shown in blue (residues 113 to 175), and the AMPlid is shown in orange (residues 31 to 73). The tetragonal bipyramidal coordination of the Mg²⁺ ion is shown for (D) the RS and (E) the PS. The capping and bridging interaction between Mg²⁺ and the substrates are indicated with “CAP” and “BRIDGE” for the RS and PS, respectively. (D) and (E) also highlight the indirect and water-mediated interaction between the Mg²⁺ ion and Asp⁸⁴ and the hydrogen bond between Asp⁸⁴ and Gln²⁸ that stabilizes the orientation of Asp⁸⁴. Water molecules that coordinate Mg²⁺ are shown as red spheres. The Mg²⁺ coordination to substrate oxygens and water is shown with yellow dotted lines. Hydrogen bonds are shown as blue dotted lines.

In both the reactant and product structures, a tetragonal bipyramidal coordination of Mg²⁺ was observed. In particular, the Mg²⁺ ion was coordinated by four water molecules and two phosphate oxygens of the substrates. While none of the protein side chains make direct interactions with the Mg²⁺ ion, the side chain carboxyl group of Asp⁸⁴ participates in indirect, water-mediated coordination to Mg²⁺ [outer sphere interactions (9)]. The position of the Asp⁸⁴’s carboxylate is stabilized by a hydrogen bond donated from Gln²⁸ (Fig. 1, D and E, and fig. S2). In the RS, Mg²⁺ is coordinated to the oxygens on the β- and γ-phosphorus atoms of ATP. We will refer to this coordination as a “capping” interaction. In the PS, Mg²⁺ bridges the oxygens on the β-phosphates of the two product ADP molecules, in what we refer to as a “bridging” interaction (Fig. 1, D and E). This coordination pattern suggests a direct transfer of the γ-phosphoryl group from ATP to AMP. Therefore, during this process, the coordination of the γ-phosphoryl oxygen to the Mg²⁺ ion is maintained, ultimately becoming the β-phosphate of ADP in the PS. While these structures provide a detailed structural description of the coordination of Mg²⁺ ions in AK_eco catalysis, it is important to note that they represent stable ground states and that additional motions or structural changes are required for catalysis. To address these motions, we used MD simulations as discussed below.

Mg²⁺-dependent alignment of reactants and products for the phosphoryl transfer reaction

The AK reaction proceeds through a nucleophilic attack of one of the phosphate oxygens on the accepting electrophilic phosphate atom. A recent quantum mechanical/molecular mechanical (QM/MM) simulation study (22) indicates that the reaction follows a concerted mechanism between the P─O bond cleavage and formation, known as the A_ND_N mechanism (50, 51). A requirement for an efficient nucleophilic attack is that the atoms involved are aligned with an optimal geometry of what we name the reaction angle. This angle in the RS is defined as the angle between the attacking oxygen on AMP, the accepting phosphate of ATP, and the oxygen atom of ATP that bridges the β- and γ-phosphates. In analogy, the reaction angle for the PS is defined as the angle between the attacking oxygen on ADP (bound in the ATP binding pocket), the accepting phosphate atom of ADP, and the oxygen atom that bridges the α- and β-phosphates of ADP in the AMP binding pocket (Fig. 2). We measured the reaction angles in the AK_eco structures in complex with AMP and ATP (RS) as well as in complex with two ADPs (PS) in the presence of Mg²⁺. The reaction angle was similar in both states: 167° and 168°, respectively (Fig. 2, A and B, and table S2). To benchmark these angles, we compare them to scenarios in the absence of Mg²⁺ and to the optimal reaction angle expected in the transition state of the catalytic reaction. Our previously solved structures of AK_eco in complex with ADP in the absence of Mg²⁺ (PDB IDs: 8CRG and 7APU) (22, 49) displayed reaction angles of ~135° (Fig. 2, C and D). Thus, in the absence of Mg²⁺, the reaction angle is suboptimal and is therefore not expected to promote catalysis to any appreciable extent. This effect then contributes to the substantially reduced catalytic activity observed in the absence of Mg²⁺ (12, 22). In the transition state, the reaction angle is approximately 170° ± 4° (22). Hence, it is evident that the Mg²⁺ cofactor induces a shift in the reaction angle to prime it for the nucleophilic attack. Consequently, the presented structures effectively capture the orientation of the reacting atoms for efficient phosphoryl transfer. The complete picture, however, has additional elements of complexity. In chain B of the PS, the reaction angle was found to be 141° (table S2) although Mg²⁺ is present which points toward a heterogeneity of the structural states and that dynamic adjustments are required for optimal configuration of the reaction angle (see the “MD simulations” section below).

Fig. 2. Reaction angles in Mg²⁺ complexes.

Display of the reaction angle for (A) the RS and (B) PS in complex with the Mg²⁺ cofactor (green sphere). Top: The atoms defining the reaction angle is indicated with arrows in the crystallographic structures. The nucleophilic attack is indicated with a purple dotted line. The Mg²⁺ ion coordination to the substrate oxygens is shown with yellow dotted lines. Bottom: Schematic illustration of the reaction angle. (C) The reaction angle in the PS for the Mg²⁺-free ADP structure [PDB ID: 7APU (22)]. The top and bottom displays follow the outline in (A) and (B). (D) Overlay of the PS in the absence [gray; PDB ID: 7APU (22)] and the presence of Mg²⁺ (colored by hetero atoms). The major adjustment and translation of atoms induced by Mg²⁺ binding are indicated with a black dashed arrow.

The findings suggest that, in addition to charge complementation at the active site (52, 53) and activation of the electrophilicity of the accepting phosphate (7), the Mg²⁺ cofactor contributes to catalysis by optimizing the reaction geometry with a substantial shift of the reaction angle. This shift is achieved by a translation of the β-phosphate group of the product ADP (i.e., the phosphoryl group that is transferred during the reaction) toward Mg²⁺ (Fig. 2D). Consequently, the β-phosphorus atom of the product ADP aligns in an in-line geometry relative to the nucleophilic oxygen of the ADP in the ATP binding pocket, which is stabilized by the coordination of the β-phosphoryl oxygen of the product ADP to the Mg²⁺ ion. Such an in-line geometry is consistent with the geometry expected at the reaction transition state, thus expected to contribute to the enzyme catalysis as proposed by the concept of the near attack conformation (NAC) (54). However, although the reaction angle is adjusted favorably by Mg²⁺, the distance between the nucleophilic oxygen and the electrophilic phosphorus atom remains unchanged (3.6 Å) both in the absence and presence of Mg²⁺. Hence, the data suggest a model for catalysis where Mg²⁺ acts to bring the two reacting groups into alignment to form the in-line geometry with an optimal reaction angle and that this Mg²⁺-bound state corresponds to a stable structural ground state. Then, thermal fluctuations are required to increase the energy of the system such that the transition state configuration, including an optimal distance for the nucleophilic attack, is achieved. The detailed interaction networks in the active site that supports the favorable reaction angles in RS and PS are mapped and discussed in the following section.

Structural map of side chain to substrate interactions in RS and PS

The structures of the RS and PS of AK_eco in complex with the Mg²⁺ cofactor allowed us to generate a structural map of the involvement of active site side chains in stabilization of the reaction angles. The active site of AK_eco is composed of six positively charged residues, which include Lys¹³, Arg³⁶, Arg⁸⁸, Arg¹²³, Arg¹⁵⁶, and Arg¹⁶⁷ (Fig. 3A) (55). The importance of these residues for catalysis has been demonstrated through both quantification of activity following replacement of these side chains (22, 56) and their impacts on the dynamics of the ATP- and AMPlids and on the catalytic barrier (28). Below, we discuss these active site residues and their interaction patterns in order of relevance, i.e., from higher to lower perturbations of k_cat following amino acid replacements (Fig. 3B).

Fig. 3. Interaction map for catalytic residues in Mg²⁺-coordinated RS and PS from the present x-ray structures.

(A) The active site residues of AK_eco are shown for the RS (left) and PS (right). Indicated is also the k_cat value for wild-type AK_eco (41). (B) The interaction patterns between catalytic side chains and the reactants (ATP and AMP) are shown in the left, and the interaction patterns to the products (ADP and ADP) are shown in the right. Only interactions with the reactant and products are shown for clarity. The relevant k_cat values are shown together with the side chain identities, and the figure is arranged from the smallest to the largest k_cat value. The indicated k_cat values are for replacement of arginine residues obtained from (22), and the value for replacement of Lys¹³ is from (56). Interactions with the terminal phosphates of both reactant and products are denoted CAP, whereas interactions across terminal phosphates and thus connecting reactant and products are denoted BRIDGE. The interaction between Arg³⁶ and the α-phosphate of a product ADP molecule is labeled “ANCHOR.”

First, Lys¹³ participates in capping interactions with the β- and γ-phosphates of ATP in the RS, while it bridges across the terminal phosphates in the PS. This interaction pattern mirrors the interaction pattern for Mg²⁺ (Fig. 1), which exhibits capping and bridging interactions in the RS and PS, respectively. A similar capping/bridging interaction pattern is observed for Arg¹²³. For Arg¹⁵⁶, a similar but reversed pattern is observed with bridging interactions to the two terminal phosphates in the RS and a capping interaction with the product ADP molecules. Arg⁸⁸ and Arg¹⁶⁷, on the other hand, display capping interactions in both the RS and PS. In particular, while Arg¹⁶⁷ follows the reaction trajectory by shifting its interaction from the γ-phosphate of ATP to the β-phosphate in the PS, Arg⁸⁸ is locked into one of the ADP molecules. Last, Arg³⁶ is involved in capping interactions with AMP in the RS, while it is left to interact with the α-phosphate of ADP in the PS, which we classify as an “anchoring” interaction. Together, a distinctive pattern emerges, wherein the residues with the largest impact on k_cat mimic the coordination pattern of Mg²⁺, involving both capping and bridging interactions. These symmetries in the interaction patterns suggest that these residues, i.e., Lys¹³, Arg¹²³, and Arg¹⁵⁶, play a direct role in the electrostatic activation of the reaction in agreement with previous studies (22). In contrast, Arg¹⁶⁷ and Arg⁸⁸, which show an intermediate relative effect (albeit large on absolute terms) on the catalytic activity, are only involved in the capping interactions, suggesting that they may contribute to the catalysis to a lesser extent. Arg³⁶, which shows the smallest catalytic contribution, seems to predominantly play a structural role, anchoring the substrates in the correct position for catalysis. In summary, the presented structures reveal distinct interaction patterns for side chains in both RS and PS, each associated with unique contributions to catalysis. The interaction pattern together with the coordination pattern of Mg²⁺ represents the structural framework that brings the reaction angle into a favorable position.

Dynamic displacement of reaction angles and conformational substates from MD

Although the crystallographic structures revealed an Mg²⁺-dependent optimized adjustment of the reaction angle, catalysis from these ground state structures requires additional structural rearrangements. To investigate the role of dynamics on the geometry of the RS and PS, we turned to MD simulations. However, instead of running microsecond-long MD simulations, we limited the simulations to 10 ns and repeated them 100 times starting from different initial geometries and velocities. This approach allows sampling of the enzyme confined to the closed and catalytically competent conformations while providing datasets with robust statistical significance. We have previously proposed that protonation of the substrates is important for the catalytic mechanism of AK_eco (12, 22). To test this further, we first examined the effects of protonation in the RS and PS in the Mg²⁺-bound complexes with MD simulations (Supplementary Results and figs. S3 and S4), using the x-ray structures determined in this study as a valuable benchmark. In particular, the structures obtained from the MD simulations in the RS and PS were compared with their corresponding x-ray structures. The protonation states tested were decided on the basis of the possible hydrogen bonds formed by the terminal phosphoryl oxygens (of substrates and products) with surrounding residues, which are summarized in table S3 (and Supplementary Results). The simulations showed that the protonation at the O1G of ATP in the RS was in best agreement with the RS x-ray structure determined (fig. S2A) and that in the PS the pronation at the O3B of ADP in the ATP binding pocket agreed with the present PS x-ray structure (fig. S2A). This supports that the present x-ray structures involve a protonation at one of the two bound ligands, which is needed to minimize repulsion between the two (negative) charges of the ligands. The MD simulations discussed below were performed based on these identified protonation states.

The results of the replicated MD simulations are presented in Fig. 4 and figs. S4 and S5, where the geometries of the reaction angle and the distance of the nucleophilic oxygen to the transferring phosphorus atom are presented and compared between the systems with and without Mg²⁺. In the systems with bound Mg²⁺, the geometries of both the substrates and products are clearly well in line with that of the expected transition state (Fig. 4 and fig. S4) (22), suggesting that the active site of AK_eco effectively maintains the orientation of the reacting species for catalysis, consistent with the idea of the NAC (54). In contrast, the systems without Mg²⁺ show substantial deviations from the transition state and/or NAC-like geometries. This deviation is characterized by a wide distribution of both reaction distance and angle, generally with increased reaction distances and decreased reaction angle (fig. S6). This observation holds true for previously resolved structures of AK in the absence of Mg²⁺ ions (12, 22, 49), corresponding to the PS mentioned earlier (Fig. 2C).

Fig. 4. Distribution of reaction angles, reaction distances, and conformational substates.

The distribution of the reaction angles and reaction distances in (A) the RS and (B) the PS. The reaction distance is defined in Fig. 2 (A and B), for RS and PS, respectively. The average structures of the major and minor substates for (C) the RS and (D) the PS, respectively, where the close-up view of their respective active sites is also shown to highlight the displacement of the transferring phosphoryl group in the minor substate. In both (C) and (D), the ATP and AMPlids of the major substate are shown in blue and orange, respectively, and the minor substate is shown in gray. Similarly, the substrates and products of the major substates are shown in color-coded sticks, and those for the minor substates are shown in gray sticks.

Furthermore, our MD simulations revealed the presence of conformational substates, characterized by distinct distributions of the reaction angles and distances, even for the systems with Mg²⁺ as shown in Fig. 4. In both the RS and PS, the major substate population exhibits reaction angles and distances closer to those of the reaction transition state, while the minor substate shows a lower value of the reaction angle due to the upward distortion of the transferring phosphoryl group away from the optimal in-line geometry. In the RS, this distortion occurs with a slight upward rotation of the Arg¹²³ side chain (Fig. 4C). Similarly, in the PS, the upward rotation of the transferring β-phosphoryl group coincides with upward lifting of the Arg¹⁵⁶ side chain (Fig. 4D). The impact of the upward lifting of the Arg¹²³ and Arg¹⁵⁶ side chains is not only localized in the active site only to displace the transferring phosphoryl group but is also propagated to the rest of the ATPlid to cause a slight upward displacement (Fig. 4, C and D). In comparison, the ATPlid of the major substates remains in a more closed conformation. In contrast, in the absence of Mg²⁺, although the enzyme remains in the closed ATPlid conformation (fig. S6, C and D), the distributions of the reaction angles and distances are more complex, dominated by increased reaction distances and decreased reaction angles (Fig. 4, A and B). This observation suggests that the Mg²⁺ cofactor suppresses the dynamic motions of the reacting species, i.e., substrates in RS and PS, to properly orient them for the reaction. On the other hand, in the absence of Mg²⁺, the substrates become more dynamic and poorly oriented relative to each other, contributing to a slower catalytic rate compared to that observed with Mg²⁺. This result clearly demonstrates the structural and dynamic role of the Mg²⁺ cofactor, which is distinct from the roles of charge complementation (52, 53) and electrostatic catalysis (7). Previously, we reported two subpopulations from long MD simulations (22) where the slight opening of the ATPlid was proposed to occur as part of the enzyme opening process. The present study further supports this observation and provides a detailed molecular basis for the displacement of the ATPlid.

Large-scale conformational dynamics induced by Mg²⁺

AK_eco and other enzymes of the large nucleoside monophosphate kinase family all have arginine-rich active sites (Fig. 3A). This opens the door to quantifying molecular motions with NMR spectroscopy by focusing on the arginine side chain H_ε-N_ε correlations (57). It has been shown that binding of the inhibitor P1,P5-Di(adenosine-5′) pentaphosphate (Ap5A) to AK_eco in the absence of Mg²⁺ induces ordering of arginine side chain N_ε atoms on the ps-ns timescale (22). Further, it has been shown from analysis of AK_eco backbone resonances that Mg²⁺ affects the dynamics of substrate binding domains with substrates present, a feature that occurs with an increase in NMR linewidths (12). Here, we set out to quantify the effects of Mg²⁺ on the dynamics of the arginine side chains in the complex with Ap5A. In contrast to the high-quality ¹H_ε-¹⁵N_ε spectrum of the Mg²⁺-free Ap5A complex, the addition of Mg²⁺ ions induces severe line broadening of all active site arginine side chain resonances even at substoichiometric Mg²⁺ concentrations (Fig. 5A). To rule out any contributions from contamination of Mg²⁺ with traces of paramagnetic ions (such as Mn²⁺), we compared the linewidths of the water signal in the absence and presence of Mg²⁺ (fig. S7). There was no observable increase in linewidths of the water signal, and therefore, we conclude that the line broadening observed is dependent on the dynamics of the system and not paramagnetic contamination. These effects preclude quantitative analysis of spin relaxation, but some degree of qualitative analysis is possible. Notably, the line broadening effect is propagated to Arg¹¹⁹, which is situated 10 Å away from the γ-phosphoryl group of ATP (Fig. 5B). Arg¹¹⁹ drives the induced-fit closure of the ATPlid by engaging in a cation-π interaction with the adenine base of ATP (41). The remote location of Arg¹¹⁹ suggests that the impact of the Mg²⁺ ion binding and its coordination to the Ap5A inhibitor must be large scale and global to induce line broadening at this distant site.

Fig. 5. Large-scale conformational dynamics induced by Mg²⁺.

Line broadening of NMR resonances of arginine side chains indicates a global change to the Ap5A-bound state. (A) The top left is an overlay of arginine side chains’ ¹H_ε-¹⁵N_ε HSQC spectra acquired for 700 μM AK_eco with 1000 μM Ap5A in the absence (black contours) and presence (orange contours) of 0.2 mM MgCl₂. The ¹H projections of catalytic arginine ¹H_ε-¹⁵N_ε peaks show the perturbation of line shapes upon titration with MgCl₂ in the concentration range between 0.01 and 0.2 mM. (B) Position of Arg¹¹⁹ in the RS structure. Arg¹¹⁹ serves to nucleate the closure of the ATPlid through a cation-π interaction between the side chain and the adenosine nucleobase (41). (C and D) Influence of Mg²⁺ on the AK_eco conformational states. DEER EPR measurements of the coupling between the nitroxide-labeled ATPlid and AMPlid. The distance distribution normalized to the maximum [P(r)/P(max)] are shown for (C) 20 μM AK_eco in the absence (gray) and presence of 20 mM MgCl₂ (red) and (D) 20 μM AK_eco with 100 μM Ap5A in the absence (gray) and presence of 100 μM MgCl₂ (red). ppm, parts per million.

To determine the relevant distances between subdomains, we turned to electron paramagnetic resonance (EPR) spectroscopy and, in particular, double electron-electron resonance (DEER) experiments. The influence of Mg²⁺ binding on the structure of AK_eco was evaluated by quantifying distances between the ATPlid and AMPlid for apo- and Ap5A-bound states in the presence and absence of Mg²⁺. To this end, both domains were labeled with 2,2,5,5-tetramethyl-1-pyrrolidinyloxy. The label was conjugated to the AMPlid at position Lys⁵⁰ and to the ATPlid at position Val¹⁴⁸ (see Material and Methods for details). It was found that the structural distribution of conformational states in the ligand-free apo state was invariant to the addition of Mg²⁺ (Fig. 5C). These data demonstrate that Mg²⁺ alone does not induce any major conformational changes to the apo enzyme which is reasonable since the enzyme has evolved to recognize Mg²⁺-loaded nucleoside phosphate complexes (e.g., Mg²⁺-ATP). For the Ap5A-bound scenario, AK_eco populates one well-defined structural state corresponding to the closed state (58) in the absence of Mg²⁺ (Fig. 5D). With added Mg²⁺, the structural distribution is markedly shifted toward that of the open, substrate-free state (Fig. 5D), where the range of the distance distribution suggests that multiple conformational states representing both open and closed states are populated. Expanded information for the EPR data is summarized in figs. S8 and S9 and table S4. The EPR data demonstrate that Mg²⁺ binding shifts the structural ensemble toward open conformations, which is consistent with the broadening of NMR side chain resonances. The simplest explanation for the NMR and EPR data is that Mg²⁺ affects the opening and closing dynamics of the enzyme, which is consistent with the previous finding that Mg²⁺ induces three orders of magnitude increase in the rate of the ATPlid opening (12). We recently proposed that the Mg²⁺ cofactor together with Lys¹³ and Arg⁸⁸ contributes to the facilitated opening of the enzyme after the reaction (28), where they hold the bound ligands so that the ATP and AMPlid easily can open. In contrast, in the absence of Mg²⁺ or replacement of the two residues, the two lids interact tightly with the substrates/products and thus open slowly. To further probe the perturbed lid dynamics, we compared the binding affinities of the inhibitor Ap5A. In particular, since substrate binding and product release are coupled to the conformational change of the enzyme (25, 26, 41, 59), the opening and closing rates must change in tandem to result in a minor overall effect on ligand binding affinities. This is indeed observed for binding of the Ap5A inhibitor, where the dissociation constant (K_d) values at 25°C are within the same order of magnitude, approximately 130 nM (25) in the absence and 200 ± 45 nM (fig. S10) in the presence of Mg²⁺. Although both K_d values are comparable, the presence of Mg²⁺ results in a slightly increased K_d (close to a factor of two). This minutely reduced binding affinity of Ap5A in the presence of Mg²⁺ is consistent with a relative increase in the opening rate relative to the closing rate which agrees with the EPR data that demonstrated an enrichment of open states in the presence of Mg²⁺.

Mg²⁺ coordination by AK_eco with a conserved functional motif

In living cells, more than 90% of the available ATP exists as a complex with Mg²⁺ (Mg²⁺-ATP) (5). Consequently, enzymes that use ATP must have evolved to interact with the Mg²⁺-ATP complex rather than with free ATP. In these enzymes, the Mg²⁺-ATP complex is activated through precise coordination in their active sites, as shown for the product and reactant complexes described here (Fig. 2). In AK_eco, a key residue for the positioning of Mg²⁺ in the active site is Asp⁸⁴, which coordinates Mg²⁺ indirectly via two water molecules (outer-sphere interactions). Asp⁸⁴ is in turn stabilized by a hydrogen bond donated from the side chain of Gln²⁸ (Fig. 1, D and E). Phylogenetic analysis (fig. S11) suggests that this stabilizing interaction pattern, specifically the indirect water-mediated Mg²⁺coordination via Asp⁸⁴ and stabilization via Gln²⁸, forms a conserved functional motif (Fig. 6A). For example, in AKs of bacterial origin, the conservation of an aspartic acid at the position corresponding to Asp⁸⁴ is high, with a percentage identity of 88%. For the position corresponding to Gln²⁸, the conservation of Gln is low (16%), while the most frequent residue in this position is histidine (58%). However, like glutamine, a histidine residue can donate a hydrogen bond through its side chain, and thus, the overall structural motif is likely a conserved feature in the AK family of enzymes.

Fig. 6. Conservation, structure, reaction angle, and activity of AK^D84A.

(A) The overall conservation of residues in AK is displayed on the substrate-free open structure of AK_eco [PDB ID: 4AKE (17)]. The ribbon view is colored according to conservation from the phylogeny shown in fig. S11. The Mg²⁺ coordinating residues Gln²⁸ and Asp⁸⁴ as well as the catalytic residues Lys¹³, Arg³⁶, Arg⁸⁸, Arg¹²³, Arg¹⁵⁶, and Arg¹⁶⁷ are indicated and colored with their degree of conservation. (B) The x-ray PS structure of AK^D84A. The ADP molecules and Ala⁸⁴ are shown with a “ball and stick” representation, and Ala⁸⁴ is also highlighted with an arrow. (C) The reaction angle in the structure of AK^D84A is shown (129°). Top: The atoms defining the reaction angle are indicated with arrows in the crystallographic structure. Bottom: Schematic illustration of the reaction angle. (D) Enzymatic activity of AK^D84A from a coupled enzymatic assay. The normalized reaction velocity (V/[E]_tot) that asymptotically approaches k_cat at high substrate concentrations is plotted versus ADP concentration. Errors are estimated from technical repetitions in triplicate, and the red line indicates the best fit to the Michaelis-Menten equation (Eq. 1).

To test the role of this proposed functional motif, we first characterized the AK_eco variant AK^D84A, in which Asp⁸⁴ was replaced with an alanine. We determined the structure of AK^D84A in complex with two molecules of ADP (Fig. 6B, fig. S1D, and table S1). Although Mg²⁺ was present in the crystallization buffer, we did not observe any corresponding electron density. This finding serves as an indication of the importance of Asp⁸⁴ for the binding and correct coordination of Mg²⁺. The reaction angle in the AK^D84A structure is 129°, which is fully consistent with expectations for a Mg²⁺-free scenario (Fig. 6C). In addition, removal of the carboxylate group of Asp⁸⁴ resulted in a reduction of k_cat from 330 s⁻¹ for the wild-type enzyme to 1.8 s⁻¹ for the AK^D84A, a change that coincided with an increase in the Michaelis constant (K_m) value from 70 to 470 μM (Fig. 6D).

On the basis of that K_m serves as a proxy for binding affinities for AK_eco (26), the binding affinity for the Mg²⁺-ATP complex in AK^D84A is reduced by only about 5 kJ mol⁻¹ relative to the wild type. On the other hand, the reduction in k_cat for AK^D84A is large and comparable to the replacement of active site arginines (Arg³⁶, Arg⁸⁸, Arg¹²³, Arg¹⁵⁶, and Arg¹⁶⁷), which are associated with k_cat values ranging from 0.28 to 2.44 s⁻¹ (22). These changes are large but still superseded by the substantial reduction in k_cat to 0.0075 s⁻¹ as observed in the absence of Mg²⁺ (22). Analysis of the k_cat values for the wild-type enzyme in the absence and presence of Mg²⁺, together with the k_cat for AK^D84A in the presence of Mg²⁺, allows us to dissect free energy contributions of Mg²⁺ binding to catalysis. Specifically, the reduction in the free energy barrier for AK^D84A relative to that of the Mg²⁺-free wild-type AK is ~14 kJ mol⁻¹ (based on k_cat values of 1.8 and 0.0075 s⁻¹). This difference in energetic barriers indicates that Mg²⁺ will enhance the reaction rate substantially even without the precise coordination mediated by Asp⁸⁴. Analogously, the increase in the free energy barrier for AK^D84A relative to the wild type both in presence of Mg²⁺ is ~13 kJ mol⁻¹ (based on k_cat values of 1.8 and 330 s⁻¹). A simple model that is consistent with these observations is that, in the absence of Asp⁸⁴, Mg²⁺ will bind to the nucleoside phosphate, but the exact active site architecture, including an optimal reaction angle, is not achieved. Therefore, it is likely that Mg²⁺ primarily interacts nonproductively with the substrate and the enzyme when lacking stabilization from Asp⁸⁴. Assuming that k_cat values directly reflect the fraction of productively bound complex, a rough estimate is that only 0.5% (1.8 s⁻¹ divided by 330 s⁻¹) of the Mg²⁺ ions is correctly coordinated in AK^D84A. Following these arguments, we propose that the contribution of Mg²⁺ to catalysis can be deconvoluted into two discrete steps corresponding to (i) general binding of the Mg²⁺-ATP complex to the active site and (ii) reduction of the number of nonproductive arrangements of Mg²⁺ through the water-mediated coordination through Asp⁸⁴. These two steps make approximately equal contributions to catalysis (13 and 14 kJ mol⁻¹, respectively), as deduced from the above analysis.

We further examined the role of the structural motif in Mg²⁺ coordination by replacing Gln²⁸ with alanine (AK^Q28A). The replacement resulted in a k_cat of 250 ± 5 s⁻¹ and a K_m of 21 ± 1 μM (fig. S12). This k_cat value is approximately 75% of the wild-type value, suggesting that replacement of the side chain of Gln²⁸ slightly perturbs the optimal Mg²⁺ coordination by Asp⁸⁴. This is an example of tertiary solvation layer interactions with the active site, where the Mg²⁺-coordinated waters are the primary layer of interactions, Asp⁸⁴ forms the second layer interactions, and lastly, Gln²⁸ forms the tertiary layer interactions. Thus, the order of their effects on the catalytic activity of the enzyme would be expected to follow the same order, which is indeed the case for the mutations of Asp⁸⁴ and Gln²⁸. Together, the experiments performed for AK^D84A and AK^Q28A demonstrate the importance of the identified structural motif underlying Mg²⁺ coordination in AK_eco.

Quantitative understanding of the phenomenal rate enhancements achieved by enzymes (60) is the focus of extensive contemporary biophysics research (61, 62). For any given chemical reaction catalyzed by an enzyme, there exists an underlying evolved reaction mechanism, as exemplified by the classic case of serine proteases, which rely on the concerted action of a catalytic triad (63). In the case of AK_eco, our QM/MM calculations have suggested that efficient phosphoryl transfer occurs through a concerted formation of a P─O bond and cleavage of an existing bond, facilitated by the collective action of the active site residues together with the Mg²⁺ cofactor (22). For a reaction mechanism to be effective, the number of unproductive conformational states during the reaction trajectory must be reduced, while unwanted side reactions are suppressed (59). Energetically, this translates into a reduction in entropy (i.e., ordering) as the reaction trajectory passes through the transition state barrier. Experimental evidence for this concept was put forward, for example, in the cases of the nucleoside monophosphate kinase UmpK (57) and the protein kinase Aurora B (64).

Here, we have presented findings from both experiments and simulations that contribute to a deeper understanding of how the Mg²⁺ cofactor affects the catalytic rate of AK_eco, adding additional data to the established electrostatic effect of Mg²⁺ in the activation of nucleoside di- and triphosphates. Notably, x-ray crystallographic cocrystal structures reveal that Mg²⁺ acts to prealign the substrates so that the reaction angle is optimally adjusted for the nucleophilic attack. A noteworthy aspect of this finding is that there is no sharp boundary between the concepts of active site preorganization and reduction of unproductive structural states, since the adjustment of the reaction angle (i.e., prealignment) occurs at the expense of unproductive reaction angles that are populated in the absence of Mg²⁺. In addition, the dynamic sampling of the reaction angle, observed from MD simulations, correlates with the subdomain opening/closing motions of the enzyme, while such correlation seems to be lost in the absence of Mg²⁺. This correlation suggests that the motion associated with sampling of the reaction angle occurs on a trajectory that overlaps with fluctuations of subdomain opening and closing that are biased toward the closed state. This correlation does not imply that the chemical transformation and the subdomain dynamics occur on the same timescale, but it does suggest a potential mechanistic linkage between these events in agreement with previous data (22, 28). This coupling between catalysis and large-scale enzyme dynamics was reinforced by observations from EPR and NMR spectroscopy where it was found that Mg²⁺ seems to increase the rates of opening and closing of substrate binding domains (12).

Mg²⁺ must be precisely coordinated in the active site for optimal catalytic function. In AK_eco, Mg²⁺ is coordinated in a tetragonal bipyramidal geometry in both the RS and PS and is anchored to the enzyme via water-mediated interactions with Asp⁸⁴. This conserved structural motif also includes the stabilization of Asp⁸⁴ via a hydrogen bond to Gln²⁸. Substitutions of both Asp⁸⁴ and Gln²⁸ with alanine established the importance of this structural motif for catalytically competent Mg²⁺ coordination. On the basis of these experiments, we propose that the structural motif has a gatekeeping role that ensures catalytically competent coordination geometry of Mg²⁺. We estimated that the effect on catalysis from reducing the number of unproductive structural states is substantial and on the order of 13 kJ mol⁻¹. The stabilizing effect of Gln²⁸ is defined as a tertiary layer of active site interactions, and the contribution to k_cat from this position highlights a challenge in de novo design of engineered enzymes (62, 65, 66). Even if an optimal active site can be generated, the failure to take outer-sphere interactions into account will severely diminish the catalytic activity. This challenge has also been demonstrated through the application of the directed evolution technique, which, in some cases, introduced mutations at remote positions from the active site (67). On a larger spatial scale, bound Mg²⁺ contributes to an increase in the rate constants of both opening and closure of the nucleotide-binding subdomains, consistent with previous findings for AK (12). Notably, this change in dynamics occurs without a modest twofold decrease in the binding affinity of the Ap5A inhibitor. This is in contrast to AK isolated from Odinarchaeota, where Mg²⁺ acts to increase the apparent affinity of substrates (45). Apparently, the role of Mg²⁺ can vary even within the same family of enzymes.

Through analysis of crystallographic structures of RS and PS, we were able to create a structural map of the interactions between catalytic side chains and the substrates. The structures uncovered a complex distribution of side chain–substrate interactions that were clustered into capping, anchoring, and bridging interactions. Arguably, the substrate and PS in complex with Mg²⁺ still represent ground-state equilibrium structures where the excursions into the catalytic transition states were captured within the ensemble of conformers in the MD simulations. Thus, the structures provide the complex atomic framework for how catalytic arginine side chains act to prealign the nucleoside phosphates in both the forward product formation and the reverse substrate formation directions.

Last, our data highlight a noteworthy similarity between the protein folding reaction and enzymatic catalysis. In protein folding, a key aspect to avoid the so-called Levinthal’s paradox, i.e., extremely slow folding (68), is an ordered folding mechanism that reduces the number of unproductive pathways, resulting in rapid folding into a well-defined, albeit dynamic, native structural state. For small two-state folding proteins, the nucleation-condensation mechanism has been found to enable efficient and rapid folding into the native state (63). In enzymatic catalysis, there exists a similar problem of reducing the number of unproductive structural states during passage over the transition state barrier. The solution is analogous to protein folding, i.e., the evolution of mechanisms that collectively act to steer the reaction along a productive reaction coordinate. Here, we have provided examples at atomic resolution of how Mg²⁺ supports an efficient reaction via prealignment and reduction of unproductive conformational states through proper Mg²⁺ coordination, optimization of the reaction angles, and specific side chain–substrate interaction networks. The intricate mechanism of AK_eco catalysis and the influence of outer solvation layer interactions, i.e., Mg²⁺ coordination via Gln²⁸ and Asp⁸⁴, highlight the complexity of enzymatic catalysis and call for further studies at the interfaces of experiment and computation that will eventually enable rational design of enzymes catalyzing novel chemistry.

MATERIALS AND METHODS

Protein production and site-directed mutagenesis

The recombinant plasmids of Escherichia coli AK were transformed into BL21(DE3) E. coli cells by heat shock and grown on a LB agar plate containing carbenicillin (100 μg/ml) at 37°C overnight. The following day, the transformed cells were inoculated into a 10-ml fresh LB medium containing carbenicillin at 37°C for approximately 4 hours. The cell culture was centrifuged at 4000 rpm for 10 min; the medium was then discarded, and the cells were gently resuspended in a liter of LB medium containing carbenicillin (100 μg/ml). Production of AK_eco was performed at 37°C for 21 hours. The plasmid is self-inducing and under the control of the endogenous AK promotor (69). The cells were harvested by centrifugation at 5500 rpm for 30 min, and the pellets were resuspended in 50 mM tris-HCl at pH 7.5. Disruption of cells was performed by using a Branson 450 Digital Sonifier (Branson Ultrasonics Corporation, USA). The cell lysate was centrifuged for 45 min at 16,000 rpm, and the soluble fraction containing the protein was subjected to further purification. Two steps of chromatographic purification methods were performed to obtain a sufficient purity of E. coli AK protein. The soluble fraction of cell lysate was first applied to a Blue Sepharose column (Blue Sepharose 6 Fast Flow, Cytiva, USA) equilibrated with 50 mM tris-HCl buffer at pH 7.5. After washing with the same buffer, the protein was eluted with 1.5 M NaCl and 50 mM tris-HCl at pH 7.5. The eluted fractions of AK were analyzed with SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and concentrated by using a 15-ml Amicon 10K ultrafilter (Merck Millipore Ltd., Ireland). Protein was then loaded onto a HiPrep 26/60 Sephacryl S-100HR gel filtration column (GE Healthcare, USA) equilibrated with 50 mM NaCl and 30 mM Mops buffer at pH 7.0. The protein purity was determined with SDS-PAGE and its concentration using the absorbance at 280 nm using an extinction coefficient of 10,430 M⁻¹ cm⁻¹. Purified protein was flash-frozen in liquid nitrogen and stored at −80°C. The plasmids for overproduction of the Asp⁸⁴Ala and Gln²⁸Ala variants were purchased from GenScript (NJ, USA).

Enzymatic activity assay

Enzymatic activity for AK^D84A was quantified with a coupled enzymatic assay in the direction of ATP and AMP production and by initiating the reaction with ADP. ATP production was coupled to the reduction of nicotinamide adenine dinucleotide phosphate (NADP⁺) into NADPH (reduced form of NADP⁺) by the combined actions of hexokinase and glucose-6-P dehydrogenase and supplementation of the reaction mixture with glucose. The buildup of NADPH was followed by the absorption at 340 nm and using an extinction coefficient (ɛ_{340 nm}) of 6220 M⁻¹ cm⁻¹. The protocol was modified from (70), and ATP production of 1 μM AK^D84A was quantified in 20 mM Hepes (pH 7.5), 100 mM potassium acetate, 200 μM NADP⁺, 400 μM glucose, 2 mM MgCl₂, bovine serum albumin (0.2 mg/ml), 3.7 U of hexokinase, and 2.8 U of glucose-6-P dehydrogenase, by varying the concentration of ADP between 50 μM and 10 mM at 25°C. For the AK^Q28A variant, the enzymatic activity was quantified as described previously (71, 72) in the direction of ADP production via the combined action of pyruvate kinase and lactic dehydrogenase and supplemented with phosphoenol pyruvate and NADH. To determine the K_m and the reaction velocity at saturating substrate conditions k_cat, the Michaelis-Menten equation (Eq. 1) was fitted to the normalized reaction velocities (V/[E]_tot) in response to increasing substrate concentrations ([S]). Errors were estimated by performing the experiment in triplicate

$$V/{[E]}_{\text{tot}}=k_{\text{cat}}[S]/(K_{\mathrm{M}}+[S])$$ (1)

Isothermal titration calorimetry

AK_eco and the inhibitor Ap5A were prepared in 30 mM Mops (pH 7.0), 3 mM MgCl₂, and 50 mM NaCl. A solution of 1 mM Ap5A was titrated into a cell containing 100 μM AK_eco at 25°C. The experiments were performed using a MicroCal ITC200 calorimeter (MicroCal-Malvern). The injection parameters were set as follows: total number of injections = 20; injection volume = 2 μl; initial delay and spacing between injections = 300 s; duration = 4 s; filter period = 5 s; and stirring speed = 300 rpm. The binding parameters including ΔG_a, ΔH_a, ΔS_a, K_a, and stoichiometry (“N”) were quantified from the titration binding curves by fitting the binding isotherms to a “one site” binding model in Origin 7 (OriginLab). Errors were estimated from technical repeats performed in triplicates.

MD simulations

The initial structures for the MD simulations were derived from the present study and deposited as 8RJ6 and 8RJ4 in the PDB repository, corresponding to the RS and PS. On the basis of the present structures of AK_eco in complex with Mg²⁺, four MD simulation systems were prepared: the RS and PS systems with and without Mg²⁺, respectively. The RS-AK_eco system with Mg²⁺ composed of AK_eco, ATP, AMP, and Mg²⁺ substrates based on the RS x-ray structure was determined in this study. Similarly, the PS-AK_eco system with Mg²⁺ contained AK_eco with two ADPs and Mg²⁺ based on the PS x-ray structure. The corresponding RS and PS-AK_eco systems without Mg²⁺ were prepared by removing the Mg²⁺ ion and its coordinated waters from the respective RS and PS-AK systems with Mg²⁺. The protonation states of all titratable residues were determined using the PROPKA program (73, 74); for the His residue, the protonation state was deduced from the hydrogen bond network in the enzyme. In all systems, one of the ligands (i.e., ATP in RS and ADP in the ATP binding pocket in PS) was protonated, which was determined by comparing the MD simulations to the orientation of the reactants and products in the corresponding x-ray structure. See the Supplementary Materials for details on the protonation states of the ATP/AMP (in the RS; fig. S3) and ADP/ADP (in the PS; fig. S3) systems.

Using the CHARMM-GUI web server (75, 76), the initial structures of these four systems were prepared, followed by solvation in a 70-Å cubic box of TIP3P-site water molecules (77). Neutralization was achieved by introducing Na⁺ and Cl⁻ ions to a final concentration of 0.15 M. Subsequently, the systems were minimized and equilibrated for 130 ps at 298 K (~25°C), using the CHARMM36m force field (78). Then, the production MD simulations were performed for 10 ns for each system, repeated 100 times by starting from different initial coordinates and velocity distributions to generate sufficient conformations. Thus, a total of 1-μs MD simulations were performed for each system. Here, we limited the length of the MD simulation to 10 ns to sample the enzyme conformation in a tightly closed state.

All MD simulations were performed using the CHARMM/OpenMM program (79, 80) (CHARMM version c47a1 and OpenMM version 7.3), at 1 atm and 298 K. Temperature was controlled using the Langevin thermostat (81) with a damping coefficient of 1.1 ps⁻¹, and pressure was controlled using the Monte Carlo barostat (82) with a piston period of 50 fs. Each MD simulation was performed with a 2-fs time integration using the leapfrog Verlet integrator and the SHAKE algorithm (83) to constrain hydrogens involved in bonds. The particle mesh Ewald method (84) with a cutoff of 12 Å was used for calculating electrostatic interactions, and the same cutoff distance was used for the evaluation of van der Waals interactions.

Crystallization and structure determination

Purified AK_eco was concentrated to 23.3 mg ml⁻¹ in 30 mM Mops, 50 mM NaCl, 10 mM MgCl₂, and 10 mM Ap4A (protein: Ap4A ratio, ~1:10). Crystals of the complex were obtained at 18°C using the sitting-drop vapor-diffusion method. Crystallization drops contained 0.5 μl of the protein-Ap4A solution mixed with 0.5 μl of precipitant and equilibrated against 0.2 ml of precipitant solution containing the following: (i) for the reactant complex (AK_eco-ATP-AMP): 24% (w/v) poly(ethylene glycol) 3350 (PEG3350), 100 mM bis-tris propane (pH 6.4); and (ii) for the product complex (AK_eco-ADP-ADP) and the PS complex of variant AK^D84A: 26% (w/v) PEG3350, 100 mM bis-tris propane (pH 7.0). Crystals were cryoprotected in 35% PEG3350 before vitrification in N₂ stream maintained at −173.5°C (100 K) using a Cryostream Cooler (Oxford Cryosystems). High-resolution synchrotron diffraction data at −173.5°C were collected at beamlines ID30B at the European Synchrotron Radiation Facility laboratory (Grenoble, France) and Biomax at MaxIV (Lund, Sweden). Data were processed and scaled using XDS, Pointless, and Aimless from the CCP4 suite (85). Data collection statistics are listed in table S1. The phases for structure determination were obtained by molecular replacement using PHASER from the PHENIX package (86) and the AK_eco structure (PDB ID: 1AKE) (15) as the search model. All structures were built and refined using Coot (87) and PHENIX REFINE.

NMR spectroscopy

NMR experiments were acquired on a Bruker Avance III HD spectrometer at 850 MHz using a triple-resonance (TXI 5 mm) cryoprobe equipped with pulsed field gradients along the x, y, and z axis. ¹H-¹⁵N HSQC spectra were measured at 298 K using a 30 mM MES and 50 mM NaCl buffer at pH 5.5. D₂O [10% (v/v)] was added to the sample for the field-frequency lock. The offset was centered to the arginine side chain region (84 parts per million). One hundred twenty-eight increments were collected in the indirect dimension, and the number of transients was set to 8 to obtain a proper signal-to-noise ratio. For NMR data acquisition and processing, Bruker’s Topspin 4.2.0 software package was used.

EPR spectroscopy

For EPR measurements, an AK_eco variant containing two point mutations at position Lys⁵⁰ and Val¹⁴⁸ was engineered, expressed, purified, and labeled as published previously (49). In short, Lys⁵⁰ was replaced by the noncanonical amino acid para-propargyloxy-l-phenylalanine, Val¹⁴⁸ was replaced by a cysteine, and the labeling was performed site-selectively over copper-catalyzed azide-alkyne cycloaddition at position Lys⁵⁰pPrF with 3-(azido)-2,2,5,5-tetramethyl-1-pyrrolidinyloxy and over cysteine-maleimide coupling at position Val¹⁴⁸Cys with 3-maleimido-2,2,5,5-tetramethyl-1-pyrrolidinyloxy (49). DEER samples were prepared by mixing the spin-labeled AK with the respective ligands (see table S4) in deuterated 30 mM Mops and 50 mM NaCl (pH = 7.0) and incubated for 15 min at room temperature. Then, 60% of d8-glycerol was added as cryoprotectant, and samples with a final concentration of 20 μM AK_eco in 3-mm quartz tubes with 3-mm outer diameter (fused quartz tubing, Technical Glass Products) were flash-frozen in liquid nitrogen and stored at 193 K (−80°C). The DEER experiments were performed at 34 GHz on a Q-band Elexsys E580 spectrometer (Bruker Biospin) equipped with a 150-W pulsed traveling wave tube amplifier (Applied Systems Engineering) in an over-coupled, commercial Q-band resonator (ER5106QT-2, Bruker Biospin with a Q value of approximately 200) at 50 K by using the standard four-pulse DEER pulse sequence (88). The experimental parameters were as follows: microwave attenuation: 0 dB; shot repetition time: 4080 μs; pulse delay between π/2 and π observer pulse: 400 ns; pump pulse position increment: 8 ns; eight-step phase cycling: [(x) (x) x_p x] (89) with typical pulse lengths of 100 ns for the pump pulse (a 90-MHz broad HS{1,1} pulse with β/t = 8 set to the resonator and nitroxide maximum frequency) and 30 ns for the rectangular observer π-pulses (set to −90 MHz with respect to the pump frequency). Nuclear modulation artifacts were suppressed by averaging eight traces with interpulse delay τ₁ varied by ∆τ₁= 16 ns. For data analysis, individual scans of the DEER measurements were phase-corrected and summed using MATLAB (version 2021b). Shown distance distribution was evaluated with the integrated deep neural network DEERNet (Generic) with uncertainties depicted as shaded areas for the 95% confidence intervals (90).

Phylogenetic tree construction and conservation analysis

AK protein sequences from reviewed proteomes were obtained from the UniProtKB database (91). The sequences were clustered at 50% identity using MMseqs 2 easy-cluster function with --min-seq-id 0.5 and -c 0.9 as parameters (92). Cluster representatives were aligned using Clustal Omega to default parameters (93). Conserved residues were extracted using the BLOSUM30 matrix with Block Mapping and Gathering with Entropy (94). The resulting data and multiple sequence alignment were used to generate a phylogenetic tree with iqtree2 (95). The phylogenetic tree was then visualized using Interactive Tree Of Life and rooted at midpoint (96). On the basis of the phylogenetic tree and the sequence alignment, the sequences belonging to the nonarchaeal clade were aligned using Clustal Omega to assess the conservation rate of individual amino acids in JalView (97). The conservation was additionally visualized by using the alignment of the full phylogeny in ChimeraX to color residues based on sequence conservation (98).

Data deposition

The atomic coordinates and structure factors were deposited in the PDB, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ, with the following accession IDs: RS complex with Mg²⁺ (8RJ6), PS complex with Mg²⁺ (8RJ4), and PS of AK^D84A in the absence of Mg²⁺ (8RJ9).

Supplementary Materials

This PDF file includes:

Supplementary Results

Figs. S1 to S12

Tables S1 to S4

References